115200 S pe bo La ci r M ResearchDigest al Is ark su e World Bank e ts on VOLUME 11 NUMBER 3 SPRING 2017 Are Active Labor Market Programs Effective in Developing Countries? Active labor market programs in to six months. Some programs involve IN THIS ISSUE developing countries have had much just classroom training, while others smaller impacts than policy makers include internships to provide on-the- Are Active Labor Market Programs Effective job training. in Developing Countries? … page 1 and participants have expected Only a third of studies testing such G programs find a significant impact on Are Automation and Trade overnments have long used employment. The simple unweighted Polarizing Developing Country active labor market programs average across the studies is equal to Labor Markets, Too? … page 2 to try to increase employment. a 2.3 percentage point increase in em- How Does Innovation Affect Employment Yet until recently there had been few ployment. That is, for every 100 people in Developing Countries? … page 3 rigorous tests of their effectiveness offered vocational training, fewer than in developing countries. The past three will find a job that they would How Is the Internet Reshaping three years have seen the release of a not otherwise have found. The result Work? … page 4 number of studies using randomized is a cost of around $17,000–60,000 experiments to test the impacts of per additional person employed. The Big Data from Online Job Portals … page 4 these programs. These compare effects on earnings are also modest, employment outcomes for a treatment with an average increase of only $19 a The Motherhood Penalty and Female group randomly assigned to participate month, and again the majority of stud- Employment in Urban India … page 5 in a program with the outcomes for a ies fail to find a significant impact. control group of similar individuals not A second approach is to give job- How Some Female Entrepreneurs in Africa participating. A critical examination seekers a wage subsidy voucher that Enter Male-Dominated Sectors … page 6 of these studies in a recent paper by they can take to employers when McKenzie reveals the limited impacts hunting for a job. This subsidy lowers Can Wage Subsidies Boost Employment in the Wake that many of these programs have the cost to the employer of hiring the of an Economic Crisis? … page 7 had on employment creation. worker for the period of the subsidy, Traditional active labor market often six months to a year. The hope is Understanding the Effects of the World’s programs are of three main kinds: that this short-term employment expe- Largest Workfare Program … page 8 programs aimed at increasing the rience may continue to have an impact employability of workers through vo- once the subsidy has ended. How Do Cities in Ethiopia Create cational training; programs designed Experiments have tested the im- Jobs? … page 9 to increase the demand for labor by pact of wage subsidies for unemployed lowering its cost to firms through wage workers in Argentina and for unem- What Is Behind the Decline in Wage subsidies; and search and matching ployed youth in Jordan and South Inequality in Latin America? … page 10 assistance programs aimed at reduc- Africa. In both Argentina and South ing frictions that prevent demand from Africa employers were required to for- Global Talent Flows … page 11 meeting supply in the labor market. mally register the workers and comply One reason people may struggle with labor laws. This made employers Recent Policy Research Working to find jobs is that they lack the tech- very reluctant to use the vouchers. In Papers on Labor Market Issues … page 12 nical skills demanded by employers. contrast, the program in Jordan, which Vocational training programs aim to did not require firms to formally reg- overcome this constraint by provid- ister the workers, resulted in a very ing training in specific occupational skills, typically over a period of three (continued on page 12) 2 World Bank ResearchDigest Are Automation and Trade Polarizing Developing Country Labor Markets, Too? Labor markets show no evidence of demand, we will see an expansion of But in developing countries on polarization in developing countries employment. Because more open de- average, operators, professionals, and on average, but there are emerging veloping countries are often assumed elementary occupations all increase at to be price takers, productivity gains around the same rate. In Vietnam, per- signs that this could change should lead to greater employment. haps the archetypal offshoring destina- T •  The degree to which automation tion, operators and assemblers have he automation and outsourcing is adopted depends heavily on a coun- increased relative to every category of routine, codifiable tasks are try’s capacity to absorb technology, the except professionals. The story is simi- seen as driving polarization in la- skill of its workforce, the ability to mo- lar in India, up to 2004. Broadly similar bor markets in high-income countries. bilize resources for large capital invest- patterns appear in Ecuador, the Arab Recent studies in the United States ments, the capacity for maintenance, Republic of Egypt, El Salvador, Ghana, and across the OECD document ex- and attention to tolerances that may Malawi, Mali, Morocco, Nicaragua, panding job opportunities in both make it less easy to substitute away Peru, and South Africa. high-skill, high-wage occupations and from labor. Yet there are important exceptions low-skill, low-wage occupations, cou- •  To the degree that automation suggesting that this may not always be pled with contracting opportunities in does occur, the likelihood of develop- the case. Indonesia shows an absolute middle-wage, middle-skill white-collar ing the kinds of upgraded jobs that are fall in the operators category along and blue-collar jobs. necessary complements to automation with a rise in technicians and service Should we expect to find this sharp would appear to be lower in countries and elementary occupations. Brazil polarization of labor markets emerg- with a weak educational base. and Mexico both show relatively slow ing in the developing world as well? A The first set of estimates of the growth of the operators category that, recent paper by Maloney and Molina combination of these effects is mod- while not the absolute hollowing out explores this question using global estly pessimistic. Using labor force sur- found in the United States, is consis- census data. It begins by arguing that veys, the World Bank’s World Development tent with polarization. Press accounts there are several reasons to expect Report 2016: Digital Dividends argues that suggest the emergence of the same weaker forces of polarization in devel- middle-skill occupations intensive in automation dynamic in China, where oping than in advanced economies: routine cognitive and manual skills robotization is proceeding rapidly. •  Developing countries begin with have decreased across the develop- This raises two concerns related to a smaller population in such codifi- ing world, except in Argentina, China, internal and international equity. In able tasks and thus with fewer jobs to Ethiopia, and Nicaragua. developing countries, as in advanced displace. In their paper Maloney and Molina economies, the new, more complex •  Jobs offshored from advanced track job categories across time for labor tasks expected to complement economies are moving to develop- 21 developing countries in Africa, automation may emerge very slowly. ing countries, so we would expect to Asia, and Latin America following In China recent college graduates re- see a complementary expansion of the approach of David Autor (“The port problems finding jobs or consider the middle—a “depolarization” of the Polarization of Job Opportunities in themselves underemployed. wage distribution in at least some host the U.S. Labor Market: Implications Globally, the concern is that as au- countries. for Employment and Earnings,” tomation eliminates routine manufac- •  New technologies may provide Center for American Progress and the turing-type jobs, we may see a short- new leverage to break down barriers to Hamilton Project, 2010). They use the circuiting of the traditional forces entry and efficiency growth and facili- Integrated Public Use Microdata Series generating the “flying geese” pattern tate information flows on markets and (IPUMS) developed by the Minnesota in which stages of the value chain are opportunities, making entirely new Population Center, which harmonizes passed down from advancing to lag- industries possible. census micro-data from around the ging economies. In this case there may •  Because information and com- world and offers a standardized set of remain a large global population who munication technology (ICT) capital occupational categories. will not be writing code and program- stocks are lower in developing coun- They first replicate Autor’s results ming robot routines anytime soon— tries, the displacement effects on jobs for France and the United States, but who will also not be inheriting directly affected by ICT adoption may where operators and assemblers as unskilled jobs. be lower. well as crafts show a decline over the •  The employment impact of past decade compared with elemen- William F. Maloney and Carlos Molina. 2016. technological progress depends on the tary occupations and more skilled “Are Automation and Trade Polarizing Developing relative product elasticity: if the fall categories. Analysis aggregating across Country Labor Markets, Too?” Policy Research in price arising from the labor savings the advanced economies confirms this Working Paper 7922, World Bank, Washington, more than proportionately increases pattern. DC. World Bank Research Digest 3 How Does Innovation Affect Employment in Developing Countries? Innovation boosts employment This unique data set, based on in- automation may actually displace la- by firms—though at a decreasing novation surveys implemented by the bor by reducing the employment elas- rate as they move toward the World Bank’s Enterprise Survey unit, ticity of product innovation when these is comparable across 53 countries. are introduced jointly. technological frontier The analysis uses a methodology that The results have important impli- I exploits the relationship between the cations for policy. Innovation policy, nnovation—the introduction of sales associated with the introduction when effective in leading to an in- new or substantially improved of innovations and changes in employ- crease in innovation activities and suc- products, processes or organiza- ment. In addition to expanding the cessful innovations, even through imi- tional structures—is key to driving the evidence to developing countries, the tation, can also be an important policy growth of firms. It increases their abil- paper explores in detail the impact of for increasing employment in the short ity to compete in international mar- process innovation and automation on run. This is especially so for those kets by improving their product mix employment as well as considering the countries furthest from the techno- and helps enhance their productivity impact in relation to the degree of nov- logical frontier, where generating new through the adoption of new technol- elty of the product innovation. sales due to innovation has the largest ogies and production processes. The results suggest that in lower- effect on employment. But for higher- While the benefits of innovation for and middle-income countries, and income countries, because of their productivity and firm growth are well especially in Africa, a dollar increase in greater ability to generate productivity established, little is known about the sales from innovative products is asso- gains in new products, these new sales short-term direct effect of innovation ciated with greater employment growth have a smaller effect on employment. on employment, especially in coun- than in high-income countries. This is These results, qualitatively similar to tries furthest from the technological mainly because innovations in lower- previous findings in advanced econo- frontier. An emerging empirical lit- income countries tend to be more mies, highlight a positive direct effect erature is documenting the economy- incremental—primarily small upgrades of innovation on the quantity of em- wide effects of new technologies on to existing products and processes— ployment—though at a decreasing rate job polarization, but there is little evi- and lead to smaller efficiency gains. as firms transition to the technological dence of the employment impact that The model estimates unity elasticities frontier. introducing new products and process- for most of these countries, suggesting es has on innovative firms. Introducing that if all products could be replaced new product or service lines can gener- by new or upgraded ones, the firm’s ate direct positive impacts on employ- overall level of employment would be ment. But introducing new processes at least as high as the previous level with enhanced and more modern tech- and would not decrease. However, for nologies can result in more efficient high-income countries in the sample, use of labor, leading to less additional especially in Eastern Europe and employment or even the replacement Central Asia, new sales attributable to of existing plant workers. innovation generate new employment Thus a critical question for policy but at lower rates, since the new or is whether innovation generates trade- upgraded products are more efficient offs between productivity increases in the use of labor. and employment growth. Determining The evidence suggests that prod- what the trade-offs might be (if any) is uct innovation is the main channel of critical, especially in developing coun- employment creation. Organizational tries facing the greatest needs to ab- innovation appears to have no impact sorb new entrants to the labor market on employment changes, whether it in formal and higher-productivity jobs. is implemented alone or with product A recent paper by Cirera and innovation. Process innovation also Sabetti addresses this gap and sheds appears to have no impact on employ- some light on the direct impact of ment, even when it takes the form of technological and organizational in- introducing automation. The main novation on firm-level employment effects of these types of innovation Xavier Cirera and Leonard Sabetti. 2016. “The growth in a sample of more than are likely on the quality of labor—skill- Effects of Innovation on Employment in Develop- 15,000 firms in Africa, South Asia, the biased technical change—rather than ing Countries: Evidence from Enterprise Surveys.” Middle East and North Africa, and on its quantity. However, the authors Policy Research Working Paper 7775, World Eastern Europe and Central Asia. find some support for the idea that Bank, Washington, DC. 4 World Bank ResearchDigest How Is the Internet Reshaping Work? implications? New technologies repre- sent both opportunities and challeng- es. More flexible work arrangements Evidence from Europe suggests that into smaller tasks, could explain these could help some sociodemographic the internet has led to a rise in part- trends. Vazquez and Winkler inves- groups facing barriers to participation time work and telecommuting tigate whether the internet has con- in the labor market, such as persons tributed to the rise of alternative work with disabilities and older workers. But T arrangements in Europe, hypothesiz- unless other barriers are also removed, he rise of alternative work ar- ing that if the internet is truly chang- technologies may not help them at all. rangements—the “gig” economy ing the labor market, we should see At the same time, technological —is a trending topic. It is often larger and more rapid changes among change poses challenges to social said that part-time and freelance posi- sectors more dependent on digital protection systems, designed during a tions are increasing while salaried full- technologies. time when a lifetime salaried job was time jobs are disappearing. The authors find evidence sup- the standard. Alternative work arrange- There is some empirical evidence porting their hypothesis. European ments may require a social protection for this. According to McKinsey, more countries that were early adopters of system where contributions and ben- than 24 million people are providing telecommunications reforms aimed efits are linked more to the individual services through sharing economy at increasing the availability and af- and not to the job. platforms in the United States and fordability of the internet experienced Finally, online platforms are gen- Europe. Using a new survey, Lawrence more dramatic labor market changes erating a process of creative destruc- Katz and Alan Krueger find that almost among internet-intensive sectors. The tion, where traditional jobs are being all the net job creation in the United earlier the reforms were implemented, replaced by new and more productive States since 2005 has been in alter- the faster the job creation in sectors ones. Policies to ease the transition native work arrangements. And in a that depend more on the internet to of displaced workers toward new jobs recent paper Vazquez and Winkler esti- conduct their operations. And workers will be crucial to smooth the disrup- mate that the share of workers in tradi- in internet-intensive sectors experi- tive effects of the internet on the labor tional work arrangements (permanent, enced more rapid growth in part-time market. full-time salaried jobs) has fallen dra- employment and telecommuting. Emmanuel Vazquez and Hernan Winkler. 2017. matically in most European countries These findings are consistent with “How Is the Internet Changing Labor Market since the early 2000s. the idea that internet adoption tends Arrangements? Evidence from Telecommunications Some scholars argue that the inter- to increase alternative work arrange- Reforms in Europe.” Policy Research Working net, by making it easier to break jobs ments in Europe. What are the policy Paper 7976, World Bank, Washington, DC. Big Data from Online Job Portals Online job portals generate big data fields of labor economics and work- One illustration is wage analysis useful in analyzing labor markets force skills development. Through an based on more than 50,000 job ad- and the demand for job skills empirical analysis of information gen- vertisements from daily postings in erated by Babajob, an online Indian 20 cities between 2011 and 2015. The E job portal, the authors highlight key analysis reveals a cyclical but upward conomists and other social sci- areas where such data could contrib- trend in wage offers. Except in 2012, entists are increasingly using ute to the development of labor market wage offers show a fall around early big data analytics to address policies and analytical knowledge, in- summer months and a rapid increase long-standing economic questions cluding observing job-search behavior, toward the end of the year. and complement existing information improving skills matching, and project- Another is text analysis based on sources. Big data produced by online ing demand for workforce skills. the words most frequently used in platforms can yield a wealth of infor- The authors’ analysis provides a describing required skills in job ad- mation. Online job portals, whether unique case study on labor market vertisements. Using word clouds to global (LinkedIn, Indeed, Monster, data analytics in a developing-country understand employers’ demands for CareerBuilder) or local, potentially context. It also reflects the diverse skills, the authors show that skill- and offer a rich and continuous stream of methodologies allowed by the Babajob qualification-related words appear labor market data that remains largely data. While online job portals are by more often in professional-level job untapped by policy makers. no means representative of the en- descriptions. For both professional In a new paper Nomura, Imaizumi, tire labor market, the granularity and and nonprofessional jobs, experi- Areias, and Yamauchi examine how uniqueness of the data they capture ence and communication are the most data from online job portals can be offer advantages over traditional sur- used for policy-relevant research in the vey instruments. (continued on page 10) World Bank Research Digest 5 The Motherhood Penalty and Female Employment in Urban India Urban women are unlikely to enter The authors anchor their analysis norms need to change, but in fact the labor market in India unless in the normative constructions of change in norms often follows strong the state, employers, and families motherhood in India, where mothers policy interventions—and as family have an exalted role and women who policy across the world has shown, weave a web of support are perceived as not fulfilling that role norms and behaviors respond well to I in the traditional sense are censured incentives. ndian women have been dropping both within and outside the home. The state could also signal a more out of the labor market since 2004, Moreover, child care is almost entirely enabling vision for women and moth- prompting a swath of empirical the responsibility of the mother. While ers in different ways. Publicly funded and theoretical work seeking to ex- Indian fathers are becoming increas- information campaigns that value plain why. One important fact is that ingly involved in child rearing, they are women as workers and project an women’s exit from the labor market perhaps more involved after the child image of child care as a shared re- has been driven largely by their with- starts school. There is little empirical sponsibility in the home are likely to drawal from agriculture, while urban evidence on the manner and extent to remove some of the guilt that women areas have remained unaffected. which women’s caregiving roles im- often experience when they leave chil- Cities and towns, having benefited pede their ability to undertake market dren behind to go out to work. Such from India’s legendary economic work. It may of course be possible that campaigns have been effectively used growth since the 2000s, might be women want to stay home with their in India and elsewhere to encourage expected to serve as hubs for jobs children, but the reality is likely to be literacy, to achieve health outcomes for men and women alike. Moreover, more complex than that. such as immunization or family plan- women’s educational attainment has Using pooled data from six rounds ning, or to enroll participants in social risen over time and they are having of the National Sample Survey, the programs such as conditional cash fewer children, aspects that typically authors examine the effects of having transfer schemes. enhance women’s labor force partici- a young child on mothers’ employ- Both the state and the private sec- pation in other settings. Yet women’s ment in urban India over the period tor in India are beginning to gingerly employment in urban areas has always 1983–2011. They focus on women in experiment with options for attracting been very low, never exceeding 25 per- the age group 25–55, a range in which more women into the labor market. cent. Some observers have argued that the younger women have completed A better understanding is needed of women are unable to get jobs com- higher education and the older ones what these initiatives are, how well mensurate with their qualifications are too young to retire. They also look they work, and how they can be scaled and aspirations, others that cultural at household structure by factoring in up. Because unless the state, employ- norms of status and seclusion inhibit the effects of other household mem- ers, and families weave a web of sup- women’s entry into the labor market. bers on women’s labor supply. port, women are unlikely to enter the Despite the conundrum of low and The authors find that while the labor market. static female labor force participation responsibility of child bearing has less- in urban areas amid rising prosperity, ened because of the secular decline the empirical work has focused mostly in fertility, the onus of child rearing on the country as a whole or on ru- appears to have increased: having a ral areas. A recent paper by Das and young child in the household reduces Žumbyte focuses on urban women in urban mothers’ likelihood of being India, looking at how motherhood af- employed, and this effect has intensi- fects their chances of being employed. fied over time. The presence of mother The authors are motivated by the fact surrogates, in the form of other women that marriage on its own, and the in the home, somewhat attenuates this presence of a young child in the fam- negative effect. But overall, the moth- ily in particular, are known to depress erhood penalty is an important factor women’s labor force participation in in women’s decisions to enter or stay India, as in several other countries. in the labor market. They draw on the demographic lit- What does this mean for policy erature on “maternal role incompat- aiming to attract more women into the ibility,” which posits that the roles of labor market? Providing affordable, mother and worker may be inherently high-quality child care and ensur- Maitreyi Bordia Das and Ieva Žumbyte. 2017. incompatible, and argue that there is a ing that women have access to safe “The Motherhood Penalty and Female Employ- “motherhood penalty” for urban Indian transportation are of course quick ment in Urban India.” Policy Research Working women. wins. It is also fashionable to say that Paper 8004, World Bank, Washington, DC. 6 World Bank ResearchDigest How Some Female Entrepreneurs in Africa Enter Male-Dominated Sectors Female entrepreneurs in Africa Industry Association; and a qualitative not know that they are making less can earn more in male-dominated and quantitative survey of women op- money than those who cross over. sectors. What explains why some erating in male-dominated sectors as Indeed, among those making less well as women operating in traditional than the crossovers, about 75 percent cross over while others do not? sectors. believe that they make as much as I The analysis examines potential female entrepreneurs in male-domi- n Africa businesses owned by differences in household demograph- nated sectors. women have a lower performance ics, wealth status, financial access, In addition, the influence of male on average than those owned by entrepreneurial characteristics, ex- role models and exposure to a sec- men—on size, profitability, survival posure, and business challenges to tor through family and friends appear rate, and growth trajectory. These gen- understand the factors associated with to be critical in helping women cir- der differences can be explained by women crossing over. These differ- cumvent or overcome the norms that the different strategic choices made ences are interpreted as correlations undergird occupational segregation. by male and female entrepreneurs. between the outcome of interest and Women who reported having a male Among these strategic choices is the a woman’s likelihood of crossing over, role model in their youth are 20–28 sector in which they operate, a con- though the direction of causality can- percent more likely to be a crossover. sistent and important determinant of not be confirmed. And women who do not cross over gender differences observed in firms’ The analysis reveals two signifi- are more likely to have been intro- performance and growth. Among cant findings. First, firms owned by duced to traditionally female sectors African entrepreneurs, women are women who cross over are about by mothers and teachers. Moreover, concentrated in catering, retail trade, three times as profitable on aver- once women engage in a tradition- garments, hairdressing, and other age as those owned by other women. ally female sector, they are unlikely to services, while men engage in a wider Second, businesses owned by women make the switch to a male-dominated range of sectors, including construc- in male-dominated sectors are just as one. Thus early influence by a male tion and manufacturing. Do these dif- profitable as those owned by men in role model is important in shaping ferences arise because most women those sectors. So there is a clear ratio- women’s professional path to a more prefer to work in certain sectors, or nale for crossing over. But what allows profitable sector. because they are constrained in their certain women to cross over, and what This analysis suggests that operat- choice of profession? prevents more women from doing so? ing in a male-dominated sector can A recent paper by Campos, While the women who cross over lead to higher profits, indicating that Goldstein, McGorman, Boudet, and may differ in some dimensions of their encouraging female entrepreneurs to Pimhidzai looks at this question in personality, they are not more likely to enter such sectors could be an im- Uganda, where only 6 percent of fe- consistently score higher on the skills portant way to help improve women’s male entrepreneurs operate in male- that would aid them in becoming economic status. It also provides dominated sectors (defined as sec- successful entrepreneurs. Analysis of avenues for piloting and testing pro- tors where more than 75 percent of scores on two cognitive tests measur- grams to address occupational gender enterprises are male-owned), while 34 ing working memory and fluid intelli- segregation through a combination of percent of male entrepreneurs have gence—indicators of an entrepreneur’s information, mentoring, and exposure businesses in these sectors. The study ability to understand and solve com- to the sectors. investigates differences in sectoral plex problems—shows that these are choice between men and women and not important in explaining women’s examines the role of factors that may entry into male-dominated sectors. help female entrepreneurs cross over The authors also considered a variety into male-dominated sectors or hinder of noncognitive tests to gauge innate them from doing so. entrepreneurial spirit, which may be The study uses a mixed-methods useful in predicting an entrepreneur’s approach to investigate the differenc- success over time. Women who cross es in enterprise size and performance over did not score consistently bet- between women who cross over to ter on these tests either. Nor did they male-dominated sectors and those have more years of education or great- Francisco Campos, Markus Goldstein, Laura who do not. It draws on two data sets: er access to finance. McGorman, Ana Maria Boudet, and Obert Pim- a baseline survey, conducted for an Two things that do seem to matter hidzai. 2015. “Breaking the Metal Ceiling: Female impact evaluation, of male and female are access to information and psycho- Entrepreneurs Who Succeed in Male-Dominated entrepreneurs in urban Uganda who social factors. Women who remain in Sectors.” Policy Research Working Paper 7503, are part of the Katwe Small-Scale female-dominated sectors simply do World Bank, Washington, DC. World Bank Research Digest 7 Can Wage Subsidies Boost Employment in the Wake of an Economic Crisis? Evidence from Mexico suggests that To get a better understanding of compared how employment evolved in wage subsidies provided during these issues, Bruhn recently studied these industries during and after the the recent economic crisis helped a wage subsidy program in Mexico. program. Under this program firms in certain du- The results show a positive but employment recover faster rable goods manufacturing industries not statistically significant effect of U were eligible to receive wage subsidies the wage subsidies on employment nemployment often rises during during the period from January 2009 to during the program, ranging from 5.7 an economic crisis, and policy August 2009. To receive a subsidy firms percent to 13.2 percent, depending on makers take a range of actions were not allowed to fire more than a the econometric specification. The size to try to mitigate this increase. During third of their workforce as a percentage of the effect increased to a statistically the 2008/09 crisis, for example, 22 of the drop in sales they had experi- significant 24 percent after the pro- countries used some form of wage enced during the economic crisis; for gram ended, and the results indicate subsidy program to promote employ- example, if sales dropped by 30 per- that employment recovered faster in ment retention. Many studies have cent, they could not fire more than 10 eligible industries than in ineligible looked at the effect of wage subsidies percent of their workforce. ones. Indeed, in eligible industries em- on employment in noncrisis times, The total funding disbursed ployment was back at precrisis levels with mixed findings. But there is not through the program was about $63 in 2011, while in ineligible industries it much evidence on whether wage sub- million, with a median of $92,000 per was still not back at precrisis levels in sidies can raise employment in the firm. In practice, firms typically re- 2013 (figure 1). wake of a crisis. ceived the subsidy many months after The impact on firms’ employment Conceptually, wage subsidies dur- they limited layoffs, in part because levels is larger than other estimates in ing a crisis make sense. Layoffs could reviewing applications took some time. the literature. This suggests that wage slow the recovery, as rehiring and Most funds were approved starting in subsidies may be particularly effective training workers may be costly for June 2009. during an economic crisis, since they firms. Facing lower demand for their Bruhn obtained monthly admin- can be paid for a short time and have products, however, firms may not have istrative data on employment at the lasting effects on employment. the financial means to keep paying industry level from the Mexican Social At the same time, the lack of a sta- their workers, particularly in the pres- Security Institute, covering nine years tistically significant effect on employ- ence of credit constraints, which are before and four years after the pro- ment during the program suggests often exacerbated during a crisis. This gram’s implementation. She used that the firing restrictions imposed by is where wage subsidies come in. But propensity score matching to construct the program were not binding. That is, do these subsidies really cause firms groups of eligible and ineligible du- even in the absence of the program, to retain workers they otherwise would rable goods manufacturing industries firms that received the subsidy would not have retained? Ultimately, we just that show statistically identical pre- not have fired more than a third of don’t know. program trends in employment, then their workforce as a percentage of their drop in sales during the crisis. Instead, the program’s effect seems Figure 1. Average Employment in Matched Industries in Mexico, 2004–14 to have operated through the pay- Average log employment ment of subsidy funds. These funds 10.2 may have provided liquidity for rehir- Ineligible industries ing workers, allowing firms to take 10.0 Eligible industries advantage of market opportunities at a time when other funding sources 9.8 were scarce. Indeed, the time when employment started to increase again 9.6 in eligible industries (mid-2009, as shown in figure 1) coincided with the 9.4 time when the subsidy funds started to 9.2 be paid out. 9.0 2004m1 2006m1 2008m1 2010m1 2012m1 2014m1 Miriam Bruhn. 2016. “Can Wage Subsidies Boost Month Employment in the Wake of an Economic Crisis? Note: The figure shows data from the Mexican Social Security Institute. The two vertical lines denote Evidence from Mexico.” Policy Research Working the start and end of the wage subsidy program (January and August 2009). Paper 7607, World Bank, Washington, DC. 8 World Bank ResearchDigest Understanding the Effects of the World’s Largest Workfare Program India’s rural workfare program has and agricultural production struc- covered by government-imposed floor boosted agricultural productivity tures. They find that in the short term prices, and an increase in the intensity as well as increasing wages and NREGS led to a marked increase in of cultivation (the number of seasons agricultural wages and higher levels in which crops are grown during a employment levels of nonfarm casual work and on-farm year). Some of these effects may be ex- A self-employment. The program trig- plained by the program’s effects in in- s the world’s largest workfare gered more intensive use of irrigation creasing farmers’ liquidity. In addition, program, India’s Mahatma and greater diversification of crop the rehabilitation of infrastructure Gandhi National Rural portfolios, especially by small farmers. and construction of new small-scale Employment Guarantee Scheme It also increased productivity, largely water conservation structures may (NREGS) has attracted much atten- by alleviating liquidity constraints and have helped support more intensive tion. Yet its impacts on agriculture improving access to insurance. land use, particularly the planting of a have been relatively neglected. A re- Results suggest that the program second or third crop beyond wheat or cent paper by Deininger, Nagarajan, increased the wage significantly with- paddy. and Singh addresses this gap by fo- out crowding out private employment. Evidence on how workfare affects cusing on the program’s effects on ag- Most of this increase can be attrib- agricultural productivity matters not ricultural productivity as well as labor uted to higher wages in agriculture, only for a better understanding of the market outcomes. which affected men and women about NREGS intervention. It also has im- The program offers unskilled em- equally. Women also experienced an plications for the broader debate on ployment, for up to 100 days a year per increase in nonagricultural wages. the comparative merits of this type household, in projects to provide local Analysis of the extent to which of approach. Workfare programs rely productivity-enhancing infrastructure. wage changes affected labor allocation on work requirements as a screening Wages are set by statute, at rates that points to insignificant impacts on ag- device based on the assumption that are equal for men and women and, it ricultural wage work in the aggregate. such screening makes these programs is hoped, not attractive enough to pre- While a significant increase in nonfarm a more cost-effective tool for social vent effective self-targeting. casual work may be attributed to the protection than, say, unconditional Workfare programs like NREGS can aggregation of NREGS and other work, cash transfers. But such programs affect agricultural productivity through there is also evidence of a program- would be less desirable if they dis- several channels. By using labor, they induced increase in on-farm self- placed existing workers rather than can affect wage rates and employment employment. But while men increased generating new jobs, if supply-side levels in the short term and produc- their labor supply to the agricultural constraints were to reduce the ef- ers’ choice of technology and the sector, women shifted away from farm fectiveness of self-targeting, or if the capital intensity of production in the to nonfarm employment and to some work done had no productive value. medium to long term. By providing extent salaried work. Concerns have been raised about implicit insurance against downside Increases in labor supply to the NREGS in each of these areas. While risk, in the form of predictable wage nonfarm sector were concentrated further study of the extent to which payments, they may allow poor farm- among landless and small to medium- such effects persist in the longer term ers to increase investment or adopt size farmers, suggesting effective is needed, the authors’ findings sug- crop portfolios with higher risk-return self-targeting. Increases in labor sup- gest that the potential productivity profiles. And by constructing or im- ply to agricultural self-employment benefits of workfare could be an im- proving local infrastructure, they may emerged only for small and, to some portant aspect to take into account in increase agricultural productivity and degree, medium-size farmers, possi- evaluating the impact or desirability of thus boost returns to land and labor, bly because some NREGS investment such programs. the main assets of small farmers and can be performed on farms. There is the rural poor. also some evidence of a reduction in The authors use unique panel data nonfarm self-employment and an in- covering the same households in crease in salaried work by the largest 1999/2000 (before the program) and landowners. 2007/08 (after some had received the Results indicate that the program program), complemented by informa- had a significant impact on agricultural Klaus Deininger, Hari K. Nagarajan, and Sudhir tion on the program’s implementa- productivity, in part by supporting di- K. Singh. 2016. “Short-Term Effects of India’s tion in a subset of villages. Using versification. Evidence suggests that Employment Guarantee Program on Labor these data and a robust method, they the program led to greater use of ma- Markets and Agricultural Productivity.” Policy assess the program’s short-term im- chinery and fertilizers, a shift beyond Research Working Paper 7665, World Bank, pacts on rural wages, labor demand, rice and wheat toward riskier crops not Washington, DC. World Bank Research Digest 9 How Do Cities in Ethiopia Create Jobs? A new study looks at how Ethiopian small to medium to large? The statis- Mekele–Dessie, Harar–Dire Dawa, and cities can create the right conditions tics suggest that rates of graduation Addis Ababa–Hawassa) would help for entrepreneurs, firms, and are extremely low: of all small enter- strengthen linkages within and be- prises (21–50 employees), only 6.5 per- tween cities. industries to become competitive cent graduate over time to large-enter- •  Identify and invest in growing I prise status (101–1,000 employees). the missing middle of companies and n Ethiopia, which aims to become Why are some firms, industries, and jobs. To increase opportunities for a middle-income country by 2040, cities able to grow employment, while wage employment, cities must identify industrialization appears to be the others seem to lag behind? Using an and invest in the factors that allow main instrument at the government’s econometric model, the author models existing firms to expand and to cre- disposal for creating employment and job growth at the firm level as a func- ate new jobs and that encourage firms growth. In a recent paper Mukim fo- tion of economic geography (agglom- to graduate, such as better-regulated cuses on understanding whether and eration), business environment, and taxes and access to infrastructure. to what extent firms in light manu- infrastructure. These variables include •  Target interventions accord- facturing industries in Ethiopia are factors that are common to the city ing to the constraints and needs of driving sustained job creation, and to as well as factors that are common to different types of firms. Access to fa- what extent their immediate spatial particular industries within the city. cilities and services would help small environment—that is, cities—deter- Agglomeration includes localization businesses create linkages and grow. mines their success over extended and industrial diversity, business en- Improvements in local business licens- periods. The World Bank Group’s 2015 vironment includes licensing fees and ing and permitting systems would help flagship report Competitive Cities for Jobs taxes, and infrastructure includes the medium-size enterprises grow. And and Growth found that if cities support costs of transport and electricity. better access to land and connective their firms and industries in creating What factors appear to drive infrastructure would help large enter- jobs, millions of additional jobs could sustained increases in jobs across prises flourish. be created every year. Are Ethiopian Ethiopian cities? The answer varies Ethiopia has a small yet significant cities creating the jobs the country somewhat depending on the outcome window of opportunity to exploit its needs—and if so, how? of interest. For job creation over time, demographic dividend. Its labor force Descriptive statistics reveal that the important factors are intra-industry has doubled in size over the past two capital, Addis Ababa, accounts for the clustering, low taxes and low licensing decades, and a very large youth bulge dominant share of firms and employ- fees, and access to better electricity will continue to enter the labor market ment across the country. But firms in and transport services. For the emer- over the next 20 years. Many of these new, emerging cities (such as Adwa, gence of “job superstars,” enterprises young people will be moving to cities Debreziet, and Sebeta) are the most that double their employment every in search of employment. Urbanization employment-intensive. And while four years or so, industry concentration could become one of the most impor- fast-growing industrial towns (such and low licensing fees are again im- tant catalysts of economic growth and as Hawassa, Mekele, and Nazareth) portant factors. This is particularly true job creation over the next few years, attract many new firms, these new for small and medium-size enterprises, even decades. Cities should aim to entrants create fewer jobs than their for which providing a boost to growth create the conditions for entrepre- counterparts in other Ethiopian cities, would help in graduating into large en- neurs, firms, and industries to become suggesting greater capital intensity in terprises. Finally, if the outcome of in- competitive. production. terest is to increase employment at the What is worrying, though unsur- city level, intra-industry clustering and prising, is that large firms account for industrial diversity matter. Keeping most of the existing employment in transport costs low also helps. Ethiopia. Large firms also create 97 How can public policy draw on percent of new jobs. And indeed it is these findings to improve economic incumbents, not entrepreneurs, that outcomes for Ethiopian cities? account for much of the sustained job •  Identify and invest in cities’ creation—new entrants create but also competitive advantage. Secondary and destroy many jobs. There is also huge smaller cities are seeing the growth variation in net job creation across of naturally forming industrial clus- cities. ters, while Addis Ababa’s advantage If large, existing firms are creat- seems to lie in its diverse industrial Megha Mukim. 2016. “How Do Cities in Ethio- ing most of the sustainable jobs, how structure. Investments in growth cor- pia Create Jobs?” Policy Research Working Paper easily can firms graduate in size from ridors (for example, Bahir Dar–Gondar, 7866, World Bank, Washington, DC. 10World Bank ResearchDigest What Is Behind the Decline in Wage Inequality in Latin America? Latin America’s most experienced distribution) have risen by more than contrast, more than half the reduction workers have seen their experience 50 percent in real terms. This is sig- in labor income inequality observed premium over younger workers fall nificantly more than the 15 percent in the region since the early 2000s growth among higher-paid workers stemmed from a reduction in the vari- by almost half since 2002 (those in the top decile of the earnings ance of the earnings of workers with L distribution) and also more than the similar characteristics—what is known atin America has been charac- 32 percent growth experienced in the in the literature as residual earnings terized as having an excess of median of the distribution. In contrast, inequality. This fact calls for further inequality for being a middle- the increase in labor income inequality investigation, particularly to clarify the income region. Indeed, the region during the 1990s was supported by a extent to which this outcome has been ranks as the second most unequal in reduction in real earnings at the bot- due to changes in the composition of the world, right behind Sub-Saharan tom of the distribution, while the rest the labor force across the region. Africa. But while income inequality in of the distribution showed no change Latin America remains substantial, in real earnings or slightly positive the region is also known to have ex- growth. Carlos Rodríguez-Castelán, Luis F. López-Calva, perienced a turning point in the early Results also highlight a steady Nora Lustig, and Daniel Valderrama. 2016. 2000s, shifting from a trend of slightly decline in the education premium in “Understanding the Dynamics of Labor Income In- increasing inequality during the 1990s Latin America since the early 2000s equality in Latin America.” Policy Research Work- to one of steady decline. This repre- that has been driven by greater growth ing Paper 7795, World Bank, Washington, DC. sents a sharp contrast with what has in labor earnings among less well- occurred in other developing regions, educated workers than among workers particularly in Europe and Central Asia with a high school or college educa- and in East Asia and the Pacific, where tion. This drop in the education pre- (continued from page 4) income inequality has been on the mium has been coupled with a steady rise since the early 2000s. fall in the experience premium. This Big Data from Online A vast literature exists on the rise and fall of total income inequality decline, which has been accelerating since the early 2000s, is a novel fact Job Portals across Latin America. In sharp con- for the region and has also been the trast is the scarce evidence about the aspect of the paper receiving the most frequently used words. Keywords re- facts behind changes in labor income attention. The most experienced work- lated to language skills and customer inequality across the region. A recent ers have seen their experience premi- care appear more often in advertise- paper by Rodríguez-Castelán, López- um with respect to younger workers fall ments for nonprofessional jobs; those Calva, Lustig, and Valderrama aims to by almost half. Indeed, after 2002 most related to problem-solving, leadership, fill this gap in knowledge. The study countries in the region experienced analytical skills, work ethic, reliability, takes advantage of harmonized data rapid reductions in these premiums. creativity, and personality attributes for 17 countries in Latin America The study also provides evidence tend to appear more often in profes- (which cover around 90 percent of that the gender wage gap has narrowed sional job descriptions. the region’s population) starting in consistently since the mid-1990s but Big data from online job portals are the early 1990s. Through analysis of was almost stagnant during the first not only very frequent and dense; they these data, it provides an overview of decade of the 2000s. This suggests a also present a greater variety of infor- the main factors that could be at play weak relationship between the gender mation and a more intense degree of in the trend reversal in labor income earnings gap and the trend reversal in granularity than existing information inequality observed across the region wage inequality. Correspondingly, the sources. This case study shows how since the early 2000s, as the unweight- urban-rural earnings gap widened in big data analytics has opened new av- ed average Gini coefficient of wage in- the 1990s but then narrowed sharply enues for objectively monitoring work- equality fell from 0.473 in 2002 to 0.410 during the 2000s, an outcome closely force skills demand with a wide array in 2013. related to what happened in overall of applications. The analysis shows that the decline labor income inequality. in labor income inequality was sup- In addition, the study finds that the ported by a substantial expansion in narrowing in the earnings dispersion Shinsaku Nomura, Saori Imaizumi, Ana Carolina real hourly earnings at the bottom of among workers with different observ- Areias, and Futoshi Yamauchi. 2017. “Toward the distribution. In particular, since able attributes (education, experience, Labor Market Policy 2.0: The Potential for Using 2002 the labor incomes associated gender, and location), though impor- Online Job-Portal Big Data to Inform Labor Mar- with relatively low-paying jobs (those tant, still accounts for less than half ket Policies in India.” Policy Research Working in the bottom decile of the earnings the reduction in earnings inequality. In Paper 7966, World Bank, Washington, DC. World Bank Research Digest11 Global Talent Flows Agglomeration factors drive the importance in economic growth, pull and regions within destination coun- patterns in high-skilled migration— factors generated by skill agglomera- tries. The United States has significant and reinforce the inequality in the tion, lower transportation and commu- concentrations of high-skilled migrants nication costs, and the rising number in Boston, New York City, and Seattle global competition for skills of international students. as well as California. These spatial H While OECD countries account for concentrations are even sharper in ighly skilled workers play a star- less than a fifth of the world’s popu- scientific fields and are also present ring role in today’s knowledge lation, they host two-thirds of high- globally. economy. They make excep- skilled migrants. Among OECD desti- What about origin countries? Many tional direct contributions, including nations the distribution is even more have limited educational capacities breakthrough innovations. As teach- skewed. Four English-speaking coun- and fiscal resources to train workers or ers, policy makers, and entrepreneurs tries—the United States, the United to replace those who have emigrated. they guide the actions of others. They Kingdom, Canada, and Australia—are Countries with particularly high emi- propel the knowledge frontier and the chosen destinations for nearly 70 gration rates of high-skilled workers spur economic growth. In this pro- percent of high-skilled migrants to the to OECD destinations in 2010 tended cess the mobility of skilled workers, OECD. The United States alone hosts to be small, low-income countries and within and across national borders, close to half of all high-skilled mi- island states, such as Guyana (93 per- becomes critical to enhancing pro- grants to the OECD. The attractiveness cent) and Haiti (82 percent). There is ductivity. Using newly available data, of these destination countries has led a strong inverse relationship between a recent paper by Kerr, Kerr, Özden, others, such as France, Germany, and country size and high-skilled emigra- and Parsons reviews the landscape of Spain, to increase their policy efforts. tion rates. These movements of high- global talent mobility and discusses Nevertheless, the volume of skilled skilled people away from certain small the causes and consequences of high- migration to these four countries, and low-income countries have raised skilled migration. coupled with the asymmetry in the controversies about “brain drain.” Much attention has been paid to concentration of leading universities, Agglomeration economies explain understanding the worldwide distribu- high-tech firms, and research centers, the high-skilled migration patterns. tion of human capital and how global implies that the global competition for The presence of high-skilled people migration flows further tilt the deck skills will continue to be unequal. in a location—whether natives or im- against poor countries. The migration Agglomeration of talent is even migrants—increases the incentives for patterns we see today are the result starker at the highest levels. The Nobel additional high-skilled people to move of a complex tangle of firms and other Prizes in Chemistry, Medicine, Physics, there because of a wide range of posi- employers pursuing scarce talent, and Economics provide powerful ex- tive externalities. At the core of this governments trying to manage these amples. Since World War II more than process is trade in knowledge services flows through policy, and individuals 65 percent of these Nobel Prizes have provided by high-skilled people. Skill seeking their best options given the been awarded to academics associ- clusters allow better technology ex- constraints imposed on them. The ated with U.S. institutions, only half of changes, deeper labor market special- central outcome, however, is clear: the whom were born in the United States. ization, and stronger complementary flows of high-skilled migrants are very Of all these Nobel Prizes, around a inputs. concentrated, both within and across third have gone to immigrants, more These agglomeration factors and national borders. than half of whom were affiliated with their implications for economic growth Overall rates of international mi- U.S. institutions. make the economics of high-skilled gration have been hovering around Many host countries end up with migration quite different from those 3 percent over the past 60 years. But higher concentrations of high-skilled of low-skilled migration. Moreover, beneath this perceived stability are immigrants in particular occupations. the critical role of actors like firms and certain strong asymmetric patterns, For example, immigrants account universities remains substantially un- especially with respect to human for some 57 percent of scientists in explored relative to their importance. capital. There were about 28 million Switzerland, 45 percent in Australia, The data necessary to analyze these high-skilled migrants (those with at and 38 percent in the United States. important patterns are just coming on- least one year of tertiary education) in Foreign-born individuals made up 27 line, and the future research potential OECD countries in 2010, reflecting an percent of all physicians and surgeons is immense. increase of nearly 130 percent since and more than 35 percent of current 1990. This exceptional rise is the result medical residents in the United States of several forces, including increas- in 2010. Sari Kerr, William Kerr, Çaglar Özden, and Chris ing efforts by policy makers to attract Stark inequalities in the concentra- Parsons. 2016. “Global Talent Flows.” Journal of human capital as they recognize its tions of talent also exist across cities Economic Perspectives 30 (4): 83–106. 12World Bank ResearchDigest (continued from page 1) Recent Policy Research large (38 percentage point) increase Justifying these programs on a cost- Working Papers on Labor in employment while the subsidy was benefit basis typically requires assum- Market Issues in effect. But when the subsidy ended, ing that impacts persist for much lon- most of these youth left the firms ger than typically measured and that 7801 Impact of Social Assistance on Labor Market and there was no lasting impact on formal employment is to be especially Mobility: The Case of Turkey employment. valued. Deniz Alcan, Raif Can, Temel Taskin, and William Wiseman A third set of studies considers Yet governments continue to 7808 Are Minimum Wages and Payroll Taxes a programs designed to reduce informa- face pressure to be seen to be doing Constraint to the Creation of Formal Jobs in Morocco? tion and search frictions and to bet- something to help people find jobs. Diego F. Angel-Urdinola, Abdoul Gadiry Barry, ter match workers and firms. These Thus unless better alternatives can be and Jamal Guennouni 7835 Deepening without Broadening? Jobs in include job fairs, transport subsidies, found, the lack of empirical evidence Ghana’s Private Sector public information services, and certi- for the effectiveness of many tradition- David C. Francis and Maddalena Honorati 7879 Are There Skills Payoffs in Low- and Middle- fication of hard and soft skills. These al programs is unlikely to be enough Income Countries? Empirical Evidence Using services tend to be much less expen- to cause them to be abandoned. The STEP Data Alexandria Valerio, Maria Laura Sanchez sive than training or wage subsidies, paper discusses several potential ap- Puerta, Namrata Tognatta, and Sebastian costing $25 or less per person assisted. proaches, including working more on Monroy-Taborda 7888 Enforcement Capacity and the Impact of These lower costs certainly lower policies to foster labor demand, and Labor Regulation: Evidence from the Russian the bar on the treatment impacts helping workers access different labor Federation Alvaro Gonzalez, Siddharth Sharma, and Hari needed in order for the programs to markets by removing barriers to their Subhash pass cost-benefit tests. But of 10 differ- moving across job sectors and to dif- 7924 Labor Drops: Experimental Evidence on the Return to Additional Labor in ent interventions, only one (in India) ferent locations. Such efforts also need Microenterprises was found to have a significant impact rigorous evaluation to determine their Suresh De Mel, David McKenzie, and Christopher Woodruff on employment, increasing it by 2.4 effectiveness. 7927 Job Quality and Poverty in Latin America percentage points over three years. The Peter Brummund, Christopher Mann, and average impact across studies is 2.7 Carlos Rodríguez-Castelán 8000 Measuring Skills in Developing Countries percentage points. Moreover, none of Rachid Laajaj and Karen Macours the studies finds a significant impact 8015 An Employment Guarantee as Risk Insurance? Assessing the Effects of the on labor earnings. NREGS on Agricultural Production Decisions The conclusion is that most tradi- Esther Gehrke 8042 Economywide and Sectoral Impacts on tional active labor market programs Workers of Brazil’s Internet Rollout have had at best only modest impacts Mark A. Dutz, Lucas Ferreira Mation, Stephen D. O’Connell, and Robert D. Willig on employment: for every 100 people 8053 The Influence of Non-Cognitive Skills on that such programs aim to help, the Wages within and between Firms: Evidence typical impact is that two to three from Bangladesh’s Formal Sector Shinsaku Nomura and Samik Adhikari might gain employment, and this ef- David McKenzie. 2017. “How Effective Are Active fect is seldom statistically significant. Labor Market Policies in Developing Countries? These impacts are much smaller than A Critical Review of Recent Evidence.” Policy Re- Working Papers can be downloaded at http://econ.worldbank.org policy makers and program partici- search Working Paper 8011, World Bank, Wash- To download the World Bank Research E-Newsletter, pants have expected in many cases. ington, DC. go to http://econ.worldbank.org/research_newsletter The World Bank Research Digest is a quarterly publica- The Research Digest is financed by the Bank’s Editorial Committee: Chorching Goh (managing editor), tion disseminating findings of World Bank research. Research Committee and managed by DECDP, the Aslı Demirgüç-Kunt, and Shiva S. Makki. Editor: Alison The views and interpretations in the articles are those research support unit of the Development Economics Strong; production: Roula Yazigi. For information or of the authors and do not necessarily represent the Senior Vice Presidency (DEC). The Research Digest is free subscriptions, send email to research@worldbank. views of the World Bank, its Executive Directors, or the not copyrighted and may be reproduced with appropri- org or visit http://econ.worldbank.org/research_digest. countries they represent. ate source attribution. The World Bank 1818 H Street, NW Washington, DC 20433, USA Printed on Recycled Paper