77349 the world bank economic review, vol. 15, no. 1 165–176 New Tools in Comparative Political Economy: The Database of Political Institutions Thorsten Beck, George Clarke, Alberto Groff, Philip Keefer, and Patrick Walsh This article introduces a large new cross-country database, the Database of Political Institutions. It covers 177 countries over 21 years, 1975–95. The article presents the intuition, construction, and definitions of the different variables. Among the novel variables introduced are several measures of checks and balances, tenure and stability, identification of party affiliation with government or opposition, and fragmentation of opposition and government parties in the legislature. Research into the institutional and political roots of government decisions has accelerated rapidly in the past 20 or more years. However, cross-country em- pirical work has been handicapped by a lack of detailed data on countries’ po- litical and institutional characteristics and on how they change over time. There- fore, comparative political economy has been restricted to small-sample case studies, which are important in their own right but impose unavoidable limits on researchers seeking to generalize their results beyond the case study sample. The lack of data has been a serious obstacle to better understanding of nu- merous areas of political economy, particularly the political conditions under which governments choose policies that promote rather than retard economic development. We are still far away from the answers to two fundamental ques- tions. First, which political institutions are most conducive to development and reform? Second, under what conditions do such institutions emerge? Unfortu- nately, cross-country data that could help to answer these questions, data that would provide a disaggregated picture of a country’s political institutions and operations, have been scarce. In this article we introduce a new data set that fills many of these data gaps, the Database of Political Institutions (dpi). It has been compiled in the Development Research Group of the World Bank, but it rests on an intellectual foundation laid by a large body of political economy research. The dpi contains 108 variables for 177 countries over the years 1975–95 (see the appendix). The variables provide details about elections, electoral rules, Thorsten Beck, George Clarke, Philip Keefer (e-mail pkeefer@worldbank.org), and Patrick Walsh are with the Development Research Group at the World Bank; Alberto Groff is with the Federal Depart- ment of Foreign Affairs of the Government of Switzerland. The authors are grateful for the helpful comments of two anonymous referees. © 2001 The International Bank for Reconstruction and Development / THE WORLD BANK 165 166 the world bank economic review, vol. 15, no. 1 type of political system, party composition of the opposition and government coalitions, and the extent of military influence on government. The dpi also contains a number of new variables compiled from the raw data, including origi- nal measures of checks and balances and political stability. In contrast to other databases, these variables are nearly all objective and their construction is en- tirely transparent. They are also disaggregated, allowing researchers to get away from such useful but broad indicators as whether countries allow elections, whether elections are “free,� or whether the executive is “constrained.� In- stead, the dpi allows researchers to use precise and concrete institutional fea- tures of countries and to combine them in ways appropriate for theoretical analysis. The remainder of the article and the appendix provide details on sources and collection methodology. The database is described in more detail in Beck and others (2000).1 I. The Political System and Electoral Competitiveness Two fundamental characteristics of a political system are the relationship of the executive and legislative branches and the competitiveness of elections of the political actors who occupy these branches. The dpi has information on both. With regard to the first, the variable SYSTEM was created to capture whether coun- tries are presidential, assembly-elected presidential, or parliamentary (see the chief executive variables in the appendix). For countries with both a president and a prime minister, the decision whether to call the system presidential or parlia- mentary is based on the power that the president has relative to the prime min- ister. Our criteria were taken, with some modification, from the literature (for example, Shugart and Carey 1992).2 However, as with the other variables in this database, the criteria on which we base this decision are stated explicitly, allow- ing researchers to determine more easily the adequacy of variables to their test- ing requirements. The database also contains two indexes that characterize the competitiveness of elections in countries, one for executive elections and one for legislative elec- tions. The core of the two indexes is the number of parties that could and did compete in the last election. Building on work by Ferree and Singh (1999), we scale countries as follows: 1. No executive/legislature 2. Unelected executive/legislature 3. Elected, one candidate 4. One party, multiple candidates 1. The database will be put on a Web site. Until then, the extended working paper, codebook, and database can be obtained by sending e-mail to Paulina Sintim-Aboagye at psintimaboagye@worldbank.org. 2. For example, we place less weight on the nonlegislative powers of the president, such as the for- mal right to choose cabinet members. Beck, Clarke, Groff, Keefer, and Walsh 167 5. Multiple parties are legal, but only one won seats (because other parties did not exist, compete, or win seats) 6. Multiple parties competed and won seats (but one party won 75 percent or more of the seats) 7. The largest party received less than 75 percent of the seats. Three other variables supplement these indexes. The first and second vari- ables indicate whether the chief executive and the minister of defense, respec- tively, are military officers. The third records whether our sources mention voting irregularities or candidate intimidation that is serious enough to affect electoral outcomes. Nonconstitutional irregularities are also recorded as present when important parties boycott elections or the election results. II. Preferences and Parties For most political economy questions, the main concern is not only the struc- ture of political decisionmaking but also the policy preferences or ideological leanings of decisionmakers. For many countries of the world, however, such information is not available, not germane, or not a reliable indicator of the pref- erences of decisionmakers. A starting point for any analysis that requires the preferences of the policy- makers is the party identification of the decisionmakers. Our data set identifies the party of the prime minister and/or president, the three largest parties in the government coalition, and the largest party in the opposition. In addition, our sources provide useful information for some countries about the policy prefer- ences of key decisionmakers on five dimensions: economic, nationalistic, regional, rural, and religious. The first dimension refers to preferences regarding more or less state control of the economy—the standard left–right scale. The other di- mensions—reported as zero/one dummies—do not necessarily correlate with each other or with the economic dimension. Coverage is far from exhaustive, unfor- tunately, but constitutes the most comprehensive characterization of parties and decisionmakers that could be extracted from the sources we used. III. Tenure, Turnover, and the Share of the Popular Vote of the Chief Executive and the Chief Executive’s Party A key hindrance to development is the tendency of many political leaders to make opportunistic decisions that entail long-run costs to society that outweigh short- run benefits. Leaders with shorter horizons in office are more likely to find such choices to be worthwhile, and there is evidence suggesting that this effect is sig- nificant, at least in less democratic countries (Clague and others 1996, Alesina and others 1996). The dpi presents indicators of the turnover and tenure of gov- ernments, and it provides opportunities to circumvent several ambiguities that arise when using existing data. 168 the world bank economic review, vol. 15, no. 1 First, this data set presents indicators of whether there are constitutional re- strictions on executive terms and whether sitting executives can be reelected. Second, executives with more popular support can be expected to have a longer horizon. We therefore record the share of votes that presidents received in their last election and the share of votes of the prime minister or president’s party. Third, it would be expected that stronger parties would rein in party members who happen to have shorter horizons. The dpi therefore reports the number of years that the party of the executive has been in office (independent of the par- ticular executive who has represented the party) and how long the party has existed under its current name. Finally, the dpi records whether parties are in the government coalition and whether the party of the executive controls the lower house. Based on this infor- mation, the data set contains two new variables on government stability—STABS and STABNS—that capture the extent of turnover of a government’s key decision- makers in any year. The two variables are calculated by dividing the number of exits between year t and year t + 1 by the total number of veto players in year t. Veto players are defined as the president and the largest party in the legislature for a presidential system and as the prime minister and the parties in the govern- ment coalition for a parliamentary system. The dpi also has several variables that provide different perspectives on ten- ure. It includes the tenure of the veto players who have been in office the longest and the shortest periods of time, assuming multiple veto players. It also includes the tenure of the political system—the length of time executive elections have been competitive (where “competitive� means at least a six for the executive index of electoral competitiveness). If elections have not been competitive, the vari- able indicates how long the executive has been in office. IV. The Legislature The role of the legislature in the political process is influenced not only by consti- tutional restrictions on the scope of its initiative and veto authority but also by its composition and the relative strength of government and opposition coalitions. The data set introduced here allows for the calculation of the traditional measure of legislative fragmentation. Our sources also allowed us to identify which parties were inside and outside of the governing coalition. We created variables that record separately the fragmentation of the government coalition and the opposition. The correlation of government and opposition fragmentation measures is only 0.27 (for all countries and all years), suggesting that average fragmentation for the entire legislature is likely to be misleading for many purposes. V. Electoral Rules The incentives of political decisionmakers and their ability to act independently of each other depend in large measure on the electoral rules under which they Beck, Clarke, Groff, Keefer, and Walsh 169 operate. Among other effects, electoral rules influence whether candidates have an incentive to pursue individual or party interests, whether small and single- interest parties have an incentive to coalesce or to remain independent, and whether party leaders exercise significant or weak influence on party candidates (see Cox and McCubbins 2001). One electoral rule that affects responsiveness to narrow interests is the mean district magnitude (given as MDMH for the House of Representatives and MDMS for the Senate in the section on electoral rules in the appendix). The database defines MDM as the average number of representatives elected by each electoral district in a country. A second electoral characteristic that the data set captures is whether the districts of the upper house are states or provinces. Voting rules are a third aspect of electoral rules. The dpi records whether leg- islators are elected using majority or proportional representation systems. For proportional systems, the data set records whether there is a threshold that par- ties have to overcome to gain any seats in the legislature. And it records whether the D’Hondt system, which is less favorable to small parties, is used to calculate the number of seats that each party receives. The dpi also contains a variable that records whether the country has an open or closed-list electoral system. The dpi classifies a system as closed if voters cannot express preferences within a list— that is, if the position on a party’s candidate list combined with the number of votes received by the party is the sole determinant of whether a candidate is elected. The dpi provides some (albeit scant) information on the candidate selection process. Where possible, the data set includes a variable that reflects whether selection takes place on the national level, the subnational level, or by primary election. Finally, the dpi records the month when presidential or parliamentary elections were held. VI. Checks and Balances A key element in the description of any political system is the number of decision- makers whose agreement is necessary before policies can be changed. The dpi significantly extends one existing measure of checks and balances and introduces new measures and the possibility of constructing others. Building on work by Roubini and Sachs (1989), we constructed the index of political cohesion, which records whether the same or different parties control the executive branch and the legislature in presidential systems. For parliamen- tary systems, the index records whether there is a minority government or a coalition of one, two, three, or more parties controlling the government. The index of political cohesion does not distinguish countries according to the effec- tiveness of electoral checks on government decisionmakers or according to elec- toral rules that influence party control over members. To adjust for these cir- cumstances, the dpi includes two new variables, CHECKS1 and CHECKS2 (see the section on stability and checks and balances in the appendix). 170 the world bank economic review, vol. 15, no. 1 These variables count the number of veto players in a political system, ad- justing for whether these veto players are independent of each other, as deter- mined by the level of electoral competitiveness in a system, their respective party affiliations, and the electoral rules. If the legislative index of electoral competi- tiveness is less than four, the two measures are set equal to one because, re- gardless of the formal constitutional arrangements in a system, where legisla- tive elections are uncompetitive, constitutional checks on officials are unlikely to be binding. For all other countries, the variable CHECKS1 is increased by the number of veto players in the system.3 In presidential systems, it is increased by one for the president and one for each legislative chamber. However, if elec- tions are conducted under closed-list rules and the president’s party is the largest government party in a particular chamber, then the dpi assumes that the presi- dent exercises substantial control over the chamber and it is not counted as a check. For parliamentary systems, CHECKS1 is increased by one for the prime minister and by one for each party in the government coalition, including the prime minister’s own party. However, if it is a closed-list system and the prime minister’s party is the largest in the government coalition, then this sum is re- duced by one. The extent to which one political decisionmaker might act as a check on another depends in part on the similarity of their policy preferences. To take into account the policy orientation of parties, we created a second checks-and- balances variable, CHECKS2. CHECKS2 has the value of CHECKS1 plus one for every veto player (defined as the largest government party in the legislative chamber in presidential systems and as all coalition members in parliamentary systems) whose orientation (left, right, or center) is closer to the opposition’s orientation than to the average of the rest of the government. The opposition’s orientation is given by the orientation of the largest opposition party. VII. Federalism Subnational political structure affects national-level policymaking in numerous ways. First, subnational units may have veto power over national-level policy decisions. Second, they may exert pressure for greater (or at least different) lev- els of redistribution than would otherwise be the case. Third, subnational units may affect the cohesiveness of national parties, which struggle for positions in both subnational and national jurisdictions, although the distribution of voter preferences in the two could be quite distinct. The dpi contains several variables that capture the extent of federalism in a country’s political structure. One indicates whether the country has contiguous autonomous or self-governing regions. A second dimension of information on 3. The values of CHECKS1 and CHECKS2 are nearly invariant to the use of the executive or legislative index of electoral competitiveness. Beck, Clarke, Groff, Keefer, and Walsh 171 subnational governments is whether the local executive and legislature are lo- cally elected. This variable is zero if neither is locally elected, one if one of the two is locally elected, and two if both are directly and locally elected. We also investigated, with limited success, whether municipal governments were locally elected. The third dimension on which we attempted to collect information is the jurisdiction of local governments. We asked whether the states or provinces had authority over taxing, spending, or legislating. VIII. Sources and Data Collection There were two main sources of data for this project, each used for a different time period. For 1975–84, we consulted The Europa Year Book (various years). For 1985–95, we used Banks’s Political Handbook of the World (various years). We used The Europa Year Book for the latter time period to plug topical gaps in coverage, mostly for electoral rule and constitutional framework variables. The consistency of data between these two periods was extensively checked. We took information on electoral rules mostly from the Web site of the International Par- liamentary Union (parline, www.ipu.org/parline-e.parlinesearch.asp). We ob- tained information on party orientation from a European Web site maintained by Agora Telematica (www.agora.stm.it/elections/parties.htm). We also cross-checked party orientation against information in Political Parties of Africa and the Middle East: A Reference Guide (1993) and Political Parties of Eastern Europe, Russia, and the Successor States: A Reference Guide (1994). IX. Conclusion This database is the product of an effort to bring systematic, cross-country in- formation on political institutions within easy reach of researchers investigating comparative political economy. We have found no other source of information that provides as much objective data on these issues. We hope that the dpi will help illuminate a wide range of questions, from the determinants of democratic consolidation, the political conditions for economic reform, and the political and institutional roots of corruption to the appropriate and institutionally sensitive design of economic policy recommendations. Appendix. Variables in the Database of Political Institutions General Variables COUNTRYC Country code IFS International Financial Statistics code YEAR Year 172 the world bank economic review, vol. 15, no. 1 Chief Executive Variables SYSTEM Direct presidential (0); strong president elected by assembly (1); parliamentary (2) YRSOFFC Number of years in office FINITTRM Is there a finite term in office? (1 if yes, 0 if no) YRCURNT Years left in current term MULTPL? If there are formal restraints on the term (NA if not), can the chief executive serve multiple terms? MILITARY Is the chief executive a military officer? DEFMIN Is the defense minister a military officer? PERCENT1 Percentage of votes the president got in the first/only round PERCENTL Percentage of votes the president got in the final round PRTYIN Length of time (years) the chief executive’s party has been in office Chief Executive’s Party EXECME Name of party, if any EXECRLC Orientation: right (R); left (L); center (C); not applicable (0) EXECNAT Nationalist (1 if yes) EXECRURL Rural (1 if yes) EXECREG Regional (1 if yes) EXECME Religion: not specified (1); Christian (CH); Catholic (CA); Is- lamic (IS); Hindu (HD); Buddhist (BD); Jewish (JW); other (O) EXECAGE Length of time (years) since the party was formed under this name ALLHOUSE Does the party of the chief executive control all relevant houses? NONCHIEF In systems with both a nonceremonial prime minister and a presi- dent, what is the party affiliation of the one not called the chief executive? Party Variables in the Legislature HERFGOV Herfindahl index for government GOVFRAC Government fractionalization (the probability that two random draws would produce legislators from different parties) NUMGOV Number of government seats Largest Government Party 1GOVME Name 1GOVSEAT Seats 1GOVRLC Orientation: right (R); left (L); center (C); not applicable (0) 1GOVNAT Nationalist (1 if yes) 1GOVRURL Rural (1 if yes) 1GOVREG Regional (1 if yes) Beck, Clarke, Groff, Keefer, and Walsh 173 1GOVREL Religion: not specified (1); Christian (CH); Catholic (CA); Is- lamic (IS); Hindu (HD); Buddhist (BD); Jewish (JW); other (O) 1GOVAGE Length of time (years) since the party was formed under this name Second-Largest Government Party 2GOVME Name 2GOVSEAT Seats 2GOVRLC Orientation: right (R); left (L); center (C); not applicable (0) 2GOVNAT Nationalist (1 if yes) 2GOVRURL Rural (1 if yes) 2GOVREG Regional (1 if yes) 2GOVREL Religion: not specified (1); Christian (CH); Catholic (CA); Is- lamic (IS); Hindu (HD); Buddhist (BD); Jewish (JW); other (O) 2GOVAGE Length of time (years) since the party was formed under this name Third-Largest Government Party 3GOVME Name 3GOVSEAT Seats 3GOVRLC Orientation: right (R); left (L); center (C); not applicable (0) 3GOVNAT Nationalist (1 if yes) 3GOVRURL Rural (1 if yes) 3GOVREG Regional (1 if yes) 3GOVREL Religion: not specified (1); Christian (CH); Catholic (CA); Is- lamic (IS); Hindu (HD); Buddhist (BD); Jewish (JW); other (O) 3GOVAGE Length of time (years) since the party was formed under this name GOVOTH Number of other government parties GOVOTHST Number of other government party seats HERFOPP Herfindahl index for opposition OPPFRAC Opposition fractionalization (the probability that two random draws would produce legislators from different parties) NUMOPP Number of opposition seats Largest Opposition Party 1OPPME Name 1OPPSEAT Seats held 1OPPRLC Orientation: right (R); left (L); center (C); not applicable (0) 1OPPNAT Nationalist (1 if yes) 1OPPRURL Rural (1 if yes) 1OPPREG Regional (1 if yes) 1OPPREL Religion: not specified (1); Christian (CH); Catholic (CA); Is- lamic (IS); Hindu (HD); Buddhist (BD); Jewish (JW); other (O) 1OPPAGE Length of time (years) since the party was formed under this name 174 the world bank economic review, vol. 15, no. 1 Second-Largest Opposition Party 2OPPME Name 2OPPSEAT Seats Third-Largest Opposition Party 3OPPME Name 3OPPSEAT Seats OPPOTH Number of other opposition parties OPPOTHST Number of other opposition party seats ULPRTY Number of parties nonaligned or with allegiance unknown NUMUL Seats nonaligned or with allegiance unknown HERFTOT Herfindahl index total FRAC Total fractionalization (the probability that two random draws would produce legislators from different parties) OPPMAJH Does one opposition party have the majority in the House? OPPMAJS Does one opposition party have the majority in the Senate? Details on the Legislature DATELEG When were legislative elections held (number records month; 13 indicates unknown month) DATEEXEC When were executive elections held (number records month; 13 indicates unknown month) MAJ Majority (number of government seats divided by total seats) PARTYAGE Average age of parties LEGELEC Is there a legislative election? (1 if yes) EXELEC Is there an executive election? (1 if yes) EXECSPEC Is executive party special interest? GOVSPEC Is first government party special interest? COALSPEC Are any coalition parties special interest? Electoral Rules LIEC Legislative index of political competitiveness EIEC Executive index of political competitiveness MDMH Mean district magnitude (House) MDMS Mean district magnitude (Senate) S/ S + H Number in Senate /(number in house + number in Senate) PLURALTY Plurality? (1 if yes, 0 if no) PR Proportional representation? (1 if yes, 0 if no) HOUSESYS If plurality and proportional representation, which governs the majority or all of the House seats? (1 if plurality, 0 if proportional) SENSYS If plurality and proportional representation, which governs the majority or all of the Senate seats? (1 if plurality, 0 if proportional) THRESH What is the vote threshold for representation? Beck, Clarke, Groff, Keefer, and Walsh 175 DHONDT Is the D’Hondt system used? (1 if yes, 0 if no) CL Are closed lists used? (1 if yes, 0 if no) SELECT Candidate selection: national (1); subnational (2); primary (3) FRAUD Were vote fraud or candidate intimidation serious enough to affect the outcome of the most recent elections? Stability and Checks and Balances TENLONG Longest tenure of a veto player TENSHORT Shortest tenure of a veto player TENSYS Tenure of system of government if democratic; tenure of chief executive otherwise IPCOH Index of political cohesion (based on Roubini and Sachs 1989) CHECKS1 Number of veto players CHECKS2 CHECKS2 = CHECKS1 + 1 for each veto player whose orientation is closer to the opposition than to the government POLARIZ Maximum difference of orientation among government parties (0–2) STABS Percentage of veto players that dropped from government be- tween year t and year t+1 (assume Senate, if it exists, also changes) STABNS Percentage of veto players that dropped from government between year t and year t+1 (assume Senate, if it exists, did not also change) Federalism AUTON Are there autonomous regions? MUNI Are the municipal governments locally elected? STATE Are the state or province governments locally elected? AUTHOR Do subnational governments have extensive taxing, spending, or regulatory authority? STCONST Are the constituencies of the senators states/provinces? References The word “processed� describes informally reproduced works that may not be commonly available through library systems. Alesina, Alberto, Sule Ozler, Nouriel Roubini, and Philip Swagel. 1996. “Political Insta- bility and Economic Growth.� Journal of Economic Growth 2:189–213. Banks, Arthur S. Various years. The Political Handbook of the World. Volumes 1986 through 1996. Binghamton, N.Y.: CSA Publications. Beck, Thorsten, George Clarke, Alberto Groff, Philip Keefer, and Patrick Walsh. 2000. “New Tools and New Tests in Comparative Political Economy: The Database of Political Institutions,� World Bank Policy Research Working Paper 2283. World Bank, Washington, D.C. Clague, Christopher, Philip Keefer, Stephen Knack, and Mancur Olson. 1996. “Prop- erty and Contract Rights under Democracy and Dictatorship.� Journal of Economic Growth 1(2):243–76. 176 the world bank economic review, vol. 15, no. 1 Cox, Gary, and Mathew McCubbins. 2001. “The Institutional Determinants of Economic Policy Outcomes.� In Stephan Haggard and Mathew D. McCubbins, eds., Structure and Policy in Presidential Democracies. New York: Cambridge University Press. The Europa Year Book. Various years. Volumes 1977 through 1994. London: Europa Publications Ltd. Ferree, Karen E., and Smita Singh. 1999. “Institutional Change and Economic Perfor- mance in Africa, 1970–1995.� Paper presented at the 1999 meetings of the American Political Science Association, Atlanta, Ga. Political Parties of Africa and the Middle East: A Reference Guide. 1993. Longman Current Affairs Series, Detroit: Longman. Political Parties of Eastern Europe, Russia, and the Successor States: A Reference Guide. 1994. Longman Current Affairs Series, Harlow, U.K.: Longman. Roubini, Nouriel, and Jeffrey Sachs. 1989. “Political and Economic Determinants of Bud- get Deficits in the Industrial Democracies.� European Economic Review 33:903–38. Shugart, Matthew Soberg, and John Carey. 1992. Presidents and Assemblies: Constitu- tional Design and Electoral Dynamics. New York: Cambridge University Press.