64525 ResearchDigest World Bank VOLUME 5 ✬ NUMBER 4 ✬ SUMMER 2011 Generating Jobs in Developing Countries: Small versus Young Firms What role do small and medium-size contribute more to employment in IN THIS ISSUE enterprises play in generating jobs low-income countries than in higher- Generating Jobs in Developing Countries: and promoting economic recovery? income ones. Earlier work has shown Small versus Young Firms … page 1 that new �rm entry is much higher in Small �rms—especially mature ones—are high-income countries. Taken together, big contributors to job creation in developing M ulti-billion-dollar aid portfo- these �ndings suggest that higher- countries lios are directed at fostering income countries have high rates of the growth of small and entry and turnover of small �rms rather FOCUS medium-size enterprises (SMEs). But than a large SME sector. Measuring Trade in a World of Fragmented there is little systematic research or Second, across country income groups, Production … page 2 data informing the policies in support small �rms not only employ the most people, Today’s increasingly integrated world of SMEs, especially in developing they also generate the most new jobs. SMEs demands new measures of trade. Experts are countries. Moreover, the empirical with 250 or fewer employees generate developing them evidence on the relationship between a median 86.01 percent of the jobs. Radio Access and Service Delivery �rm size and growth has been mixed. They generate a larger share (93.05 in Benin … page 3 Recent work in the United States percent) in the countries with net job Does community radio help citizens hold suggests that start-ups and surviving creation (�gure 1). government accountable for service delivery young businesses contribute dispro- Even in countries with net job loss- in northern Benin? portionately to job creation and net es, SMEs with 250 or fewer employees growth—but shows no systematic create a substantial share of jobs Trends in Adult Mortality around the relationship between �rm size and (81.51 percent). Young �rms generate Developing World … page 4 growth after controlling for �rm age. It only 14 percent of net jobs in countries A new study suggests that adult mortality is unclear whether these �ndings apply with net job creation and only 5.39 warrants a more central place in analysis of in developing countries, where there percent in those with net job losses. population health are greater barriers to entrepreneur- Third, small �rms and young �rms have Eight Questions about Brain Drain … page 5 ship and venture capital markets are the highest employment growth rates in regres- less developed. sions controlling for country, industry, and A recent wave of research helps address key In a recent study Ayyagari, year �xed effects. But small �rms’ higher questions underlying the brain drain debate. But better data are still needed Demirgüç-Kunt, and Maksimovic put employment growth is not accompa- together a database with consistent nied by higher sales or productivity Who Uses Bottled Gas in Developing and comparable information on the growth. Large �rms and young �rms Countries? … page 6 contribution of �rms to employment, have higher productivity growth. Thus Raising awareness, rather than subsidizing job creation, and growth across 99 while SMEs employ a large share of prices, is likely to promote a long-term switch developing economies. The sample workers and create most jobs, their to cleaner fuels consists of 47,745 �rms surveyed in contribution to productivity and 2006–10. The authors then examine growth is uncertain at best. How Much Does Infrastructure Contribute the relationship between �rm size, age, Overall, the authors’ �ndings con- to GDP Growth? … page 7 employment, and productivity growth trast with those of studies using U.S. Increasing the quantity and quality of and how this varies with country in- data. In the United States large mature infrastructure, while worthwhile, is costly come. The results lead to three main �rms have the largest share of em- and can take decades �ndings. ployment; the authors �nd that while First, SMEs are the biggest contributors to employment across countries. But they (continued on page 8) 2 World Bank ResearchDigest FOCUS Measuring Trade in a World of Fragmented Production A recent workshop featured as well as methods to integrate and exports processed in multiple coun- research on developing better reconcile trade and national accounts tries before reaching their �nal measures of trade for today’s statistics in a multicountry framework. consumers. And the �nal session discussed future increasingly integrated world work on integrating value added trade Robert Stehrer, Marcel Timmer, and statistics with the system of national G lobal production has become accounts. input-output database, found that for increasingly fragmented, with Initial empirical insights emerging most countries an increasing part of different stages of production from the work presented at the work- domestic �nal output is captured as now regularly taking place in differ- shop include the following examples: income by foreign production factors. ent countries. Because inputs cross Between 1995 and 2006 the foreign in- borders multiple times, traditional Noguera showed that bilateral imbal- come share of �nal manufacturing out- statistics on trade values—measured ances measured in value added differ put in China increased from 14 to 21 in gross terms—no longer reflect eco- from gross trade imbalances. Most percent. This share is now on par with nomic reality on the value added in prominently, the U.S.-China imbal- that in the United States but smaller any particular country. International ance in 2004 is 30–40 percent smaller than that in any European country. economists and trade statisticians are when measured in value added. China’s share in foreign �nal output developing new measures of trade to is increasing even faster, though it re- meet the increasing demand from pol- mains smaller in value. icy makers and the public for a better of trade in net value added closely Advances in measuring trade in val- understanding of cross-border trade in resemble net trade flows but that ue added could influence how policy today’s increasingly integrated world. there are distinct patterns when look- makers deal with a range of issues. Top experts in the �eld presented ing at individual factors. For example, Trade imbalances. The reallocation their research on developing better North American Free Trade Agreement of bilateral trade de�cits and surpluses measures of trade at a Development (NAFTA) members are net exporters between partner countries could allevi- Economics workshop organized by of highly educated labor, mostly to ate protectionist pressures. Aaditya Mattoo, Francis Ng, and Zhi the EU-15 (the 15 European Union Trade disputes. Accounting for do- Wang and held at the World Bank on members before the accessions in May mestic value added in foreign products June 9–10, 2011. Besides building con- 2004)—but have increasingly become could modify notions of “them� and sensus and identifying opportunities net importers of less educated labor “us� and create a greater awareness of for collaboration among experts in dif- from China. mutual dependence. ferent groups, the workshop identi�ed Trade and industrial policy. If there data gaps, sketched the direction of Koopman, and Zhi Wang showed that is a case for proactive policy, the focus future work in this area, and contrib- Mexico’s manufactured exports have may need to be less on the industry uted to the formation of a network of a foreign value added share of about and more on occupations and tasks. researchers and international agencies. 66 percent on average. Industries with Trade and macroeconomic shocks. The workshop focused on both con- a foreign content share of 50 percent An empirical study of global value ceptual and practical issues. The �rst or more account for 80 percent of the chains may help in understanding how session was devoted to the concep- country’s manufactured exports. crises are transmitted through trade tual framework and methodology for channels. computing trade in value added terms Powers, Zhi Wang, and Shang-Jin Wei Trade and employment. A study of based on statistics now available. The showed that among emerging markets, global value chains could also help in second focused on the statistical chal- East Asian countries are more tightly identifying the links between trade and lenges in measuring value added trade integrated with international produc- job creation or destruction in different and in improving statistics. The third tion chains. They have greater foreign locations. discussed potential policy insights content in their exports and a larger from the proposed new measures of share of intermediate inputs sent to trade. The fourth provided evidence on third countries for further processing. value added trade from two developing Among major developed economies, countries, China and Mexico. The �fth the United States uses the most im- The workshop program, papers, and concentrated on speci�c data issues, ported intermediate inputs to produce presentations can be found at such as identifying intermediate and exports, while Japan has the most http://go.worldbank.org/R156ABXQQ0 or �nal demand in goods and services value added embodied in intermediate http://econ.worldbank.org/programs/trade. World Bank Research Digest 3 Radio Access and Service Delivery determined their location. The study’s in Benin conclusions, however, are based on within-commune comparisons of vil- lage access to out-of-commune com- Access to community radio Households with greater access to munity radio broadcasts. improves household decisions but community radio are signi�cantly more Surely, though, one might argue not government accountability in likely to pay for mosquito bed nets. that radio stations’ programming But they appear to have lower access choices are related to the characteris- villages of northern Benin to free, government-provided bed nets: tics of the villages in their broadcast in areas where household demand for area. After all, it is well known that I n many developing countries bed nets is higher, more bed nets are community radio stations cater to citizens are unable to hold gov- diverted from the free program into a the linguistic and cultural demands ernments accountable for better gray market, where households pay for of their listeners. But they receive service delivery, but they also under- them. little or no revenue from advertising. estimate the bene�ts of the services Is the relationship between com- Instead, they depend heavily on the that government offers. Donors have munity radio access and service de- �nancing they receive from donors and long believed that access to mass livery a spurious association, or does government ministries in exchange for media addresses both problems. They the �rst cause the second? This is a broadcasting programs that suit the �nance community radio, especially serious concern: community radio priorities of the donors and ministries in Africa, to reinforce accountability stations broadcasting about health (health and education), not those of in new democracies, and they support and education may prefer to locate in - mass media programming to encour- areas where households are already nity radio access therefore have greater age households to change their health more concerned about these issues. exposure to broadcasts touting the practices and to send their children Unobserved levels of household con- virtues of better health practices and to school. A new study by Keefer and cern about health and education, the importance of education. Khemani, using information from rather than community radio access, These results have three policy 4,200 households in 210 villages would then explain the association implications. First, lack of policy infor- across northern Benin, examines for between community radio access and mation is often not the only impedi- the �rst time whether media access health and education outcomes. In ment to collective action needed to affects service delivery outcomes and, fact, the conditions of the radio market hold governments accountable. As in if so, whether through changes in in northern Benin make it clear that Benin, citizens may lack organizations, household behavior or in government village access to community radio is particularly political parties, that can accountability. unrelated to village characteristics. mobilize even fully informed citizens Among the sample villages, re- Small community radio stations for collective action. Second, com- sults of literacy tests given to 2,100 with a limited broadcast radius dot munity radio stations that depend on second-grade children were signi�- northern Benin. Even within com- donor and government �nancing may cantly better in villages with greater munes, villages vary substantially in be particularly unlikely to broadcast community radio access. But the study the number to which they have access. “accountability� programming. Third, could identify media effects only on This variation is unrelated to village even in the absence of an account- household behavior: households with characteristics that might also influ- ability channel, media interventions greater media access spend more on ence service delivery. When radio sta- can still influence household behavior. their children’s education. In contrast tions have weak signals, insigni�cant But media outlets that can support to previous research and to the main topographic and geologic differences themselves with advertising may be rationale for donor efforts to improve between villages can have a substan- reluctant to broadcast the health and mass media, the study found that tial effect on their radio access. As a education programs that influence be- community radio access did not ap- result, as the sample of villages shows, havior in northern Benin. pear to increase the ability of villages a village may receive the broadcasts to demand greater government ac- of fewer radio stations than an other- wise identical neighboring village. It Philip Keefer and Stuti Khemani. 2011. “Mass education were not signi�cantly differ- is implausible that such differences Media and Public Services: The Effects of Radio ent across villages with more and less would affect service delivery outcomes. Access on Public Education in Benin.� Policy Research Working Paper 5559, World Bank, access to community radio. In addition, historical accounts of the Washington, DC. These results are consistent with founding of community radio stations ———. Forthcoming. “The Role of Mass Media those of a companion investigation indicate that highly local factors (such in Developing Countries: The Impact of Radio by the authors, using data from the as intracommune power struggles), on the Distribution of Anti-malaria Bed Nets in same survey but focusing on house- affecting only villages in the com- Benin.� Policy Research Working Paper, World hold acquisition of mosquito bed nets. munes where the stations are located, Bank, Washington, DC. 4 World Bank ResearchDigest Trends in Adult Mortality around the Developing World In Sub-Saharan Africa mortality Swaziland, Zambia, and Zimbabwe) education. Women ages 15–24 with rates in countries with the highest exceed those in countries that expe- less than primary education have high- HIV prevalence exceed those in rienced episodes of civil war. Excess er mortality rates than all other groups mortality during episodes of genocide of the same age. This is consistent with countries that had civil war is readily apparent in the data—with aggregate adult mortality rates ap- with �ndings by others that younger, I n a new paper de Walque and proaching 15 percent (that is, among more educated women appear to be Filmer combine data from 84 those alive at the beginning of a �ve- engaging in more protective behaviors Demographic and Health Surveys year period before the genocide, the from 46 developing countries to probability of death was 15 percent disseminated. analyze trends and socioeconomic over that �ve-year period). But mortal- Finally, and perhaps unsurpris- differences in adult mortality. They ity rates decline at the end of these ingly, countries that have experienced calculate mortality based on the sib- extreme mortality events. In contrast, conflict or episodes of internationally ling mortality reports collected from adult mortality in countries with high recognized genocide display heteroge- female respondents ages 15–49. The - neous patterns in time trends and so- analysis yields �ve main �ndings. ing over the period ending in 2004. cioeconomic patterns in mortality. The First, adult mortality is different Third, even in Sub-Saharan coun- experience in several countries sug- from child mortality. While under-�ve gests that male mortality is typically mortality shows a de�nite improving mortality rates appear to be at best more responsive to these events—and trend over time, adult mortality does stagnating and in several cases even in several cases it is urban or more not. Moreover, the cross-sectional as- increasing. It is unclear whether this educated males for whom mortal- sociation between under-�ve mortal- ity increases the most. At the same ity and national income is quite a bit are translating into higher mortality time, it is important to recognize that stronger than that for adult mortality. in these countries or because of some increases in mortality during these epi- In addition, while under-�ve mortal- other underlying cause. sodes of conflict are not restricted to ity has fallen over time conditional on Fourth, the main socioeconomic men. In all cases women’s mortality is national income, this is not the case dimension along which mortality ap- affected as well. for adult mortality. Indeed, in Sub- pears to differ in the aggregate is This research suggests that analysis Saharan Africa the trend is the oppo- gender. In Sub-Saharan Africa adult of population health in developing site, with adult mortality rising at any mortality rates have risen substantially countries should give more central given level of income. higher for men than for women— consideration to adult mortality. Other Second, adult mortality has in- indicators, such as child mortality, are creased dramatically in Sub-Saharan prevalence. On the whole, the data do poor measures of the dimensions of African countries, especially in those not show large gaps between urban health captured by direct measure- - and rural residence or by school at- ment of adult mortality. While the demic (�gure 1). Mortality rates in tainment. To the extent that there are analysis of the data created for the the most affected countries of south- some differences, the data suggest study shows considerable hetero- ern Africa (for example, Namibia, that urban women and more edu- geneity in mortality patterns across cated women had smaller increases in countries, these publicly available mortality. Among the groups studied, data will allow other analysis to in- Figure 1. Predicted Male Adult Mortality urban women and educated women vestigate country-speci�c patterns in by Period and Location 7 also have the lowest overall mortality more depth, supporting the design of tailored policies and programs that are 6 better informed. 1980–84 prevalence. 2000–04 Male adult mortality (%) 5 Men with less than primary edu- 4 cation had the largest increase in 3 prevalence these men have mortal- ity rates that are substantially higher 2 than those for men with more educa- Damien de Walque and Deon Filmer. 2011. 1 tion and for women. One age group “Trends and Socioeconomic Gradients in Adult for which the higher mortality among Mortality around the Developing World.� Policy 0 men does not hold is among those Rural Urban Rural Urban Research Working Paper 5716, World Bank, Sub-Saharan Africa Countries in other regions ages 15–24, especially those with less Washington, DC. World Bank Research Digest 5 Eight Questions about Brain Drain New evidence counters some of developed countries. This migrant flow to be stronger contributors to brain the myths and eases some of the also has a rising skill level, so that gain, particularly in small countries common concerns about brain drain brain drain is increasing in absolute with limited possibilities for tertiary terms. But except in Sub-Saharan education. Africa, rising levels of tertiary educa- Another potential bene�t of brain T he emigration of high-skilled tion have offset this skilled migration drain is return flows of income, in- labor is an emotive issue that to ensure that the rate of emigration vestment, and expertise in the send- in popular discourse is often has been very stable. ing country. In India remittances referred to as brain drain, conjuring Brain drain rates vary widely across have reached $55 billion a year, and images of extremely negative effects countries, with smaller, unstable returned migrants have shaped the on developing countries. Recent dis- countries (outside of Africa) having development of the information tech- cussion of brain gain, diaspora effects, greater brain drain. The microeco- nology sector. Smaller or unstable and other advantages of migration nomic literature shows that broader countries face a different picture. The has been used to argue against this, career concerns—such authors �nd that less but much of the discussion has lacked as the quality of op- than half of migrants evidence. portunities to research, Twice as many with tertiary education Fortunately brain drain has been learn, and work—as articles on brain send remittances. But enjoying a renaissance as a subject of well as lifestyle and economic study: according to Econlit, family reasons, are more drain were published there is a strongbetween tive correlation nega- 247 articles on brain drain were pub- important than income between 2005 and income levels and the lished between 2005 and 2009—about gaps in determining the likelihood that skilled twice as many as over the previous decision to migrate. 2009 as in the migrants send remit- The authors analyze previous 15 years. tances, so poor coun- McKenzie builds on this wave of re- stocks of migrants to ex- search to answer eight key questions amine how these evolve But data limitations tries are most likely to bene�t from such underlying much of the brain drain over time. They �nd that hamper our remittances. debate: What is brain drain? Why countries that send rela- should economists care about it? Is tively many high-skilled In understanding of its �scal assessing theand production brain drain increasing? Is there a posi- migrants to one coun- consequences externalities of brain tive relationship between skilled and try also send relatively drain, three factors unskilled migration? What makes brain many low-skilled migrants to the same of return need to be considered: the drain more likely? Does brain gain ex- country. Data from the Organisation welfare of the individual, the �scal ist? Do high-skilled workers remit, in- for Economic Co-operation and losses (estimated to be $500–1,000 vest, and share knowledge back home? Development show strong evidence of per migrant in the Federated States What do we know about the �scal and of Micronesia and Tonga, and $5,500– production externalities of brain drain? low-skilled migration, and vice versa. Brain drain is used to refer to the Consequently, destination countries drain affects the production and health outmigration of a nation’s most highly must be aware that attempts to limit externalities that may arise through skilled individuals, making up a dis- one form of immigration may have ef- education. proportionately large share of total fects on the other form. The new evidence outlined by the migration. Consider this example: Recently, theoretical papers have authors counters some of the myths among Cambodians age 25 and above, suggested that high-skilled migra- and eases some of the common con- 2.6 percent of those with primary edu- tion can lead to a rise in human capi- cerns about brain drain. But a recur- cation live abroad, compared with 5.9 tal levels in the originating country. rent message is that data limitations percent of those with secondary edu- Individuals take into account the remain a huge challenge. Better data cation and 18.3 percent of those with prospect of future migration when as well as specialized surveys are need- tertiary education. investing in education, but may not ed to improve the understanding of the The international migration cap- eventually migrate. The authors argue consequences of these movements. tured by the term brain drain is of for caution with these claims, because tremendous policy concern in many empirical evidence suggests shifts not countries. Between 1960 and 2010 the in the level of attainment but in what John Gibson and David McKenzie. 2011. “Eight global stock of migrants grew only is studied—for example, languages. Questions about Brain Drain.� Policy Research slightly faster than the world popula- Other channels, such as remittances Working Paper 5668, World Bank, Washington, tion, but the flow has increasingly and the return of migrants who ac- DC. Also forthcoming in Journal of Economic been from less developed to more quire education abroad, are likely Perspectives. 6 World Bank ResearchDigest Who Uses Bottled Gas in Developing Countries? Educating the public, especially continue to use biomass in a number to choose clean and convenient fuels women, about the costs and bene�ts of developing countries. because women are directly affected by of fuel choice could promote a A recent study by Kojima, Bacon, fuel choice. Once education levels of and Zhou, surveying 110 developing women and men were separately ac- switch to cleaner fuels counted for, however, the gender of the the primary cooking fuel for more than head of household was not statistically M ore than a third of the world’s half the households in 33 countries. signi�cant in most cases. Where it was population relies on biomass signi�cant, results showed that male- (wood, dung, crop residues, quintile in 51 of the 63 countries for headed households were more likely charcoal) for cooking and heating. Indoor air pollution from burning bio- use by quintile. stands in contrast mass in traditional stoves can cause The study further Household choice to other study �nd- serious health damage. Collecting examined national ings, may suggest that biomass takes time away from such household surveys of LPG increases female-headed house- productive activities as schooling and with household holds face unmeasured child care. And concentrated use of Indonesia, Kenya, economic disadvan- fuelwood and charcoal can threaten Pakistan, and Sri Lanka expenditure and level tages that make less tree resources. Yet a great many fami- to gain a better under- of education lies in developing countries will con- standing of what factors purchase. tinue to use biomass for the foresee- influence households’ Some developing able future because it typically has the lowest �nancial cost. for the households that have decided to encourage fuel switching. But re- Switching to a gaseous fuel, such to use it, how much to use (consump- selection mean that the subsidies needed would one way of reducing the problems increased with household expenditure have to be so large as to be �scally and the highest levels of education unsustainable. Monthly household the most commonly used clean fuel attained by female and male house- expenditure of more than $300 would alternative for cooking in developing consumption be needed to enable a household (kilograms per month per household) - households tends to rise with income increased with household expenditure tain its use. These �ndings suggest that in promoting household use of of choice for convenience and cleanli- ness: urban areas without access to level of infrastructure development households that have an income high natural gas and all rural areas. But evi- (electricity connection, urban rather dence in developing countries shows than rural residence) increased selec- subsidies and that already live in areas tion and consumption, while engage- continue to use biomass and that use ment in agriculture (broadly associated Education is likely to be a proxy of multiple fuels is common except at with biomass availability) reduced for the level of awareness about the high income levels. them. Rising prices of �rewood and - Besides income, other factors also influence patterns of household fuel raising awareness about the bene�ts of income and relative fuel prices were the two most important determinants good road infrastructure. There are of fuel use patterns. among women, might be effective in economies of scale in cylinder man- Interestingly, the higher the educa- shifting households away from solid tion level attained by female and male a critical mass of regular consumers to members of a household, the more become viable. These factors limit the This effect was larger for women than developing world, particularly in rural for men. Education affected consump- areas. Cultural preferences and inad- tion less than selection, and men’s ed- Masami Kojima, Robert Bacon, and Xin Zhou. equate information about different fuel ucation increased consumption more 2011. “Who Uses Bottled Gas? Evidence from options—including fear of explosion than women’s in India and Sri Lanka. Households in Developing Countries.� Policy Re- and �re—also influence fuel use, and It is often speculated that female- search Working Paper 5731, World Bank, Wash- many high-income urban households headed households are more likely ington, DC. World Bank Research Digest 7 How Much Does Infrastructure Contribute to GDP Growth? A big push in infrastructure signi�cant positive effect of infrastruc- principal component procedure. The investment may entail trade-offs and ture on output, productivity, or their econometric approach deals explicitly distortions that could substantially growth rate. Results are less conclu- with the nonstationarity of infrastruc- sive among studies using pecuniary ture and other productive inputs, reduce the net effect on growth measures such as public investment reverse causality from output to infra- flows or their accumulation into public structure, and potential cross-country A mong empirical studies examin- capital. There is a good reason for this: heterogeneity in the contribution of ing the long-run effect of public the lack of a close correspondence infrastructure to aggregate output. infrastructure on total factor between public capital expenditure The authors’ empirical strategy productivity and growth in develop- and the accumulation of public infra- involves estimating a production func- ing countries, few are convincing. This structure assets or the provision of tion relating output per worker to non- literature began with David Aschauer’s infrastructure services, owing to inef- infrastructure physical capital, human “Is Public Expenditure Productive?� �ciencies in public procurement and capital, and infrastructure inputs. Their (Journal of Monetary Economics 23 [1989]: outright corruption—issues probably estimates, based on heterogeneous 177–200). Aschauer, using annual more important in developing than in panel time-series techniques, place time-series data for the United States, developed economies. the output elasticity of infrastructure calculated that with private inputs The recent literature tends to �nd in a range between 0.07 and 0.10 (de- held constant, the annual marginal smaller (and more plausible) effects pending on the speci�cation used). product of public capital was in excess than those reported in Aschauer and The estimates are highly signi�cant of 100 percent. subsequent studies. Among recent and robust to a variety of experiments Subsequent research estimating studies using a production function involving alternative econometric production functions used pooled approach, the midpoint estimate of speci�cations and different synthetic U.S. state data in levels, disaggregated measures of infrastructure. Some il- public capital into its main compo- infrastructure capital lies around 0.15 lustrative calculations show that the nents, or used industry data and found for developed countries, implying that output contribution of infrastructure similar effects, particularly for road a doubling of infrastructure capital implied by these results is also eco- and highway capital. Time-series esti- nomically signi�cant. The estimates mates in levels might simply capture from recent studies using broader of the output contribution of human common trends. And pooled state country samples are similar. But all capital and noninfrastructure physical data in levels might simply capture these capture only the direct effect of capital are similarly signi�cant and in underlying persistent state character- infrastructure on output, given the use line with those reported by the earlier istics; that is, richer states invest more of other productive inputs. There may literature. in public capital. In fact, when the be additional indirect effects accru- Tests of parameter homogeneity production function is estimated with ing through changes in the use of the reveal little evidence that the output aggregate U.S. data in differences or other inputs because of complemen- elasticity of infrastructure varies across with state data with �xed effects, zero tarities with infrastructure. countries. Observed differences in the or even negative marginal products of A recent paper by Calderón, ratio of aggregate infrastructure to public capital become typical. Moral-Benito, and Servén represents output offer a useful guide to the dif- This literature measures infrastruc- the state of the art in the empiri- ferences in the marginal productivity ture in terms of physical stocks (for cal literature on the contribution of of infrastructure. The paper also points example, kilometers of roads) or pecu- infrastructure to aggregate output. out that increases in the quantity and niary stocks constructed by accumu- Using an infrastructure-augmented quality of infrastructure, while worth- lating spending flows. The underlying production function approach, the pa- while, are costly and take decades to assumption is that the flow of produc- per estimates the output elasticity of implement. The �scal distortions as- tive infrastructure services is directly infrastructure on a large cross-country sociated with such an effort, and the related to the size of the stock of infra- panel data set covering 88 countries, trade-offs it would entail with other structure assets. A number of empiri- spanning the years 1960–2000, and needed investments, could well reduce cal studies using various approaches containing more than 3,500 annual the net growth effect substantially. also �nd that the output contribution observations. of infrastructure exceeds that of con- The paper addresses several limita- ventional capital, which suggests the tions of the earlier literature. It uses César Calderón, Enrique Moral-Benito, and Luis presence of externalities associated a multidimensional concept of infra- Servén. 2011. “Is Infrastructure Capital Produc- with infrastructure services. structure, combining power, trans- tive? A Dynamic Heterogeneous Approach.� Policy The literature using physical mea- port, and telecommunications into a Research Working Paper 5682, World Bank, sures of infrastructure stocks reports a synthetic index constructed through a Washington, DC. 8 World Bank ResearchDigest (continued from page 1) Figure 1. Job Creation Shares across Countries by Size and Age of Firms Recent Policy Research 30 Working Papers 25 5714 Does Female Empowerment Promote Economic Development? 20 Matthias Doepke and Michèle Tertilt Percent 5715 The Effects of Conflict on Fertility in Rwanda 15 Kati Schindler and Tilman Brück 5719 Population, Poverty, and Sustainable 10 Development: A Review of the Evidence Monica Das Gupta, John Bongaarts, and 5 John Cleland 5722 The Air Connectivity Index: Measuring 0 Integration in the Global Air Transport Network 5–99 Jean-François Arvis and Ben Shepherd 100–249 5727 Productivity Effects of Land Rental Markets Size in Ethiopia: Evidence from a Matched Tenant- 250–499 (employees) Landlord Sample 11+ 6–10 500+ Klaus Deininger, Daniel Ayalew Ali, and Tekie 3–5 Alemu Age (years) <=2 5730 Political Economy of the Mining Sector in Ghana Joseph Ayee, Tina Søreide, G. P. Shukla, and Tuan Minh Le 5732 Mixtures of g-priors for Bayesian Model large �rms have a substantial share of With countries around the world Averaging with Economic Application employment in developing economies, struggling to recover from the crisis, Eduardo Ley and Mark F. J. Steel 5736 Ef�ciency in Public Procurement in Rural Road small mature �rms have the largest job creation is a priority for policy Projects of Nepal share. The U.S. evidence suggests that makers. The authors’ results suggest Radia Benamghar and Atsushi Iimi 5738 The Labor Market, Education and Armed small mature �rms have net job losses; that the challenge is not only to create Conflict in Tajikistan the authors �nd that in developing more jobs but to create better-quality Olga N. Shemyakina countries small mature �rms have the jobs. Overall, they show that while 5744 The Role of Inventory Adjustments in Quantifying Factors Causing Food Price largest share of job creation. Moreover, SMEs employ a large number of peo- Inflation in countries with net job losses, it is ple and create more jobs than large Gal Hochman, Deepak Rajagopal, Govinda Timilsina, and David Zilberman only the small �rms—especially small �rms, they contribute less to produc- 5747 Governance in State-Owned Enterprises mature �rms—that have net job gains. tivity growth than large �rms do. Revisited: The Cases of Water and Electricity in Latin America and the Caribbean And while the U.S. evidence shows no Luis Alberto Andrés, José Luis Guasch, and systematic relationship between �rm Sebastián López Azumendi 5748 Commitments to Save: A Field Experiment in size and growth once age is controlled Rural Malawi Meghana Ayyagari, Aslı Demirgüç-Kunt, and Vo- for, the authors �nd that in developing Lasse Brune, Xavier Giné, Jessica Goldberg, jislav Maksimovic. 2011. “Small vs. Young Firms and Dean Yang countries small �rms are signi�cant across the World: Contribution to Employment, contributors to employment growth Job Creation, and Growth.� Policy Research Work- even after controlling for age. ing Paper 5631, World Bank, Washington, DC. Working Papers can be downloaded at http://econ.worldbank.org To download the World Bank Research E-Newsletter, go to Data & Research at http://www.worldbank.org The World Bank Research Digest is a quarterly publica- The Research Digest is �nanced by the Bank’s Editorial Committee: Jean-Jacques Dethier (managing tion disseminating �ndings of World Bank research. Research Committee and managed by DECRS, the editor) and Aslı Demirgüç-Kunt. Research assistance: The views and interpretations in the articles are those research support unit of the Development Economics Julia Bird; editor: Alison Strong; production: Roula of the authors and do not necessarily represent the Senior Vice Presidency (DEC). The Research Digest is Yazigi. For information or free subscriptions, send views of the World Bank, its Executive Directors, or the not copyrighted and may be reproduced with appropri- email to research@worldbank.org or visit countries they represent. ate source attribution. http://econ.worldbank.org/research_digest. The World Bank 1818 H Street, NW Washington, DC 20433, USA Printed on Recycled Paper