44990 Information and Public Choice FROM MEDIA MARKETS TO POLICY MAKING Roumeen Islam, Editor Information and Public Choice Information and Public Choice From Media Markets to Policy Making Roumeen Islam, Editor The World Bank Washington, D.C. © 2008 The International Bank for Reconstruction and Development / The World Bank 1818 H Street NW Washington DC 20433 Telephone: 202-473-1000 Internet: www.worldbank.org E-mail: feedback All rights reserved 1 2 3 4 11 10 09 08 This volume is not a product of the staff of the International Bank for Reconstruction and Development / The World Bank. The findings, interpretations, and conclusions expressed in this volume do not necessar- ily reflect the views of the Executive Directors of The World Bank or the governments they represent. The World Bank does not guarantee the accuracy of the data included in this work. The boundar- ies, colors, denominations, and other information shown on any map in this work do not imply any judgement on the part of The World Bank concerning the legal status of any territory or the endorse- ment or acceptance of such boundaries. Rights and Permissions The material in this publication is copyrighted. Copying and/or transmitting portions or all of this work without permission may be a violation of applicable law. The International Bank for Reconstruction and Development / The World Bank encourages dissemination of its work and will normally grant permis- sion to reproduce portions of the work promptly. For permission to photocopy or reprint any part of this work, please send a request with complete information to the Copyright Clearance Center Inc., 222 Rosewood Drive, Danvers, MA 01923, USA; telephone: 978-750-8400; fax: 978-750-4470; Internet: www.copyright.com. All other queries on rights and licenses, including subsidiary rights, should be addressed to the Office of the Publisher, The World Bank, 1818 H Street NW, Washington, DC 20433, USA; fax: 202- 522-2422; e-mail: pubrights. ISBN-13: 978-0-8213-7515-0 eISBN: 978-0-8213-7516-7 DOI: 10.1596/978-0-8213-7515-0 Library of Congress Cataloging-in-Publication Data Information and public choice : from media markets to policymaking / edited by Roumeen Islam. p. cm. Includes bibliographical references and index. ISBN 978-0-8213-7515-0 -- ISBN 978-0-8213-7516-7 (electronic) 1. Mass media--Economic aspects. 2. Mass media--Political aspects. 3. Mass media and public opinion. 4. Communication in politics. 5. Communication in economic development. I. Islam, Roumeen. P96.E25.I54 2008 338.4'730223--dc22 2008011011 Cover design by Drew Fasick. Contents Acknowledgments ix About the Editor and Authors xi Abbreviations xv 1 Overview: From Media Markets to Policy 1 Roumeen Islam 2 The Media's Influence on Public Policy Decisions 17 David Strömberg and James M. Snyder, Jr. 3 National Media and Local Political Participation: 33 The Case of the New York Times Lisa M. George and Joel Waldfogel 4 Minority-Targeted Local Media and Voter Turnout: A Summary 49 Joel Waldfogel 5 I'm News, Are You? Newspaper Coverage of Elected vs. Appointed Officials 59 Riccardo Puglisi and James M. Snyder, Jr. 6 The Political Impact of Media Bias 79 Stefano DellaVigna and Ethan Kaplan 7 Market Forces and News Media in Muslim Countries 107 Matthew Gentzkow and Jesse M. Shapiro 8 Political Economy of Media Capture 121 Maria Petrova 9 Fostering an Independent Media with a Diversity of Views 139 Joseph Stiglitz 10 Media Regulation in the United States 153 Jonathan Levy 11 Aspects of Two Media Models: France and the United Kingdom and EU Media Governance 177 Pierre-Yves Andrau 12 Three Countries: Three Stories 197 Edetaen Ojo, Ziad Majed, Bambang Harymurti v vi Contents Tables 2.1 Newsworthiness of Disasters 27 3.1 Sample Statistics 37 3.2 Does the New York Times Depress Voting among the College Educated? 39 3.3 Robustness of Voting Results (Statewide Elections, Residency, and Internet) 41 3.4 Does the New York Times Affect Knowledge of Congressional Candidates? 45 4.1 Hispanic and Non-Hispanic Voter Turnout and the Presence of Spanish-Language Local Television News 54 4.2 Percent of Black and White Voter Turnout in Locales with and without Black Weekly Newspapers, 1998 55 4.3 Percent of Black and White Voter Turnout in Locales with and without Black Radio Stations, 1994 and 1998 56 5.1 Average Newspaper Coverage of Elected vs. Appointed Superintendents 63 5.2 Newspaper Coverage of Elected vs. Appointed Superintendents 66 5.3 Newspaper Coverage of Education Superintendents and Governors 68 5.4 Election-Time Newspaper Coverage of Elected Education Superintendents 70 5.5 Newspaper Coverage of Elected vs. Appointed Superintendents 72 6.1 Survey of Studies on Effect of Media Bias on Political Decisions 83 6.2 Impact of Fox News on Voting 92 6.3 Persuasion and Mobilization Rates 99 7.1 Consumer Beliefs and Satellite Television Viewership 112 7.2 Consumer Beliefs and Assessments of Quality 113 7.3 Consumption of Local Media 118 8.1 Regional Media Freedom and Advertisement Revenues, Russian Media 132 8.2 Description of Variables and Data Sources 133 12.1 Freedom of the Press in Indonesia 217 12.2 Global Corruption Perception Index 218 12.3 Number of Complaints to the Press Council 218 12.4 Number of Publications in Indonesia 219 Contents vii Figures 7.1 Country-by-Country Newspaper Coefficients 117 8.1 Inequality and Media Freedom (Freedom House) for Democratic Countries 124 8.2 Inequality and Media Freedom (Freedom House) for Autocracies 124 8.3 Media Freedom (Freedom House) and Public Spending on Education for Democracies 125 8.4 Media Freedom (Freedom House) and Public Spending on Education for Democracies 126 8.5 Media Freedom (Freedom House) and Public Spending on Health for Democratic Countries 126 8.6 Media Freedom (Freedom House) and Public Spending on Health for Autocracies 127 8.7 Fraction of Independent Newspapers in U.S. Counties as a Function of Advertising Rates in 1881 134 Acknowledgments I would like to acknowledge the participants who contributed to the con- ference, The Effects of Mass Media on Public Policy, held on April 12­13, 2007, which initiated this volume: Caroline Anstey, Shanta Devarajan, Alan Gelb, Steen Lau Jorgensen, Daniel Kaufmann, Steve Knack, Brian Knight, Mustapha Nabli, Rakesh Nangia, Sina Odugbemi, Sanjay Pradhan, and Janet Steele. I would also like to acknowledge the contributions of several colleagues who worked toward the production of this book and assisted in the administrative arrangements of the conference: Dulce Afzal, Lucas A. Bossard, Juan Castilo, Blair Ann Corcoran, John Didier, Maxine Alonso Pineda, Sunetra Puri, and Kaitlin Tierney. ix About the Editor and Authors Roumeen Islam is manager of the World Bank Institute's (WBI) Poverty Reduction and Economic Management Division. Prior to joining WBI, she was staff director of the World Bank's World Development Report 2002: Building Institutions for Markets. She was advisor to the chief economist and senior vice president in the Bank's Development Economics group. Ms. Islam has worked in World Bank Operations in several regions. Her pro- fessional expertise includes international trade and exchange issues, mac- roeconomic and fiscal stability, public expenditure rationalization, country growth strategies, sovereign debt rationalization, financial sector reform, governance and institutional reform, and private sector development. Pierre-Yves Andrau graduated with degrees in law and business. He joined the European Parliament in 1994 after a career in the private sector. In 1996, he moved to the European Commission, where he contributed to policy development in various policy teams responsible for international trade law and competition law. He participated in the drafting of the white paper on governance in Europe and contributed to the working group on building the European public space. He is a member of the Media Task Force in the Information Society and Media department and deals with media pluralism; freedom of expression; and press freedom; and with the legal, social, and economic effects of technological changes in the media field. Stefano DellaVigna has been an assistant professor in the economics department at the University of California, Berkeley, since 2002. He spe- cializes in applications of psychology and economics. He has published in such international journals as the American Economic Review, the Quarterly Journal of Economics, the Journal of Finance, and the Journal of Labor Economics. He has been a principal investigator for a National Science Foundation grant (2004­07) and is an Alfred P. Sloan Fellow for 2008­10. He has a PhD from Harvard University. His work on the economics of the media has focused on two areas: the impact of the media on voting, in particular, the impact of the Fox News Channel; and the impact of violent media on violent crime. xi xii About the Editor and Authors Matthew Gentzkow is an associate professor of economics and the John Huizinga Faculty Fellow at the University of Chicago's Graduate School of Business. He is also a faculty research fellow in industrial organization at the National Bureau of Economic Research. He received his PhD in economics from Harvard University in 2004. Lisa M. George is an assistant professor of economics at Hunter College, City University of New York, and has previously taught in the department of eco- nomics at Michigan State University. She is an applied empirical economist specializing in industrial organization and political economy. Her current research examines the economies of media markets. Bambang Harymurti is a leading Indonesian journalist and activist for clean government. Since 1999, he has led TEMPO weekly newsmagazine. He is a member of the Southeast Asia Regional Board of Advisors for the Open Society and the Anti Corruption and Assets Recovery Network. Mr. Harymurti holds a master's of public administration degree from the Kennedy School of Government, Harvard University. Ethan Kaplan is assistant professor of economics at the Institute for Inter- national Economic Studies and visiting assistant professor in the economics department of the University of California, Berkeley. He received his PhD in economics from Berkeley in 2005 and specializes in political economy and international macroeconomics. Jonathan D. Levy is the deputy chief economist at the Federal Commu- nications Commission (FCC) in Washington, D.C. In addition to advising the chief economist principally on media policy issues, Dr. Levy serves as a liaison between the chief economist and economists in the commission's bureaus and offices. Dr. Levy is also a senior economist in the commission's Office of Strategic Planning and Policy Analysis. Prior to joining the FCC in 1980, he taught economics at the University of Wisconsin­Milwaukee. He holds a PhD in economics from Yale University. In 1993, Dr. Levy was a Fulbright Senior Scholar in the School of Humanities at the University of Technology, Sydney, studying U.S.-Australian trade in television program- ming and film. Ziad Majed is a freelance researcher on social and political issues in the Arab world and consultant for the International Institute for Sustained Dialogue About the Editor and Authors xiii on the multicultural dialogue project among Arab, European, and American scholars and politicians. He is a candidate in political science at L'Institut d'Etudes Politiques de Paris and holds a master of arts degree in Arabic litera- ture from the American University of Beirut. He has worked at Internews in Beirut and at International IDEA in Stockholm. Edeaten Ojo is executive director of Media Rights Agenda, director of the Africa Freedom of Information Centre in Lagos, Nigeria, and chairman of the editorial board of Media Rights Monitor in Lagos. He holds a master's degree in international journalism from City University of London and a bachelor's degree from Oafemi Awolowo University in Nigeria. Maria Petrova is a PhD candidate at Harvard University. She is working on political economy of mass media and on politics of international investment. Riccardo Puglisi is a Marie Curie post-doctoral fellow at the ECARES, Uni- versité Libre de Bruxelles. An alumnus of Ghislieri College, he holds a PhD in economics from the London School of Economics and a doctorate in public finance from the University of Pavia. Jesse M. Shapiro is an assistant professor of economics at the University of Chicago Graduate School of Business and a faculty research fellow in labor studies at the National Bureau of Economic Research. He received his PhD in economics from Harvard University in 2005, and was the inaugural Becker Fellow at the Becker Center on Chicago Price Theory from 2005­07. James M. Snyder, Jr., is the Arthur and Ruth Sloan professor of political science and professor of economics. He has written on elections and cam- paign finance, legislative behavior and institutions, direct democracy, and mass media. Joseph Stiglitz is an economist and professor at Columbia University, teach- ing in the Columbia Business School, the department of economics, and the School of International and Public Affairs. He is a recipient of the John Bates Clark Medal (1979) and the Nobel Prize in economics (2001) and former senior vice president and chief economist at the World Bank. He founded the Initiative for Policy Dialogue and chairs the University of Manchester's Brooks World Poverty Institute. xiv About the Editor and Authors David Strömberg is Assistant Professor at the Institute for International Economic Studies, Stockholm University, and holds a PhD in Economics from Princeton University. He is the author of Workbook to Accompany Politi- cal Economics: Explaining Economic Policy (with Isabelle Brocas, Michael Cas- tanheira, and Ronny Razin). Joel Waldfogel is the Ehrenkranz Family Professor and Chair of the Business and Public Policy Department at the Wharton School of the University of Pennsylvania and a research associate of the National Bureau of Economic Research. Prior to arriving at Wharton in 1997, he served on the faculty of the Yale University Economics Department (1990­1997). His research interests span law and economics and industrial organization, with media a particular focus. Within media issues, he has studied the efficiency of free entry in radio markets; the relationship between public and private broadcasting; the effects of ownership consolidation on programming variety; how media markets serve minority consumers in radio, television, newspapers, and the Internet; local media markets and political participation; the effect of national media on local media and political participation; intellectual property piracy; and other topics. Abbreviations ABC Audit Bureau of Circulations APA Administrative Procedure Act CEO chief executive officer CJEC Court of Justice of the European Communities CPS Current Population Survey CRED Centre for Research on the Epidemiology of Disasters DBS direct broadcast satellite EM-DAT Emergency Disaster Database EU European Union FCC Federal Communications Commission Fed Federal Reserve FTA Free Trade Agreement GDP gross domestic product GRP gross regional product HHI Herfindahl-Hirschmann Index ICT information and communications technologies IPI International Press Institute MSA Metropolitan Statistical Area MVPD multichannel video program distributors NBC National Broadcasting Commission NGO nongovernmental organization NPRM notice of proposed rulemaking NTA Nigerian Television Authority OFDA Office of Foreign Disaster Assistance PCC Press Complaints Commission PEG public, educational, and governmental R&O report and order TVWF Televisions Without Frontiers U.K. United Kingdom USAID United States Agency for International Development xv 1 Overview: From Media Markets to Policy Roumeen Islam I nformation providers such as the media potentially influence political and economic outcomes by reducing society's cost of obtaining informa- tion. For business, households, and most others, these providers are the main source of information on public policy choices and current social and economic conditions. What news the media choose to gather, analyze, and disseminate, and what slant they choose to put on what they report are issues of consequence. These choices are not always simple ones. Information provision is costly, and media businesses must cater to dual markets: owners/advertisers versus consumers.1 For those concerned about the impact of the media on economic and political outcomes, it is important to assess media performance from the standpoint of economic efficiency as well as from its impact on other markets. This book addresses factors affecting the content and reach of news coverage and its impact on public policy through its influence on politics and economics.2 The chapters in this book address both market constraints that affect the media, particularly news and content, and the impact that news report- ing3 has on economic and political choices. The chapters consider the following: · evidence that media reporting affects politics, economics, and public policy outcomes or the process of policy making · the impact of bias in media reporting · the impact of market and nonmarket factors on news reporting, includ- ing sources of bias · the objectives of government regulation of the media sector. The final pages of the book cover the environment in which the media operate in developing countries. These examples help to distinguish the 1 2 Roumeen Islam differences in country and market conditions that the media and regulators face in developing and developed countries. The ability of the media to affect outcomes in economic and politi- cal markets has been discussed in several recent publications (World Bank 2002a, 2002b). News reporting and advertising may directly affect con- sumer behavior in goods and services markets in several ways: · revealing information about a product · acting as agenda setters to influence consumer demand (for example, goods produced with child labor may be negatively portrayed) · enhancing competition in markets by alerting consumers about substi- tutes or enhancing demand by revealing complementary goods · alerting consumers and producers about overall market conditions, such as those affecting intertemporal income profiles, for example, stock market conditions · alerting consumers and producers about changes in policy affecting goods or services markets or other nonmarket factors affecting their well- being, such as changes in monetary policy. News reporting and advertising may affect behavior in political mar- kets as well: · informing voters about politicians' or leaders' views, actions, and the consequences of their actions · informing citizens about outcomes of public policy · "making news" or taking stances on political, social, or economic issues to influence the behavior and views--and ultimately the votes--of consumers of media products. Media reporting can help garner support for government and for public policy and help or hinder implementation of new policy by raising aware- ness of potential benefits or costs associated with policy. Media reporting can increase the demand for change in public policy. For example, through media reporting, citizens become informed of special employment or social service plans provided by the government, or public expenditure increases in different areas. Media reporting has highlighted the costs of trade liber- alization and Free Trade Agreements in countries as diverse as Mexico, the United States, and the Dominican Republic. News reporting has an impact on policy if it changes voting patterns or affects reputation because individuals care about their reputation even if they are not elected to office. The influence of information on reputation Overview: From Media Markets to Policy 3 and its value in promoting certain types of behavior are well documented (World Bank 2002a, 2002b). Those in power have an incentive to dissemi- nate information flattering to themselves and to repress unflattering infor- mation. Individuals may care about their reputation in the workplace as well as in social circles. Thus penalties or rewards brought about by media reporting can be expected to produce some change in behavior and thus accountability. Finally, reporting on achievements of other public entities or individuals can heighten incentives for better performance and improved policy choices. Media content not only helps monitor activities citizens care about; media organizations may also highlight issues and be agenda setters (McCombs and Shaw 1972).4 Given the large number of ways that media reporting could influence outcomes, it is surprising that evidence on such influence is still relatively sparse. The first condition for media reporting to affect public policy is that it reaches the relevant audience--"access" to media. Access to media in many developing countries is limited and may depend on factors beyond the con- trol of media companies: it may depend on infrastructure, public policy limits on competition, or market conditions affecting private sector media profitability. In the case of both public and private provision of news, finan- cial returns may not always determine who gets news. The state may restrict or expand media reach for nonfinancial reasons. For example, governments may take conscious steps to increase access to the media in order to adver- tise their policy actions. Similarly, private businesses and nongovernmental organizations (NGOs) may take steps to raise media access even if these steps are not financially profitable in order to advance their nonfinancial agenda or to promote other business operations. A second condition is that the media cover information relevant to policy decisions over which voters and the general public have an interest, rather than focus on entertainment. A third is that media can affect policy if citizens have a way of influencing what public servants do based on the information they receive. In chapter 2, Strömberg and Snyder examine how media access and news content affect political behavior and policy. They contend that greater media access has affected the pattern of government expenditures. A num- ber of studies relate media access to government expenditure patterns or even government corruption. Strömberg and Snyder also show that where the press cover local politics, voters are better informed and politicians are more active. The public policy result is that more government funds are spent in those areas where there is greater news coverage. 4 Roumeen Islam Strömberg and Snyder address the divisions of content between news and entertainment and between local and national news. These divisions depend partly on consumer demand. If an issue is not newsworthy, then media outlets tend not to cover it since they cannot be guaranteed an audi- ence. The media may become agenda setters and cover news that the partic- ular owners of media organizations believe should be brought to society's attention or may "create" news to capture audiences. In the case where news is covered over entertainment, the question of impact turns to whether the news delivered is relevant to the stakeholder receiving the news and whether the stakeholder in turn can use the information in making decisions. In chapter 3 George and Waldfogel examine how technological and cost factors in the media industry's production function affect media con- tent and political markets. They contend that technological advances have made it easier and cheaper to aggregate consumers over large areas. More- over, due to large fixed costs incurred in developing a product, the larger the audience is, the lower the per-audience cost of production will be. These economic considerations tend to bias production content toward items that are attractive to many audiences. As a result, media products may reflect local tastes less than would be the case if products could not be cheaply shared among many audiences and production had low fixed costs. A growing literature discusses the merits of local content and local pro- gramming over national content in terms of encouraging civic participation. Most individuals are passive consumers of policy-related information and will find it prohibitively costly to find out about all the local affairs that affect their welfare. News reporting featuring local policy content reduces the cost of obtaining information. When local news is reduced, civic awareness and political participation should fall; similarly, they should rise when local news content rises. Individuals who consume less information relevant for local affairs may be less engaged in local affairs and local civic participation.5 Looking at the penetration of the New York Times, a national paper, in local markets and the political participation among its readers in these markets, George and Waldfogel conclude that participation in congressional elections in the United States fell with greater penetration of the New York Times. Greater penetration is associated with more national and less local news in the paper. Lower participation in congressional elections may indicate a less active role in local public affairs. This case illustrates that even if entry of national media in local markets may increase competition and enhance diversity in general programming content (particularly when local media groups are consolidat- ing), it may not necessarily increase local news content. Overview: From Media Markets to Policy 5 Waldfogel in chapter 4 considers another size-related aspect of media markets that may affect news reporting. Empirical evidence suggests that preferences for content vary widely among consumers of different types, such as consumers from different ethnic or income groups. Waldfogel examines how local media targeting of minority groups affects voter turn- out. Given the high fixed costs of production in the media sector, minori- ties will generally receive more news targeting their special interests when their communities are large. Waldfogel finds that the presence of Spanish- language local news increases Hispanic voter turnout in both presidential and nonpresidential (congressional) election years in the United States. Similarly, he finds that metropolitan areas with black-targeted local radio stations see more black voter turnout rates than areas without such stations. He also finds that the introduction of a black-targeted station raises voter turnout over time. While Waldfogel focuses on ethnic differences, other sources of differences--such as income--may also have an impact on news reporting. Advertisers prefer consumers with purchasing power, so certain media reporting would tend to target these consumers with the content they produce; but advertisers also prefer large audiences ensuring content for the middle class. Another example of how the cost of information may affect news sto- ries is captured by Puglisi and Snyder in chapter 5. These authors discuss the use of national school test scores in newspaper stories; using standard- ized test scores is a relatively easy way to assess the quality of the education provided in schools. News articles discussing school quality or the gov- ernment role in public education are to be expected around the time test scores are released because the costs of obtaining information on school performance declines when test scores are released, and it is "new" informa- tion or an "event." The information is "suddenly" revealed.6 While access may reduce the cost of obtaining relevant information, news may be biased toward what the particular source can provide. The authors confirm that the cost of obtaining information matters by showing that when this cost falls, news coverage tends to be greater. They relate the extra coverage to what they call an "event"--something new of note. Actually, it may be expected that people with children in school always care about performance, not only when scores are released. How- ever, it is usually costly to get an overall view of how a school--versus the individual child--is performing. Release of test scores is an event precisely because it is a one-time drop in the cost of obtaining information. Puglisi and Snyder confirm that coverage of school performance increases during 6 Roumeen Islam the time that test results are released. Studying newspaper coverage of the quality of education in public schools in the United States, they also find that newspapers give more coverage to elected superintendents rather than appointed ones.7,8 The additional coverage for elected superintendents is substantial during election years. Further, "bad" news tends to be covered more in states that elect their superintendents. The authors explain this higher coverage of bad news for elected officials as attempts by elected offi- cials to justify their performance in the media and attempts by the oppo- sition to use this information to their advantage in the media (proactive newsmaking). This latter phenomenon is an example of another source of slant in news reporting. Much news cannot be represented by one set of standardized numbers. There is usually no single or ideal set of data that allows objective com- parison of the issues voters care about. Also, the value of much of the data depends on how they are analyzed for consumer use (the storyline) and whether all the relevant data available are shared. For example, while one school district may deliver better results than another, it may also spend many times more per student than other districts and may receive more federal grants than another. Clearly, reporting that focuses on only part of the story could convey a different message to readers from one that uses all the relevant data and information. Omission is one of the methods used to create bias in news reporting. Other ways use words or pictures that audiences associate with undesirable outcomes, evoke negative images, or simply magnify the negative aspects of an incident. (In chapter 9 Stiglitz discusses another factor that reduces the cost of obtaining information: access to special "sources.") If news reporting has an effect in economic and political markets, then the bias or slant put on news reporting is important for consumers. In chapter 6, DellaVigna and Kaplan nicely complement the coverage of elected officials in chapter 5. These authors show that media slant can have a significant impact in political markets by affecting voting patterns. They study the impact of the introduction of the Fox News Channel in the United States--a right-wing channel ("right wing" being used to define the policies of the Republican party in the United States and "left wing" used to define the policies of the Democratic party in the United States) that generally supports the values and policies of the Republican party--on the Republican vote share, and voter turnout during both the U.S. senate and presidential races. The authors find that not only did the Fox News Channel convince more voters to vote Republican than would have been Overview: From Media Markets to Policy 7 the case otherwise, but it also mobilized more voters overall to vote. The results of the chapter seem generalizable since there is nothing inherently U.S.-centric in the general proposition about media behavior and response to media behavior that they explain. In fact, many countries have regula- tions governing the use of media coverage during elections or by different political parties. In chapter 7, Gentzkow and Shapiro ask what forces shape media con- tent, including slant/bias, more generally. The many potential sources of influence include both economic (profit related) and noneconomic (non- financial benefits) ones. The different pressures on media content come from market structure, ownership, regulation, advertisers and other finan- cial contributors, as well as consumer characteristics. The authors contend that in free markets the main influence on content seems to come from consumers. Markets will only produce accurate information when consum- ers want to hear it. This is because private media respond to advertisers, who aim to maximize the audience to which they expect to sell their prod- ucts. Maximizing audience means producing content that consumers want. Finally, they contend that consumers of news wish to hear reporting that reflects their prior biases. In a comparison of CNN and Al Jazeera viewers in countries where both channels are broadcast, the authors show that these television chan- nels tend to broadcast news with a slant or bias that reflects the bias of the majority of their viewers. They argue that news consumers tend to watch sources with a perspective that reflects their own independently acquired perspectives rather than sources that have a different perspective. Specifi- cally, in the Muslim countries studied, they contend that more religious people are more likely to watch Al Jazeera rather than CNN and these view- ers are more likely to be biased against the United States9 Their study also highlights that opening markets will foster viewpoint diversity on the sup- ply side, but that supplying different views does not necessarily ensure that consumers will watch different views being presented. The analysis of CNN and Al Jazeera viewers follows a study conducted in the United States in which the same authors show that readers preferring the Democratic party also prefer news tilted to the left (liberal), and readers aligned to the Repub- lican party prefer news tilted to the right (conservative). They conclude from this work and the extension to other countries in this book that media reporting is biased toward what consumers want to hear. They qualify this statement by acknowledging that there are sev- eral situations in which these conclusions would not be valid. For example, 8 Roumeen Islam where state ownership and/or state control of media is prevalent and the media are used to propagate views supporting the state's policies, consumer views may have little effect on media content. This is the case in some devel- oping countries.10 State ownership and capture of the media are addressed in a previous volume (World Bank 2002a) and are also demonstrated in the country examples presented in chapter 12 of this volume. In chapter 8, Petrova addresses a different source of media slant/bias; she considers the ability of special interest groups to manipulate public opinion through the media.11 She focuses on societies with a degree of inequalityinincomewherespecialinterestgroupscapturethemediainorder to disseminate information that benefits them and/or helps them retain their economic/political power. She contends that markets with more inequality and smaller advertising markets are more prone to capture. In democratic societies, the purpose of this capture would be to effect redistribution through policy change or to maintain existing inequalities by influencing public opinion and, ultimately, voters. Even in autocratic societies, public opinion manipulation matters to rulers; the role of media reporting is to help keep the current government in power. Petrova finds that inequality in income distribution and media freedom tends to be nega- tively related. Her work also suggests that smaller advertising markets are more prone to capture. Petrova addresses the impact that political parties and advertisers have on media content. She discusses what other authors have also described as the "stages" through which media financing evolves in economies as they develop and the consequences of this type of evolution for news content. In the earlier stages of development when advertising markets tend to be less developed and smaller, the state and political parties tend to play an important role in financing the media--either directly through ownership or indirectly through advertising. Ownership by the state and political par- ties has been used to disseminate views favorable to owners. In the early stages, advertising markets are small and are controlled by a few large com- panies. In addition, business owners whose primary financial interest is not in the media sector may establish media houses to influence policy or conditions that will support their other businesses. This has been the case in some countries in Latin America, for example.12 In larger, more devel- oped markets, the potential negative effects on news slant may be mitigated by the greater degree of potential competition in advertising and consumer markets and the larger size of markets. Overview: From Media Markets to Policy 9 Petrova sees the transition to a more developed market economy as one in which media reporting responds to consumers' interests. In the mod- ern economy where most revenues come from advertising, this is tanta- mount to producing what advertisers want, since they aim to maximize consumer reach. In this stage, advertisers constitute a much broader group who primarily aim to sell their goods and services. Even in larger developed markets, advertisers sometimes have preferences regarding content that are distinct from consumer preferences; they may want to act as agenda setters in order to influence policy, although Petrova does not address this issue directly. They may prefer positive coverage of public policy that furthers their business interests on the supply side--for example, policies regarding taxation, the environment, wages, or trade liberalization. These two goals of audience maximization and slant in news may not be in conflict. Media companies may tend to favor bigger advertisers' views regarding policy and slant their news reporting accordingly. In chapter 9 Stiglitz notes that news may be captured another way, specifically, when the public gets news from media who rely on particular "sources" for their information. He highlights the possibility that informa- tion may be provided as part of a gift exchange, or quid pro quo. He uses the example of government officials having special relationships with media personnel; media will obtain "new" information in exchange for favorable political/policy coverage. In other words, the supplier of information has "power" over the intermediary that needs the information. Reliance on spe- cial sources for information may also be related to the costs of obtaining information. This cost is lower when a reporter has a "special" relationship with a source. He advocates legislation like the U.S. Freedom of Informa- tion Act that reduces the artificially created scarcity of public information and therefore reduces the power to manipulate news. Stiglitz contends that private firms, like agents in the public sector, have incentives to provide only information favorable to their businesses. Fur- ther, media firms may have ties to business executives similar to those with public officials in which they have access to news early. Stiglitz advocates diversity in the media industry; multiple media can provide checks and bal- ances. Stiglitz sees a strong rationale for promoting public media and media run by nontraditional actors such as NGOs. The discussion in the chapters presented here and in other vol- umes indicates that a free or unregulated media or alternatively, state- dominated media, may not produce independent and diverse news 10 Roumeen Islam coverage of economics or politics. Further, they may not even provide news coverage relevant for important segments of the population. Some of the chapters speak to how the demand side of the market affects content, point- ing out that companies supply what consumers want. In such a scenario, increasing supply of a certain kind of output does not ensure that the audi- ence will consume it. Yet, by understanding the importance of informa- tion in markets and in democracies, many governments attempt to ensure diversity or pluralism in news reporting on the supply side to promote the interests of minorities, to ensure local/relevant content in news, and to encourage competition in the industry. To achieve their goals, countries have adopted various regulations of the media sector. In chapters 10 and 11, Levy and Andrau discuss regula- tions adopted in the United States and countries in the European Union, respectively. Levy, focusing primarily on broadcast television, sees three basic policy objectives for regulators: promoting economic competition, diversity, and localism, where diversity refers to programming content as well as viewpoints. The localism goal of the United States is intended to promote programming that is responsive to the communities in which the license is granted. Television licensees in the United States have a general obligation to offer public-interest programming, and the United States reg- ulates how political candidates are represented on television. For example, stations that provide or sell time to one candidate for office at national, state, or local levels must provide or sell time to all candidates, and time must be offered at the station's lowest unit charge during a specified num- ber of days before elections. Levy points out that regulations on market structure are being used to promote diversity--as in other countries--but that calibrating the impact of market structure on content diversity is not a simple matter. It is probably harder to trace the impact to viewpoint diversity than it is to programming diversity. Among the various rules set by the United States, the Federal Com- munications Commission has limited the number of television and radio stations in a given market that may have a single owner, the cross owner- ship of media, and established national broadcast rules limiting market reach and mergers.13 Regulators in the United States continue to assess the impact of their oversight in terms of its ability to promote its objectives. Andrau discusses the French and British models of regulation. In France, private television has been allowed only since 1986. The laws gov- erning the media sector in France espouse independence, transparency, and pluralism. The French public service broadcaster is financed by license fees Overview: From Media Markets to Policy 11 and advertising and is called upon to ensure free expression for all political and social representatives of French society. Specific programming obliga- tions are imposed on both private and public broadcasters to ensure plural- ism. While the French government provides financing for the press and the audiovisual sector, a stringent legal framework prevents state authorities from linking aid to political affiliation. To enhance transparency, publishers are required to provide information on the identity of the main owners and editors, and above a certain threshold, publishers are required to publish all financial information in their publications. All terrestrial broadcasters in the United Kingdom also have public service obligations. Traditionally, the public service programming contribution has been made by commercial broadcasters to fund content deemed to be in the public interest. However in Britain, the funding provided to public television, namely the BBC, has often come under attack as creating an unfair advantage for the company. Both Andrau and Levy point to how regulations need to be reassessed as technology changes. As illustrated by the French, U.K., and U.S models, some governments seek to safeguard the public interest by adopting regulations that promote diversity in viewpoints. However some governments do the opposite. Simi- larly, private incentives may be aligned toward "fair and accurate" repre- sentation, toward agenda-setting behavior, or toward playing an active role in politics by supporting certain politicians, policies, or leaders. A variety of pressures is at work on media companies. For example, markets where there are large inequalities in income, several ethnic/cultural groups, or less competition are different in their economic and political characteristics than markets in which the opposites hold true. Much of the available empirical evidence linking media reporting to public policy focuses on developed countries and particularly the United States. This is partly due to the availability of systematic data needed to link various aspects of industry structure to industry behavior and partly due to researcher interest. Although the structure of the media industry varies from country to country, as does the historical context in which regula- tions are formed, it is generally accepted that media-reported information is important for both economic and political well-being. It is probably rea- sonable to expect that some lessons related to incentives and behavior that may lead to less-than-optimal production and dissemination of informa- tion are broadly applicable to other countries. What differs from country to country are the defining characteristics that make one effect greater than another. Which incentives dominate in a particular country is an empiri- 12 Roumeen Islam cal question. What the literature helps to highlight are the many possible ways in which news reporting can be affected and how it in turn can affect economic and political markets. In order to understand media industry behavior in developing countries, it is necessary to think of how differ- ent initial conditions--such as market size, consumer tastes, technological constraints, literacy, the role of the state, and the capacity to regulate--may affect market incentives and news reporting. This book highlights how more empirical analysis would benefit our understanding of the ways in which the media affect policy. In chapter 12, three practitioners in the media industry provide a snap- shot of media market characteristics and government intervention in the media sector giving a flavor of how conditions differ among countries. Ojo points out that the Nigerian media operate in small markets for products and advertising in a country where illiteracy is high, infrastructure is poor, and the regulatory/ownership structure gives the state a great deal of con- trol over the media. He contends that the dominance of state ownership in media (particularly television) gives the state an advantage in dissemi- nating information that is pro-government. Further, in terms of the legal framework, he notes that the constitution does not specifically guaran- tee freedom of the media, that the Freedom of Information Bill has yet to become law, and that the Official Secrets Act and the criminal code for media operations are overly restrictive. In discussing the Lebanese media, Majed highlights the turbulent times that media companies have undergone in recent decades, related mostly to Lebanon's political turmoil. His account mentions state censorship, but also self-censorship by media owners afraid of political reprisals. At times, the Lebanese media have been overtly used to further the interests of dif- ferent political groups. Harassment of journalists has been common. Yet, as he writes, the Lebanese constitution provides for freedom of speech, unless news "harms religious beliefs, destabilizes the country or damages relations with allies." Unfortunately, this restriction may be interpreted broadly. The Lebanese press is privately owned. Political parties still own some media-- reminiscent of the situation in the United States some decades ago. The Indonesian story related by Harymurti brings to the fore the impor- tance of an independent judiciary in overseeing and guaranteeing legislated media freedoms. Harymurti believes that the legislative framework for media operations is very important, as it may protect the media from undue interference exerted by special interest groups. Harymurti highlights some Overview: From Media Markets to Policy 13 concerns with concentrated media markets in Indonesia and nontranspar- ency with respect to the ownership of media companies. All three authors refer to political history, social attitudes, and laws supporting press freedom in defining how much media freedom exists. They acknowledge that media independence or freedom cannot be taken for granted and that media independence and pluralism evolves along with economic and political conditions. All three relate greater diversity in the media sector to a stronger private sector that may provide checks and bal- ances to the public sector. While only the Indonesian story mentions some concerns with concentrated private ownership of media, in other countries around the world, such as Russia, oligarchic private media markets may also not deliver a diversity of viewpoints. Many of the chapters in this book discuss how private media ownership and market conditions are critical factors affecting the content, slant/bias, and reach of news reporting. Just as there is no presumption that government-owned and -operated media automatically produce independent and pluralistic news reporting, there is no guarantee that free markets will do so either. Independence of viewpoint and diversity of views is affected by many factors: ownership, financing, or consumer biases. In addition, other market conditions such as language and ethnic differences affect content. Although more open markets, includ- ing foreign channels, tend to foster competition in views, nevertheless sup- plying diversity may not always promote greater diversity in consumption of content. Some general differences among countries in economic conditions and political institutions are relevant for media operations. Developing country markets tend to be small both for advertising and consumption, and gov- ernments in developing countries tend to own or control larger segments of the media industry than they do in developed countries.14 Institutions sup- porting media freedoms are more developed in democratic countries and also tend to be more developed in countries with a vibrant private business sector and competitive markets. Developing countries cannot take these supporting institutions for granted. Experiences around the world indicate that a strong and diverse media sector capable of producing checks and bal- ances to market actors and government is not something that can be taken for granted. Finally, much more research needs to be done to better under- stand the impact that news reporting by the media has on public policy and to better understand the benefits of media and information regulation in a changing landscape. 14 Roumeen Islam Notes 1. The choices made by privately owned media depend to a large extent on market characteristics that influence costs and revenues but also nonmarket characteristics such as the prestige or influence associated with owning media or affecting media coverage. Publicly owned media may have different motiva- tions for their choices. In some cases, gains in political markets may influence media choices. 2. Public policy decisions are the result of both economic and political forces and result in economic, social, and political consequences and thus the choice of wording "economic and political markets." 3. News reporting is a broad term that includes news production, analysis, and dissemination. 4. This volume focuses mostly on the impact of news in democracies. Democra- cies are presumed to foster the exchange of information since information flows between citizens and government officials is critical for the proper functioning of a democracy. However, this is not to say that information is not viewed as a critical agent of policy change in nondemocracies. It is precisely because of the power of information to change support for leaders in non- democracies that information--and the media--are controlled. 5. The same individual may be expected to also have preferences over national issues as well; to the extent that they receive relevant national news, their civic participation will be affected. 6. This is regardless of whether test scores are a good measure of school performance. 7. In the United States, there is variation among the states. In some states the superintendent of schools is elected, and in some states the superintendent is appointed by the (elected) governor. 8. However, they find that once they account for the extra coverage devoted to the campaign of elected officials (and the extra coverage during appointment periods), this difference is not important. 9. In practice, how religious a society is may be influenced by exposure to certain media. It may be hard to distinguish between situations where the introduc- tion of a station increases religiosity versus one in which more religious people are found to be watching that station. 10. The Washington Post reports the case of Al Raheel, a private newspaper that is posted on trees in Sudan (and does not charge subscribers!). See McCrummen (2007). 11. See also Herman and Chomsky (1988) and Lippman (1922) for various inter- pretations of how mass media influence politics and public policy. 12. Hughes and Lawson (2005) mention the cases of Mexico and República Boli- variana de Venezuela, among others. 13. For example, no firm may own television stations reaching more than 39 per- cent of U.S. households with televisions. 14. World Bank 2002a and 2002b covers issues related to state control in greater depth. Overview: From Media Markets to Policy 15 References Herman, Edward, and Noam Chomsky. 1988. Manufacturing Consent: A Propaganda Model. New York: Pantheon Books. Hughes, Sally, and Chappell Lawson. 2005. "The Barriers to Media Opening in Latin America." Political Communication 22(1): 9­25. Lippmann, Walter. 1922. Public Opinion. Glencoe, Illinois: Free Press Paperbacks. McCombs, M. E., and D. L. Shaw. 1972. "The Agenda-Setting Function of Mass Media." Public Opinion Quarterly 36 (Summer), 176­87. McCrummen, Stephanie. 2007. "In Darfur, A Journalist Branches Out." Washington Post. March 4. World Bank. 2002a. The Right to Tell: The Role of Mass Media in Economic Develop- ment. Washington, DC: World Bank. ------. 2002b. Building Institutions for Markets: World Development Report. Washing- ton, DC: World Bank. 2 The Media's Influence on Public Policy Decisions David Strömberg and James M. Snyder, Jr. Introduction T his chapter will discuss recent work by economists on the influence of the mass media on public policy. It will concentrate on the recent work by the authors rather than provide a full review of the literature and will examine the extent to which the following factors may affect policy: · differences in media access across groups · active media coverage of political content · publication of individual news stories. We first discuss the mechanisms by which the media may influence policy. Media content may affect policy because it influences the weight that people place on different issues when voting. If the media devote a large amount of attention to unemployment--for example, publishing stories about how high the rate is, how it has grown, how particular firms are laying off large groups, and how unemployment affects people's daily lives--then people may attach more weight to unemployment when voting. Anticipat- ing this, politicians may devote more resources to dealing with this issue. This is called the "agenda-setting" model of politics, developed especially by McCombs and Shaw (1972). Although the model relates policy to media content, it is silent about the effects of media access. In other models, media access may matter if the information provided by media influences collective behavior--such as voting decisions--because policy makers will have a strong incentive to accommodate the concerns of the informed. Consequently, citizens with media access may benefit dispro- portionately from government programs. Media content may matter for the same reason. If the media do not cover certain issues, then they do not help the affected voters to acquire the information necessary to take appropriate 17 18 David Strömberg and James M. Snyder, Jr. actions on those issues. Finally, media access and content only matter if citizens take interest in the information. Information transmitted by the media may influence both the ongoing policy process and electoral campaign promises. For example, in the 2001 election in Thailand the Thai Rak Thai party promised inexpensive univer- sal health care through a 30-Bath Gold Card program; in the 2002 election in Brazil, Luiz Inacio Lula da Silva promised to reduce hunger through a "zero-hunger" program. These types of promises are only politically pro- ductive if those who stand to gain learn about them. The electoral incen- tives for politicians to make these types of promises are small if many poor voters lack access to the mass media, or if the media fail to cover the prom- ises because attracting a poor audience generates little extra advertisement revenues. Similarly, after the election, politicians have lower incentives to actually deliver policies favorable to voters without media access. Such vot- ers are less able to recognize and give credit to politicians who provide policies that benefit them. In other words, these voters are less able to hold politicians accountable for policies. This information-driven model of the policy effects of media access and content was developed by Strömberg (1999). The model is more empir- ically convenient since media access data are more readily available than media content data. Studies of the effects of media access can easily relate mass-media penetration rates--from, for example, census data--to politi- cal outcomes in geographical areas. The three issues identified in the opening paragraph are clearly related. Citizens with access to mass media may benefit politically, but they will do so only if the media carry information that is politically relevant and useful or entertaining enough that people attend to it. Only then can the media elicit collective responses, inducing policy makers to accommodate people's concerns. Perhaps the clearest example of the combined effect of media access and content is the work by Reinikka and Svensson (2005).1 They studied the effects of a Ugandan newspaper campaign informing people of entitlements to government educational spending. Although some people had better access to newspapers because they lived closer to newspaper outlets, these people did not experience higher levels of school spending prior to the campaign because the media did not carry the rel- evant information. The government subsequently launched an advertising campaign to communicate this information; people with better newspaper access experienced larger increases in school funds, consistent with a posi- The Media's Influence on Public Policy Decisions 19 tive media impact. This case contains all three ingredients that make media effects likely: access, relevant content, and reader interest. However, this is not a study of "ordinary" media content. First, it measures the effects of advertisements rather than journalistic content. Second, the advertiser is a government seeking change in the implementation of policy. The policy effects of regular media content, written by reporters and less aligned with the agenda of the government, might be quite different. How Does Media Access Affect Policy? Other studies have tried to assess the policy impact of the regular stream of information through the media. They ask whether policy outcomes depend on media penetration (access) rates, taking media content as given. The first example of this type of study is Strömberg (1999, 2004b), who found that American counties with high radio penetration rates received more unem- ployment relief in the 1930s. The conditions for finding media effects were good. The introduction of radio in the 1930s caused a substantial increase in people's access to information, particularly in rural areas where daily newspapers were scarce. The radio stations carried ample and politically relevant news, both at the state and federal levels. For example, Franklin Roosevelt gave radio addresses during his tenure as governor of New York; later as president, he used this medium for his series of "fireside chats" with the American public. These "chats" were used to explain the policy issues of the day, such as the bank crises of the early 1930s, unemployment and the New Deal programs, and drought conditions. Farmers in rural areas were motivated to listen to the news so that they could learn about weather and crop prices. Moreover, special programs packaged news about politics together with weather and market reports. For example, a popular agri- cultural radio program, the "National Farm and Home Hour," combined political news of interest to farmers, such as developments in government programs, with crop reports and prices, educational talks on technical agri- cultural procedures, and music and features (see Brunner 1935). In the model of Strömberg (2004b), better-informed voters receive pol- icy benefits because they are in a better position to hold politicians account- able. Informed voters know whom to credit for relief and only reelect those who have provided sufficient relief. Politicians give relief with an eye to reelection. They realize that providing relief can increase support, more so in areas where people are better informed. The relief can only swing 20 David Strömberg and James M. Snyder, Jr. voters in three groups: those who need relief, those who are informed that this particular politician is responsible for the relief, and those who are sufficiently unattached ideologically that relief funds may swing their votes. Consequently, expenditures will increase in the interaction of these charac- teristics. The following equation describes relief expenditures to a county: (relief per capita) = const. · (share voting) · (share informed) · (share in need of relief) · (share marginal voters) Radio penetration matters directly because it increases the share of vot- ers who are informed, and also indirectly because it increases the share who votes. The above equation implies that relief expenditures are increasing in response to increasing radio penetration. It also implies that in areas where more people are informed, relief should be more responsive to (more corre- lated with) need. Strömberg estimates a linear form of the above equation, after taking logarithms. However, Besley and Burgess (2002) later explicitly tested for the positive interaction effect. Empirically, Strömberg finds that areas where many people have radios are more successful in attracting unemployment relief. These effects are particularly large in rural areas. More relief is also spent in areas where more people vote. Further, voter turnout in both presidential and guberna- torial elections is increased by rising radio use. How do we know that media access caused the increased voter turnout and higher relief spending? Obvi- ously, radio ownership is related to socioeconomic factors that may also drive relief. Additional evidence is that unemployment relief and turnout are also systematically higher in counties where the physical conditions for radio reception are good. It is harder to find alternative explanations for this fact. While Strömberg focuses on how successful different groups of voters are in attracting government funds, Besley and Burgess (2002) focus on the conflict between all voters and policy makers. Using a model similar to Strömberg's, they find that relief should increase in the share who votes, is informed, is in need of relief, and lives in areas where the election is close. Empirically, Besley and Burgess analyze the interaction-term between the share who is informed and the share who is in need of relief. A positive term means that relief expenditures are more responsive to need in places with many informed people. In a study of Indian states from 1958­92, Besley and Burgess find that both public food distribution and calamity relief were more responsive to needs in states where newspaper penetration was high. The Media's Influence on Public Policy Decisions 21 In particular, circulation of newspapers in local languages seems to be a significant factor. As in the United States, the conditions in India are favorable for the media to have an impact. India has a relatively free and independent press that often covers the events that trigger relief efforts, such as dramatic natu- ral disasters and food shortages. These events also evoke reader interest. In fact, disaster relief in India is one of the archetypical examples of press influence on government policy. Sen (1990) notes that in the case of Indian famines: An uncensored and active news media can have a very important part to play in alerting the government as well as the public about impend- ing threats of famines, by reporting early cases of starvation which often serve as tell-tale indicators of things to come, unless prevented by decisive public action. Other papers have studied more directly whether politicians seek ben- efits for themselves at the expense of the voters. For example, Svaleryd and Vlachos (2005) find that Swedish municipalities with higher local news- paper penetration have lower public financial support of political parties, a legal form of "political rents." Ferraz and Finan (2005a, 2005b) study Brazilian municipalities and find that the presence of a local radio sta- tion makes voters react more strongly when corruption by their mayor is exposed. They also find evidence suggesting that mayors tend to be less cor- rupt in such areas, but the estimates are imprecise and statistically insignifi- cant. Olken (2006) finds that corruption in Indonesian village-level road projects is uncorrelated with television access. This is not surprising, since television broadcasts are national in scope and rarely cover issues of rel- evance to individual villages. In fact, Olken argues that television is more likely to increase local corruption than reduce it, since watching television makes people politically passive. A number of chapters in this book study the influence of mass media on political outcomes, such as voter turnout or vote choice. Results from many of these chapters have policy implications. As noted in Strömberg (2004b), U.S. counties where voter turnout was higher in the 1930s received more unemployment relief. The ability of the media to increase or decrease turn- out may change government policies. In addition, policy may be affected if the media change the election outcome. In general, the media may influ- ence policy both by inducing politicians to change policy stands and by changing the types of politicians elected. In areas where politicians have 22 David Strömberg and James M. Snyder, Jr. strong personal motivations, such as abortion or corruption, the media may mainly change policy by changing the politicians elected; the opposite may be true in policy areas where politicians are less strongly attached to particular positions (Strömberg 1999). These studies have looked at differences in media access, taking media content as given. Media content, defined as the amount of political cover- age, may have effects as well. Effects of Active Press Coverage of Politics The media must be free and independent in order to actively cover politics. A number of papers have investigated how these characteristics correlate with good governance at the country level. Some key findings are that cor- ruption is typically lower in countries with a free press (Ahrend 2002; Bru- netti and Weder 2003), with fewer state-owned newspapers (Djankov and others 2003), and with more foreign ownership of the media (Besley and Prat 2006). While suggestive, however, it is difficult to conclude from these studies that a free and independent press lowers corruption. Whether the media are free is influenced by governmental actions, and more corrupt governments have stronger incentives to silence the media. Accordingly, correlations between corruption and media structure may simply reflect that these variables are jointly determined, as discussed by Besley and Prat (2006). One promising approach to identifying effects that are more likely causal is to focus on variation in the intensity of political coverage within a country. However, also within countries, it is difficult to identify the effect media coverage has on policies, voters, and politicians. The relationships among media coverage and political knowledge, politicians' actions, and policy outcomes are complex. Many forces are at work, and it is hard to control for them well. For example, voters know more about politicians who are frequently in the press. This not only reflects the effect of the arti- cles on knowledge, but also that the media tend to cover politicians who are better known. Similarly, the implementation of federal projects or the award of grants attracts media attention and coverage. For this reason, a correlation between, say, media coverage of a certain area and government funds to that area is unlikely to capture the effect of media coverage on government funds. SnyderandStrömberg(2006)analyzetheeffectofhavingapressactively covering politics in the United States, using the fact that the amount of cov- The Media's Influence on Public Policy Decisions 23 erage is driven to a large extent by the coincidental match between media markets and political districts. Snyder and Strömberg (2006) study elec- tions for the U.S. House of Representatives. They find that in areas where the media are expected to cover congressional politics closely, voters are better informed, politicians act more frequently and perhaps more in their constituencies' interests, and more federal funds per capita are spent. To identify media effects, Snyder and Strömberg (2006) use a measure of the match, or congruence, between media markets and U.S. congres- sional districts. The measure is based on the share of a newspaper's read- ership that lives in a certain congressional district. Intuitively, newspaper coverage of a congressman should be increasing in this readership share. Snyder and Strömberg find that this is indeed the case, after analyzing the content of a large set of U.S. newspapers. Since more than one newspaper sells in each district, they define congruence as the circulation-weighted average of readership for all of the newspapers sold in a district. It is higher in rural areas; once this is controlled for, however, it is not strongly related to other variables important for voter information or the other variables of interest. Snyder and Strömberg (2006) find that more people are able to cor- rectly name the candidates in the congressional elections in areas where congruence is high. The estimates imply that respondents are around 25 percent more likely to correctly name candidates in congressional races in areas where congruence is 1 compared to races in which it is 0. To arrive at this estimate, Snyder and Strömberg analyze the answers to open-ended questions asking respondents the names of the candidates running in their congressional races, from the American National Election Studies (www. electionstudies.org) over the period 1990­98. Why should we believe that these estimates measure the causal effects of news coverage? The correlations could arise because congressional repre- sentatives or voters are systematically different in areas where congruence is high. To address the concern that candidates are different, Snyder and Strömberg (2006) include a dummy variable for each congressional race. This way Snyder and Strömberg essentially compare only respondents who live in the same congressional district and are reporting about the same congressional race, involving the same pair of candidates--but from differ- ent counties with different levels of congruence. To address the concern that voters are different, Snyder and Strömberg include county-fixed effects and only allow congruence to vary because of changes in congressional district boundaries due to redistricting between 1990 and 1992. The estimates are 24 David Strömberg and James M. Snyder, Jr. similar across all specifications and are positive and statistically significant in all cases. Finally, Snyder and Strömberg find no correlation between con- gruence and the measure of respondents' more general political knowledge, such as the ability to name their senators. Snyder and Strömberg also find that evidence that high congruence leads representatives to work harder and to work more in line with the inter- ests of constituents. Representatives from highly congruent districts appear more frequently as witnesses in congressional committee hearings. These representatives are also less loyal to their party's leadership when voting in Congress, perhaps because they give more consideration to their constituen- cies' interests. Finally, they are less likely to serve on committees that have little influence over targeted spending (so-called "policy committees"). Does this behavior by the representatives lead to more federal funds being spent in highly congruent areas? To test this, Snyder and Strömberg assembled annual data on federal expenditure allocations across counties from the Consolidated Federal Funds Report (U.S. Census Bureau, Gov- ernments Division) for the period 1983­2004. The expenditures analyzed include grants, procurement contracts, salaries and wages, direct payments for retirement and disability, and other direct payments--loans, insurance, and social security payments are excluded. The total value of the expendi- tures studied is roughly 10 percent of the U.S. gross domestic product. Snyder and Strömberg (2006) find that more congruent counties receive significantly more federal funds. In magnitude, going from minimal to maximum congruence is associated with a 10 percent increase in federal funds. Snyder and Strömberg try to carefully account for observable differ- ences among counties. They also reach similar results when they compare only neighboring-county pairs on different sides of congressional bound- aries. Finally, including county-fixed effects causes the estimated effect of congruence to drop to about 5 percent, but it is still statistically significant. One possible reason for the drop is that expenditures adjust slowly over time, so Snyder and Strömberg may only have observed part of the full adjustment to changes in congruence during the period studied. The Effects of Publishing Individual News Stories The final question to consider is how the news content matters. Can news editors, by publishing a news story, influence government policy? Eisensee and Strömberg (2007) analyze this question, specifically, whether U.S. relief The Media's Influence on Public Policy Decisions 25 to natural disasters abroad is affected by whether the disaster was covered by the television network news. Although it is widely believed that network news is an important determinant of government policy, little conclusive evidence has been produced (Graber 1997; Robinson 1999). Convincing evidence is hard to find, since disaster relief and news cov- erage would be correlated even if there were no effect. This happens since more severe disasters are both more likely to be in the news and to receive relief, and because political agendas drive both media coverage and relief. For example, one of the most cited cases of media influence on foreign policy is the 1992 U.S. intervention in Somalia. Daily, televised pictures of starving children in Somalia allegedly compelled the U.S. government to intervene for humanitarian reasons. However, Mermin (1997, 388), writes, "[B]efore television made the decision to cover the crisis in Somalia, influ- ential politicians had spoken out on it, indicating to journalists [...] that Somalia constituted a significant concern of American foreign policy and warranted consideration for space in the news." Political agendas may drive media coverage rather than vice versa. Eisensee and Strömberg (2007) study 5,212 natural disasters that took place worldwide between 1968 and 2002. These disasters claimed 63,000 lives and affected 125 million people per year. Data on the disasters have been collected in the Emergency Disaster Database as provided by the Cen- tre for Research on the Epidemiology of Disasters. Eisensee and Strömberg (2007) study whether these disasters received relief from the United States Agency for International Development Office of Foreign Disaster Assistance (OFDA). Approximately one in five disas- ters received relief. To see whether news coverage affected this decision, Eisensee and Strömberg also collect data on whether the disaster was cov- ered by the U.S. television network news. They used the Vanderbilt Televi- sion News Archives, which has compiled data on the content of the evening news broadcasts of the major U.S. television networks (ABC, CBS, NBC), and cable station CNN since 1968. Disasters covered in the news are signifi- cantly (around 30 percent) more likely to receive relief than other disasters. This partly reflects that more severe disasters both receive relief and are in the news. To test whether news coverage influences policy, Eisensee and Ström- berg (2007) use the fact that disaster news may be crowded out of the news broadcasts by other newsworthy material. For example, it seems unlikely that disasters that happen to occur at the same time as the Olympic Games 26 David Strömberg and James M. Snyder, Jr. are systematically more severe or higher on political agendas than other disasters, after accounting for seasonal effects. However, news coverage of a disaster may be crowded out by news coverage of the Olympics. If so, then disasters that occur during the Olympics, while otherwise similar to disasters occurring at other times, will nonetheless receive less coverage. If news affects policy, then the disasters that occur during the Olympics will receive less relief. This is indeed what Eisensee and Strömberg (2007) find. There is no correlation between the occurrence of Olympic games and the severity of natural disasters; however, disasters occurring at the same time as the Olym- pics receive less news coverage and are also less likely to receive OFDA relief. The crowding out of disaster news makes sense given that the Olympics are covered so thoroughly by the networks. In the sample period, 2,443 news stories have "Olympic" in the headline, covering the 18 Olympic games, ranging from the 1968 summer Olympics in Mexico City to the 2002 winter games in Salt Lake City. Eisensee and Strömberg (2007) find similar results using another, more general, measure of the amount of available news material, which they call "news pressure." This consists of the median number of minutes a news broadcast devotes to the top three news segments in a day. When a large media event takes place, that story is usually placed as one of the first news stories in a broadcast, and more time is devoted to it. For instance, on October 3, 1995, a jury found O. J. Simpson not guilty of two counts of murder. That night, ABC, CBS, and NBC devoted all of their first three news segments to that story. The top three news segments comprised an average of 16 minutes and 30 seconds--the highest value of that year. The Simpson verdict effectively crowded out other news. NBC only covered this story, while CBS also reported on one other story, and ABC included only four other stories. This suggests that the amount of time devoted to the first three news segments is a good indicator of how much newsworthy material is available on a given day. Eisensee and Strömberg use the median value across news broadcasts in a day, rather than the mean, to reduce the effect of measurement error in the reported time for news segments. Eisensee and Strömberg (2007) argue that a causal effect of news is the only plausible explanation for the negative correlation between the availability of other newsworthy material and disaster relief. Under this assumption, the impact of news on relief can be estimated using the instru- mental variables technique. The estimated effect of news on disaster relief is significant. The Media's Influence on Public Policy Decisions 27 TABLE 2.1 Newsworthiness of Disasters Casualty ratio to make news coverage equally likely Disaster type Continent Volcano 1 Europe 1 Earthquake 2 South and Central America 3 Fire 12 Asia 43 Storm 280 Africa 45 Flood 674 Pacific 91 Landslide 882 Epidemic 1,696 Drought 2,395 Cold wave 3,150 Food shortage 38,920 Source: Eisensee and Strömberg 2007. The effect of news biases relief in favor of certain disaster types and regions. To have the same estimated probability of being covered by the television network news, a drought must have 2,395 times as many casual- ties as a volcano, all else being equal (country, year, month, and number affected) (see table 2.1). This implies that media coverage biases U.S. relief in favor of volcano and earthquake victims at the expense of drought vic- tims. Similarly, it requires 45 times as many people to be killed in a disaster in Africa to achieve the same probability of media coverage as a disaster in Europe. Thus, media coverage biases U.S. relief against African disaster victims. There may be an underlying affinity between the United States and Europe that makes relief more likely in Europe, even absent media. How- ever, the unequal media coverage aggravates this tendency. Conclusions and Discussion We have examined three questions: how policy is affected by differences in media access across groups, how policy is affected by the amount of media coverage of politics, and how policy is affected by differences in news content. The first question is the most thoroughly researched. A common find- ing is that mass-media access empowers people politically in the sense of increasing their benefits from different government-run programs. This 28 David Strömberg and James M. Snyder, Jr. type of influence has been found in a variety of government programs, for example, unemployment relief in the United States, public food provision and calamity relief in India, and educational spending in Uganda. These results imply that differences in access to the mass media may induce dif- ferences in political benefits. For example, poor people use media less and may end up benefiting less from public spending programs. The results also point to a way to avoid part of this problem by increasing media access in this group. However, effects of the media have only been found under cer- tain conditions. Providing poor people with media access may have little impact on policies if the media do not cover issues that are politically salient to the poor, or if they do but this information is not of interest to them. To address the last point, packaging of news may be important. For example, news that was politically relevant to American farmers in the 1930s was packaged in programs that also contained weather and market reports. Regarding the second question, having a press that actively covers poli- tics seems to improve electoral accountability. Snyder and Strömberg (2006) find that more active media coverage of the U.S. House of Representatives makes voters better informed and makes the representatives work harder for their constituencies, resulting in more policy benefits for the voters. Can these results be generalized to other settings or other countries? This seems all the more likely given that the media have been found to influence policy in countries such as India and Uganda. However, the size of the effects could of course differ significantly, depending on the political environment. The results in Snyder and Strömberg (2006) imply that a mismatch between media markets and political jurisdictions lowers political account- ability. For example, having a mainly international media, with a large share of viewers living abroad, may be bad for accountability. Similarly, the rise of television as the dominant media in many countries may have lowered political accountability at more local levels, since television market areas are typically large--even larger than newspaper markets--for technologi- cal reasons.2 There are other types of mismatch as well. For example, while many important political decisions in Europe are made at the level of the European Union, the media markets are mainly national, and so is most of the political coverage. Practically, media regulators might be advised to consider how media mergers influence the match between media markets and various political districts. Regarding the final question, media content does seem to influence policy. The evidence discussed above suggests that having a disaster covered The Media's Influence on Public Policy Decisions 29 by the network news influences U.S. government relief. Disturbingly, the Eisensee and Strömberg (2007) study finds that the mass media make U.S. foreign policy dependent on events such as the O. J. Simpson trial and also bias foreign policy toward dealing with spectacular events. Can these findings be applied to other policy areas? It seems likely that the mechanisms of media influence--information, agenda setting, or pub- licity management--would be equally important in other policy areas. In the same way that the media draw attention to some natural disasters, they also draw attention to problems, for example, in health care and education (see Cook and others 1983). An indication that media coverage may affect other policy areas is that media access--for example, radio access--affected unemployment benefits in the United States of the 1930s. It seems likely that variation in information flows, not only from access but also in cover- age, would matter; for example, regions where the radio covered unemploy- ment relief less would receive less relief. If media content affects policy, then the media may be one additional reason why the poor are politically disadvantaged. Issues that are salient to the poor may be covered less thoroughly by the media, because the poor do not have much purchasing power and are less attractive to advertisers, the media's main source of revenue. In other words, various biases in what the media cover may influence policy, as argued by Strömberg (2004a). While the literature has produced some important findings and our knowledge may be starting to accumulate, there are obvious areas where we know little. For example, does bias in the news affect policy? Are local policies affected by the relative amount of space the media devote to, say, welfare programs and property taxes? Existing studies of the effects of media bias look instead at the impact on voter preferences and behavior (DellaVigna and Kaplan 2007). The next step--policy outcomes--has yet to be taken. Notes 1. Francken, Minten, and Swinnen (2005) conduct a similar analysis in Madagascar. 2. Relatedly, Gentzkow (2006) finds that the rising use of television in America crowded out newspapers and this depressed voter turnout in congressional elections; George and Waldfogel (2006) find similar effects of a national news- paper crowding out local newspapers and depressing voter turnout in local elections. However, how increasing voter turnout relates to accountability is not obvious. 30 David Strömberg and James M. Snyder, Jr. References Ahrend, Rudiger. 2002. "Press Freedom, Human Capital, and Corruption." Work- ing Paper No. 2002­11, DELTA (Ecole normale supérieure). Paris. Besley, Timothy, and Robin Burgess. 2002. "The Political Economy of Government Responsiveness: Theory and Evidence from India." Quarterly Journal of Econom- ics 117(4): 1415­51. Besley, Timothy, and Andrea Prat. 2006. "Handcuffs for the Grabbing Hand? Media Capture and Government Accountability." The American Economic Review 96(3): 720­36. Brunetti, Amyo, and Beatrice Weder. 2003. "A Free Press Is Bad News for Corrup- tion." Journal of Public Economics 87(7­8): 1801­24. Brunner, Edmund deS. 1935. A Symposium on the Relation of Radio to Rural Life. New York: The Radio Institute of the Audible Arts. Cook, Fay Lomax, Tom R. Tyler, Edward G. Goetz, Margaret T. Gordon, David Protess, Donna R. Leff, and Harvey L. Molotch. 1983. "Media and Agenda Set- ting: Effects on the Public, Interest Group Leaders, Policy Makers, and Policy." Public Opinion Quarterly 47(1): 16­35. DellaVigna, Stefano, and Ethan Kaplan. 2007. "The Fox News Effect: Media Bias and Voting." Quarterly Journal of Economics 122(3): 1187­1234. Djankov, Simeon, Caralee McLiesh, Tatiana Nenova, and Andrei Shleifer. 2003. "Who Owns the Media?" Journal of Law Economics 46(2): 341­83. Eisensee, Thomas, and David Strömberg. 2007. "News Floods, News Droughts, and U.S. Disaster Relief." Quarterly Journal of Economics 122(2): 693­728. Ferraz, Claudio, and Frederico Finan. 2005a. "Reelection Incentives and Political Corruption: Evidence from Brazilian Audit Reports." Unpublished paper, Uni- versity of California at Berkeley. ------. 2005b. "Exposing Corrupt Politicians: The Effect of Brazil's Publicly Released Audits on Electoral Outcomes." Unpublished paper, University of California at Berkeley. Francken, Nathalie, Bart Minten, and Johan F.M. Swinnen. 2005. "The Impact of Media and Monitoring on Corruption in Decentralized Public Programs: Evi- dence from Madagascar." Unpublished paper. Katholieke Univeriteit Leuven. George, Lisa M., and Joel Waldfogel. 2006. "The New York Times and the Market for Local Newspapers." The American Economic Review 96(1): 435­47. Gentzkow, Matthew. 2006. "Television and Voter Turnout." Quarterly Journal of Economics 121(3): 931­72. Graber, Doris A. 1997. Mass Media and American Politics. Fifth Edition. Washington, DC: CQ Press. McCombs, M., and D. Shaw. 1972. "The Agenda-Setting Function of Mass Media." Public Opinion Quarterly 36(2): 176­87. Mermin, Jonathan. 1997. "Television News and American Intervention in Somalia: The Myth of a Media-Driven Foreign Policy." Political Science Quarterly 112(3): 385­403. Olken, Benjamin A. 2006. "Do Television and Radio Destroy Social Capital? Evidence from Indonesian Villages." Working Paper No. W12561. National Bureau of Economic Research, Cambridge, MA. The Media's Influence on Public Policy Decisions 31 Reinikka, Ritva, and Jakob Svensson. 2005. "Fighting Corruption to Improve Schooling: Evidence from a Newspaper Campaign in Uganda." Journal of the European Economic Association 3(2­3): 259­67 Robinson, Piers. 1999. "The CNN Effect: Can the News Media Drive Foreign Policy?" Review of International Studies 25(2): 301­09. Sen, Amartya. 1990. "Public Action to Remedy Hunger." Arturo Tanco Memorial Lecture, August 2, 1990. London. http:www.thp.org/reports/sen/sen890.htm Snyder, James M., Jr., and David Strömberg. 2006. "Media Effects on Politics." Unpub- lished manuscript, Massachusetts Institute of Technology, Cambridge, MA. Strömberg, David. 1999. "The Politics of Public Spending." Ph.D. dissertation, Princeton University. ------. 2004a. "Mass Media Competition, Political Competition, and Public Pol- icy." Review of Economic Studies 71(1): 265­84. ------. 2004b. "Radio's Impact on Public Spending." Quarterly Journal of Economics 119(1): 189­221. Svaleryd, Helena, and Jonas Vlachos. 2005. "Political Rents in a Non-Corrupt Democracy." Unpublished manuscript, Stockholm University. U.S. Census Bureau. Consolidated Federal Funds Report. Government's Division. 3 National Media and Local Political Participation: The Case of the New York Times Lisa M. George and Joel Waldfogel A growing literature documents the importance of shared consumer preferences in shaping the characteristics of media products available in local markets. When aggregation of consumers over large areas is dif- ficult because information distribution is costly, available products tend to be targeted toward local tastes. As information technology reduces the costs of aggregating consumers and expands the geographic range of tra- ditionally local media, the link between media offerings and local tastes weakens. The availability of outside media is clearly beneficial to indi- viduals whose tastes are not well-represented in local markets. However, individuals attracted to outside products might consume less local infor- mation. One result is that these individuals may become less informed about local policies and local candidates than individuals who continue to consume local media offerings. Distraction from local affairs is a potential negative consequence of modern information technology. While concern about the consequences of outside media expanding into local markets is not new, empirical work has only recently begun to systematically document the effect of media technology on local media products and local behavior. Research to date suggests that entry of the national media does affect consumption of local products. The growth of radio, television, and foreign-language newspapers has also been shown to affect local political affairs. This chapter estimates whether outside media narrowly targeted at highly educated consumers distract these individuals from local political participation. Our vehicle for studying the effect of targeted media on political par- ticipation is the national expansion of the New York Times (hereafter referred to as the Times). Although the Times began publishing a national edition in 33 34 Lisa M. George and Joel Waldfogel 1980, the company initiated a major national expansion in the mid 1990s, increasing home delivery in over 100 markets between 1996 and 2000. We use this expansion to estimate whether the Times depresses political partici- pation among its readers. Using longitudinal data on Times penetration and political participation, we find that highly educated readers targeted by the Times are less likely to vote in congressional elections in areas with greater penetration. These results contribute to the debate over localism in media, specifi- cally, the merits of policies to promote local content and local program- ming. The potential for outside media to weaken interest in local affairs may provide a rationale for promotion of locally tailored products. Yet the results of this research highlight the importance of targeting in media con- sumption. Entry of outside products does not need to reduce the supply of local media to dampen its impact. Supporting content demanded by groups with specialized tastes may be more important than promoting local con- tent alone. This research also supports a broader literature on the role of media markets in public policy. While the effect of media messages on political outcomes is extensively explored in a large body of literature, little empir- ical evidence documents how the underlying economics of media mar- kets can affect the behavior of voters and policy makers. As information technology continues to evolve, understanding how technological change translates into social change can help individuals and policy makers con- front new challenges. The chapter proceeds as follows. Section one outlines our approach and links our work to the literature. Section two describes the data used in the analysis. Section three presents our empirical strategy and results. Section four concludes the chapter. Background and Literature We estimate the effect of national media on local political participation by studying the effect of the Times' national expansion on voting in congres- sional elections. While many national media products have the potential to divert attention from local affairs, the Times offers a particularly suitable platform for evaluating the effect of national media for several reasons. First, the Times is a narrowly targeted newspaper aimed at highly educated readers. Readership surveys indicate that the paper strongly appeals to col- lege-educated consumers, particularly outside of New York.1 The narrow National Media and Local Political Participation 35 targeting of the paper is useful in our identification strategy. Second, George and Waldfogel (2006) demonstrate that national expansion has reduced local newspaper circulation among highly educated readers targeted by the Times. With evidence that this targeted group has already been distracted from local newspapers by the Times, it is a logical step to look for distraction from political affairs among this same group. Finally, the national expansion that began in the 1990s offers a useful empirical environment for studying the effect of national media on politi- cal participation. National expansion was accomplished through satellite transmission of the paper to printing plants across the county combined with a network of home delivery contracts with local newspapers. Between 1996 and 2000, the Times substantially expanded home delivery in more than 100 markets. The New York Times Company does not systematically report the timing and location of expansion markets, so we cannot con- struct instruments for Times penetration. Instead, we take the national expansion strategy as a broad source of variation in Times' penetration that is exogenous to interest in local media and local political affairs. While this is a strong assumption, a number of robustness checks documented in George and Waldfogel (2006) support its validity. This chapter contributes to a growing body of research on the eco- nomic factors that determine product position and product variety in media markets as well as the consequences of media markets for political behavior. Waldfogel (2003), George and Waldfogel (2003), and Waldfogel (2004) demonstrate the importance of shared preferences in determining both the nature of products available in a market and who consumes them in radio, newspaper, and television, respectively. Works by Oberholzer-Gee and Waldfogel (2005, 2006), Prat and Strömberg (2005), and Gentzkow (2006) establish links among shared preferences, media offerings, and political participation. Strömberg (2004) shows that the media can also drive policy attention toward targeted groups. Data We analyze the effect of the Times' penetration on political participation using individual-level voting and demographic data drawn from the 1994­ 2000 Current Population Surveys (CPSs). We link individual voting data to market-level Times' circulation in these years. Circulation data are available at the market level from the Audit Bureau of Circulations (ABC), an indus- try association that audits circulation statements for use by advertisers. 36 Lisa M. George and Joel Waldfogel We study circulation outside of the home market, excluding the New York metropolitan area from the analysis. Circulation is reported for Met- ropolitan Statistical Areas (MSAs) defined by the U.S. Census, so we link MSA circulation of the Times to CPS voting data for residents of the MSAs. Geographic reporting and MSA definitions have changed over time, so we include only those markets with data available over the entire period of our study. The CPS data include voting information and socioeconomic charac- teristics for about 80,000 eligible voters each year. Market identifiers are available for about 55,000 individuals each year. Differences in market definitions between ABC and CPS data, changes in market definitions over time, and restriction of the sample to eligible voters old enough to have completed their education reduce the usable sample to about 35,000 indi- viduals in 178 markets. While we cannot link individuals over time, we treat CPS data as a market-level panel. We note that CPS data are collected only in even-numbered years. Accordingly, we cannot study most local elections that take place in odd years; instead, we study congressional races that take place in even- numbered years. However, all that is required for distraction to occur is that the Times offers less coverage of congressional races than that offered by local newspapers. Simple counts of references to individual members of the U.S. House of Representatives in the Times versus local newspapers show overwhelmingly that this is the case.2 Sample Characteristics Table 3.1 reports summary statistics for circulation, voting, and demo- graphic data. Times' circulation varies considerably across markets, with penetration (the ratio of circulation to population) ranging from 0.003 percent at the 1st percentile to 1.6 percent at the 99th percentile. (We study circulation outside of the home market, hence the New York consolidated metropolitan statistical area is not included in the table or analysis that follow). The tendency to vote differs across education groups and between presidential and nonpresidential elections. The difference in the tendency to vote between individuals with and without a college degree is about 22 percentage points in both presidential and nonpresidential years. Overall turnout is between 13 percentage points and 16 percentage points higher in presidential than nonpresidential elections, with average turnout 68 per- cent and 55 percent, respectively. National Media and Local Political Participation 37 TABLE 3.1 Sample Statistics MSA New York Times sales Year N Mean SD 1% 99% Per capita NYT sales 1994 178 0.0020 0.0026 0.003% 1.15% Per capita NYT sales 1996 178 0.0020 0.0028 0.003% 1.38% Per capita NYT sales 1998 178 0.0021 0.0028 0.003% 1.37% Per capita NYT sales 2000 178 0.0023 0.0031 0.003% 1.64% Source: Audit Bureau of Circulations. 1994 1996 1998 2000 N = 39,666 N = 36,503 N = 35,891 N = 32,875 Individual voting data Mean SD Mean SD Mean SD Mean SD Voting probability (all) 0.56 0.56 0.67 0.47 0.54 0.56 0.70 0.46 Voting probability (no degree) 0.50 0.50 0.61 0.49 0.48 0.50 0.49 0.48 Voting probability (degree) 0.71 0.45 0.82 0.38 0.68 0.47 0.86 0.35 Fraction college degree 0.26 0.44 0.28 0.45 0.29 0.45 0.30 0.46 Source: Current Population Surveys. Empirical Strategy and Results This section first describes the estimation strategy and presents results. An examination of the robustness of the results follows. Empirical Strategy We adopt three approaches to estimating the effect of Times' penetra- tion on local behavior. The first approach uses cross-sectional data to identify the effect of penetration in a market on political participation among the target relative to the non-target audience. The simplest cross- sectional approach to documenting the effect of penetration would be to ask whether educated persons are less likely to vote in markets with higher pen- etration. Yet, the tendency for members of the target audience to vote may be determined by market characteristics related to local penetration, such as the quality of other local media, casting doubt on a causal interpretation of the correlation. However, if unobserved determinants of local behavior are constant across the target and non-target audience, then the inclusion of market fixed effects allows us to identify the effect of the Times from the relationship between penetration in a market and the local behavior 38 Lisa M. George and Joel Waldfogel among the target audiences relative to the non-target group. We first esti- mate the effect of penetration on voting using this approach. With longitudinal data, we can separately identify the effect of Times' penetration on the target and non-target audiences. This second approach produces valid inference as long as the market fixed effect is common to both groups. Our third approach takes into account that unobservable determinants of local behavior may not be common across the target and non-target audience. Using longitudinal data, we estimate models that allow for group-specific MSA fixed effects. We then identify the effects of Times penetration on voting from the relationship between changes in pene- tration and changes in local behavior for the target relative to the non-target group. Stated another way, we examine the relationship between changes in penetration and changes in the voting gap between individuals with and without a college degree. This method produces valid inference even if the unobserved determinants of local newspaper readership and voting are not common across the target and non-target groups. Because the Times includes less coverage of House members outside of New York than local and regional papers, we expect the Times to depress political participation in nonpresidential elections. We do not expect to observe distraction in presidential election years. More specifically, we first examine the effect of Times' penetration on the probability that an individual votes using cross-sectional data with MSA fixed effects, estimating the following: Vi = 0 + 1NYTM + 2edi + 3NYTM edi + M + i (1) where Vi indicates whether the individual voted, edi indicates whether the individual has a college degree, mM is an MSA-specific time-constant effect, and ei is an individual error. With this cross-sectional specification, the effect of Times penetration, g3, is identified by the differential effect of penetration on the voting tendencies of the target relative to the non-target group. The first four columns in table 3.2 show results of cross-sectional regressions for each year in the sample. The effect on voting among the target relative to the non-target group, g3, is negative and significant in 1994 and 1998. The magnitude of the effect is smaller in 1996 and 2000, insignificantly differ- ent from zero in 1996. The results indicate that an increase in Times penetra- tion of 0.01, which corresponds approximately to the difference between the 5th and 95th percentile, reduces the gap in the tendency to vote between college- and non-college­educated individuals by about 6 percentage points in congressional elections. National Media and Local Political Participation 39 TABLE 3.2 Does the New York Times Depress Voting among the College Educated? Voting probability 1994 1996 1998 2000 Pooled Pooled (1) (2) (3) (4) (5) (6) Per capita NYT (g1) -- -- -- -- ­3.045 ­2.603 -- -- -- -- (1.02) (0.87) NYT*pres race -- -- -- -- ­2.173 ­1.931 -- -- -- -- (1.59) (1.47) NYT*high ed (g3) ­7.733 ­0.775 ­5.103 ­2.904 ­6.537 ­9.376 (3.10)** (0.41) (2.43)* (2.45)* (3.76)** (1.85) NYT*high ed*pres race -- -- -- -- 4.012 3.295 -- -- -- -- (2.15)* (1.85) High ed (g2) 0.569 0.639 0.566 0.569 0 0.470 -- (18.73)** (15.22)** (12.66)** (12.66)** (23.57)** -- High ed*pres race -- -- -- -- 0.007 0.008 -- -- -- -- (0.91) (0.94) 1996 year dummy -- -- -- -- 0.101 0.101 -- -- -- -- (11.55)** (11.53)** 1998 year dummy -- -- -- -- ­0.049 ­0.049 -- -- -- -- (6.25)** (6.28)** 2000 year dummy -- -- -- -- 0.105 0.105 -- -- -- -- (12.00)** (12.05)** Constant (g0) ­0.407 ­0.340 ­0.228 ­0.266 ­0.248 ­0.145 (13.03)** (7.38)** (5.14)** (4.70)** (9.52)** (6.45)** Other variables See See See See See See notes notes notes notes notes notes Fixed effects MSA MSA MSA MSA MSA MSAxEd MSA's 178 178 178 178 178 178 Observations 39,666 36,503 35,891 32,875 144,935 144,935 Source: Authors. Notes: Linear probability models with standard errors clustered by MSA for cross-sectional estimates and MSA × Year for pooled estimates. T-statistics in parentheses: * significant at 5% level; ** significant at 1% level. Constants in fixed effects regressions represent the average value of the fixed effects. All regressions include year dummies, controls for race, sex, and age as well as categorical variables for income and education. Pooling years of data allows separate identification of the effect of Times' penetration on individuals with and without a college degree (both g1 and g3). This approach requires the assumption that both the target and non-target groups have the same unobservable effect mM. We can relax that 40 Lisa M. George and Joel Waldfogel assumption and include group-specific MSA fixed effects, again estimating the effect of changes in Times' penetration on changes in the tendency of college-educated relative to non-college­educated individuals to vote. We interact the baseline and high education Times' effect with a dummy vari- able for presidential races to determine whether the distraction shrinks in presidential years. The last two columns of table 3.2 present results of the longitudinal specifications. Results are consistent with the cross-sectional estimates. The effect on individuals without a college degree is negative but not statisti- cally significant. As in the cross-section results, the effect on highly edu- cated individuals is negative and statistically significant. Increasing sales by 0.01 reduces the gap in the tendency to vote between college- and non- college­educated individuals by about 6 percentage points, from 22 percent to 16 percent. Allowing for group-specific fixed effects in column six, the distraction effect is higher. Increasing sales by 0.01 reduces the gap by 9 percentage points. Also, as suggested in the cross-sectional specifications, the distraction effect interacted with presidential race is positive. Distrac- tion drops significantly in presidential years. The effects in table 3.2 are large. Even a more modest increase in pen- etration of 0.005 reduces the difference in the tendency to vote by 3 per- centage points to 5 percentage points. Another way to consider this is to note that the fraction of individuals with a college degree in the CPS is about 0.25 percent. With an assumption that all copies of the Times are sold to individuals with a college degree, an increase of 0.01 paper per capita raises per capita circulation among the target group by 0.04. The estimates in table 3.2 thus indicate that an increase of 0.01 paper per capita among the target group would reduce the voting differential between college- educated and non-college­educated individuals by 6 ¥ 0.0025 = 0.015, which, loosely speaking, is more than one vote per paper. Given the size of the effect, we consider the robustness of our results to other factors that might differentially reduce voting among college-educated individuals. Robustness Our identification strategy ensures that the documented effects are not driven by unobservable market characteristics that may differ by education group but do not change over time. In this section we consider the robust- ness of our results to market factors that may be correlated with changes in Times' penetration and differentially affect the tendency of college and National Media and Local Political Participation 41 non-college­educated individuals to vote. A large literature in political sci- ence considers factors that affect turnout, and close races consistently play an important role. Although we do not expect changes in the closeness of races to be correlated with the timing of the Times' expansion, we repeat the analysis of table 3.2, including categorical variables and the margin of victory for presidential, gubernatorial, and senate elections reported in the Almanac of American Politics3 (Barone and Ujifusa 1997; Barone, Cohen, and Ujifusa 2001). We include an education and presidential race interaction to allow the effect of a close race to differ by education and across presi- dential and nonpresidential elections. Coefficients of interest are virtually unchanged with these controls. Results with pooled years are presented in columns one and two of table 3.3, corresponding to the MSA and group- specific MSA specifications in table 3.2. TABLE 3.3 Robustness of Voting Results (Statewide Elections, Residency, and Internet) Voting probability Pooled Pooled Pooled Pooled Pooled Pooled (1) (2) (3) (4) (5) (6) Per capita NYT ­3.450 ­2.649 ­4.472 ­3.636 ­3.580 ­3.026 NYT*pres race ­2.188 ­2.110 ­2.586 ­2.522 ­3.063 ­2.969 NYT*high ed ­5.916** ­9.379* ­5.872** ­9.695* ­6.527** ­9.483* NYT*high ed*pres race 4.747* 4.642** 4.479* 4.510** 5.236* 5.153** High ed 0.442** -- 0.468** -- 0.471** -- High ed *pres race 0.036* 0.054* 0.083** 0.084** 0.007 0.007 1996 year dummy 0.159** 0.153** 0.142** 0.142** 0.153** 0.153** 1998 year dummy ­0.046** ­0.046** ­0.047** ­0.047** ­0.046** ­0.046** 2000 year dummy 0.167** 0.160** 0.104** 0.104** 0.107** 0.148** Senate race 0.026* 0.023* Governor race 0.054* 0.049* Senate race*pres race ­0.027* ­0.025 Governor race*pres race ­0.033 ­0.024 Senate race*high ed 0.012 0.019 Governor race*high ed 0.012 0.030 Sen race*high ed*pres race ­0.011 ­0.017 Gov race*high ed*pres race ­0.007 ­0.038 Senate race margin ­0.001 0.011 (Continued on next page.) 42 Lisa M. George and Joel Waldfogel TABLE 3.3 (continued) Voting probability Pooled Pooled Pooled Pooled Pooled Pooled (1) (2) (3) (4) (5) (6) Governor race margin ­0.096** ­0.093** Presidential race margin ­0.038 ­0.033 Sen race margin*high ed 0.084* 0.047 Gov race margin*high ed ­0.005 ­0.020 Pres race margin*high ed ­0.032 ­0.042 Sen race margin*pres race ­0.055 ­0.062 Gov race margin*pres race ­0.028 ­0.010 Sen race mgn*high ed*pres race ­0.058 ­0.035 Gov race mgn*high ed*pres race 0.045 ­0.008 Residency > 1 year 0.069** 0.069** Residency > 3 years 0.106** 0.106** Residency > 5 years 0.083** 0.083** High ed*residency > 1 year 0.060** 0.060** High ed*residency > 3 years ­0.025 ­0.024 High ed*residency > 5 years ­0.030** ­0.029** Residency > 1 yr* pres race 0.009 0.009 Residency > 3 yr* pres race ­0.058** ­0.057** Residency > 5 yr* pres race ­0.016 ­0.017 High ed* res > 1 yr* pres race ­0.074** ­0.073** High ed* res > 3 yr* pres race 0.004 0.003 High ed* res > 5 yr* pres race ­0.033* ­0.033* High speed Internet ­0.010 ­0.010 Internet * pres race ­0.128 ­0.179 High speed * pres race 0.022 0.033 Constant ­0.299** ­0.204** ­0.305** ­0.199** ­0.248** ­0.145** Other variables See notes See notes See notes See notes See notes See notes Fixed effects MSA MSAxEd MSA MSAxEd MSA MSAxEd MSA's 178 178 178 178 178 178 Observations 144,935 144,935 144,935 144,935 144,935 144,935 Source: Authors. Notes: Linear probability models with standard errors clustered by MSA for cross-sectional estimates and MSA×Year for pooled estimates. T-statistics suppressed: * significant at 5% level; ** significant at 1% level. Constants in fixed effects regressions represent the average value of the fixed effects. All regressions include year dummies, controls for race, sex, and age as well as categorical variables for income and education. Margins for presidential, senate, and gubernatorial races are measured as the fraction of votes to the winning candidate minus the fraction of votes to the second place candidate. National Media and Local Political Participation 43 We also consider the possibility that causality in our model is reversed: the Times expands first in markets where growing populations of college- educated individuals are becoming less engaged in local affairs relative to the remaining population. Since there is evidence that length of residence in an area affects voter turnout (Squire, Wolfinger, and Glass 1987), we examine the possibility that tenure explains both Times' circulation and disinterest in voting. The CPS asks individuals how long they have lived at their current address, and we can re-estimate our model with education-specific controls for length of residence. If mobility explains both disinterest in voting and the appeal of the Times, then the residency variables should absorb the effects we attribute to penetration. We also allow residency effects to vary by group. We estimate the pooled specifications of table 3.2 with controls for length of residency. Results are presented in columns three and four of table 3.3. In both specifications, the baseline residency variables are posi- tive and significant. Residency effects also vary with education. However, coefficients measuring the effect of Times' penetration on participation are again virtually identical to those in table 3.2. While length of tenure is an important determinant of voting and a plausible determinant of interest in the Times, our results are apparently not attributable to mobility. Because we have no instruments for the Times' expansion, we are most concerned about potential bias from unobserved factors correlated with penetration that are changing over time and that differentially affect col- lege- and non-college­educated individuals. One potential candidate is the spread of the Internet. Although the Internet was not widely available in 1994 or 1996, growth in Internet-based alternatives to local news might be a concern by 1998 and 2000. We can test whether the distraction we measure is explained by the spread of the Internet by including market- level internet penetration for college and non-college­educated individuals in the specifications of table 3.2. Internet penetration is available for 1998 and 2000 from the March Computer Use Supplement to the CPS. Results for longitudinal specifications are reported in the last two columns of table 3.3. As with the election and residency controls, coefficient estimates are stable across specifications. We conclude that the spread of the Internet is not driving our results. As a final confirmation that the distraction we measure is indeed driven by the Times and not other unobservable social trends, we consider whether the Times affects political knowledge as well as turnout. If highly educated individuals are familiar with congressional candidates in 44 Lisa M. George and Joel Waldfogel markets with greater penetration, this would further support the effect over other unobservables. The National Election Studies documents knowledge of political candidates and challengers during even-numbered election years. We consider whether an increase in Times' penetration in a market reduces the likelihood that an individual with a college degree can name at least one candidate in their congressional race. We use a specification similar to those in table 3.2, with a new dependent variable for political knowledge. As expected, individuals with a college degree are more likely to correctly name at least one candidate. We also find that the gap in knowl- edge between college- and non-college­educated individuals shrinks as penetration increases. An increase in sales of 0.01 reduces the gap in the knowledge of congressional candidates between college and non-college­ educated individuals by about 10 percentage points, from 20 percent to 10 percent. Results are presented in table 3.4. Conclusion The results above show that the presence of the New York Times in local markets depresses political participation among highly educated individu- als targeted by the publication. The distracting effect is less in presidential races, which are covered heavily by the Times. In demonstrating that national media consumption can affect local political participation, the results pro- vide empirical support for popular arguments that media affect engagement in local communities. However an important caveat is in order. We examine only the distracting effect of the Times. Distraction by other nonlocal media may compound the effect measured here. Indeed, it would be interesting to examine the impact of Internet programming and cable television on engagement in local affairs. The effect on its target audience may be small compared with the overall distraction of nonlocal media generally. Is the distraction identified in this study a problem? Political and schol- arly debate over the virtue of localism reveals deep concern that media from large markets suppress diversity and foster disengagement from local affairs.4 This study offers empirical evidence that local media play a role in local political participation and that this role is undermined for some citizens by outside media. Yet we know from the literature that, while tar- geted consumers are distracted from local affairs, local products grow more local when exposed to competition from the Times. Hence, limiting the flow of outside media into local markets to promote local engagement may be counterproductive. TABLE 3.4 Does the New York Times Affect Knowledge of Congressional Candidates? Can the respondent name at least one congressional candidate? (1994, 1996, 1998, 2000) (1) (2) Per capita NYT 9.017 7.089 (0.69) (0.54) High ed 0.222 0.222 (10.37)** (10.45)** NYT * high ed ­11.180 ­11.166 (2.19)* (2.17)* High income 0.095 0.096 (2.85)** (2.88)** Hispanic 0.002 0.002 (0.07) (0.05) Male 0.062 0.061 (3.88)** (3.88)** Black ­0.103 ­0.104 (4.61)** (4.59)** Other race ­0.083 ­0.083 (3.05)** (3.04)** Year = 1996 0.142 0.140 (7.17)** (7.21)** Year = 1998 ­0.086 ­0.084 (3.70)** (3.34)** Year = 2000 ­0.145 ­0.143 (5.13)** (4.98)** Senate election 0.161 0.160 (5.81)** (5.76)** Uncontested house election ­0.022 (0.83) No House incumbent running ­0.018 (0.54) Uncontested Senate election ­0.035 (1.13) No Senate incumbent running ­0.011 (0.44) Constant ­0.010 0.002 (0.20) (0.04) Other variables 7 age categories 7 age categories Fixed effects MSA MSA Observations 4,484 4,484 Source: Authors. Notes: Linear probability models with standard errors clustered by MSA for cross-sectional estimates and MSA x Year for pooled estimates. T-statistics in parentheses: * significant at 5% level; ** significant at 1% level. Constants in fixed effects regressions represent the average value of the fixed effects. 45 46 Lisa M. George and Joel Waldfogel Results presented here suggest a number of directions for further research. Entry of outside media into local markets is likely to impact the market structure of local media markets, with potential implications for antitrust policy. Indeed one reason that consolidation in local media mar- kets has not led to higher prices or lower consumption may be growing competition from national products.5 Continued expansion of the Internet is also likely to impact both the supply and demand for local media.6 The distracting effect of non-local media may also affect not just voter turnout but other political behavior that depends on local media and local infor- mation. Information has been linked, for example, to party loyalty (Jones and Hudson 1998). As the cost of informing particular groups of citizens about individual candidates increases, members of these groups may be more likely to vote along party lines. Media markets may also affect politi- cal outcomes. If incumbents have lower-cost alternatives to newspapers for communicating policy positions, the loss of readers to outside media may affect the relative balance between incumbents and challengers.7 Finally, if politicians draft policy platforms to appeal to individuals easily reached through the mass media, media markets may affect substantive policy out- comes as well as participation.8 Our results suggest the promise of these and other avenues of research. Notes 1. The New York Times Company periodically publishes readership surveys con- ducted by Scarborough Research according to procedures established by the Audit Bureau of Circulations. The 2005 Readership Profile Study shows the fraction of Times readers with a college degree to be 59 percent versus 24 per- cent with a college degree in the national sample. Results for previous years are similar. 2. A citation count of references to 21 representatives in 11 states shows that more than 90 percent of references to congressional representatives are made by regional papers rather than the Times. 3. We cannot identify house districts with this sample, since MSA boundaries typically encompass multiple districts. 4. See Stephen Labaton (2001). According to the Times' account of Senate Com- merce Committee hearings on relaxing media cross ownership rules, Sena- tor Ernest Hollings "said the rules had already been relaxed too much and what remained of them was central to preserving local programming and [a] diverse array of voices on the air." For international examples, see the United Nations Educational, Scientific and Cultural Organization (UNESCO) report of the November 2000 roundtable on Cultural Diversity (available at http:// portal.unesco.org/culture/en/ev.php-URL_ID=35258&URL_DO=DO_ TOPIC&URL_SECTION=201.html, accessed February 25, 2008.) Arguing that National Media and Local Political Participation 47 "respect for cultural diversity hinges on respect for the individual," the French culture minister advocates strong trade barriers against U.S. media. 5. For a detailed discussion of the effect of ownership concentration on local newspaper markets, see George (2007), which shows that consolidation in the 1990s led to greater product variety without higher prices or lower readership. 6. See George (2006) for a discussion of the effect of the Internet on daily news- paper readership. 7. An extensive literature documents that campaign spending by challengers has a larger effect on outcomes than spending by incumbents. See in particular Jacobson (1978, 1985) and Ansolabehere and Gerber (1994). A notable excep- tion is Levitt (1995). One interpretation of this finding is that incumbents have more channels through which to inform citizens about both policy plat- forms and expected performance than do challengers. See Fiorina (1977). 8. Strömberg (2004) uses data on U.S. radio penetration during the Depression to study whether regions with higher radio ownership attracted a larger share of relief funds. The effect of radio ownership turns out to be quite large, with a 1 percent increase in radio ownership increasing public relief spending by 0.62 percent. References Ansolabehere, Stephen, and Alan Gerber. 1994. "The Mismeasure of Campaign Spending: Evidence from the 1990 U.S. House Elections." The Journal of Politics 56(4): 1106­18. Barone, Michael, and Grant Ujifusa. 1997. The Almanac of American Politics, 1998. Washington, DC: National Journal Group, Inc. Barone, Michael, Richard E. Cohen, and Grant Ujifusa. 2001. The Almanac of Ameri- can Politics, 2002. Washington, DC: National Journal Group, Inc. Fiorina, Morris P. 1977. Congress, Keystone of the Washington Establishment. New Haven: Yale University Press. Gentzkow, Matthew. 2006. "Television and Voter Turnout." Quarterly Journal of Economics 121(3): 931­72. George, Lisa. 2006. "The Effect of the Internet on the Market for Daily Newspa- pers." Hunter College Working Paper, New York. ------. 2007. "What's Fit to Print: The Effect of Ownership Concentration on Prod- uct Variety in Daily Newspaper Markets." Information Economics and Policy, 19(3­4): 285­303. George, Lisa, and Joel Waldfogel. 2003. "Who Affects Whom in Daily Newspaper Markets?" Journal of Political Economy 111(4): 765­85. ------. 2006. "The New York Times and the Market for Daily Newspapers." American Economic Review 96(1): 435­47. Jacobson, Gary C. 1978. "The Effects of Campaign Spending in Congressional Elec- tions." American Political Science Review 72(2): 469­91. ------. 1985. "Money and Votes Reconsidered: Congressional Elections, 1972­ 1982." Public Choice 47(1): 7­62. 48 Lisa M. George and Joel Waldfogel Jones, Philip R., and John Hudson. 1998. "The Role of Political Parties: An Analysis Based on Transaction Costs." Public Choice 94(1­2): 175­89. Labaton, Stephen. 2001. "Broadcasters Ask Senate to Let F.C.C. Loosen Rules." New York Times, July 18. Levitt, Steven D. 1995. "Policy Watch: Congressional Campaign Finance Reform." Journal of Economic Perspectives 9(1): 183­93. Oberholzer-Gee, Felix, and Joel Waldfogel. 2005. "Strength in Numbers: Group Size and Political Mobilization." Journal of Law Economics 48(1): 73­91. ------. 2006. "Media Markets and Localism: Does Local News en Español Boost Hispanic Voter Turnout?" Working Paper 12317, National Bureau of Economic Research, Cambridge, MA. Prat, Andrea, and David Strömberg. 2005. "Commercial Television and Voter Infor- mation." Discussion Paper 4989, Center for Economic and Policy Research, Washington, DC. Putnam, Robert. 2000. Bowling Alone: The Collapse and Revival of American Commu- nity. New York: Simon and Schuster. Sapiro, Virginia, Steven J. Rosenstone, and the National Election Studies. 2001. 1948­2002 Cumulative Data File [dataset]. Ann Arbor: University of Michi- gan, Center for Political Studies. Squire, Peverill, Raymond E. Wolfinger, and David P. Glass. 1987. "Residential Mobility and Voter Turnout." American Political Science Review 81(1): 45­66. Stromberg, David. 2004. "Radio's Impact on Public Spending." Quarterly Journal of Economics 119(1): 189­221. U.S. Dept. of Commerce, Bureau of the Census. Current Population Survey: Voter Supplement File, 1994, 1996, 1998, 2000 (Computer files). ICPSR version. U.S. Dept. of Commerce, Bureau of the Census, 1995­2001. Ann Arbor: Inter-Uni- versity Consortium for Political and Social Research. Waldfogel, Joel. 2003. "Preference Externalities: An Empirical Study of Who Ben- efits Whom in Differentiated Product Markets." RAND Journal of Economics 34(3): 557­68. ------. 2004. "Who Benefits Whom in Local Television Markets?" Brookings- Wharton Papers on Urban Affairs. 5(1): 257­305. 4 Minority-Targeted Local Media and Voter Turnout: A Summary Joel Waldfogel A growing body of evidence indicates that larger groups of consumers face more options in markets with high fixed costs, incurred regard- less of the number of units produced. When preferences differ across con- sumers, the availability of an appealing product depends not simply on the number of people in the market but rather on the number of persons with similar preferences (Waldfogel 2003, 2004, 2007, 2008; George and Waldfogel 2003). Local media markets provide compelling examples of this phenom- enon. Fixed costs are large relative to market size, and preferences differ sharply across groups of consumers. Even relatively large metropolitan areas typically have only one daily newspaper, a few television stations, and per- haps several dozen radio stations. Because preferences differ substantially across groups of consumers, the absolute number of products offered locally overstates the number of media products targeting particular groups of con- sumers. For example, blacks and whites tend to listen to different radio stations, watch different television programs, and prefer different types of newspapers. Similar differences exist for Hispanics and non-Hispanics. Thewelfareofconsumersoflocalmediaproductsdependsonthegroup's size locally, particularly for small groups. Only places with large concentra- tions of black or Hispanics can support black- or Hispanic-targeted media outlets. If group-targeted information sources make it easier for targeted individuals to be informed about civic matters, then the benefit that people derive from concentrations of like-minded consumers may affect whether people vote, as well as what they consume. Preference minorities--small groups of people with atypical preferences--may be doubly disadvantaged both as consumers and as citizens. This chapter addresses this possible disadvantage directly by summa- rizing research on whether the presence of black- and Hispanic-targeted local media outlets affects the voting turnout of targeted individuals. 49 50 Joel Waldfogel How might the availability of media content that people find appeal- ing affect whether people vote? It is not even clear why people vote in the first place.1 In an electorate that includes millions of voters, the probabil- ity of affecting the outcome of an election is essentially nil. In terms of its effect on outcomes, the benefit of voting is thus zero; and voting has costs. Given the time it takes to become informed about the issues and candidates, as well as the time needed to go to the polls and vote, citizens who value both their time and the election outcome might well choose not to vote. However, many people do vote. Citizens obviously derive some ben- efit from the act of voting. It seems reasonable, if simplistic, to suppose that citizens choose whether to vote by comparing the benefit of participat- ing (along with the infinitesimal benefit of raising their preferred candi- date's win probability) against the cost of participation. In addition to the time spent actually voting, the main cost arises from the need to become informed. The availability of appealing media products with local informa- tion can play a role. Citizens who prefer Spanish- to English-language media content pro- vide an important example of differing preferences in media products. Spanish-language­preferring citizens living in a locale with few fellow Hispanics and no Spanish-language local media outlets find it harder to become informed. Compared to either English-speaking local counterparts or Spanish-speaking persons in locales with larger Hispanic populations, they would face fewer, and less appealing, media options and might be less likely to vote.2 Similarly, black citizens living in a metropolitan area with few blacks will face little local black-targeted media content. Like an iso- lated Spanish-speaker, an isolated black citizen would face less appealing local media content and may be less likely to vote. It is important to note that the media can affect citizens through either content or advertising. Even if the programming itself has no informational content, the availability of programming targeting a particular group allows advertisers to reach the group in a cost-effective manner. Political television advertising spending that targets Hispanics has become significant. In the 2004 presidential race in the United States, for example, candidate John Kerry spent US$1.3 million, and George Bush spent US$3.2 million target- ing Hispanic voters on Spanish-language television (Segal 2006). To determine whether the availability of targeted media products affects voter turnout among blacks and Hispanics in local elections, two types of data are required: data on the availability of targeted media outlets Minority-Targeted Local Media and Voter Turnout: A Summary 51 by locale and over time; and data on whether people vote, by type of per- son, by locale, and over time. This information could help to answer the question in several differ- ent ways. First, do preference minorities turn out to vote more, relative to the typical difference between their turnout and the general popula- tion turnout, in locales where the minority group in question has targeted local media products? Second, does the appearance of group-targeted local media outlets over time raise the tendency for group members to vote, relative to the time pattern for others? The investigation is shaped by the availability of data and by institutional features, such as which media have outlets targeting which groups (for blacks, radio stations and weekly news- papers; for Hispanics, local television news), and which voting data are systematically available.3 Group-Targeted Media Which media have group-targeted local outlets potentially relevant to vot- ing? Until recently very few metropolitan areas had group-targeted local daily newspapers; the Spanish-language Miami Nuevo Herald is a notable exception. None has local television news specifically targeted to black people. On the other hand, local Spanish-language news production and broadcasts occur in a substantial number of metropolitan areas. Indeed, in the past 15 years there has been substantial growth in two major Span- ish-language television networks in the United States: Univision and Tel- emundo. Both operate as national cable channels in metropolitan areas. In addition, both networks have local affiliates that produce local news in a large and growing number of U.S. metropolitan areas. Univision is the largest Spanish-language television network in the United States, operating 17 full-power and seven low-power stations. National programming on the Univision Network is also distributed through 56 broadcast television affiliates and 1,789 cable affiliates nation- wide (Univision 2004). Providing local, national, and international news is a company goal. According to the company, "Univision has the high- est rated local news broadcasts among Hispanics in the U.S." (Univision 2004). The company has devoted resources to covering local news, and "the Univision full-power stations broadcast local news twice every week- night." Because Univision dominates the Hispanic segment while affiliates of the English-language networks tend to split local audiences, Univision stations' local news is often dominant (Univision 2004). 52 Joel Waldfogel Univision reaches 98 percent of Hispanic television households, and it has more viewers than its largest competitor, Telemundo. Telemundo reaches 91 percent of U.S. Hispanic viewers in 118 markets through its 15 owned and operated stations, 32 broadcast affiliates, and nearly 450 cable affiliates. Telemundo is wholly-owned by General Electric and is an operat- ing subsidiary of NBC, the nation's leading broadcast network.4 While the national programming of both Univision and Telemundo is widely available on cable systems, they offer local news in only metro- politan areas with the largest concentrations of Hispanics. By contacting all of the local affiliates currently operating with queries about their broad- cast history, Felix Oberholzer-Gee and I obtained data from 1994­2002 on which metropolitan areas have local news broadcasts in Spanish and when they began airing. Both networks have grown substantially in the past few decades. In 1994, 14 metropolitan areas among the 265 studied had local Spanish-language television news. By 2002, the number had risen to 25. While there are very few black-targeted daily newspapers and no fully black-targeted local television broadcasts, there are dedicated black-targeted outlets in other media, in particular, weekly newspapers and radio stations. In addition to the data on Spanish-language local news, I also have data on black-targeted radio stations by metropolitan area in 1993 and 1997. I have data on black-targeted weekly newspapers at a single point in time, 1998. Finally, I have information from the Current Population Survey (CPS) on general-election voter turnout for each of roughly 60,000 individuals who live in metropolitan areas in each of the even-numbered years from 1994­2002. Analysis of these data provides insights into how the avail- ability of group-targeted media outlets affects the voter turnout of targeted preference minorities. Spanish-Language Local Television News and Voting How much does the availability of local television news in Spanish affect voter turnout? CPS data analyzed in Oberholzer-Gee and Waldfogel (2006) show that overall, voter turnout in even-numbered years varies between about two thirds of eligible U.S. voters in presidential election years and half in other even-numbered years, when congressional candidates and some statewide (senate and gubernatorial) races are at the top of the ballot. Among Hispanics in the CPS, turnout varies between about half in presi- dential years and just over a third in other even-numbered years.5 Minority-Targeted Local Media and Voter Turnout: A Summary 53 The higher turnout in presidential election years stems from the heightened interest in the presidential contest, which, in turn, is supported by the uniformly high level of media coverage of the presidential contest in the national media, particularly national television news. What would we expect to find if local news in Spanish affects turnout? First, if Spanish-language local television news raises Hispanic turnout, then we should see higher turnout rates among Hispanics in metropoli- tan areas with Spanish-language local television news broadcasts than in areas without. Second, if the elevated Hispanic turnout is caused by Spanish-language local news, then the turnout of persons who do not watch Spanish-language news--non-Hispanics--should not be higher in places with local news in Spanish. Third, this targeted local news should elevate Hispanic turnout, relative to non-Hispanic turnout, in elections that are not covered by national news media, which are available everywhere in English and Spanish. Thus, the increase in Hispanic voter turnout brought about by local Spanish television news should be larger in off-year congressional elections and smaller in presidential election years. The data in table 4.1 confirm all three of these predictions. First, in the nonpresidential election years, Hispanic voter turnout is higher in the met- ropolitan areas with Spanish-language local news than in the locales without such broadcasts. Hispanic turnout averages 40 percent in locales with His- panic local television news but only 31 percent in those without in nonpresi- dential years. In presidential years, Hispanic turnout in Spanish local news cities is 54 percent, compared with 46 percent in cities without. Second, the turnout difference between metropolitan areas with and without Spanish-language local television news is absent for non-Hispanic citizens. While the gap between Hispanic turnout in the locales with and without Spanish-language local television news was 8 percentage points to 11 percentage points, the gap for non-Hispanics was only 1 percentage point to 2 percentage points. Thus, the difference is not attributable to some deter- minant of voting common to both Hispanics and non-Hispanics. Finally, the effect is smaller or nonexistent in the presidential election years. Third, the data in table 4.1--along with more sophisticated analyses-- indicate that the presence of Spanish-language local television news raises turnout from an average of 31 percent to about 40 percent in nonpresiden- tial election years, or by about a third. In presidential election years, the presence of Spanish-language local television news raises Hispanic turnout by about one-sixth. These are large effects. 54 Joel Waldfogel TABLE 4.1 Hispanic and Non-Hispanic Voter Turnout and the Presence of Spanish-Language Local Television News a. All years Area has . . . Hispanics Non-Hispanics No Spanish-language local news 36.8 58.2 Observations 5,441 173,328 Spanish-language local news 45.2 59.8 Observations 11,732 79,330 b. Nonpresidential election years (1994, 1998, 2002) Area has . . . Hispanics Non-Hispanics No Spanish-language local news 30.8 52.9 Observations 3,265 107,837 Spanish-language local news 40.0 54.5 Observations 7,235 49,565 c. Presidential election years (1996, 2000) Area has . . . Hispanics Non-Hispanics No Spanish-language local news 45.7 67.0 Observations 2,176 65,491 Spanish-language local news 53.5 68.5 Observations 4,497 29,765 Source: Oberholzer-Gee and Waldfogel 2006. Note: Table shows percent of Hispanic and non-Hispanic population turning out to vote in places with and without Spanish-language local television news. Calculations based on the Current Population Surveys, 1994­2002. A second way to measure the effect of Spanish-language local news on Hispanic voter turnout is to ask how turnout changes over time in places with a Spanish-language local newscast during the sample period. Of the metropolitan areas for which we have data on both Spanish news and vot- ing over time, 14 had Spanish-language local news in 1994 and 25 had it in 2002. Spanish news becomes available in 11 places over this period. What happens to the tendency for Hispanics in these places to vote? Studying this is complicated by two factors. First, different places get Spanish-language local news at different times. Second, the elections of different years attract different levels of interest. Analyzing these data requires a simple statistical model that explains the tendency for Hispanic individuals to vote as the Minority-Targeted Local Media and Voter Turnout: A Summary 55 sum of a metropolitan-area effect--an effect attributable to the election years and a variable reflecting whether the metropolitan area has Spanish- language local news. Beyond the overall Hispanic time pattern of voter turnout, when a met- ropolitan area gets Spanish-language local television news, Hispanic turn- out increases 5 percentage points. As expected, in nonpresidential years, the increase is even larger--8 points--while the increase in presidential years is only 2 points (and not statistically significant). The effects on the non- Hispanic voting tendency can serve as a reality check. Given the ubiquity of English-language local television news, the presence or absence of Spanish- language local television news should have no effect on the non-Hispanic vot- ing tendency. Indeed, there is no discernible change in non-Hispanic turnout with the local appearance of Spanish-language local television news. The two approaches give very consistent estimates for the nonpresi- dential years: an increase of 8 percentage points in Hispanic voter turnout. The estimates reflect a large and statistically significant effect of Spanish- language local news on Hispanic voter turnout. Black Radio and Weekly Newspapers and Voting Do black-targeted radio stations and weekly newspapers affect black voter turnout? My data on these outlets are slightly less complete than the Spanish-language television data. The voting data are the same, but here we have information on black-targeted radio stations in each locale at only two points in time, 1993 and 1997, as well as information on black weekly news- papers in one year, 1998. Nevertheless, similar questions may be asked.6 Table 4.2 shows black and white voter turnout in 1998 in two groups of metropolitan areas, those with and those without black-targeted weekly TABLE 4.2 Percent of Black and White Voter Turnout in Locales with and without Black Weekly Newspapers, 1998 Black White Locales with black weeklies 52.8 53.2 Observations 4,532 25,014 Locales without black weeklies 41.4 52.0 Observations 713 12,050 Source: Oberholzer-Gee and Waldfogel 2005. 56 Joel Waldfogel newspapers, from the CPS. Turnout averages 41 percent among blacks in the metropolitan areas without black weekly newspapers. By contrast, turn- out averages 53 percent among blacks in metropolitan areas with at least one black-targeted weekly. Moreover, metropolitan areas with more than one or two black-targeted weekly newspapers have higher rates of black voter turnout than those with fewer. Does this difference reflect an effect of the presence of black-targeted media? It is always hard to attribute causality with confidence. For one thing, the relationship between elevated black turnout and black media outlets may reflect some other feature of the metropolitan areas related to both higher black turnout and the presence of black weeklies. A major fac- tor supporting the causal interpretation, however, is the absence of a white turnout differential between metropolitan areas with and without black- targeted weeklies (53 versus 52 percent). We have data on the presence of black-targeted radio stations in each metropolitan area in both 1993 and 1997, two years that are nearly aligned with the 1994 and 1998 election years. We can use these data, again in conjunction with the voter turnout data, to ask whether black turnout is higher in places with black radio stations. Indeed, it is higher. As table 4.3 shows, in both 1994 and 1998, blacks in metropolitan areas with black-tar- geted local radio stations have turnout rates of 49 and 53 percent, respec- tively, while blacks in metropolitan areas without black local radio turn out at rates of 41 and 34 percent, respectively. White turnout rates are similar across places with and without black local radio. In fact, white turnout is lower in the places with black local radio. TABLE 4.3 Percent of Black and White Voter Turnout in Locales with and without Black Radio Stations, 1994 and 1998 1994 1998 Black White Black White Locales with black radio stations 49.1 55.0 53.1 51.7 Observations 5,548 28,443 4,722 25,488 Locales without black radio stations 40.6 57.2 34.4 55.2 Observations 635 12,572 523 11,576 Source: Oberholzer-Gee and Waldfogel 2005. Note: The voter turnout data refer to 1994 and 1998, while the radio station data cover 1993 and 1998. Minority-Targeted Local Media and Voter Turnout: A Summary 57 Because I have data on the availability of black-targeted radio stations by locale, at two points in time, we can ask whether the appearance of black-targeted radio, as happens in 11 of our sample metropolitan areas between 1994 and 1998, raises turnout among blacks relative to whites. Estimates based on statistical models like those describing Hispanic voter turnout over time indicate that this is so. Black turnout increases by a large and significant 16 percentage points in metropolitan areas that get a first black-targeted radio station, over and above the average growth in black turnout over this period of 3 percentage points. By contrast, white turnout changes only slightly in places that get a first black-targeted radio station. Hence, the growth in black turnout in locales getting black radio stations is not attributable to some more general phenomenon affecting both blacks and whites in those places. The availability of group-targeted media outlets for local information-- markets that deliver only to groups that are sufficiently large--increases those groups' turnout in local elections by large and statistically significant amounts.7 Markets are inegalitarian for both traditional reasons--wealthier per- sons have more purchasing power and can afford more of the available products--and for the additional reason that only groups of sufficient size face appealing targeted products. When fixed costs are large and preferences differ across consumers, markets deliver products and satisfaction in rough proportion to disparate groups' sizes. Hence, we can add the numerous (those in large groups of persons sharing their preferences) to the wealthy as those favored by the way that markets operate. Because of the way that media markets function, the same groups disadvantaged as consumers also find themselves disadvantaged in becoming informed voters. Notes 1. See Riker and Ordeshook (1968) for an argument that voting is puzzling behavior. 2. It is possible that minorities who choose to live in places with few fellow minorities have preferences more like those of the general population. It should be noted, however, that minorities listen overwhelmingly to minority- targeted radio stations in markets with minority populations large enough to support such outlets; they listen to generally targeted programming in the mar- kets with minority populations not large enough to attract minority targeted entry. This suggests that the preference differences between minorities and the general population exist even in outposts with small minority populations. 58 Joel Waldfogel 3. The evidence in this chapter on Spanish-language local news and Hispanic voter turnout is drawn from Oberholzer-Gee and Waldfogel (2006). 4. Telemundo was formed in 1986 by Saul Steinberg and Henry Silverman of Reliance Capital Group, L.P. See Telemundo (2008). 5. Most of the difference is attributable to the difference between the Hispanic and non-Hispanic age and education distributions. After accounting for edu- cation and age, the gap falls from 15 to about 5 percentage points. 6. The evidence on black-targeted local media and black voter turnout is described in more detail in Oberholzer-Gee and Waldfogel (2005). 7. It is possible that the effects attributed directly to the presence of group- targeted media outlets arise instead because of other institutions (such as clubs) that arise only with sufficient concentrations of minority groups. Oberholzer-Gee and Waldfogel (2005) provide some further discussion of this point. References George, Lisa, and Joel Waldfogel. 2003. "Who Affects Whom in Daily Newspaper Markets?" Journal of Political Economy 111 (4): 765­84. Oberholzer-Gee, Felix, and Joel Waldfogel. 2005. "Strength in Numbers: Group Size and Political Mobilization." Journal of Law & Economics 48(1): 75­91. ------. "Does Local News en Español Raise Hispanic Voter Turnout?" 2006. Work- ing Paper 12317, National Bureau of Economic Research, Cambridge, MA. Riker, William H., and Peter C. Ordeshook. 1968. "A Theory of the Calculus of Vot- ing." American Political Science Review 62(1): 25­42. Segal, Adam. 2006. "Total 2004 Spanish-Language TV Ad Spending by Market and Campaign." Hispanic Voter Project, Johns Hopkins University. http:// advanced.jhu.edu/government/hvp/. Telemundo. 2008. "Corporate Information--English." http://tv.telemundo.yahoo. com/corporativo/articulo/29092006/71/corporativo-corporate-information- english.html. Univision. 2004. "Univision and Telefutura Television Groups: Our Story." http:// www.univision.net. Waldfogel, Joel. 2003. "Preference Externalities: An Empirical Study of Who Ben- efits Whom in Differentiated Product Markets." RAND Journal of Economics 34 (3): 557­68. ------. 2004. "Who Benefits Whom in Local Television Markets?" Brookings-Whar- ton Papers on Urban Affairs 5 (1): 257­84. ------. 2007. The Tyranny of the Market. Cambridge, MA.: Harvard University Press. ------. 2008. "The Median Voter and the Median Consumer." Journal of Urban Eco- nomics 63(2): 567­82. 5 I'm News, Are You? Newspaper Coverage of Elected vs. Appointed Officials Riccardo Puglisi and James M. Snyder, Jr. T he mass media are thought to play a crucial role in the political and electoral process of representative democracies. Media outlets provide readers and viewers with information about issues and events that are relevant to public decision making. Although citizens can rely on direct experience for their information on some issues, for many issues the media constitute the main source of information. Relying on the mass media for information may be problematic. In particular, according to the agenda-setting hypothesis, the mass media can influence citizens' perceptions about which issues are important and to what extent. Whether consciously or not, editors and journalists decide what is newsworthy; they decide which topics to cover and which to ignore. Given the severe constraints on time, space, and interest in the news, the set of topics covered is invariably highly selective. Cohen (1963 page 13) states the argument eloquently: The press "may not be successful much of the time in telling people what to think, but it is stunningly successful in telling its readers what to think about." This gives mass media outlets con- siderable power, and we cannot assume they will always use this power in socially beneficial ways.1 The role of mass media as information providers and agenda setters is particularly relevant when viewed through the lens of a political agency model. Citizens must learn about the performance of incumbent politi- cians in order to properly decide whether to reelect or to oust them. If media outlets provide more information about this performance, then citizens The authors thank Roumeen Islam, Stephen Knack, Daniel Kaufman, and seminar participants at the World Bank Institute's Conference on the Effects of Mass Media on Public Policy for their comments. 59 60 Riccardo Puglisi and James M. Snyder, Jr. will be better able to assess it and compare it with the promises put forward by challengers. If this is the case, incumbent politicians will have stronger ex ante incentives to exert effort, as it is more difficult to keep citizens uninformed about poor performance. Besley and Burgess (2002) develop a simple game-theoretical model of political accountability, in which the effort a reelection-seeking incumbent puts forth is an increasing function of media access by citizens. The authors test the model using data on the 16 major states in India during 1958 to 1992; they find that the responsiveness of state governments to decreases in food production and flood damage, as proxied by the public distribution of food and expenditure on relief, is stronger where newspaper circulation is higher.2 Political agency models are especially useful for evaluating political institutions. Key issues arise: · Which decisions should be made directly by citizens, and which should be made by others? · Which public officials should be elected, and which should be appointed? · How widely should direct democracy be used, and for which decisions and which types of issues? Besley and Coate (2003) analyze the question of institutional choice in the context of regulation of public utilities in the United States.3 They note that when citizens elect a governor who in turn appoints the regulator, the degree to which the regulator adopts a pro-consumer or pro-industry policy is only one of many issues on which citizens must assess the performance of the incumbent governor. On the one hand, when the regulator is directly elected by voters, the regulatory issue is unbundled from other issues rel- evant to the overall evaluation of the governor, because citizens cast two votes. If most citizens are consumers rather than shareholders or employees of the regulated public utility, and if the regulator is directly elected, then it seems likely that the majority will be able to elect a pro-consumer candidate. On the other hand, if the regulator is appointed and utility regulation is not a highly salient issue, then both political parties might choose gubernato- rial candidates who would appoint a pro-industry regulator in order not to alienate the votes and other resources of the industry and its employees. Using panel data on electricity prices in U.S. states, Besley and Coate show that electricity prices in states with elected public utility commissioners are significantly lower than in states with appointed commissioners, and that there is systematically less pass-through of costs into prices in the former. I'm News, Are You? 61 While plausible, it is not clear how the direct election of regulators actually works. Does direct election work the way Besley and Coate and others postulate, even in the cases they study? We know almost nothing about the intermediate steps between the institutional arrangements and final policy outcomes. Do voters actually reward and punish elected regu- lators on the basis of their performance on the specific policy areas under their jurisdiction? Do voters even have the necessary information to vote in such a manner? Could voters plausibly acquire this information at low cost--that is, do the media, political campaigns, and opinion leaders pro- vide sufficient information? In addition to the lack of direct evidence on these questions, circumstantial evidence suggests that voters do not always hold elected regulators accountable. Besley and Coate (2003) survey the lit- erature on elected versus appointed officials and find that, overall, the evi- dence that policy outcomes differ according to the institutional mechanism at play is somewhat mixed.4 If the mixed evidence in fact reflects mixed performance, then it may also reflect variation in the degree to which voters are actually able to assess, reward, and punish their elected regulators. This chapter begins in a very limited way to fill the gap in our knowl- edge about the intermediate steps. We focus on the third issue noted above--the behavior of the mass media. We study state superintendents of education rather than public utility regulators for several reasons.5 · First, there is considerable variation in the institutional arrangements for selecting education superintendents; they are directly elected in 14 states and appointed by the governor, school board, or state assembly in the remaining states. · Second, education is perceived by citizens to be one of the most impor- tant public policy issues, so it is likely that the media devote a notice- able amount of attention to it and that variation in coverage will be detectable even with fairly crude measures.6 · Third, we exploit the fact that we know the exact dates of some regular, education-related events, namely, the release of key standardized test score results.7 Using an online newspaper archive, we gather data about the press cov- erage of education superintendents in a large number of newspapers and ask a simple set of questions. Do newspapers devote more coverage to elected superintendents than appointed ones? Does this happen only around elec- tion time or throughout the year? Does newspaper coverage respond to 62 Riccardo Puglisi and James M. Snyder, Jr. events such as the release of test score results and school "report cards"? How much? Is coverage of elected superintendents more sensitive to events than coverage of appointed ones? Is the coverage more sensitive to "bad" news--for example, a decline in test scores or an increase in dropout rates-- when the superintendent is elected? What about the coverage of the gover- nor, who typically appoints the superintendent and who also has a large responsibility for education policy through the budget-making authority? Are governors covered more on education issues in states with appointed superintendents? In the sections below we provide some answers. Coverage of Elected vs. Appointed Superintendents We study newspaper coverage of state education superintendents over the period from 1998 to 2006.8 Our sample includes 48 states plus the District of Columbia. We use the Newslibrary.com archive (http://nl.newsbank.com) and include all newspapers for which the archive's coverage begins during or before 1999; the sample includes 245 newspapers.9 We exclude newspapers covered only after 1999 in order to keep our panel fairly well-balanced.10 We begin by asking whether, on average, elected superintendents sys- tematically receive more press coverage than appointed ones. We let hist be the total number of stories that mention the superinten- dent of education in state s in newspaper i (which is based in state s) during month t. Since different superintendents hold office in a given state during our period of study, hist is calculated as the sum of stories for each given month that mention all these different superintendents.11 This way we cap- ture stories about the sitting superintendent as well as those about recent past and future superintendents, if different; this is especially important during election campaign periods or when a newly appointed superinten- dent takes over. To control for the size of each newspaper, we search for the word "and" and use the corresponding number of hits as a proxy for the total number of stories appearing in each newspaper during each month,Hist.12 We did analogous searches for governors, using a search string analogous to the one described in note 11. For each state in each month we calculate the overall coverage of super- intendents as an average of the relative coverage devoted by the different newspapers which are based in that state. We do this in two ways. The first gives equal weights to the different newspapers, while the second weights I'm News, Are You? 63 newspapers by their size, that is, by the total number of stories they pub- lish.13 The unweighted relative frequency is as follows: is H hist cst = u ist ns where with a slight abuse of notation "i s" stands for all newspapers based in state s, and ns is the total number of newspapers that are based in state s. The size-weighted relative frequency is: cst = w is hist isHist We calculate weighted and unweighted relative frequencies of men- tions for governors as well. Let gust and gwst be the resulting variables. The state education superintendent was elected for the entire period of our sample in 14 states.14 In all other states and the District of Columbia the education superintendent was appointed. We begin with a simple comparison of the overall difference in the cov- erage of superintendents between states that elect and those that appoint them. We also look at the differential coverage of governors. For each state we average across time the relative frequencies cust, cwst, gust, and gwst in order to obtain the state-specific frequencies cus, cws, gus, and gws. We then perform (unpaired) t-tests. The results are shown in table 5.1. While appointed superintendents are on average mentioned by 0.2 percent of total articles, TABLE 5.1 Average Newspaper Coverage of Elected vs. Appointed Superintendents Appointed Elected Difference p-value Superintendents Unweighted frequencies 0.206 0.318 0.112 0.077 Weighted frequencies 0.199 0.319 0.120 0.050 Governors Unweighted frequencies 1.049 0.887 ­0.162 0.388 Weighted frequencies 0.997 0.871 ­0.126 0.506 Number of observations 35 14 Source: Authors. Note: Percentage frequencies are reported. 64 Riccardo Puglisi and James M. Snyder, Jr. elected ones appear on 0.3 percent of articles. Thus, on average, elected superintendents receive about 50 percent more coverage than appointed superintendents. The difference is mildly significant at the 10 percent level when looking at the unweighted frequencies cus, and at the 5 percent level in the case of size-weighted frequencies. The overall number of articles shows a similar pattern. On average, there are 2.83 stories per month per news- paper in states with appointed superintendents and 3.57 stories in states with elected superintendents. The third and the fourth rows of table 5.1 show results about governors. The overall coverage of governors is much greater than for superintendents in both electing and appointing states; for example, the number of articles mentioning the governor is nearly three times as large even in electing states. Of course, many of these articles may simply mention the governor in passing, with little or no discussion of specific education issues or poli- cies. More interestingly for our purposes, there is no statistically significant difference in the coverage of governors in electing and appointing states. The point estimate is larger in appointing states, but the difference is small in percentage terms. These simple t-tests suggest that an institutional feature--in this case the distinction between states that elect and those that appoint their educa- tion superintendent--is correlated with press coverage of that institution, that is, the extent to which that institution "is news." We now examine the nature of this relationship in more detail. In particular, we investigate whether superintendents are differentially covered across time in electing and appointing states when education-relevant events take place. Also, we investigate whether the processes themselves through which superinten- dents are chosen--appointments and elections--are differentially covered in the two types of state. Exploiting the panel nature of our data allows us to reach firmer conclusions than what is possible in simple cross-sectional comparisons. For states with elected superintendents we collected data on whether the general or primary election was contested, as well as primary election dates, from each state's election Web site (usually that of the secretary of state or state board of elections).15 We also collected the date of appoint- ment for all appointed superintendents, from a variety of sources but mainly from the states' department of education Web sites. The other newsworthy events we consider are the release of test scores. There are three main nationwide standardized tests. The National Associa- tion of Educational Progress tests, also called "the Nation's Report Card," I'm News, Are You? 65 are administered to samples of students in the fourth, eighth, and twelfth grades on a variety of subjects. Reports on these tests are released at vari- ous times. Since 1998, the major results for math, science, reading, writ- ing, U.S. history, and geography were released in the following months: March 1999, September 1999, April 2001, August 2001, November 2001, May 2002, June 2002, June 2003, July 2003, November 2003, December 2003, October 2005, December 2005, May 2006, and June 2006.16 The SAT and ACT tests are taken by college-bound high school seniors.17 The results of the ACT are released in reports every year in mid-August, and the SAT results are reported at the end of August.18 The reports on these tests provide important information to politi- cians, administrators, teachers, policy analysts, and ordinary citizens to help them gauge how well their states and school districts are perform- ing relative to other states and school districts, as well as how well their states and schools are performing relative to previous years. The release of test results is usually accompanied by widespread discussion, including critiques of educational policies, fingerpointing, and occasional bragging. Briefly, our empirical strategy is as follows. We perform a linear regres- sion where the dependent variable is hst, the average number of stories men- tioning the education superintendent during month t across newspapers that are based in state s. The set of regressors includes Elected, a dummy that equals one for electing states; Test, a dummy for the months during which key test scores are released; Elected × Test, an interaction between those variables; Appointment Month, a dummy for the month during which a new regulator was appointed (and the month before); and a set of dummies for the primary and general election campaigns in years during which elected superintendents were actually elected. To control for the average size of newspapers, we also include Hst, the average number of stories containing the word "and" during month t across the newspapers based in state s, i.e., Hst = isHist. The equation is as follows: hst = a + b1Electeds + b2Testst + b3Electeds × Testst + Appointment Monthst (1) + b5Primary Campaignst + b6General Campaignst + b7Hst + est Regarding the two campaign periods, we include three dummy vari- ables for each of them--one for the month of the election, one for the month before (October for the general elections), and one for two months before (September for general elections). For ease of notation we have not explicitly accounted for this in the equation above. In addition, we include year dummies and a linear time trend. 66 Riccardo Puglisi and James M. Snyder, Jr. Note that in equation (1) all cross-state, time-invariant heterogeneity is absorbed by the Elected dummy. We also estimated a specification with state-specific fixed effects. In all specifications we calculate standard errors clustered by state. The results are reported in table 5.2. Column one presents the analysis with the Elected dummy, and column two presents the results with state- TABLE 5.2 Newspaper Coverage of Elected vs. Appointed Superintendents (1) (2) Elected 1.072 -- (1.58) Test 0.366*** 0.365*** (2.87) (2.84) Elected test ­0.028 ­0.019 (­0.13) (­0.09) Appointment month 1.556*** 1.429*** (2.64) (2.67) Primary election month ­2 1.196 0.950 (1.14) (1.03) Primary election month ­1 2.031** 1.784** (2.12) (2.34) Primary election month 4.313** 4.060*** (2.67) (2.89) General election month ­2 2.481*** 2.530*** (3.71) (3.79) General election month ­1 6.952*** 6.981*** (3.85) (3.84) General election month 4.877*** 4.908*** (3.63) (3.55) Total hits (Hst) 0.001** 0.001 (3.03) (1.07) Observations 5,196 5,196 R-squared 0.11 0.42 Source: Authors. Note: State-specific fixed effects included in column 2. Year dummies and linear time trend included in both specifications. Robust t-statistics in parentheses, standard-errors clustered by state. * significant at 10% level; ** significant at 5% level; *** significant at 1% level. I'm News, Are You? 67 specific fixed effects. Both columns tell the same story. First, newspapers based in states that elect their superintendents do not appear to give sig- nificantly more coverage to them, as compared to those based in states with an appointed superintendent (although the point estimates are positive), once we allow for a separate effect of campaign and appointment periods. Second, although the point estimates for the Test dummy are positive and significant, the interaction term Elected × Test dummy is always small and statistically insignificant.19 Third, and perhaps most interesting, all of the variation between appointing and electing states seems to be concentrated during periods that are related to the careers of superintendents themselves. The coeffi- cients on all but one of the General Campaign and Primary Campaign vari- ables are large and statistically significant. During the month prior to the general election in electing states, there are seven additional articles about the superintendent; and in the months of the primary and general elections, there are four to five additional stories. In relative terms these effects are quite large, since the average number of stories each month is only three. In appointing states there are about 1.5 additional stories per month about the superintendent during the period when the new one is appointed. The dependent variable in table 5.2 has the advantage of providing immediate measures of the effects of the explanatory variables on the count of articles mentioning the superintendent. As a robustness check, we also estimated models identical to equation (1) above but substituting hst with cust and cwst. The results are qualitatively the same as those reported in table 5.2. In all cases more than half of the General Campaign and Primary Cam- paign variables are statistically significant, and the Appointment Month vari- able is significant as well but smaller. As discussed, it is important to analyze how governors are covered in the press when dealing with the education issue, and compare this with the coverage of education superintendents. In particular, one would like to check whether--in states where the superintendent is appointed-- governors dealing with the education issue receive more press coverage than in electing states, especially during the election period. Table 5.3 broadly replicates the specifications shown in table 5.2, with the first two columns on the coverage of superintendents and the last two on the cover- age of governors. Columns one and three report the results with the Elected dummy, while columns two and four show results with the state-specific fixed effects. The campaign dummies are lumped together, so that there is only one dummy for the primaries' campaign and one for the general 68 Riccardo Puglisi and James M. Snyder, Jr. TABLE 5.3 Newspaper Coverage of Education Superintendents and Governors Superintendents Governors Elected 1.084 -- ­4.130** -- (1.60) (2.05) Test 0.386*** 0.379*** ­2.668*** ­2.592*** (3.05) (2.97) (6.26) (6.56) Elected × test ­0.253 ­0.241 0.448 0.341 (­0.99) (­0.93) (0.41) (0.32) Appointment month 1.915*** 2.026*** ­4.140** ­0.411 (3.15) (4.19) (­2.43) (­0.50) Primary election in electing state 2.384** 2.139** 2.306 ­0.011 (2.14) (2.33) (1.66) (­0.01) Primary election in appointing state ­0.493** ­0.405** 2.012 1.365 (­2.64) (­2.53) (1.16) (1.14) General election in electing state 4.774*** 4.851*** 9.081*** 9.105*** (4.30) (4.32) (6.93) (6.62) General election in appointing state ­0.107 ­0.108 11.095*** 10.871*** (­0.39) (­.050) (10.44) (11.22) Total hits (Hst) 0.001*** 0.001 0.001 0.007*** (3.0) (1.08) (0.51) (3.16) Observations 5196 5196 5196 5196 P-value on general election variables 0.0001 0.0000 0.23 0.27 R-squared 0.1 0.42 0.22 0.51 Source: Authors. Note: State-specific fixed effects included in columns 2 and 4. Year dummies and linear time trend included in both specifications. Robust t-statistics in parentheses, standard-errors clustered by state. * significant at 10% level; ** significant at 5% level; *** significant at 1% level. The p-value on general election variables is for the test of the null hypothesis that the coefficients on general election in electing state and general election in appointing state are equal. election campaign. Also, there are separate dummies for the campaigns in electing and appointing states. Regarding superintendents, the results shown in table 5.2 are con- firmed, with the additional finding that there is no statistically significant variation in their coverage during the general election campaign in appoint- ing states. In fact, a t-test of equal effects of the general election campaign I'm News, Are You? 69 in appointing and electing states strongly rejects the null, as shown by the small p-values at the bottom of the table. Governors receive less coverage with respect to education in electing states (column three). The effect is large (almost four stories per month less in electing states) and significant at the 5 percent confidence level. This should be compared with an average of 16 articles per month mentioning the governor's name and the education issue. It is also the case that there are significantly less mentions of the governor when test results are released, and this effect is robust to the inclusion of state fixed effects (column four). Column three shows that there are fewer articles on the governor when the superintendent is appointed, but the significance of the effect disappears when controlling for state fixed effects. Finally, there is a large and strongly significant increase in the coverage of governors during the general election campaign in both electing and appointing states, with a higher point estimate for the latter. However, the reported p-value on the t-test of equal effects shows that the two effects are not significantly different, contrary to what was seen for superintendents. Is Campaign Period Coverage Different? We checked a few cases to see whether the additional election-time coverage of superintendents in states where they are elected is actually substantive. Some of it clearly is, as table 5.4 shows. The second column in the table gives the number of news stories that present information about the candidates, including biographical information, qualifications, issue positions, poll results, and commentary on debates among the candidates. The third column gives the number of editorial page articles that contain explicit newspaper endorsements of the candidates or follow-up discussions of these endorse- ments. The fourth column gives the number of editorial columns or letters to the editor regarding the candidates. The fifth column gives the number of news stories in which the incumbent is "on the job." As in the nonelection period, some of these report on important education policy choices or propos- als. However, many are stories about the presentation of various teaching and student awards in which the state education superintendent is simply men- tioned as a participant in the ceremony or a commentator on the award. The sixth column gives the number of election-related stories in which the educa- tion superintendent is simply mentioned in passing--for example, a sample ballot that lists the offices up for election and the candidates' names. 70 Riccardo Puglisi and James M. Snyder, Jr. TABLE 5.4 Election-Time Newspaper Coverage of Elected Education Superintendents Candidate Editorial Doing In Total Official or race Endorsement or letter job passing hits Kathy Cox 8 3 0 16 5 32 Terry Bergeson 11 11 5 21 3 51 Elizabeth Burmaster 10 5 1 5 2 23 Susan Castillo 4 5 6 14 0 29 Source: Authors. Note: For Kathy Cox (GA) the period covered is 10/1/2006­11/5/2006 and the newspapers are the Atlanta Journal-Constitution and Augusta Chronicle. For Terry Bergeson (WA) the period covered is 9/1/2004­11/2/2004 and the newspapers are the Seattle Times, Seattle Post-Intelligencer, Spokane Spokesman Review, and Tacoma News Tribune. For Elizabeth Burmaster (WI) the period covered is 3/1/2005­4/5/2005 and the newspapers are the Milwaukee Journal Sentinel, Capital Times, and Wisconsin State Journal. For Susan Castillo (OR) the period covered is 4/1/2006­5/13/2006 (primary election) and the newspaper is the Portland Oregonian. Including endorsements, there are as many articles about the cam- paigns as there are about the superintendents doing their jobs: 57 to 56, respectively. More importantly for our purposes, table 5.3 demonstrates that the extra articles appearing during campaign periods do not con- sist mainly of stories about other races or campaigns in which the race for education superintendent is merely mentioned in passing. While we would hesitate to give the campaign coverage a passing grade (let alone an "A"), careful readers could definitely extract useful information about the candidates. Readers could learn, for example, that Kathy Cox had tried but failed to remove the word "evolution" from Georgia's science curricu- lum;20 that the previous Georgia superintendent, Linda Schrenko, "didn't even come to work for the last six months of her term and later received an eight-year prison term for stealing education funds" (Downey 2006); and that Terry Bergeson strongly supported the Washington Assessment of Student Learning tests while her opponent, Judith Billings, wanted to de-emphasize the tests (Roberts 2004). It is not evident that the campaign coverage is any less thorough than the day-to-day coverage in noncamp- aign periods. Of course, the data shown in table 5.4 are merely illustrative. In future research we will conduct a more systematic analysis. This requires a careful search for keywords and phrases that serve as "sufficient statistics" for sub- stantive content, which is a difficult challenge we have not yet met. I'm News, Are You? 71 Bad News vs. Good News Next, we examine briefly whether the press is "tougher" on elected super- intendents than appointed ones. As noted, coverage of education super- intendents increases during months when test scores are released than at other times. Here, we examine whether there is an especially large amount of coverage when the superintendent is elected and the test score results are "bad." To measure "good" versus "bad" scores, we calculate the percentage change in a state's score from the previous year. If test scores increased the news is "good" and if they decreased the news is "bad." We interact the change in test scores with the Elected Superintendent dummy variable. We focus on the SAT and ACT and use the test that is more commonly taken in the state.21 These test results are released in mid- or late August, and some- times discussed into September, so we include only the observations for the months of August and September. The equation is as follows: hst = a + b1Electeds + b2DTest Scorest + b3Electeds × DTest Scorest + (2) + b4General Campaignst + b5Primary Campaignst + b6Hst + est As above, we also include year dummy variables and a linear time trend. Table 5.5 presents the results. We find that in states with an elected superintendent, the superintendent receives more coverage when the news is relatively bad, compared to states with an appointed superintendent. While statistically significant, the substantive significance of the effect depends on how it is viewed. For example, we consider the first column and suppose, for example that the average SAT or ACT score in a state falls by .5 percent (the standard deviation of DTest Score is .52). If the state's superintendent is elected, then the number of stories increases by about 0.7; if the superinten- dent is appointed, the number of stories increases by only about 0.15, for a net difference of 0.55 stories. In raw terms this effect is small. In relative terms, however, it is substantial. Since the average number of stories is only about three, the 0.55 stories represent an increase of nearly 20 percent.22 There is also some evidence that the differences in the treatment of elected and appointed representatives are driven by elections themselves. Columns three and four of table 5.5 add an interaction term for election years versus nonelection years, Elected × DTest Score × General Campaign. The coefficient on this term compares the number of hits in September in election years and nonelection years for the states with an elected superintendent. The coefficient on this variable is large, although imprecisely estimated espe- cially in the fixed-effects specification. Given the short period studied and the crudeness of the measures, these results must be treated with caution. We 72 Riccardo Puglisi and James M. Snyder, Jr. TABLE 5.5 Newspaper Coverage of Elected vs. Appointed Superintendents (1) (2) (3) (4) Elected 0.885 -- 0.858 -- (1.31) (1.25) Test score ­0.347** 0.116 ­0.342** 0.122 (­2.21) (0.70) (­2.17) (0.74) Elected × test score ­1.041** ­1.202*** ­0.769 ­0.899** (­2.48) (­3.48) (1.58) (­2.38) Elected × test score -- -- ­2.193** ­2.303* × general campaign (­2.02) (1.89) General campaign 1.455** 1.156** 1.453* 1.166* (2.08) (2.03) (2.00) (1.91) Primary campaign ­0.123 ­0.698 ­0.319 ­0.871 (­0.15) (­0.93) (­0.40) (­1.09) Appointment month 3.368 2.501*** 3.356 2.497*** (1.38) (3.15) (1.37) (3.16) Total hits (Hst) 0.001 0.001 0.001 0.001 (1.53) (1.24) (1.53) (1.26) Constant 80.36 -- 83.08 -- (1.53) (1.59) Observations 858 858 858 858 R-squared 0.08 0.52 0.08 0.52 Source: Authors' analysis; see text for description of sources. Note: State-specific fixed effects included in columns 2 and 4. Year dummies and linear time trend included in all specifications. Robust t-statistics in parentheses, standard-errors clustered by state. * significant at 10% level; ** significant at 5% level; *** significant at 1% level. plan to do more work on this in the future. For example, we can do additional searches to estimate how many of the additional stories in August and Sep- tember actually discuss the test results rather than other education issues. Discussion Observers, especially incumbent politicians, frequently complain that the media report bad news more than good news. One possible explanation is I'm News, Are You? 73 that "bad news sells" and good news does not. The results above suggest an alternative possibility: the media appear to pay more attention to bad news because bad news leads politicians and other newsworthy figures to create more news. The idea is as follows. At any moment in time there are political figures who hold positions of power, and at certain times there are also credible candidates seeking such positions. Whenever these "newsworthy" people do or say things--pass legislation, issue reports, propose new policies, or criticize or compliment existing policies--editors, journalists, and even readers view these actions and statements as newsworthy events. If some good news comes along on some issue, the incumbent politician in charge of the issue says, "Look what a great job I am doing." This produces one story, then silence from most other newsmakers. Perhaps a sympathetic press can keep the story alive another day or two with commentary and a few letters to the editor; but soon it will be stale news. If, however, some bad news comes along, especially around election time, incumbent politicians' opponents seize on this and do their best to keep the issue alive. They criticize the incumbents and their policies and put forth their own solutions to the problems. Interest groups dissatisfied with the incumbents jump in as well, piling on the criticism and advancing their own proposals. The incumbent's defenders must respond in kind to limit the damage. All of this is news and must be covered by the press. Our results on education superintendents are broadly consistent with this "table-tennis" theory of bad news. We find that a decline in test scores is newsworthy especially--perhaps only--in states that elect their education superintendents. Moreover, within electing states, this decline is especially newsworthy if it happens during the general elections campaign. The incen- tives for political actors and interest groups to step in and comment on poor test performance are likely to be stronger if the education superintendents are elected rather than appointed, because this commentary may have a direct effect on the election outcome. These incentives are stronger still when the bad news occurs during a campaign period, since at that time voters may be more focused on acquiring electorally relevant information and acting on it when they vote. Clearly, the next step is to look more closely at the actual content of the stories, especially those around election time, to see if articles published after a bad test result indeed feature a chorus of political actors criticizing or defending the incumbent superintendent. 74 Riccardo Puglisi and James M. Snyder, Jr. Notes 1. See, for example, DellaVigna and Kaplan (2007) for an estimate of the effects of the introduction of Fox News on the Republican vote in presidential elec- tions, and Bergan, Gerber, and Karlan (2006) for an estimate of the differen- tial impact of the Washington Post and the Washington Times on voting behavior in Virginia. 2. Snyder and Strömberg (2004) estimate the effects of the media environment on voter information, voter behavior in U.S. congressional elections, and the behavior of congressmen. 3. See also Maskin and Tirole (2004) and Alesina and Tabellini (2007) for related models. 4. A number of early studies failed to find significant differences between elect- ing and appointing states in the pricing of regulated services. See Berry (1979), Boyes and McDowell (1989), Costello (1984), Crain and McCormick (1984), Harris and Navarro (1983), Navarro (1982), and Primeaux and Mann (1986). However, several recent studies find noticeable differences; in particular, they find more pro-consumer policies under elected regulators than under appointed ones. See Formby, Mishra, and Thistle (1995) on electric utility bond ratings; Fields, Klein, and Sfiridis (1997) on insurance companies; and Smart (1994) on telephone rates. One methodological issue is that very few states switch between the two institutional regimes, so that unobserved cross- state heterogeneity might deliver spurious results. This is less of a concern for the analysis of cost pass-through as a function of the regulatory regime, to the extent that the cost shocks symmetrically hit states. 5. We are in the process of gathering data on other offices as well. Twelve states elect their commissioner of agriculture, and even among agricultural states there is considerable variation. For example, the agricultural commissioner is elected in Iowa and North Dakota but not in Kansas, Nebraska, or South Dakota; the commissioner is elected in most southern and border states but not in Arkansas, Tennessee, or Virginia. Ten states elect their insurance com- missioner, and eleven states elect public utility commissioners. 6. For example, a survey conducted by The Pew Research Center for the People and the Press in January 2006 lists a variety of items and asks respondents to assess their importance. The exact wording was, "I'd like to ask you some ques- tions about priorities for President Bush and Congress this year. As I read from a list, tell me if you think the item that I read should be a top priority, impor- tant but lower priority, not too important or should it not be done." The item "improving the educational system" ranked second in importance, behind only "defending the country from future terrorist attacks." It ranked slightly ahead of such items as "strengthening the nation's economy," "improving the job situation," "taking steps to make the Social Security system financially sound," and noticeably ahead of "reducing crime," "providing health insur- ance to the uninsured," "dealing with the nation's energy problem," and "pro- tecting the environment." I'm News, Are You? 75 7. It is likely that, regardless of the role played by the media, social groups differ in the importance they attach to test score results, with parents of children of the relevant age attaching the highest salience to the issue (together with their teachers). These "natural" levels of salience for different social groups are the baseline against which one must evaluate the agenda-setting effects of media coverage. 8. The name of this office varies across states. The Council of Chief State School Officers has a complete list, as well as other information and links to states' department of education Web sites. 9. Arkansas and Florida are excluded. We exclude Arkansas because there is only one, tiny, newspaper in the Newslibrary.com archive for this state. We exclude Florida because it switched from electing to appointing its education superintendent in 2000. We would like to exploit switches from elections to appointments but cannot do so with any degree of confidence since we have only one case. We hope to bring Arkansas back into the analysis by searching other news databases. 10. We make one exception: Rhode Island. To keep this state in the sample we include two newspapers that are covered beginning in 2000 and 2001, respectively. 11. The search string we use is (["education" OR "school"] AND nnn), where nnn is the name of the superintendent (sometimes including variants of his or her name). We search for all persons holding the office of education superinten- dent during the period from 1995­2006. 12. We are interested in obtaining a good proxy for newspaper size and we use the word "and" because it appears in almost all articles. On the Newslibrary. com archive it is in fact possible to search for the blank ("") string and get all available articles in a given newspaper during a given time interval. However, some other archives like LexisNexis and Factiva do not allow a search for the blank string; for the purposes of replicability on other news archives, we have decided to use "and" as our proxy. Using a blank string instead of "and" pro- duces essentially identical results. 13. We weight newspapers by size for two main reasons. First, the data on coverage are likely to be noisier for small newspapers than for large ones, so weighting is a straightforward way to take into account heteroskedasticity. Second, news- paper size is strongly correlated with circulation, so weighting by size is likely to produce estimates similar to those one would obtain weighting by circula- tion. Weighting by circulation may provide better estimates of the impact of coverage in terms of the number of people exposed. We weight by size because we do not have circulation data for all newspapers in our sample. If we restrict attention to the subsample of newspapers for which we have circulation data, we obtain estimates that are similar to the size-weighted estimates. These additional results are available upon request. 14. These are Arizona, California, Georgia, Idaho, Indiana, Montana, North Carolina, North Dakota, Oklahoma, Oregon, South Carolina, Washington, Wisconsin, and Wyoming. 76 Riccardo Puglisi and James M. Snyder, Jr. 15. Like other statewide elected officials in the United States, candidates for edu- cation superintendent are nominated in primary elections. These are held at various times, depending on the state. In some states they are held in August or September, shortly before the general election. In other states they are held in the spring. With the exception of Wisconsin, which elects its superintendent in April in the odd-numbered year immediately following presidential elec- tion years, all states hold the general election for superintendent in November. In most states the office is partisan, but in a few--California, North Dakota, Oregon, Washington, and Wisconsin--it is nonpartisan. 16. Reports released during the last two days of a month are counted for the sub- sequent month. 17. The SAT is more widespread in the northeast and West Coast, and the ACT is more widespread in the midwest and southeast. 18. In addition, since the No Child Left Behind Act of 2001 (Public Law 107-110), all states must now administer comprehensive standardized tests every year. Many states have administered tests since some time in the 1990s, and a few for even longer. Reports on these tests are released at various times, and we have so far tracked down the release dates for about 20 states. We will incor- porate these in future analyses, after collecting all of the data. 19. We do not investigate why the coverage of superintendents is higher when test scores are released. Demand factors probably play an important role, since readers want to know about test results and what the superintendent has to say about them. On the supply side, the cost for journalists to report about education-related events is lower when test scores are released, because the scores themselves are publicly available. 20. Kathy Cox was quoted as follows: "Evolution is a buzzword that causes a lot of negative reactions." She then ordered to have the word "evolution" struck from the state's proposed new science curriculum, and replaced it with the phrase "biological changes over time." After an uproar, "evolution" was restored (MacDonald 2004). 21. In one state, Nevada, this changed over the course of our sample. Up until 2002 more students in Nevada took the ACT, but after 2002 more took the SAT. We use the ACT results for Nevada. The results are essentially the same if we use the SAT instead. 22. We also ran specifications identical to those in table 5.5 for hits on governors. We find no significant difference in the coverage of good vs. bad news, that is, the coefficients on Test Score and Elected × Test Score are always statistically indistinguishable from zero. References Alesina, A., and G. Tabellini. 2007. "Bureaucrats or Politicians? Part I: A Single Policy Task." American Economic Review 97(1): 169­79. Bergan, D., A. Gerber, and D. Karlan. 2006. "Does The Media Matter? A Field Exper- iment Measuring the Effect of Newspapers on Voting Behavior and Political Opinions." Unpublished manuscript, Yale University, New Haven, CT. I'm News, Are You? 77 Berry, W. 1979. "Utility Regulation in the States: The Policy Effects of Professional- ism and Salience to the Consumer." American Journal of Political Science 23(2): 263­77. Besley, T., and R. Burgess. 2002. "The Political Economy of Government Respon- siveness: Theory and Evidence from India." Quarterly Journal of Economics 117(4): 1415­52. Besley, T., and S. Coate. 2003. "Elected versus Appointed Regulators: Theory and Evidence." Journal of the European Economic Association 1(5): 1176­1205. Boyes, W. J., and J. M. McDowell. 1989. "The Selection of Public Utility Commis- sioners: A Re-examination of the Importance of Institutional Setting." Public Choice 61: 1­13. Cohen, B. 1963. The Press and Foreign Policy. Princeton, Princeton University Press. Costello, K.W. 1984. "Electing Regulators: The Case of Public Utility Commission- ers." Yale Journal on Regulation 2: 83­105. Crain, M. W., and R. E. McCormick. 1984. "Regulators as an Interest Group." In The Theory of Public Choice II, ed. J. M. Buchanan and R. D. Tollison. Ann Arbor: University of Michigan Press. DellaVigna, S., and E. Kaplan. 2007. "The Fox News Effect: Media Bias and Voting." Quarterly Journal of Economics, 122(3): 1187­1234. Downey, M. 2006. "Making the Grade with Experience." Atlanta Journal-Constitution, July 11. Fields, J. A., L. S. Klein, and J. M. Sfiridis. 1997. "A Market Based Evaluation of the Election versus Appointment of Regulatory Commissioners." Public Choice 92(3­4): 337­51. Formby, J. P., B. Mishra, and P. D. Thistle. 1995. "Public Utility Regulation and Bond Ratings." Public Choice 84(1­2): 119­36. Harris, M. C., and P. Navarro. 1983. "Does Electing Public Utility Commissioners Bring Lower Electric Rates?" Public Utilities Fortnightly 112(5): 23­88. MacDonald, M. 2004. "Cox: `Evolution' a Negative Buzzword: State Schools Super- intendent Defends Purge of Word from Proposed Biology Curriculum." Atlanta Journal-Constitution, January 30. Maskin, E., and J. Tirole. 2004. "The Politician and the Judge: Accountability in Government." American Economic Review 94: 1034­54. Navarro, P. 1982. "Public Utility Commission Regulation: Performance, Determi- nants, and Energy Policy Impacts." Energy Journal 3(2): 119­39. Pew Research Center for the People and the Press. 2006. "January 2006 News Inter- est Index." Primeaux, W. and P. Mann. 1986. "Regulator Selection Methods and Electricity Prices." Land Economics 63(1): 1­13. Roberts, Gregory. 2004. "In Race to Oversee State's Schools: A Contrast in Styles." Seattle Post-Intelligencer, October 19. Smart, S. R. 1994. "The Consequences of Appointment Methods and Party Control for Telecommunications Pricing." Journal of Economics and Management Strategy 3(2): 301­23. Snyder, J. M., Jr., and D. Strömberg. 2008. "Press Coverage and Political Account- ability." NBER working paper W13878. Cambridge, MA. 6 The Political Impact of Media Bias Stefano DellaVigna and Ethan Kaplan Introduction I n a representative system of government, policy outcomes are affected by the political preferences and the beliefs of the voters. The media play a key role in shaping these preferences and beliefs. They collect, summarize, and frame the information that voters use in their voting decisions. As a result, many have expressed concern that the media may be able to systematically manipulate political beliefs. Media slant may bias voters and thereby bias policy decisions. Concerns of this type are relevant in the United States, given that over 70 percent of Americans believe that there is a great deal or a fair amount of media bias in news coverage (Pew 2004). Media bias is at least as com- mon, if not more common, in countries with less media freedom than the United States. Is media bias necessarily a problem? The effect of media bias depends on how the audience processes the information presented by the media. If the audience is aware of the media bias and filters it from the information, distor- tions in reporting are unlikely to have large effects on voter beliefs (Bray and Kreps 1987). In this rational world, media bias does not persuade voters. Alternate theories hold that media bias does persuade voters. This may occur because voters do not sufficiently account for bias in the media (De Marzo, Vayanos, and Zwiebel 2003). This, in turn, may be a direct effect of the framing of the news (Lakoff 1987). Ultimately, understanding the impact of media bias on voter beliefs and preferences is an empirical task. In this chapter, we first review some of the papers that have provided a measure of this impact. Most of these papers Bran Knight, Roumeen Islam, and seminar participants at the World Bank Insti- tute's Conference on the Effects of Mass Media on Public Policy provided useful comments. Anitha Sivasankaran provided excellent research assistance. 79 80 Stefano DellaVigna and Ethan Kaplan indicate that the media have a large impact. However, some of the findings can also be explained by self-selection of voters into preferred media. For example, right-wing voters are more likely to expose themselves to right-wing media, giving an impression that the right-wing media persuades them.* Other stud- ies provide evidence of an impact on self-reported voting, or stated voting in a laboratory experiment, as opposed to voting in actual elections. In the rest of the chapter, we summarize the result of a natural experi- ment that addresses the question of the impact of media bias on political preferences. We draw on DellaVigna and Kaplan (2007), which examines the timing of the entry of the Fox News Channel into local cable markets and considers the impact of that entry on voting. Relative to DellaVigna and Kaplan (2007), we present new results on voter turnout for U.S. senatorial elections, as well as a more general analysis of persuasion rates. Rupert Murdoch introduced the 24-hour Fox News Channel in October 1996. The channel expanded rapidly to reach 20 percent of U.S. cities and an audience of 17 percent of the population by June 2000 (Scarborough Research data). The decentralized nature of the cable industry induced sub- stantial geographical variation in access to the Fox News Channel. Since the channel is significantly to the right of all other mainstream television networks (Groseclose and Milyo 2005),** its introduction into a cable mar- ket is likely to have had a significant effect on the available political infor- mation in that cable market. This is true whether the Fox News Channel represents the political center and the rest of the media the liberal wing, or whether it represents the right and the rest of the media the middle. The entry of the Fox News Channel into the U.S. media market makes it likely that, on the one hand, the impact was plausibly large enough to be detected and, on the other hand, that it is possible to identify that impact separately from other confounding factors affecting elections. In this chapter, we discuss our findings on the impact of the Fox News Channel on voting patterns. The key finding is that we detect a significant impact on voting for the Republican candidates. Media bias, therefore, affected voting, at least in the case of the Fox News Channel's expansion. We discuss a variety of results ranging from the impact on the Republican *In this paper, by "right-wing" or "right-leaning" media, we mean media that is supportive of the Republican party or ideas espoused by the Republican party; by "left-wing" or "left-leaning" media, we mean media that is supportive of the Demo- cratic party or ideas espoused by the Democratic party. **Groseclose and Milyo (2005) provide evidence that Fox News is to the right of ABC, CBS, CNN, and NBC. This differentiation may occur because Fox News is right-wing or because the other networks are left-wing. The Political Impact of Media Bias 81 vote share, the impact on voter turnout, regional variation in the impact, and the impact over a longer time horizon and on races that the channel did not explicitly cover. To apply these results to other media markets, such as those in develop- ing countries, it is useful to obtain quantitative estimates of the persuasive impact of the media that are able to be generalized to other contexts. We use our estimates of the impact of the Fox News Channel to compute persua- sion rates, that is, the share of Democratic voters that switched to voting for Republican candidates because of exposure to Fox News. We also compute mobilization rates, that is, the share of nonvoters that turn out to vote because of exposure to the Fox News Channel. This section expands substantially on the discussion of persuasion rates in DellaVigna and Kaplan (2007). In our baseline calibration, we estimate that 4 to 8 percent of the audi- ence was persuaded to vote Republican because of this exposure. When we allow for a separate effect on nonvoters, we find that the mobilization effect of the Fox News Channel may have accounted for one-sixth to one- hundred percent of the impact. We obtain similar persuasion rates for the channel's effect on U.S. senatorial elections. These estimates imply a size- able impact of the media on political decisions. We conclude by discussing some limitations of our approach and some questions for future research on the impact of media bias on politics. This chapter relates to the empirical literature on media bias (Herman and Chomsky 1998; Hamilton 2004; Groseclose and Milyo 2005; Puglisi 2004), as well as the theoretical literature on it (Mullainathan and Shleifer 2005; Gentzkow and Shapiro 2004). We provide evidence that exposure to media bias persuades voters, an implicit assumption underlying most of these papers. Theoretical Predictions We summarize here the key results of a model (DellaVigna and Kaplan 2006) that allows for two channels through which exposure to media news can affect voting. The first channel captures rational learning and predicts that exposure to the media may have an impact on beliefs and voting only in the short run. The second channel captures nonrational persuasion and implies that exposure to the media may affect beliefs and voting in the long run. We present first the rational updating channel in the presence of a new media source whose bias may not be known. A media source injects bias into its coverage of a political candidate. For example, it reports more positive and less negative news about the Republican candidate. Rational 82 Stefano DellaVigna and Ethan Kaplan viewers, knowing the exact extent of the bias, realize that bad news often is not reported and good news often is exaggerated. If the viewers have a good sense of the degree of the media source's bias, they will take into account the bias and discount the news about the candidate. They will not on aver- age be persuaded by the biased news source. The prediction differs if the bias of the media source is unknown. This is the case for television viewers who watch a new news source for the first time. As in the case of the Fox News Channel, we consider the case of a new media source that is more positive to the Republicans than other media sources. Viewers watch reports about Republican candidates and find the reports to be positive relative to what had been expected. Therefore, they alter their beliefs, thinking that the candidates are possibly high quality choices; also, however, they leave room for the possibility that the new media source might be biased to the right. Over time, as the viewers see a large number of positive reports about Republican candidates in com- parison with other media sources, they start to realize that the new media source's bias is to the right of the average media source. Therefore, they take the updated bias into account when evaluating candidates. In the short run, therefore, they are persuaded by the new media source; in the long run, they learn about the bias and are no longer affected. A second possibility is that nonrational viewers do not properly filter out the bias. For example, viewers may be able to learn the degree of the bias but do not realize the degree to which bias impacts reporting. Systematically then, these behavioral viewers place too little weight on the media source being biased and too much weight on the news reports of the media source. These behavioral viewers eventually learn the degree of bias of the media source but are nonetheless persuaded because they underweight the degree to which the bias of the source impacts news reports. In this behavioral scenario, the media has a permanent persuasive impact that does not decrease over time. The two different theories--rational and behavioral--have similar short-run predictions but different long-run predictions. The first predicts that the Fox News Channel's effect will be temporary, and the second pre- dicts that it will be more lasting. Estimates of the Impact of Media Bias Table 6.1 summarizes a small number of key studies that examine the impact of media bias on political behavior and voting. The studies are grouped into four groups by the methodologies used: surveys, laboratory experiments, field experiments, and natural experiments. TABLE 6.I Survey of Studies on Effect of Media Bias on Political Decisions Variable: persuasion rate f (share of listeners convinced by media) Election Control type or group Treatment Exposure Persuasion Treatment question Variable t Year Place Sample size tT group tc rate eT-ec rate f Paper (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Surveys Kull et al. Respondent watches Did U.S. Share of yes 2003 USA N = 8,634 0.220 0.330 1.000 0.141 (2003) Fox News find WMD answers in Iraq? Gentzkow Respondent watches Did Arabs Share of yes 2002 Arab N = 2,457 0.215 0.280 1.000 0.083 and Shapiro CNN carry out answers 2002 countries N = 2,457 0.215 0.133 1.000 0.105 (2004) Respondent watches 9/11 attack? Al Jazeera Laboratory experiments Ansolabehere Laboratory exposure Governor Vote share 1990 southern N = 1,716 0.530 0.568 1.000 0.082 and Iyengar to 30-second election for party 1992 California (1995) political ad Senate sponsoring 1993 election ad Mayoral election Field experiments Gerber, Free subscription to Governor Democratic 2005 Virginia N = 1,011 0.291 0.363 0.940 0.109 Karlan, and Washington Post election vote share Bergan (2006) Natural experiments DellaVigna Fox News exposure, Presidential Republican 2000 28 U.S. N = 66,372,804 0.556 0.560 0.127 0.034 and Kaplan district field election vote share states 0.556 0.563 0.086 0.084 (2007) experiment Fox News exposure, county field 83 experiment Note: Calculations of media effect by the authors is based on data from the papers cited. Columns (7) and (8) report the share of Republican voters in the control and treatment groups. Column (9) reports the exposure rate, that is, the difference between the treatment and the control groups in the share of people exposed to the treatment. Column (10) computes the estimated persuasion rate f as (tT-tC)/((eT-eC)*(1-tC)), except in the first row (see text). The persuasion rate denotes the share of the audience that was not previously convinced and that is now convinced by the message. The data for this paper refer to the estimates obtained using the (predicted) recall audience measure and the diary audience measure, respectively. The data for the Gerber, Karlan, and Bergan (2006) study are courtesy of the authors. For the Ansolabehere and Iyengar (1995) study, we use the data in tables B1.1 and B2.4 neglecting voters that state the intention not to vote. We obtain the baseline share of voters t_{C} from table B1.1 as the weighted average share of the subjects with the same party affiliation as the sponsoring party: (50/(50+38)) * 46/(46+18)+(38/(50+38)) * 18/(46+18). 84 Stefano DellaVigna and Ethan Kaplan Surveys Following Lazarsfeld, Berelson, and Gaudet (1944), political scientists have widely used surveys to assess the impact of the media. Several of these sur- veys have pointed out that the people who watch a given media source tend to share a common political viewpoint with that source. For example, a survey of 8,634 U.S. respondents in 2003 (Kull, Ramsay, and Lewis 2003) finds that 33 percent of the Fox News Channel viewers believes erroneously that weapons of mass destruction were found in Iraq by October 2003, compared to 22 percent for the overall sample. The find- ing holds even after controlling for the political affiliation of respondents. Taken at face value, these estimates imply that the Fox News Channel per- suaded 14 percent of the respondents who did not previously believe that such weapons were found. Findings of this type suggest that exposure to the media may swing voter opinions in very significant ways. Other studies find similar results. Gentzkow and Shapiro (2004) exam- ine the effect of media exposure in nine predominantly Muslim countries based on a survey of 2,457 respondents. Members of the CNN audience were 30 percent more likely to believe, and members of the Al Jazeera audience were 40 percent less likely to believe, that Arabs carried out the 9/11 attacks, compared to survey respondents who did not watch either source. If, again, we translate these responses into persuasion rates, this study implies that the media persuaded from 8 to 10 percent of the audience. These studies clearly document that media audiences differ in their politi- cal beliefs and opinions. They do not, however, necessarily imply that the media persuades voters. An alternative interpretation of this evidence is that peoplechoosemediasourcesthatmatchtheirownpoliticalviews.Thisalterna- tive interpretation of the findings would lead to different policy implications, since it does not imply that media bias shapes the preferences of voters. Laboratory Experiments Politicalscientistshavetakenasecondapproach--laboratoryexperiments-- to measure the impact of the media on voting. In particular, they have examined the impact of political advertisements on stated voter prefer- ences. The impact of political advertisements on voting is similar to the impact of media bias on voting in that both attempt to provide information to voters. However, they differ in that political advertisements claim to be partisan while the media do not. Methodologically, a key difference from the survey studies is that the laboratory allows researchers to separate self-selection from persuasion. By The Political Impact of Media Bias 85 randomly assigning subjects to groups watching different advertisements, researchers can estimate the causal impact of exposure to different political information. In a classical study in this literature, Ansolabehere and Iyengar (1995) expose experimental subjects to 30-second political advertisements sup- porting a candidate or criticizing the opposite candidate. They elicit beliefs and voting intentions at the end of the experiment. The advertisements are embedded in longer news clips to make the exposure to the advertisement more credible and more externally valid. While the impact of political advertisement differs for positive and negative advertisements and depends on the content, in general Ansolabehere and Iyengar (1995) find substantial effects of persuasion. We summarize the results of three experiments run in southern California involving advertisements for a gubernatorial race in 1990, a senatorial race in 1992, and a mayoral race in 1993. When the data are aggregated for the 1,716 subjects in these three experiments, the results indicate that expo- sure to one advertisement increases the stated vote share for the sponsoring party from 0.530 to 0.568. This is a sizeable persuasion effect, implying that the advertisements convinced 8 percent of the subjects who would not otherwise have done so to support the sponsoring party. These experiments capture the causal effect of exposure to the media on voting intentions in the laboratory. It is less clear, however, that these findings would translate into similar persuasion effects of the media in the field. In the experiments, subjects state their voting intention immediately following the advertising. If the impact of exposure to advertising is tem- porary, advertisements in the field would have a much more muted impact. In addition, these experiments do not measure actual voting. Statements of voting in the laboratory do not readily translate into actual votes. For example, survey respondents generally report much higher voting rates than appear in voting records. Finally, subjects may also respond differently in a laboratory setting, compared to the response in an election campaign. For these reasons, while these experiments suggest very interesting pat- terns of the impact of the media, it is important to also collect evidence in the field from media exposure in actual campaigns. Field Experiments Recently Gerber, Karlan, and Bergan (2006) performed a randomized experiment to look at the impact of media bias on voting patterns. In the fall of 2005, they randomly selected three groups of people from a county 86 Stefano DellaVigna and Ethan Kaplan in Virginia at the time of the 2005 Virginia gubernatorial election. They gave a free subscription to the Washington Post (a purportedly left-leaning newspaper) to members of the first group. They gave a free subscription to the Washington Times (a purportedly right-leaning newspaper) to members of the second group. The third group was a control group whose mem- bers did not receive free subscriptions. A few months later, they surveyed the subscription recipients and the control group members with respect to knowledge of current events, political viewpoints, and voting patterns. They found little statistically significant evidence on the impact of media bias on knowledge or viewpoints, but they did find a significant impact on self-reported voting. The group assigned a subscription to the left-wing newspaper stated that they voted more heavily for the Democratic candidate in the Virginia gubernatorial election of 2005. They also find that also the group assigned a subscription to the right-wing newspaper also voted more for the Demo- cratic candidate, albeit insignificantly so. They interpret the decrease in sup- port for the Republicans by the group receiving the right-wing newspaper as due to an information provision (rather than bias) role of the media. The experiment took place at a time when a number of scandals reduced Republican popularity; exposure to these scandals reduced support for the Republicans. A study of this type has a double advantage: it controls for self- selection by randomly allocating the newspapers, and it measures the impact in a real election. In this sense, it combines the advantages of the surveys and the advantages of the laboratory experiments. However, this study also has two drawbacks. First, the study does not measure actual voting, but only self-reported voting, which displays some known biases. Second, since field experiments such as this are expensive to run, the sample size is necessarily small (1,011 subjects) and geographically concentrated in one county. This implies that the impacts of voting are assessed with substantial uncertainty and may not represent the impact over a different population. The Impact of the Fox News Channel We now report the results of a natural experiment on the exposure to media bias and its effects on voting. We summarize the results from the staggered timing of the entry of the Fox News Channel in local cable systems from DellaVigna and Kaplan (2007). In our view, natural experiments address the issues raised above for the other studies. As with the field experiments and The Political Impact of Media Bias 87 the surveys, we consider the impact of actual political information in the field, avoiding the artificial setting of the laboratory. Unlike the above stud- ies, we measure the impact using actual voting as opposed to self-reports. Finally, while the assignment of the media is not random as in the laboratory or field experiments, we argue that it is quasi-random, allowing us to study the causal impact of media bias. Conditional on a set of controls, the avail- ability of the new media of the Fox News Channel appears to be random. Introduction to the Market Rupert Murdoch introduced the 24-hour Fox News Channel in October 1996 to compete with CNN. Like CNN, it was offered via cable and, to a smaller extent, via satellite. The introduction has three features that make it a par- ticularly appropriate case to study to estimate the impact of media bias. First, the expansion of the Fox News Channel was very fast. Thanks to an aggressive marketing campaign, a number of cable companies added the channel to their programming over the next four years. The geographical expansion was accompanied by a corresponding increase in the audience share. By June 2000, less than four years after the introduction, 17.3 percent of the U.S. population reported watching the Fox News Channel regularly. The speed of the expansion implies that the pre-Fox News Channel period (year 1996) and the post-Fox News Channel period (year 2000) are reason- ably comparable. Second, the expansion was geographically differentiated. Cable mar- kets are natural monopolies with capacity constraints on the number of channels. The availability of the Fox News Channel in 2000 in a town depended on whether the local cable company decided to add it to the pro- gramming, possibly at the expense of another channel. Cable companies in neighboring towns adopted the Fox News Channel in different years, cre- ating idiosyncratic differences in access. This allows us to compare voting patterns in neighboring towns that are similar except for the availability of the channel. The comparison at a very fine geographical level makes it less likely that confounding factors affect the estimates. While we do not have an instrument for the availability of the Fox News Channel, we document below that its introduction appears to be idiosyncratic, conditional on a set of controls. Third, the expansion altered the political news coverage in a cable market. Even given the sudden expansion and popularity of the Fox News Channel, and the variations in its diffusion, it is unclear whether the 88 Stefano DellaVigna and Ethan Kaplan addition of any single media source could have a significant impact on the political beliefs of voters. The Fox News Channel's coverage, however, is unique among the television media. The channel is significantly to the right of CNN and all of the mainstream television networks (ABC, CBS, and NBC). This difference is agreed upon in popular discussions as well as aca- demic ones (Groseclose and Milyo 2005). The introduction of the channel into a cable market, therefore, had a systematic and significant effect on the available political information in that cable market. This is true whether the Fox News Channel represents the political center and the rest of the media the liberal wing, or whether the Fox News Channel represents the right and the rest of the media the middle. The channel did not carry politi- cal advertisements and political ads carried by local cable companies were uncommon in 2000. Hence, the impact of the exposure to the new channel is purely due to exposure to the content of the programming. The three features of the expansion are unusual and are the main reason we focused on the Fox News Channel rather than other politically biased news sources. For example, it would be hard to estimate the impact of the introduction in radio programming of right-wing and left-wing talk shows, since radio waves spread over a very large geographical area. Similarly, it is nearly impossible to study the impact of the coverage of the major networks (ABC, CBS, and NBC), which are now available virtually everywhere in the United States. Selection We take advantage of the three features of the expansion of the Fox News Channel and estimate the impact of its availability in 2000 on voting in the 2000 elections at the town level. The data set includes 9,256 towns with voting records in both 1996 and 2000. Before we estimate the impact of Fox News on voting, we assess whether the towns offering the channel in 2000 are ex-ante comparable to the towns not offering it in 2000. We first compare the two groups of towns without including any con- trols. We find that the towns offering the Fox News Channel in 2000 are substantially more likely to vote Republican in the pre-period in 1996 and more likely to go to the polls in 1996. They are also more likely to live in larger towns. This implies that we cannot compare the two groups of towns directly. This kind of comparison, however, does not take advantage of the rich set of town-level controls that we assembled. In particular, the comparison The Political Impact of Media Bias 89 is not limited to geographical neighbors and cable systems with a simi- lar number of channels. Next, we exploit the detailed town-level controls and introduce controls for the cable system and for county fixed effects or congressional district fixed effects. The introduction of these detailed geo- graphical controls (fixed effects) implies that we compare towns with and without the Fox News Channel within a county and within a U.S. congres- sional district. When we make the comparison conditional on these controls, there is no evidence that towns with a higher Republican vote share in 1996, or with higher turnout, are more likely to offer the Fox News Channel in 2000. Moreover, once we control for geographic heterogeneity and size of the cable system, the availability of the Fox News Channel in 2000 is uncor- related with town-level demographic controls from the 1990 and 2000 U.S. Census, such as population, income, ethnic composition, education, and the unemployment rate. To summarize the findings, while overall the availability of the Fox News Channel is highly selective--the channel enters into larger markets and, given the town size, into more Republican areas--conditional on cable market size, the assignment to towns within an area (county or congressio- nal district) is essentially random. This implies that, as long as we include the controls for geography and cable size, we can estimate the causal effect of the introduction of a new media by comparing towns with and without the Fox News Channel in 2000. Impact on Voting in Presidential Elections Next, we come to the main analysis. We consider the impact of the entry of the Fox News Channel on the change in the Republican vote share between 1996 and 2000 at the town level, conditional on the control variables described. This strategy exploits the timing of the entry. By the November 1996 elections, the Fox News Channel had been launched in only a few markets; even in those markets, the launch was only one month before the elections. By the November 2000 elections, the channel had an audience that was smaller than, but nonetheless comparable to, that of CNN. We compare the change in Republican vote share between 1996 and 2000 for towns with the Fox News Channel in 2000 and towns without the channel in 2000, weighting for number of voters. This uses a standard differences-in-differences methodology in that it compares the change over time (first difference) for the towns with the Fox News Channel versus 90 Stefano DellaVigna and Ethan Kaplan the towns without (second difference). This tests whether or not exposure to the channel, and more in general to politically biased media, leads to persuasion. The results are reported in Table 6.2, column one. Formally, we esti- mate the specification as follows: vkR,Pres (1) ,2000- v k R,Pres ,1996= + Fd k FOX ,2000+ Xk + g + k where vk R,Pres denotes the change in the two-party Republican vote ,2000- v k R,Pres ,1996 share between the year 1996 and the year 2000. The set of controls Xk includes town-level demographic variables from the 1990 and 2000 U.S. Census, as well as controls for features of the cable system in the town (number of channels provided and in the number of potential subscribers). In addition, the specification includes a set of geographical fixed effects hg, at the U.S. congressional district level in panel A and at the county level in panel B. The fixed effects and the control help to ensure the comparability of towns with and without the Fox News Channel. In the specification with district fixed effects, we compare towns in the same congressional district, served by cable companies with similar features, and with similar demo- graphics. In the specification with county fixed effects we make the same comparison for towns within a county. Geographic neighbors are more likely to be comparable, in particular, if they share similar demographics and cable system features. Our main finding is that the Fox News Channel had a significant impact on the 2000 elections. The entry increased the Republican vote share in presidential elections by 0.4 percentage points with district fixed effects (panel A) or 0.7 percentage points with county fixed effects (panel B). The difference between the specifications with congressional district (panel A) and county fixed effects (panel B) reflect different geographic comparisons. In both specifications, the result is statistically significant and robust to a variety of alternative specifications, alternative samples, and placebo speci- fications, documented in DellaVigna and Kaplan (2007). Column two in table 6.2 presents one such robustness check: We obtain very similar results if we control for the vote share in 1996, vk R,Pres, instead of taking the first dif- ,1996 ference as in equation (1). Altogether, these results imply that exposure to the media shifted people's voting in the direction of the media content. How large is this effect of the media? Since the Fox News Channel was available in 2000 in about 35 percent of households, the impact is esti- mated to be 0.15 to 0.20 percentage points, or approximately 200,000 votes nationwide. While this vote shift is small compared to the nationwide shift The Political Impact of Media Bias 91 toward the Republicans of 3.5 percentage points between 1996 and 2000, it is still likely to have been decisive in the close 2000 presidential elections. Moreover, this impact may become larger over time as the channel's audi- ence and diffusion grow. Town Characteristics We examine how the Fox News Channel's effect interacts with town char- acteristics, namely the number of channels, the share of population that is urban, and the political orientation of the congressional district (Della- Vigna and Kaplan 2007). The impact on voting was smaller in towns with more cable channels, which is consistent with a moderating effect of competition (Mullainathan and Shleifer 2005). The lower impact result could reflect exposure to more balanced reporting (although CNN and the network news are available in all towns in the sample) or merely lower audience rates for the Fox News Channel when more channels are available. In either case, this suggests that the impact of media bias on voting would be larger in countries with a small number of media sources, as is the case in most developing countries. We also find that the impact of the Fox News Channel was significantly smaller in rural towns, in the South, and in more Republican districts. All these results may be explained by the fact that in rural towns, in the South, and in Republican districts most people already voted Republican and the share of the population that could be convinced was smaller. Persistence of Effects A prediction of the model of persuasion described earlier is that the expo- sure to the Fox News Channel would have a persistent effect on voting rather than a temporary one. Instead, the model of rational learning pre- dicts that over time the effect should decay, as voters learn about the previ- ously unobserved bias. We therefore study whether the impact of the Fox News Channel per- sists between the 2000 presidential election and the 2004 presidential elec- tion. In column three of table 6.2, we estimate the specification as follows: vk R,Pres (2) ,2004- v k R,Pres ,2000= + Fd k FOX ,2000+ Xk + g + k. We find that the availability of the channel in a town in 2000 is associ- ated with an insignificant 0.2 percentage point vote share increase between 2000 and 2004. The result is essentially identical with district fixed effects 92 TABLE 6.2 Impact of Fox News on Voting Vote share Turnout Main result--presidential Persistence Senate Presidential U.S. Senate Change in Pres. Pres. Rep. U.S. Senate Log change Log change Rep. 2-party Pres. Rep. 2-party vote Rep. 2-party in Pres. in Senate vote share, 2-party vote share, 2004­ vote share, turnout, turnout, 2000­1996 share, 2000 2000 2000 2000­1996 2000­1996 Panel A: U.S. House district (1) (2) (3) (4) (5) (6) fixed effects Availability of Fox News 0.0042 0.0041 0.0021 0.0072 0.0046 0.0054 via cable in 2000 (0.0015)*** (0.0016)*** (0.0020) (0.0026)*** (0.0039) (0.0044) Republican vote share in 1996 0.9362 0.8295 1996 presidential race (0.0079)*** (0.0111)*** Fox News in 2000 * (New York race) 0.0039 (0.0067) Control variables: Census controls: 1990 and 2000 X X X X X X Cable system controls X X X X X X U.S. House district fixed effects X X X X X X Log change in voting-age pop.: 2000­1996 X X R2 0.7533 0.9824 0.6281 0.9768 0.6151 0.6993 N N = 9,256 N = 9,256 N = 8,605 N = 8,192 N = 9,256 N = 8,186 Panel B: county fixed effects (1) (2) (3) (4) (5) (6) Availability of Fox News 0.0069 0.0068 0.0019 0.0071 0.0178 0.0158 via cable in 2000 (0.0014)*** (0.0014)*** (0.0024) (0.0028)** (0.0051)*** (0.0056)*** Republican vote share in 1996 0.9432 0.8432 1996 presidential race (0.0092)*** (0.0146)*** Fox News in 2000 * (New York race) ­0.0017 (0.0060) Control variables: Census controls: 1990 and 2000 X X X X X X Cable system controls X X X X X X County fixed effects X X X X X X Log change in voting-age pop.: 2000­1996 X X R2 0.8119 0.9865 0.6941 0.9829 0.6863 0.7474 N N = 9,256 N = 9,256 N = 8,605 N = 8,192 N = 9,256 N = 8,186 Note: An observation in the OLS regressions in columns (1)­(3) and (5) is a town in one of the 28 U.S. states in the sample. Panel A is estimated with US House district fixed effects and Panel B is estimated with county fixed effects. In columns (4) and (6), an observation in the OLS regression is a town in one of the US states with a Senate election in the year 2000. In column (1), the dependent variable is the change in the two-party Republican vote share for the 2000 presidential election. In column (2), the dependent variable is the two-party Republican vote share in 2000. In column (3), the dependent variable is the two-party Republican vote share for the 2004 presidential election minus the same variable for the 2000 election. In column (4), the dependent variable is the two-party Republican vote share for Senate in 2000. In column (5), the dependent variable is the log of turnout in the presidential election in 2000 minus the log of turnout in 1996. In column (6), the dependent variable is the log of turnout in U.S. Senate elections in 2000 minus the log of turnout in the presidential election in 1996. The specification in column (6) drops six outliers (observations with a change in log turnout larger than one in absolute value). In columns (4) and (6), the change in the log of voting-age population between 1996 and 2000 is an (unreported) control variable. The variable "Availability of Fox News via cable in 2000" is a binary variable that equals one if Fox News was part of the town's local cable package in 2000. The census controls are 12 demographic variables from the census, present both in the 2000 values and in differences between 2000 and 1990. The cable system controls are deciles in the number of channels provided and in the number of potential subscribers. Robust standard errors clustered by local cable company in parentheses. The 93 observations are weighted by total votes cast in the 1996 presidential election. * significant at 10 percent; ** significant at 5 percent; *** significant at 1 percent. 94 Stefano DellaVigna and Ethan Kaplan (panel A) and with county fixed effects (panel B). The effect therefore appears to be persistent, if not increasing over time. Persistence is consis- tent with the predictions of a model of nonrational persuasion; however, this result could also be due to greater audience exposure to the Fox News Channel over the period from 2000 to 2004. Ideology versus Popularity The previous findings suggest that the channel had a significant effect on the Republican vote share and on turnout in the presidential election. We now consider whether the effect extends to local politics not covered by the channel. This allows us to test whether the effect is candidate-specific and does not extend to local elections, or a general ideological shift, and should affect local elections. Senatorial elections in the United States are a good test in this respect, because a large majority of these races fail to get national coverage. These elections are similar to local elections, for which unfortunately no town-level data set is available. As a test of the ideology shift, therefore, we estimate whether exposure to the Fox News Channel affected the two-party vote share in the senatorial elections. In addition, one or two senatorial races per year attract substantial national coverage, almost like presidential races. This allows us to compare the effect on races that were not covered, where only ideological shifts should matter, to the effect on covered races, where candidate-specific coverage also could matter. In 2000, the senatorial race that got the most coverage on the Fox News Channel by a wide margin was the Hillary Clinton-Rick Lazio race in New York. These two candidates had 99 mentions on the O'Reilly Factor and the Hannity & Colmes show in the two months prior to the 2000 elections, with most mentions critical of Hillary Clinton. All other senate candidates in the 2000 campaign combined received a total of 73 mentions, with Joseph Lieberman, who was typically mentioned because of his vice- presidential race, getting the lion's share of these mentions. We examine whether the Fox News Channel had an impact on the vote share in the senate elections, and whether it had a differential effect for the Clinton-Lazio race. In table 6.2, column four, we estimate as follows: vkR,Sen (3) ,2000= v kR,Pres ,1996+ + Fd kFOX ,2000+ Fd k FOX ,2000 dNY + Xk + g + k , where vk R,Sen is the two-party vote share in the senatorial elections in ,2000 2000. The coefficient bF indicates the effect of the Fox News Channel on senatorial races other than New York, and fF indicates the differential The Political Impact of Media Bias 95 effect for the featured New York race. This specification controls for the 1996 presidential vote share.1 We find that the Fox News Channel signifi- cantly increased the Republican vote share for the senate by 0.7 percentage points b^ F = 0.0072 (panel A) and b^ F = 0.0071 (panel B). Interestingly, the effect is as large as that on the presidential elections. Additionally, the effect is not significantly larger for the one senatorial race that the Fox News Channel covered heavily, the New York race between Hillary Clinton and Rick Lazio (f^F = 0.0039 in panel A and f^F = ­0.0017 in panel B ). Thus the channel appears to have induced a generalized ideological shift, as opposed to a candidate-specific popularity effect. Impact on Turnout The significant impact of the Fox News Channel on voting in presidential and senatorial elections could have occurred through two mechanisms. First, the channel's entry could have convinced Democratic voters to vote for Republican candidates. Second, the entry could have turned latent Republican non-voters out to vote. To provide evidence on the two mecha- nisms, we study the impact of the Fox News Channel on voter turnout, as measured by the number of people going to the polls. To the extent that the persuasion effect was purely due to a change in the minds of Democratic voters, we would not expect an increase in turnout. In table 6.2, column 5, we estimate as follows: t Pres - t Pres = + Fdk FOX k,2000 k,1996 ,2000+ ln (Poopk,2000) - ln(Pop ) + k,1996 Xk + g + k , (4) where t Pres is the log total votes in town k in year t: t Pres = ln Vk The k,t k,t ( TOT,Pres ,t ). change in this measure over time is the percent change in total votes cast. This specification controls for the percentage change in the voting-age town population over time, ln(Popk ), since increases in popu- ,2000) ­ ln(Popk ,1996 lation increase the number of votes cast. We obtain somewhat different answers using our two benchmark spec- ifications. The estimates with county fixed effects (panel B) imply that the availability of the Fox News Channel increased turnout to the polls by 1.78 percent, a large and significant effect. This estimate would imply that the effect on voting was mainly though mobilization of Republicans. The effect is still positive, but smaller and statistically insignificant using congressio- nal district fixed effects (panel A). This latter estimate would imply that the impact operated mainly through convincing Democratic voters. 96 Stefano DellaVigna and Ethan Kaplan In table 6.2, column six, we estimate the impact on turnout in U.S. senatorial elections. We estimate an equation parallel to equation (4) with the change in turnout between the senatorial elections in 2000 and the presidential elections in 1996, t Sen - t Pres , as a dependent variable. We k,2000 k,1996 find that the entry increased turnout in senatorial elections by 0.54 percent with district fixed effects (panel A) or by 1.58 percent with county fixed effects (panel B). These estimates parallel the estimates of turnout for presi- dential elections, with a significant impact in the specification with county fixed effects. Overall, the Fox News Channel's entry into a market appears to have mobilized voters. However, the evidence for this is not as consistent as for the effect on vote share. Persuasion Rates of the Media Overall, we find a sizeable impact on the vote share for Republicans and on turnout. These estimates, however, do not tell us how effective the Fox News Channel was in convincing Democrats who were exposed, nor does it tell us how effective the channel was in mobilizing latent Republicans. Measures of the persuasiveness of the media depend, among other things, on the size of the audience in 2000. The smaller the audience, the larger the persuasion effect associated with the half percentage point impact on vote share. To generalize the results to other media markets, includ- ing possibly those in developing countries, it is useful to obtain quanti- tative estimates of effective persuasiveness of the media per individual exposed.2 What share of the public exposed to a media source changes its opinions in the political direction of the media source? While the impact may not easily generalize to very different media markets, in principle the estimates of persuasion rates can be applied to other similar media markets. In this section, we compute the effectiveness of the Fox News Channel in convincing non-Republican viewers to turn out and vote Republican. This substantially extends computations in DellaVigna and Kaplan (2007) where we assumed that the Fox News Channel convinced the same percent- age of Democrats and nonvoters to vote Republican, and where we used only the vote share and not the turnout estimates to compute the persua- sion rate. We generalize the previous approach by (1) allowing for differen- tial influence rates on Democrats and on nonvoters and (2) using turnout estimates in addition to vote share estimates. The Political Impact of Media Bias 97 Setup We compare the vote share vj in treatment towns exposed to the channel (j = T) and control towns not exposed (j = C). Before the exposure, a share r of the voting-eligible population votes Republican, a share d votes Demo- crat, and the remaining share (1 ­ r ­ d) does not vote. Since the two types of towns have similar political outcomes in the pre-Fox News Channel period conditional on a set of controls, we assume that r and d are the same in towns T and C. A fraction e of the town population is exposed to the Fox News Chan- nel after the nationwide introduction. Exposure e is higher in treatment towns, that is, eT > eC 0. We allow for nonzero exposure eC in control towns because, for example, of the availability of satellites that broadcast the channel to subscribers in both towns. The key parameters we use to capture the effectiveness in affecting political behavior are the persuasion rate f and the mobilization rate m. The Fox News Channel persuades a fraction f of the Democrats in the audi- ence, ejd, to vote Republican. In addition, the channel mobilizes a fraction z of the nonvoters in the audience, ej(1 ­ r ­ d) , inducing them to vote. Of these mobilized voters, fm is the percentage of who turn out to vote for Republicans, with 0 fm 1. ( ) vj = r + fejd + mej fm 1 - r - d r + d + mej 1 - r - d ( ) (5) The number of Republicans in town j is equal to the number of Republicans in the town before the entry, r, plus the percent of exposed Democrats who were persuaded, fej d, plus the share of the mobilized voters who turn out for the Republicans, mej fm(1 ­ r ­ d). The denominator in equation (5) is the turnout in town j: tj = r + d + mej(1 ­ r ­ d). (6) The turnout in town j is affected by the entry through the mobilization effect m on nonvoters. Using equation (6) for the turnout tj, we can compute the mobilization rate m. Subtracting tC from tT and re-arranging, we obtain the following: m = tT - tC (eT (7) - eC 1 - r - d )( ) . This equation is easily interpretable. The percent of those mobilized by the Fox News Channel to vote is equal to the difference in turnout across 98 Stefano DellaVigna and Ethan Kaplan treatment and control towns, divided by the differential in the number of treated individuals (the differential exposure rate multiplied by the size of the nonvoting population). We can calculate the persuasion rate f given the mobilization rate m of Fox News, provided that we make assumptions about fm. The other variables vj, tj, r, d, and ej are observed. We report the solution for f in the appendix. Persuasion Results We now provide results for the mobilization rate m and the persuasion rate f for different specifications. In particular, we estimate mobilization and per- suasion rates for both presidential elections and senatorial elections, using the specifications with district fixed effects (table 6.2, panel A) and the specifications with county fixed effects (table 6.2, panel B). This provides a broad array of estimates of the impact of the media. To obtain these estimates, we need measures for the parameters vT , VC, tT, tC, r, d, eT , and eC. We use the specifications in table 6.2 and summary statistics reported in DellaVigna and Kaplan (2007) to estimate the vote shares VT and vC and the turnout rates tT and tC. We estimate the pre-Fox News Channel share of Democrats and Republicans r and d using the aver- age voting patterns in the data. Finally, we document the audience rates eT and eC using measures of the audience of the Fox News Channel accord- ing to Scarborough Research data. According to the benchmark audience measure of the recall audience and using the estimates with district fixed effects, the exposure to the Fox News Channel e is 8.9 percentage points in the control towns and 21.7 percentage points in the treatment towns. The availability of the channel via cable thus increased its audience by about 12.8 percentage points. The estimated increase in audience is about 8.6 per- centage points for the specification with county fixed effects. We document further the estimates of these parameters in the appendix. We estimate the persuasion rates and the mobilization rates under three different scenarios and report the results in table 6.3. The first sce- nario, "Mobilization=Persuasion," assumes that the persuasion rate f and the mobilization rate m are equal, that is, the effect on nonvoters is the same as the effect on Democratic voters. This is the assumption used for the estimates in DellaVigna and Kaplan (2007). For presidential elections, these assumptions imply that the Fox News Channel persuaded 3.4 per- cent of voters in the specification with district fixed effects or 8.4 percent of voters in the specification with county fixed effects to vote Republican. TABLE 6.3 Persuasion and Mobilization Rates Geographic controls Estimated effect Estimated effect Implied Implied (fixed effects) used of Fox News of Fox News persuasion mobilization Assumptions for calibration for the estimates on vote share on turnout rate f of the media rate m of the media (1) (2) (3) (4) (5) Panel A: presidential elections Mobilization = persuasion District fixed effects 0.0042 0.0046 0.034 X Mobilization 100% for Rep. District fixed effects 0.0042 0.0046 0.025 0.046 Mobilization 56% for Rep. District fixed effects 0.0042 0.0046 0.054 0.046 Mobilization = Persuasion County fixed effects 0.0069 0.0178 0.084 X Mobilization 100% for Rep. County fixed effects 0.0069 0.0178 ­0.052 0.263 Mobilization 56% for Rep. County fixed effects 0.0069 0.0178 0.115 0.263 Panel B: U.S. Senate elections Mobilization = persuasion District fixed effects 0.0072 0.0054 0.054 X Mobilization 100% for Rep. District fixed effects 0.0072 0.0054 0.062 0.045 Mobilization 56% for Rep. District fixed effects 0.0072 0.0054 0.096 0.045 Mobilization = persuasion County fixed effects 0.0071 0.0158 0.079 X Mobilization 100% for Rep. County fixed effects 0.0071 0.0158 ­0.025 0.196 Mobilization 56% for Rep. County fixed effects 0.0071 0.0158 0.122 0.196 Note: This table reports the estimated persuasion rate and mobilization rate of the media implied by the Fox News estimates. The persuasion rate is defined as the share of Democratic voters who are convinced to vote Republican due to exposure to Fox News. The mobilization rate is defined as the share of nonvoters who are convinced to vote due to exposure to Fox News. The table presents the result for three types of estimates. The first estimate, "Mobilization=persuasion," assumes that mobilization rates equal persuasion rates (that is, the effect of Fox News on nonvoters is the same as the effect on Democrats). The second estimate, "Mobilization 100% for Rep.," assumes that mobilization rates can differ from persuasion rates, and that all mobilized nonvoters vote for the Republicans. The third estimate, "Mobilization 56% for Rep.," assumes that mobilization rates can differ from persuasion rates, and that 56% of mobilized nonvoters vote for the Republicans and 44% for Democrats (this is based on the composition of the Fox News audience). The formulas and parameters used to calculate the implied mobilization rate and the implied persuasion rate are detailed in the appendix. In Panel A, the estimated effect of Fox News on the vote share is the coefficient on the Fox News variable in the presidential vote share regression (column 2), and the estimated effect on 99 turnout is the coefficient on the Fox News variable in the presidential turnout regression (column 3). We present separate estimates using U.S. House district fixed effects (Panel A) and county fixed effects (Panel B). In Panel B, the estimated effect of Fox News on the vote share is the coefficient on the Fox News variable in the U.S. Senate vote share regression (column 2), and the estimated effect on turnout is the coefficient on the Fox News variable in the U.S. Senate turnout regression (column 3). 100 Stefano DellaVigna and Ethan Kaplan Intuitively, to obtain the estimate of the persuasion rate we rescale the effect on the vote share (0.42 and 0.69 percentage points, respectively) by the 12.8 percentage point differential audience rate.3 For senatorial elections, the estimates imply persuasion rates of 5.4 percent for district fixed effects and 7.9 percent for county fixed effects. These estimates indicate sizeable persuasive effects of the media. A drawback of this first approach is that it predicts an increase in turnout due to the Fox News Channel that is significantly smaller than the observed large increase in the specifications with county fixed effects (table 6.2, panel B, column five). The larger impact on turnout may be due to higher mobilization rates m compared to the persuasion rates f. In addi- tion, the newly mobilized voters may have in part voted for the Democratic party, implying that fm is smaller than 1. In this section, we extend our previous work to separate out a mobilization effect on nonvoters from a persuasive effect on Democrats.4 In the second scenario, which we label "Mobilization 100% for Rep.," we allow for a different persuasion rate f and mobilization rate m. We also assume that all the nonvoters that the Fox News Channel mobilizes vote Republican, that is, fm = 1. The results are quite similar for presidential and senatorial elections, but differ depending on the unit of the fixed effects (district or county). The estimated mobilization rates m are large with county fixed effects (26.3 for presidential and 19.6 percent for senatorial) and sizeable with district fixed effects (4.6 percent for presidential and 4.5 percent for senatorial). The estimates of the mobilization rates in turn affect the estimates of the persuasion rates. In the specification with county fixed effects, the large mobilization rates render the persuasion rates small, or even slightly negative. If the Fox News Channel had a large effect of con- vincing nonvoters to vote Republican, this fully explains the vote share results, even without any effect on converting Democratic voters. The esti- mates of persuasion rates with district fixed effects are less affected by this scenario since the estimated mobilization rates are lower. This second scenario, while allowing for a separate turnout and con- version effect, requires the turnout effect to benefit only Republican can- didates. However, it is possible that a fraction of the nonvoters that is mobilized votes for Democratic candidates. To quantify this, in the third scenario, "Mobilization 56% for Rep.," we assume that 56 percent of mobi- lized nonvoters vote for the Republicans and 44 percent for Democrats, that is, we assume fm = 0.56. This breakdown, while arbitrary, is based on The Political Impact of Media Bias 101 the observation that, according to Scarborough data, 56 percent of the Fox News Channel's audience is self-declared Republican. We assume that this breakdown also holds for the newly mobilized voters. Under this scenario, we obtain persuasion rates that are typically higher than under the other scenarios, ranging from 5.4 percent in presidential elections with dis- trict fixed effects to 12.2 percent in senatorial elections with county fixed effects. The reason is that under these assumptions, the increase in Republi- can vote share due to the Fox News Channel cannot be due to the effect on nonvoters, since nonvoters divide themselves fairly evenly across parties. The effect, therefore, has to be due to a large conversion effect of Democrats into Republicans. Exposure to more conservative coverage had a sizable effect on politi- cal choices of voters. Most scenarios imply a substantial role of the media in persuading Democratic voters to vote Republican. However, if we take at face value the estimates indicating large turnout effects (and hence high mobiliza- tion rates), the data are also consistent with pure mobilization and no persua- sion. While our best guess based on the different estimates is that exposure affected both margins, we leave the task of fully differentiating between per- suasive impacts of the media and mobilizing impacts of the media to future research. Conclusions The study on the impact of the Fox News Channel discussed in this chapter provides evidence on the extent to which the political content of a media source persuades and mobilizes potential voters. We have compared this study to other studies in the literature that take different approaches to answering a similar question. We have argued that natural experiments in media exposure provide a combination of two desir- able features, quasi-random assignment of the media and a natural setting. In comparison, surveys also examine voting in the field, but they cannot separate sorting from causal effect. Laboratory experiments provide a clean randomization, but they do so at the cost of an artificial setting. Field exper- iments can also provide randomization in the context of a real election, but it is often difficult to map the outcomes to real election variables. Other studies use natural experiments to address the impact of the media on voting. Expansions of the New York Times in the 1990s (George and Waldfogel 2006) and of television between 1940 and 1972 (Gentzkow 102 Stefano DellaVigna and Ethan Kaplan 2006) decrease turnout, while radio entry between 1920 and 1940 increases turnout (Strömberg 2004). These studies analyze the link between media and voting from other vantage points. A number of important questions are left unanswered, or only partially answered, by this and other studies on the impact of the media. We outline a few that we consider to be particularly important. · First, do the media mostly mobilize the "already convinced" or do they persuade voters to switch parties? We find evidence that the effect of the Fox News Channel was at least partly due to increased turnout of latent Republicans, the "already convinced," but we cannot evaluate the extent of this channel with precision. · Second, does media bias affect other behavior beyond voting? It would be interesting to consider the impact on other politically charged deci- sions, such as the degree of political activism, propensity to contribute money to political causes, or military enrollment rates. · Third, who is most likely to be persuaded by the media? A large litera- ture in political science tries to determine when political preferences are formed, including whether young people are most affected by polit- ical messages. In this chapter, we did not have access to individual data and hence could not test these hypotheses. · Fourth, does exposure to the media change policy? We have not directly examined the impact on policy making. · Fifth, why do the media have an effect on voting? We have provided some evidence to distinguish rational updating from nonrational per- suasion, but we cannot draw firm conclusions. Understanding the exact channels of media influence is important both from policy and research perspectives. Appendix Using equations (8) and (9), we can derive the difference in the vote shares as follows: vT - vC = r + feTd + meT fm 1 - r - d ( ) - r + feCd + meC fm 1 - r - d ( ) (8) tT tC Multiplying by tTtC and subtracting off r(tC ­ tT), we get the following: (9) The Political Impact of Media Bias 103 Subtracting the terms involving fm and dividing by d(eTtC ­ eCdtT), we get the following: f = d eTtC - eCtT (v( T - vC tTtC ) )- r tC - tT ( ) d eTtC - eCtT ( ) - mfm(1 - r - d) (10) d Finally, using the definition of tj, we note that eTtC ­ eCtT = (eT ­ eC)(r + d). Substituting this expression, we can simplify, combine terms, and solve for the influence rate f as follows: (vT r - vC ) fm - f = ) r tTtC m 1 - r - d ( + d d eT - eC ( ) - r + d (11) d Equation (11) has two components, and is roughly interpretable as the effect of the Fox News Channel on vote share vT ­ vC per exposed Democrat, minus the increase due to Republican turnout. The first term says that the higher the impact of the Fox News Channel on the vote share per exposed Democrat, the higher the influence rate f. The second term subtracts the impact of mobilized nonvoters. This second term can be positive or nega- tive depending upon whether mobilization is biased toward the Republi- cans or the Democrats (that is, whether fm ­ r/(r + d) is greater or less than zero and how large its magnitude is). As mentioned in the text, if we restrict fm = 1 and impose f = m, we can simplify (11) to the formula we used in DellaVigna and Kaplan (2007), which also corresponds to our "mobilization = persuasion" case in table 6.3. This formula is as follows: f = (vT(- vC tTtC ) d eT - eC ) (12) Estimation We compute mobilization and persuasion rates for different specifications using different assumptions. We measure vT ­ vC as the impact of the Fox News Channel on the two-party Republican vote share. In our county fixed effects specifications, we use 0.0069 for presidential elections and 0.0071 for senatorial elections. In our congressional district fixed effects specifi- cation, we use 0.0042 for presidential elections and 0.0072 for senatorial elections. We measure TC, turnout in the control towns, as 0.5600 for the presidential elections and 0.5167 for senatorial elections. We measure TT, turnout in the treatment towns, as the turnout in the control town plus the turnout effect of the Fox News Channel. For presidential elections, 104 Stefano DellaVigna and Ethan Kaplan it is (1 + 0.0178)*0.56 = 0.5700 for county fixed effect specifications and (1 + 0.0046)*0.56 = 0.5626 for district fixed effects specifications. Similarly, for senatorial elections, this is (1 + 0.0158) * 0.5167 = 0.5247 for county fixed effect specifications, and for district fixed effect specifications it is (1+0.0054) * 0.5167 = 0.5195. The exposure rates eT and eC do not depend upon whether we are look- ing at senatorial or presidential elections. The exposure rate for control towns eC was 0.0262 * 3.43 = 0.089866, whereas the exposure rate for treat- ment towns eT is (0.0262 + 0.0371)*3.43 = 0.2171 with district fixed effects and (0.0262 + 0.0251) * 3.43 = 0.1760 with county fixed effects. We take our estimates of Republicans (r) and Democrats (d) from the population-weighted average of Republicans and Democrats in our sample. The variable d is the share of Democrats in the population before the entry of the Fox News Channel, computed as the two party-vote share of Demo- crats multiplied by turnout. The same is true for calculating r, the share of Republicans in the population. In presidential elections, d is equal to 0.547*0.56 = 0.2537 and r is equal to 0.453*0.56 = 0.3063. Thus 1 ­ r ­ d is equal to the percentage of eligible voters that does not turn out, which is equal to 0.44. In senatorial elections, d is equal to 0.5469*.0.5167 = 0.2826 and r is equal to 0.4531 * 0.5167 = 0.2341. Thus 1 ­ r ­ d is equal to the percentage of eligible voters that does not turn out, which is equal to 0.48. Notes 1. The results are similar if we control for the 1994 senatorial vote share instead. The disadvantage of this specification is that it restricts the sample to 2,037 towns in five states. 2. Of course, the effective persuasiveness per individual of the media can vary across countries due to differences in political systems, educational systems, competitiveness of media markets, political orientation of the media, and many other factors. 3. A more restrictive audience measure implies that availability of the Fox News Channel via cable increased the channel's audience by 2.5 to 3.7 percentage points. These audience numbers imply persuasion effects between 11 and 28 percent. We report results using those measures in DellaVigna and Kaplan (2007). 4. The mobilization effect on nonvoting Republicans could also be a persuasive effect on latent nonvoting Democrats. The Political Impact of Media Bias 105 References Ansolabehere, Stephen, and Shanto Iyengar. 1995. Going Negative: How Attack Ads Shrink and Polarize the Electorate. New York: Free Press. Bray, Margaret, and David Kreps. 1987. "Rational Learning and Rational Expecta- tions." In Arrow and the Ascent of Modern Economic Theory, ed. G. Feiwel. Lon- don, Macmillan. DellaVigna, Stefano, and Ethan Kaplan. 2006. "The Fox News Effect: Media Bias and Voting." Working paper W12169. Cambridge: National Bureau of Eco- nomic Research. ------. 2007. "The Fox News Effect: Media Bias and Voting." Quarterly Journal of Economics. 12(3): 1187­1234. DeMarzo, Peter, Dimitri Vayanos, and Jeffrey Zwiebel. 2003. "Persuasion Bias, Social Influence, and Uni-Dimensional Opinions." Quarterly Journal of Econom- ics. 115(3): 910­968. Gentzkow, Matthew. 2006. "Television and Voter Turnout." Quarterly Journal of Economics. 121(3): 931­72. Gentzkow, Matthew, and Jesse Shapiro. 2004. "Media, Education, and Anti- Americanism in the Muslim World." Journal of Economic Perspectives. 18(3): 117­33 George, Lisa, and Joel Waldfogel. 2006. "The New York Times and the Market for Local Newspapers." American Economic Review. 96(1): 435­47 Gerber, Alan S., Dean Karlan, and Daniel Bergan. 2006. "Does the Media Matter? A Field Experiment Measuring the Effect of Newspapers on Voting Behavior and Political Opinions." Yale Economic Applications and Policy discussion paper No. 12. Groseclose, Timothy, and Jeff Milyo. 2005. "A Measure of Media Bias." Quarterly Journal of Economics. 120(4): 1191­1237. Hamilton, James T. 2004. All the News That's Fit to Sell. Princeton: Princeton Uni- versity Press. Herman, Edward, and Noam Chomsky. 1998. Manufacturing Consent: The Political Economy of the Mass Media. New York: Pantheon Books. Kull, Steven, Clay Ramsay, and Evan Lewis. 2003. "Misperceptions, the Media, and the Iraq War." Political Science Quarterly. 118(4): 569­98. Lakoff, George. 1987. Women, Fire and Dangerous Things. Chicago: University of Chicago Press. Lazarsfeld, Paul, Bernard Berelson, and Hazel Gaudet. 1944. The People's Choice. New York: Duell, Sloan & Pearce. Mullainathan, Sendhil, and Andrei Shleifer. 2005. "The Market for News." Ameri- can Economic Review. 95(4): 1031­53. Puglisi, Riccardo. 2006. "Being The New York Times: The Political Behaviour of a Newspaper." Working paper 573801, March 10, 2006. SSRN. Strömberg, David. 2004."Radio Impact on Public Spending." Quarterly Journal of Economics. 119(1): 189­221. 7 Market Forces and News Media in Muslim Countries Matthew Gentzkow and Jesse M. Shapiro Introduction A growing body of evidence--reviewed by James Snyder and David Strömberg in chapter two--highlights the media's impact on pub- lic policies relevant to economic development. Governments are more likely to channel resources to consumers who have access to information through the media. Elected officials appear to work harder in support of their constituents interests when their activities are scrutinized, and the media may play an important role in disciplining corruption. Evidence that media content affects policy raises the question of what forces shape media content. A central issue is whether private ownership and competition in markets for news will foster accuracy, or merely create a tendency for news outlets to reflect the preexisting beliefs of their respec- tive audiences. Recent events have made this tension salient in the context of Muslim countries, especially in the Middle East. State control of media in these countries has eroded rapidly, through the rise of transnational satellite television networks such as Al Jazeera, the growth of the Internet, and the flourishing of free newspapers in Iraq, among other factors. Far from unam- biguously applauding these changes, however, Western policy makers have been wrestling with the fact that the media content that a free market actu- ally produces can differ greatly from what its proponents would prefer (Pincus 2003). In terms of American policy, the result has been a clash between stated support for freedom of the press on the one hand, and a series of policies such as the planting of unacknowledged American- produced news items in Iraqi newspapers (Mazzetti and Daragahi 2005) that could be seen as undermining it. 107 108 Matthew Gentzkow and Jesse M. Shapiro For free markets to produce accurate information requires three things: that consumers want to hear the truth, that markets provide incentives to give consumers what they want, and that firms respond to these incentives. None of these is a given. Each may depend on a particular country's legal and institutional environment. Understanding where and when each holds is an important input to formulating policy and understanding the links between press freedom and development more generally. In this chapter, we present evidence on these questions based on an analysis of media consumption in nine predominantly Muslim countries, drawing on data from the 2002 Gallup Poll of the Islamic World (Gal- lup Organization 2002). We first show that consumers tend to gravitate toward news sources with a perspective similar to their own. Those who report relatively negative attitudes toward the United States are more likely to report watching Al Jazeera (an Arab-language station frequently criti- cal of the United States) and less likely to report watching CNN (a station whose coverage of the United States tends to be more favorable). Moreover, a similar relationship holds between news viewership and a predictor of favorability toward the United States--the reported importance of religion in a respondent's life--that is relatively unlikely to be determined by news content. These correlations do not establish a causal relationship, but they are suggestive of a strong tendency of consumers to seek out like-minded news sources. We then discuss explanations for this tendency. The theoretical litera- ture has identified two possible mechanisms. Mullainathan and Shleifer's (2005) model posits an innate preference for confirmatory information. By contrast, Gentzkow and Shapiro (2005) propose a reputation-based mecha- nism in which consumers come to regard like-minded sources as being of higher quality.1 Using data on consumers' assessments of the quality of different news sources, we show that consumers with relatively positive atti- tudes toward the United States tend to regard CNN as significantly higher quality and Al Jazeera as significantly lower quality. Although the mag- nitude of the effect is small, this finding is at least consistent with quality inference as a possible causal mechanism. Finally, we ask whether media firms respond to consumer demand by tailoring their news to the prior beliefs of their customers. Although we do not have direct measures of the political content of news in Muslim countries, we adopt an indirect approach to studying the choices of media firms. Following George and Waldfogel (2003), we note that, because of the fixed costs in news production, newspapers tend to reflect the tastes Market Forces and News Media in Muslim Countries 109 of a wide geographic area rather than an individual consumer. As a result, targeting behavior by news outlets can be inferred from consumer demand. We test whether consumers are more likely to read the newspaper if others in their country have beliefs similar to their own. We find strong evidence for such effects in the context of newspaper and radio news, sug- gesting that political targeting of content may indeed be important for these formats. By contrast, we find no evidence of these effects for local television news, suggesting that television news is less tailored to the political prefer- ences of the local population. One possibility is that the vast majority of television stations in these countries are state owned and thus have weaker incentives to cater to the tastes of consumers. This chapter contributes to a growing literature on the economic deter- minants of news content. The most closely related paper is Gentzkow and Shapiro (2007), which studies the way market forces determine the politi- cal content of U.S. newspapers. That paper makes three main contributions. First, it proposes and implements a direct measure of newspapers' political content. The measure uses phrases spoken much more by members of one party than the other (the phrase "global war on terror" for Republicans, for example, or the phrase "tax cuts for the wealthy" for Democrats) to iden- tify, for each newspaper, the ideology to which its language is most similar. Second, Gentzkow and Shapiro (2007) present an analysis of newspaper demand that is closely related to the analysis in this chapter. Using a similar methodology to the one we implement here, they show Democratic read- ers have a relative preference for left-leaning (liberal) news and Republi- can readers have a relative preference for right-leaning (conservative) news, and the cost to newspapers of failing to match the political content their consumers prefer is large in dollar terms. Finally, Gentzkow and Shapiro (2007) show that the average level of slant in the sample is neither to the right nor to the left of the profit maximizing point, that the variation across newspapers in slant tracks the preferences of consumers fairly closely, and that news content is not strongly related to the identity of a newspaper's owner. The extent to which these results apply in other countries with differ- ent institutional environments remains an open question. State ownership, less developed markets, weaker democratic institutions, and the absence of explicit protection for press freedom could all change these conclusions substantially. The chapter provides some initial evidence on this question, suggesting that the broad outlines of the conclusions from the analysis of the United States may apply more broadly. 110 Matthew Gentzkow and Jesse M. Shapiro The 2002 Gallup Poll of the Islamic World Our data come from the 2002 Gallup Poll of the Islamic World (Gallup Organization 2002). The survey consists of 10,004 responses from nine pre- dominantly Muslim countries: Indonesia (1,050), Islamic Republic of Iran (1,501), Jordan (797), Kuwait (790), Lebanon (1,050), Morocco (1,000), Pakistan (2,043), Saudi Arabia (754), and Turkey (1,019). Although there is a slight oversampling of urban households, the samples are designed to be representative of the population ages 18 and over in each country. Interviews were conducted in person in the respondents' homes by local survey com- panies. Importantly, the methodology was designed to be sensitive to local norms, for example, by having female interviewers for female respondents. In analyses where we focus on the Arabic-language network Al Jazeera, we restrict attention to countries where Arabic is the primary language. These countries are Jordan, Kuwait, Lebanon, Morocco, and Saudi Arabia. We will use two key measures of attitudes toward the United States The first uses responses to the question: "In general, what opinion do you have of the following nations? . . . The United States." Respondents answers range from 1 for "very unfavorable" to 5 for "very favorable." We rescale the response to vary from 0 to 1. The second measure captures respondents' religiosity. This variable will be useful because it is a strong predictor of attitudes toward the United States and it is less likely than the first measure to be directly affected by news content. We use responses to a question that asked individuals to rate the importance of religion in their lives relative to four other concepts (own fam- ily/parents, extended family/local community, country, and self). The rank varies from 1 to 5, and we have rescaled it so that the measure varies from 0 to 1, with 1 implying that religion is the most important among the list. We also use several demographic measures as control variables: dum- mies for gender, education, age, urban/rural status, marital status, and fam- ily size. We also include missing data dummies for all controls. What Do Consumers Want from the News Media? The evidence cited above from the United States highlights the importance of understandingthepreferencesconsumersbringtomediamarkets.Afreepress will be most likely to achieve socially desirable outcomes when the incentives of individual news consumers and of society are aligned. When they are not, relaxing press controls could in principle make outcomes worse. Market Forces and News Media in Muslim Countries 111 What do the Gallup data say about the incentives of consumers? Two facts stand out. The first is that, as in the United States, consumers in Muslim countries tend to choose information sources that share their own views. The second is that consumers also believe that the media they choose are more likely to deliver accurate, timely information. We discuss evidence for each of these facts in turn, and we then consider what they imply about the welfare impact of a free press. Consumer Beliefs and Media Consumption Although views differ on the extent to which either Al Jazeera or CNN satis- fies objective notions of accuracy or objectivity, most would agree that the former is relatively less favorable toward the United States in its coverage than the latter. Gentzkow and Shapiro (2004) present a variety of evidence in support of this proposition. The raw data reveal a strong correlation between respondents' personal views and the media they consume. Of those who say they turn to CNN first for information about world affairs, the average favorability toward the United States is 0.43. In contrast, the average favorability among those who turn to Al Jazeera first is 0.34. The difference in the mean favorability between these two groups is equal to about a third of a standard deviation and is strongly statistically significant. Of course this correlation could be caused in part by causation running from media consumption to consumer beliefs rather than the other way around. Indeed, in Gentzkow and Shapiro (2004), we argue that the effects of CNN and Al Jazeera on viewers' beliefs appear to be large. In order to isolate the effect of consumer beliefs on media consumption, we need some predetermined characteristic of consumers that predicts their favorability toward the United States--a characteristic that is unlikely to be changed by exposure to media. We use the respondents' ranking of the importance of religion in their lives described above. Although we cannot rule out the possibility that religiosity is directly affected by the press, we expect this link to be much weaker than the impact of the press on attitudes about world affairs. The first column of table 7.1 shows the coefficient from a regression of favorability toward the United States on the importance of religion, includ- ing a fixed effect for each country and the vector of demographic character- istics described above. The table shows that religious beliefs have a strongly significant effect on favorability toward the United States (p < 0.001), with 112 Matthew Gentzkow and Jesse M. Shapiro TABLE 7.1 Consumer Beliefs and Satellite Television Viewership (1) Favorability (2) (3) toward the Turn to CNN Turn to Al Jazeera United States first? first? Importance of religion ­0.1519 ­0.0314 0.0909 (Mean = 0.76, SD = 0.30) (0.0118) (0.0076) (0.0248) Country fixed effects? Yes Yes Yes Demographic controls? Yes Yes Yes N 8,551 8,910 4,391 R2 0.1612 0.0552 0.1158 Note: Dependent variables are reported favorability toward the United States (mean = 0.33, SD = 0.33), a dummy for whether respondent turns first to CNN for information about world affairs (mean = 0.04, SD = 0.20), and a dummy for whether respondent turns first to Al Jazeera (mean = 0.40, SD = 0.49). Respondents with missing data on dependent variable or importance of religion have been omitted from the regressions reported. Results are weighted as recommended by the data providers. Demographic controls include dummies for gender, education, age, urban/rural status, marital status, and family size. Missing data dummies are included for all demographic controls. a one-standard-deviation increase in the importance of religion decreasing favorability by .14 standard deviations. The second and third columns of table 7.1 present similar regressions where the dependent variable is a dummy equal to 1 if the respondent reports that the station they turn to first for news about world affairs is CNN and Al Jazeera, respectively. In the Al Jazeera regression, we include only those respondents in Arabic-speaking countries. The results show that those who view religion as more important are significantly less likely to turn to CNN (p < 0.001) and significantly more likely to turn to Al Jazeera (p < 0.001). Taken together, these results suggest that consumers who are predis- posed to be critical of the United States are significantly more likely to choose news sources that echo this view. Of course our measure of the importance of religion will be correlated with a wide range of preferences, and some of the correlation with news choices may come from these other dimensions rather than favorability toward the United States per se. For example, it may be that Al Jazeera includes more religious programming and that accounts for some of its appeal to those who report valuing reli- gion highly. Nevertheless, the data strongly suggest that the same pattern of selecting belief-consistent news sources that we observed in the United States operates in the Islamic world as well. Market Forces and News Media in Muslim Countries 113 Consumer Beliefs and Quality Assessments An advantage of the Gallup Poll data relative to the data we analyzed for U.S. newspapers is that it includes direct measures of the way consumers perceive the quality of news sources. Respondents in all countries (except the Islamic Republic of Iran) were asked to report whether each of the fol- lowing five descriptions applies to both CNN and Al Jazeera: · comprehensive news coverage · good analyses · onsite coverage of events · daring, unedited news · unique access to information. We have constructed an overall measure of perceived quality equal to the share of these characteristics respondents feel each network possesses. This measure has a correlation of over 0.7 with each individual component and therefore seems like a good proxy for the respondents' overall attitude toward the quality of news coverage. In table 7.2 we present evidence on the way consumer beliefs influence their quality assessments. The regressions have the same controls as those in table 7.1: country fixed effects and our vector of demographic controls. TABLE 7.2 Consumer Beliefs and Assessments of Quality (1) (2) Rating of Rating of CNN quality Al Jazeera quality Favorability toward the United States 0.0380 ­0.1159 (Mean = 0.33, SD = 0.33) (0.0079) (0.0176) Importance of religion ­0.0299 0.0332 (Mean = 0.76, SD = 0.30) (0.0093) (0.0207) Country fixed effects? Yes Yes Yes Yes Demographic controls? Yes Yes Yes Yes N 8,265 7,446 4,259 4,391 R2 0.1647 0.1785 0.1315 0.1529 Note: Dependent variables are perceived quality of CNN (mean = 0.10, SD = 0.24) and Al Jazeera (mean =0.66, SD = 0.41), constructed as described in the text. Respondents with missing data on dependent variable or importance of religion have been omitted from the regressions reported. Results are weighted as recommended by the data providers. Demographic controls include dummies for gender, education, age, urban/rural status, marital status, and family size. Missing data dummies are included for all demographic controls. 114 Matthew Gentzkow and Jesse M. Shapiro We present results with both favorability toward the United States and the reported importance of religion as independent variables. The dependent variables in columns one/two and three/four, respectively, are the respon- dent's ratings of the quality of CNN and Al Jazeera. The coefficients show that consumers relatively predisposed to share the views expressed by Al Jazeera--those who are less favorable toward the United States or say religion is important in their lives--rate its qual- ity significantly higher. These same consumers view CNN as lower quality. A one-standard-deviation decrease in favorability toward the United States increases the perceived quality of Al Jazeera by 0.1 standard deviations (p < 0.001) and decreases the perceived quality of CNN by 0.06 standard deviations (p < 0.001). A similar pattern holds for increases in the impor- tance of religion, although the effect on the perceived quality of Al Jazeera is not statistically significant (p < 0.110). Implications Arguments for a free press generally center on the proposition that it will lead to a more informed citizenry. The two facts revealed by the Gallup data are somewhat contradictory from this point of view. On the one hand, the tendency of consumers to seek out like-minded information sources is often seen as prima facie evidence that they are not primarily interested in accuracy. Discussion of this pattern in the psychology literature (Lord, Ross, and Lepper 1979) and explicit models of it by economists (Mullaina- than and Shleifer 2005; Yariv 2005) both interpret it as evidence of psycho- logical "bias" and posit that consumers are willing to trade accuracy for a greater likelihood of having their personal beliefs confirmed. On the other hand, the fact that consumers consider the sources they choose to have access to better information, better analyses, more comprehensive cover- age, and so forth hints that accuracy may play an important role in their decisions. A model developed in Gentzkow and Shapiro (2005) provides one way of reconciling these facts. The key observation is that prior beliefs about the issues reported in the news will have an important effect on the way a rational consumer evaluates the quality of an unknown news source. A con- sumer who believes strongly that something is true will rationally infer that a media source that reports otherwise is more likely to have poor informa- tion, execute poor analysis, or have motives other than accurate reporting. This will lead to a situation where consumers believe that news sources that share their views are of higher quality than those that do not. Market Forces and News Media in Muslim Countries 115 As a concrete example, we suppose that there are two possible states of the world: either U.S. troops routinely target civilians, or they never inten- tionally inflict civilian casualties. There are also two types of news firms in the world: high-quality firms that almost always report the events that actually happened, and low-quality firms that frequently give distorted or inaccurate reports. Next we suppose that rational consumers turn on a new satellite news station to which they have never been exposed and see a report alleging that a U.S. soldier executed innocent civilians. How will they judge its quality? If they begin with a strong prior belief that U.S. troops often target civilians, they will think it is more likely that this report came from a high-quality source than a low-quality one, and they will raise their estimate of the qual- ity accordingly. However, if they begin with a prior belief that U.S. troops never target civilians, they will think it unlikely that the report is true. The most likely explanation is that it comes from a low-quality news source, and so they will reduce their estimate of quality accordingly. Even in a world where consumers only value accurate information, there- fore, we would expect to see exactly the pattern revealed by the Gallup data: consumers choose the sources that they believe to be highest quality, and these are also the sources that tend to agree with their own beliefs. This is not the only possible explanation for the data. It is possible, for example, that con- sumers seek out information that reinforces their beliefs and then convince themselves that the sources they chose are high quality as a form of ex-post rationalization. We have argued elsewhere (Gentzkow and Shapiro 2005), however, that evidence from a variety of contexts supports the view that the demand for accurate information is central to the demand for news, and that this mechanism is likely to explain at least a significant part of the correlation between prior beliefs and consumption choices observed in the data. Do Firms Give Consumers What They Want? Thus far, we have not presented any evidence on the way firms in the Mus- lim world respond to consumer demand. The evidence cited earlier from Gentzkow and Shapiro (2007) suggests that U.S. newspapers do respond and that this explains a large share of the variation in political coverage across papers. Whether this should also be true in the Muslim world is far from obvious. The countries in our sample do not generally have strong protections for press freedom. Many media firms are controlled directly by the state. Those that are not may be subject to strong state pressure. We do not have direct measures of content for news outlets in Muslim countries 116 Matthew Gentzkow and Jesse M. Shapiro and therefore cannot study firm responses directly. However, the Gallup data contain some suggestive indirect evidence. We focus on the content choices of local news outlets--those produced separately in each country. The differences in content of international sat- ellite networks such as CNN and Al Jazeera are certainly consistent with firms positioning to satisfy their consumers. However, local media provide a richer set of experiments, since the characteristics of their audiences vary in ways we can observe, and they also vary in the extent to which they are subject to direct state control. If local media outlets did respond to consumers, we would expect their content to track the average characteristics of the population, at least to a first approximation. In a country where most consumers had relatively pro- Western views, for example, news outlets should produce more pro-Western content. This means, in turn, that consumers whose views are similar to the average views in their country will find that the news suits their tastes and should consume more as a result; consumers who hold minority views (who does not support the West, in this example) should consume less local news. This is the pattern we will test for in the data. To do so, we again focus on what respondents report about the impor- tance of religion in their lives as the key independent variable. As discussed, this will serve in part as a proxy for the extent to which a consumer holds favorable views of the United States. Clearly, it will also be correlated with a range of other beliefs and preferences that could affect their media choices. To test the prediction that consumers will devote more time to local media when their own views are similar to the average views in their country, we focus on the interaction between what a given consumer reports about the importance of religion and the average value of this variable among all respondents from the same country. A positive coefficient on this interaction will provide sug- gestive evidence that firms are responding to consumer demand. The main dependent variables measure consumption of newspapers, radio, and television separately. Two facts will be relevant to interpreting the results. First, we expect that most radio and newspaper consumption is likely to be from local sources (although we cannot verify this directly). Television is more likely to be a mix of local and nonlocal sources. Second, Djankov and others (2003) present data on ownership of media in eight of the nine countries we study (Lebanon is not included in their study). In these coun- tries, 100 percent of television stations are owned directly by the state, while the share of state ownership in newspaper markets only exceeds 50 percent in Jordan (60 percent) and the Islamic Republic of Iran (100 percent). They do not present detailed data on radio station ownership in these countries, Market Forces and News Media in Muslim Countries 117 although they note that the level of state ownership of top radio stations internationally is high. Differences in the response of newspaper and televi- sion firms may provide evidence on the impact of state ownership. More precisely, the main dependent variables come from the following question: "How frequently do you {read daily newspapers/watch television/ listen to the radio} these days regardless of how much time you spent . . . in an average day?" Possible responses were seven days a week, six days a week, . . . , one day a week, less often than one day a week, or do not read/watch/listen. We have coded measures of media use to indicate the number of days a week the medium is used, with "less often" coded as 0.5 and "do not use" coded as 0. As a first look at the data, we run separate country-by-country regres- sions of newspaper consumption on the reported importance of religion as well as the usual demographic controls. The coefficients on importance of religion from these regressions are graphed in figure 7.1 against the average importance of religion in each country. The results are striking. The graph FIGURE 7.1 Country-by-Country Newspaper Coefficients 1.0 Kuwait Saudi Arabia religion 0.5 Pakistan of 0 importance on Iran, Islamic Morocco Rep. of ­0.5 Jordan Coefficient Lebanon Turkey ­1.0 0.6 0.7 0.8 0.9 1 Average importance of religion in country Note: Values on the y-axis are the coefficients on the importance of religion from a regression of the number of days respondent reported reading a newspaper in the last week on this variable and demographic controls, run separately by country. Values on the X-axis are the average reported importance of religion among respondents from each country. 118 Matthew Gentzkow and Jesse M. Shapiro displays a strong positive relationship: the more respondents in a country value religion on average, the more positive the relationship between the importance of religion and the consumption of newspapers at the indi- vidual level. This pattern is consistent with a significant response by firms to consumer demand. Table 7.3 presents the same facts in a regression context, along with results for the other dependent variables. The first three columns present regressions of newspapers, radio, and television consumption of the key interaction between individual and average importance of religion, a main effect of the individual importance of religion, country dummies, and the usual vector of demographics. Because average religiosity varies only at the country level, we cluster the standard errors by country. The coefficients on the interaction terms for newspapers and radio are strongly significantly positive, confirming the pattern apparent in figure 7.1. The coefficient for television, on the other hand, is insignificant and negative. The final column of table 7.3 presents an analogous regression that provides more detail on television consumption. The dependent variable in column four is a dummy equal to 1 if the television network that consum- ers report turning to first for news on world events is a local station. The TABLE 7.3 Consumption of Local Media (1) (2) (3) (4) Newspaper Radio TV Turn to local consumption consumption consumption TV first Importance of religion 3.924 2.768 ­2.081 ­0.1002 Average importance in country (1.1134) (0.9534) (1.3825) (0.2370) Importance of religion ­3.036 ­1.865 1.355 .0653 (0.8302) (0.6304) (0.9324) (0.1664) Country fixed effects? Yes Yes Yes Yes Demographic controls? Yes Yes Yes Yes N 8,910 8,910 8,910 8,910 R2 0.3607 0.2015 0.2741 0.4684 Note: Dependent variables are number of days in the last week respondent read a newspaper (mean = 2.48, SD = 2.79), listened to the radio (mean = 2.90, SD = 2.99), and watched television (mean = 5.87, SD = 2.28), and a dummy for whether respondent turns first to local stations for information about world affairs (mean = 0.50, SD = 0.50). Respondents with missing data on dependent variable or importance of religion have been omitted from the regressions reported. Results are weighted as recommended by the data providers. Demographic controls include dummies for gender, education, age, urban/rural status, marital status, and family size. Missing data dummies are included for all demographic controls. Market Forces and News Media in Muslim Countries 119 coefficient confirms that there is no significant relationship between the match between own and country characteristics and the tendency to turn to local stations for news. A possible conclusion is that state-owned local tele- vision stations respond less to consumer demand for information, although interpretation is clouded by the fact that we do not have detailed data on the ownership of radio stations. Conclusions The evidence presented here sheds some light on the way market forces drive news content in Muslim countries. The results are far from conclu- sive. Nevertheless, they suggest two themes that are likely to be important in assessing the links among media freedom, the dissemination of accurate information, and economic development. The first is that consumers appear to value media more when they believe they deliver accurate, high-quality information. Moreover, there is evidence consistent with the view that local media respond to this demand by shaping their content to appeal to consumers. These facts sup- port a larger body of evidence from a variety of sources that the desire for accuracy plays a large role in the demand for news and that this is an important driver of firm choices (Gentzkow and Shapiro 2007). All of this supports the view that strengthening press freedoms will tend to produce better information and ultimately advance social welfare. The second theme is that consumers' assessments of what is accurate are highly influenced by their own beliefs about the world. For this rea- son, the news outlets that consumers in Muslim countries believe are most reliable--that is, those whose coverage is consistent with the views of the majority of consumers in those countries--often appear to Western observ- ers to present distorted views of the facts. What does this mean for the link between a free press and an informed citizenry? On the one hand, the fact that consumers' beliefs influence their assessments of quality means that firms may have an incentive to cater to these beliefs even if this means distorting the facts. If the beliefs that citi- zens start with are wrong, this dynamic may inhibit the ability of the press to correct them. On the other hand, the fact that the ultimate driver of the market is a demand for accurate information means this distortion can only go so far. If there is some chance that the truth will be revealed even- tually, firms may prefer to contradict consumer beliefs in the short run in order to be vindicated in the long run. Such vindication is most likely when 120 Matthew Gentzkow and Jesse M. Shapiro consumers can receive information from a variety of sources, reinforcing the view that it will be desirable to have news markets that are not only free but also competitive. Furthermore, firms will be especially likely to report stories that specifically demonstrate their quality and independence; challenging incumbent governments and exposing corruption are likely to satisfy these criteria. Notes 1. Suen (2004) provides an alternative Bayesian account of confirmation-seeking behavior, based on the observation that news outlets report only a coarse sum- mary of their underlying information. References Djankov, Simeon, Caralee McLiesh, Tatiana Nenova, and Andrei Shleifer. 2003. "Who Owns the Media?" Journal of Law and Economics 46: 341­82. Gallup Organization. 2002. "The 2002 Gallup Poll of the Islamic World Report." Gallup Poll Tuesday Briefing, February. Gentzkow, Matthew, and Jesse M. Shapiro. 2004. "Media, Education and Anti- Americanism in the Muslim World." Journal of Economic Perspectives 18(3): 117­33. ------. 2005. "Media Bias and Reputation." Working paper, National Bureau of Economic Research, Cambridge, MA. ------. 2007. "What Drives Media Slant? Evidence from Daily U.S. Newspapers." Working paper, University of Chicago. George, Lisa, and Joel Waldfogel. 2003. "Who Affects Whom in Daily Newspaper Markets?" Journal of Political Economy 111 (4): 765­84. Lord, Charles G., Lee Ross, and Mark R. Lepper. 1979. "Biased Assimilation and Attitude Polarization: The Effects of Prior Theories on Subsequently Consid- ered Evidence." Journal of Personality and Social Psychology 3(11): 2098­109. Mazzetti, Mark, and Borzou Daragahi. 2005. "The Conflict in Iraq: U.S. Military Covertly Pays to Run Stories in Iraqi Press." Los Angeles Times, November 30. Mullainathan, Sendhil, and Andrei Shleifer. 2005. "The Market for News." Ameri- can Economic Review 95(4): 1031­53. Pincus, Walter. 2003. "Free Iraqi Media Called a Worthy Onus." Washington Post, November 22. Suen, Wing. 2004. "The Self-Perpetuation of Biased Beliefs." Economic Journal 114(495): 377­96. Yariv, Leeat. 2005. "I'll See It When I Believe It--A Simple Model of Cognitive Con- sistency." Draft, University of California, Los Angeles. 8 Political Economy of Media Capture Maria Petrova Introduction I n most countries, the media are the main sources of information for the majority of the population. Although the media can promote the accountability of elected politicians, they can also provide a vehicle for manipulating public opinion, even when voters understand that the media can be biased. As a result, the existence of the media creates opportunities for the rich, special interest groups, political parties, or governments to use the media to affect public opinion. The mechanisms of this influence may include investment in media firms, political advertising and paid articles, subsidies, and bribes. The situation in which media behavior is heavily influenced by special interest groups, political parties, governments, or any actors other than consumers is called media capture. However, by influencing the voting decisions of the electorate, the media can also play a positive role and constrain the behavior of elected politicians and thereby facilitate representative democracy. As Hamilton (2004) puts it, "If too many voters lack information on too many topics . . . politicians can enjoy too much freedom to pursue policies which their constituents would reject if they were actual decision makers." Gentzkow, Glaeser, and Goldin (2006) show that at the end of the nineteenth century, after most newspa- pers in the United States had become independent from political parties, the press became significantly more informative. Media content may also be affected by the patterns of media ownership. As demonstrated in Djankov and others (2003), countries with higher shares of state-owned media have less media freedom. This chapter discusses the conditions under which capturing the media can be profitable, as well as the conditions under which media capture can be facilitated. I show that countries with high inequality or small advertis- ing markets are especially likely to lack media freedom. 121 122 Maria Petrova Inequality and the Media Media coverage of issues on welfare and inequality provides a good exam- ple of issues I talk about. The mass media may be used by the rich to affect public opinion about taxation, redistribution, and the size of the welfare state. In this chapter, I describe the theoretical and empirical foundations for this argument. Classical theoretical approaches, such as the Meltzer-Richard model (Meltzer and Richard 1981) and the concept of social insurance (Diamond and Mirrlees 1978), imply a positive relationship between inequality and redistribution. Higher inequality leads to a larger gap between mean and median incomes; this gap in turn increases the popular demand for income redistribution in societies in which public policy is determined by the median voter.1 Empirically, however, higher inequality is associated with a lower level of taxation and redistribution (Perotti 1996; Alesina and Glae- ser 2004). Among different possible explanations for this phenomenon,2 this chapter emphasizes the influence of the media. As an important source of information about public policy, the media can become a tool used by the rich to influence public preferences with respect to income redistribution. If aggregate income stays constant, the greater inequality in the economy implies the lower income of the median voter, which leads to a higher popular demand for redistribution. This, in turn, increases incentives for the rich to affect the preferences for taxation and to provide for the public good of low-income voters. In Petrova (2008), I develop a model of media capture in which the rich can influence the news provided by a media outlet. Forms of influence include buying media firms, paying for political advertising, and bribery. Theory The model describes two groups of voters: high income and low income, that is, rich and poor. By assumption, the proportional income tax provides public good. The tax rate in the economy is determined by the median voter with the probability, which depends on the strength of democratic institu- tions in the country; in a perfect democracy, this probability is equal to 1. The productivity of public good provision is ex-ante unknown, but the poor can get this information from the media. As a result, the rich have an incentive to pay the media to affect voters' beliefs about this productivity and, consequently, their preferred levels of taxation. Voters are rational and understand that the media can be captured, but they do not know it with Political Economy of Media Capture 123 certitude. If the aggregate income in the economy is held constant, higher inequality implies larger gaps between the mean and the median income and between the preferred tax rates of the rich and the poor. An increase in inequality makes media capture more profitable for the rich, and the prob- ability of media capture grows. In general, the main result of the model is that higher inequality in the economy increases the willingness of the rich to spend money influencing the media. In addition, the level of public good provision should be higher if the media are not captured. The logic of the model works better in democratic societies in which policy outcomes are responsive to the position of a median voter. In less democratic regimes, the rich can use other methods to achieve desired pol- icy outcomes (such as lobbying, regulatory capture, threats, and corrup- tion). It is only when politicians are accountable to voters that "educating the society" may be an efficient way to affect policy outcomes. The media are just as likely to be captured in a more autocratic regime, but the extent of media capture would not depend on inequality in such regimes. Captured media in autocracies can also "educate society," but the focus would be on keeping the current regime in power rather than on income redistribution. The theoretical model in Petrova (2008) shows that the effect of inequality on media capture is stronger in more democratic regimes. The model also suggests that the provision of public goods in the economy is higher in countries in which the media are not captured. The model predicts that the rich have less incentive to influence the tax- ation preferences of the general public in less democratic regimes, because in such regimes they can rely on other channels of influence. This statement can be illustrated by the repeal of the inheritance tax in the Russian Federa- tion. The Russian inheritance tax was permanently abolished in 2005 after President Vladimir Putin proposed this act in his Annual Address to the Federal Assembly. Even before the final voting for this act, most articles in the press discussed the repeal as an already accomplished task that "Putin gave to the government."3 There was no discussion of the consequences of this repeal in the media except those that quoted the address.4 Empirical Results The theoretical model discussed has two important verifiable predictions. First, media freedom is negatively associated with inequality, and this effect is greater in democracies. Figures 8.1 and 8.2 illustrate corresponding empirical relationships, controlling for gross domestic product (GDP) per FIGURE 8.1 Inequality and Media Freedom for Democratic Countries, 1994­2003 3.0 NOR USA DEU AUTCAN DNK x BEL NLD IRL deni 2.5 SWE GBR FIN AUSFRAITA om NZL ederf ESP ISR a SVN die 2.0 CZE M MNG HUN ZAF POL LTU CRI PNG 1.5 THA JAM PAN 20 30 40 50 60 Gini, WIDER database Sources: Freedom House, WDI World Bank. Note: Democracy Score > ­ 9; Controlling for GDP per Capita N of Observations = 30. The democracy score is from the Polity IV dataset; all figures use average values of all variables for 1994­2003. FIGURE 8.2 Inequality and Media Freedom for Autocracies, 1994­2003 3.0 xedni 2.5 modeerf BFA UGA a KGZ di 2.0 UZB Me MRT GMB CMR VNM AZE GIN 1.5EGY MAR ZWE CHN KAZ DZA 40 50 60 70 80 Gini, WIDER database Sources: Freedom House, WDI World Bank. Note: Democracy Score <­1; Controlling for GDP per Capita N of Observations = 16. The democracy score is from the Polity IV dataset; all figures use average values of all variables for 1994­2003. Political Economy of Media Capture 125 capita.5 In democracies (figure 8.1), media freedom is a negative function of the level of inequality (p-value of t-statistics is 0.05), while the corre- sponding relationship in autocracies (figure 8.2) is slightly positive but not significant (p-value is 0.11). The relationships between media freedom and public spending on health and education as a fraction of GDP are shown in figures 8.3, 8.4, 8.5, and 8.6. The relationship is positive for democracies (p-values are 0.14 for education and 0.07 for health) and negative for autocracies (p-values are 0.17 for education and 0.97 for health). These figures illustrate that the extent of media freedom has a positive effect on the amount of public good provision in democratic countries. Corresponding regression results reported in Petrova (2008) go in line with these figures. These results show that inequality has a negative effect on media freedom, and this effect is stronger in democracies. Also, the level of public spending as a percentage of GDP is higher in those countries in which media freedom is higher, controlling for GDP per capita, level of democ- racy, and several control variables. This effect is stronger in democracies. FIGURE 8.3 Media Freedom and Public Spending on Education for Democracies, 1994­2003 3.0 NOR ISL CHE NLD DNK JPN IRL AUT CAN xedni BEL ITA 2.5 AUS FRA SWE DEU FIN GBR modeerf NZL ESP ISR CYP a GRC PRT SVN di 2.0 CZE MNG Me HUN URY MUS POL ZAF PNG CRI LTU 1.5 TTO PAN BWA THA JAM 2 4 6 8 Public spending on education (% of GDP) Sources: Freedom House, WDI World Bank. Note: Democracy Score > ­9; Controlling for GDP per Capita N of Observations = 39. The democracy score is from the Polity IV dataset; all figures use average values of all variables for 1994­2003. FIGURE 8.4 Media Freedom and Public Spending on Education for Democracies, 1994­2003 3.0 BDI YEM xedni 2.5 TCD ERI BFA modeerf ARE RWA UGA KWT a BHR LAO di 2.0 TGO KGZ UZB OMN SDN Me GMB SAU VNM MRT GIN CMR AGO AZE MAR ZWE 1.5 GABTKMEGY GNQ TUN CHN SYR DZA KAZ 0 5 SWZ 10 15 20 Public spending on education (% of GDP) Sources: Freedom House, WDI World Ban. Note: Democracy Score < ­1 ; Controlling for GDP per Capita N of Observations = 43. The democracy score is from the Polity IV dataset; all figures use average values of all variables for 1994­2003. FIGURE 8.5 Media Freedom and Public Spending on Health for Democratic Countries, 1994­2003 3.0 JPN NOR CHE AUT DNK AUS CAN ISL IRL NLD BEL FRA DEU 2.5 xedni FIN GBR ITA SWE ISR ESP NZL modeerf PRT CYP GRC SVN 2.0 MNG CZE a HUN di MUS ZAF POL URY Me TTO PNG 1.5 LTU CRI PAN THA BWA JAM 1 2 4 6 8 Public spending on health (% of GDP) Sources: Freedom House, WDI World Bank. Note: Democracy Score > ­9); Controlling for GDP per Capita N of Observations = 39. The democracy score is from the Polity IV dataset; all figures use average values of all variables for 1994­2003. Political Economy of Media Capture 127 FIGURE 8.6 Media Freedom and Public Spending on Health for Autocracies, 1994­2003 3.0 BDI YEM xedni 2.5 TCD ERI BFA ARE modeerf RWA UGA KWT BHR a KGZ di 2.0 LAO UZB TGO OMN Me GMB SAU VNM GIN MRT AGO GAB ZWE 1.5 EGY TKM MAR SYR TUN CHN SWZ DZA 1 GNQ 2 KAZ 3 4 5 Public spending on health (% of GDP) Sources: Freedom House, WDI World Bank. Note: Democracy Score < ­1) ; Controlling for GDP per Capita N of Observations = 44. The democracy score is from the Polity IV dataset; all figures use average values of all variables for 1994­2003. The results on public good provision imply that media freedom has a real effect on policy outcomes in these countries. Media and Special Interests Special interest groups can also be interested in influencing media. In Petrova (2007), I develop a theoretical model of interaction between media outlets and special interest groups and show how the structure of media revenues affects how news coverage is framed. The theoretical model shows that the size and the structure of an advertising market affect news coverage in the economy. This chapter is a contribution to a growing literature about relation- ships among media outlets and interest groups that have their own prefer- ences for media content. Herman and Chomsky (1988), Baker (1994), and 128 Maria Petrova Hamilton (2004) argue that media news is biased in favor of advertisers. Ellman and Germano (2005) describe this bias as arising from advertisers' commitment to punish publications that contain negative stories. Media bias can also arise as a result of capture by government or incumbent poli- ticians (Besley and Prat 2006; Egorov, Guriev, and Sonin 2006); interest groups (Herman and Chomsky 1988; Grossman and Helpman 2001); or journalists (Baron 2006; Puglisi 2006). Other scholars (Mullainathan and Shleifer 2005; Gentzkow and Shapiro 2006, 2007; Dyck, Moss, and Zin- gales 2005) focus on the demand side of the problem and analyze how consumer preferences for a certain type of content affects the choices made by the media. In a sense, my chapter combines supply-side and demand- side approaches and shows how media outlets frame content in response to both subsidies or advertising payments by special interest groups and the preferences of their audience. Although I primarily focus on the interplay between advertisers and media outlets, the framework of the model is more general; it can be used to describe a game among any kind of special interest group or politi- cal party and media outlets, and it can also be related to the relationship between governments and the media. In less democratic countries, most media is owned by the state. In the set of countries with democracy scores less than 8, according to Polity IV (Marshall and Jaggers 2000), the median share of state ownership of television is 1.0 (data from Djankov and oth- ers 2003). In such circumstances, government itself may choose between the higher extent of media bias and larger potential revenues from its media firms. This does not always mean that censorship in these coun- tries is explicit. According to a survey of journalists conducted by Public Expertise (2004) in 2004 in Russia, 40 percent of Russian journalists state that it does not feel "external censorship or pressure" but rather practices "self-censorship." In addition to the effects of advertising, I also show that if the prefer- ences of different special interest groups are aligned, then the resulting bias, on average, is higher than when the preferences of different special interest groups are misaligned. This proposition formalizes an argument made by Herman and Chomsky (1988):6 there is significant bias about those issues for which preferences of special interest groups are similar (for example, foreign policy), and there is smaller aggregate bias about issues for which these preferences differ significantly (for example, support for a candidate from a particular party in presidential elections). Political Economy of Media Capture 129 Audience Constraint and Special Interest Groups In my model, special interest groups7 offer subsidies to media outlets. These subsidies depend on the size of the affected audience and the extent of media bias. However, the size of the media audience decreases with the extent of bias. Some people no longer want to consume a media product if it is biased. Special interest groups face a trade-off between larger audiences and less biased coverage, and smaller audiences and more biased coverage; special interest groups are willing to pay for a bias in the direction that increases the payoff. The extent of the bias in a media product depends on the size of the advertising market and the number of competitive special interest groups. Theoretically, for media outlets there is an audience constraint that arises because of the trade-off between larger audiences and the greater extent of a bias. This constraint should be taken into account by both special interest groups and profit-maximizing media outlets. Lippmann, as early as 1922, observed that "a newspaper can flout an advertiser, it can attack a powerful banking or traction interest, but if it alienates the buying public, it loses one indispensable asset of its existence." There is some evidence that media owners deliberately choose between their audience and associated profits on the one hand, and biased news coverage on the other hand. Advertisers as powerful special interest groups have substantial lever- age over media outlets. As General Motors spokesman Brian Arke said (when General Motors stopped its advertising contract with the Los Angeles Times after a negative article by Dan Neil), "We recognize and support the news media's freedom to report and editorialize as they see fit. Similarly, General Motors and its retailers are free to spend our advertising dollars where we see fit" (BBC 2005). This example illustrates that the media can be influenced by special interest groups, for example, via implicit advertis- ing contracts. It exemplifies the existence of implicit advertising contracts, in which a media outlet not only sells its advertising space but also com- mits not to cover its advertisers negatively. A threat point here is canceling the contract, precisely as the General Motors's story shows. Herman and Chomsky (1988) provide substantial evidence about these implicit adver- tising contracts. They highlight the importance of advertising as one of "fil- ters" that "information passes before becoming the news," inducing a bias toward special interest groups. The trade-off between the audience and bias is also important for polit- ical media bias, in which politicians may capture the media to guarantee 130 Maria Petrova favorable media coverage (Besley and Prat 2006). It is true that a typical government can affect media coverage not only via subsidies but also via regulation or direct censorship. However, it can also use economic methods for its goals. Even if regulation or censorship is chosen as a way to affect media content, the trade-off between the bias and the audience limits the propaganda ability of the state; the greater the extent of the bias, the smaller the size of the affected audience. One example of political media capture is a case of NTV, the last inde- pendent national TV channel in Russia in 2001. At an April 2001 board meeting, natural gas company Gazprom effectively gained control of the channel, replacing management and appointing several of its own staff to key positions in the company. The style and content of news coverage immediately changed. As the Los Angeles Times writes, this action "gives the Kremlin control over a leading independent television network" (Dan- iszewski 2003). Change in news coverage caused a decrease in NTV's share of audience (yearly averages) from 17.9 percent in 2000 to 12.6 percent in 2001.8 The channel never received such a viewership again; its share of audi- ence in 2005 was only 11.2 percent, without significant changes in its area of coverage. Such a change in media policy is inconsistent with audience- maximization behavior (even though Boris Jordan, its new director after the change in ownership, claimed in the Los Angeles Times that he "wanted only to put NTV on its feet financially and promised to retain the station's high standards and as much of the news team as he could"). However, this change in media coverage is consistent with the theory of media capture. Advertising The theoretical model in Petrova (2007) also shows that low potential advertising revenues available to media outlets increase their dependence on the state or the political parties. Development of an advertising market, therefore, should stimulate growth of independent mass media. When the advertising market is large, media outlets are expected to closely follow the interests of their consumers, in contrast to the interests of political par- ties and the government. Owners of media outlets may be viewed as just another special interest, and larger potential advertising revenues increase the monetary part of their payoff, so the theoretical proposition holds even if owners themselves are politically biased (see also Gabszewicz, Laussel, and Sonnac 2001). Political Economy of Media Capture 131 My theory is about transition from Lippmann's (1931) first or second stage of development of newspapers to the third stage: from the media sys- tem in which newspapers depend on the state or the political parties to the media system in which newspapers deliver news preferred by consumers. The main prediction of the theory is that growth of advertising stimulates growth of independent media. I illustrate this prediction by evidence from development of the Russian media and from American newspapers of the nineteenth century. Advertising Revenues and Press Freedom in Russia Freedom of the press in developing countries can be constrained not only by deliberate censorship policies of the state but also by the structure of adver- tising and media markets. Russia provides a particularly good example, since there is no official censorship or strict formal media regulation, and media outlets are left to operate as commercial profit-maximizing firms. However, the existence of media freedom in Russia is questionable. For example, for International Press Institute (IPI), Russia was the first country to be placed on the IPI Watch List (IPI 2006). Freedom House named Russia among those countries having "not free" media since 2003 (Freedom House 1994­2005). Reporters without Borders ranked it as 138 out of 167 in 2005 (with the highest number being the worst, 167 corresponds to North Korea) (Reporters without Borders 2005). A number of analysts and policy makers emphasize the importance of advertising revenues as a prerequisite to the development of an inde- pendent media in Russia. As Henrikas Yushkiavitshus notes in his annual Andersen lecture, "It will be possible to root out this phenomenon [system of paid-for articles masquerading as straight news reports] only if and when journalists can earn decent wages and when newspapers can be profitable from such normal income sources as advertising and circulation sales" (Yushkiavitshus 2004). Similarly, Michail Lesin (2002), former Russian minister of press, television, and telecommunications, said in an interview in 2002 that "neither advertisement nor subscription revenues can provide enough support for all existing media outlets. As a result, many of them depend on political parties, power and business structures, and, often, on criminal groups." Some suggestive quantitative evidence in favor of an advertising-media freedom hypothesis is presented in table 8.1. In 2000, the nongovern- mental organization Obshchestvennaya Expertiza constructed indexes of 132 Maria Petrova media freedom for different Russian regions and estimated advertising revenues of regional media outlets. Table 8.1 shows that media freedom in Russian regions depends on advertising revenues, state subsidies, con- trol variables (including gross regional product) per capita, log of popula- tion, and the consumer price index).9 Aggregate advertising revenues have a significant positive effect on regional press freedom, as predicted by the model, while state subsidies have a negative though insignificant effect. The size of the sample is restricted to 49 out of 89 Russian regions, and this restriction is driven by data availability. Column two of the table has the same regression results with the city of St. Petersburg excluded, to demon- strate that results are not driven solely by this outlier (St. Petersburg has extremely high advertising revenues compared with other regions; Moscow is not in the sample because the data were not available). TABLE 8.1 Regional Media Freedom and Advertisement Revenues, Russian Media Media freedom index, 2000 Obshchestvennaya Expertiza Whole sample Excluding St. Petersburg Advertisement revenues, 0.05 0.06 regional media [3.32]*** [2.08]** State subsidies, ­0.026 ­0.017 regional media [0.45] [0.27] Log (gross regional product) ­20.842 ­22.219 [2.36]** [2.43]** Level of unemployment ­1.222 ­1.333 [1.81]* [1.87]* Average monthly nominal wage 0.009 0.01 [2.58]** [2.66]** Log (population) 18.879 19.516 [2.07]** [2.11]** Density of population 0.011 0.02 [0.14] [0.25] Consumer price index ­0.441 ­0.488 [0.89] [0.97] Observations 49 48 R-squared 0.22 0.18 Note: Robust t statistics in brackets * significant at 10%; ** significant at 5%; *** significant at 1%. Political Economy of Media Capture 133 Advertising Revenues and the Rise of Independent Media Starr (2004) in his historical analysis of development of the media in the nineteenth century, comments on the importance of potential advertising revenues: "In one case [that of Europe], taxes on advertising limited the ability of newspapers to become commercially self-sufficient and indepen- dent of the state; in the other [the U.S.], tax-free advertising enabled news- papers to grow with the expansion of commerce." This can be interpreted as evidence in favor of the theoretical model, since the technology of media production was quite similar in Europe and the United States at that time (except for taxes/subsidies for production of one copy of a newspaper, which TABLE 8.2 Description of Variables and Data Sources Regional media Index of media freedom for regional media, compiled on the freedom basis of three indices: freedom of access to information, freedom of production of information, and freedom of distribution of information. Source: Obshchestvennaya Expertiza; www.freepress.ru Advertising Aggregate amount of money spent on advertising in regional media revenues, outlets, expert estimates regional media Source: Obshchestvennaya Expertiza; www.freepress.ru State subsidies, Aggregate amount of money spent by regional governments to regional media subsidize media outlets; data extracted from official regional budget reports Source: Obshchestvennaya Expertiza; www.freepress.ru Gross regional Gross regional product (in rubles) product Source: Russian regions 2005, Goskomstat Average Average monthly nominal wage (in rubles) monthly Source: Russian regions 2005, Goskomstat nominal wage Population Population (in thousands of persons); data for January 1, 2000 Source: Russian regions 2005, Goskomstat Population Density of population (persons per sq. km); data for January 1, 2000 density Source: Russian regions 2005, Goskomstat Level of Level of unemployment (% of economically active population) unemployment Source: Russian regions 2005, Goskomstat Consumer Consumer price index (December to December previous year, %) price index Source: Russian regions 2005, Goskomstat 134 Maria Petrova Starr and Hamilton consider to have a separate effect). In addition, accord- ing to Starr (2004) and Hamilton (2004), it was the growth of advertising that eventually allowed the press to become independent of political par- ties. The growth of commercial advertising "virtually eliminated the role of political patronage in [the] economics of daily newspapers" and "provided an economic basis for press autonomy and power that early printers and editors never enjoyed" (Starr 2004). Examples from U.S. experiences in the nineteenth century are useful for a theory about modern developing countries, as the political and economic environments in the United States at that time were similar to environments in many modern developing countries. Figure 8.7 shows how the number of independent newspapers was different in U.S. counties with high advertising rates and low advertising rates. The data is from Ayer's Directory of Newspapers, 1881­1885, which for each newspaper indicated its political affiliation (for example, "Independent," "Republican," "Democratic"). As figure 8.7 illus- trates, the counties with higher advertising rates at that time had a larger frac- tion of independent newspapers. This finding also supports the hypothesis that development of advertising stimulated growth of the independent press. FIGURE 8.7 Fraction of Independent Newspapers in U.S. Counties as a Function of Advertising Rates in 1881 0.29 0.28 High advertising in 1881 0.27 0.26 0.25 Low advertising in 1881 0.24 0.23 0.22 1881 1882 1883 1884 1885 Source: Ayer's Directory of Newspapers, 1881­1885. Political Economy of Media Capture 135 Conclusion Media outlets can be influenced by different players: by the rich, by spe- cial interest groups, by political parties, and by the state itself. Even when media consumers are rational and understand that media can be biased, the lack of accurate information prevents them from making informative voting choices. The results of this chapter show that media freedom is negatively asso- ciated with inequality and the preferences of the median voter with respect to social spending. Also, media freedom and the growth of independent media are more likely when advertising markets are developed. When potential advertising revenues are high enough, media should pay atten- tion to the interests of the general audience in order to attract advertisers. However, when advertising revenues are high, there is a bias toward special interest groups. Media freedom has a real effect on policy outcomes: figures 8.3, 8.4, 8.5, and 8.6 illustrate that democratic countries with a greater extent of media freedom have higher spending on health and education, controlling for GDP per capita. This effect disappears in autocracies. This chapter shows the conditions under which media capture may be profitable. Economies with high inequality or small advertising markets are especially prone to political media capture and are likely to suffer from the lack of media freedom. Notes 1. In the theoretical literature on voting, the median voter is a voter with prefer- ences in the middle of one-dimensional policy space. Theoretically, if two par- ties choose a policy platform to compete for people's votes, their best choice is to choose the policy preferred by the median voter, if they want to maximize their vote share and assume other party's rationality. 2. A number of explanations were offered recently to explain this puzzle; the explanations stress the differences between high-inequality and low- inequality countries in cultural heritage (Alesina and Fuchs-Schundeln 2006); in the perceptions of a social mobility (Piketty 1995; Alesina and Glaeser 2004); in the preferences for fairness (Alesina and Angeletos 2005; Benabou and Tirole 2006); and in the coercive power of the state (discussed in Boadway and Keen 2000). 3. "Polnoe bezvozmezdie," Vremya novostei, May 12, 2005. ("Full retaliation," Time of the News). 4. Content analysis of articles in more than 100 Russian national and local newspapers, 2005 (ISI Emerging Markets media archive www.securities.com) 136 Maria Petrova revealed only three articles discussing possible benefits and drawbacks of the tax; two of them were interviews. 5. Data on media freedom are from Freedom House (1994­2005); data on inequality are from WIDER (2005). 6. Herman and Chomsky (1988) in their famous book Manufacturing Consent develop a "propaganda model" of mass media which challenges basic prem- ises of operation of the media and suggests that the media serve antidemo- cratic ends. They claim that as information passes five different "filters" before becoming news, the resulting media policy is biased toward the rich and the powerful. These filters are: (1) the influence of ownership, managers, and shareholders together; (2) the interests of advertisers; (3) the regulation of access to information by government agencies; (4) the ability of the power holder or government authorities to punish deviators; and (5) the ideology (communism versus anticommunism). However, the authors explain the rela- tive absence of aggregate media bias about such issues as which party to sup- port, by the misalignment of preferences of interest groups in the economy. 7. Here special interest groups may be broadly defined; they can include journal- ists, media owners, government officials, political parties, and advertisers. 8. Data are from the analytic report "Television in Russia 2005" prepared by Video International, a leading Russian advertising agency. 9. These results may be more precise than possible cross-national comparisons, as legal, economic, and social environments in Russian regions are very similar. References Alesina, A., and G. M. Angeletos. 2005. "Fairness and Redistribution." American Economic Review 95(4): 960­80. Alesina, A., and N. Fuchs-Schundeln. 2006. "Good-bye Lenin (or Not?): The Effect of Communism on People's Preferences." Working paper, Harvard University, Cambridge, MA. Alesina, A. and E. Glaeser. 2004. Fighting Poverty in the US and Europe: A World of Difference. Oxford: Oxford University Press. Baker, C. E. 1994. Advertising and a Democratic Press. Princeton: Princeton Univer- sity Press. Baron, D. 2006. "Persistent Media Bias." Journal of Public Economics 90(1­2): 1­36. BBC News. 2005. "GM stops advertising in LA Times," April 8. http://news.bbc. co.uk/2/hi/business/4426107.stm. Benabou, R., and J. Tirole. 2006. "Belief in a Just World and Redistributive Politics." Quarterly Journal of Economics 121(2): 699­746. Besley, T., and R. Burgess. 2002. "The Political Economy of Government Respon- siveness: Theory and Evidence from India." Quarterly Journal of Economics 117(4): 1415­52. Besley, T., and A. Prat. 2006. "Handcuffs for the Grabbing Hand: Media Capture and Government Accountability." American Economic Review 96(3): 720­36. Political Economy of Media Capture 137 Boadway, R., and M. Keen. 2000. "Redistribution." In Handbook of Income Distri- bution, ed. A.B. Atkison and F. Bourguignon. North-Holland, Amsterdam, Netherlands. Daniszewski, J. 2003. "The World Firing of Media Chief Points Up Kremlin Poli- tics." Los Angeles Times, January 21. Diamond, P. A., and J. A. Mirrlees. 1978. "A Model of Social Insurance with Variable Retirement." Journal of Public Economics 10(3): 295­336. Djankov, S., C. McLeish, T. Nenova, and A. Shleifer. 2003. "Who Owns the Media?" Journal of Law and Economics 46(2): 341­81. Dyck, A., D. Moss and L. Zingales. 2005. "Mass Media versus Special Interests." Working paper. Egorov, G., S. Guriev, and K. Sonin. 2006. "Media Freedom, Bureaucratic Incen- tives, and the Resource Curse." Discussion paper no. 5748, Centre for Eco- nomic Policy Research, London. Ellman, M., and F. Germano. 2005. "What Do Papers Sell?" Working paper no. 800, Department of Economics and Business, Universitat Pompeu Fabra, Barcelona. Freedom House. 1994­2005. Freedom of the Press. Washington, DC: Freedom House. Gabszewicz, J., D. Laussel, and N. Sonnac. 2001. "Press Advertising and the Ascent of the "Pensee unique." European Economic Review 45: 645­51. Gentzkow, M., and J. Shapiro. 2006. "Media Bias and Reputation." Journal of Politi- cal Economy 114(2): 280­316. ------. 2007. "What Drives Media Slant? Evidence from U.S. Daily Newspapers." Working paper, University of Chicago. Gentzkow, M., E. L. Glaeser, and C. Goldin. 2006. "The Rise of the Fourth Estate: How Newspapers Became Informative and Why It Mattered." In Corruption and Reform: Lessons from America's Economic History, ed. Glaeser and Goldin. Cambridge, MA: National Bureau of Economic Research. Glaeser, E., J. Scheinkman, and A. Shleifer. 2003. "The Injustice of Inequality." Journal of Monetary Economics 50(1): 199­222. Grossman, G. M., and E. Helpman. 2001. Special Interest Politics. Cambridge, MA: MIT Press. Hamilton, J. T. 2004. All the News That's Fit to Sell: How the Market Transforms Infor- mation into News. Princeton: Princeton University Press. Herman, E. S., and N. Chomsky. 1988. Manufacturing Consent: The Political Economy of the Mass Media. New York: Pantheon. IPI (International Press Institute). 2006. Watch List Country Report, Russia. Vienna, Austria: IPI. Lesin, Michail, 2002. Proceedings of the conference, "The Power of the Press and the Press of Powers." Moscow. Transcript in Russian available at http://www. radioportal.ru/ed/18.shtml. Lippmann, W. 1922. Public Opinion. New York: Macmillan. ------. 1931. "Two Revolutions in the American Press." In The Essential Lippmann, ed. C. Rossiter and J. Lare, 402­406. Cambridge, MA: Harvard University Press. 138 Maria Petrova Marshall, Monty G., and Keith Jaggers. 2000. "Polity IV Project: Political Regime Characteristics and Transitions, 1800-1999 Dataset Users' Manual." College Park, MD: University of Maryland. Meltzer, A. H., and S. F. Richard. 1981. "A Rational Theory of the Size of Govern- ment." Journal of Political Economy 89(5): 914­27. Mullainathan, S., and A. Shleifer (2005) "The Market for News." American Economic Review 95(4): 1031­53. Perotti, R. 1996. "Growth, Income Distribution and Democracy: What the Data Say." Journal of Economic Growth 1(2):149­87. Petrova, M. 2008. "Inequality and Media Capture." Journal of Public Economics, Feb- ruary, 92 (1­2): 183­212. ------. 2007. "Mass Media and Special Interest Groups." Working paper, Harvard University, Cambridge, MA. Piketty, T. 1995. "Social Mobility and Redistributive Politics." Quarterly Journal of Economics 110(3): 551­84. Public Expertise. 2004. Moscow. http://www.freepress.ru/publish.shtml. Puglisi, R. 2006. "Being the New York Times: The Political Behaviour of a Newspa- per." Political Economy and Public Policy working paper 20, STICERD, Lon- don School of Economics, London. Reporters without Borders. 2005. Annual Worldwide Press Freedom Index. Washing- ton, DC: Reporters without Borders. Starr, P. 2004. The Creation of the Media: Political Origins of Modern Communications. New York: Basic Books. Strömberg, D. 2004. "Radio's Impact on Public Spending." Quarterly Journal of Eco- nomics 119(1): 189­221. Yushkiavitshus, Henrikas. 2004. "Why Press Freedom Is Disappearing in the Post- Communist World." Address at the 18th Annual Harold W. Andersen Lecture Series, World Press Freedom Committee, Reston, VA, December 9. 9 Fostering an Independent Media with a Diversity of Views Joseph Stiglitz I have been interested in the question of how to promote a competitive and pluralistic media for a very long time. In a way, it is a very natural outgrowth of my work on the economics of information. My research ana- lyzed how information is important for the economy to function well and how there are systematic distortions in incentives to provide information that lead to information asymmetries and a whole set of other problems. My more recent interests have been in information and political pro- cesses, that is, how imperfect information affects political processes, just as my earlier work focused on how imperfect information affects economic processes. Here again I have attempted to analyze incentives: why there are often incentives to provide distorted information and how we can make political systems work more effectively, both through the provision of infor- mation and by the adoption of certain regulatory and legal frameworks. This chapter will deal primarily with the impact of information pro- vided by the media on public policy. However, much of what I will discuss also applies to the importance of information in the functioning of mar- kets, for instance, in identifying market failures, including abuses of market power and failings in corporate governance. The view that good policies require good information is widely accepted. In a democratic society, the views of citizens are critical. Democ- racy is not meaningful if people do not know what they are voting about. Voters' beliefs are shaped by what they see and read in the media. The author is a professor at the Columbia University and Chair of the Brooks World Poverty Institute at Manchester University. He is indebted to the Ford, Mott, and MacArthur foundations for their financial support. Research assistance was pro- vided by Dan Choate and Jill Blackford. The personal views expressed in this paper are those of the author and not the World Bank. 139 140 Joseph Stiglitz The consequence is that distortion of information can contribute to views that are distanced from reality. We saw a dramatic example of this in the months prior to the commencement of the Iraq war, where there were markedly different perceptions of reality in different countries, each shaped by the distinctive media coverage in the respective countries. The differences in perceptions were, to a large extent, a result of differences in media coverage.1 The further consequence is that different views of reality lead to support for different positions and policies. Those countries where media coverage was particularly distorted adopted policies that were not only adverse to global interests but arguably were adverse even to national interests. One way of focusing more narrowly on the role of information in dem- ocratic societies is to observe that underlying well-functioning democracies is a system of checks and balances. British historian Lord Acton said that power corrupts and absolute power corrupts absolutely. There is another aphorism, "knowledge is power." The absence of information gives those who are in policy making positions a level of power that would be limited by greater access to information. Corruption provides another example illustrating the role of information--and the media--in political processes. The subject of corruption has recently received considerable attention at the World Bank, and rightly so. Corruption is an "agency" issue, that is to say, individuals with delegated authority advance their own interests in particularly egregious ways at the expense of those whose interests they should be representing. Agency problems arise because of imperfections of information; they would not occur in a world with perfect information. If we could observe what was happening instantaneously all the time, these agency problems would not arise and thus the scope for corruption would be limited. People would realize that they would be caught and would act appropriately. Monitoring provides a check, and it is an important way of dealing with agency problems. The problem is that monitoring is a public good, and, more broadly, the public good is a public good. Managing society in ways that benefit every- one is a public good. In some ways, it is the most important public good. Ensuring that the government acts in ways consistent with the public good is itself a public good, and monitoring, which is essential to this process, is a public good. It follows, then, that since in general there is an undersupply of public goods, there will be an undersupply of monitoring. In democratic societies, the media are one of the key monitors. But if the media are to perform their role, two "rights" have to be recognized: the Fostering an Independent Media with a Diversity of Views 141 right to tell--or freedom of the press--and the right to know, highlighting the importance of having access to information, reflected in Freedom of Information Acts. There has been great progress in recent years in the pas- sage of such acts around the world. After leaving the World Bank, I helped create a think tank at Columbia University called "The Initiative for Policy Dialogue," and we have pushed these Freedom of Information Acts. Many countries have adopted these laws; the real challenge now is enforcement. Even in the United States we have trouble with the government understand- ing the intent and the purpose of this law. Perhaps it would be more accu- rate to say that those in power do understand not only the purpose of the law but also its effects; it is because they know it limits their power and their opportunities for corruption by making them more accountable that they have worked to limit the scope of the law. I do not mean to imply that circumscribing corruption and strength- ening the system of checks and balances so essential for a functioning democracy are the only reasons that it is important to have good informa- tion. Even without corruption, all individuals are fallible, and the conse- quence of human fallibility is that there has to be shared decision making. A number of years ago my colleague Raaj Sah and I described a framework for thinking about human fallibility and collective decision making.2 For example, we explored how the structure of decision making could affect the tradeoffs between type one and type two errors. (Type one errors are also known as false positives. To use a medical example, if tests for cancer reveal that persons have cancer when in fact they do not, these would be type one errors. Type two errors are known as false negatives. Returning to our example, this would describe situations where the tests failed to iden- tify those who actually had cancer.) Our work provides another argument for having widespread participation in decision making by reducing the likelihood of these errors, but effective participation requires information. Again, the media play a key role as an information intermediary, and there is a problem of the undersupply of the public good of information. I have talked about the role of information in monitoring and in enabling shared decision making. I have suggested that there will be an undersupply of information in either case. However, the issue is not only the level of information supplied but also the nature of the information supplied, or in other words, distortions in the information supplied. The problem is that those who supply information often have an incentive to supply distorted information, and this is true both in the public and the private sectors. 142 Joseph Stiglitz The problems in the public sector are obvious--politicians do not want citizens to know of the corruption in which they are engaged, to have their power circumscribed or their mistakes to be exposed--but let me empha- size that these are just as important in the private sector. The point is that decisionmakers in both the public and private sectors do not like checks and balances. People like to have their way and operate without someone looking over their shoulders. People particularly do not like to be monitored when they are making mistakes, even if the mistakes are honest and not corrupt. The problem is that we often have a difficult time distinguishing between honest mistakes and corrupt mistakes. If there were any question about that difficulty, we can look at Iraq and Hurricane Katrina, where massive mis- takes have occurred and massive amounts of money have been spent with outcomes that are disappointing, to say the least. The question is to what extent these outcomes are due to corruption or to incompetence. The general answer is that there were some of both, but the difficulty lies in trying to parse out what fraction of the failures is due to corruption and what fraction is due to incompetence. In the end, it may not be that important to distinguish between the two, as long as we arrive at the correct decision of what to do about the situation. Government officials want to increase support for their positions by enhancing the belief of likely success, and so they have a strong incentive for supplying distorted information. Unfortunately, we have seen ample evidence of this in recent years. But this is only part of the problem. There are also opportunities for more pervasive corruption in the supply of information. This is another agency problem that is widespread throughout the government: the mar- ginal cost of providing information should be zero, but an artificial scarcity can be created through secrecy, thereby reducing the flow of information and creating something of value. Free information has no value because everybody has it. Restricting the supply of information, however, gives it value, and those who have that information then can trade it. Now, for the most part in the United States, we do not have outright corruption in the sense that it exists in some other countries, where offi- cials actually take bribes. We do have an American form of corruption that is possibly more corrosive of our society, and it is larger in magnitude in terms of absolute dollars (largely because the economy is larger). It is traded in what is sometimes called "gift exchange": I create something of scarcity, which somebody then wants, meaning there is room for a deal. I give him what he wants, and he gives me what I want. What the government official Fostering an Independent Media with a Diversity of Views 143 wants from the typical business person is a campaign contribution. What the government official wants from the media are stories written in the way that the official wants. This may include giving more weight to the slant of the government or putting the government official in a more pivotal role than is the case. It may include, as part of the package, a couple of puff pieces about how great the government official is. (When we see a puff piece in the newspaper, we can be fairly sure that it was part of a gift exchange, a return for getting some piece of information.) When I was in the White House, first as a member of the Council of Eco- nomic Advisers (1993­95) and then as its chairman (1995­97), it was obvi- ous who "owned" which reporters and what they were getting in exchange for their stories. Often there are well-established relationships, with each reporter owned by a particular member of the administration and providing "services" in return for which he gets access to so-called "inside informa- tion." I understood the market in information that was at work; but most of us do not see it as clearly. This makes it more difficult to evaluate the infor- mation that has been provided and to understand the ways in which it may have been distorted. In other cases, access to information is an exchange for campaign contributions, although this is probably not the major source of corruption in campaign contributions; far more important, sources are pieces of legislation, like our prescription drug benefits that cost Americans amounts probably in the hundreds of billions of dollars. These are good investments, high returns for relatively small campaign contributions. On the supply side of the market, are the government officials who create an artificial scarcity of information; but there are lots of people on the demand side, both businesses and journalists. Knowledge is power, and access to information can be converted into profits. More specifically, hav- ing information before anyone else can make a great deal of money. How- ever, this does not necessarily have any social return. Let me make it clear: a lot of this information has very little social return, but knowing something before somebody else could give a high private return. Obviously, there is also demand from the media, with kudos and sales for those who break news. In both cases, there are high private returns to having information just slightly before others, and there is a large disparity between social and private returns. This problem of artificially creating scarcity is one of the most impor- tant reasons why it is necessary to have a Freedom of Information Act, as well as to have an independent enforcement agency, because the govern- ment has strong incentives not to comply or to comply in a minimal way. 144 Joseph Stiglitz Even without corruption, there can be honest differences of opinion about what is in the national interest; even without corruption, governments still want to distort the flow of information, especially when government poli- cies do not seem to be working in the way advertised. This is a problem in the private sector as well. The cigarette industry pro- vides one of the obvious cases. It tried to distort the public's beliefs about the amount of carcinogens in light cigarettes and the effects of smoking on can- cer and on health more broadly. The members of the cigarette industry had the information; they knew what the effects of smoking were; and they were in a position to engage in research to get still more information. They knew that light cigarettes delivered just as many carcinogens as regular cigarettes, but they also knew that it was not in their interest to have this information disseminated. Instead, they created information saying that there was uncer- tainty about the science leading to these conclusions, even though there was, in fact, no uncertainty that cigarettes were bad for health. More recently, Exxon paid money to so-called "think tanks" to dissemi- nate the same kind of uncertainty about climate change. There are some uncertainties, for example, about the pace of change. There are even uncer- tainties about how serious it is--whether it will be equivalent to the loss of 5 percent or 7 percent of gross domestic product. However, there is no doubt that global warming is real, that it will have a devastating effect on many developing countries, that there have been significant effects from our use of fossil fuels, and that it is an issue we ought to do something about. These two cases illustrate the worst in corporate practices; but the prob- lems are more pervasive (Stiglitz 2003). Each firm wants potential investors to believe that its expected returns are high. This problem is exacerbated by stock options where the compensation of the chief executive officer (CEO) increases with the stock price; this gives CEOs an incentive to provide dis- torted information to shareholders and the media. (There was a more per- vasive web of distorted incentives: analysts, the investment banks, and the accountants all had incentives to provide distorted information.) Thus, the underlying problem, to which the media have at last caught on, is that those who have control of information often have incentives to provide distorted information. There is no perfect solution to these problems. The problem is what can be done to mitigate these risks, for instance, by changing incentives or cre- ating various systems of checks and balances in the provision of informa- tion. In the private sector, the Sarbanes-Oxley Act3 has changed incentives for accounting firms. Before the law was enacted, accounting firms were Fostering an Independent Media with a Diversity of Views 145 hired by the CEO of the company rather than by the audit committee, and thus the accounting firms' incentives were to please the CEO. This meant that they were not as concerned about getting accurate information as they were about getting information that the CEO liked. The CEO wanted posi- tive news that would lead to an increase in the stock prices--and thus in the value of options. Before the Sarbanes-Oxley Act, most accounting firms also got most of their money from consulting, which distorted these incentives even more. At the root of the problem is the way CEOs are compensated; stock options provide a strong incentive for distorted information. The media play a critical role in monitoring and in providing infor- mation relative to public decision making. The problem is that, while the media in principle can play this role, they too have incentives that may not be well-aligned with the interests of society. For instance, the reporter or newspaper that gets the news first gains more attention and more advertis- ing dollars, which in turn contributes to the patterns of distorted informa- tion described earlier. This is where diversity comes in. Multiple media can provide an impor- tant set of checks and balances. In other words, each reporter or newspaper has an incentive to uncover the mistakes or distortions of others. I want to emphasize that what is relevant is not just the total number of media outlets; one has to look closely at the structure of the industry. Anyone who has been engaged in ascertaining the degree of competition in markets for purposes of antitrust is familiar with this kind of problem. For instance, in banking, one does not look at the total number of banks. There are a lot of banks in the United States, but that does not necessarily mean that the market is highly competitive. Instead, one must look at each market by region and sector, for instance, competition in the supply of loans for small businesses in rural areas within the state of Washington. Despite the large number of banks, few banks serve this particular market, and competition is very limited. The same thing is true in the context of the media. What is important is how many sources of news or information any one viewer has, or how many sources of coverage any one company has. How many checks are there, how many people are providing independent information to any one person, and how many people are looking at the behavior of a company or a particular politician? We also have to examine closely what the differences are within media, that is, whether there are significant differences in incentives and perspec- tives. Both the amount and the kind of information are of concern, and one of the problems with privately funded or privately owned media is that 146 Joseph Stiglitz their business is to make money. This leads to a whole set of issues: they worry about offending advertisers or a large number of readers, and they worry about offending government officials and thereby losing access. Let me be clear: these are real worries. I saw what happened when jour- nalists do not pay sufficient attention to these concerns. A reporter for a prestigious U.S. newspaper who covered the Federal Reserve (Fed), with very good access, thought he was supposed to write about what was going on at the Fed, and he thought that meant occasionally writing critically about what was happening. However, the Fed did not view that as the job of reporters covering the Fed, especially those who were given "access"; rather, the Fed viewed the job of the reporter as being more of a public relations outlet. This young reporter, who was getting front-page news stories in one of the major newspapers, made a mistake: he wrote a critical article. It was not long before he was reporting about automobiles in Detroit.4 He had lost access, and without access, he had lost his value covering the Fed. His loss of access revealed the dangers of offending government officials. Another set of problems arises when the owners of media outlets have economic interests that might be affected by how they cover a story. For instance, Americans are debating why the Iraq war coverage was so bad-- and so distorted--in the United States before the war. There are a number of explanations, but one is that several of the newspapers were trying to get television licenses at the time and did not want to offend the regu- latory authorities, where presidential appointees predominated. Whether this is true or not, it at least alerts one to the possibility that politics can color news coverage, even apart from outright censorship and even in a democratic society. It should make us particularly sensitive to the media being subjected to economic pressures from politically controlled regula- tory authorities, particularly when discretionary decisions are being made by regulators. Imposing restrictions (such as that newspapers cannot own television and radio stations) reduces the scope for such self-censorship and may increase media diversity. The problem is exacerbated by the fact that many people, including rich business people, want to shape the world. There is nothing wrong with this, and they should be commended for their public spiritedness. But they are able and willing to spend a great deal of money to do so. Some owners of media outlets clearly view their control as a mechanism for advancing particular political positions. That this is a concern is evident in the own- ership restrictions on media that many countries have. The United States has ownership restrictions on television, and one of the reasons is the Fostering an Independent Media with a Diversity of Views 147 worry that foreign owners might promulgate a view that is un-American. These concerns highlight the fact that ownership matters. But if the rich disproportionately control the media, and if the rich disproportionately share common views (for example, on the desirability of low taxes, the elim- ination of the estate tax, and fears of big government), the media may not be able to do their job: the media may supply distorted news coverage and/ or may be perceived as doing so. At the very least, there will not be a diverse set of perspectives that should underlie active democratic processes. There have been real concerns about media ownership in the Russian Federation. In the Russian election of 1996, the oligarchs clearly used their positions and wealth to advance the reelection of Boris Yeltsin, in return for which they received large ownership shares in what was called the "Sale of the Century" (Freeland 2000). Some estimates put the value of state assets that were transferred to the oligarchs at bargain-basement prices at between US$1 trillion and US$1.5 trillion, and this has had a long-term effect on Russian society. People in Russia say they went from a state- controlled media that gave them very little diversity to an oligarch- controlled media that gave them very little diversity. Now they are return- ing to a more mixed system, where censorship may be more subtle, or at least less formal, but no less repressive. If the media are to perform their critical role in democratic societies, it is important to have a diversity of views. This means that competition is more important in the media than in other sectors. The usual standards for com- petition in other sectors, such as monopoly power (evidenced by the ability to charge prices above marginal costs), are not the appropriate ones for this sector. What is at stake here is more fundamental: the effective functioning of our democratic society. Thus, the criteria for assessing excessive concen- trations of media power need to be markedly different and markedly tighter. Also, because different media reach different audiences, the definition of "market concentration" needs to be very attentive to the issues of both cov- erage and audience. We need diversity not only within each media but also diversity across media. We should accordingly worry about ownership of television and radio by newspapers because it reduces the overall scope for diversity. This may be a problem of increasing concern, as economies of scope act as a natural force leading to increased media concentration. There are other important aspects of a program for designing a more effective and diverse media, in addition to strong competition and diver- sity laws. For instance, we should encourage the entry of foreign media, for this broadens the set of perspectives that confront citizens. Moreover, 148 Joseph Stiglitz foreign media may be less subject to some of the political pressures to which I alluded earlier. Also, as we have noted, some countries have strong restric- tions on foreign ownership of media. Worry about foreign domination of media is understandable, but foreign views can be an important check. In the United States, in the months preceding the war in Iraq many Americans found that the only accurate sources of information, judged both ex ante and ex post, were provided by the BBC and the Financial Times. The Ameri- can media did not do their job; but, even if the information provided by the foreign media turns out to be wrong, having a diversity of views on a global issue like war is important. At the very least, it is important for citizens to see the information that is being provided to those in other countries, unfiltered through domestic media.5 There is another reason that it is important to have this foreign perspec- tive: foreign media often have a different set of incentives and a different kind of accountability. The kinds of "biases" that exist in one country may also be different from those of another. They may have their own biases, but at least they are different, which is what leads to a diversity of perspectives. Another example that illustrates this is the New York Stock Exchange scandal involving then-chairman of the Exchange, Dick Grasso, who received an extraordinarily large pay package approved by the Exchange's Board of Directors. The New York Stock Exchange provides "self- regulation" of its members; there is an obvious conflict of interest in the "regulated" setting the compensation for the "regulator." Perceived inadequacies in certain aspects of this self-regulation have heightened concerns about these potential conflicts of interest.6 When news of the magnitude of Grasso's compensation broke, it became a scandal, and he was forced to resign. The interesting thing about the episode was that the story was evidently broken by the Wall Street Journal about six months before the Financial Times began its series. However, in the eyes of the Wall Street Journal the compen- sation was perfectly normal, or at least not scandalous. Perhaps the editors did not want to offend their "customers," many of whom are people receiv- ing similarly generous compensation packages, and some of whom may have engaged in practices that might, in appearances at least, look similarly scandalous. It took an outsider to point out that it was not acceptable to have the compensation of a self-regulator be decided by people who are engag- ing in practices that clearly ought to be circumscribed. The Financial Times began an attack on what was going on at the New York Stock Exchange; as is so often the case in such scandals, every day new aspects were reported. Fostering an Independent Media with a Diversity of Views 149 Grasso and the Board of the New York Stock Exchange may have hoped that the story would go away, but it did not. Eventually, it is safe to say that the Financial Times brought about Grasso's resignation. A third example where foreign media played an important role in cover- ing an American story is the scandal involving Harvard University and Andre Shleifer. The Harvard Institute for International Development was given a large contract by the United States Agency for International Development to give advice to Russia about privatization. The principals in the contract were Shleifer and a colleague. Whatever can be said about the quality of their advice, what is clear is that they used the opportunity to advance their own financial interests. Harvard eventually had to pay US$28 million, and Shleifer and his colleague had to pay the U.S. government between US$2 million and US$3 million. The story is described at great length in the Institutional Investor (McClintock 2006).7 However, what was reported in the Institutional Investor is not the full story; it is much worse than that. This is another model case of cor- ruption, with enormous ramifications within Russia, but my point here is to call attention to the nature of the media coverage. The coverage in the Financial TimeswasagainmuchstrongerthaninAmericannewspapers,perhapsbecause the incentive structures are different. Perhaps the American newspapers felt uneasy about rattling an establishment institution in this way.8 Perhaps it was because the Financial Times was more sensitive to the global consequences of this instance of American corruption and how it affected perceptions of the nature of Russia's transition from communism to a market economy. It is important not only that there be a diversity of media but also that media have an incentive to provide information that is relevant for public decision making. The private incentives in gathering and providing content may not be in accord with social returns. We may get more advertising rev- enues by having Mickey Mouse on the air 24 hours a day than by having any news programming, and the private media will therefore have 24 hours of Mickey Mouse programming. This is an example where there is a large deviation between private and social returns; one way of dealing with this would be to require all televi- sion and radio stations to provide programming about matters of public concern, such as political debates. This is a legitimate restriction; after all, the spectrum is owned by the public and so the public has the right to allo- cate it in ways that advance the public good. People will disagree about the best way of doing so, but because the spectrum is a public resource, there is an obligation to ensure that this resource is used in ways that are consonant with the public interest.9 150 Joseph Stiglitz There is also a role for public media; the obsession with privately owned media is misplaced. Public media can have different incentives and, there- fore, different perspectives. There are many examples that demonstrate that one can have high quality, publicly controlled media that are relatively politically independent in well-functioning democracies. Obviously, the institutional design is important. Not every country has succeeded in hav- ing this, but some have; the BBC is an excellent example. In thinking about how to get more media diversity, one ought to think not only about public ownership but also about a variety of ways of sub- sidizing nongovernmental organizations to run community-based media and other kinds of media. For instance, the costs associated with radio may be very low, which suggests that one could provide support for a large array of perspectives at relatively low costs. This is why a lot of countries are now expressing interest in public radio. Some governments, concerned about the control of the media by a small group of people, are supporting community radio stations, and this is a positive development. The rationale for the role of government in media is clear. Information is a public good, and as a public good it ought to be publicly supported. The public good is also a public good, as are democratic participation and citi- zenship. Without public support, there would be an undersupply of these public goods. Monitoring and maintaining competition and pluralism will be diffi- cult, but a variety of instruments are available through which we can attain these objectives, including antitrust laws, government subsidies, spectrum subsidies, state-owned radio and television, and requirements for public interest programming. It is important, although difficult, to monitor diver- sity; examples of ways to do so include measuring the diversity of coverage of key issues, constructing a diversity index, and comparing media across locales and over time. In conclusion, good information is essential to the effective function- ing of a market economy and economic processes. Without good informa- tion, prices will not provide appropriate information to guide resource allocations. My own work has explained why there are incentives to provide distorted information. By the same token, good information is essential for the functioning of democratic political processes. Without good information, citizens cannot participate meaningfully. Their votes would not guide resource allocations and other decisions in appropriate ways, and there will be an ineffective system of checks and balances. Fostering an Independent Media with a Diversity of Views 151 The implication is that we need a diverse and competitive media present- ing a variety of viewpoints. We need effective and strong laws and regula- tions ensuring the right to know and the right to tell. But we need more than that: we need a diverse media as a complement. Developing and maintain- ing pluralism in the media will not be easy because there are strong politi- cal and economic forces pushing the media in the other direction. There are people who do not want the kind of open and democratic society that a diverse media helps support; but through a variety of mechanisms that I have outlined in this chapter, I believe we can achieve more competition and more pluralism. Doing so will create more efficient economies. Even more importantly, doing so will strengthen democracy around the world. Notes 1. Evidence that television can influence beliefs has been around since Iyengar and Kinder's News That Matters: Television and American Opinion. While it is dif- ficult to prove that public opinion, as mediated by television, drove countries' decisions to either participate in or oppose the war in Iraq, it is certainly the case that such differences of opinions existed. The most obvious example per- tains to the question of whether Saddam Hussein possessed weapons of mass destruction or aided Al Qaeda prior to the invasion. Even as late as the 2004 presidential elections in America, 75 percent of Bush supporters believed that at least one of these was true (Wills 2004). 2. See, in particular, Sah and Stiglitz (1985, 1986, 1988a, 1988b, 1991). 3. The full text of the act is available: http://www.sarbanes-oxley.com/section .php?level=1&pub_id=Sarbanes-Oxley. 4. The transfer could, of course, have just been a coincidence. The worries about loss of access are, however, real. 5. The Internet, for the first time, has increased access to a diversity of viewpoints enormously. However, it is not a substitute for access to television, radio, and print. 6. In addition, some members of the board may have thought the extraordinary compensation was "normal," because compensation was perhaps not totally out of line with (admittedly widely criticized) practices of executive compen- sation in their own and other companies. 7. In addition, in litigating the case Harvard faced huge legal bills, estimated in the millions of dollars. Harvard's defense--that they were under no obligation not to have had a conflict of interest--hardly set the moral tone for what one would have expected of one of America's leading institutions. 8. This is especially so because of the strong links between Shleifer and Harvard's president, who in turn had close links with other "establishment" figures. 9. By the same token, most cable stations make use, in laying their cables, of pub- lic rights of way; and it is a perfectly legitimate demand of the state to request, in return, public interest programming. 152 Joseph Stiglitz References Freeland, Chrystia. 2000. Sale of the Century: Russia's Wild Ride from Communism to Capitalism. New York: Crown. Iyengar, Shanto, and Donald R. Kinder. 1989. News That Matters: Television and American Opinion. Chicago: University of Chicago Press. McClintick, David. 2006. "How Harvard Lost Russia." Institutional Investor, January 13. Sah, R., and J. E. Stiglitz. 1985. "Human Fallibility and Economic Organization." American Economic Review 75(2): 292­96. ------. 1986. "The Architecture of Economic Systems: Hierarchies and Polyarchies." American Economic Review 76(4): 716­27. ------. 1988a. "Committees, Hierarchies, and Polyarchies." The Economic Journal 98(391): 451­70. ------. 1988b. "Qualitative Properties of Profit-Maximizing k-out-of-n Systems Sub- ject to Two Kinds of Failures." IEEE Transactions on Reliability 37(5): 515­20. ------. 1991. "The Quality of Managers in Centralized Versus Decentralized Orga- nizations." Quarterly Journal of Economics 106(1): 289­25. Stiglitz, Joseph. 2003. Roaring Nineties. New York: W.W. Norton. Wills, Gary. 2004. "The Day the Enlightenment Went Out." New York Times, November 24, A25. 10 Media Regulation in the United States Jonathan Levy Introduction T he media in general, and television in particular, play an increas- ingly important role in virtually every nation in the world. Although the entertainment provided is significant, it is clear that the media also provide valuable information that strengthens the social and political fabric of the country and assists residents in discharging the obligations of citizenship. This chapter presents a brief overview of media regulation in the United States, with a primary emphasis on broadcast television. The purpose is not to advocate a mechanical copy of the U.S. model to any other country.1 Rather, it is to highlight some of the issues and questions that arise in designing a media regulatory structure. Indeed, some of these issues, particularly in the area of media ownership, are currently under reexamination. Media regulation in the United States is relatively limited. The Federal Communications Commission (FCC) was created by an Act of Congress, the Communications Act of 1934.2 That statute states the basic policy goal of the agency as "to make available, so far as possible, to all the people of the United States a rapid, efficient, nation-wide and world-wide wire and radio communications service." The title of the Act that deals with broadcasting also includes the instruction to "encourage the larger and more effective use of radio." In both cases, the term "radio" must be understood as encom- passing television and other spectrum-based distribution systems as well. Any opinions expressed herein are those of the author and do not represent those of the Federal Communications Commission or any other members of its staff. 153 154 Jonathan Levy Additionally, the section of the Act dealing with broadcast licensing directs the commission to determine with respect to each license application com- ing before it "whether the public interest, convenience, and necessity will be served by the granting of such application." Moreover, in considering radio license applications, the FCC must "make such distribution of licenses, fre- quencies, hours of operation, and of power among the several states and communities as to provide a fair, efficient, and equitable distribution of radio service to each of the same." Three Policy Objectives These general injunctions have been translated, with respect to television, into the following three basic policy objectives: competition, diversity, and localism. These objectives may, in some instances, be in conflict with one another. The most recent statements of these objectives by the FCC are in the 2002 Biennial Regulatory Review, Report and Order and Notice of Pro- posed Rulemaking in MB Docket 02-277 (Biennial Review Order) and the Broadcast Localism Notice of Inquiry in MB Docket No. 04-233 (Broadcast Localism NOI).3 Competition "Competition" refers to economic competition. In the case of commer- cial broadcast television, which in the United States is solely advertiser- supported, viewers do not pay a price for programming. Rather, advertis- ers pay broadcasters for access to viewers. For this reason, past analyses of competition in broadcasting focused on the advertising market. Although the 2002 Biennial Review Order did not completely reject this approach, it did shift the focus to the end user. The idea was to assess competition and innovation in the provision of programming to viewers. Diversity "Diversity" is a term that has several possible meanings. The 2002 Biennial Review Order enumerated five. One is diversity of program choice. View- ers benefit from a wide and varied choice of programming and informa- tion alternatives, one that matches as closely as possible the varied tastes of the audience. This concept has some similarity to competition in the provision of programming, to the extent that providers competing for the patronage of viewers are led to provide a wide and attractive menu of programming and to seek innovative program types that cater to viewer preferences. Media Regulation in the United States 155 The term also connotes diversity of viewpoint. This concept of diversity is relevant to free political expression and the value to the public of being able to access a wide range of political perspectives. Outlet diversity refers to the range of independent owners of media outlets in a relevant media market. One virtue of this concept is that it is easily measured. Also, it is thought that the degree of outlet diversity influences the degree of program and viewpoint diversity, making this a prime candidate for regulation. The Biennial Review Order also discussed source diversity, "the avail- ability of media content from a variety of content producers," and minority and female ownership diversity. In this Order, the Commission empha- sized the importance of program and viewpoint diversity, chose to main- tain regulation of outlet diversity, decided not to consider source diversity as a separate goal, and deferred consideration of minority and female own- ership diversity to a later proceeding. The current 2006 quadrennial regula- tory review explicitly addresses minority and female ownership diversity.4 The FCC will be free in the current proceeding to revise its previous findings with regard to the other categories of diversity. Localism In its Broadcast Localism NOI, the FCC defined localism as the trans- mission by stations of "programming that is responsive to the needs and interests of their communities of license." Thus, localism means program- ming tailored to the needs and tastes of a specific local geographical com- munity. This could be coverage of local as well as national political debate and elections. It could be news and information about local sports teams, weather, traffic, politics, or crime. Localism also comes into play when the manager of a television station selects the programming from outside sup- pliers, whether it is a national network or another source, to exhibit on the local station. In the United States, broadcasting has the connotation of free television--viewers make no direct payment for the service--and this gives rise to another implicit policy goal. As noted, the Communications Act directs the FCC "to make available, so far as possible, to all the people of the United States a rapid, efficient, nation-wide and world-wide wire and radio communications service," and to "provide a fair, efficient, and equi- table distribution of radio service" across states and communities. This is a kind of "universal service" goal for broadcasting, analogous in some respects to the universal service goal for access to basic telecommunications service. Two other subsidiary goals are worth noting; they will be discussed in more 156 Jonathan Levy detail later under the heading of content regulation. The U.S. law specifically provides for certain access to the broadcast media by political candidates, and it provides certain protections for children from unsuitable content. Before turning to the discussion of media regulation, it will be use- ful to provide a brief sketch of media markets in the United States.5 The United States has about 110 million television households and over 1,700 full power television stations. Roughly 1,300 are commercial; 400 are non- commercial educational stations. The larger population centers have more television stations assigned to them than the small markets do. While the average television household has access to more than 17 stations off-air, viewers in the largest markets can receive over 20 stations off-air. Television stations produce some of their own programming, but they acquire much of their material from outside suppliers--networks and syndicators. There are four major commercial networks--ABC, CBS, Fox, and NBC--and several smaller networks, for example, CW, MyNetwork TV, and two major Spanish language networks, Univision and Telemundo. There is also a noncommer- cial network, the Public Broadcasting Service. Roughly 86 percent of U.S. television households subscribe to a multi- channel pay television service. Cable television is the leader in this area, with 65 million television households subscribing; direct broadcast satel- lite service is second with 26.1 million subscribers. In considering diversity both in terms of entertainment and in terms of political expression, it will be of interest to know something about other media; there are over 13,000 terrestrial local radio stations in the United States, of which more than 2,000 are noncommercial. A substantial amount of information and enter- tainment content, including audio and video, is available via the Internet. A 2006 survey by the Pew Internet and American Life Project indicates that 42 percent of Americans have a broadband connection at home, and fully 73 percent are Internet users (Pew 2006). The remainder of this chapter consists of four additional sections. Sec- tion two addresses the three major broadcast media regulatory functions-- technical, content, and structural (ownership). Section three suggests that it is difficult to assess the impact of media regulation. Section four provides a status report on the FCC's quadrennial regulatory review, including a dis- cussion of some important methodological questions that come up in any attempt to evaluate broadcast ownership regulations. Section five briefly addresses other matters, such as the structure of the FCC, an overview of regulation of nonbroadcast electronic media in the United States, and the possible impact of technological change on media policy. Media Regulation in the United States 157 Broadcast Media Regulatory Functions Broadcast media regulation in the United States is conveniently catego- rized under three headings--technical, content, and structural. These are described in turn below. Technical Regulation For the purposes of this discussion, technical regulation encompasses spectrum management, transmission standards, and licensing. The FCC isresponsibleformanagingallradiofrequencyspectrumintheUnitedStates, except that allocated to the federal government or the national military. The FCC has assigned certain portions of the radio frequency spectrum to terrestrial television service. It has created a table of allotments for televi- sion, which lists all of the channels available by channel number and per- missible transmitter location. The rules include limits on antenna height and power level that, together with the specified transmitter location, determine the geographic area that a station's signal may reach. Interference among stations on the same, adjacent, or other channels is limited by the spacing of station assignments. The effective service area of a station may be limited at the edges by interference from another station. Stations are required to place a signal of a specified strength over their community of license, but their effective service areas generally extend well beyond that community. As dis- cussed briefly, cable television systems are generally required to retransmit local television stations, and this has an important effect on their actual service areas, especially for ultra high frequency stations. The FCC has established transmission standards for radio and televi- sion services. A detailed discussion is beyond the scope of this chapter, but it is worth noting that the transition from analog to digital television is now underway in the United States. The FCC has adopted a digital television transmission standard known as the Advanced Television Systems Commit- tee, and it has created a digital television table of allotments, parallel to the analog table, so every analog television licensee has available a paired digital allotment to use alongside the analog channel during the transition period. At the end of that period (which Congress has set at February 17, 2009), each television licensee will return one of its allotments to the FCC.6 Each television station is assigned to a particular community of license. This assignment methodology is based on the statutory instruction that the FCC assign licenses equitably across states and communities. It would have been technically possible to create a table of allotments with a smaller 158 Jonathan Levy total number of stations operating at higher power levels with larger ser- vice areas. In this scenario, each region of the country would have been served by a larger number of stations than is available in many markets under the current plan. But, pursuant to its interpretation of the statute, the FCC chose to ensure that more cities and towns had their own local stations. Turning to the administrative aspects of television licensing, through- out most of its history, the FCC awarded broadcast licenses via comparative hearings, what many have referred to as "beauty contests." Since 1997, pur- suant to congressional instructions, all new commercial television licenses are awarded via auction.7 Applicants may apply for a vacant channel in the table of allotments or petition the FCC to amend the table. If only one applicant expresses interest in an allotment, then, provided that the appli- cant meets some basic eligibility criteria, it receives the license for free. If more than one eligible applicant expresses interest, then the license is auc- tioned.8 The license term for television is eight years and there is an expec- tancy of renewal. The basic eligibility criteria for holding a television license are as follows: · First, the licensee must not be in violation of FCC media ownership rules, including the statutory limit on foreign ownership of broadcast licenses (no more than 20 percent direct or 25 percent indirect ownership). · Second, the licensee must adhere to the relevant technical rules regard- ing signal transmission. · Third, applicants for a new license must certify that they meet certain minimum financial requirements. Television licenses are freely transferable, provided that the transferee meets the basic eligibility criteria. Licenses are renewed if the licensee has met three statutory criteria: it has served the public interest, convenience, and necessity; it has committed no serious violations of the Communica- tions Act or the rules of the FCC; and it has not violated the Communica- tions Act or the rules of the FCC in a way that constitutes a "pattern of abuse."9 Members of the public may participate in the renewal process by filing a petition to deny the renewal. Only if the FCC finds that a station is not eligible for license renewal is the agency permitted to consider new applications for the license. The FCC may issue a forfeiture against a station (that is, require a licensee to make a payment to the FCC) for violation of FCC rules. In Media Regulation in the United States 159 extreme and very rare circumstances, the FCC may revoke a station's license. Content Regulation Media content regulation in the United States in general and by the FCC in particular is limited. The First Amendment to the Constitution of the United States guarantees freedom of speech and of the press and this provides sig- nificant protection to television broadcasters, although not all speech is protected by the First Amendment.10 With regard to broadcast television, the major statutory provisions regarding content address indecent program- ming, children's educational programming, and access to advertising time by political candidates. As a threshold matter, federal law prohibits the utterance of "any obscene, indecent or profane language by means of radio communica- tion."11 Federal prohibitions on obscenity are enforced by the U.S. Depart- ment of Justice, not by the FCC. In order to protect children, the FCC has regulations, authorized by statute,12 limiting the broadcast of "indecent" speech. Indecent speech is "language or material that, in context, depicts or describes, in terms patently offensive as measured by contemporary com- munity standards for the broadcast medium, sexual or excretory organs or activities." Indecent speech is constitutionally protected and cannot be prohibited, but it is subject to certain time, place, and manner restrictions. Specifically, indecent speech may not be broadcast between the hours of 6:00 a.m. and 10:00 p.m., hours when the FCC has determined that there is a reasonable risk that children may be in the audience.13 Congress has asked the FCC to consider whether it can constitutionally regulate vio- lent programming on broadcast and cable television. This issue is being debated. Pursuant to the Children's Television Act of 1990,14 the FCC requires commercial television licensees to serve the educational and informational needs of children 16 years of age and under, by providing programming specifically designed to meet those needs.15 While there is no single pre- scribed way of meeting the requirement, the FCC has adopted a processing guideline indicating that three hours per week of "core" children's educa- tional programming will fulfill the requirement. Advertising on children's television programming is also regulated.16 The limits are no more than 12 minutes of commercial matter per hour on weekdays and no more than 10.5 minutes per hour on weekends. This 160 Jonathan Levy provision applies to broadcast and to cable programming for children. In this context "children's programming" means programming aimed at chil- dren 12 years of age and under. Other limitations on children's advertising include prohibitions on program-length commercials and on host selling. A program-length commercial is a program that includes advertisements for a product associated with the program. "Host selling" refers to the use of program talent, such as the host of a children's show, to deliver commercial messages. The Telecommunications Act of 1996 included a provision that would have required the FCC to develop a video program rating system if the industry failed to do so voluntarily. The industry did develop such a system, and the FCC approved it in 1998. The system includes six categories ranging from "TV-Y--All Children" to "TV-MA--Mature Audiences Only." The ratings consider the presence of violence, coarse language, sexual situations or activity, and suggestive dialogue. However, not all networks use the ratings and many if not most shows are not rated. New television receivers with screens greater than 13 inches are required to include the "V-Chip" circuitry that recognizes the rating information and allows pro- gramming to be blocked based on the rating. Political candidates are also protected by U.S. communications law and regulations. In general, television stations are not required to accept adver- tising from any particular source. However, legally qualified candidates for federal office--presidency, vice presidency, U.S. Senate, and U.S. House of Representatives--are entitled to purchase or be given a reasonable amount of time on all commercial broadcast outlets. Candidates for federal, state, and local office are entitled to "equal opportunities" in the use of broadcast stations; if a station provides or sells time to one candidate for an office, it must provide or sell time to all. Additionally, all candidates entitled to purchase advertising time must be offered it at the station's "lowest unit charge" during the 45 days before a primary election and the 60 days before a general election. Broadcast licensees have a general obligation to provide programs in the public interest. To document how they meet this obligation, television licensees must prepare and put in a public file a quarterly issues/programs list. The list contains a statement of the issues of importance to its com- munity of license that a station has identified, together with a list of the programming transmitted to address those issues.17 The legal and analytic factors that come into play in content regulation are well illuminated by an examination of a piece of content regulation that the FCC repealed in 1987--the so-called "Fairness Doctrine."18 In the Media Regulation in the United States 161 Order repealing the doctrine, the FCC described it as placing a two-part obligation upon broadcast licensees. First, broadcasters have an affirmative obligation to cover vitally important controversial issues of interest in their communities. Second, they are obligated to provide a reasonable opportu- nity for the presentation of contrasting viewpoints on those controversial issues of public importance that are covered.19 In repealing the fairness doctrine, the FCC found that it "chills speech and is not narrowly tailored to achieve a substantial government interest" and that "the Constitution bars us [the Commission] from enforcing the Fairness Doctrine."20 The Fairness Doctrine, while it was in effect, made stations subject to claims from interested parties that a station had dis- cussed a controversial issue of public importance without sufficient cover- age of contrasting viewpoints. If the station disagreed, it was vulnerable to a complaint to the FCC, with potential significant costs of defending the station's decision before the agency. Alternatively, the station could accede to the complaint and transmit additional material on the issue, in place of some content that the station otherwise would have considered more important. Although the fairness doctrine did not mandate access to a sta- tion's air time by any outside party, if the station was unwilling or unable to provide the contrasting viewpoints itself, it would need to seek an outside party to do so. If no outside party was available that could pay for the neces- sary air time, then the station could be in the position of having to provide free air time. The FCC's analysis suggested that, in order to avoid potentially costly and time consuming disputes of this nature, stations would simply shy away from coverage of certain controversial issues.21 Structural Regulation The United States has relied heavily on structural regulation, that is, media ownership rules, to pursue its media regulatory goals. This is in part due to the Constitutional limits on media content regulation mentioned above. This subsection describes the current FCC broadcast ownership limits. There are four local market rules and two national rules.22 The congressio- nally mandated review of these rules that took place in 2002 and the one now underway are discussed below. The range of media outlets actually available to a viewer/listener is contained in a local geographic area, and FCC rules limit the number of television stations and the number of radio stations one entity can own in a single market. There are also cross-ownership limits involving radio/ television and newspaper/broadcast combinations. 162 Jonathan Levy No company can own more than two television stations in the same market. A company may acquire a second station in a market, provided that there are at least eight independently-owned local television stations post- merger and that both of the merged stations are not ranked among the top four in the market in audience share. There is also a sliding scale limit on the number of radio stations per market that a company may own. The 1996 Telecommunications Act established the rules, permitting ownership of up to eight stations in markets with 45 or more total stations, and correspond- ingly fewer in markets with fewer stations. In the smallest size class (fewer than 15 stations), a firm is permitted to own up to five stations, provided that no one owns more than one half of the stations in the market. There are also limits on common ownership of radio and television sta- tions in the same market. A company may own one commercial television station and up to seven radio stations in a market, provided that at least 20 independent media voices would remain post-merger. The limit is six radio stations if the company owns two television stations in the market. If at least 10 independent media voices would remain post-merger, a company may own up to four radio stations in the market. Regardless of the number of independent media voices in the market, a company owning television stations may also own one radio station. In each case these combinations are permitted only provided they do not violate the separate local television and local radio ownership rules. In any market, a company could own one radio and one television station regardless of the number of post-merger media voices. Total media voices in a market include local television and radio stations, daily newspapers, and cable television (a cable television system counts as one "voice").23 Additionally, common ownership of a tele- vision (or radio) station and a daily newspaper of general circulation in the same area is prohibited. There are some newspaper/broadcast combinations in place that pre-date the rule and were "grandfathered" in, and there are a limited number of combinations for which the FCC has waived the rule. There are two national broadcast ownership rules. First, no firm may own television stations reaching more than 39 percent of U.S. television households. This limit, which was set by Congress in 2004, revised a deci- sion made in the FCC's 2002 Biennial Review Order.24 The Commission also has a "dual network rule," which prohibits a merger among the top four commercial television networks (ABC, CBS, Fox, and NBC). The Communications Act limits foreign ownership of U.S. radio and television licenses. The ceiling is 20 percent for direct ownership and 25 percent for indirect (for example, via a holding company) ownership.25 Media Regulation in the United States 163 Challenges of Assessing the Impact of Regulation The FCC's broadcast regulations focus primarily on the welfare of citizens/ consumers, and what the citizens/consumers actually view or listen to is content. Unfortunately, it is difficult to assess the value of content. From a legal point of view, the government has very little discretion to make judgments based on content. From an economic point of view, broadcast programming is a public good with no market price paid by viewers. This makes it difficult to place a value on broadcast programming, whether it is the offerings of a single outlet or the full range available to viewers in a par- ticular market. The situation is somewhat better with regard to subscription television, since citizen/consumers reveal their willingness to pay for the service. However, because subscription services in the United States, such as cable and direct broadcast satellite (DBS), generally sell bundles of chan- nels, it is often not possible to determine the value of particular channels. With regard to diversity, the situation may be even more complex. First, as explained, there is more than one concept of diversity that is potentially of interest, for example, program diversity and viewpoint diversity. It is possible that a change in the structure of a local media market may have different effects on different types of diversity. Increased concentration may actually foster diversity of program choice while at the same time constraining diversity of viewpoint, thus requiring a balancing of these considerations.26 Beyond this, to the extent that viewpoint diversity is a goal because it enhances good citizenship, two additional issues come into play. First, there is an "externality" here. The awareness of diverse viewpoints that a citizen gains from the media makes her a better citizen, who votes in a more informed way, for example. This benefits not only this citizen herself, but also the other citizens of the same jurisdiction. Since citizens cannot cap- ture the benefits they confer on others, they may "under-consume" infor- mation on diverse political viewpoints. Moreover, it is difficult to measure the actual effect on citizenship behavior associated with the consumption or availability of diverse viewpoints. Some interesting work has been done relating political participation (voting) in local elections to the increased availability of national newspapers in an area (George and Waldfogel 2007), and there is some empirical work available relating political partici- pation to the availability or usage of broadcast media.27 Broadcast ownership regulation is, in a sense, "instrumental" in nature; it is regulation of market structure in order to influence indirectly the range 164 Jonathan Levy of content provided. Calibrating changes in the range of content to specific changes in market structure is a challenge.28 The next section addresses the FCC's recent attempts to grapple with this challenge and describes some of the important methodological questions raised by the attempts. Media Ownership Regulation: The FCC's Quadrennial Review The 1996 Telecommunications Act required the FCC to review its broadcast ownership regulations every two years. The FCC's 2002 Biennial Review Orderincludedchangesinfiveofthesixbroadcastownershiprulesdescribed above, leaving the dual network rule intact. The Order was appealed to the U.S. Court of Appeals for the Third Circuit. While it was under appeal, Con- gress passed a law setting the national television station ownership cap at 39 percent of U.S. television households (down from the 45 percent speci- fied in the 2002 Biennial Review Order).29 The legislation also changed the frequency of the periodic review from biennial to quadrennial. The Order's changes in the four local rules were stayed and remanded by the Court to the FCC, thus leaving in effect the rules as they stood prior to the Biennial Review Order.30 Those rules are the subject of the 2006 quadrennial review now underway at the FCC. The court considered the FCC's actions with regard to four rules: the local television ownership limit, the local radio ownership limit, the local limit on common ownership of television and radio stations, and the flat prohibition of common ownership of a broadcast station and daily newspa- per of general circulation in the same market. The court affirmed the FCC's authority to regulate media ownership, affirmed the decision to eliminate the flat prohibition on newspaper/broadcast cross-ownership in the same market, affirmed the prohibition of mergers among the four most highly- rated television stations in any market, affirmed the revision of the geo- graphic market definition procedure for local radio markets, and affirmed the inclusion of noncommercial radio stations in the counts of numbers of radio stations in a local market. The court's remand included directions to the FCC to justify numerical limits on the number of television or radio stations that one entity could own in a single market. In the case of the local television station rule, the court remanded the Order "for the FCC to support and harmonize its ratio- nale," and in the case of the local radio station rule, the court remanded Media Regulation in the United States 165 with an instruction to "develop numerical limits that are supported by a rational analysis."31 Thus, it appears that the court did not reject the use of numerical limits in local television and radio station ownership rules. The Biennial Review Order had replaced the television-radio and newspaper/broadcast cross-ownership rules with combined "cross-media limits," which were based in part on a diversity index, calculated by analogy with the Herfindahl-Hirschmann Index (HHI) used in antitrust analysis. That index was designed to reflect the relative importance of radio, broad- cast television, newspapers, and the Internet as sources of news and current affairs information. The court remanded the cross-media limits because "the Commission gave too much weight to the Internet as a media outlet, irrationally assigned outlets of the same media type equal market shares, and inconsistently derived the Cross-Media Limits from its Diversity Index results."32 The FCC recently released a decision in the 2006 quadrennial review proceeding.33 The decision amends the former ban on newspaper/ broadcast cross-ownership. The revised rule presumptively allows a news- paper to own one television station or one radio station in the 20 largest markets, subject to strict criteria and limitations. The decision makes no changes to the other broadcast ownership rules that were under review, namely, the local television "duopoly" rule, the local radio ownership rule, the local radio-television cross-ownership rule, and the dual network rule. Regardless of the particular rule or policy goal, when promulgating a media ownership rule, certain methodological questions arise. The fol- lowing discussion is intended to introduce some of these issues without prejudging the appropriate methodological choice. Although the FCC is not an antitrust agency, its ownership rules have frequently drawn on antitrust concepts and tools of analysis. The idea of a relevant product and geographic market carries over quite naturally to ana- lyzing approaches to achieving the FCC's competition goal. To the extent that viewpoint diversity is considered important, the FCC has utilized the con- cept of "competition in the marketplace of ideas" to help frame its analysis. In any of these situations, certain questions come up: how to define the product and geographic market; how to measure market shares; where to draw the line between permitted and prohibited combinations; whether to analyze combinations on a case-by-case basis or adopt "bright-line" rules; and whether, in the presence of multiple rules and multiple policy goals, to assign certain policy instruments (rules) specifically to certain policy goals. 166 Jonathan Levy Analysts have generally found it useful to have some summary measure of market structure to use in their work. To illustrate, the federal antitrust authorities in the United States have published Horizontal Merger Guidelines34 that employ the HHI, defined as the sum of squared market shares of all firms in the relevant market. If the index is below 1,000 (equivalent to 10 firms of equal size), the market is considered unconcentrated and mergers are unlikely to be challenged. If the index is above 1,800 (equivalent to six firms of equal size), the market is considered highly concentrated, and a merger is highly unlikely to be approved. If the index is between 1,000 and 1,800, the market is moder- ately concentrated and there is no presumption. The Merger Guidelines are, of course, much more detailed than this description of one part of them indicates.35 The point of the description is to indicate some methodological possibilities. First, the antitrust authori- ties enforce the law on a case-by-case basis. In principle, even a merger with an HHI over 1,800 could be approved, depending on the more detailed analysis performed by the authority. By contrast, the FCC has generally opted for bright-line rules, for example, a firm can acquire a second televi- sion station in a market only if there will be eight independent television voices in the market post-merger and the two stations to be combined are not both among the top four in the market in audience share.36 Second, possible summary measures of market structure include those based on market share, as well as a simple count of firms in the market, and there is a kind of correlation across those two measures. The FCC's current local market broadcast ownership rules, described above, are expressed in terms of the number of firms in the relevant market. The revisions proposed in the Biennial Review Order and rejected by the court were also expressed in terms of the number of firms in the relevant mar- ket. However, the Biennial Review Order did draw on an additional summary measure of market structure in its development of cross-media limits. Other Matters This section covers several topics that provide some additional context for understanding media regulation in the United States and related policy issues. It begins with a brief institutional discussion of the workings of the FCC, followed by a description of the noncommercial educational portion of the U.S. television sector. Next is a brief overview of the regulation of Media Regulation in the United States 167 other electronic media such as cable television and DBS service. Finally, this section considers the impact of technical change on media regulation. The Federal Communications Commission The FCC is an independent regulatory agency, created by the U.S. Congress via the Communications Act of 1934, as amended. The FCC has five com- missioners (including the chairman) who are appointed by the president for five-year terms, subject to confirmation by the U.S. Senate. No more than three commissioners may belong to the same political party. The FCC, like all independent federal regulatory agencies, operates under the Administrative Procedure Act (APA). The APA requires a "notice and comment" procedure for decision making. Thus, if the agency wishes to add, change, or eliminate a regulation, it must first issue a Notice of Proposed Rulemaking (NPRM) specifying the options under consider- ation. NPRMs are by nature filled with questions. The Commission for- mally makes its decisions by adopting a Report and Order (R&O). The R&O contains not only the decision but the reasoning and analysis supporting it, including discussion of arguments made by commenters and why the agency chose one approach over another. The FCC is also subject to the Government in the Sunshine Act, which requires it to make its major decisions in a public meeting and to give seven days' advance notice to the public of the matters to be considered. Inter- ested parties who disagree with an FCC decision can file a "petition for reconsideration." Ultimately, FCC decisions are subject to review by the federal courts. Noncommercial Educational Television and Radio The U.S. government has chosen to reserve a substantial portion of the spectrum allocated to television for noncommercial educational stations. There is a similar reservation of radio spectrum for noncommercial use. As mentioned, somewhat less than one-fourth of U.S. television stations are noncommercial. They are licensed to local educational institutions and other nonprofit entities. These stations cannot ignore audience preferences because they receive some of their financial support from viewer contribu- tions, but they are not subject to the same overall incentives that commer- cial stations face. The result is a different programming mix than is found on commercial stations and an additional source of program diversity. 168 Jonathan Levy Public television and radio receive limited federal funding, amounting to 19.1 percent of their total revenues in fiscal year 2005. The other major funding sources are individual subscribers (26.5 percent), state and local governments (15.9 percent), businesses (15.3 percent), colleges and univer- sities (10.3 percent), and foundations (6.8 percent).37 Overview of Multichannel Video Program Distributor Regulation This subsection provides a brief overview of regulation of the two major multichannel video program distributors (MVPDs)--cable television and DBS. Cable television regulation is shared between local franchise authori- ties and the FCC.38 Local authorities are entitled to require prospective cable operators to get a franchise, although federal law prohibits exclusive franchises and the unreasonable refusal to grant a competitive franchise. The local authority is permitted to charge a franchise fee of up to five per- cent of gross revenues from cable service; it is permitted to require the cable operator to set aside channels for public, educational, and governmental (PEG) access. The local authority may regulate the rate for the basic tier of cable service (the tier containing retransmitted broadcast signals, and per- haps others; all subscribers to cable must purchase this tier), provided that the cable system is not subject to "effective competition."39 Federal regulation of cable television includes two provisions govern- ing access of programmers to cable channels. Pursuant to explicit provi- sions of the Communications Act of 1934, FCC regulations require cable operators to retransmit all local analog television signals, subject to certain limitations.40 The signal carriage rules require that, every three years, each commercial television station must elect either must-carry or retransmis- sion consent status for their signals. Must-carry status means that, subject to the limitations referred to above, a cable operator is required to retrans- mit the station's signal without compensation and to provide it with cer- tain channel positioning rights. Retransmission consent means that a cable operator is prohibited from carrying the signal without the consent of the station licensee. Compensation may be paid by either party as part of a retransmission consent agreement.41 The United States is in the midst of a transition from analog to digital terrestrial television. Most stations are transmitting in both modes now. When the transition is completed on February 17, 2009 (a date set by Con- gress), the analog signals will be turned off, and must-carry rules will apply to the remaining digital signals. Very few stations currently transmit only digitally; their digital must-carry rights are in effect now. Media Regulation in the United States 169 The Communications Act also requires cable operators to set aside a fraction of their capacity for lease to unaffiliated programmers, the so- called "commercial leased access" provision. The setaside is on a sliding scale, with low capacity (fewer than 36 "activated channels") systems not required to provide commercial leased access channels. Systems with over 100 channels must set aside 15 percent of capacity for commercial leased access. If there is not sufficient demand for leased access channels to occupy the full setaside, the cable operator may use the remaining channels for its own purposes. The FCC also has statutorily mandated program access rules. These rules prohibit discriminatory practices by cable operators, broadcast satel- lite program vendors, and vertically integrated satellite cable programmers. A vertically integrated satellite cable programmer is a programming network that is owned by a cable operator and is distributed to outlets via satellite. Vertically integrated satellite cable programmers are currently prohibited from signing exclusive distribution contracts. The effect of the rules is to require cable operators who own program networks to make them available to rival distribution platforms (for example, DBS) on nondiscriminatory terms. The FCC is currently considering whether to extend the exclusive contract limit, which is scheduled to expire in October 2007.42 Pursuant to congressional instruction, the FCC has adopted a national cable television ownership cap. No cable operator may own an attributable interest in entities that serve more than 30 percent of U.S. MVPD subscribers. The cable rule was adopted out of concern that excessive concentration might reduce or impede the flow of programming to cable subscribers. This rule was appealed to a federal court, which reversed and remanded it. There is an FCC rule pending to address the court remand and promulgate a new rule.43 There are no limits on foreign ownership of U.S. cable television systems. Direct broadcast satellite service is subject to even less regulation than cable television service. There is no rate regulation, no ownership cap (although, of course, any proposed merger involving a DBS company is subject to case-by-case review), no program access regulations, no PEG channels, and no commercial leased access requirement. DBS carriers are required by FCC regulation to set aside four percent of their channel capac- ity for noncommercial informational and educational programming. DBS carriers are subject to the must-carry/retransmission consent regime, but the DBS signal carriage rules do not require retransmission of all local analog television signals.44 DBS carriers are subject to the requirement to carry all local analog signals in a market if they choose to carry one local analog signal pursuant to a compulsory copyright license. Additionally, 170 Jonathan Levy unlike cable, there is no requirement that DBS customers receive the local signals as part of a basic tier. The subscriber may choose to receive them but at an additional charge. Digital must-carry rules for DBS have not been finalized yet. The last point to be made is that few of the (limited) content regulation provisions that apply to broadcast television and radio apply to cable televi- sion or DBS. The equal time rule for political advertising (which provides that, if an outlet offers advertising time to one candidate in an election, it must offer time to all the other candidates on the same terms) does apply to cable, but only to "origination cablecasts."45 With respect to DBS, the equal time rule applies to "DBS providers" for "DBS origination programming"; the rule requiring reasonable access for legally qualified candidates for fed- eral office applies to "DBS providers."46 The Impact of Technical Change on Media Regulation Over the past 30 years or so, technical change has substantially expanded the range of video programming available to U.S. citizens and reduced the costs of production and distribution.47 Cable television, which began a rapid expansion in the mid-1970s, and DBS, launched in the mid-1990s, are now almost universally available in the United States, and over 85 percent of television households subscribe to one or the other. Digital compression technology has made it possible for these distribution platforms to offer hun- dreds of distinct channels. The transition from analog to digital terrestrial television has expanded the capacity of traditional television broadcasters, making it possible for them to deliver high definition images and/or multiple channels of "standard definition" video. The resulting fragmentation of the video audience is putting pressure on the advertiser support business model that has sustained commercial broadcast television. The development of the personal video recorder not only makes it possible for viewers to time shift their viewing but to skip advertisements entirely, which may also erode the traditional business model.48 Last but not least, the Internet offers access to a huge amount of online information--text, audio, and video. What do these developments mean for media regulation? As a general matter, competition among independent providers of media services is supe- rior to regulation as a means of achieving media policy goals, provided that sufficient competition is present. It is certainly true that the video market in particular and media markets in general have become more competitive over the past 30 years. However, the range of providers varies according to Media Regulation in the United States 171 the specific product or service under consideration. For example, it appears clear that cable television and DBS have increased the range of sources of national and international news available in many local U.S. markets. However, these distribution platforms have not had the same effect on the range of sources of local news. Moreover, the fact that these subscription media are available to all or almost all households may not be sufficient for policy makers to count them on an equal footing with broadcast media. As noted, in the U.S., roughly 14 percent of television households do not subscribe to an MVPD. The Internet is also a promising medium from both a competition and a diversity point of view, but the same caveats apply there.49 A massive amount of information and entertainment matter is available online. A substantial amount of the news content on the Internet appears to origi- nate with other media such as newspapers, television networks, and cable networks. While this undoubtedly expands the availability of national and international news and information available in a market, the impact on local news and information is less clear. Moreover, in order to access Inter- net content effectively, particularly video content, it is necessary to have a broadband connection. Broadband is widely available, but it is not yet ubiquitous; perhaps due to the price of broadband, a substantial majority of households do not have a home broadband connection. Thus, as U.S. media markets continue to become more competitive, it is quite possible that the case for media regulation will become weaker. How- ever, the market may not work perfectly to provide the appropriate menu of diverse programming and range of political viewpoints, so the need for media regulation is unlikely to disappear. It remains unclear the extent to which media such as the Internet will become sufficiently widely available and widely used to be considered strong substitutes for traditional media such as television. Notes 1. The statute setting up the Federal Communications Commission is the Com- munications Act of 1934, as amended; the Commission's rules are in the Code of Federal Regulations, Title 47. See: http://www.fcc.gov/Reports/1934new. pdf and http://www.access.gpo.gov/nara/cfr/cfr-table-search.html. 2. 47 U.S.C. §151 et seq. The quotations in this paragraph are from §§151,303(g), 309(a), and 307(b). 3. See Biennial Review Order, 18 FCC Rcd 13620 (2003) and Broadcast Localism NOI, 19 FCC Rcd 12425 (2004). The FCC is required by Congress to review 172 Jonathan Levy broadcast ownership rules at fixed intervals (previously every two years, now every four years). 4. See Further Notice of Proposed Rulemaking in MB Docket No. 06-121, 21 FCC Rcd 8834 (2006) (2006 Quadrennial Regulatory Review Further Notice). 5. The figures on number of television and radio stations in this subsection are from Federal Communications Commission, "Broadcast Station Totals as of June 30, 2005," http://hraunfoss.fcc.gov/edocs_public/attachmatch/ DOC-260747A1.doc. The figures for television households, cable, and DBS subscribers, also as of June 2005 are from the FCC's 2005 Video Competition Report. 12th Annual Report, 21 FCC Rcd. 2503 (2006), table B-1. The figures on number of stations available to U.S. television households are from Nielsen Media Research, "Television Audience 2006" (2007), at 12. 6. A very small number of stations are now transmitting only in digital mode. 7. See 47 U.S.C. §309(j). 8. Applications for noncommercial educational broadcast stations are not subject to competitive bidding procedures. See 47 C.F.R. §73.5000(b). 9. See 47 U.S.C. §309(k)(1). 10. For example, commercial speech, that is, advertising, is regulated by the Federal Trade Commission. The Federal Trade Commission Act prohibits "Unfair methods of competition in or affecting commerce, and unfair or deceptive acts or practices in or affecting commerce," and also prohibits false advertising under a wide variety of circumstances. See 15 U.S.C. sec.45(a)(1) and sec. 52. See http://straylight.law.cornell.edu/uscode/html/uscode15/usc_ sec_15_00000045----000-.html and http://caselaw.lp.findlaw.com/scripts/ ts_search.pl?title=15&sec=52. 11. See 18 U.S.C. §1464. 12. See Public Telecommunications Act of 1992, Pub. L. No. 102-356, 106 Stat. 949 (1992) [Section 16(a)]. This provision was upheld in Action for Children's Television v. FCC, 58 F.3d 654 (D.C. Cir. 1995). In 1978, the Supreme Court upheld the constitutionality of regulation of indecent broadcast programming in FCC v. Pacifica Foundation, 438 U.S. 726 (1978). 13. See Federal Communications Commission, Enforcement Bureau, "Obscenity, Indecency, and Profanity; Frequently Asked Questions." http://www.fcc.gov/ eb/oip/FAQ.html#TheLaw. 14. Pub. L. No. 101-437, 104 Stat. 996 (1990). 15. See 47 C.F.R. §73.671. 16. See 47 C.F.R. §§73.670 and 76.225. 17. See 47 C.F.R. §73.3526(e)(11)(i). 18. The repeal decision is in Memorandum Opinion and Order In re Syracuse Peace Council, 2 FCC Rcd 5043 (1987) ("Fairness Doctrine Repeal Order"). For a detailed analysis and history of the Fairness Doctrine, see Krattenmaker, Thomas G., and Lucas A. Powe, Jr., (1994, 61­65 and chapter 9). The FCC repealed two related rules (the personal attack and political editorial rules) in 2000. See Order in MM Docket 83-484. 15 FCC Rcd 20697 (2000). 19. Fairness Doctrine Repeal Order, ftn. 2. 20. Fairness Doctrine Repeal Order, paragraph 98. Media Regulation in the United States 173 21. See Inquiry into Section 73-1910 of the Commission's Rules and Regulations Concern- ing Alternatives to the General Fairness Doctrine Obligations of Broadcast Licensees in Gen. Docket No. 84-282, 102 FCC 2d 145 (1985). Hazlett and Sosa studied radio formats before and after the elimination of the fairness doctrine. They found an increase in the share of informational formats post-elimination and they interpret this as evidence that the fairness doctrine had, indeed, "chilled" presentation of news about important social issues. See Hazlett and Sosa (1997, 279­301). 22. See 47 C.F.R. §73.3555 for all of these except the dual network rule, which is found in 47 C.F.R. §73.658(g). 23. For the specifics, see 47 C.F.R. §73.3555(c)(3). 24. See Order in the Matter of Implementation of Section 629 of the Consolidated Appropriations Act, 2004 (released March 1, 2007). http://hraunfoss.fcc.gov/ edocs_public/attachmatch/FCC-06-117A1.doc. 25. See 47 U.S.C. §310(b). 26. The Biennial Review Order stated that common ownership of television sta- tions in the same market could "facilitate efficiencies and cost savings . . . and increase opportunities for counter-programming." Biennial Review Order at paragraph 147. However, excessive concentration of media ownership could adversely affect the opportunity for the expression of diverse viewpoints. 27. See Oberholzer-Gee and Waldfogel (2006). Additionally, a study by Strömberg found that in the 1930s, U.S. counties with greater radio penetration received greater distributions of funds from a New Deal emergency relief program. See Strömberg (2004). 28. Part of the challenge lies in the fact that citizen/viewer demand has an impact on media content in addition to the influence of "supply side" market struc- ture conditions that can be influenced by regulation. A recent study of "media slant" in newspapers found that the impact of consumer preferences on slant was significant and apparently of greater magnitude than the supply side factor of newspaper owner identity. See Gentzkow and Shapiro (2007). 29. See footnote 24, supra. 30. See Prometheus Radio Project v. Federal Communications Commission (U.S. Court of Appeals for the Third Circuit: June 24, 2004) ("Remand"). The court later gave the Commission permission to implement the new geographic market definition for the local radio rule. See Prometheus Radio Project v. FCC, No. 03- 3388 (3d Cir. Sept. 3, 2004) (granting petition for rehearing in part). 31. See Remand at 94, 123. 32. Remand at 58. The FCC's Further Notice of Proposed Rulemaking in the 2006 quadrennial review states that "we tentatively conclude that the DI [diversity index] is an inaccurate tool for measuring diversity." 2006 Quadrennial Regu- latory Review Further Notice at para. 32. 33. See Report and Order on Reconsideration in MB Docket No. 06-121 (released February 4, 2008) http://hraunfoss.fcc.gov/edocs_public/attachmatch/FCC- 07-216A1.doc. In connection with this proceeding, the FCC commissioned 10 special studies, three by FCC staff and seven by outside experts. They are available at http://www.fcc.gov/ownership/studies.html. 174 Jonathan Levy 34. U.S. Department of Justice and the Federal Trade Commission, "Horizontal Merger Guidelines" (issued April 2, 1992; revised April 8, 1997) http://www. usdoj.gov/atr/public/guidelines/horiz_book/hmg1.html. 35. For example, the Guidelines provide for analysis of the "potential adverse competitive effects of mergers," of entry prospects, and of efficiencies associ- ated with prospective mergers. 36. It must be noted that this rule and, indeed, all FCC rules, is subject to waiver, although the hurdle is often stringent. 37. See Corporation for Public Broadcasting, "Public Broadcasting Revenue Fiscal Year 2005," table 2. http://www.cpb.org/stations/reports/revenue/2005Public BroadcastingRevenue.pdf. 38. Statutory provisions for FCC cable regulation are found in Title VI of the Com- munications Act of 1934, as amended, 47 U.S.C. §521 et seq. 39. Effective competition essentially means that there is a rival MVPD offering service in the franchise area. Four specific criteria are used to determine if effective competition is present. See 47 U.S.C. §543(l)(1). 40. See 47 U.S.C. §§534 and 535, esp. 534(b) and 535(b). Cable operators are not required to devote more than one third of their "usable activated channels" to local commercial television stations, nor are they required to carry more than one affiliate of the same commercial broadcast network or to carry a station that substantially duplicates the signal of another local commercial television station carried by the cable system. Cable operators are also not required to carry stations that do not provide a good quality signal to the principal head- end of the cable system. 41. The retransmission consent provisions of the Communications Act are found in 47 U.S.C. §325(b)­(e). 42. See Notice of Proposed Rulemaking in MB Docket 07-29 (adopted February 7, 2007). Note that under current rules it is possible for a vertically integrated cable programmer to apply for a waiver of the exclusivity prohibition. 43. See Second Further Notice of Proposed Rulemaking in MM Docket 92-264. 20 FCC Rcd 9374 (2005). 44. For DBS signal carriage rules, see 47 C.F.R. §76.66. For the noncommercial educational and informational programming setaside, see 47 C.F.R. §25.701. 45. See 47 C.F.R. §76.205. "Origination cablecasting" is defined as programming "subject to the exclusive control of the cable operator," 47 C.F.R. §76.5(p). 46. See 47 C.F.R. §25.701. DBS providers are the entities with the license to trans- mit on satellite frequencies and DBS origination programming is program- ming "subject to the exclusive control of the DBS provider." 47. For a more detailed discussion of technical change in video delivery, see Levy and Levine (2008). 48. On the other hand, the personal video recorder could offer new advertising revenue opportunities by enabling advertisers to target their messages to view- ers who value those messages highly. 49. There is some preliminary evidence of substitution between television and the Internet for accessing news. See Waldfogel (2002). Media Regulation in the United States 175 References Gentzkow, Matthew, and Jesse Shapiro. 2007. "What Drives Media Slant? Evidence from U.S. Daily Newspapers." Working paper. http://faculty.chicagogsb.edu/ matthew.gentzkow/biasmeas081507.pdf. George, Lisa, and Joel Waldfogel. 2007. National Media and Local Political Partici- pation: The Case of the New York Times. Working paper. http://urban.hunter. cuny.edu/~lgeorge/Research/NYTVote.pdf. Hazlett, Thomas W., and David W. Sosa. 1997. "Was the Fairness Doctrine a `Chill- ing Effect'? Evidence from the Postderegulation Radio Market." Journal of Legal Studies 26(1): 279­301. Krattenmaker, Thomas G., and Lucas A. Powe, Jr. 1994. Regulating Broadcast Pro- gramming. Cambridge, MA: MIT Press. Levy, Jonathan, and Anne Levine. 2008. "Evoluzione strutturale e cambiamento del mercato televisivo statunitense nell'era digitale" ("The Evolving Structure and Changing Boundaries of the U.S. Television Market in the Digital Era"). In La nuova televisione: Economia, mercato, regole, ed. Antonio Nicita, Giovanni B. Ramello, and Francesco Silva. Bologna, Italy: Società editrice il Mulino. Oberholzer-Gee, Felix, and Joel Waldfogel. 2006. "Media Markets and Localism: Does Local News en Espanol Boost Hispanic Voter Turnout?" Working paper 12317, National Bureau of Economic Research, Cambridge, MA. Pew Internet and American Life Project. 2006. "Internet Penetration and Impact." http://www.pewinternet.org/pdfs/PIP_Internet_Impact.pdf. Strömberg, David. 2004. "Radio's Impact on Public Spending." Quarterly Journal of Economics 119(1): 189­221. Waldfogel, Joel. 2002. "Consumer Substitution Among Media." Media Ownership Working Group Study 3, Federal Communications Commission, Washing- ton, DC. 11 Aspects of Two Media Models: France and the United Kingdom and EU Media Governance Pierre-Yves Andrau Introduction F reedom of the press has a very special status, even within the hallowed ranks of fundamental rights and freedoms. The media are par excel- lence the places for exchanges of ideas, voices, and opinions. The media are also the places where criticism can be formulated against any establish- ment, and not only public authorities. There is no true democracy without the media. The media also play other potentially conflicting roles that must be reconciled. In effect, the media are also increasingly regarded as a substi- tute for traditional institutions--such as schools, family, religion, and the state--transmitting social values, including the integration of different citizens, such as minorities, into a common and homogeneous domain of public opinion, often the national one. Finally, the media are for the most part also for-profit businesses acting in a competitive environment gov- erned by trade and financial interests. This means that any media organism has to compete in the marketplace.1 These three functions--informative, educative, and commercial--confront and often contradict each other. Permitting the free exercise of these three functions, and providing a framework to reconcile them, are the objectives of media law. Hence, it is generally admitted in Western democracies that regulation and public action in the media sector are deemed to be the "reasonable" expressions This chapter expresses the personal views of the author and in no way con- stitutes a formal position of the European Commission. The author is especially grateful to Adam Watson Brown for his contribution. 177 178 Pierre-Yves Andrau of essential political, economic, and cultural choices made by a society at a certain point during its existence. In many countries, the fundamental rights commonly recognized are shaped by regulations aimed at softening the apparent competition between individual and collective freedoms. In democracies, regulations and public policy should ensure that media prac- tices and behavior support the public interest. In certain countries, the law regulates the media strongly. In others, the lack of state intervention targets the same outcome, but through means such as voluntary self-regulation. This chapter deals with media regulations and the general situation in France and the United Kingdom. Aspects of the U.K. situation will mainly supply elements of comparison and stress the divergence of two models intended to achieve similar outcomes. The development of the new infor- mation and communications technologies (ICT) and market liberalization further obliges one to take into account the European dimension and all European regulation of the media, either from the European Union (EU) or from the Council of Europe. The media are operating more and more across borders thanks to ICT and commercial pressure. In this respect, EU regula- tion and policies, in particular in the audiovisual field, play an increasing role in the formation of the national regulations of the EU member states. The Council of Europe and the European Court of Human Rights provide the European standards in terms of fundamental rights, which are often integrated in the case law of the Court of Justice of the European Com- munities. This continental European dimension is a crucial one. However, the discussion in this chapter is limited to the legal framework covered or governed by the EU that applies to the written press and audiovisual media; online media are not included in our reflections. The rules governing the media, as a specific branch of the law, have suffered and still suffer from the hybrid legal nature of the media, mixing several spheres of legal activity and transcending the traditional frontiers of the law. This is especially true in France and the United Kingdom. The chap- ter will only evoke nonsectoral and horizontal laws (business, competition, civil, criminal, tax, contractual, or social) when necessary to illustrate the exceptions applied to the media. The weight of media regulation in a country depends on the cultural con- ception of the state. This is one of the most important sources of the differ- ence between the French and British models. The regulatory framework and the self-regulation covering the media come from different conceptions of the role of the state. Certain specific instruments--legal or paralegal--applied Aspects of Two Media Models 179 to the media are implemented in France and the United Kingdom to reach the general objectives assigned to the media. For EU member states, the EU level has become the relevant place for common actions in order to respond to certain global challenges such as the development of ICT, new consumer behavior arising from ICT, or the preservation of national cultures. Two Different Views of the State In France and the United Kingdom, as in all European countries, the media were concentrated for a long time in the hands of a minority of influential interests that used it as an instrument for advancing their own economic or political interests. As a result, the media were regarded as the primary tool by critics of established authority of any kind--religious, economic, governmental--and they were considered as the symbol of democracy and freedom. From the period of the Enlightenment onward, the media were nothing less than the battering ram used by democratic forces against the Anciens Régimes around Europe. Naturally, the written press, and all other later media, were frequently the subject of repressive regulations designed to silence freedom of expression; very rarely, they were the subject of pro- tective regulations. In any case, the media have been for a long time at the mercy of the power of the state or economic forces. All the regulations accu- mulated through the ages--whether for more freedom or less freedom-- have influenced the present situation. Dealing with the freedom of the media and regulation implies first a need to tackle the role of the state in the democratic society. Regulation and policy should reflect the political conception of the freedom of the media and thus the media's role in society. The role attributed to the media in modern democracies is to enliven democratic debate. The link between the state and the media and the position of state authorities on the inde- pendence of media companies are not only matters of law but also have a significant effect on the economic structure of the media. The development of the media certainly marches hand in hand with the development of civil liberties, as well as with the progress of the media as industries. Scholars write of "media models" that characterize the relations, more or less nar- row, between media companies and the state. In France and the United Kingdom, the respective models reflect more than simply diverging concep- tions of the media. They are above all the result of substantive differences in the conception of the state, and notably of its weight and role in society. 180 Pierre-Yves Andrau The United Kingdom is the first state in the world that conquered the obstacles to freedom of the press and enjoyed the benefits of a free press for many years before others could catch up. The starting point of this new freedom was the repeal of the Licensing Act in 1695; it had imposed ex- ante authorization to create a newspaper, licensing in other words. In the U.K. model, the idea of the state that predominates is a minimalist con- ception pleading for the weakest state possible. The state lacks sufficient legitimacy--both in the strictly legal sense and in the view of society at large--to intervene in the business of the media. From the beginning, the media were considered as a quasi-institutional power shaping public opin- ion. Freedom and independence from the state constitute the fundamental bases of media activities and directly result from the limited role played by the state. The independence of the press is not provided or granted by the state; rather this independence predates the modern state. In the United Kingdom, the press is known as the Fourth Estate, coming after Lords Tem- poral, Lords Divine, and the Commons. The state is not regarded as the guarantee of the independence of the press, but rather as a potential threat to that independence. This independence depends on commercial freedom that in turn guarantees freedom of expression. This concept of the state is still a very strong instrument for supporting the freedom of the media, even if recent examples have shown that the reality is quite different. The French model differs dramatically in that the nature of the rela- tionship between the state and the media is much less clear. It was only with the law of 1881 on press freedom that the media became independent from political authority. This late advent of a liberal regime in France, compared with that in the United Kingdom, brought France into the modern age with a law that defined the administrative and contentious framework of the press; that law is still in force. Even if France created one of the most liberal regimes, the media became free only because the state partially refrained from intervening. However, despite this very liberal regime, the French media are in a situation that the British would find alien. The state enjoys great prestige in France: through its unceasing efforts in support of liberty, equality, and fraternity, it supports the peaceful social coexistence of the French people and resolves social conflicts. Therefore, the state is regarded as the natural protector of the freedom of the media. The legal status of French media companies is strongly based on missions of "general interest"--also rendered as "public interest" in English--that authorities impose. From this perspective, "information" as a raw material can be perceived in France as a kind of service of general interest. Undertakings devoted to information Aspects of Two Media Models 181 are like para-public entities; they are not submitted to normal market law. The influence of financial or economic interests on the media is considered in France to be more dangerous than the influence of the state. This view would be controversial in the United Kingdom. The French Paradox and British Self-Regulation French media regulations are based on the principle of the freedom of expression and the traditional liberal2 concept existing in most pluralistic regimes and Western democracies. Nevertheless, certain pieces of French legislation would be closer to those of a nonliberal system. For instance, it is accepted that free trade, business freedom, and contractual freedom do not alone guarantee freedom of the media and the rights of the public to the information. Freedom of expression is too often reduced to busi- ness freedom; this reduction is not completely satisfactory in France, where the theme of social injustice is very popular. Business freedom would be necessary but not sufficient. The liberal concept generally demands as the mandatory condition for freedom that the state abstains from interven- tion and allows free play for private initiative. It is commonly thought now in the world that this concept is the mandatory condition for freedom. In theory, this view would imply no specific law other than business law. In the French model, state forbearance, one of the characteristics of the lib- eral system, is clearly called into question. Justified by the need to give more reality to freedom of expression as proclaimed in the Constitutional text, state interventions are solicited. The broad support for the media granted by the state is based on the new theory of the "right to informa- tion." This apparent inconsistency between the liberal and interventionist schemes is a source of legal uncertainty for media undertakings, whether private or public. Nevertheless, this contradictory situation also helps to reconcile conflicting interests and satisfy certain principles of diversity and pluralism. In France, one of the objectives of media legislation is to ensure that pluralistic media are available for all citizens on an equal basis. This con- cern for a greater equality, based on the right to information, diverges fundamentally from the traditional liberal conception of freedom. In this rather hybrid system, the liberal regime is not considered sufficient to guar- antee freedom of expression. However, press freedom is not endangered in France. This kind of debate is possible in France because no one seri- ously disputes the fact that the press should be independent from the state, 182 Pierre-Yves Andrau and few dispute the role of commercial freedom.3 The objectives of media regulation and other public actions are in fact directed toward strengthen- ing freedom of expression via the right to information and not limiting it. Indeed, the system has the ambition to satisfy the real meanings and objec- tives of this freedom. These arguments justify an alternative legal approach intended to cre- ate better conditions for the exercise of freedom of expression. Actually, the French are not completely satisfied with the mere affirmation of the principle of freedom of expression, because this freedom has often proved to be only formal. Several elements of French media regulation refer to the theory known as the "right to information." Although this right is not expressly mentioned in any statute, the principle inspires numerous instru- ments. In moving from the "freedom of" toward the "right to," the general or public interest becomes the main objective and the essential criterion for assessment. Even in the media field, the French state is called to intervene so that it takes over or contributes to the financing of certain activities of services of general interest. Public service in this respect is one means to implement the right to information. The publishers themselves call for state intervention, arguing that only state aids maintain pluralism, protect their freedom, and sustain their existence. The liberal character of U.K. media regulation manifests itself most obviously through one key feature: delegation of sensitive issues of media governance to the media industry. However, this approach is often criti- cized. Various codes of conduct have evolved over the years, in particular, codes of conduct for journalists. For instance, the British Press Complaints Commission enforces the ethical codes developed by the British National Union of Journalists. Codes are mainly intended to ensure high ethical and technical standards across the sector, as well as guarantee freedom of the press and other media. Clashes between freedom of expression and the right to privacy are frequent. The Department of Culture, Media, and Sport shares competence for media with the Department of Trade and Industry. These ministries are restricted to supervision of self-regulation procedures carried out by the British Press Complaints Commission. The BBC or the Guardian newspa- per, for example, like other media, have their own internal systems of self- regulation, codes of conduct, and sometimes even their own ombudsmen. Public authorities try, and occasionally succeed, in influencing and manip- ulating the media. Many U.K. commentators fiercely denounce government "spin doctors" deployed to optimize the presentation of government busi- Aspects of Two Media Models 183 ness. It is clear that a system based on self-regulation and codes of conduct does not eliminate the weight that politics can impose on the media. Common law characterizes the whole approach. Statutes not only reflect English common law principles but are interpreted in the light of common law tradition. Many essential principles seem absent from the raw texts; but they are already understood from the point of view of preexist- ing case law and custom. Knowledge of precedents is essential. Given that media regulation is not a homogeneous field but a portmanteau containing rules from different sources, the role of common law becomes very clear in media regulation the more one studies it. Statutory guarantees for freedom of expression were enacted very recently as part of the Human Rights Act (1998); this transposed the European Convention of Human Rights into national law. In light of the increasing number of European rules, the Brit- ish legal system is moving more and more toward a statutory framework for media regulation and governance. Regulatory Framework: Some Examples In France, the very substantial state aid to the written media and the audio- visual public services can or should only be justified by the right of the pub- lic to information. A certain number of regulations are directly inspired by this right, including the obligation to open the audiovisual public services to a variety of beliefs or thoughts, the obligation to report on the campaign for national elections, the right to reply, the mandatory single retail price for books, and the cultural exception. The system comprises obligations and supports the written and audiovisual media. Primary Obligations As far as the written media are concerned, the French system is based on a specific status for publishing companies. This status gives advantages (mainly state aids) and requires the respect of certain obligations. The French law of 1986 that amended the law of 1881 contains three main obli- gations to be respected by publishing companies: transparency, indepen- dence, and pluralism. With respect to transparency, the publishers must provide the reader- ship with information on the staff, the name of the main owners, and the identity of the editors. In the same way, above a certain threshold publish- ers are required to publish all financial information in the columns of their publications, for instance, sales of share capital. 184 Pierre-Yves Andrau The independence of publishers is an essential feature of editorial freedom. In France, this concern is intended to secure independence from financial interests, whether domestic or foreign. A certain number of provi- sions of the 1986 law ensure this independence: prohibition on receiving money from foreign governments, limitation of the participation of for- eigners from countries outside the EU in the share capital, and the obliga- tion to make a clear distinction between editorial content and advertising or business communication. The political influence of state authorities on certain editorial content was once a reality in France even if today the inde- pendence of the written media is unquestioned. Yet, in the French mind, the dependence on the state is less threatening to press freedom than the risk posed by financial and private interests. With respect to pluralism, the provisions of Article 11 of the law of 1986 are more restrictive but concern nonspecialized daily newspapers only. Pluralism is treated from the perspective of concentration of ownership. Everything hinges on determining ownership thresholds; the law also takes into account all types of media in its elaboration of ownership rules. Yet media pluralism is a concept that embraces a number of other aspects, including diversity of ownership, but also variety in the sources of informa- tion and the range of contents available in different EU member states. For many analysts or observers, in France in particular, media pluralism has come to mean plurality of ownership. In any case, the system is less restric- tive than it appears. Indeed, the law is completely silent on concentration among publishing companies of any publications, even daily newspapers, on one side, and press agencies, advertising agencies, book publishers, or online publishers on the other side. In the same way, no restriction exists on the right for a journalist to work for rival publishers under multiple names. This is a quite frequent phenomenon in France and this could in fact be regarded as a threat to pluralism. In fact, the status of press publishers is not as restrictive or specific as it is presented and does not impose many more requirements than those imposed on other sectors of the economy. In reality, the relevance and the scope of the special press status has been reduced with the application of principles inspired by business freedom and neo-liberal economic theory. Despite this relative freedom granted to publishers and shareholders, their status does not prevent them from claiming state aids, which represent a crucial aspect of the media law in France. They constitute the positive form of public intervention in favor of freedom of expression and the concrete implementation of the right to information. Officially at least, such state aids must contribute to media Aspects of Two Media Models 185 pluralism and should ensure the survival of certain publications. In reality, the conditions for granting the aid, the determination of the beneficiaries and the lack of ex-post monitoring, all combine to contradict the primary intentions. This complex system of state aids accumulated over the years has been a windfall for publishers and may be questioned on the ground that it may compromise their independence from the state authorities. The many state aids to the press can be grouped in three main categories: reduced distribution tariffs (for example, reduction of postage fees and transport); tax advantages (for example, reduced professional tax, reduced company tax, reduced value-added tax for printing and for sales); and subsidies (for example, aids to publications in difficulty, aids for local newspapers, sub- sidies for export, and exceptional aids offsetting reduction of advertising resources). The exact amount of these reductions and advantages is difficult to determine; it is believed that this financial ecosystem costs the public a huge amount. This public support maintains the press in a situation of economic dependency. This system does not stimulate rapid adaptation of the structure of the French press or any changes in its management. It has generated such a high level of interdependence between the media and state authorities that reform is problematic. Whatever its perverse effects in practice, this state aid regime is well organized from the legal and administrative points of view. The stringent legal framework prevents state authorities from attributing or refusing aids for political reasons. Indeed, the system does not allow discrimination of any kind among publications requesting public aids. All of these advantages are granted automatically without compromise and without compensation of any kind from the beneficiaries. Therefore, at the legal and administrative levels, the system should not endanger the independence of the editorial content. This nondiscriminatory system helps certain marginal publica- tions survive, for example, the communist newspaper L'Humanité, whereas the United Kingdom's sole communist newspaper, The Morning Star, went out of business many years ago. Newspapers in the United Kingdom are traditionally segmented into two types. Broadsheets are larger in size and are seen as being more intel- lectual and upmarket; tabloids are smaller and seen as more downmarket than broadsheets. Also known as "red tops" because of the color surround- ing their mastheads, they contain more stories about celebrities and gossip. Broadsheet titles have lower circulation but compensate with high-priced advertising, given the higher earnings of their readership. Tabloids have 186 Pierre-Yves Andrau much higher circulation. This distinction is still fundamental in terms of heritage, even if several broadsheets have changed to tabloid format and may have altered their traditional style.4 The media in the United Kingdom are subject to self-regulation. The Press Complaints Commission (PCC) handles complaints from the public about the editorial content of newspapers and magazines. The editors' Code of Practice provides a framework for investigating complaints. The code binds all national and regional newspapers and magazines and was devel- oped by the editors themselves. The code covers the way news is gathered and reported. It is intended to protect vulnerable groups, such as children, hospital patients, and those at risk of discrimination. Self-regulation works because the publishers are committed to it and because the state authorities do not intervene officially.5 The PCC is funded by the industry but operates independently. Newspapers and magazines pay an annual levy to the Press Standards Board of Finance. This model derives from the self-regulatory body established by the advertising industry in 1974; it guarantees both the PCC's financial base and its complete independence, given that nonindus- try members constitute a structural majority. The PCC receives no govern- ment funding and is free from statutory control. The aim of self-regulation is to ensure the freedom of the press. It seeks to combine high standards of ethical reporting with a free press. According to the British view, and con- trary to the French one, statutory controls would undermine the freedom of the press. The current standards achieved through self-regulation will be in any case higher than the ones that would be reached with statutory control, according to the system's proponents. In the same way, any private law regulating privacy or the right of the personality would not be viable in the British context and would be regarded as an intolerable infringement on freedom. As far as the audiovisual media is concerned, it is clear that even in the most liberal country, the regime governing the audiovisual media is rather more restrictive and detailed than the one for the written press. This situation is historically due to the technical characteristics of broadcasting and it persists because of the political importance of television. As for the written press, the private audiovisual sector has three obligations: transpar- ency, independence, and pluralism. As in many countries, broadcasters are submitted to a licensing process. The development of French broadcasting and the evolution of French politics have been closely linked. In 1982 the state monopoly on broadcast- ing was abolished, and in 1986 private television channels were allowed. Aspects of Two Media Models 187 Today, broadcasting is a dual public-private system. State authorities are very active in the implementation of regulation. The government is in charge of designing broadcasting policies and laws. The Parliament con- trols the funding of public broadcasters. The High Council for Broadcast- ing is responsible for granting licences to private broadcasters, appointing the heads of public broadcasters, and supervising the programming of all broadcasters. The French public service broadcaster has been financed by license fees and advertising until the end of 2008. From 2009, advertising may no longer be a resource. According to recent political reflections, the specific roles of the public service broadcaster assigned by law are to ensure free expression for all political and social representatives of French soci- ety. Specific programming obligations are imposed on public and private broadcasters, aimed at ensuring pluralism and diversity of opinions, pro- tecting young audiences, and limiting advertising. Provisions are intended to protect the French language and culture, such as programming quotas and restrictions, and support the production of films and other audiovi- sual works. Rules on cross-ownership apply to commercial broadcasters, but in practice they do not prevent concentration and the emergence of large multi-media groups. The development of ICT proves to be rather dif- ficult compared with some other European countries because new media are expected to fit existing market structures rather than define their own structures. Cable television has developed and satellite pay television is extremely successful. The French state has successfully promoted digital terrestrial television. The principle of public service has strongly shaped the audiovisual land- scape in the United Kingdom. All terrestrial broadcasters have public service obligations. This principle has traditionally been supported by a political consensus in the country. The British model of broadcasting was considered as the most stable and innovative, originally based on relatively few chan- nels, well-funded by different, noncompeting sources, such as licence fees and advertising. The British broadcasting industry--including independent program production--is one of the largest and most dynamic in Europe and hasoneofthehighestpenetrationsofdigitaltelevisionservices.However,the implementation of public policy objectives is shifting, given that the impo- sition of public service obligations on commercial, terrestrial broadcasters no longer seems sustainable. The advent of multichannel television avail- able on many platforms means that terrestrial broadcasters no longer have privileged access to the public, and advertising revenues are fragmenting. Accordingly, commercial broadcasters are less willing to accept expensive 188 Pierre-Yves Andrau public service programming obligations in exchange for the terrestrial radio spectrum. The Office of Communication is developing a mechanism to replace the public service programming contribution previously made by commercial broadcasters, the so-called Public Service Publisher. This would fund desirable programming and other content deemed to be in the public interest, although the government has not yet endorsed the concept. These changes mark the passing of the United Kingdom's unique model, whereby the terrestrial broadcasters as a whole were legally required to provide a television service that not only entertains but also educates and informs the public across a wide range of subject areas with quality programming. There is a clear recognition at the policy level that the market structure has already changed in the multiplatform, multichannel era and will change even more as broadband penetration rises. Terrestrial television--the free view platform--still has a special status for many U.K. citizens and interest groups and is managed by a BBC-led consortium. Meanwhile, in other countries, including France, terrestrial television is still at the heart of the audiovisual system.6 The market structure is defined by public authorities, justified by the need to maintain pluralism and use of the rare spectrum resource. Every effort is being devoted to maintain its status of primus inter pares, with the terrestrial market structure mirrored on other platforms, notably through must-carry obligations.7 There is still some commonality between the United Kingdom and France, thanks to the role of terrestrial broadcasting, but this is likely to change. The lower level of fund- ing of France Télévision compared with the BBC remains a much more sub- stantive difference between the two systems. It remains to be seen whether the combination of the BBC and new funding arrangements for public service programming8 will sustain the successful implementation of public service objectives across the United Kingdom's television market, or whether the United Kingdom will increasingly resemble the more commercial continen- tal European broadcasting markets. In these markets, the whole system has become highly commercial, including much of the output of public service broadcasters, seeking to guarantee their future through high audiences. European Media Governance With the development of satellite television and online communication, the European media landscape has altered significantly over the past few years. This internationalization and globalization of the means of com- munication have had substantial influences on both national regulation Aspects of Two Media Models 189 and the actions of the authorities. Media regulation has acquired an inter- national dimension, and national regulation has had to adapt to this par- adigm shift. Without reform, national regulations would in any case be partially deprived of their efficiency and utility. In this context of increas- ing internationalization, European law has had a growing impact on the national regulatory frameworks of the EU member states.9 It is worth recalling that European law cannot be reduced to an interna- tional law applied to a precise geographical area. Its influence is much more subtle and arguably much more effective. European law is adopted according to a very original process and includes sanctions. In a nutshell, the sources of the applicable laws to media in the EU framework originate in all the trea- ties; these are further elaborated through secondary rules (regulations and directives) and the case law of the Court of Justice of the European Com- munities (CJEC). The fundamental principles founding the EU--the free circulation of goods, persons, services, and capital--and competition law-- rules on unfair agreements and the state aids--apply to all activities and the media as well. The judicial control of the respect of these principles is undertaken by the CJEC. The role assigned to the European Commission by the treaties is to propose and implement Community legislation. The Com- mission has a duty to ensure the treaties and laws are upheld; this means that the Commission, together with the CJEC, is responsible for making sure EU law is properly applied in all the member states. This role is infor- mally known as the "guardian of the treaties." The Commission is the only institution of the EU with the right to propose legislative initiatives, which then have to be adopted in most cases through a co-decision procedure by the European Parliament and the Council. The Directorate General Information Society and Media within the Commission is responsible for media and audiovisual policies. The Euro- pean Information Society sector has grown partly due to initiatives such as the creation of the Single Market, the Television without Frontiers Directive, the adoption of harmonized standards such as the Global System for Mobile communications, DVB television standards, and the liberalization of the telecommunications sector. Today, there are two main areas of regulation in the information society sector: electronic communications (networks) and content. One of the areas of "hard law" regulation is the "Television With- out Frontiers" Directive10 (TVWF); only the general laws governing the EU apply to written media. The TVWF is complemented by certain "soft" laws, such as the Recommendation on the Protection of Minors and other policy initiatives, or initiatives such as the initiative on media pluralism. 190 Pierre-Yves Andrau Community policy in the area of the regulation of media content is essentially internal market policy and is governed and limited by the internal market objective of freedom of movement for goods (including newspapers and magazines, for example) and services (including radio and television broadcasts). Adopted in 1989, the TVWF Directive estab- lished the country of origin principle for television services. The directive was revised in 1997 to take account of technological and market develop- ments. As confirmed by the 2003 review, the directive has provided a stable and secure legal framework for broadcasting services in the Community ever since. The TVWF Directive lays down the minimum standards that national regulation of television broadcast content licensed by the member states must guarantee. Television services of broadcasters that are subject to the jurisdiction of a member state and that meet the legal requirements applicable in that member state may be freely received and rebroadcast throughout the Community. In accordance with the initial focus on the importance of the media and communication for a free, democratic, and social society, this minimum standard essentially comprises the following obligations: · promote the production and dissemination of European television programs · protect consumers with respect to advertising, sponsoring, and teleshopping · protect minors and the public order · protect the right of reply · adopt measures to ensure that events of major importance for society are not broadcast on the basis of exclusivity in such a way that significant sections of the public in the relevant member state are prevented from following the event in a freely accessible television broadcast. Community law at present distinguishes between television broad- casting services subject to the TVWF Directive and information society services subject to the e-Commerce Directive. The regulatory approach of the TVWF Directive is more detailed than the regulatory approach of the e-Commerce Directive. This is due to the impact of television broadcasting on society, giving it paramount importance in political and cultural terms. Digital technologies and convergence will increasingly involve new forms of content, new means of providing it, and new business models to finance it. Regulatory policy in the sector has to safeguard certain public interests, such as cultural diversity, the protection of minors, and consumer protec- Aspects of Two Media Models 191 tion now and in future. The wide-ranging consultation that the Commis- sion conducted in 2003 confirmed that these objectives are not called into question by technological or market developments. What is questioned are the means by which these objectives can be achieved in a changing environment. The modernization of the TVWF Directive was launched in December 2005 by a new legislative proposal for a modernized audiovisual services directive. The new directive is based on the distinction between linear (television broadcast) and nonlinear (on demand) services. Because users/consumers have greater control and choice in the case of nonlinear services--the possibility to decide at which moment in time a specific pro- gram chosen in a catalog is transmitted--the rules for these services are lighter. The rules for on-demand services contain mainly provisions on the protection of minors; prohibition of incitement to hatred; identification of the media service providers; and basic rules for commercial commu- nication, such as identification of commercial communication, prohibi- tion of surreptitious advertising, clear rules on product placement and sponsoring, as well as some qualitative restrictions on advertising. Linear services have to comply with more specific rules, mainly due to the unpar- alleled impact of television broadcasting on the way people form their opinions. The written media are not submitted so far to any specific texts at EU level. All liberal economic freedoms apply to this special sector, which constitutes the basis of the EU, that is, the free circulation of goods, per- sons, services, and capital, and the competition law. In the same way, no European Treaty text derogates from these principles, which would be able to take into account the specific democratic and cultural role of the writ- ten media.11 This is a huge difference compared with audiovisual media, whose characteristics and development make action at the EU level more necessary. Moreover, the audiovisual media are not confronted by the same obstacles; in the 1980s it was generally admitted among the member states that special attention should be paid at the EU level to the audio- visual licensed media. On the contrary, the member states never agreed unanimously on the application of liberal economic principles to the writ- ten media. Notably, the most interventionist member states such as France want to keep their specific media legislation, conceived as a national guar- antee of the freedom of expression and the right of the public to informa- tion and not limited to the application of liberal economic principles. A draft directive on the press was proposed in the 1960s but withdrawn in 1974.12 Therefore, as far as EU regulations on written media are concerned, 192 Pierre-Yves Andrau all the rules and general principles included in the treaties apply. In the same way, all general derogations apply horizontally to all sectors, includ- ing the media. The sources of EU law applicable to the written media are to be found in the EC Treaty as interpreted by the CJEC. Maintaining media pluralism is an essential condition for preserving the right to information and freedom of expression that underpins the democratic process. The EU is committed to protecting media pluralism as well as the right to information and freedom of expression enshrined in Article 11 of the Charter of Fundamental Rights. Similar provisions are included in Article 10 of the European Convention for the Protection of Human Rights and Fundamental Freedoms.13 The European Commission has long faced political pressure to legislate from the European Parliament, but years of consultation indicate that it would be politically inappropriate for the Commission to harmonize media pluralism. Subsidiarity is a strong consideration in this field. However, the question of media pluralism is a crucial one in the Inter- net age. The debate within the EU has been too political and it does not take into account the need for the media to change their structure as a result of new technologies and business models. Moreover, even if plural- ism guarantees an intrinsic element of European democracies, there must also be favorable economic conditions so that European media companies can withstand competition at the international level. The risk, as stressed in the French example, is that every deal on media pluralism is reduced to the sole issue of concentration. The approach followed by the Commission for measuring media pluralism must take into account how new develop- ments in the media might contribute to the dynamism of the public sphere. An original and ambitious new round of monitoring activity to reinforce the Commission's support for media pluralism was launched in January 2007. It is obvious that the Commission, together with other European stakeholders, sorely lacks full knowledge of the diversity of the practices and models for media pluralism in the member states as well as how to avoid locking media pluralism into supporting legacy industry structures. Without a deep understanding of the "real" media pluralism taking place, any initiative clearly courts failure. The three-step approach launched by the Commission responds to this situation as follows: · First, a Staff Working Paper14 set out a preliminary framework and described the various ways of achieving pluralism at the national level. Aspects of Two Media Models 193 The Commission's working hypothesis is that pluralism is a complex matter. It should not be limited just to the question of media concentra- tion, that is, ownership. Therefore, the Commission follows a twin-track approach: the external pluralism of the media and internal pluralism within individual media companies. The approach to media pluralism is broad and includes analysis of all measures that ensure citizens' access to a variety of information sources, opinions, and voices in order to build public opinion, without the undue influence of one dominant opinion- forming power. · The second step will be a comprehensive study on media pluralism in order to develop indicators. A number of available studies analyzes legal instruments and some aspects of real media pluralism. None of these studies systematically identifies concrete indicators to mea- sure media pluralism in the member states. An independent study will analyze possible indicators for assessing pluralism. The multiple dimensions of this issue requires a multidisciplinary approach. The identification of concrete indicators marks a new approach and will enable citizens and all interested parties to assess media pluralism in the member states more objectively. · As a third step, the indicators developed in the study will be presented to the general public. This will enable citizens to have a say, together with European institutions, regulators, professionals, and nongovern- mental organizations in the framework of their respective competences. This process will enhance our understanding of media pluralism and set some realistic indicators for measuring it in a future that will be very different from the past. Conclusion This chapter has reviewed certain media regulations in two member states and some aspects of EU law and initiatives in this field. However, the media industry is subject to some horizontal laws. This is the case at the national level and EU level. At the EU level, media governance is not limited to the hard or soft law as described above. In the framework of the competence granted to the EU by the member states, certain policies are implemented, including those for agriculture, public health, employment, the environ- ment, and the economy. The implementation of the single market and EU competition law has created common legal standards and a very deep 194 Pierre-Yves Andrau harmonization. Moreover, the EU has acquired new competences in fields like criminal and civil laws and antiterrorism initiatives. All of these poli- cies or regulations may have an impact on the media, including the written, audiovisual, online, or traditional press. Even if some EU policies or regula- tions have an impact on the media, it is important to stress that none of the EU policies or initiatives is designed to encroach upon the media's editorial freedom. None is meant to interfere with journalists' investigative freedom or exert control on media content. Whatever the intention of the policies or regulation and whatever the media models in force in various member states, either inspired by the French or British model, the European Commission has developed an original approach toward governance of the media industries. Within the Directorate General Information Society and Media of the European Commission, the Media Task Force acts as a sounding board for other Com- mission services by prescreening policy proposals at the conception stage. This helps to establish whether they will affect the editorial freedom of the media or have unintended or policy initiatives affecting the economy of the media industry, thereby affecting editorial freedom. The approach of assessing the total weight of policy and regulation affecting the media across the whole of the institution's output is entirely different from the normal approach within the Commission, which is normally confined to policies and regulation affecting a particular sector and essential interfaces with other policies. The Media Task Force is also developing an economic research role in order to prime media policy, given the scale of changes foreseen to industry structure in the long term. The first priority is to develop the Commission's understanding of publishing to the same level as for audiovisual media. In addition, the Media Task Force is engaged in a fruitful dialogue with the publishing industry in order to reflect on the adaptation of this indus- try to the digital age and especially the shaping of new business models.16 The establishment of the Media Task Force is a new form of collaboration between public authorities and the media; it is indeed unique as no other sector benefits from such a coordination mechanism. The partnership is not "a one-way street" and maintains a constant dialogue between the media industries and the Commission; it is sufficiently robust for a frank exchange of views on any topic if judged necessary. This approach is pos- sible because the European Commission does not act within a national public sphere and is therefore better able to take the necessary distance for improved governance. Aspects of Two Media Models 195 Notes 1. Although Europe's publicly funded public service broadcasters do not have to compete to make a profit, they must still attract audiences. 2. The term liberal is used in its European sense to mean pro-market and minimal regulation, rather than in the U.S. sense of left-leaning or progressive. 3. Free trade is not put in question in France even if a recent survey shows that more than 60 percent of the French people are against the market economy as a general opinion. 4. Some commentators claim that smaller pages lead to shorter and more super- ficial coverage. 5. This does not prevent political concern from being expressed. During March 2007, the BBC Parliament Channel carried coverage of the relevant House of Commons Select Committee into certain practices of the U.K. press, inter alia paying third parties for certain stories. 6. This is also true in the United States, where the local structure of terrestrial television is intended to provide a form of pluralism and reflect the diverse views of different communities. 7. Must-carry rules make it compulsory for network operators (such as cable companies or telecom operators) to carry specified radio and television broad- cast channels and services where a significant number of end-users of such infrastructure use them as their principal means to receive radio and televi- sion broadcasts. Member states may, under the EU's Universal Service Directive, lay down reasonable must-carry obligations for the transmission of specified broadcast channels and services on the network operators under their jurisdiction, for legitimate public policy reasons. However, such obligations should only be imposed where they are (a) strictly necessary to meet clearly defined general interest objectives, (b) proportionate and transparent, and (c) subject to peri- odic review. 8. One of the stated objectives is to ensure the contestability of this new form of public funding and to decouple it from one institution. The BBC is cur- rently the sole beneficiary of the license fee paid by all television-owning households. 9. The chapter does not treat the important role played by the Council of Europe and the European Court of Human Rights in the protection of the fundamen- tal freedoms, notably the freedom of expression and freedom of press. 10. Council Directive 89/552/EEC as amended by Directive 97/36/EC of the Euro- pean Parliament and of the Council. 11. Article 20 of the Merger Regulation does, however, permit member states to apply supplementary criteria to concentrations involving media companies beyond the provisions of competition law. This covers both licensed and unli- censed media. 12. 65/106/EEC. 13. The European Court of Justice underlined the importance of maintaining press diversity. "Such diversity helps to safeguard freedom of expression, as 196 Pierre-Yves Andrau protected by Article 10 of the European Convention on Human Rights and Fundamental Freedoms, which is one of the fundamental rights guaranteed by the Community legal order" (Case C-368/95, paragraph 18). 14. SEC (2007) 32 of 16.1.2007. 15. For example, the directive on tobacco advertising, the directive on unfair com- mercial practices, the proposal for an EU regulation on food claims, the direc- tive on equal treatment between women and men in access to and supply of goods and services, the e-commerce directive, copyright rules, environmental regulations, social policy, VAT rules, and Rome II. 16. http://europa.eu.int/information_society/media_taskforce/index_en.htm. 12 Three Countries: Three Stories M edia market conditions differ greatly among countries. In this chap- ter, three practitioners in the media industry provide snapshots of market and government interventions in the media sector. The Nigerian media operate in small markets for products and advertising--and the state is a dominant owner. Ojo contends that the dominance of ownership gives the state an advantage in disseminating pro-government information. He underscores the role of government regulation in setting an uneven playing field between the private and public sectors. Majed highlights the turbulent phases that media companies in Leba- non have undergone in recent decades, related primarily to political tur- moil. His account mentions state censorship as well as self-censorship by media owners afraid of political reprisals. The Lebanese press is privately owned, but political parties still own some media. The Indonesian story related by Harymurti brings to the fore the impor- tance of an independent judiciary in overseeing and guaranteeing media freedoms that may be written into the law. Judicial attitudes influence how "free" a country's media sector is; the Indonesian judiciary has recently become a more active supporter of greater media freedom. Harymurti high- lights some concerns with concentrated media markets in Indonesia and nontransparency with respect to ownership of the media. All three authors refer to political history, social attitudes, and laws in defining how much media freedom exists in their respective countries. Majed focuses on political history in light of Lebanon's particular history. Ojo refers to the culture of secrecy in Nigeria. Harymurti mentions recent changes in the judiciary and perhaps social attitudes that allow for greater freedom of the media in Indonesia. All three authors acknowledge that media independence or freedom cannot be taken for granted. They relate greater diversity in the media sector to a stronger private sector that may provide checks and balances to the public sector. 197 Media and Governance in Nigeria Edetaen Ojo T he Nigerian media are regarded as the most advanced and well- developed in West Africa and among the best on the continent. While the Nigerian media have a robust tradition, media organizations in Nige- ria can hardly be said to be strong and viable. On the contrary, the media face a range of challenges that raise serious questions about the accuracy of its characterization as the most well-developed in West Africa. These challenges include lack of access to adequate capital; a poorly devel- oped market for media products arising from the high level of illiteracy and widespread poverty; poor economic conditions of media establish- ments in the face of high production costs; poor media facilities and infrastructure; widespread corruption and unethical practices in the media industry; and inadequate legal, regulatory, and policy frameworks for media practice. The Structure of the Nigerian Media The Nigerian media landscape presents a complex scenario. Despite a popu- lation of about 140 million, the nation's largest newspaper has an estimated average daily circulation figure of less than 100,000 copies. No reliable cir- culation or readership figures are available, since newspapers guard their circulation figures jealously. However, a study by the Baltic Media Centre in 2001 (Bonde and others 2001) estimated that the combined daily circula- tion of all the national, regional, and provincial newspapers in the country is between 350,000 and 500,000. There is little concentration in the ownership of newspapers in Nige- ria. Entities claiming to own groups of newspapers frequently refer to their regular daily editions, Saturday (or weekend), and Sunday editions as con- stituting the "group." Apart from these, hardly any newspaper group pub- lishes more than one or two different titles. This feature of the 1980s and early 1990s, when publishers such as the Guardian Newspapers Limited 198 Media and Governance in Nigeria 199 published up to five different titles, has died with the current harsh eco- nomic environment in the country. The newspaper and newsmagazine industry is dominated by private entities. The federal government currently owns only one newspaper, the New Nigerian, while some state governments own various provincial news- papers. Although the New Nigerian lays claim to being a national newspa- per, it is not available in many major urban centers around the country. No state-government owned newspaper circulates nationally, and many only appear on the newsstand intermittently. About 10 national newspapers are circulated regularly on a daily basis. These are all privately owned news- papers, but even these "market leaders" are plagued by low readership and poor distribution. They face the challenge of survival, arising from high production costs and a poor market. As a result of the poor economic conditions of most media organiza- tions, journalists are poorly paid. On average, the vast majority of journalists earn less that US$400 monthly. Even these low salaries are not paid regularly. It is common in Nigeria to find news organizations in arrears on their work- ers' salaries from three to 12 months. Many journalists blame the high level of unethical practices and corruption in the media on the poor pay and the arrears; journalists' economic circumstances make them susceptible to eco- nomic pressures from government officials, businesses, and other groups. The high cost of producing newspapers in Nigeria is worsened by the poor state of the nation's infrastructure, particularly the electricity supply and communications, as well as the unstable economic environment. This situation has consequently resulted in high costs for buyers of newspapers. Most newspapers sell at between N80 (US$0.62) and N200 (US$1.56) per copy, while newsmagazines sell at between N150 (US$1.17) and N200 (US$1.56) per copy. Given that about 70 percent of Nigeria's population is estimated by the United Nations Development Programme to be living on less than one dollar a day (UNDP 2003), it is apparent that there is no meaningful market for newspapers and newsmagazines. The newspaper industry has tried to recover production and opera- tional costs through advertising. The industry has published advertis- ing rates that should enable its members to make profits or at least break even; periodically, the industry revises these rates upward whenever the costs of newsprint and other production materials rise. The result is that the advertising rates published by many newspapers and newsmagazines have become prohibitive in themselves and are hardly justified by their low 200 Three Countries: Three Stories circulation figures. As a consequence, only a handful of newspapers are able to attract advertising to sustain their operations. Many news organizations are unable to provide the necessary resources and facilities for their reporters to operate effectively. In many cases, report- ers working for such organizations depend on facilities and resources provided by government officials, business concerns, or self resources to perform their reporting functions. Typically, journalists based in states out- side the headquarters of their organizations frequently rely on the commu- nication facilities of the press secretary to the governor or other officials to transmit their reports to their head offices as their organizations are unable to provide these facilities. This naturally undermines independent report- ing on the part of such journalists. The current 1999 Nigerian Constitution provides that: Every person shall be entitled to own, establish and operate any medium for dissemination of information, ideas and opinions, provided that no person other than the government of the Federation, or of a State or any other person or body authorized by the President, Commander-in-Chief of the Armed Forces shall own, establish or operate a television or wireless broadcasting station for any purpose whatsoever. This provision, which had been replicated in all previous Nigerian con- stitutions, ensured that while private entities could own newspapers and newsmagazines, only government at the national or state level could own radio or television stations. As a result, all the radio and television stations in Nigeria were owned and controlled by either the federal or state govern- ments. This situation persisted until 1992 when the military government of President Ibrahim Babangida enacted the National Broadcasting Com- mission Decree No. 38 of 1992, authorizing the licensing of private entities to own and operate radio and television stations. The first privately owned broadcast station, the RayPower 100 FM radio station, was licensed in 1993 and began transmission in 1994. Television Stations There are about 144 television stations in Nigeria. Of this number, 98 sta- tions are owned by the federal government and consist of a national chan- nel, the Nigerian Television Authority (NTA) and 97 local NTA affiliate stations spread around the country. In addition, state governments account for the ownership of 32 television stations, while 14 terrestrial television stations are owned by private entities. Media and Governance in Nigeria 201 Radio Stations There are about 104 radio stations in Nigeria. Out of this number, the fed- eral government owns 40 stations, spread across the different states. Five of the federal government-owned stations are network stations located in Abuja, the federal capital; in Lagos, the commercial capital; and in Ibadan in the southwest, Kaduna in the north, and Enugu in the southeast. The state governments own 39 radio stations, while private entities own 25 radio stations. Of the 25 private stations, 20 are commercial broadcasters while five are specialized radio stations. The specialized radio stations are the Atlantic FM station, which broadcasts French programs; the Spectrum FM station, which is a whole news station; BRILLA FM, which is a sports station; UNILAG FM, owned by the University of Lagos's mass communica- tion department and is an experimental station for educational and aca- demic programs; and Crowder FM, owned by the Anglican Church. Government Dominance Despite the emergence of broadcast stations owned and operated by pri- vate entities, it is evident that the media landscape in the broadcast sector continues to be dominated by the government. Nigeria does not have any comprehensive antitrust laws. The only attempt at addressing the issue of concentration and monopolies is contained in section 9(5) of the National Broadcasting Commission Act, which states: "It shall be illegal for any per- son to have controlling shares in more than two of each of the broadcast sectors of transmission." The dominance of the broadcast media by the government has signifi- cant implications for fairness of coverage of a broad range of issues and sec- tors in the society. Whether at the state or federal level, media coverage is almost always lopsided, especially regarding political issues in favor of the ruling party and the chief executive. There is a tendency by those in power at both the federal and state levels to view the state-controlled broadcasting stations as the propaganda arms of their governments and to use them to advance their partisan or personal interests while preventing access to these media by their opponents and opposition parties. As the ruling party at the federal level, the Peoples Democratic Party and President Olusegun Obasanjo control the federal government-owned news organizations, NTA and the FRCN. Control is exercised directly by the minister of information. Critical voices hardly ever come through any of these stations and opponents are seldom covered. When they are covered, 202 Three Countries: Three Stories they are frequently presented in a negative light. This pattern is replicated in almost all the states, regardless of which political party is in control. Media Laws and Regulations The state of media laws and regulations in Nigeria today is no different from what it was at the end of military rule eight years ago. No new media laws have been enacted since the restoration of democratic rule on May 29, 1999, and none of the objectionable media laws in existence as of that date have been repealed. In the 15 years of military rule preceding the restoration of democratic rule, the promulgation of restrictive decrees was one of the main tools used by the successive military governments to control the media. Although some of the decrees were not specifically aimed at the media, they were decrees nonetheless used against journal- ists. Some of the general decrees used against journalists are the State Security (Detention of Persons) Decree No. 2 of 1984 as amended and the Treason and Other Offences (Special Military Tribunal) Decree No. 1 of 1986. In addition to these were several decrees established specifically for the media, such as the Offensive Publications (Proscription) Decree No. 35 of 1993 and the Newspapers Decree No. 43 of 1993, the latter being one of the most draconian media laws ever made in Nigeria. They existed along- side longer standing laws, many of them dating back to the colonial period, which had negative implications for media freedoms in Nigeria. Nearly all the decrees were abrogated by the military government of General Abdulsalami Abubakar a few days before it handed over power in 1999. Many others, however, remain in force. Nigerian Press Council (Amendment) Decree One of the most potentially dangerous media laws in Nigeria today remains: the Nigerian Press Council (Amendment) Decree No. 60 of 1999, which was promulgated by the government of General Abdulsalami Abubakar on May 26, 1999, three days before it handed over power to an elected civilian gov- ernment. The decree introduced a draconian regulatory process designed to subject the media to governmental control; it incorporated many of the elements of its controversial predecessor, the Newspapers Decree No. 43 of 1993, which had been the subject of widespread criticism and which the Abubakar regime repealed. Media and Governance in Nigeria 203 Under Decree No. 60, the Press Council is empowered to register jour- nalists before they can practice in Nigeria and to strike their names off the register of journalists when they contravene a code of professional ethics approved by the Council. The Press Council is also authorized to register all newspapers, magazines, and journals in Nigeria before they can pub- lish. Failure to comply with any of these provisions amounts to a crimi- nal offense for which the affected journalist or publisher is liable to be imprisoned for up to two years, pay huge amounts in fines, or be sanctioned through both imprisonment and fines. The law also provides that the registration of newspapers and maga- zines be subject to renewal annually, which apparently means that publica- tions that have been critical of the government or too independent stand the risk of not having their registration renewed. The implementation of this law has been frustrated by resistance from newspaper proprietors and media freedom organizations.1 National Broadcasting Commission Decree While the newspapers and newsmagazines battle with the potential effects of the amended Newspapers Decree, the broadcast sector faces a licens- ing and regulatory framework that inhibits independent broadcasting. The promulgation of the National Broadcasting Commission Decree No. 38 of 1992 by the government of General Ibrahim Babangida was in many respects a positive development, despite some of the shortcomings in the decree. The decree deregulated the broadcast industry by providing for the licensing of private entities to own and operate radio and television sta- tions, which previously had been monopolized in all of Nigeria's history by the state and federal governments. However, the independence of the National Broadcasting Commis- sion (NBC), which is established by the decree to consider applications for licenses, revoke licenses already approved, and to regulate the industry gen- erally, is undermined by the fact that a provision in the decree empowers the minister of information to "give the commission directives of a general character relating generally to particular matters with regard to the exercise by the commission of its functions" and imposes a duty on the commission to comply with such directives.2 Moreover, contrary to the provisions of the decree that the commis- sion should be run by a board, the commission is in reality run by an indi- vidual designated as director-general, who is answerable to the minister of 204 Three Countries: Three Stories information and the president. The result has been the tendency to stifle private broadcasters under the guise of regulation. The commission imple- ments a discriminatory regime of licensing fees for state-owned broadcast stations as against privately owned broadcast entities. For instance, while the stations owned by federal and state governments pay a flat fee of only N500,000 (about US$3,800), the private radio and television stations are made to pay fees ranging from N10 million (about US$78,125) to N20 mil- lion (about US$156,250), depending on the type of station, the system, and the locations. These fees are payable both for the original license and for renewals. These fees undermine the viability of private broadcast entities. Many private broadcast entities are frequently unable to pay the fees for the renewal of the licenses and there have been many instances where the NBC has either shut down stations or threatened to shut them down for their inability to pay their license fees. Yet, such private broadcast entities com- pete with state-owned broadcasters for commercials. Besides, their com- petitiveness is undermined by the fact that they are not allowed to run network programs, while the NTA and the Federal Radio Corporation of Nigeria run network programs. 1999 Constitution In addition to these critical issues facing the Nigerian media, several other laws exist whose provisions inhibit media freedom. Chief among them is the 1999 constitution itself, which has failed to specifically guarantee press freedom or media freedom, access to information, editorial independence, and the confidentiality of journalistic sources. Instead, section 39 simply rehashed section 36 of its predecessor, the 1979 constitution, which guar- anteed a general right to freedom of expression to every Nigerian citizen. In effect, the 1999 constitution, by its section 22 imposes obligations, duties, and responsibilities on the media without according them any right or pro- tection in the discharge of these constitutional responsibilities. Other Laws Other laws whose provisions affect media freedoms include the Official Secrets Act of 1962 and section 97 of the Criminal Code, which deal with the unauthorized disclosure, in whole or in part, of documents classified or tagged as official secrets. The laws punish both the public officer disclosing the information and any other person (obviously including the journalist) Media and Governance in Nigeria 205 who either receives the information or further discloses it. These laws repre- sent the greatest hindrance to access to information held by government or public officers since virtually all official documents are routinely classified as "secret" or "top secret," thus bringing them under the protection of these laws. This means that the disclosure, abstraction, transmission, reproduc- tion, or retention of such documents is a crime, punishable with as much as 14 years' imprisonment if it is under the Official Secrets Act, or one to two years imprisonment if it falls under section 97 of the Criminal Code. Sections 373 to 381 of the Criminal Code (which is applicable in the southern part of Nigeria) and sections 391 to 392 of the Penal Code (appli- cable in northern Nigeria) criminalize defamation and make it punishable under the Criminal Code, depending on its nature, with periods of impris- onment ranging from one to seven years, while the Penal Code provides for up to two years imprisonment as well as a fine. Other worrisome provisions of the Criminal Code are section 59, which makes the publication of false news an offense punishable with three years imprisonment and section 60, a provision of colonial origin that criminalizes the defamation of foreign princes or sovereigns and makes it punishable with two years of imprison- ment. Contrary to the provisions of section 36(5) of the 1999 constitution, which presumes an accused person innocent until the contrary is proved, under the provisions of section 233 of the Criminal Code accused persons are presumed guilty until proven innocent. There is also the Defamatory and Offensive Publications Act No. 44 of 1966, which provides punishment for the producers and purveyors of both audiovisual and print publications that are considered offensive to society with terms of imprisonment of up to three months or a fine or both. A simi- lar law is the Obscene Publications Act of 1962, which is now applicable in Abuja and the southern states by virtue of sections 232 and 233 of the Criminal Code. It prohibits the publication, reproduction, distribution, or exhibition of any material that "tends to deprave and corrupt," and it pro- vides that the distribution or projection of any article defined under the law as obscene is an offense punishable with up to three years of imprisonment, or a fine, or both. In addition, other laws were enacted during the colonial period but remain as part of the statutes. These include the Sedition's Offences Act (1909), which introduced the offense of sedition into Nigerian law and is now incorporated into the Criminal Code for southern Nigeria and the Penal Code for northern Nigeria and some other state laws. The definition of "seditious intention" used by the laws is broad and can be interpreted 206 Three Countries: Three Stories by government officials to include any expression of dissent of which they disapprove. There are many examples of the government prosecuting jour- nalists under charges of sedition. There are worrisome provisions in the laws establishing some federal government-owned media organizations such as the Nigerian Television Authority Act of 1976, the News Agency of Nigeria Act of 1976, the Federal Radio Corporation of Nigerian Act of 1978, and the Voice of Nigeria Corpo- ration Act. These laws provide for governing boards and financial structures controlled by the government and give the federal government powers to intervene in the editorial decision making of these organizations. These laws also give the minister of information powers to give directives of a general character relating to particular matters and also imposes a duty on the organizations to comply with such directives. The consequence of this situation is that rather than being public service media, these media organizations see themselves and operate as propaganda arms of the government and frequently deny critics of the gov- ernment access to the media. Access to Official Information Another challenge the Nigerian media face is the pervasive culture of secrecy that characterizes the conduct of official business at all levels of government and continues to hamper the free flow of information. Many public officials and civil servants do not appreciate the role of citizens in the governance process and continue to place obstacles in the way of media practitioners and members of the public seeking to access government-held information. The environment for public access to official information is hostile, and legal and institutional frameworks impede public access to informa- tion rather than facilitate it. There is no law in Nigeria giving members of the public access to official records, documents, or information. Even in some cases where it is apparent that a given law--such as the Code of Con- duct Bureau and Tribunal Act--recognizes that members of the public have a role to play in the effective implementation of the law and the achieve- ment of its purpose, the mechanisms for effective public participation are either absent or are so vague that the overall effect is that such mechanisms do not exist.3 On the other hand, Nigeria continues to retain dozens of laws with secrecy clauses that prohibit public officers and civil servants from dis- Media and Governance in Nigeria 207 closing information, even when there is no apparent justification for such prohibition.4 In some of these laws, even the courts are precluded from compelling the public officers to disclose such information. Besides these secrecy clauses, many categories of government officials are obliged upon taking office to subscribe to an oath of secrecy under which they undertake not to disclose any information that comes to their knowledge or custody in the course of the performance of their duties, unless specifically autho- rized to do so. In addition to the general oath of secrecy to which public officers and civil servants are obliged to subscribe, an additional undertak- ing is required for persons taking up appointments in some government departments, agencies, and bodies.5 There have even been instances where journalists who are employed in state-owned media organizations have also been required to subscribe to such oaths of secrecy.6 This situation has resulted in a severe restriction on the ability of members of the public, including the media, to access information about governmental activities, programs, and policies. The situation has engendered a culture of secrecy in government institutions, with the result that governments and their actions are insulated from public scrutiny. Nigeria's bicameral legislature, the National Assembly, has passed a Freedom of Information bill that seeks to give citizens, including media professionals, access to official records and documents, subject to certain exemptions. On February 14, 2008, both chambers of the National Assem- bly harmonized the different versions of the bill and transmitted it to the president on March 23 for him to sign into law. If the bill became law, it would significantly enhance the free flow of information in Nigeria. Under the provisions of the 1999 Constitution, the president had 30 days to indi- cate whether he intended to assent to the bill or withhold his assent; he declined to assent, and the bill did not become law. Notes 1. For instance, both the Newspapers Proprietors Association of Nigeria and Media Rights Agenda, a nongovernmental organization that works to defend media freedom and freedom of expression in Nigeria, filed separate suits at the federal High Court challenging the constitutionality of the law. The suits are yet to be determined but have been put in abeyance to enable a negotiated resolution of the dispute. 2. See section 6 of the National Broadcasting Commission Decree No. 38 of 1992. 208 Three Countries: Three Stories 3. An example of this is the Code of Conduct Bureau and Tribunal Act, Cap 56, Laws of the Federation of Nigeria, 1990, which established a Code of Conduct Bureau and entrusted it with the responsibility to, among other things, receive assets declarations by public officers and to receive complaints from members of the public about noncompliance with, or breach of, the provisions of the act. It is clear both from simple logic and from the provisions of the act that its effective implementation would require the active participation of members of the public who can bring to the notice of the bureau cases of false declaration of assets. But the act contains no provision or mechanism by which members of the public can ascertain the extent to which public officers have complied with the provisions of the act as there is no right of public access to informa- tion provided under the act. Thus, members of the public cannot contribute to the effective administration of the assets declaration process. This situa- tion is also replicated in the third schedule to the 1999 constitution, which contains no mechanism by which members of the public can have access to assets declarations made by public officers pursuant to the provisions of the constitution. 4. Instances of such secrecy clauses can be found in legal provisions such as sec- tion 168 of the Evidence Act; section 2 of the Federal Commissions (Privileges and Immunities) Act, Cap 130, LFN, 1990; section 10 (2) of the Public Com- plaints Commission Act; section 12 (2) of the Architects (Registration, etc.) Act; sections 53 and 55 of the Nigerian Railway Corporation Act; and section 13 of the Statistics Act, Cap 416, LFN, 1990. 5. Such a requirement can be found in section 28 of the National Population Commission Act, Cap 270, LFN, 1990; section 12 of the Statistics Act, Cap 416, LFN, 1990; and section 3 of the Tribunals of Inquiry Act, Cap 447, LFN, 1990. 6. Journalists working with the Ondo state government-owned media organi- zations say they are required upon appointment to subscribe to an oath of secrecy not to disclose information that comes to their knowledge in the course of their work. References Bonde, B.N., Waseem Mahmood, and Abim Onasanya. 2001. "Nigerian Media: Overview, Evaluation and Draft Programmes for Change." Baltic Media Cen- tre, Copenhagen. UNDP (United Nations Development Programme). 2003. Human Development Report. UNDP: New York. Lebanon: An Overview of the Media Ziad Majed T he media in Lebanon have a long history of freedom and diversity. They also have a long history of "martyrdom" and face many internal and external challenges. This section focuses on aspects of the Lebanese media sector today, including its recent history and the difficulties the media face. Four Major Political Phases Since 1943 A general overview of the evolution of the media sector in Lebanon since the independence of the country in November 1943 reveals four major phases. The first phase occurred from 1943 to the beginning of the civil war in 1975. It is a phase characterized by dynamism; diversity; relative freedom; and the attraction of Arab intellectuals and dissidents to Beirut, its press, and its publishing houses. It is also the phase of pan-Arabism, leftist thoughts, the rise of the Palestinian cause, and the birth of papers and reviews analyz- ing political struggles and partisan propaganda. In this phase, newspapers were at the heart of media life, since radio and television were state-owned channels of communication. The second phase started with the civil war in April 1975 and lasted until 1994, three years after the end of the conflict. The political diversity that characterized the first phase remained a major characteristic of this second one. To diversity were added aspects of censorship practiced by the divided national authorities, as well as militias and warlords in different regions. Political war and clashes covered by the media gained considerable attention. Parties from different camps invested in newspapers and radio and television stations to circulate their messages. By 1993, the country had dozens of private television and radio stations and 10 functioning daily newspapers. The third phase was the post-war first phase from 1994 to 2005, which started when the Lebanese government issued a new law "organizing" the media. The majority of television and radio stations were closed, and a 209 210 Three Countries: Three Stories quota system was established to satisfy large parties representing different communities. This phase witnessed for the first time a large-scale interven- tion in editing news, censoring it, threatening journalists, and manipulat- ing their sources of information through security officers loyal to the Syrian regime that controlled the country until March 2005. Nevertheless, media freedoms and dynamism survived. The long tradition of diversity was main- tained, and it supported the image of the Lebanese media in general, even if the majority of editorialists began to engage in autocensorship. The fourth phase was inaugurated in March 2005, with the withdrawal of the Syrian military. This phase witnessed less autocensorship and fewer interventions with journalists and journalism. However, fears of assassina- tions increased after June 2005, when Samir Kassir, one of the most coura- geous intellectuals, a historian and journalist for the daily Annahar, was assassinated. In September, another journalist, television presenter May Chidiac, was targeted and badly injured; in December, a journalist and a member of Parliament and director general of the daily Annahar, Jobran Tweini, was assassinated. It is true that throughout the four phases, Lebanese journalists have been assassinated as well as subjected to various political pressures and threats by different local groups and regional regimes,1 but it is also true that the assassinations after the Syrian era in Lebanon have a differ- ent significance. Not only has there been revenge against journalists and institutions opposing the Syrian hegemony, but there is zero tolerance with journalists and public opinion makers who would "intervene" in "Syrian affairs," criticize the Assad regime, or discuss democratization in Syria.2 The Legislative Environment The Lebanese constitution provides for freedom of speech. Article 13 states that "Freedoms of expression, oral, written, and published are protected by law." Lebanon has also ratified all human rights conventions and other relevant international accords protecting public and individual freedoms of expression and creativity. Theoretically and legally, no censorship should exist on expression or speech in Lebanon, except according to the 2004 media law in cases of news or positions that Lebanese laws consider as "harming religious beliefs, destabilizing the country, or damaging its rela- tions with its allies." Lebanon: An Overview of the Media 211 The Lebanese press law enacted in 1962 states that any newspaper or periodical wanting to publish political news must obtain a legislative decree insuring a Category 1 license; a Category 2 license is necessary to publish social, cultural, or entertainment news (without the right to cover political ones). At present, to obtain a Category 1 license, a publisher must acquire an already existing license from a publisher going out of business. The Lebanese press is privately owned by political parties and independent entities. The 1994 Audio-Visual Media Law ended the state monopoly on electronic broadcasting and allowed for private radio and television broad- casting to be produced and distributed in Lebanon. Category 1 and Cat- egory 2 licenses are also applicable for the audiovisual sector; there are different licensing fees for the two categories. In terms of the other regulations of note, a television station is required to broadcast at least 4,000 hours per year with 40 percent of the time allotted to locally produced programs and to transmit through Lebanon. A television license is granted for 16 years. Private radio sta- tions are required to have shareholders from major religious communi- ties of the country. According to article 35 of the 1994 law, the ministry of information can close broadcasting for corporations not conforming to the law. As examples of government intervention, in 2002 MTV, a televi- sion station covering an opposition electoral campaign, was closed; during 2004 the government censored television and radio broadcasts on a case- by-case basis.3 With respect to the Internet and the electronic media services, the min- istry of telecommunications announced in 2004 a telecommunications law to regulate the sector. Parties, specialized organizations, and private enti- ties have Web sites and regularly use the Internet and electronic mail as tools for political, journalistic, or news messages and coverage. The Educational Context Journalism and media-related topics are taught in Lebanon at the public Lebanese University as well as in five private educational institutions. Ara- bic, English, and French are languages used in these different institutions that have different levels and quality of teaching services. Media entities with resources provide occasional inhouse training to journalists when the need arises. International specialized organizations that operate nationally and regionally also provide training to journalists. 212 Three Countries: Three Stories The Media Today The Lebanese population residing in the country is estimated today at 3.6 million. The literacy rate among the Lebanese is close to 88 percent. Approximately 600,000 Palestinian refugees and foreign workers live in the country as well. Newspapers In 2006, there were 105 Category 1 licenses for 53 daily newspapers, 48 weekly, and four monthly magazines. Only 20 percent of these licenses are attributable to newspapers or magazines that function. There are 12 daily newspapers: one is published in French (L'orient le jour), one in English (The Daily Star), and 10 in Arabic. It is estimated that daily newspaper sales are 80,000 copies. Television In 2006, there were six private television stations broadcasting, including both terrestrial and satellite stations, in addition to two public stations broadcasting in Arabic (Télé Liban) and in French (Le neuf, which has an agreement with the French TV5 to broadcast its programs). Lebanon also receives hundreds of Arab satellite television channels (around 300), and more than 1,000 coded and "free to air" foreign satellite channels. Some Lebanese regions receive channels from neighboring countries such as Cyprus, Israel, Jordan, Syria, and Turkey that can be viewed with ordinary antennas. Studies reveal that 100 percent of the residential units in the coun- try own a television. More than 48 percent of these units own more than one television. Other studies conducted in 2006 show that the number of residential units with subscriptions to satellite networks averages around 730,000 out of the 850,000 households.4 The satellite channels broadcasting through street cable operators cover households in the various Lebanese regions, where each Lebanese house- hold can be connected to different encrypted television networks in addi- tion to dozens of other channels for a fee ranging between US$7 and US$15 a month; the cost to watch these channels in European and Mediterranean countries would exceed US$150 a month. These street cable operators con- trol which terrestrial and satellite channels they offer to their subscribers, and are thereby able to practice censorship for political or "moral" reasons that they judge appropriate. Lebanon: An Overview of the Media 213 Radios In 2006, four AM and 12 FM private radio stations broadcast news. Two dozen FM stations broadcast music and entertainment programs under Category 2. Studies on radio for the period between 1994 and 2006 show that, on aver- age, a Lebanese person listens to the radio two hours a day. Approximately 60 percent of drivers turn their radios on while driving their vehicles.5 Internet The number of users in Lebanon was around 500,000 in 2005.6 It is the highest user-to-habitant ratio in an Arab country. There are 81 personal computers per 1,000 people, which is the highest ratio in Arab countries as well. No studies reveal the nature and the quality of Internet usage, but a considerable number of users visit political and news Web sites.7 Challenges Facing the Media There are two sets of challenges: external and internal. The external chal- lenges are those in the form of political pressures, religious attempts to impose censorships on social questions, and insufficient financial resources in the advertising market. In addition are fears resulting from assassina- tions and attacks. The internal challenges are autocensorship and financial dependence on politicians or political movements, in addition to unclear job descriptions and work guidelines within institutions. Conclusion If the overall situation is acceptable in Lebanon today compared to what is in place in the region and to what the minimum of professional and techni- cal standards presume, it is important to focus on the steps and measures necessary to develop the media sector. These are as follows: · signing and respecting codes of ethics based on the Munich convention and other relevant texts for journalistic deontology · replacing the ministry of information with an independent and cred- ible media council · ending all forms of censorship and allowing the media council to assess any law violations through the judicial authorities · developingnewacademicprogramsforcontinuoustraining,andorganiz- ing seminars to foster journalistic skills, as well as technical and manage- rial experience within media institutions. 214 Three Countries: Three Stories Finally, due to the tense situation in the country and the deepening divisions, it is urgent to gather media representatives around a table to agree on a professional pact that avoids the circulation of aggressive messages, biased terminology, and unverified facts that worsen the situation and feed the ongoing antagonist political campaigns. Notes 1. More than 20 journalists were targeted or assassinated between 1958 and 1991 in Lebanon. Syrian forces, regional security services, and some Lebanese factions during the civil war were behind the majority of these crimes. In July 2006, Israeli airplanes attacked and destroyed buildings used by the Al-Manar television station and Annour Radio Station. Both of these stations operate legally in Lebanon, and attacks against them were severe violations of laws protecting journalists. The airplanes also targeted and killed a journalist cov- ering the fights in south Lebanon. 2. Annahar opened its opinion pages to Syrian writers, especially opponents, who criticized the baathist regime. Samir Kassir himself wrote many editorials on reforms and democratization in Syria. It is widely believed that these were the reasons behind his assassination. In 2001, he was already harassed and threatened by Lebanese security services loyal to the Syrian regime. 3. The case of MTV was an arbitrary application of the law. While other "politi- cally loyal" television stations covered elections and supported different candidates, this station was closed because it was accused of intervening in political campaigning during a partial legislative election that was won by the opposition candidate. In 2005, the first post-Syrian government cancelled the decision to close the station. MTV now has the right to go on air again, but it is still incapable of doing so for financial and technical reasons. 4. Statistics Lebanon Ltd. 5. "Statistics Lebanon" study on radio station audiences in the country, 2006. 6. www.internetworldstats.com. 7. It should be noted that the dynamism of the private sector detaining media institutions in Lebanon (and the implications of that in terms of political diversity and different political affiliations) does not hide the fact that major newspapers as well as television and radio stations are owned directly or indi- rectly by families, parties, businessmen, and religious institutions. This fact reduces the public participation in funding and monitoring the institutions; in addition, it often protects these groups against legal charges, or transforms them in moments of crises to forums for political propaganda with violent terminology. The Media Environment in Indonesia: Bright Light at the End of a Long Tunnel Bambang Harymurti T he freedom of the press has been at the vanguard of democracy in Indonesia. Unfortunately, the record of its maintenance has not been sterling. Indonesia has been a newly democratic country only since the resignation of President Suharto in 1998, and its journey to full democracy is far from over. Indonesia had tried to become a functioning democracy twice before 1998 and failed. In both cases, freedom of the press was the first political right taken from the people. The absence of such freedom made it much easier for the government to take away other political rights. Hopefully, Indonesia will be more successful in attaining its demo- cratic goals this time. Some major progress has been achieved, primarily in the formation of formal political institutions. However, there have been many setbacks as well, primarily in finding the proper balance between the rights of the majority versus those of the individual. A recent trial of Erwin Amada, the chief editor of Playboy magazine in Indonesia, is an example of such a struggle. Although the magazine's pic- torial contents are tame, even by local standards, a group of Muslim hard- liners filed a criminal charge against its publisher at the South Jakarta Police Station in 2006. The charge claimed that indecent pictures, as defined by article 282 of the Indonesian criminal code--a legacy of the Dutch colonial period--had been published in the magazine. The police and the prosecutor's office yielded to this political pressure and brought Mr. Amada to trial, which received local and international media attention. The court quickly became a battleground between the pro- ponents of freedom of the press and conservative religious groups. The Press Council, an independent body tasked to protect press freedom in Indonesia by the press law, took the position that since Playboy is a press product, the police and prosecutor should apply only the press law in their charges. The press law, made when public euphoria about democracy was still strong 215 216 Three Countries: Three Stories in 1999, does not allow journalists to be jailed for their journalistic work. Instead, the court may impose a fine of up to Rp 500 million (US$55,000) in the event that a breach of the Indonesian Journalists Code of Ethics has occurred; such a breach could include publication of indecent materials. The prosecutor did not heed the Press Council and persisted in charging Mr. Amada with violating article 282 of the penal code. The charge carries a maximum sentence of 32 months in jail. In the year-long trial, attended by many supporters of the hardline groups, the prosecutor requested that the judges give a two-year jail term to the defendant; the hardline supporters denounced this request as too lenient. Ultimately, the judges decided to dismiss the case, ruling that the pros- ecutor should have brought charges under the press law rather than the criminal code. This decision was an important legal landmark for freedom of the press in Indonesia, particularly so since it was issued by the South Jakarta Court, which was known in the past as the "press graveyard" due to its legal judgments against the press. The turnaround in court attitude toward press freedom started at the top. Bagir Manan, the current Chief of the Supreme Court, has made clear to his subordinates the need to preserve freedom of the press in order to maintain the independence of the judiciary. His position was publicly known before he was appointed to his current position. A distinguished law profes- sor from University Pajajaran, Mr. Manan was one of the legal advisors of the Indonesian Press Society when this organization drafted the press law in 1999. In 2006 the Supreme Court overturned the decision of the High Court of Jakarta in the case of Tempo Weekly Newsmagazine. The publication had been accused of criminally defaming Tomy Winata, a powerful business- man known to have close ties with the military and police. The Supreme Court acquitted Mr. Bambang Harymurti, the chief editor of Tempo, who had been given a one-year jail sentence by the lower court. In its landmark decision, the Supreme Court stated that the press law should be applied in cases involving press matters. Advocates of a free press hope that the Tempo case will have a profound impact on the Indonesian media legal environment, similar to the Ameri- can experience with Sullivan vs. the New York Times case in 1960 or Germa- ny's Der Spiegel vs. Minister of Defense Franz-Josef Strauss in 1962. The South Jakarta Court decision in the Playboy case on April 5, 2007, is an encouraging sign that this hope is not just wishful thinking. The Media Environment in Indonesia 217 After Suharto: A Fast Decline, a Slow and Haphazard Recovery Losing freedom is easy, but regaining it again is difficult. The Reporters sans Frontiers annual press freedom index clearly shows the rapid decline and the slow beginnings of press freedom recovery in Indonesia relative to other nations (table 12.1). In 2001 Indonesia had been ranked as the nation in Asia with the greatest degreeoffreedomofthepress;in2002therankingdroppedto57of139coun- tries, probably due to threats from political organizations to media offices. The situation worsened in 2003 (111 of 166 countries), when the speaker of the parliament and President Megawati Sukarnoputri filed criminal charges against Rakyat Merdeka Daily, and again in 2004 (117 of 167 countries), when courts in many parts of Indonesia gave jail sentences to journalists. Change of government at the end of 2004 turned the situation around. Press freedom started to improve gradually. In 2005 the press freedom ranking was 105 of 167 countries, when newly elected president Susilo Bambang Yud- hoyono publicly supported press freedom. In 2006 the Supreme Court issued several decisions in favor of the press, and Indonesia's press freedom index rose to 103 of 168 countries. In 2007 the ranking improved slightly to 100 of 169 countries. Unfortunately, 2007 did not end on a good note. The Supreme Court fined Time Inc. about US$108 million to be paid to former president Suharto, and two journalists were sent to jail on separate criminal libel cases. It is interesting to note that as the condition of press freedom improved with the change of government in October 2004, the score of Transparency TABLE 12.1 Freedom of the Press in Indonesia Countries Year Rank surveyed President 2001 Most free in Asia Asia only B.J Habibie and Abdurrahman Wahid 2002 57 (no 4 in Asia) 139 Abdurrahman Wahid 2003 111 166 Megawati Sukarnoputri 2004 117 167 Megawati Sukarnoputri 2005 105 167 Susilo Bambang Yudhoyono 2006 103 168 Susilo Bambang Yudhoyono 2007 100 169 Susilo Bambang Yudhoyono Source: Reporters sans Frontiers (http://www.rsf.org). 218 Three Countries: Three Stories TABLE 12.2 Global Corruption Perception Index Year 2004 2005 2006 CPI Score 2.0 2.2 2.4 Source: Transparency International (www.transparency.org). International's annual Global Corruption Perception Index also improved. A higher score means less perception of corruption (table 12.2). Guarding Press Freedom: Empowering Self-Regulation The press law in Indonesia tasks the Press Council, an independent body elected by media organizations, to protect the freedom of the press. The council has nine members, each serving a three-year term: three are represen- tatives of journalists' unions, three are representatives of publishers' associa- tions, and three are representatives of the public. Journalists are prevented by an internal regulation from becoming chairman in order to avoid the impres- sion that the Press Council is an institution created to defend journalists. Since its creation in 2000, the Press Council has established a national code of ethics for journalists, organized discussions about press freedom, provided ethical training for journalists all over the country, and medi- ated disputes between the public and the media. President Susilo Bambang Yudhoyono and his ministers have also asked the Press Council to mediate their disputes with the press. The Press Council does not have the power to impose any sanctions, but it does have the legal authority to determine whether a breach of the code of ethics has occurred. It also may make recommendations on the proper way to amend editorial errors. The council stays away from any media dispute if at least one of the parties intends to settle through the court system. Most of the cases handled by the Press Council have been well received by both com- plainants and the involved media, especially the major media (table 12.3). TABLE 12.3 Number of Complaints to the Press Council Year 2003 2004 2005 2006 Complaints 101 153 127 207 Source: Press Council 2007. The Media Environment in Indonesia 219 Some members of the council have provided expert testimony, usually in the defense of freedom of the press cases. Emerging Threat: Concentration of Ownership The resignation of President Suharto in 1998 was followed by radical dereg- ulation of the media sector. License for the right to publish was no longer required. New publishers proliferated, and many did not survive the mar- ket. According to the Press Council, more than 1,300 of these new publica- tions ceased publication within three years. Some continue to do well. One regional newspaper, Jawa Pos, publishes a chain of at least 120 local daily newspapers throughout Indonesia. Concentration of ownership in the media industry is inevitable, espe- cially in the absence of clear regulation of cross-ownership of print and broadcast media. The only regulation to prevent the cartelization of media is the generic antitrust law. It is not very effective for this purpose, however, because real ownership of any radio or television station can be disguised easily. Most of the established private television stations are believed to be still under the control of Suharto's family members and cronies. Despite this suspicion, these television stations enjoy public respect for a very sim- ple reason: their reporters and editors are well-paid professionals compared to hundreds of other, much smaller media whose underpaid journalists often practice "envelope journalism" and even "blackmail journalism" for economic survival. Lack of professionalism in the media is a common com- plaint, although it was also a problem when the media were tightly con- trolled by the state. Control of the market is a more serious problem. Many media observ- ers have voiced their apprehension about the increasing concentration of media ownership in Indonesia. This concentration is a legitimate concern even though its political implications in the real world have not been clearly identified (table 12.4). The Press Council is currently looking at this issue. TABLE 12.4 Number of Publications in Indonesia Year Dailies Weeklies Magazines Bulletins Total 2001 305 132 127 2 566 2006 284 327 237 3 851 Source: Press Council. 220 Three Countries: Three Stories The Future: Light at the End of a Long Tunnel Indonesia is a large country consisting of more than 17,000 islands; more than 3,000 of these are inhabited. It is the fourth most populous nation in the world, with almost 240 million people. Although Indonesia is a secu- lar republic, it is the home of the largest Muslim community in the world. Islam is the stated religion observed by approximately 87 percent of its population. More than half of the population lives on Java Island, whose size is only 7 percent of the country land area. About 12 million live in the capi- tal city of Jakarta, which is similar in size to Singapore. In the western tip of the nation is the island of Papua, which is divided between Indonesia and Papua New Guinea. The Indonesian Papua, more than three times the size of Java, has only a little over 4 million inhabitants; half is composed of indigenous people spread over more than 250 tribes, each with its own language. A few of these tribes still live in a Stone Age culture. The media in this large and diverse nation are almost as diverse as its population. Jakarta is the home of 10 national private television sta- tions, one state television station, and one local television station; many provinces have no local station. Only large cities like Surabaya, Medan, Solo, Yogya, Bandung, Semarang, and Denpasar have their own local sta- tions. Most are newly established, and their economic future is still in question. As Indonesia slowly moves toward a more democratic system, interest in the media is growing, especially since the introduction of a new law that obliges all provinces, regencies, and cities to elect their leaders by direct election. More than a thousand private radio stations are now in service throughout the country, and new ones keep emerging. A new regulation requiring all television stations to change to digital by 2009 will open hun- dreds of new local television channels as one analog channel can be con- verted to at least six new digital channels. The sound of democracy will get louder, and the role of the media in influencing public policies in Indonesia will increase. The role of the media, however, can be a double-edged sword. It is not by coincidence that television, radio stations, and media offices are one of the first targets of coup d'etats in any country. It is imperative that basic infrastructure for the media be provided to make sure that the fourth pillar of democracy will remain the true representative of the whole population and not the mouthpiece of a few elites. The Media Environment in Indonesia 221 The Press Council and the Indonesia Broadcasting Commission, both independent public institutions, are joining forces to encourage the gov- ernment to create an enabling environment in which the media can play its role as a provider of credible and relevant information to the public in a timely fashion. Judging from the experiences of other more developed countries, this will be a long and arduous journey. However, no matter how long the pas- sage through the tunnel may take, it is a worthwhile endeavor. There is the bright light of functioning democracy at the end of it. Index A Appointed officials Access to media amount of press coverage of, versus conditions for media influence, 3 elected officials, 62­64, 66­67 content factors in media influence biased coverage of elected officials and, 18­19 versus, 71­73 determinants of, 3 coverage of, in election years, 64, 67, in developing countries, 3 68­69, 71 effects of national media access, 33 coverage of appointing politician, 64, Internet as news outlet, 43 67­68, 69 mechanism of media influence on coverage of appointments process, 64 policy outcomes, 17, 18, 19­22 coverage of institutional performance research opportunities, 18 and, 64­65 significance of, in policy outcomes, duration of press coverage of, versus 27­28 elected officials, 69­70 socioeconomic status and, 28 media coverage issues, 60, 61­62 U.S. market characteristics, 156 policy choices of, versus elected voter turnout and, 20, 50 officials, 60­61 Accountability of politicians political agency model, 60 game theory model, 60 Arke, Brian, 129 intensity of political coverage and, 28 public access to information for, 61 B public access to media and, 18, 19­20 Babangida, Ibrahim, 203 Administrative Procedures Act, 167 Bad news vs. good news, 71, 72­73 Advertising Biased coverage children's programming restrictions, benefits of foreign media presence, 159­160 147­149 influence on consumer behavior, 50 as catering to prior biases of influence on media content, consumers, 7­8, 84, 108­109, 127­128, 129 115­119 market conditions as factor in media concerns about, 79 bias, 8­9, 127, 130­134, 135 consumer assessments of accuracy of, minority targeting, 50 113­115 Nigerian market, 199­200 cost of information and, 5 political, regulation of, 160 economic significance, 139 political, voter behavior and, 84­85 effective persuasiveness, 96­101 Agency theory, 59­60, 140 effects of political advertising, 84­85 Agenda-setting function of media, 17, 59 of elected officials vs. appointed Al Jazeera, 7, 84, 108, 111, 112, 113, officials, 6, 71­73 114, 116 good news vs. bad news, 71, 72­73 Amada, E., 215­216 government dominance of Nigerian Antitrust law, 46, 150, 165­166, 201 media, 201­202 223 224 Index influence of information suppliers, 9, public attitudes and beliefs about, 142­144 122­123 in Lebanese media, 209­210 regulation of media ownership, 10 market conditions as factor in, 8­9, risk of biased coverage, 128 127­128 socioeconomic conditions favoring, 8 measurement, 109 trade-offs between biased coverage mechanism for influencing voter and audience share, 129 behavior, 81­82, 95 Chidiac, May, 210 media preferences of consumers, Children's programming, regulation of, 111­112 159­160 motivation of consumers to prefer, 108 Children's Television Act, 159 motivation of media owners, 145­147 CNN, 7, 84, 111, 112, 113, 114, 116 nonrational persuasion model, 82, Cohen, B., 59 91­94 Common law, 183 persistence of effects, 91­94 Communications Act (1943), 153, 162, political outcomes, 6­7, 79­80, 168, 169 139­140, 141­142 Consumer bias production cost factors, 4 biased coverage as market response public perception of extent of, 79 to, 7­8, 84, 108­109, 115­119 rational learning model of response implications for media content, to, 81­82, 91 119­120 research needs, 29, 102 media preference of consumers and, research on effects on political 110­111 behavior, 82­86 motivation to seek confirmatory sacrifice of audience share for, 129 media, 108 socioeconomic status of consumers perceptions of media quality and, 5 influenced by, 113­115, 119 special interest capture of media, Content of programming 8, 121 access to media and, in determining through omission of information, 6 media influence, 18­19 See also Fox News advertisers' influence on, 127­128 Brazil, 18, 21 capacity to influence policy, 24­27, Britain, 11 28­29 Bush, George W., 50 cost of information as factor in, 5­6 coverage of elected vs. appointed C officials, 6, 60, 61­73 Cable television regulation, 168­169, determinants of, 4, 7, 107, 109, 170, 212 115­120, 127­128 Campaign promises, 18 disaster coverage, 24­27, 28­29 Capture of media by special interest effects of imperfect information, definition, 121 139­140 economic inequality and risk of, 123 entertainment vs. news, 4 goals of, 8 Fairness Doctrine, 160­161 for influencing perceptions of wealth goals of government regulation of distribution, 122­127 media, 10 market conditions as factor in, 8­9, good news vs. bad news, 71, 72­73 127­128, 130­134, 135 importance of media diversity, 147 political capture, 129­130 influence of information suppliers, 9, power of information suppliers, 9, 142­145 142­144 influence of media owner, 4, 109, 121 Index 225 local news vs. national news, 4, 28 role of media, 59­60, 121, 139, market characteristics and, 5, 8­9, 28, 140­141, 177 49, 127­128 Developing countries measurement, 109 access to media in, 3 mechanism of media influence on effect of local conditions on media policy outcomes, 17­18 behavior and outcomes, 12 media freedom and, 119 lessons from U.S. nineteenth century media self-censorship, 128 media, 134 motivation for accuracy, 119­120 Direct broadcast satellite service, in Muslim countries, 115­119 169­170 omission of relevant information, 6 Disaster relief, 24­27, 28­29 political coverage effects on political Diversity of media outcomes, 22­24 definition, 145, 154, 155 politicians' influence on, 25 foreign media presence, 147­149 pressure for incumbent politicians, goals of government regulation, 59­60 10­11, 146, 150, 151, 154­155, regulation in U.S., 159­161, 163, 170 184, 192­193 socioeconomic status of consumers measurement of, 150 and, 5, 29 outcomes of media regulation in See also Biased coverage; Local news U.S., 163 coverage ownership regulation and, 146­147 Corporate­media relationship, 9 in preventing information control of information access, 144­145 distortions, 145 motivation to distort information, public media and, 150 145­147 rationale, 151, 154, 163 Corruption strategies for promoting, 150 agency theory, 140 incompetence versus, 142 E in Indonesia, 217­218 Educational attainment information imperfections in, 140, 141 local political awareness and, 43­44 media ownership and, 22 political participation and, 36, political motivation, 142 38­40, 41 press freedom and, 22 Elections, U.S. public access to media and, 21 1996 (presidential), 89 in withholding and controlling 2000 (congressional), 94­95, 96 information, 142­144 2000 (presidential), 89, 90­91, 95 Cost of information, 5­6 2004 (presidential), 50, 91 Court of Justice of the European Electric utilities, 60­61 Communities, 189 European Union future of media regulation, 194 D goals of government regulation of Democratic governance media, 10 conceptualization of role of horizontal regulation effects, 193­194 government, 179­181 media regulation, 178, 188­194 economic inequality and freedom of pluralism goals for media, 192­193 media, 123­127 size of media markets, 28 importance of media diversity, 147 See also France, media regulation in Indonesia, 215, 220 in; United Kingdom, media information imperfections and, regulation in 139­140, 141, 150 Exxon, 144 226 Index F quality of content and, 119 Fairness Doctrine, 160­161 rationale, 177 Federal Communications Commission, role of judiciary in protecting, 197 153, 154, 156, 157­158, 159, in Russia, 131­132 160­161, 162, 163, 164­166, See also Government regulation of 167, 169 media Financial Times, 148­149 First Amendment, 159 G Fox News, 6­7 Game theory, 60 effective persuasiveness, 96­101 General Motors, 129 influence on voting behavior, 80­81, Government­media relations 90­96 influence of information suppliers, 9, origins and development, 80, 87­88 142­144 political orientation, 80, 88 international differences, 12 prior biases of consumers, 84 Nigeria's state-owned media, 197, research methodology for studying 199, 200 influence of, 86­87, 88­89, political corruption and, 22 97­98, 102­104 rationale for media monitoring of France, media regulation in, 10­11 government, 140­141 audiovisual broadcast media, See also Government regulation of 186­187, 188 media; Government spending, conceptual basis, 178­179, media influence on 180­182, 183 Government regulation of media historical development, 179 challenges in assessing impact of, independence of media and, 184 163­164 media ownership transparency, 183 common law principles in, 183 objectives, 181­182 conceptual basis, 177­179 primary obligations of media, conceptualization of role of 183­185 government and, 179­181 print media, 183­185 consumer influence on media state aid to media and, 184­185 content, 8­9 Freedom of Information Act, 9, 141, content regulation, 159­161, 163 143­144 determinants of public access to Freedom of media media, 3 Constitutional protections, 159 European governance structure, consumer bias and, 113, 114 188­194 in democratic societies, 140­141 foreign ownership rules, 148, 162 determinants of, 13 future prospects, 171 economic inequality and, 123­127 in Lebanon, 209­211 effects on governance, 22 localism as goal of, 10, 155­156 historical development, 179, 180 media self-censorship, 128, 210 in Indonesia, 215­219 in Muslim countries, 107 international differences, 12­13 in Nigeria, 12, 202­206 Lebanese law, 210 objectives, 10­11, 154­156 market conditions as factor in, 121, ownership restrictions, 10, 161­162, 130­134, 135 163­166, 169 in Muslim countries, 107 ownership transparency, 11, 183 policy outcomes, 135 private sector­public sector preconditions for, 108 relationships and, 197 Index 227 to promote competition, 154 intensity of coverage and, 22­24 to promote diversity, 10­11, 146, 150, international differences, 11­13 151, 154­155, 163 local issues, 4 public service broadcasting market factors in, 1, 49, 91 requirements, 150, 187 mechanisms of, 1, 2­3, 17­18, state aid to media and, 184­185 19­22, 121 state-owned media, 22, 116­117, 119, media content factors, 17­18, 24­27, 121, 128 28­29 technical regulation, 157­159 minority voter behavior, 5 technology development and, public access to media and, 3, 17, 18, 170­171 19­22, 27­28 U.S. structure, 153­154, 166­170 research challenges, 22 See also France, media regulation research methodologies, 101­102 in; United Kingdom, media research needs, 102 regulation in significance of national media Government spending, media influence distribution, 33­34, 44, 46 on, 107 See also Biased coverage; Capture extent of political coverage and, 24 of media by special interest; mechanism of influence, 3, 19­21 Government spending, media significance of public access to media, influence on; Voter behavior 27­28 Internet, 171 Grasso, Dick, 148­149 access, 156 in Lebanon, 211, 213 H as news outlet, 43, 46 Hamilton, J. T., 121 Iran, 116 Harvard University, 149 Harymurti, Bambang, 216 J Herfindahl-Hirschmann Index, 166 Jordan, 116 Jordan, Boris, 130 I Incumbent politicians K biased coverage of, 6, 72­73 Kassir, Samir, 210 pressure of media coverage, 59­60 Kerry, John, 50 India, 21, 28, 60 Indonesia, 12­13, 21, 197 L corruption index, 217­218 Lebanon, 12, 197 freedom of media in, 215­219 assassination of journalists, 210 future prospects, 220­221 audiovisual media, 212­213 media ownership in, 219 challenges for media, 213 media self-regulation in, 218­219 future prospects, 213­214 political evolution, 215 historical evolution of politics and population, 220 media, 209­210 Influence of media on political and media laws and regulations, 210­211 policy outcomes media ownership, 211 combined effects of access and population, 212 content, 18­19 print media, 212 concerns about, 79 training for journalists, 211 generalizability of media Lesin, Michail, 131 research, 109 Lippmann, W., 129 228 Index Local news coverage significance of, in media content, 49 access to national media, political See also Minorities, media content participation and, 4, 33, 35, targeted to 38­40, 41, 44 Mermin, J., 25 determinants of media decisions, 4 Minorities, media content targeted to, 5 effect of market characteristics, 28, 49 black-targeted media, 5, 52 goals of government regulation, 10, influence of advertising, 50 155­156 linkage to voter behavior, 50 influence of consumer bias on size of minority and, 5, 49, 50 content, 116­119 Spanish-language television local media response to national programming, 5, 51­55 media penetration, 44 spending in 2004 presidential minority-targeted media effects on election, 50 voter turnout, 52­55 voter behavior research methodology, political participation and, 4 50­51 Spanish-language television voter turnout outcomes and, 5, 49, programming, 51, 52 52­57 technology development as factor in, Multichannel video program 33, 170­171 distribution, 168­170, 187­188 Murdoch, Rupert, 80, 87 M Muslim countries Manan, Bagir, 216 determinants of media content, Market characteristics 115­119 assessment methodology, 36 independence of media in, 107 black-targeted media, 52 media ownership, 116­117 competitiveness of media market, media preferences of consumers, 7, 145, 165­166, 170­171 84, 108, 110­111 current U.S., 156 survey of public attitudes and effects of biased coverage and, 91 beliefs, 110 generalizability of media research, 109 local effects of national media N penetration, 46 New York Times, national distribution of media capture risk, 8­9, 127­128, consumer characteristics, 34­35, 43 130­134, 135 effect on local news consumption, 35 media characteristics influenced by, 35 growth of, 33­34, 35 media freedom and, 121, 135 influence of, in presidential vs. political participation and, 35 nonpresidential elections, 38, 41 quantification, 165­166 local political coverage, 36 size. See Market size methodology for assessing political Spanish-language television influence of, 35­44 programming, 51­52 political participation affected by, 4, structural regulation in U.S., 10, 35, 38­40, 41, 44 161­162, 163­167 political significance, 44 Market size Nigeria, media in, 12 evolution of media bias, 8 access to information, 206­207 media capture risk, 8 advertising market, 199­200 media content targeted to minorities, 5 audiovisual media, 200­201 media influence on policy outcomes economic status, 199 and, 28 future prospects, 207 minority access to media affected government dominance, 201­202 by, 50 media laws and regulation, 202­206 Index 229 ownership structure, 197, 198, 199, historical evolution in Lebanon, 200, 201 209­210 print media, 198­200 impact of biased coverage, 6­7 significance of, 198 imperfect information and, 139, 141­142 O mechanism of media influence, 1, 2, Obscenity rules, 159 17­18, 19­22, 130 Olympic games, 25­26 media capture by politicians, Ownership of media 129­130 corruption and, 22 media influence on minority voter foreign ownership, 147­149, 162, 184 behavior, 5 importance of media diversity, 147 motivation to distort information, 142 in Indonesia, 219 pacifying effect of television, 21 influence on content, 4, 109, 121 political agency model, 59­60 in Lebanon, 211 political coverage effects on political motivation to distort information, outcomes, 22­24 145­147 politicians' influence on media in Muslim countries, 116­117 content, 25 in Nigeria, 197, 198, 199, 200, 201 role of media, 59, 79, 121, 153 pluralism goals of European media television broadcast regulations, 10, regulation and, 184, 192­193 160­161 in Russia, 147 See also Appointed officials; transparency requirements in France, Government spending, media 11, 183 influence on; Incumbent U.S. regulation, 158, 161­162, politicians; Voter behavior; Voter 163­166, 169 turnout Public media, 9, 150, 167­168 P Public perception and understanding Policy development and for assessment of regulator implementation performance, 61 influence of information suppliers, 9, consumer assessment of media 142­144 quality, 113­115, 119 mechanism of media influence, 1, coverage of local issues and, 4 2­3, 17­18, 19­22 effective persuasiveness of biased performance of elected vs. appointed media, 96­101 officials, 60­61 effects of national media presence significance of public access to media, on local news awareness, 27­28 43­44 Political functioning on extent of biased media agenda-setting function of media, coverage, 79 17, 59 goals of media capture, 8 bad news coverage vs. good news government spending and, 3 coverage, 72­73 local news consumption affected by campaign promises, 18 national media presence, 33 coverage of local news and, 4 media capture to influence economic inequality and risk of perceptions of wealth media capture, 123 distribution, 122­127 effects of biased coverage, 79­80 research findings on effects of biased effects of national media presence coverage, 82­86 on local political awareness, survey of Muslim countries, 110 43­44 See also Voter behavior; Voter turnout 230 Index Public service programming, 11, 149 Telemundo, 51­52, 156 in France, 187 Television in Lebanon, 211 British regulation, 187­188 in United Kingdom, 187­188 European media governance, 190­191 French regulation, 186­187, 188 R in Lebanon, 211, 212 Race/ethnicity. See Minorities, media multichannel video program content targeted to distribution, 168­170, 187­188 Reputation of public figures in Nigeria, 200 mechanism of media influence on on-demand services, 191 policy outcomes, 2­3 political passivity and, 2121 Role(s) of media program rating system, 160 contradictions in, 177 significance of market size, 28 in democracies, 59­60, 121, 139, Spanish-language programming, 140­141, 153, 177 51­52 Roosevelt, Franklin D., 19 tailoring content to political beliefs of Russia, 128, 130, 131­132, 147, 149 consumers, 109 U.S. regulatory system, 157­162, S 167­170 Sarbanes-Oxley Act, 144­145 See also Fox News School performance, media coverage Thailand, 18 of, 5, 64­65, 71, 73 Shleifer, Andre, 149 U Simpson, O. J., 26 Uganda, 18­19, 28 Socioeconomic status of consumers United Kingdom, media regulation in access to media and, 28 audiovisual broadcast media, 187­188 economic inequality and freedom of common law principles in, 183 media, 123­127 conceptual basis, 178­179, 180 media capture risk, 8 ethical and technical standards, 182 media capture to influence historical development, 179 perceptions of wealth industry self-regulation, 182­183, 185 distribution, 122­123 newspaper styles, 185­186 media content and, 5, 29 print media, 185­186 media freedom and, 8 United States Somalia, 25 attitudes toward foreign media, 107 Sukarnoputri, Megawati, 217 disaster relief, media influence on, Sunshine Act, 167 24­27, 28­29 Sweden, 21 generalizability of media research, 109 goals of government regulation of T media, 10, 154­156 Technology of media production and impact of media regulation, 163­166 distribution media market characteristics, 156 European media governance, 191 media regulation system, 153­154, influence on content, 4 157­162, 166­170 Internet as news outlet, 43 Muslim consumer attitudes toward national media distribution range, 33 and media preferences, 111­112, regulatory issues, 170­171 113­114 significance of, 34 nineteenth century media, 133­134 Telecommunications Act (1996), 160, political coverage effects on political 162, 164 outcomes, 22­24 Index 231 public media, 167­168 persuasion rate of biased media, survey of attitudes toward, in Muslim 96­101 countries, 110 political coverage effects on political Univision, 51­52 outcomes, 22­24 research on media bias effects, 82, V 85­86 V-Chip technology, 160 response to political advertising, 84­85 Voter behavior See also Voter turnout cost-benefit analysis in, 50 Voter turnout effects of slanted media coverage, access to national media and, 34, 35, 80­81, 95­96 38­40, 41, 44 in election or appointment of competitiveness of campaign and, 41 regulators, 60­61 effects of media bias, 80­81 Fox News access and, 90­96 historical patterns, 36 influence of media, 79 media access and, 20, 50 information for assessment of elected minority-targeted media presence official, 60­61 and, 49, 52­57 length of residency and, 43 voter characteristics and, 36, 38­40, 41 local effects of national media See also Voter behavior presence, 46 mechanism of biased media coverage W effects, 81­82 Wall Street Journal, 148 media influence on minority Winata, Tomy, 216 voters, 5 media market characteristics and, 35 Y methodology for assessing national Yudhoyono, Susilo Bambang, 217, 218 media influence on, 35­44 Yushkiavitshyus, Henrikas, 131 E C O - A U D I T Environmental Benefits Statement The World Bank is committed to preserving Saved: endangered forests and natural resources. · 12 trees The Office of the Publisher has chosen to · 565 lbs. of solid waste print Information and Public Choice: From Media Markets to Policy Making on 60-pound · 4,397 gallons of waste water Roland Opaque, a recycled paper made with · 1,059 lbs. of net greenhouse 30 percent post-consumer waste, in accor- gases dance with the recommended standards for · 8 million BTUs of total paper usage set by the Green Press Initiative, energy a nonprofit program supporting publishers in using fiber that is not sourced from endan- gered forests. For more information, visit www.greenpressinitiative.org. T HE ABILITY OF THE MEDIA TO AFFECT OUTCOMES IN ECONOMIC AND POLITICAL markets has been well documented. News reporting and advertising influence consumer behavior in goods and services markets by revealing (or selectively revealing) information about a product, acting as agenda setters to influence consumer demand, or enhancing competition in markets by alerting consumers to substitutes. In political markets, they can affect behavior by informing voters about a politician's views or actions, enlightening citizens to outcomes of public policy, or taking a stance on political, social, or economic issues. For businesses, households, and most others, the media is the main source of information on public policy choices and current social and economic conditions. As a result, what news the media chooses to gather, analyze, and disseminate--and the slant they choose to put on what they report--is of consequence. Information and Public Choice addresses the factors that affect the content and reach of news coverage as well as its impact on public policy. The book addresses both market constraints that affect media--particularly news content--and the impact that news reporting has on economic and political choices. The authors examine a range of issues, including bias or slant in media reporting, the impact of market and nonmarket factors on news reporting, and the role of government regulation of the media sector in developing countries. The studies in this volume provide new evidence and a good summary of previous research on the power of the media. An invaluable guide for those concerned about the impact of media on economic and political outcomes, Information and Public Choice draws attention to an under-researched yet important area of economics. "Economists have long ignored the role played by media in shaping public policy. Not anymore. This books brings together some of the most exciting research in this area, highlighting the potential distortions created by the way the media market works. It is a must read for policy makers and concerned citizens alike." -- LUIGI ZINGALES, Professor, University of Chicago ISBN 978-0-8213-7515-0 SKU 17515