68944 A gender (R)evolution in the making Expanding women’s economic opportunities in Central America: A decade in review A GENDER (R)EVOLUTION IN THE MAKING? EXPANDING WOMEN’S ECONOMIC OPPORTUNITIES IN CENTRAL AMERICA: A DECADE IN REVIEW         [2]    THE WORLD BANK  EXECUTIVE SUMMARY.................................................................................................................................. 5  1  THE EVOLUTION OF LABOR MARKET PARTICIPATION AND EARNINGS IN CENTRAL AMERICA .......................................................................................................................................................... 9  1.1  THREE KEY TRENDS FROM THE DECADE (1997-2006).................................................................................. 10  1.1.1  TREND 1: Insignificant Changes in Female Labor Force Participation ............................................................ 10  1.1.2  TREND 2: For the First Time, Women Earn More than Men....................................................................... 12  1.1.3  TREND 3: The Economic Contribution of Women has Increased .................................................................... 13  1.2  STUDY ROADMAP .......................................................................................................................................... 14  2  UNDERSTANDING FEMALE LABOR FORCE PARTICIPATION................................................... 16  2.1  WEAK LABOR MARKET DEMAND, CONCENTRATION AND SEGREGATION .................................................... 19  2.2  EDUCATION KEY BUT NOT ENOUGH ............................................................................................................ 24  2.3  FAMILY DYNAMICS—FEMALE HEADSHIP AND CHILDREN ............................................................................ 31  3  EXPLAINING CHANGES IN GENDER EARNINGS GAPS .............................................................. 36  3.1  DECOMPOSING THE EARNINGS GAP ............................................................................................................. 36  4  HETEROGENEITY AND THE POOR: REVISITING FLFP AND EARNING GAPS ...................... 42  4.1  HETEROGENEITY IN FLFP............................................................................................................................ 42  4.2  HETEROGENEITY IN EARNINGS GAPS ........................................................................................................... 46  5  UNDERSTANDING TRENDS THROUGH POLICY-BASED EVIDENCE ...................................... 49  5.1  INCREASING FEMALE LABOR MARKET PARTICIPATION ................................................................................. 49  5.1.1  Skill Formation and FLFP ...................................................................................................................... 50  5.1.2  Expanding Child Care: Present and Future Generational Gains....................................................................... 55  5.2  INCOME GENERATION AND PROMOTING ENTREPRENEURSHIP .................................................................... 59  5.3  CHANGING GENDER NORMS ........................................................................................................................ 65  5.3.1  Female Migration and Intra-household Resource Allocation .............................................................................. 66  5.3.2  Changing Intra-household Resource Allocation via Cash Transfers..................................................................... 71  5.3.3  Inducing Behavioral Changes Through Social Interactions................................................................................. 71  5.4  CONCLUSIONS ............................................................................................................................................... 74  6  THE ROAD AHEAD .............................................................................................................................. 75  BIBLIOGRAPHY ............................................................................................................................................. 80  APPENDIX A1: ADDITIONAL TABLES ....................................................................................................... 87  APPENDIX A2: WAGE DECOMPOSITION METHODOLOGY.................................................................. 94  [3]    Acknowledgments “A Gender (R)evolution in the Making? Expanding Women’s Economic Opportunities In Central America: A Decade In Reviewâ€? was prepared by Renos Vakis (Task Team Leader), Ana Maria Munoz Boudet and Barbara Coello from the Poverty and Gender Unit of the Latin America and the Caribbean Region (LCSPP). Elizabeth Katz (University of San Francisco) and Florencia Castro-Leal (LCSPP) provided additional contributions and suggestions, while Alejandro Hoyos and Hugo Ñopo from the Inter-American Development Bank led the analysis on gender wage gaps. The report was supervised by and benefited from comments, suggestions and advice from Laura Frigenti (Country Director for Central America), Louise Cord (Sector Manager, LCSPP), and Humberto Lopez (Sector Leader, PREM). The peer reviewers were Trine Lunde (Gender and Development Network), Carolina Sanchez (Poverty Network) and Manohar Sharma (East Asia, Poverty). Laura Chioda (Chief Economist Office, Latin America), Pedro Olinto (LCSPP) and Rodrigo Garcia Verdu (LCSPP) provided valuable contributions and comments. Ane Perez Orsi de Castro, Lucy Castro and Anne Pillay (all LCSPP) provided excellent support to the team. Finally, the team is grateful to Eliana Rubiano (LCSPP) and Matthias Schmidt (LCSPP) for support with the data preparation. [4]    EXECUTIVE SUMMARY Central America made considerable progress during 1997-2006: stronger economic and political stability, overall positive albeit modest economic growth and improvements in socio-economic outcomes. Nonetheless, the panorama for women in the region, and in particular their ability to work and generate income is mixed. Labor force participation of women in Central America remained unchanged over the decade, with one in every two women of working age still not participating economically. This is lower than the rest of Latin America, a region with the lowest labor force participation in the world, although it has improved significantly in recent years. And yet, women in the region have made major advances over the period in closing the gender earnings gaps and increasing their ability to earn income. Women’s contribution to overall income generation in Central America rose significantly over the decade. This report aims to identify the factors behind these three trends—stagnant labor force participation rates for women contrasted with narrowing earnings gaps and a greater contribution to overall income—apparent in Central America from 1997 to 2006. The insights from this analysis can inform policy options to overcome remaining obstacles and accelerate improvements to women’s ability to participate in economic activity, to the benefit not only of themselves but also their families and society as a whole. For example, bringing the labor force participation of Central American women up to the world average for their level of GDP per capita could result in up to 2 million people exiting poverty. The report shows that improved education has been a driving force behind female decisions to enter the labor market, especially among younger women. For the first time, gender gaps in education have closed by 2006; in fact, in most countries in the region women have actually surpassed men in education attainment. Family structure dynamics, such as an increasing female-headed household population and lower fertility rates over the decade, have also played a role in reducing constraints or changing preferences vis-à-vis labor market participation. Despite these positive changes, accessing labor markets has been hindered by limited dynamism in the labor market and weak labor demand, linked to the slow overall growth in the region over the period. Gender concentration and segregation across sectors and industries remained unchanged, while overall job creation was limited. [5]    Only Panama and Costa Rica (the largest and most diverse economies in the region) have seen modest increases in women’s labor market entry. The potential for improvement in this area is great. Regarding earnings, for the first time women in Central America have equalized or even exceeded the earnings of men, and are ahead of the rest of Latin America. These gains can be partly attributed to the overachievement of women in human capital accumulation over the decade and to changes in family structure dynamics. Even more important, the findings show that in the majority of countries, earnings equality has been driven largely by improvements among the poorer segments of the populations (poorer, rural, with more children, less educated), and evidence on the causes of those improvements suggest that one of the channels through which change has happened may be linked to changes in attitudes and social norms. A closer look at the welfare distribution reveals heterogeneous performances. While poor women in El Salvador, Honduras, Nicaragua and Panama have experienced faster reductions in earnings gender gaps, poor women in Costa Rica, El Salvador and Panama have reduced gender gaps in female labor force participation. The best performers are El Salvador and Panama, where poor women have caught up with men in both labor force participation and earnings. Guatemala, on the other hand, is the only country where poor women experienced reductions in labor force participation and a widening earnings gap. These results to a large extent mirror urban-rural comparisons, as the poor are predominantly rural. The report probes these results further by reviewing evidence from innovative programs in Central America and the rest of the Latin America region over the last decade. Three broad lessons emerge. First, policies aiming to improve the returns of female labor by reducing labor market constraints and enhancing skills have shown promise, particularly among the poor. These include skills formation schemes, especially those that combine skills with job placement services that induced female entry in the labor force. Similarly, the availability of childcare has facilitated the ability of young mothers to work more by reducing time constraints at home. Micro-finance schemes have also helped poor female entrepreneurs by alleviating credit constraints and improving business- related skills through training. Second, policies that have changed intra-household resource allocations by putting money in the hands of women have facilitated investments in human capital accumulation and future gender equality, potentially inducing changes in social and behavioral norms and preferences. These [6]    programs include conditional cash transfers targeted at poor women that have proliferated in the region over the decade. The emerging role of female migration in search of better work opportunities has also had an important impact on intra-household resource allocation and social norms, and should be more studied more closely to assess its implications on policy. Third, many of the programs targeting poor women directly or indirectly facilitated social interactions through increased communication among female beneficiaries, leaders and community members. The small but growing body of evidence shows that such interactions can promote knowledge exchange and induce important changes in behavior and attitudes, which can in turn translate to better economic outcomes in the form of improved investments and incomes. Closer attention is therefore needed to program elements that can facilitate or enhance such social interactions in shifting aspirations and social norms and improving development outcomes. For a region like Central America, renowned for its plethora of small-scale private sector-led initiatives and pilots run by small NGOs and other local organizations, these impacts on behavior and social norms are promising areas of future research and policy innovation. Despite these advances, existing program coverage, limited fiscal space and modest economic growth in Central America constrain additional advances. A policy roadmap should therefore carefully consider how to prioritize and expand income generation opportunities via: (i) policies to facilitate labor market entry and job intermediation; (ii) expanded provision of childcare services; (iii) skills formation that are better linked to labor market demand; (iv) interventions that allow women to develop new businesses and expand existing ones; and (v) policies that promote empowerment and aspirations, which can bring changes in decision-making and affect development outcomes. Finally, policies should be responsive to ongoing changes in gender roles and intra-household resource allocation, which may be affecting social norms. Policy actions to encourage the development of new market niches with the capacity for job creation for women, such as in the service sector where much of the female labor force is employed, could be explored. Finally, the normative framework regulating economic activity may also need to be adapted to prevent unintended biases or gender disincentives. Beyond these specific policy considerations, a set of broad policies related to human capital development (universal education, expansion of secondary education and improved education quality), need to be revised and [7]    strengthened. Similarly, industrial and labor market policies capable of generating important scale effects also may be needed to stimulate across-the-board demand for additional female labor. On the research front, the report highlights several areas where existing knowledge gaps are limiting effective policy dialogue and design. First, the growing prevalence of seasonal and permanent migration in the region is a phenomenon that merits further analysis. Beyond income and labor market implications, the evidence shows that it may also be affecting social norms, attitudes and intra-household dynamics, which in turn can have additional impacts on welfare and implications for policy. Understanding in more depth the costs and benefits related to migration is therefore important. Second, shedding light on intra-household time allocation dynamics could lead to a better understanding of the way work and family responsibilities have evolved along gender dimensions. It could also help understand their links to inefficient outcomes such as youth delinquency, shorter career paths, flatter earnings curves and intergenerational transfer of poverty. Third, a deeper understanding of the links between empowerment and economic development is urgently needed. The emerging evidence suggests that changes in aspirations can play a major role in improving decision-making, investments and income generation potential for women. More work on this is needed to better understand its potential and inform policy decisions and program design. [8]    1 THE EVOLUTION OF LABOR MARKET PARTICIPATION AND EARNINGS IN CENTRAL AMERICA During the last two decades, Central America has made steady but modest progress in economic growth and poverty reduction. The 1990s marked a critical turning point for the Central American countries, both politically and economically. The civil conflicts in El Salvador, Guatemala and Nicaragua were largely brought to an end with the signing of national peace accords, and the severe economic decline experienced throughout Latin America in the 1980s began to reverse itself. As a result, the economies grew modestly in the six Central American countries over the 1997-2006 period. In fact, the majority of countries in the region have outpaced the other Latin American countries over the decade. Costa Rica and Panama have led the way, growing at an annual rate of 3 percent between the mid-1990s and mid-2000s (Table 1.1). El Salvador, Honduras and Nicaragua, all with significantly smaller per capita GDPs, have also grown at an annual rate close to 2 percent. The only laggard in the period was Guatemala, which grew at an annual rate of less than 1 percent. Partly as a consequence of this economic growth, poverty rates fell during the period. While poverty rates are higher in Central America than in other Latin American countries, poverty fell by more than ten percentage points in Honduras, while the rest of the countries experienced positive but smaller rates of poverty reduction. Consistent with the lower rates of growth observed and persistently high levels of inequality, Guatemala and El Salvador saw very little reduction in the poverty rate between 1997 and 2006. In parallel with these economy-wide advances, Central America also achieved significant progress on various dimensions of gender equality. For example, women’s average educational attainment has increased, despite the fact that they are still significantly lower than in the rest of Latin America. In four of the six Central American countries, women are now more educated than men, reversing the gender gap existing at the end of the 1990s. Similarly, fertility rates during this period dropped at least as much as in the rest of Latin America, with the exceptions of Guatemala and Panama (Table 1.1). Other improvements can be seen with respect to life expectancy and indicators related to [9]    Millennium Development Goal 3 on gender equality, with progress across the board in Latin America, including Central American countries.1 Table 1.1: Central America Key Socio-Economic Indicators Sources: (a) World Development Indicators, (b) Socio-Economic Database for Latin America and the Caribbean (SEDLAC). 1.1 THREE KEY TRENDS FROM THE DECADE (1997-2006) Reviewing the evolution of income-related gender gaps over the decade, three trends stand out: (i) female labor force participation rates are almost unchanged over the decade; (ii) women for the first time earn on average more than men; and (iii) the overall economic contribution of women has increased. The explanation and interpretation of these three trends form the motivation of this report. 1.1.1 TREND 1: INSIGNIFICANT CHANGES IN FEMALE LABOR FORCE PARTICIPATION Between 1997 and 2006, female labor force participation (FLFP) in Latin America increased, reaching an average of 53 percent in 2007. By contrast, less than 50 percent of Central American women of working age (15-60) were in the labor force in 2007. This ranges from a low of 40 percent in Honduras to a high of 51 percent in Costa Rica (Figure 1.1a). By contrast, the participation rate of men is around 90 percent in all countries of the region. Figure 1.1a: Labor Force Participation Rates (Circa 2006)                                                         1Examples include the ratio of girls to boys in education, the share of women in non-agricultural paid employment and women with parliamentary seats. [10]    100%  80%  60%  level  40%  20%  0%  female  male  female  male  female  male  female  male  female  male  female  male  Costa Rica   El Salvador  Guatemala   Honduras  Nicaragua  Panama  Source: SEDLAC and World Bank. Perhaps more importantly, FLFP in Central America has experienced very small increases since the mid-1990s. The annual rate of FLFP growth was zero for El Salvador, Honduras and Nicaragua and less than 1 percent for Guatemala (Figure 1.1b). This situation takes place side by side the largest increase of the Latin America region –Costa Rica- and a general sustained increase in most Latin American countries. Figure 1.1b: LFP Annual Rate of Growth (1997-2006) 3%  2%  growth  1%  0%  â€?1%  female  male  female  male  female  male  female  male  female  male  female  male  Costa Rica   El Salvador  Guatemala   Honduras  Nicaragua  Panama  Source: SEDLAC and World Bank. As a result, gender gaps in labor force participation in Central America remain extremely high: men are on average 70 percent more likely to be in the labor force than women (Figure 1.2a), a gap that is [11]    10 percentage points larger than the average for Latin America. Given the small increases over the decade, it is no surprise that these gaps have changed very little (Figure 1.2b). The two countries that did manage to reduce the gap over the decade are Costa Rica and Panama, where the gaps have been closing at an average annual rate of 2 and 1 percent respectively. Figure 1.2a: Female to Male LFP Ratio (Circa 2006) 0,8  0,6  level  0,4  0,2  0,0  Honduras  Guatemala   Nicaragua  Costa Rica   Panama  El Salvador  Source: SEDLAC and World Bank. Figure 1.2b: Female to Male LFP Ratio Annual Growth Rate (1997-2006) 3%  2%  growth  1%  0%  â€?1%  Honduras  Guatemala   Nicaragua  Costa Rica   Panama  El Salvador  Source: SEDLAC and World Bank. 1.1.2 TREND 2: FOR THE FIRST TIME, WOMEN EARN MORE THAN MEN Despite the lack of change in LFP, earnings gaps between men and women have decreased dramatically in Central America over the decade. In fact, while men earned on average 9 percent more than women in 1997, women earned 3 percent more than men by 2006 (Figure 1.3). 2                                                         2 Earnings are measured in hourly wages in the principal activity for workers between 15 and 64 years. The data comes from nationally representative household surveys. [12]    Interestingly, this occurred in a period where real wages were in decline for both men and women, suggesting that real wages fell faster for men (see Table A1.4 in Appendix). Figure 1.3: Gender Earnings Gaps Over the Decade (Ratio of Male to Female Earnings) Source: SEDLAC and World Bank. These trends have been widespread across the region, with some variation. El Salvador, Guatemala and Honduras have been the best performers, while wage gaps in Costa Rica have not changed over the period. While earnings gaps narrowed in Nicaragua, the country still has earnings gaps favoring men. As a whole, Central America has surpassed the rest of Latin America in reducing earnings gaps. 1.1.3 TREND 3: THE ECONOMIC CONTRIBUTION OF WOMEN HAS INCREASED The contribution of Central American women to income generation, both at the national and household level, has been on the rise. An estimate of women’s income contribution can be calculated by proxying the total national income in a country as the sum of all incomes reported by men and women in national representative household surveys. In Central America, about a third of total income is earned by women (Figure 1.4a), ranging from 20 percent in Nicaragua to 40 percent in El Salvador. Over the last decade, this contribution has grown at an annual rate that ranges from 1 to 3 percent (Figure 1.4b). Figure 1.4a: Women’s Share of Total Income (Circa 2006) [13]    0,6  0,4  level  0,2  0,0  Costa Rica   El Salvador  Guatemala   Honduras  Nicaragua  Panama  Source: SEDLAC and World Bank. Figure 1.4b: Annual Change of Women’s Share of Total Income (1997-2006) 3%  2%  growth  1%  0%  Costa Rica   El Salvador  Guatemala   Honduras  Nicaragua  Panama  Source: SEDLAC and World Bank. 1.2 STUDY ROADMAP The three trends outlined above present a puzzle. On one hand, the overall trends indicate that during the 1997-2006 period little has changed in labor force participation rates. And yet, women in the region have made large advances in both closing the earnings gaps as well as increasing their contribution in terms of income generation. What explains these contrasting findings? The report aims to answer this by exploring the evolution of these trends over the decade. The main focus is to document and identify the factors behind the changes—and in some cases the lack of changes—in the decade regarding men’s and women’s economic participation and income generation. The report focuses on three questions motivated by the trends: 1. What explains female labor force participation patterns? 2. What are the factors behind the reversal of the gender earnings gaps? 3. Is there evidence that policies may have played a role in any of these trends? The report is organized as follows. Chapter 2 discusses the evolution of female labor force participation and explores demand and supply factors associated with the observed trends. Chapter [14]    3 explores the reasons behind the reduction in gender earnings gaps in the region over the decade. Chapter 4 explores heterogeneity in gender gap changes across welfare dimensions, and in particular among the poor. Chapter 5 provides insights for guiding policy design efforts in Central America by extracting key lessons from programs that aim directly or indirectly to improve gender equality with respect to income generation and poverty reduction. Chapter 6 presents the main conclusions and policy recommendations, and suggests ways to move the gender agenda forward in the region as well as how to fill remaining knowledge gaps. The analysis uses nationally representative household data from all six countries in Central America that include detailed and comparable (across and within countries) information on poverty, income and employment trends by gender. Two rounds from each country (circa 1997 and 2006 depending on data availability) are used for the analyses of the evolution of trends over the decade and thus offer a unique dynamic setting to study gender gaps (see Appendix Table A1.1 for details). [15]    2 UNDERSTANDING FEMALE LABOR FORCE PARTICIPATION Is Central America where it should be with respect to female labor force participation (FLFP)? In general, increases in GDP per capita go hand in hand with increases in the presence of women in the labor force.3 Despite the large variation in empirical findings in both the rate of increase and the level of FLFP, the existing literature points to links between greater educational opportunities among women and higher growth rates or levels of per capita income.4 One simple way to explore this is by looking at the relationship between female labor force participation and economic development. Goldin (1995), Mammen and Paxson (2000) and Psacharopoulos and Tzannatos (1989) among others found a U–shape relationship. The hypothesis is that at low levels of income—where agriculture is the dominant form of economic activity— women participate in the labor force but tend to do so as unpaid family workers. As economic activity starts shifting from home-based production to market-oriented activities, due to the expansion of markets or the introduction of new technology that reduces the number of workers needed for agricultural production, income rises and the FLFP rate drops. As women’s education improve so does their expected wages, increasing their labor force participation. Finally, expansion of the tertiary/service sector increases the demand for women’s labor further.5 Several insights stand out when this relationship is plotted visually for Central America (Figure 2.1).6 First, the figure confirms the U-shape relationship between female participation and GDP described above. Second, since the mid-1990s, this relationship has been shifting, with more women                                                         3 World Bank (2001 and 2006) and Inter-American Development Bank (1999 and 2002) present evidence from different countries at different periods of time on the relationship between gender equality and economic growth. Besides the direct effect on economic growth from women’s lower labor supply (relative to men), occupational segregation and vertical gender segregation are other dimensions of gender inequality in the labor market that generate economic efficiency losses. Occupational and vertical segregation by gender are discussed in section 2.1 of this chapter. 4 Examples include King and Hill, 1993, and Knowles, Lorgelly and Owen, 2002. GDP per capita growth and gender relation has been explored by Klasen 1999, 2002, 2003; Dollar and Gatti, 1999; Forbes, 2000; and Appiah and McMahon, 2002. 5 Income and substitution effects are hypothesized to operate during this process. The declining portion of the U-shaped curve suggests that a strong income effect dominates a small own-substitution effect. In the rising portion of the U- shaped curve, the substitution effect of higher wages (away from home to market activities) dominates the small income effect. 6 The figure plots the log of GDP per capita based on purchasing power parity (PPP) with the share of females in total labor force (ages 15 and older). Data are in constant 2005 international dollars. For the figure, a quadratic equation was estimated for the share of females in total labor force, as function of GDP per capita for each five-year period for an unbalanced sample of 228 countries and 1,596 observations. [16]    participating in the labor force for a given level of GDP. Third, although Latin America has on average overachieved compared with other countries with similar levels of GDP, Central America is lagging behind, with the exception of El Salvador. In fact, given GDP levels, Central American countries are below what they should have been by the end of the 1990s! An average of five percentage points would be needed for Central American countries to get them to the average level. Figure 2.1: Female Share of Total Labor Force and Per Capita GDP 50  Labor force, female (% of total labor force)  45  SLV  LAC  40  NIC  GTM  PAN  CRI  35  HND  30  5  6  7  8  9  10  11  LOG GDP per capita, PPP (constant 2005 internaRonal $)  1996â€?2000  2001â€?2005  2006 and aQer  Observed values for 2006 and aQer    Source: Adapted from World Bank (2010a). Estimates based on data from the World Development Indicators 2009, World Bank.  The fact that Central America is not where the U-shape relationship between participation and GDP levels would otherwise imply indicates an important potential for further growth and poverty reduction. In order to asses this further, a simple model of labor force participation and income earnings can be used to estimate the additional welfare impacts of an increase in FLFP.7,8 When the                                                         7 The simulation exercise consists of three steps. First, country-models of FLFP are estimated based on individual and household characteristics. Predicted probabilities of participation for everyone in the sample can be then calculated (participants and non-participants). Second, an income regression of participants is estimated controlling for the predicted probability of labor force participation. Third, the income effect of an increase in the labor force participation [17]    model is applied to Central America, the results are telling: an increase of women in the labor force by a moderate 10 percent (enough to bring countries to their implied level on the U-curve) is associated with a significant increase in overall incomes in the region, ranging between 4-8 percent (Figure 2.2a). This in turn is associated with additional poverty reduction in the region of 5-8 percent (Figure 2.2b). Based on this, more than 2,000,000 people in Central America would exit poverty. In other words, bringing FLFP to the level predicted by the model according to the region’s GDP levels (Figure 2.1) would result in a sizeable reduction in poverty. While such results do not constitute a motivation for arguing for higher female labor force participation per se, they do underscore the potential welfare gains from an increased female participation in the economy. In a context where women are contributing an ever-increasing share of household income, the point is even more noteworthy. Figure 2.2: Income and Poverty Reduction Effects From a 10% Increase in FLFP 2.2a: Household Income Increase (%) 2.2b: Decrease in Poverty Rates (%) 8% Costa Rica El Salvador Honduras Nicaragua Panama 0% 6% 4% -4% 2% -8% 0% -12% Moderate poverty Extreme poverty Source: SEDLAC and World Bank. Based on the above, this chapter explores the factors that may have influenced the lack of FLFP improvements over the decade in Central America, in particular: (i) labor market conditions, such as                                                                                                                                                                                 is estimated by calculating income predictions for non-participant women based on their predicted probabilities of participation. That is, to simulate a 10 percentage-point increase of labor force participation in a given country, income is calculated based on step 2 for the 10 percent of non-participant women with the highest predicted participation probabilities (among non-participants). This assumes that the returns to education and to other assets are not affected despite the increase of the labor force participation. Guatemala does not appear on the figure as official poverty lines were used and none were available for the country. 8 In a recent study of Chile, the results show that a rise in the country’s FLFP rate from 37 to 50 percent is associated with Chile’s growth advantage vis-à-vis other countries of same income level from 2.7 to 3.2 percentage points and significantly reduced poverty levels (World Bank 2007). Similarly, using the same methodology, increasing Turkish FLFP to 29 percent (from the also low 23 percent) would reduce the country’s poverty rate by 15 percent (World Bank 2010b). [18]    overall labor demand and sectoral concentration and segregation; (ii) the evolution of skills and human capital accumulation; and (iii) changes in preferences and constraints linked to fertility and family formation, affecting competing demands between productive and reproductive roles. 2.1 WEAK LABOR MARKET DEMAND, CONCENTRATION AND SEGREGATION Where do men and women in Central America work? Has this has changed over time? Almost 80 percent of Central American women work in four sectors: commerce, health and education, domestic service and low-skill industry (see Appendix Table A1.3). Commerce employs about a quarter of all women in Costa Rica and Panama, about a third in Nicaragua, Guatemala and Honduras, and almost half in El Salvador. By contrast, half of working men are employed in agriculture and commerce, while the other half are spread over other sectors. The decade in review was not a great one in terms of job creation. Overall growth was modest, constraining increases in labor demand. In addition, even the modest expansion in most of the sectors that employ the majority of women (commerce, services, manufacturing) was not associated with increases in FLFP. No link is apparent between sectoral growth (in GDP levels) and changes in female participation rates in the sectors that employ the majority of women, with the one exception of Guatemalan manufacturing (Figure 2.3). On the whole, the decade does not seem to have been one of increases in labor demand. Figure 2.3: Annual Change in FLFP and Sectoral GDP Growth (1997-2006) [19]      Source: SEDLAC, World Bank and CEPALSTAT. Notes: Serv = services (includes public sector, education, health and domestic service), Com= commerce, Ind= manufacturing. All indicators of concentration and segregation in the Central American labor market over the decade in review reveal the same pattern: men and women have been and continue to be concentrated and segregated across sectors and industries. A large share of these jobs is in the informal sector, a trend that persists. Central America has relatively high levels of informality, reaching almost 60 percent in Nicaragua, Guatemala and El Salvador (Figures 2.4a). Women are somewhat more likely to work in the informal sector. These patterns have barely changed over the decade (Figures 2.4b). A notable exception is Honduras, where the informal sector contracted over the period. Figure 2.4a: Share Working in Informal Sectors, Circa 2006 (Adults 15-64) [20]    80%  60%  percentage  40%  20%  0%  females  males  females  males  females  males  females  males  females  males  females  males  Costa Rica   El Salvador  Guatemala   Honduras  Nicaragua  Panama  Source: SEDLAC and World Bank. Figure 2.4b: Annual Growth of Informal Sector Participation 1997-2006 (Adults 15-64) 3%  0%  growth  â€?3%  â€?6%  â€?9%  females  males  females  males  females  males  females  males  females  males  females  males  Costa Rica   El Salvador  Guatemala   Honduras  Nicaragua  Panama  Source: SEDLAC and World Bank. Sectoral segregation also persists. The Duncan index is a simple metric of establishing whether a certain type of occupation is more likely to employ men or women. In the case of Central America, the Duncan index suggests a relative high level of segregation, ranging from 0.38 to 0.55 (Figure 2.5a)9. While there are some variations across countries, these trends have again changed very little over the last decade (Figure 2.5b). In fact for two countries, Panama and Guatemala, segregation has increased. Figure 2.5a: Sectoral Gender Segregation—Duncan Index (2006)                                                         9 The Duncan index compares the share of men to women in a given sector. The index ranges from zero (no segregation) to 1 (full segregation). An index of 0.75 would suggest that in order to obtain equal distribution of employment one would need to move 75 percent of women to another sector for the distributions between men and women to be the same. [21]    60%  40%  20%  0%  Costa Rica   Panama  Guatemala   El Salvador  Honduras  Nicaragua  Source: SEDLAC and World Bank. Figure 2.5b: Sectoral Gender Segregation—Duncan Index Annual Change (1997-2006) 2,0%  1,0%  growth  0,0%  â€?1,0%  Costa Rica   Panama  Guatemala   El Salvador  Honduras  Nicaragua  Source: SEDLAC and World Bank. A closer look at manufacturing is consistent with these findings. Much attention in recent years has been paid to the potential of the textile and clothing (maquila) sector, a key sector for female employment. This dynamic sector, driven by foreign direct investment, has been responsible for most of the growth of manufactured exports and foreign exchange earnings, as well as for most of the formal sector employment generated in the region, since the late 1980s (Morley et al. 2007). About 90 percent of maquila production in Central American is concentrated in Guatemala, Honduras, and El Salvador. Despite its relatively recent origins, maquila has quickly become a leading export in every Central American country except Costa Rica (Table 2.1). Why would this matter for women? The textile and clothing industries tend to be labor intensive, and maquila therefore is a major provider of employment. In Honduras, approximately 70 percent of all maquila workers are women; the estimate for El Salvador is 80 percent and for Nicaragua, 90 percent (De Hoyos et al. 2009; Quintana et al. 2002; Gutierrez et al. 2008). Earnings are, on average, higher in the maquila sector than in other comparable employment. Marcouiller and Robertson (2007) estimate that, controlling for human capital and demographic characteristics, workers in the [22]    apparel industry in Honduras earn 10-20 percent more than the average wage. De Hoyos et al. (2009) find an even higher wage premium—38 percent—for women employed in the maquila sector relative to female workers in non-maquila jobs. Moreover, their results indicate a significantly lower gender earnings gap in the maquila industries relative to the rest of the economy: whereas women earn on average almost 30 percent less than men in non-maquila industries, this gap is narrowed to 12 percent in the maquila sector. Despite the potential benefits for employment and income generation, the size of the maquila sector continues to be small: less than 2 percent of the employed work there. As such, it is unclear whether the sector can offer large employment generation channels in the short term. Table 2.1: Importance of Textile Maquila in Central America Country Maquila Change Growth Maquila Maquila Maquila Approximate exports to from year rate 1995- exports to employment employment/total number of US before 2002 (%) US as a % (thousand employment in maquila (10 USD)a (%) of total persons)d manufacturing firmsd exportsc (%)d,e Costa Rica 478.6 -6.7 1.3 5.6 13 8 40 El Salvador 1445.4 -18.2 193.8 33.6 87 20 250 Guatemala 1717.1 -14.2 150.0 37.3 142 500 Honduras 2461.8 -9.9 174.2 80.3 129 27 200 Nicaragua 753.6 11.5 502.7 45.6 60 30 70 Notes: a/ Data refer to June 30, 2005-June 30, 2006 b/data for 2002 c/data from a as a percentage of total exports in 2004 d/data year 2003 except Costa Rica which is same as a. e/Total imported inputs for maquila are approximated as the sum of the imports in STIC codes number 26 (textile fiber), 65 (tarn, thread and fabrics) and 84 (clothing) Source: Jensen et al (2007) Other structural factors on the demand side, such as labor legislation, may also make women more expensive than men for the employer or prevent them from hiring them (see Box 2.1). Similarly, employer preferences as well as market discriminatory practices can affect women’s decisions and options regarding labor force participation, sectoral choices, investments in education and choice between self or wage employment. Some of these demand issues are explored later in the report. Box 2.1: Keeping Women Out of Work—Creating Adverse Effects via Legislation in Costa Rica One of the ways demand for female labor can be reinforced is by removing constraints and barriers in legislation that may make employers prefer male over female employees. While in some cases the legislation is explicit about gender equality, in other places potential discriminatory practices may be articulated. For example, since 1990 Costa Rica has the Law for the Promotion of Women’s Social Equality. The law mandates equality in the political sphere and the exercise of public office, social and economic rights— [23]    including mandatory childcare—and reform of existing laws to remove gender bias. While its implementation has been successful in increasing parliamentary representation, in economic participation less progress has been achieved. Article 88 of the law states that night work (between 7 p.m. and 6 a.m.) for women is only allowed for domestic workers, nurses, social workers and similar professions, while women working in sales or office jobs cannot work after midnight. All other firms are required to apply for authorization from the Ministry of Labor. Although intended to protect women from exploitation and exposure to risk, Article 88 may prevent employers from hiring women, given these constraints. For example, new opportunities like the introduction of call center services in Costa Rica that serve countries on different time zones cannot be taken advantage of by women due to the law. 2.2 EDUCATION KEY BUT NOT ENOUGH One of the most noteworthy phenomena in terms of gender gaps is the closing of the education gap. In Latin America, women are outperforming their male peers in both attainment and enrollment. Between 1990 and 2005, female enrollment rates in primary and secondary education increased significantly. By 2005, female enrolment rates exceeded male enrolment in several countries in the region (World Bank 2007). Like most countries in Latin America, educational attainment in Central America has also been on the rise over the last decade. The average adult in Central America has six years of education, ranging from five years in Guatemala to almost 10 in Panama (Figure 2.6a and Figure 2.6b). Figure 2.6a: Average Years of Education (Circa 2006) 12  10  8  level  6  4  2  0  female  male  female  male  female  male  female  male  female  male  female  male  Costa Rica   El Salvador  Guatemala   Honduras  Nicaragua  Panama  Source: SEDLAC and World Bank. [24]    Figure 2.6b: Annual Growth of Years of Education (1997-2006) 4%  3%  2%  growth  1%  0%  â€?1%  female  male  female  male  female  male  female  male  female  male  female  male  Costa Rica   El Salvador  Guatemala   Honduras  Nicaragua  Panama  Source: SEDLAC and World Bank. Rising education attainment has been largely driven by increases in overall school attendance, with primary completion rates for women ranging from universality in Panama to almost 80 percent in Nicaragua and Guatemala, and a significant increase in secondary school attendance (Figure 2.7a and Figure 2.7b). Figure 2.7a: Secondary School Attendance by Sex (Circa 2006) 100  80  60  level  40  20  0  Female  Male  Female  Male  Female  Male  Female  Male  Female  Male  Female  Male  Costa Rica   El Salvador  Guatemala   Honduras  Nicaragua  Panama  Source: SEDLAC and World Bank. [25]    Figure 2.7b: Annual Growth of Secondary School Attendance by Sex (1997- 2006) 4%  3%  growth  2%  1%  0%  Female  Male  Female  Male  Female  Male  Female  Male  Female  Male  Female  Male  Costa Rica   El Salvador  Guatemala   Honduras  Nicaragua  Panama  Source: SEDLAC and World Bank. Based on this, it is not surprising to find that over the last decade, women in Central America have not only closed the education gap but in some cases are ahead of men. In Honduras and Nicaragua, women have on average 20 percent higher educational attainment levels (Figures 2.8a and 2.8b). These trends are similar for both urban and rural areas, although the levels of educational attainment are significantly higher in urban areas. Figure 2.8a: Female to Male Ratio in Education Attainment (2006) 1,4  1,2  1,0  level  0,8  0,6  0,4  0,2  0,0  Guatemala   El Salvador  Costa Rica   Panama  Honduras  Nicaragua  Source: SEDLAC and World Bank. Figure 2.8b: Annual Growth of Female to Male Ratio in Education Attainment (1997-2006) [26]    0,6%  0,4%  0,2%  growth  0,0%  â€?0,2%  â€?0,4%  â€?0,6%  Guatemala   El Salvador  Costa Rica   Panama  Honduras  Nicaragua  Source: SEDLAC and World Bank. Consistent with the above, age cohort analysis suggests that younger generations of women have experienced the most marked improvements in educational attainment and are now more educated than their male counterparts. For example, while actual levels are still low, secondary school completion rates have dramatically increased in the last few decades, with completion rates among younger generations of women surpassing those of men (Figure 2.9). Figure 2.9: Secondary Education Completion Rates by Birth Year and Gender [27]    Source: SEDLAC and World Bank. Women’s advances in educational attainment are also widespread along the welfare distribution. Education gaps with respect to years of education favor women across the whole income distribution, with gaps slightly higher for wealthier households (Figure 2.10). The notable exception is Guatemala and to some extent El Salvador, where the overall educational gap favor men (in the case of Guatemala, diminishing along the income distribution). Figure 2.10: Gender Gaps in Education Attainment by Income Percentile (2006) [28]    Source: SEDLAC and World Bank. Did the elimination (and even reversal) of education gender gaps in the last decade result in a boom in labor force participation? In line with Psacharopoulos and Tzannatos (1992) and Duryea and Szekely (1998), the trends in Central America suggest a strong link between education and female labor force supply. Still, the results also indicate some nuances. Proxying skill and experience by levels of formal education, labor market participation significantly increases with education attainment, especially as women enter and complete secondary schooling (Figure 2.11).10 However, participation is flat for women with less than six years of education. In fact, while participation rates never reach those of men, they reach almost 80 percent for the majority of the countries among women with university degrees. The correlation between years of education for men is more muted, although there is also an upward trend. Figure 2.11: Education Attainment and Labor Force Participation Levels, by Sex                                                         10 Secondary schooling starts in general after year six of education. [29]    Source: SEDLAC and World Bank. These results are consistent with earlier work on the determinants of increasing FLFP in Latin America. Duryea et al. (2001) decompose the participation rate changes into two parts: the [30]    proportion explained by increases in women’s education, and the proportion attributable to rising participation rates within schooling groups. The results indicate that increases in FLFP in Latin America are driven by increased participation rates within schooling groups. Increases in female schooling account for only 30 percent of the overall increase in FLFP, while the remaining 70 percent is explained by increases in participation rates at given schooling levels. Taken together, the results in this section show that women in Central America have caught up with men in terms of educational attainment. Despite these gains, women in Central America do not have the same level of labor force participation as men at any level of education, even though education is positively associated with higher FLFP (especially at higher levels of educational attainment). Hence, education alone cannot explain the observed levels of FLFP. Other factors seem to be at play that may be related to constraints or preferences. The next section explores family dynamics over the decade. 2.3 FAMILY DYNAMICS—FEMALE HEADSHIP AND CHILDREN Accompanying the positive changes in the stock of female human capital in the last decade were changes in family structure, which both allowed and compelled more women to enter the paid workforce. For example, as mentioned above, fertility rates in Central American over the decade have decreased by between 0.5 and 1 child per woman. These changes have been associated with increased FLFP (Cruces and Galiani 2007, Piras and Ripani 2005). One of the most salient demographic factors contributing to Central America’s rising FLFP is the increase in female household headship. The latest household survey data indicate that around 2006, approximately one-third of all Central American households were headed by a woman (Figure 2.12a). Female-headed households are particularly prevalent in El Salvador, Honduras and Costa Rica, driven by a combination of male migration, divorce and spousal abandonment or death. Female headship has been increasing rapidly over the last decade, with annual growth rates of as high as 4 percent in the case of Nicaragua (Figure 2.12b). Figure 2.12a: Share of Female-headed Households (Circa 2006) [31]    40%  30%  level  20%  10%  0%  Nicaragua  Guatemala   Panama  El Salvador  Costa Rica   Honduras  Source: SEDLAC and World Bank. Figure 2.12b: Annual Growth in the Share of Female-headed Households (1997-2006) 4%  3%  growth  2%  1%  0%  Nicaragua  Guatemala   Panama  El Salvador  Costa Rica   Honduras  Source: SEDLAC and World Bank. A division of women based on marital status and whether they have small children is particularly revealing in how these factors may affect labor force participation, as revealed by mapping labor force participation among four “typesâ€? of women for the case of Panama, which shows a pattern mirrored by all six countries (Figure 2.13). The figure shows that irrespective of type, life-cycle patterns of labor force participation with respect to age exist: FLFP increases until the age of 40, when it starts decreasing. Figure 2.13: Female Labor Force Participation and Family Dynamics (Circa 2006) [32]    Source: SEDLAC and World Bank. As well, single women have almost exactly the same labor force participation trajectories as households headed by females with small children (Figure 2.13). In fact, labor force participation trends are the highest for these two types of women along the whole age distribution. One interpretation is that single women may face fewer constraints with respect to work, while work for female household heads with children is a necessity. Interestingly, the peak for these groups at age 40 is close to 80 percent, almost as high as male participation rates. One related issue with female-headed households is teenage pregnancy. For example, Nicaragua has the highest share of young women (15 to 19 years old) who are mothers in the entire Latin America region. Early motherhood is closely linked to forgone education opportunities (teenage mothers are less likely to complete primary or secondary education), which is critical for their labor market participation and employment opportunities, as well as their families’ wellbeing. A recent study in Costa Rica (World Bank 2007) finds that the highest share of poor women with low skills working part time is among single mothers. This partially reflects the increase in the share of female-headed households between 1987 and 2004: from 17.0 percent of all households to 26.4 percent; and from [33]    19.7 percent to 33.6 percent among poor households (also Box 2.2). While the study does not explore this further, such trends deserve further analysis. Box 2.2: Female Headship and Poverty While there is an ongoing debate about the best ways to measure and assess the relationship between gender and poverty (cf. Marcelo and Costa 2008, Deere et al. 2009), disaggregating poverty rates by headship status provides some useful insights. Headcount poverty ratios by household headship for Central America based on a $2.50/day poverty line indicate that poverty rates for female-headed households fell by 15-30 percent in all of the countries during the past decade. Interestingly and somewhat unexpectedly, by circa 2006 poverty rates among female-headed households were also significantly lower than those of male-headed households in all of the Central American countries except Costa Rica. In the case of Honduras, where poverty rates among female-headed households are a full ten percentage points below those of male-headed households (31 percent compared to 41 percent), the gap can be partially explained by the fact that female-headed households are more likely to be located in urban areas, to have migrants abroad and receive remittances.11 In Nicaragua, poverty rates among households headed by older (over 35 years) women are lower than male- headed households, but younger female household heads (under age 35) were on average 9-13 percent poorer than their male-headed counterparts (World Bank 2008). The general finding that female-headed households in Central America are not on average poorer than male-headed households is consistent with the emerging literature in Latin America, which finds little empirical evidence of a “feminization of povertyâ€? in the region (Marcelo and Costa 2008). Finally, married women without children have participation rates ranging between 40 and 60 percent (Figure 2.13). The presence of children lowers the probability for labor market participation. For these women, additional household resources may be less important, since presumably the husband is more likely to work. Adding child rearing responsibilities further reduce their probability to work. The sharp reduction in fertility and increases in household headship over the decade, along with differential relationship with labor force participation, highlight the complicated role of constraints and individual preferences and attitudes with respect to employment and family decisions. The need for thinking through policies that can provide incentive, remove constraints or introduce flexibility to balance competing demands and preferences is urgent.                                                         11In Honduras, 23.7 percent of female-headed households have migrants abroad and 43 percent of these receive remittances, compared to just 7.9 percent of male-headed household with migrants (of which 21 percent receive remittances). Sixty-two percent of Honduran female-headed households are located in urban areas, where poverty rates are lower, compared to 49 percent of male-headed households (World Bank 2006). [34]    To summarize, the chapter explore the current level and trends in FLFP in Central America over the decade. While factors associated with increases in FLFP in other developing countries also affect Central American women, FLFP in Central America is below what would be expected given the changes in women’s education and fertility. Women have managed to close the gender gap in education, but in spite of these achievements their labor force participation remains low, what it appears to be associated with their role within the household and family reproduction. On the other hand, the economies in the region have not matched the supply of educated female workers. Little employment opportunities have been created in the region in all economic sectors, independent if they employ more men or women. Hence, nor women or men have benefited of an increased labor demand or changed their economic sector clustering to take advantage of newly created opportunities. The next section turns to earnings. [35]    3 EXPLAINING CHANGES IN GENDER EARNINGS GAPS Women contribute up to 40 percent of the total income generated in Central America, a figure that has increased over the decade. Despite the limited changes in female labor force participation, earnings gaps over the decade have decreased dramatically in Central America. A decade ago men earned on average 9 percent more than women, but now women earn 3 percent more than men. These trends have been widespread across the region, with some variations. El Salvador, Guatemala, Honduras and Panama were the best performers, while wage gaps in Costa Rica have not changed over the period. Nicaragua still has an earnings gap favoring men, although it has narrowed. As a whole, though, Central America has surpassed the rest of Latin America in reducing earnings gaps. This chapter explores the factors driving changes in the earnings gap between 1997 and 2006. 3.1 DECOMPOSING THE EARNINGS GAP Recent work by Ñopo et al. (2008) and a companion piece by Hoyos and Ñopo (2010) suggest that while wage gaps in Latin America and Central America have reduced over time, they remain large even after controlling for observable characteristics such as education, age and occupational choice. To explore this further, this chapter replicates the approach used by these authors to analyze the evolution of wage gaps in Central America over the last decade. The basic principle is to extend the traditional Blinder-Oaxaca decomposition using matching techniques, to break down wage gaps into three main components: (i) the component of the wage gap explained by the gender differences in observed individual characteristics; (ii) the component that corresponds to differences between those men (women) that cannot be matched to women (men) using the procedure; and (iii) an unexplained component (see Appendix A2 for more details on this methodology). The results for the region suggest that some of the improvements in gender equality indicators discussed earlier, such as women’s schooling overachievement with respect to males, are reflected in the observed wage differences. For example, women receive a 10 percent premium over men as a result of having higher education or better demographic characteristics, depending on the specification (Figure 3.1 and Appendix Table A1.5). This premium has doubled over the decade, [36]    suggesting that education explains a larger part of the earnings gap than before. Similarly, factors such as household composition, residence, occupation and demographics also help explain earning gaps. Figure 3.1: Decomposition of Earnings Gaps 20% 10% 0% -10% -20% Overall earnings Unexplained factors Male specific Female specific Demographic gap factors factors factors Circa 1997 Circa 2006 Source: Hoyos and Ñopo (2010) with data from SEDLAC and World Bank. Demographic factors include age, education, presence of children in the household, presence of other wage earner and urban/rural location. See Appendix A2 for more details on the model. The decomposition approach also provides an estimate for the component of earnings gaps unexplained by observable characteristics. A number of interpretations can be given to the unexplained component, ranging from discrimination, cultural and social norms, unobserved preferences or market-based dynamics not captured by observables such as changes in the wage structure (Blau et al. 2006). The results suggest that this unexplained component gap is large and offsets the gains driven by increases human capital accumulation. Men receive on average about 12 percent more than women for factors that cannot be captured by the data, depending on the specifications used in the model. This is also true after comparing males and females with the same observable characteristics (residence, age, education, marital status, being a household head and occupation). The unobserved component varies across countries, ranging from 20 percent in Nicaragua to close to zero for Guatemala (Figure 3.2). [37]    Interestingly, the analysis also shows that this unexplained part of the earnings gap has decreased significantly over the last decade. While the unexplained gaps and changes over the decade vary across countries, the patterns are similar (with the exception of Costa Rica, where they have increased). El Salvador, Guatemala and Honduras have been the best performers in reducing this unexplained gap. The unexplained components of the earnings gaps in Central America are now lower than the average for the rest of Latin America. Figure 3.2: Changes in Unexplained Earnings Gap: Still Large But Declining Source: Hoyos and Ñopo (2010) with data from SEDLAC and World Bank. Note: “Rest of LAâ€? includes 18 countries outside of the six Central America countries. Period 1 is circa 1997 and Period 2 is circa 2007. The demographic set for “Rest of LAâ€? does not include dependency ratio. Exploring the potential channels behind the reduction of this unexplained component of earnings gaps reveals a number of interesting insights. First, family structure matters. For example, the unexplained component of earnings gaps was higher among families with children at the beginning of the decade, but it has reduced the most and is now lower than those without children (Figure 3.3). Similarly, the unexplained earnings gap is the lowest and has reduced the most among the youngest women (Figure 3.4). [38]    Figure 3.3: Changes in Unexplained Earnings Gap, by Presence of Children Source: Hoyos and Ñopo (2010) with data from SEDLAC and World Bank. Figure 3.4: Changes in Unexplained Earnings Gap, by Age Groups Source: Hoyos and Ñopo (2010) with data from SEDLAC and World Bank. [39]    Sectoral and occupational differences between men and women also matter. As the trends in Chapter 1 suggest, the level of gender sectoral concentration and segregation is high and has changed very little over the decade. Central American women are concentrated in sectors like commerce (general trade, tourism, related services), health, education and low-skill industry, and a large share of these jobs are in the informal sector, where women are likely to be self-employed in small-scale retail businesses. Occupational segregation tends to be positively linked to lower educational attainment and welfare status, as well as lower FLFP (the more women in the labor market, the more gender equality in the competition for available jobs; see Blau et al. 2006, Chang 2004). It has also been linked to social conventions and stereotypes, human capital formation, employer preferences and women’s preferences—in particular, preferences for shorter or flexible hours of work because of the unequal care burden and differential income roles (Bettio 2008, Anker 1998, Arias et al. 2005). Additional analysis suggests that earnings gap improvements happened across all sectors. This can be seen by the decreasing importance of the unexplained component of the decomposition (Figure 3.5). The higher reduction in commerce and trade are related to the fact that these two sectors employ more women than men. Given the fact that sectoral composition, concentration and segregation in Central America have remained stable over the decade, these results suggest that idiosyncratic changes in the labor market composition over the decade are not driving the earnings gap reduction. Figure 3.5: Changes in Unexplained Earnings Gap, by Economic Sectors [40]    Source: Hoyos and Ñopo (2010) with data from SEDLAC and World Bank. To summarize, examining the drivers behind changes in earnings gaps in Central America over the decade reveals a number of insights. Earnings gaps have decreased significantly and women in most countries have leveled or even surpassed men in earnings and are now ahead of the rest of Latin America. El Salvador, Guatemala, Honduras and Panama have been the best performers in wage gap reduction over the decade. Overachievement of women in human capital accumulation and other observable characteristics help explain these wage gaps dynamics, suggesting that increases in education have been paying off for women. However, the full potential of these gains are offset by unexplained factors, which are large and affect earnings gaps negatively, although their importance has declined over time. These unexplained factors are higher among segments of the population that are typically more disadvantaged (rural, with children, less educated). At the same time, those same segments of the population contributed to the highest decline in the unexplained component of wage gaps over the decade. What explains these trends? The results show that earnings gaps improvements happened across the board (family structure, education levels and economic sectors). At the same time, most of the reduction in unexplained gender earnings gaps happened within segments. This is consistent with recent findings by Cunningham and Jacobsen (2008) that within-group rather than between-group inequality is the key factor underlying high earnings inequality in four Latin American countries (including Guatemala). This suggests that changes in labor market composition were less of a factor [41]    in driving these trends, consistent with weak demand in the region as well as the fact that sectoral composition, concentration and segregation in Central America have remained stable over the decade. Traditionally, the segments with the largest gender earnings gaps in Central America are typically at the bottom of the welfare distribution and linked to economic sectors with lower productivity and greater flexibility. Hence, the results could be capturing factors such as preferences for added flexibility (for example to raise children), which come at the cost of lower earnings than equally productive males who do not need such flexibility. The final hypothesis is that the wage gap results are consistent with changes in cultural norms and attitudes by women or towards women that may affect labor market earnings. Chapter 5 explores these issues in more detail. 4 HETEROGENEITY AND THE POOR: REVISITING FLFP AND EARNING GAPS The results in Chapters 2 and 3 suggest that women in all six Central American countries have made significant advances over the decade by reducing labor market constraints and balancing work-home needs and preferences. These advances include increases in human capital accumulation and changes in family dynamics and household structures. While these may have allowed women to catch up in terms of earnings (with the exception of Costa Rica), improvements in accessing labor markets has been offset by limited dynamism in the labor market and labor demand, linked to the slow overall growth in the region over the period. This chapter explores whether these trends are distributed differentially along the welfare distribution. While the chapter is descriptive and does not provide country-specific analysis, exploring heterogeneity is useful to better understand the type of policies that may be more effective in reducing gender gaps. 4.1 HETEROGENEITY IN FLFP As in the rest of Latin America, variation in labor force participation levels and annual growth among Central American men is minimal across the income distribution (Figure 4.1). By contrast, women in the higher parts of the income distribution are in general significantly more likely to work [42]    across all countries, also consistent with trends in the rest of Latin America. Participation rates for women in less-poor households are as much as double those for poorer women. While gender gaps in LFP are close to zero among the wealthiest, they are the highest among the poorest in Central America. A closer look at the annual growth rates of FLFP reveals two types of performance across Central America. On one hand, in Costa Rica, El Salvador and Panama, women in the poorest households have made the most progress over the decade. In these countries, FLFP rates have been growing significantly faster than those of men (Figure 4.1). By contrast, in Guatemala, Honduras and Nicaragua, women above the 80th percentile of the distribution have entered the labor force faster than men and poorer women. In fact, labor force participation among poor women has declined in these three countries. [43]    Figure 4.1: LFP (2006) and Annual Growth (1997-2006), by Gender and Income Percentile Source: SEDLAC and World Bank. [44]    A number of circumstances make it difficult for women from low-income households to enter the labor market: they tend to have more children, less education and fewer opportunities to find help with domestic care tasks. They also have access to a limited range of jobs, most of which are characterized by low salaries and poor working conditions. Nonetheless, the narrowing gap in participation rates in three of six countries is a positive development. These happen to be the largest and most diverse countries of the group, which may help explain these results. Further exploration of participation rates between rural and urban areas reinforces this finding. Unlike the rest of Latin America, where only around 20 percent of the population resides in rural areas, almost half of the population in Central America is rural. In addition, while Latin America has experienced high urbanization rates over the last decade, this has not been the case for Central America. Since rural areas are significantly poorer than urban areas, exploring labor market trends between urban and rural women provides additional insights. For example, rural women are up to 40 percent less likely to be in the labor force compared to urban women, while participation between rural and urban men is the same or in some cases higher for rural men (Figure 4.2). Still, over the last decade, gender gaps have closed faster in rural areas, even more among the rural poor, especially in Costa Rica, El Salvador and Panama, mirroring the results from across the income distribution (Figure 4.3). Figure 4.2a: Ratio of Rural to Urban LFP (2006) 1,2  1  0,8  level  0,6  0,4  0,2  0  female  male  female  male  female  male  female  male  female  male  female  male  Costa Rica   El Salvador  Guatemala   Honduras  Nicaragua  Panama  Source: SEDLAC and World Bank. Figure 4.3: Annual Growth of FLFP by Income Percentile and Area of Residence (1997-2006) [45]         Source: SEDLAC and World Bank. 4.2 HETEROGENEITY IN EARNINGS GAPS Consistent with the results for LFP, the analysis of earnings gaps along the welfare distribution suggests heterogeneous performances. In El Salvador, Honduras, Nicaragua and Panama, poor women experienced the largest gains in earnings (Figure 4.4). By contrast, the earnings gaps reduced the fastest among the less poor in Costa Rica and Guatemala. [46]    Figure 4.4: Convergence in Earnings: Annual Growth Rates for Male to Female Earnings (1997-2006) Source: SEDLAC and World Bank. Changes in the unexplained component of the wage gaps within urban and rural areas reveal similar patterns. For example, the unexplained wage gap was higher in 2006 among urban residents (Figure 4.6). While gaps have in general declined across groups, the highest reduction in the unexplained component is among the rural segment of the population, typically seen as more disadvantaged and historically with the highest unexplained gender disparities. [47]    Figure 4.6: Change in Unexplained Component of Gender Earnings Gap (by Area of Residence) Source: Hoyos and Ñopo (2010) with data from SEDLAC and World Bank. To summarize, this chapter analyzes heterogeneous performance in FLFP and earning gaps along the welfare distribution. The results indicate that poor women in El Salvador, Honduras, Nicaragua and Panama have experienced faster reductions in gender gaps in earnings, while poor women in Costa Rica, El Salvador and Panama (the largest and more diverse economies in the region) have reduced gender gaps in FLFP. Out of the six countries in the study, poor women in El Salvador and Panama have caught up with men on both participation and earnings. Guatemala, on the other hand, is the only country where poor women experienced reductions in FLFP and a deterioration in earnings gaps. The results are to a large extent mirrored by urban-rural comparisons, as the poor are predominantly rural. While the next chapter explores potential reasons behind these differences, future research should deepen the analysis to better explore the channels by which these trends evolved. [48]    5 UNDERSTANDING TRENDS THROUGH POLICY-BASED EVIDENCE Three broad themes emerge from the analysis so far. First, human capital accumulation as well as changes in family dynamics in the form of family structure, female headship and presence of children can help explain some of the improvement in female labor force participation and the reduction of earnings gaps between 1997 and 2006. Second, modest growth and weak labor demand seem to have offset some of these gains by constraining employment creation and income generation. Third, poor women in the region made uneven progress in closing gender gaps in both participation and earnings. In some countries, it was the poorest (traditionally rural) segments of the population where women experienced the highest labor force participation growth and saw the earnings gaps reduce the fastest over the decade. In other countries, improvements occurred among the non-poor. This chapter further explores these heterogeneous results across countries by reviewing evidence from policies in the region that directly or indirectly may have played a role in easing constraints or changing preferences and attitudes with respect to income generation. The fact that many of the interventions examined are targeted towards the poor can also help explain the differential outcomes discussed above. Further, the growing role of female migration may be changing social norms and intra-household allocation dynamics, which also may be impacting women’s labor market outcomes. The focus of this chapter is to review policies for which rigorous evidence from existing evaluations can be drawn to understand potential channels by which observed changes and trends may have occurred. As such, the chapter emphasizes cases where causal inferences can be made to provide input for policy formation towards enhancing and sustaining gender gap improvements in the region, especially among the poor. Hence, the research is limited to available evidence, and does not claim to be either complete or to fully link to the previous chapters. 5.1 INCREASING FEMALE LABOR MARKET PARTICIPATION Women in Central America still lag behind other women in Latin America in terms of labor force participation, and gains have been much slower. This may relate to barriers to enter the labor market that go beyond educational achievements or difficulties reconciling productive and reproductive [49]    roles. Two areas of policy may have a potential impact on both the supply of and demand for women in Central America’s labor market: skills formation and vocational training to increase human capital and change employers’ perception on women’s abilities, and childcare provision to free women’s time and help conciliate productive and reproductive roles.. 5.1.1 SKILL FORMATION AND FLFP As the previous chapters show, the education attainment achievements of women compared men has not translated into large increases in labor force participation or changes in sectoral mobility. This is particularly true among the youth in Central America: while young adults (15-25 year olds) have benefited from the gains in educational attainment and schooling progress in 1997-2006, they are half as likely to participate in the labor force than adults across the region (Figure 5.1a). These trends have improved slightly over the decade in Costa Rica, Nicaragua and Panama, but have declined in El Salvador and Honduras (Figure 5.1b). While increased schooling and related delays in labor market entry may explain part of low youth participation rates, they also are suggestive of higher entry costs for youth to find jobs and benefit from their educational attainment. Figure 5.1a: Labor Market Participation (Circa 2006) 1  0,8  0,6  level  0,4  0,2  0  youth  adults  youth  adults  youth  adults  youth  adults  youth  adults  youth  adults  Costa Rica   El Salvador  Guatemala   Honduras  Nicaragua  Panama  Source: SEDLAC and World Bank. Figure 5.1b: Labor Market Participation Annual Growth (1997-2007) [50]    2,0%  1,0%  growth  0,0%  â€?1,0%  â€?2,0%  youth  adults  youth  adults  youth  adults  youth  adults  youth  adults  youth  adults  Costa Rica   El Salvador  Guatemala   Honduras  Nicaragua  Panama  Source: SEDLAC and World Bank. To explore the impact of job-related skills on labor force participation, this section reviews how training programs in Latin America have evolved over time and focuses on gender-specific evidence from existing interventions on skills formation and labor market insertion. When data exists, the section discusses evidence from Central America. The policy response in Latin America to improve the skills and transitions of workers has changed over time. In the past, the traditional supply-driven model focused on specialized worker training through centralized public providers. Traditional public sector training institutions were not designed with gender-specific objectives, but in practice, young women compose a large share of the beneficiaries. Despite their proliferation, evidence on the effectiveness of such training-only schemes is mixed at best. There are three key criticisms of these schemes. First, a number of studies conclude that while women may benefit more from training than men, the impacts on labor market entry have been small (Kluve et al. 2005, Greenberg et al. 2003, Betcherman et al. 2007). Chapters 1 and 2 corroborate these findings, considering the limited change in FLFP over the decade and the fact that these programs have been typical in the region. Second, training programs for women have tended to focus on the provision of skills related to “feminineâ€? activities. This may have reinforced the trends in segregation and lack of sectoral composition change discussed earlier.12 Third, even for                                                         12Sectoral or occupational concentration of women has been typically explained by various factors, including social conventions and stereotypes, gender roles, human capital formation or the preferences of employers or women (Blau et [51]    training that could potentially provide skills to attract women to male-dominated activities, the evidence suggests that in many cases it encouraged women to enter sectors that men were exiting in search for better opportunities—e.g. manual, technical work in manufacturing (Bettio 2008). Conversely, Tzannatos (2006) shows how legislative changes to improve job skills and quality allowed men to take available jobs in sectors traditionally dominated by women. This is especially evident in Central America’s maquila industry in the CAFTA free trade zones: as conditions improved, men started entering the sector, displacing female employment. A newer trend in skill formation programs has been to combine training with job placement and in many cases accompany trainees with additional support services and financial incentives (Betcherman et al. 2007). While the approaches vary, these combinations seem to have had better results, especially in promoting labor force participation by women (typically youth). Perhaps the most widespread program of this type has been Jóvenes, implemented in Chile, Argentina, Uruguay, Paraguay, Peru, Colombia, the Dominican Republic and Venezuela. Evidence from Jóvenes evaluations suggests that it has been successful in improving both job placement and earnings. Peru’s version (ProJoven) is a successful example. Started in 1996, the program finances the technical and practical training of youth between 16 and 24 with low levels of formal education, none or minimal labor market experience, and who are unemployed, underemployed or out of the labor force at the moment of enrollment in the program. Operating on a first-come first-served basis, by 2008, ProJoven had trained 20,000 young workers through more than 240 training entities and more than 1,000 courses. 13 An evaluation of the program showed that year and a half after starting their participation in the program women were 18 percent more likely to participate in the labor force than women that had                                                                                                                                                                                 al. 2006, Chang 2004). For example, women may be drawn to certain jobs that are easier to enter and exit, thus facilitating a better work/family balance (Maning 2006, Arias et al. 2005), or they may prefer shorter or flexible work hours (Bettio 2008, Anker 1998, Solber and Laughlin 1995).   13 A variation of this model is Mexico’s Probecat, which started in 1984. Although it is not centered on the youth or disadvantaged populations, the eligibility criteria favors these groups. Probecat does not rely on training institutions to provide courses. Instead, the program provides a stipend equivalent to the minimum wage to beneficiaries while the private sector businesses provide training and the internship (for a minimum of three months) and cover the training costs: this is the in-firm training modality. To participate in the program, private firms have to agree to keep for one year at least 70 percent of the trainees. A variant of this model was implemented in Central America (Honduras and El Salvador). [52]    not participated in the program. This compares with an 11 percent reduction for male beneficiaries14. This is most noticeable considering that women beneficiaries were less likely to be in the labor force than men beneficiaries, who were already in the labor force but unemployed. The increases in income from were also high among women. After the same time period of 18 months following program completion, women earned almost twice (93 percent) the income of those in the control group. Compared with a positive but muted impact among men (11 percent), it shows a positive impact of the program on women. Both men and women beneficiaries of ProJoven earned 20 percent higher hourly wages than non beneficiaries. Women participants were more likely to have formal sector jobs and evaluation results also found that the program reduced occupational segregation by encouraging training entities to focus on increasing female worker participation in occupations with a high prevalence of male workers. Success rates can be associated with the fact that the type of training offered was based on matching local firm needs with the capabilities of participating trainees. Guaranteed internship opportunities were provided for participants who successfully complete the training, reducing barriers of entry, and allowing for firms to decide after the three months if they wanted to commit or not to a longer term contract. Similar evidence on skills formation programs in Central America is scarce. Panama created Procajoven in 2002. With the objective to improve labor market prospects for jobless youth and disadvantaged groups, it had a similar design to Peru’s Projoven. The program offered two modalities: The “insertion modalityâ€? provided short-term training for low-income, unemployed youth of 18-29 combining technical training and soft skills (job readiness) training for the labor market, and a month’s internship in a firm. The “transition modalityâ€? targeted first-time job seekers with complete secondary education (age 16-23) and provided only the soft skills and a two-month internship. In both modalities, participants also receive a cash incentive. Up until 2006, 3,700 beneficiaries had received training.                                                         14 Theseresults were obtained using a non-experimental design based on pre-program matching and double difference techniques (Ñopo and Saavedra 2007, 2002) [53]    . A significant effect on job placement among women (but not among men) was found by an evaluation conducted in 200615. Labor force participation among women beneficiaries was higher by at least 12 percentage points 9-20 months after program completion for participants of both modalities (from 26% to 42% for the transition modality, and from 35% to 47% for the insertion modality) . These effects were larger in the capital city. The evaluation also finds an increase of more than 40 percent in earnings among women as well as a rise in hours worked (increase of seven hours per week) for the insertion modality. A final interesting finding from the evaluation is that cost of the training—excluding the transfers—is recovered in about a year for all participants, but that recovery time for women in the transition modality is only three months. One final example of related training schemes in Central America is the Entra 21 program. This program started in 2002 providing businesses with skilled ITC workers while improving the employability of disadvantaged youths (age 16-29 with complete or about to complete high school). .16 Lasting about two years, the program includes ITC skills training, an internship with a firm, life- skills training, continuous tutoring and stipends. Although no impact evaluation results exist yet, job placement rates have been reported to range from about 68 percent in Peru and 41 percent in Paraguay. In Panama, women have experienced lower post-program employment rates than men (34 compared to 64 percent). Overall, these results suggest that training and job placement programs can work, especially for younger, poor, women. In considering ways to learn from these schemes, a number of challenges need to be considered. First, the cost may be prohibitive for some countries in Central America, despite potential cost recovery. Second, these schemes assume that training institutions are capable of developing and providing quality training, conducting efficient matching with potential employers in the private sector, adjusting course content to local demands and ensuring effective follow-up with beneficiaries. In the case of Panama, implementation was slower than planned due in part to institutional factors. Third, policy design needs to consider the high level of labor market                                                         15 While no evaluation design was included in the program itself, a series of delays in the rollout of the program created a natural experiment that allowed for impact evaluation (Ibarrarán and Shady 2006) 16The program is implemented by NGOs and local businesses in 18 countries including El Salvador, Guatemala, Honduras, Nicaragua and Panama Implementing institutions and number of beneficiaries in Central America are: AGAPE El Salvador (480), PoA/CADI Guatemala (590), CADERH Honduras (320), Asociación Cristiana de Jóvenes Honduras (350), Don Bosco Nicaragua (400) and Cospae Panama (600).  [54]    concentration and segmentation in the region. Careful consideration about potential effects of skill formation schemes on gender market concentration and segmentation will be important. Finally, from a conceptual point of view, a remaining central issue with these schemes is the underlying assumption that unemployment or inactivity is due to the lack of skills that are demanded by the productive sector. It therefore assumes that basic skills already exist and that short course durations are sufficient to allow beneficiaries to learn about specific market demands. As noted earlier, overall human capital accumulation in the region has increased in the last decade but it is still low, and these schemes may not allow the poorer, less educated segments of the populations to benefit from them. As such, in exploring expansion of these schemes, careful consideration of the local context is needed. Pilots exploring program modifications in course content, duration, target groups (urban, rural women, sectors), local demand, benefits and linking to existing education curriculum could potentially provide countries in Central America better instruments to expand the labor force and income generation potential of women. 5.1.2 EXPANDING CHILD CARE: PRESENT AND FUTURE GENERATIONAL GAINS As Chapter 2 indicates, the rapid change of family structures in Central America—high levels of single-parent households and declining access to extended family networks as a source of care—may be increasing the need for childcare support. Childcare usage has increased in the last decade across all countries in the region (Figure 5.2a and Figure 5.2b). While utilization is still low, more than 40 percent of young children in Costa Rica, Honduras and Panama attend some form of childcare service, while in Guatemala less than 18 percent do. Still, all countries have experienced high growth rates over the decade, which can be linked to a combination of expanding services and increased demand. Figure 5.2a: Under 5 Childcare Assistance (2006) [55]    60%  40%  level  20%  0%  Guatemala   Nicaragua  El Salvador  Panama  Honduras  Costa Rica   Source: SEDLAC and World Bank, Figure 5.2b: Under 5 Childcare Assistance Annual Growth (1997-2006) 10%  8%  6%  growth  4%  2%  0%  Guatemala   Nicaragua  El Salvador  Panama  Honduras  Costa Rica   Source: SEDLAC and World Bank. Women with access to childcare are 10 to 30 percent more likely to be in the labor force (Figure 5.3a). Interestingly, the gap between women with and without childcare has somewhat decreased over the decade (Figure 5.3b). These trends are suggestive of potential benefits of access to childcare on labor market outcomes for women. Early studies have found that the cost of childcare has a significant impact on FLFP (Connelly 1992), and estimate elasticities of FLFP to a change in the average cost of child care at -0.20, and in many cases closer to -1.0. This section explores existing program evidence. [56]    Figure 5.3a: FLFP Ratio by Access to Childcare, Females 15-64 (2006) 1,4  1,2  level  1,0  0,8  Costa Rica   El Salvador  Nicaragua  Honduras  Panama  Guatemala   Source: SEDLAC and World Bank. The graph shows that ratio in FLFP rates between women with children and access to childcare as opposed to those without access to childcare. Figure 5.3b: FLFP Ratio Annual Growth by Access to Childcare, Females 15-64 (1997-2006) 0%  â€?1%  growth  â€?2%  â€?3%  Costa Rica   El Salvador  Nicaragua  Honduras  Panama  Guatemala   Source: SEDLAC and World Bank. A growing body of research suggests that pre-primary school can improve educational outcomes for children in the short and long run (Blau and Currie 2006; Schady 2006). In addition to the potential direct benefits of early childhood interventions on gender gaps for future generations, can they also have large impacts on mothers’ labor market outcomes? The literature on this is slowly emerging for Latin America, and while not conclusive, a number of interesting findings exist. Evaluating public childcare provision in Riol de Janeiro, Brazil, Deutsch (1998) suggests that lack of flexibility and limited opening hours of publicly-provided childcare significantly restrict the impact on women’s earnings capacity. Women who can afford the more costly private childcare receive significantly higher wages than women using public childcare. In a more causal setting a recent study by, Carvalho, Lunde, Olinto and Paes de Barros (2010) finds that, Rio women not previously in the [57]    labor market are 8 percentage points more likely to work if they have access to free daycare. Nonetheless, the authors do not find significant impacts of daycare access on overall FLFP or on hours worked, even though the coefficient is positive. Evidence from Argentina (Berlinski and Galiani 2007) also shows how a large expansion of free pre- school services is positively correlated with women’s employment, with estimated impacts ranging from 7 to 14 percent. A related study finds that women whose children had just made the cutoff age for pre-school eligibility were 13 percent more likely to be working compared to those whose children had not made the cutoff (Figure 5.4), although the results were not statistically significant. The study also finds that these women worked on average eight hours longer and were more likely to work full time17 . In Colombia, Attanasio and Vera-Hernandez (2004) evaluate the program Hogares Comunitarios, a community nursery program in rural Colombia, and find that the program increased the average probability that mothers work by 31 percent. Figure 5.4: Maternal Labor Outcomes and Child’s Day of Birth, by Age Group Source: Berlinski and Galiani, 2007.                                                         17 The study uses egression discontinuity to test this results. [58]    The evidence in Central America is scarce. One existing study in Guatemala (Quisumbing et al. 2003) finds that higher prices of informal care increases the utilization of formal daycare. No labor force participation effect is found. When they control for endogeneity of labor market participation and formal daycare use, the price of formal daycare has negative but insignificant impacts on mother’s earnings and hours worked. Taken together, the emerging evidence suggests potential benefits of childcare in reducing time constraints among women, especially for those in the labor force, by allowing them to work more, improve their incomes and transition to better jobs. Beyond addressing the limited supply of quality childcare programs in the region to improve early childhood development outcomes, additional evidence from ongoing and new interventions, especially from Central America, is urgently needed to help shed light on these potential impacts and to inform policy makers on how to best design and expand such schemes. 5.2 INCOME GENERATION AND PROMOTING ENTREPRENEURSHIP Self-employment and microenterprise employment are the most important sources of remunerated work for poor women in most Central American countries. Close to 40 percent of employed women are self-employed in El Salvador, Guatemala, Honduras and Nicaragua (Figure 5.5a). Female self- employment levels have increased in the last decade and are higher than male self-employment rates in the same four countries (Figure 5.5b). Figure 5.5a: Percent of Self-employment, Adults 15-64 (2006) 60  40  level  20  0  females  males  females  males  females  males  females  males  females  males  females  males  Costa Rica   El Salvador  Guatemala   Honduras  Nicaragua  Panama  Source: SEDLAC and World Bank [59]    Figure 5.5b: Self-employment Annual Growth, Adults 15-64 (1997-2006) 6%  4%  2%  growth  0%  â€?2%  â€?4%  â€?6%  females  males  females  males  females  males  females  males  females  males  females  males  Costa Rica   El Salvador  Guatemala   Honduras  Nicaragua  Panama  Source: SEDLAC and World Bank. Though generally labeled as self-employment, the term covers an heterogeneous sector, including “independent workersâ€? in one-person establishments or those employing only unpaid family members, employers of micro-establishments (2-5 workers) or employers of microenterprises (6-10 workers) (Table 5.1). More than 80 percent of men and women work in these smaller businesses and women represent up to 50 percent of small firm owners.18 This suggests that policies and programs targeted to these smaller businesses—such as microfinance—may be key to improve women’s overall employment and income generation opportunities.                                                         Bruhn (2009), Female-Owned Firms in Latin America—Characteristics, Performance, and Obstacles to Growth, 18 World Bank Policy Research Working Paper No. 5122. [60]    Table 5.1: Self-employment Based on Firm Size and Gender, 2007 (%) Country Independent Micro-establishments (2-5 Microenterprises (6-10 Small, medium, and Workers (one- workers) workers) large person/unpaid establishments family workers) Men Women Men Women Men Women Men Women Costa Rica 18.4 17.1 23.5 18.0 4.9 4.0 40.3 28.5 El Salvador 22.4 35.8 33.8 21.2 7.8 4.4 26.3 20.0 Honduras 39.0 42.3 29.9 15.3 1.6 1.2 n.d. n.d. Nicaragua 32.9 36.6 30.2 15.4 8.3 5.3 20.8 18.7 Panama 29.5 21.2 16.4 12.5 7.8 6.9 33.3 26.0 Source: ILO (2009). Notes: Rows do not add to 100 because unclassifiable private sector workers and public sector employees are excluded. The ILO defines small establishments as those employing between 11 and 50 workers; medium establishments 51-100 workers, and large establishments over 101 employees. The majority of Central American microenterprises do not have access to credit from any source, formal or informal. In fact, a recent report suggests that only 11 percent of Guatemalan households with microenterprises reported having received credit, compared to 23 percent in Nicaragua and 29 percent in Panama (Tejerina and Westley, 2007). Women micro-entrepreneurs were even less likely to have received credit: only 5 percent in Guatemala and 17 percent in Panama. However, over the past 10 years, the influx of microcredit programs into the region has accelerated rapidly. A 2005 inventory recorded 95 microfinance institutions operating in the six Central American countries, with a total portfolio in excess of US$1 billion and a client base of over 1 million borrowers (Table 5.2). The average loan in Central America is close to $1000, although average loan size varies across the six countries. [61]    Table 5.2: Supply of Microfinance in Central America (Circa 2005) Country Number Portfolio Borrowers of MFIs (US$ Millions) Nicaragua 21 261 399,614 Guatemala 24 273 363,286 El Salvador 11 138 143,461 Honduras 14 80 143,118 Costa Rica 19 341 45,607 Panama 6 16 28,103 TOTAL 95 1,109 1,123,189 Source: Navajas and Tejerina (2006). What has been the impact of microfinance programs on women’s employment and income in Central America, and what are the prospects for expanding and reforming such programs to optimize their benefits? While the literature on the gender impacts of microfinance has focused on Asia and Africa (e.g. Pitt et al. 2006, Ashraf et al. 2007), several studies from Central and Latin America exist. We focus on three issues: (i) the short-term human capital impacts; (ii) the longer- term effects on enterprise growth and employment; and (iii) the generation of positive community- level externalities. There is some evidence that Latin American microfinance programs may have a positive effect on health and schooling outcomes for children, particularly when they are bundled with non-financial education and services. This in turn implies better labor market outcomes in the future. In Honduras and Ecuador, participation in village banks integrated with health promotion activities targeted to low-income women is associated with lower incidence of child diarrhea, greater probability of breast-feeding and higher rates of maternal cancer screening (Smith 2002). In Bolivia, children from households that have been microfinance clients for at least one year have a greater chance of being kept in school compared to children from households just entering the program (Maldonado and Gonzalez-Vega 2008). In Guatemala, better credit access also has a positive effect on child schooling, but this impact may be mitigated when hired labor and child labor are not easily substitutable—for example, in enterprises characterized by a high potential for moral hazard by hired labor, and in households where parents display a preference for imparting skills to child labor rather than to hired labor (Wydick 1999). Children are also more likely to be withdrawn from school [62]    to work in a household enterprise when borrowers use funds for the purchase of capital equipment instead of strictly for working capital purposes, since as the business becomes more heavily capitalized, the marginal product of family labor increases, raising the opportunity cost of schooling. Are these generally positive impacts on child welfare achieved at the expense of investment and increased profitability of the microenterprise? In a randomized field experiment that provided grants to microenterprise owners in Sri Lanka, de Mel et al. (2009) found that female owners invested little of the money in their businesses and showed no increase in profits as a result of the injection of new capital. Male micro-entrepreneurs, on the other hand, turned the grants into a sustained source of income by making profitable investments in their enterprises. A longitudinal study of microenterprise borrowers in Guatemala shed more light on this. Kevane and Wydick (2001) argue that lower employment growth in female-owned enterprises is primarily a function of the high marginal value of home time for women during their childbearing years, rather than high supervision costs of hired labor. They find that while young male entrepreneurs tend to be more aggressive in using loans to generate employment than older male entrepreneurs, job growth is positively correlated with age of the business owner for women (Figure 5.6). This suggests that targeting microenterprise lending to younger women (under 35) may have the greatest returns with respect to child welfare, while loans to older female entrepreneurs have more potential to contribute to enterprise and employment growth. Indeed, over a five-year period, Guatemalan microenterprises operated by female entrepreneurs were more stable than those operated by men: drop-out rates from the credit program were much lower among female entrepreneurs, as were rates of transition back into the non-entrepreneurial labor market and foreign emigration (Wydick 2002). On average, female-owned enterprises continued to grow in terms of hired labor, albeit slowly, while growth in male enterprises was slightly negative five years after microfinance borrowing. [63]    Figure 5.6: Predicted Increase in Employees by Age of Entrepreneur (by Gender) Source: Kevane and Wydick (2001). An evaluation conducted with group of female micro-entrepreneurs from Peru receiving a combination of micro-lending, business training (general business skills and strategy training), and client-specific advice (how to target customers, identify competitors and position products, and, promotional and planning strategies)19. The study found significant benefits of the business skills training program. Female entrepreneurs improved their knowledge of business processes and increased sales, revenues and profits. Positive impacts were also observed among the subgroup of women that had a lower interest in receiving the business training, presumably by relaxing borrowing constraints and lowering the need for a buffer stock of liquidity, allowing an increase in consumption. In addition, children from mothers in the                                                         19 Using a randomized control trial, Karlan and Valdivia (2006) analyze the effect the business training as an add-ons provision to women working with Foundation for International Community Assistance in Peru (FINCA), a microfinance institution that supports poor, female entrepreneurs in Lima and Ayacucho through village banks. Both the treatment and control groups received loans under the program (US$203 on average). In addition, the treatment groups received 30 to 60 minutes of training during their normal weekly or monthly meetings with the bank over a period of one to two years. [64]    group receiving the training component spent significantly more time studying and less time on leisure-related activities. In addition to the gender-specific household- and firm-level impacts of microfinance, there is reason to believe that increasing the supply of credit for women may generate positive externalities and multiplier effects. An innovative study in rural Paraguay demonstrates that the introduction of a microfinance program had the unexpected consequence of increasing women’s excess demand for entrepreneurial capital via reference group, which made it more acceptable for women to pursue non-traditional entrepreneurial livelihoods (Fletschner and Carter 2008). Women in the credit program were exposed to more entrepreneurial social reference groups, which influenced their conception of the range of appropriate activities they could undertake, driving up their demand for capital. This interesting finding suggests a seldom-measured potential dynamic impact of gender- targeted credit programs: relaxing supply constraints may reconstruct gender norms and have a social multiplier effect, shifting an entire group or community to a higher-income equilibrium. Given the multitude of small-scale microfinance intervention in Central America, these results may be already at play and can help explain the earlier results on large improvements of gender gaps among poor women. The next section explores some further evidence related to changes in gender norms. 5.3 CHANGING GENDER NORMS As the analysis in Chapters 2 and 3 suggest, a number of unobserved factors seem to explain a large part of earnings gaps and female labor participation decisions, especially among the poor. Social norms regarding gender roles and stereotypes could very well have a strong say in shaping preferences of women and their families, as well as their potential employers. Norms can directly regulate women’s activities, or can influence their intra-household bargaining. Norms that directly regulate women’s activities may prescribe social penalties for not conforming (for example, failing to adequately perform housework activities), while norms operating more directly over the household bargaining include competitive advantages like skills and human capital endowments, legislation or gender-specific cash transfers. The challenge is to measure this in a credible way. [65]    This section looks at three types of interventions where credible evidence suggests that norms could be changing in Central America. These are of interest because they alter the bargaining position of women in the household and may thus change their economic empowerment and outcomes. 5.3.1 FEMALE MIGRATION AND INTRA-HOUSEHOLD RESOURCE ALLOCATION Since the 1980s, there has been a dramatic upsurge in migration within and out of Central America. As the previous sections suggest, low growth and few income opportunities may have induced this surge. This sizeable permanent or seasonal relocation of both men and women has important ramifications for household formation and livelihood strategies, as parents and children may be separated over long distances for significant periods of time, and the role of remittances has steadily grown as a source of income for many families. International migration is an increasingly defining characteristic of how Central American households survive, and the migration phenomenon is having profound implications both for women who go and women who stay. For women who decide to migrate seasonally or permanently, understanding the impact on the children they leave behind, how effective female remittances are at substituting for mother care and whether the change in intra-household resource allocation leads to differential household investment decisions all become relevant for policy design. In the case of women who remain in their countries of origin, migration of family members raises issues of their access to remittance flows and the impact of migration and remittances on female labor supply, entrepreneurship and the ability to provide for their families as de facto female household heads. Both suggest changes not just in resources but also in evolving social and gender norms. The largest flow of international migrants from Central America is to the United States. The 2000 U.S. Census counted close to 2 million Central Americans, and more recent estimates put the number closer to 2.8 million (Pew Hispanic Center, Statistical Portrait of the Foreign-Born Population in the United States, 2008). While about half of all Central American migrants in the U.S. are women, the gender composition varies by country of origin, as does the household composition of the migrants. For example, women comprise over 50 percent of migrants from Panama, Nicaragua, Costa Rica and Honduras, while migrant flows from El Salvador and Guatemala are dominated by men. The International Organization for Migration estimates that of the almost 1.5 million Guatemalans living abroad in 2007, approximately 28 percent were women (IOM/INSTRAW 2007). Female migrants from Honduras, Nicaragua and Panama are more likely [66]    to form their own households without husbands, while the rates of female headship are lower for Costa Ricans, Salvadorans and Guatemalans.20 Remittances from Central American migrants have been an increasingly important source of income for families in origin countries (Table 5.3). In 2008, Guatemalans sent home almost US$4.5 billion per year (11.4 percent of GDP), while Salvadorans remitted US$3.8 billion (17.2 percent of GDP). Remittances are also significant as a share of GDP in the poorest Central American countries of Honduras (US$2.8 billion, or 20.1 percent of GDP) and Nicaragua (US$818 million, or 12.4 percent of GDP). While most remittance data are not disaggregated by gender, it is certain that, given female migrant labor force participation rates of over 50 percent, a considerable proportion of these resources are coming from women. In Guatemala, over 25 percent of remittance senders are estimated to be women (IOM/INSTRAW 2007). It is also certain that a significant amount of remittances are sent to women who remain in the countries of origin, as siblings, children and spouses working abroad supplement domestic sources of income. A 2003 survey indicated that 63 percent of remittance receivers in Honduras, Guatemala and El Salvador were women (MIF/PHC 2003). Data from Guatemala suggest that of the almost US$4 million dollars in remittances sent to approximately 1 million households in 2007, more than two-thirds was received by women (IOM/INSTRAW 2007). Table 5.3: Migration in the US and Remittances to Central America Total % Women % Female Female labor Remittances Remittances number of household, force (US$ as % GDP migrants no husband participation millions) living in present rate in the the US U.S. Costa Rica 71,870 52.8 12.5 55.0 605 2.0 El Salvador 817,335 48.5 15.1 55.1 3,804 17.2 Guatemala 480,665 44.4 12.8 51.8 4,451 11.4 Honduras 282,850 50.6 17.9 54.3 2,824 20.1 Nicaragua 220,335 53.8 16.4 54.1 818 12.4 Panama 105,175 62.8 17.9 58.5 196 0.9 TOTAL 1,978,230 12,698 Sources: U.S. Census Bureau Foreign Born Profiles; World Bank Migration and Remittances Factbook 2008. Most research on the determinants of migration has not explored the potentially different incentives and constraints impacting the migration decision for women and men. The exception to this is in                                                         The Pew Hispanic Center tabulations of 2008 American Community Survey indicate that 19.2 percent of all Central 20 American households in the U.S. are headed by women. [67]    Mexico, where the availability of detailed migration data has allowed for analysis of the gender- specific impacts of individual, household and community-level factors on migration flows. For example, Richter and Taylor (2008) find that schooling is positively associated with international migration from Mexico by females but not by males, and that the effect of education for women is significant only for migration to non-agricultural jobs. The results also suggest that women migrate at older ages than men, and that family migration networks, which have a more important effect on migration decisions than macroeconomic and policy variables, are both gender and sector-specific. One of the few studies that examined gender differences in international migration for Central America (Massey et al. 1998) argues that prevailing household structures in origin countries generate distinct patterns of female relative to male migration. Comparing lifetime migration rates in Costa Rica, which has relatively high rates of marriage, with Nicaragua, which is characterized by high prevalence of informal consensual unions, divorce and separation, the study finds that Costa Rican women are less likely to migrate, especially if they are married. When married women do migrate, it is often in the context of being “tied movers,â€? accompanying other family members. In contrast, Nicaraguan women’s propensity to migrate is largely unaffected by their marital status or their spouse’s migration status, but strongly related to the woman’s own prior migratory experience, suggesting that “in matrifocal societies … women are more likely to migrate autonomously as active participants in international labor markets and to contribute to the process of cumulative causation, both in comparison to women in patriarchal societies and in comparison to men in their own settingâ€? (ibid. p. 84). When women migrate from Central America (and other low-income regions), one of the most common occupations they find is that of domestic worker. Census data from the 25 largest U.S. cities indicate that while low-skilled (less than high school education) female immigrants make up less than 2 percent of the total labor force, they filled over 25 percent of the jobs in the “private householdâ€? sector (Cortes and Tessada 2007). Numerous case studies of Central American communities in the United States have found that Central American women predominate in private domestic jobs in cities such as San Francisco, Washington, D.C., Houston, and Los Angeles (cf. Hondagneu-Sotelo 2007, Menjivar 2003). More than one-third of female Nicaraguan immigrants in Costa Rica are employed in the domestic service sector (Gindling 2008). [68]    Central American migrant women’s entry into the domestic service sector of host countries has been shown to have a positive impact on the labor force participation of more highly skilled women in those economies. In Costa Rica, which has received approximately 8,000 new Nicaraguan migrant workers every year since 2000, the increase in demand for Nicaraguan immigrant women in the domestic services sector may in part be due to the growth of more highly paid employment in the new export industries (apparel, electronics, and tourism) for low-skilled Costa Rican women. Gindling’s (2008) analysis of the impact of Nicaraguan immigration on the Costa Rican labor market finds a significant negative effect on the earnings of the least-educated Costa Rican-born women, but a significant positive effect on the wages of more highly educated Costa Rican female workers. One interpretation of these results is that Nicaraguan immigrants compete with less educated Costa Rican women in the market for domestic servants, while at the same time the increased supply of Nicaraguan women willing to work as domestic servants complements more skilled Costa Rican women, aiding them in obtaining employment at higher wages. Women’s migration also has an impact on the households they leave behind in their communities of origin, both by virtue of their absence from the family and through the additional income generated by the remittances they send home. Few studies differentiate between the effects of male and female migration and remittances on origin households, but we do have some evidence that these effects are gender-specific. In Mexico, for example, male but not female migration is associated with declines in non-staple crop production and wage income, while remittances from women but not from men are linked to lower expenditures on education and higher health spending (Pfeiffer and Taylor 2008). The researchers interpret the unexpected negative effect of female migration on education spending as either a response to the signal of low returns to education in migrant labor markets, or women’s loss of ability to monitor their children’s education and the household’s schooling investments. A recent study of seasonal migration and early childhood development in Nicaragua sheds further light on the potential impact of women’s migration and remittances on the children they leave behind (Macours and Vakis 2010). The analysis suggests that mothers’ migration has a uniquely beneficial effect on the cognitive development (measured by a standardized language test) and nutritional status of preschool-age children, implying that the relatively large household income gains from temporary employment in other regional labor markets more than compensate for the potentially negative effects of their absence (Figure 5.7). This positive impact of female migration on [69]    child outcomes may be attributable to a combination of intra-household empowerment associated with women’s larger contributions to household income, and the ability of extended family members to care for the children while the mother is away. In this sense, changes in intra-household resource allocation and bargaining can have impacts on human capital investments and therefore promote inter-generational gender equity Figure 5.7: The Effect of Female Migration on Inter-generational Gender Equity Source: Macours and Vakis 2010. It is also worth noting the effects of the migration of Central American men on the women who remain behind. Two studies from El Salvador have contrasting findings. Acosta (2006), using cross- sectional data and standard instruments for migration, concludes that remittances are negatively related to adult female labor supply—a finding consistent with the literature from Mexico. However, using panel data, location- and year-specific information on political violence and detailed time use records, Damon (2009) suggests that women reallocate their work hours away from housework and into on- and off-farm labor, as well as into family business activities. These findings highlight the evolving importance of seasonal and permanent migration as a viable income strategy among Central American households. The existing evidence is limited, but suggests [70]    that the increased migration over the decade is likely to be having profound implications for gender roles, intra-household resource allocation and welfare in general. Additional research is urgently needed to explore this emerging issue. 5.3.2 CHANGING INTRA-HOUSEHOLD RESOURCE ALLOCATION VIA CASH TRANSFERS Conditional cash transfer (CCT) programs to increase household asset endowments have proliferated in the last decade, especially in Latin America. . CCTs typically aim at increasing household investment in education, health and nutrition by providing cash to poor families conditional on keeping children in school, maintaining regular health care for family members or attending training programs for adults. By design, CCT programs do not address constraints for enhancing the income portfolio options of the poor. In fact (and as expected), the majority of impact evaluations find small to no impacts of the programs on labor force participation or hours worked by men or women (Fizbein and Schady 2010). Still, with the size of CCT transfers typically around 10-30 percent of household income, they have the potential to improve households’ physical asset base and short-term income generating potential by relaxing liquidity constraints. Equally important, since CCT programs are targeted to women, they might also facilitate asset accumulation by those household members that otherwise have the least control over household assets. Indeed, intra-household bargaining models, as well as related empirical evidence, indicate that a mother’s increased control over cash flows can have important effects on household investments in human capital, consumption patterns and income strategies (Duflo 2003; Quisumbing 2003; Duflo and Udry 2004). The evidence from CCTs confirms this: since women receive the benefits, this change in intra-household resource allocation is translated in improved investments in children in the form of improved nutrition, health and stimulation (see Fizbein and Schady 2009 for summary of impact evaluation findings). In this sense, CCT provide great instruments for promoting future gender equality. 5.3.3 INDUCING BEHAVIORAL CHANGES THROUGH SOCIAL INTERACTIONS An additional dimension of CCT that can lead to changes in social norms and attitudes is their role in social interactions. Appadurai (2004) suggests that mobilization by social movements can expand [71]    the capacity to aspire, in part through regular social gatherings and sharing ideas and experiences about future-oriented activities (such as savings or investments) among the poor. Ray (2006) proposes the concept of an aspiration window, where an individual, through social comparisons, draws her aspirations from the achievements and attitudes of her peers and near-peers. A widening of this aspiration window can influence investments and help raise future standards of living. Directly and indirectly, CCT integrate a number of opportunities for social interactions and exchanges among beneficiaries and non-beneficiaries. An emerging literature on the topic suggests that the effects of these interactions are not trivial. By exposure to their neighbors participating in the programs households can change their behavior. Angelucci and De Giorgi (2008) find that direct transfers to some households causes food consumption increases of even among those not eligible to receive such transfer, with participant households increasing their consumption in 20 percent, and the non beneficiaries doing so by 10 percent. Similar peer effects have been found in secondary school enrolment (Bobonis and Finan, 2008, Lalive and Catteneo, 2006, and Angelucci et al. 2009) indicating that social interactions are an important channel for changing attitudes. A recent evaluation of a cash transfer scheme in Nicaragua provides additional evidence on the role of social interactions, in this specific case in enhancing productive investment.21 The year-long Atención a Crisis pilot, combined a traditional CCT aimed at improving health, education and nutrition with additional interventions to increase the asset base and risk management capacity of rural poor households exposed to weather risk (droughts). While all beneficiary women received the cash transfers, one-third of them received an additional scholarship to send one household member to a vocational training course, and another third received a business grant and technical assistance for productive investments in livestock or non-agricultural activities. . The evaluation of Atencion a Crisis showed that, besides the expected impacts on education, consumption and nutrition -such as children of beneficiary households being more likely to attend school; increased per capita expenditures and quality of food intake of young children- on the productive side, income from non-agricultural self-employment for beneficiaries of the productive                                                         21Implemented by the Ministerio de la Familia between 2005 and 2006, the program targeted women in about 3,000 households that had experienced a severe drought shock in the preceding year. The pilot integrated an experimental evaluation framework to learn about both the short-term impacts of the program as well as the mid/long-term impacts More information about the pilot and the results can be found at www.worldbank.org/atencionacrisisevaluation and Macours and Vakis (2010). [72]    investment package almost doubled. Furthermore, the evaluation found that large and significant impacts on investments can be attributed to increased social interactions among beneficiaries. In particular, social interactions with female leaders who randomly received the productive transfer substantially increased program impacts on both human capital investments and income diversification and affected women’s attitudes towards the future. In fact, women who received the productive grant and worked in non-agricultural self-employment increased their incomes by an additional 100 percent (in addition to the overall program impact) when they were in a setting where all leaders around them had also received the productive grant (Figure 5.8) Figure 5.8: Additional Impact of Women’s Social Interactions on Income from Non- agricultural Activities (in Cordobas) 600 400 200 0 Control Women with productive grant Women with productive grant PLUS additional impact due to social interactions Source: Macours and Vakis, 2010. While social effects may result from a variety of mechanisms, the evidence indicates that changes in aspirations facilitated by increased communication and motivation by female leaders were important in this context. These results suggest that witnessing local success stories of upward mobility can change women’s investment behavior through motivation and encouragement, and also by providing positive examples that women can aspire to follow. From the policy point of view, these results point to the importance of ensuring that development program designs leave room to enhance social interactions with natural leaders by introducing program elements that create the space to share positive experiences, while at the same time providing resources to follow those [73]    examples. The evidence from the Atención a Crisis pilot suggests that these channels can be successful in inducing a change in women’s aspirations window. Such changes are likely to have implications for the feasibility and sustainability of cash or asset transfer programs in low-income countries, directly relevant in the policy debate in Central America. All countries in the region have been implementing some version of a CCT program in the last decade, although they have tended to be small and in some cases discontinued under resource and political economy pressures (e.g. Nicaragua). It is often argued that these countries simply cannot afford to distribute transfers to all poor households for long periods of time in a context of significant resource and capacity constraints. In this context, interventions that change household’s attitudes towards the future and social norms become key, and the experience from Nicaragua already suggests ways to launch households on a sustainable pathway out of poverty. 5.4 CONCLUSIONS Existing evidence from a range of policies in the region suggests a number of potential policy directions that could affect female labor force participation and improve earnings gaps: (i) policies that directly help women enter the labor market and improve incomes, such as skill formation, job placement services, micro-finance schemes and childcare; (ii) policies like conditional cash transfers targeted at poor women and policies that take into account the emerging role of female migration, which have both changed intra-household resource allocations and thus facilitated investments towards human capital accumulation and future gender equality and potentially inducing changes in social and behavioral norms and preferences; and (iii) policies and programs targeted at poor women that facilitate social interactions, knowledge exchange and changes in behavior and attitudes, leading to higher investments in human or physical capital and higher incomes. For a region like Central America, renowned for its plethora of small-scale private sector-led initiatives run by NGOs and other local organizations and targeted at the poor, these hypotheses open up new directions for policy dialogue and research. [74]    6 THE ROAD AHEAD During the period 1997-2006, Central America experienced economic and political stability, overall modest but positive economic growth and improvements in socio-economic outcomes. However, a review of the evolution of gender gaps related to income generation in the region presents a mixed picture. Labor force participation of women in Central America has remained unchanged, with one in every two women of working age still not participating economically. This is lower than the rest of Latin America, a region with the lowest labor force participation in the world, despite significant improvements in recent years. And yet, during the same period women in the region have made major advances in closing the earnings gaps and increasing their overall contribution in terms of income generation. The report’s findings suggest that improved education has been a driving force behind women’s decisions to enter the labor market, especially among younger women. For the first time, gender gaps in education closed in the last decade; in fact, women have actually surpassed men in educational attainment in most countries of the region. In addition, family structure dynamics, such as an increasing female-headed household population and lower fertility rates over the decade, have also played a role in reducing constraints linked to participation. Despite these positive changes, accessing labor markets has been hindered by limited dynamism in the labor market and labor demand, linked to the slow economic growth in the region over the period. Estimates indicate that bringing the labor force participation of Central American women up to the world average for their level of GDP per capita could result in up to 2 million people exiting poverty. Gender concentration and segregation across sectors and industries remained unchanged, while overall job creation was limited. Only Panama and Costa Rica (the largest and more diverse economies in the regions) have seen some modest increase in women’s labor market entry. Hence, the potential for improvement is great. On the earnings front, for the first time women in Central America have equalized or even exceeded men’s earnings, and are ahead of the rest of Latin America. These gains can be partly attributed to the overachievement of women in human capital accumulation over the decade and to changes in family structure dynamics. Even more important, the findings show that in most countries, earnings [75]    equality has been driven mainly by improvements among the poorer segments of the population (poorer, rural, with more children, less educated). Evidence from decomposing the changes in earnings gaps suggest that one of the channels may be linked to changes in attitudes and social norms. A closer look along the welfare distribution reveals heterogeneous performances. While poor women in El Salvador, Honduras, Nicaragua and Panama have experienced faster reductions in gender gaps in earnings, poor women in Costa Rica, El Salvador and Panama have reduced gender gaps in female labor force participation. The best performers are El Salvador and Panama, where poor women have caught up with men in both labor force participation and earnings. Guatemala, on the other hand, is the only country where poor women experienced reductions in labor force participation and a widening earnings gap. The results are to a large extent mirrored by urban-rural comparisons, as the poor are predominantly rural. In exploring further potential channels for these results, the report reviews existing evidence from innovative programs from Central America and the rest of the Latin America region. Three broad insights emerge. First, policies aiming at improving the returns of female labor by reducing labor market constraints and enhancing skills have shown promise, particularly among the poor. These include skills formation schemes, especially those that combine skills with job placement services that induced female entry in the labor force. Similarly, the availability of childcare has facilitated the ability of young mothers to work more by reducing time constraints at home. As well, microfinance schemes have helped poor female entrepreneurs by alleviating credit constraints and improving business-related skills through training. Second, policies that have changed intra-household resource allocations by putting money in the hands of women have facilitated investments towards human capital accumulation and future gender equality, potentially inducing changes in social and behavioral norms and preferences. These programs include conditional cash transfers targeted at poor women that have proliferated in the region over the decade. The emerging role of female migration is also a major shift in recent years, and should be taken into account in policy design. Third, many of the programs above targeted at poor women directly or indirectly facilitated social interactions through increased communication among female beneficiaries, leaders and community members. A growing body of evidence shows that such interactions can promote knowledge [76]    exchange and induce important changes in behavior and attitudes, which can in turn translate into better economic outcomes in the form of improved or higher investments and incomes. For a region like Central America, renowned for its plethora of small-scale private sector-led initiatives run by small NGOs and other local organizations, these findings open up new directions for policy dialogue and research. Taken together, these findings help inform choices facing government policymakers in the region, although the limited fiscal space and ad hoc supply of such policies in Central America may constrain additional advances. For example, Costa Rica’s 47 social insurance and assistance programs are run by 20 government agencies, with looming inefficiencies and coordination failures. And while conditional cash transfers and related safety net programs may be larger and better targeted, they do not explicitly address income generation issues. The report highlights a few policy directions to expand women’s income generation opportunities, including: (i) Policies to facilitate labor market entry. Evidence highlights the potential impact of small changes in the sectors where women are developing their income generation activities. Removing barriers to enter the market may address both sector and formal/informal work allocation. Skill formation schemes that combine soft skills and on the job training (with emphasis on skills demanded by high-productivity sectors) can potentially provide a successful transition to employment. Job intermediation services—in the form of job offices or employment agencies—may also help to link women with existing labor opportunities by acting as a ‘one-stop shop’ providing information on employment opportunities (lowering search costs), training and skills formation, and labor regulation on benefits and protection. Provision of childcare services can enhance the ability of working women to take advantage of labor market opportunities and change employer perceptions of the reliability of their female workers. This includes both daycare services for younger children under five years, and also after school activities for older children. (ii) Interventions that allow women to develop or expand existing businesses. For women who have opted for self-employment, increasing the productivity of their businesses may result in higher returns. Policies supporting female entrepreneurship should be tailored to the different types of business ownership by allowing female [77]    entrepreneurs grow their businesses and improve productivity. The evidence suggests the need for policy emphasis on improving growth of female-owned firms, rather than on firm creation. Promoting improved access to finance, high-quality business training, networks and new markets can open new paths for firm productivity growth and expansion. (iii) Policies that promote empowerment and aspirations, which can bring changes in decision-making and affect development outcomes. The emerging evidence shows a large potential for income generation by changing attitudes via social interactions and enhanced social networks. Closer attention is therefore needed to program elements that can facilitate or enhance such social interactions and further encourage shifting aspirations and social norms with a view to improving development outcomes. Policy actions to encourage growth and development of new market niches with the capacity for job creation for women, such as in the service sector where much of the female labor force is employed, could be explored. Further, the normative framework regulating economic activity may need to be adapted to prevent unintended biases or gender disincentives. Beyond these, broader policies that deal with human capital development (universal education, expanded secondary education and improved education quality) need to be reviewed and strengthened. Similarly, industrial and labor market policies capable of generating important scale effects may also be needed to stimulate across- the-board demand for additional female labor. On the research front, a number of areas present large knowledge gaps that are limiting effective policy dialogue and design. First, the growing prevalence of seasonal and permanent migration in the region is a phenomenon that merits further analysis. Beyond income and labor market implications, the evidence shows that migration may be affecting social norms, attitudes and intra- household dynamics, which in turn can have additional impacts on welfare. Understanding in more depth the costs and benefits related to seasonal and permanent migration is therefore important. Second, shedding light on intra-household time allocation dynamics could allow a better understanding of the way work and family responsibilities have evolved along gender dimensions. This could also help explore links to inefficient outcomes such as youth delinquency, shorter career paths, flatter earnings curves and the intergenerational transfer of poverty. Third, a deeper [78]    understanding of the links between empowerment and economic development is urgently needed. The emerging evidence suggests that changes in aspirations can play a large role in improving decision-making, investments and income generation potential for women. [79]    BIBLIOGRAPHY Acosta, Pablo. 2006. “Labor Supply, School Attendance, and Remittances from International Migration. The Case of El Salvador.â€? World Bank Policy Research Paper 3903. Angelucci, Manuela, and Giacomo De Giorgi. 2008. “Indirect Effects of an Aid Program: How do Cash Transfers Affect Ineligibles’ Consumption.â€? American Economic Review, forthcoming. Angelucci, Manuela, Giacomo De Giorgi, Marcos Rangel, and Imran Rasul. 2009. “Family Networks and School Enrollment: Evidence from a Randomized Social Experiment.â€? Mimeo, University of Arizona. Anker, Richard. 1998. “Gender and jobs: Sex segregation of occupations in the world.â€? Geneva, ILO. Appadurai, Arjun. 2004. “The Capacity to Aspire.â€? In Rao, V. and M. Walton, eds., Culture and Public Action. Palo Alto, CA: Stanford University Press. Appiah, Elizabeth N. and Walter W. McMahon. 2002. “The social outcomes of education and feedbacks on growth in Africa.â€? Journal of Development Studies 38: 27-68. Arias, Omar, Andreas Blom, Mariano Bosch Mossi, Wendy Cunningham, Ariel Fiszbein, Gladys Lopez-Acevedo, William F. Maloney, Jaime Saavedra, Carolina Sánchez-Páramo, Mauricio Santamaria, and Lucas Siga. 2005 “Pending issues in protection, productivity growth, and poverty reduction.â€? Policy Research Working Paper Series No 3799. The World Bank, Washington, D.C. Armendariz, Beatriz, and Nigel Roome. 2008. “Gender empowerment in microfinance.â€? In S. Sundaresan ed, Microfinance: Emerging trends and challenges. Cheltenham, U.K. and Northampton, Mass. Ashraf, Nava, Dean S. Karlan, and Wesley Yin. 2007. “Female empowerment: Impact of a commitment savings product in the Philippines.â€? C.E.P.R. Discussion Papers, DP6195. Attanasio, Orazio, and Marcos Vera-Hernández. 2004. “Nutrition and Child Care Choices: Evaluation a Community Nursery Program in Rural Colombia.â€? Centre for the Evaluation of Development Policies. The Institute for Fiscal Studies, EWP04/06. November 2004. Revised March 2007. Bayard, Kimberly, Judith Hellerstein, David Neumark, and Kenneth Troske. 2003. “New Evidence on Sex Segregation and Sex Differences in Wages from Matched Employee-Employer Data.â€? Journal of Labor Economics 21(4): 887–922. [80]    Baydas, Mayada M., Richard L. Meyer, and Nelson Aguilera-Alfred. 1994. “Discrimination against women in formal credit markets: Reality or rhetoric?â€? World Development 22(7):1073-1082. Berlinski, Samuel, and Sebastian Galiani. 2007. “The Effect of a Large Expansion of Pre-Primary School Facilities on Preschool Attendance and Maternal Employment.â€? Labour Economics 14: 665-80. Berlinski, Samuel, Galian Sebastian and Patrick J. McEwan. 2008. “Preschool and Maternal Labor Market Outcomes: Evidence from a Regression Discontinuity Design.â€? IFS Working Papers W09/05, The Institute of Fiscal Studies, London. Betcherman, Gordon, Martin Godfrey, Susana Puerto, Friederike Rother, and Antoneta Stavreska. 2007. “Global Inventory of Interventions to Support Young Workers: Synthesis Report.â€? The World Bank. Bettio, Francesca, and Alina Verashchagina. 2008. “Gender segregation in the labor market: root causes, implications and policy responses in the EU.â€? EU Expert Group on Gender and Employment (EGGE). September. Blau, David, and Janet Currie. 2004. “Preschool, Day Care, and Afterschool Care: Who's Minding the Kids?â€? NBER Working Papers 10670, National Bureau of Economic Research, Inc. Blau, Francine D., Mary C. Brinton and David B. Grusky, eds. 2006. The declining significance of gender. New York: The Russell Sage Foundation. Bobonis, Gustavo J., and Frederico S. Finan. 2008. “Endogenous Social Interaction Effects in School Participation in Rural Mexico.â€? Review of Economics and Statistics, forthcoming. Bruhn, Miriam. 2009. “Female-Owned Firms in Latin America – Characteristics, Performance, and Obstacles to Growth.â€? TheWorld Bank Policy Research Working Paper No. 5122. Calonico, Sebastian, and Hugo Ñopo. 2008. “Gender Segregation in the Workplace and Wage Gaps: Evidence from Urban Mexico 1994-2004.â€? RES Working Papers 4579, Inter-American Development Bank, Research Department. Carvalho, Mirela de, Trine Lunde, Pedro Olinto and Ricardo Paes de Barros. 2010. “Impact of Free Childcare on Female Labor Force Participation: Evidence from Low-Income Neighborhoods in Rio de Janeiro.â€? Mimeo, The World Bank. Chang, Mariko Lin. 2004. “Growing Pains: Cross-National Variation in Sex Segregation in Sixteen Developing Countries.â€? American Sociological Review 69(1): 114-137. Connelly, Rachel. 1992. “The Effect of Child Care Costs on Married Women's Labor Force Participation.â€? The Review of Economics and Statistics, MIT Press 74(1): 83-90. Cortés, Patricia and José Tessada. 2007. “Cheap Maids and Nannies: How Low-Skilled Immigration Is Changing the Labor Supply of High-skilled American Women.â€? Working Paper. [81]    Cunningham, Wendy, and Joyce P. Jacobsen. 2008. “Earnings Inequality Within and Across Gender, Racial, and Ethnic Groups in Four Latin American Countries.â€? World Bank Policy Research Paper 4591. Cruces, Guillermo, and Sebastian Galiani. 2007. “Fertility and Female Labor Supply in Latin America: New Causal Evidence.â€? Labor Economics 14(3): 565–73. Deere , Carmen Diana, Cheryl Doss, and Caren Grown. 2009. “Collecting individual level asset data for gender analysis of poverty and rural employment.â€? Paper presented at the FAO-IFAD-ILO workshop on gaps, trends and current research in gender dimensions of agricultural and rural employment: differentiated pathways out of poverty, 31 March - 2 April 2009, Rome. Deutsch, Ruthanne. 1998. “Does Child Care Pay? Labor Force Participation and Earnings: Effects on Access to Child Care in the Favelas of Rio de Janeiro.â€? RES Working Papers 4140, Inter- American Development Bank, Research Department. Deutsch, Ruthanne, Andrew Morrison, Hugo Nopo, and Claudia Piras. 2005. “Working Within Confines: Occupational Segregation by Gender for Three Latin American Countries.â€? The ICFAI Journal of Applied Economics, ICFAI Press, 0(3). Dollar, David, and Roberta Gatti. 1999. “Gender inequality, income, and growth: are good times good for women?â€? World Bank Policy Research Report Working Paper 1. Washington D.C.: The World Bank. Duflo, Esther. 2005. “Gender Equality in Development.â€? (http://econ-www.mit.edu/files/799) Duflo, Esther and Christopher Udry. 2004. “Intrahousehold Resource Allocation in Cote d'Ivoire: Social Norms, Separate Accounts and Consumption Choices.â€? NBER Working Papers 10498, National Bureau of Economic Research, Inc. Duflo, Esther. 2003. “Grandmothers and Granddaughters: Old-Age Pensions and Intrahousehold Allocation in South Africa.â€? World Bank Economic Review, 17(1): 1-25. Duryea, Suzanne, Alejandra Cox Edwards and Manuelita Ureta. 2001. “Women in the LAC Labor Market: The Remarkable 1990s.â€? Women in Development Working Paper. , Washington, D.C.: Inter-American Development Bank. Duryea, Suzanne, and Miguel Székely. 1998. “Labor Markets in Latin America: A Supply-Side Story.â€? RES Working Papers 4120, Inter-American Development Bank, Research Department. Enamorado, Ted, Carolina Izaguirre and Hugo Ñopo. 2009. “Gender Wage Gaps in Central American countries: Evidence from a Non-Parametric Approach.â€? Inter-American Development Bank Working Paper Series, 111. Fiszbein, Ariel, and Norbert Schady. 2010. “Conditional Cash Transfers. Reducing Present and Future Poverty.â€? Washington, D.C.: The World Bank. [82]    Fletschner, Diana, and Michael R. Carter. 2008. “Constructing and reconstructing gender: Reference group effects and women’s demand for entrepreneurial capital.â€? The Journal of Socio-Economics 37: 672–693. Forbes, Kristin J. 2000. “A reassessment of the relationship between inequality and growth.â€? American Economic Review 90: 869–887. Gindling, Thomas H., and Luis Oviedo. 2008. “Single Mothers and Poverty in Costa Rica.â€? IZA Discussion Paper No. 3286. Bonn: Institute for the Study of Labor. Gindling, Thomas H. 2008. “South-South Migration: The Impact of Nicaraguan Immigrants on Earnings, Inequality and Poverty in Costa Rica.â€? IZA Discussion Papers 3279, Institute for the Study of Labor (IZA). Greenberg, David H., Charles Michalopoulos, and Philip K. Robins. 2003. “A Meta-Analysis of Government-Sponsored Training Programs.â€? Industrial & Labor Relations Review. 57(1) 2003 Article 2 Hondagneu-Sotelo, Pierrette. 2007. Domestica: Immigrant Workers Cleaning and Caring in the Shadows of Affluence. (New edition, with new preface "The Domestic Goes Global," in 2007). University of California Press. Ibarrarán, Pablo, and David Rosas Shady. 2008. “Evaluating the Impact of Job Training Programs in Latin America: Evidence from IDB funded operations.â€? Inter-American Development Bank. International Labour Organization. 2009. “2008 Labour Overview. Latin America and the Caribbean.â€? Lima: ILO/Regional Office for Latin America and the Caribbean. Kevane, Michael, and Bruce Wydick. 2001. “Microenterprise lending to female entrepreneurs: Sacrificing economic growth for poverty alleviation?â€? World Development 29(7): 1225-1236. Klasen, Stephan. 1999. “Does gender inequality reduce growth and development?â€? World Bank Policy Research Report Working Paper 7. Washington D.C.: The World Bank. Klasen, Stephan. 2002. “Low schooling for girls, slower growth for all.â€? World Bank Economic Review 16: 345–373. Klasen, Stephan. 2005. “Economic growth and poverty reduction: measurement and policy issues.â€? OECD Development Centre, Working Paper 246. Klasen, Stephan. and Francesca Lamanna. 2003. “The impact of gender inequality in education and employment on economic growth in the Middle East and North Africa.â€? Washington D.C.: The World Bank. Klasen, Stephan. 2006. “Pro-Poor Growth and Gender Inequality.â€? Ibero America Institute for Economic Research (IAI) Discussion Papers 151. [83]    Mammen, Kristin and Christina Paxson. 2000. “Women's Work and Economic Development.â€? The Journal of Economic Perspectives 14(4): 141-164. Kluve, J., D.Card, M. Fertig, M. Góra, L. Jacobi, P. Jensen, R. Leetmaa, L. Nima, E. Patacchini, S. Schaffner, C. M. Schmidt, B. van der Klaauw, A. Weber. 2007. Active Labor Market Policies in Europe: Performance and Perspectives. Berlin and Heidelberg: Springer. Knowles, Stephen, Paula K. Lorgelly and P. Dorian Owen. 2002. “Are educational gender gaps a brake on economic development? Some cross-country empirical evidence.â€? Oxford Economic Papers 54: 118–149. Lalive, Rafael, and Alejandra Cattaneo. 2006. “Social Interactions and Schooling Decisions.â€? IZA Discussion Paper No. 2250 Macours, Karen, and Renos Vakis. 2008. “Changing households’ investments and aspirations through social interactions: Evidence from a randomised transfer program in a low-income country.â€? Mimeo, Johns Hopkins University and World Bank. Macours, Karen, and Renos Vakis. 2008b. “Seasonal Migration and Early Childhood Development.â€? World Development forthcoming. Maldonado, Jorge H. and Claudio Gonzalez-Vega. 2008. “Impact of Microfinance on Schooling: Evidence from Poor Rural Households in Bolivia.â€? World Development 36(11): 2440–2455. Manning, Alan, and Barbara Petrongolo. 2006. “The Part-Time Pay Penalty for Women in Britain.â€? IZA Discussion Paper No. 2419. Massey, Douglas S., Joaquin Arango, Graeme Hugo, Ali Kouaouci, Adela Pellegrino, and J. Edward Taylor. 1998. Worlds in Motion: Understanding International Migration at the End of the Millennium. Oxford: Oxford University Press. Medeiros, Marcelo and Joana Costa. 2008. “What Do We Mean by Feminization of Poverty?â€? One Pager 58, International Policy Centre for Inclusive Growth. de Mel, Suresh, David McKenzie and Christopher Woodruff. 2009. “Are women more credit constrained? Experimental evidence on gender and microenterprise returns.â€? American Economic Journal: Applied Economics 1(3): 1-32. Mitchell, Maureen. 1999. “From survivor to entrepreneur: Gendered dimensions of microenterprise development in Peru.â€? Environment and Planning A 31(2): 251-271. Morrison, Andrew, Dhushyanth Raju, and Nistha Sinha. 2007. “Gender Equality, Poverty and Economic Growth.â€? Policy Research Working Paper No. 4349. Washington, D.C.: The World Bank. [84]    Navajas, Sergio, and Luis Tejerina. 2006. “Microfinance in Latin America and the Caribbean: how large is the market?â€? Sustainable Development Department Best Practices Series MSM-135. Washington, D.C.: Inter-American Development Bank. Ñopo, Hugo, Jaime Saavedra-Chanduví, and Miguel Robles. 2007. “Occupational Training to Reduce Gender Segregation: The Impacts of ProJoven.â€? RES Working Papers 4553, Inter-American Development Bank, Research Department. Ñopo, Hugo, Miguel Robles, and Jaime Saavedra. 2002. “Una medición del impacto del Programa de Capacitación Laboral Juvenil PROJoven.â€? Documentos de Trabajo dt36, Grupo de Análisis para el Desarrollo (GRADE). Ñopo, Hugo, and Alejandro Hoyos. 2010. “Gender Wage Gaps in Central America.â€? Background paper. Ñopo, Hugo, Juan Pablo Atal, and Natalia Winder. 2009. “New Century, Old Disparities: Gender and Ethnic Wage Gaps in Latin America.â€? RES Working Papers 4640, Inter-American Development Bank, Research Department. Ñopo, Hugo. 2008. “Matching as a Tool to Decompose Wage Gaps.â€? Review of Economics and Statistics 90(2): 290-299. Piras, Claudia, and Laura Ripani. 2005. “The Effects of Motherhood on Wages and Labor Force Participation: Evidence from Bolivia, Brazil, Ecuador and Peru.â€? Sustainable Development Department Technical Papers Series WID–109. Pitt, Mark M., Shahidur R. Khandker, Jennifer Cartwright. 2006. “Empowering Women with Micro Finance: Evidence from Bangladesh.â€? Economic Development and Cultural Change 54(4):791-831. Preston, Jo Anne. 1999. “Occupational Gender Segregation: Trends and Explanations.â€? The Quarterly Review of Economics and Finance 39. Psacharopoulos, George and Zafiris Tzannatos. 1992. “Latin American women's earnings and participation in the labor force.â€? Policy Research Working Paper Series 856. The World Bank, Washington, DC. Quisumbing, Agnes R., Kelly Hallman, and Marie T. Ruel. 2003. “Maquiladoras and market mamas.â€? FCND discussion papers 153, International Food Policy Research Institute (IFPRI). Ray, Debraj. 2006. “Aspirations, Poverty and Economic Change.â€? In Abhijit V. Banerjee, Roland Benabou, and Dilip Mookherjee, eds., Understanding Poverty. Oxford: Oxford University Press. Richter, Susan, and J. Edward Taylor. 2008. “Gender and the Determinants of International Migration from Rural Mexico over Time.â€? In Andrew Morrison, et al, eds., The International Migration of Women. The World Bank and Palgrave Macmillan [85]    Schady, Norbert. 2006. “Early childhood development in Latin America and the Caribbean.â€? Policy Research Working Paper Series 3869. Washington, D.C.: The World Bank. Solberg, Eric, and Teresa Laughlin. 1995. “The Gender Pay Gap, Fringe Benefits, and Occupational Crowding.â€? Industrial and Labor Relations Review, 48(4): 692-708. Smith, Stephen C. 2002. “Village Banking and Maternal and Child Health: Evidence from Ecuador and Honduras.â€? World Development 30(4): 707–723. Tejerina, Luis and Glenn D. Westley. 2007. “Financial services for the poor: household survey sources and gaps in borrowing and saving.â€? Sustainable Development Department Technical Papers series POV-117. Washington, D.C.: Inter-American Development Bank. Tzannatos, Zafiris. 1998. “Women and Labor Market Changes in the Global Economy: Growth Helps, Inequalities Hurt and Public Policy Matters.â€? The World Bank, Social Protection Advisory Service, Washington, D.C. Valdivia, Martin, and Dean S. Karlan. 2006. “Teaching Entrepreneurship: Impact of Business Training on Microfinance Clients and Institutions.â€? Working Papers 941, Economic Growth Center, Yale University. World Bank. 2010a. “Regional Study on Gender in LAC: Linking Labor Market Outcomes and Intra-Household Dynamics.â€? World Bank. 2010b. “Turkey Country Gender Assessment.â€? Draft. World Bank. 2009. “Global Financial Crisis: Assessing Vulnerability for Women and Children, Identifying Policy Responses.â€? World Bank. 2008. “Nicaragua Poverty Assessment.â€? Central America Country Management Unit. World Bank. 2007. “Millennium Development Goals: Confronting the Challenges of Gender Equality and Fragile States.â€? Global Monitoring Report. World Bank. 2006. “Honduras Poverty Assessment. Attaining Poverty Reduction.â€? Central America Country Management Unit Report No. 35622-HN, June. World Bank. 2001. Engendering Development: Through Gender Equality in Rights, Resources, and Voice. New York: Oxford University Press Wydick, Bruce. 1999. “The effect of microenterprise lending on child schooling in Guatemala.â€? Economic Development and Cultural Change 47(4): 853-869. Wydick, Bruce. 2002. “Microfinance among the Maya: Tracking the progress of borrowers.â€? Development and Change 33(3): 489-509. [86]    APPENDIX A1: ADDITIONAL TABLES Table A1.1: Data Sources [87]    Table A1.2: Socio-economic Statistics by Gender, Circa 2006 [88]    Table A1.3: Sectoral Distribution by Gender, Circa 2006 [89]    Table A1.4: Relative Wages, by Characteristics (Excludes Guatemala) [90]    Table A1.5: Raw and Unexplained Gender Earnings Gap, by Country *** Statistically different than zero at the 99% level. ** Statistically different than zero at the 95% level. * Statistically different than zero at the 90% level. [91]    Table A1.5: Earnings Gap Decomposition—Demographic Characteristics [92]    Table A1.6: Earnings Gap Decomposition—Job-Related Characteristics [93]    APPENDIX A2: WAGE DECOMPOSITION METHODOLOGY The methodology used in the paper to decompose the components of wage inequality follows the matching methodology introduced by Ñopo (2008). This method is an extension to the traditional Blinder-Oaxaca decompositions and is a non-parametric approach based on matching by individual characteristics. The matching procedure consists of creating combinations of characteristics for which males and females from the sample can be found. After the matching procedure is complete, the sample of males and females is divided into three groups: (i) unmatched females, (ii) unmatched males, and (iii) matched females and males. On the third group, males’ observations are weighted in a way such that their joint distribution of observable characteristics mimics that of matched females. The total gender wage gap (∆) is defined as the difference between average male and female wages, as a proportion of average females’ wages. It is then decomposed into four additive elements: ∆=(∆x+∆f+∆m)+∆0 ∆0 is the unexplained component of the wage gap and is calculated as the difference between average males’ and females’ wages as proportion of the average female wage on the matched (and hence reweighted) sample. ∆x is the component of the wage gap that is explained by the gender differences in individual characteristics. It is calculated as the difference between the average wage of males in the common support (matched males) using the original sample weights and average wage of males in the common support using the adjusted sample weights. ∆f and ∆m are the components of the wage gap that corresponds to differences in the common support of characteristics. ∆m captures the differences between males outside and within the common support, and ∆f captures the differences between females outside and within the common support. [94]    The advantage of using this methodology instead of the traditional Blinder-Oaxaca decomposition is that it is not necessary to assume any sort of functional form for the relationship between characteristics and wage (Mincerian model), and also it constrains the wage gaps comparison to males and females with comparable characteristics. This study presents information for six Central American countries in two periods. In order to have a pooled dataset with representative information of these six countries we have to ensure that wages are measured with the same units and the weights (expansion factors) of each observation in the pooled dataset corresponds to size of each country. Wages were expressed in 2002 dollars in purchasing power parity. The original wages as reported in each individual dataset are in current prices in the national currency. We use the nominal GDP deflators in each country to express wages in constant 2002 prices and the purchasing power parity (PPP) exchange rates in each country to transform the constant wages in national currency into dollars. [95]  Â