WPS4071 BACKGROUND PAPERTOTHE2007 WORLDDEVELOPMENT REPORT How are Youth Faring in the Labor Market? Evidence from Around the World Jean Fares, World Bank Claudio E. Montenegro, World Bank Peter F. Orazem* Iowa State University Abstract. This paper uses a new standardized micro database for a large set of developing countries to (i) describe the patterns of labor market outcomes for youth, and (ii) to explain the contributions of supply and demand factors to youth outcomes. The paper shows that youth face various difficulties in transitioning to work. This is reflected in their relatively higher unemployment rate, higher incidence of low paying or unpaid work, and a large share of youth who are neither working nor in school. This is especially true for young girls who are found outside the labor market, some engaged in home production. Finally, the paper also finds that cross-country estimates show that changes in the youth relative cohort size is unlikely to have a large effect on how youth are faring in the labor market. World Bank Policy Research Working Paper 4071, November 2006 The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the view of the World Bank, its Executive Directors, or the countries they represent. Policy Research Working Papers are available online at http://econ.worldbank.org. *We are very thankful to Mehmet Ziya Gorpe, Natsuko Kiso, Anette Richter, and Victor Sulla, who provided excellent research assistance. 1 I. Introduction Youth transition to work has important implications on youth development and the development of their communities, nations and the world. In order to assess what policies can best assist youth in this transition, we need to know how to measure youth activities in and out of the labor market; what can we learn from existing data on youth time allocations across countries; and what factors affect these time allocations. This paper provides some preliminary answers to these questions. It uses a new standardized micro data base for a large set of developing countries to (i) describe the patterns of labor market outcomes for youth and (ii) explain the contributions of supply and demand factors to youth outcomes. There are several indicators that are relevant to describe youth outcomes. A description of the time spent in the labor market needs to take into account the important share of young people who are still in school, combine school and work, or temporary withdrawn from the labor market. The types of work youth are doing and the difficulties they are facing in their search activities are most important. To get a full picture of how well youth are faring, it is important to examine more than one dimension of the youth labor market. No one measure provides a complete picture of the health of the labor market for youth, and so multiple measures are needed to analyze youth labor markets in developing economies. According to the level of country development and the gender and education of youth the relevant indicator could vary. We will list the various indicators and their possible caveats below, but we briefly outline some of the difficulties to keep in mind in defining youth time allocations in developing countries: · The unemployment rate is typically viewed as a measure of the difficulty finding work, but it also reflects willingness to accept work that is offered. In middle income countries the ratio of youth to adult unemployment rate is high and suggests greater difficulty youth face is finding work. In low income countries, the youth unemployment rate can be very low, except that in urban areas, the more educated and better off portions of the population can have high unemployment rates. It is doubtful that the more educated youth really are facing greater hardships in finding work than are the poorer and less educated youth in developing countries. · The fraction of the youth who are employed does not account for school enrollment, for the quality of the jobs available, or for some types of work that are unpaid and informal. Almost everywhere young female's employment rate is lower than young male's employment rate, due in part to the lack of measures for household work. Unpaid work is a frequent employment option for youth, especially those in poor households. · Because of the lack of measures for household work, an important share of youth, particularly young women, are considered out of school and out of work. Others have called this state idleness. However, idleness is difficult to distinguish from productive but unpaid activities conducted in the home. 2 Empirical evidence of the effect of cohort size and labor demand conditions on youth employment outcomes show that estimates vary with youth characteristics and according to the indicator used. In general, the cohort size seems to have little effect on youth outcomes, while demand conditions have disproportionate effect on youth unemployment, particularly among low skilled and low income households. Adverse demand conditions also reduce employment among youth. However, the resulting slow transition from school can have a benefit by increasing school enrollments . The next two sections describe the outcome indicators for youth and how they are interrelated. Sections 4-8 present descriptive statistics of individual indicators in a large set of countries disaggregated by gender, skills, income and sector. Section 9 presents the cross country estimates of a simple empirical model of the effect of youth cohort size and labor market demand conditions on time allocations. Section 10 concludes. II. Labor Market Indicators What is the time use pattern among young people? Because youth is a period of transition, a description of the time spent in the labor market needs to take into account the important share of young people who are still in school; who are combining school and work; or who are temporarily withdrawn from the labor market. Conclusions regarding the strength of the youth labor market can then differ depending on how youth time allocations are measured. Youth who are not employed in the formal market may be spending time productively in school or in informal production activities, they may be actively seeking work, or they may be total withdrawn from the labor force. Figure 1: An illustration of youth time use In the labor force Not in the labor force Employed Not in the labor force Formal and not in school Informal Enrolled in School Working Not Working Unemployed Not in the labor force and not in school Population Figure 1 is an illustration of different activities youth might be engaged in at any point in time during their transition to work. For those very young or of secondary school age, a large share remains in school. The older of these often combine school and work. Others have exited school and are found employed, either in the formal or informal sector. A significant share of youth, particularly females, does not enter the labor force, some due to their increased home responsibilities (child care, elderly care, 3 domestic work, etc.). Others find difficulties entering the labor force and become discouraged. Several young people in the labor force are also found unemployed, searching for work. The standard labor market indicators include the unemployment rate (UR), the employment rate (ER), and the labor force participation rate (LFPR). Box 1 provides the definitions of these states. To be considered unemployed, an individual must not be employed but actively seeking work. The LFPR is the share of the population that is either employed or unemployed, and the unemployment rate is the share of the labor force that is unemployed. The employment rate is the share of the population that is employed. Because the unemployment rate is measured with respect to the labor force while the employment rate is measured with respect to the population, the two will not be perfectly inversely correlated. In practice, it is can be difficult to distinguish between being unemployed and not being in the labor force because verifying job search for the unemployed is costly. It can even be difficult to distinguish between being employed and being out of the labor force because individuals employed without formal contracts such as in day labor situations or who work for household enterprises may not be counted in official labor market statistics. Box 1 Definitions · Population (P): The civilian, non institutionalized population aged 15-24 that is potentially available for work · Employed (E): During the reference period*, the respondent was either self- employed, working for others for pay, or working for a household enterprise without pay. · Unemployed (U): Not employed but actively seeking work during the reference period. · Labor Force (L = U + E): The sum of employed and unemployed. · Not in the labor force (N = P-L): The rest of the population not in the labor force. · Unemployment Rate: UR = U/L · Employment Rate: ER = E/P · Labor Force Participation Rate: LFP = L/P · Home Status Rate: HR = {N ­ [S&N]}/P · Jobless Rate: JR = HR + [U/P] = [P ­ E ­ S&N]/P · School Enrollment Rate: SR = S/P · School and Work Rate: S&LR = (S&L)/P * Reference period: In labor market surveys, questions regarding respondent time allocation are typically framed with respect to activities in the past week, the past month, or the past year. Two other measures are used in developing country settings, the home status rate (HR) and the jobless rate (JR). The home status rate is the proportion of the population that is neither in the labor force nor in school. It measures the group that has left school 4 but does not appear to be transiting into the world of formal paid work. Our use of the term "home status" reflects that individuals in this group may be engaged in unpaid work in the household or in home-based enterprise. The jobless rate is the proportion of the population that is either unemployed or idle. It avoids arbitrary distinctions between being out of the labor force and being unemployed. III. The Interrelationship of Alternative Measures of Youth Time Use To what extent do these indicators reveal consistent inferences about youth time allocation? Do they reveal unique information or are some essentially measuring the same thing? Are the indicators consistent across countries, or do they mean different things in different settings? The data used in this study come mostly (although not exclusively) from the World Bank's Microdata Development Data Platform (DDP). This platform is a compendium of Household Surveys that comprises most of the countries in the world. From this DDP data base we selected those countries with adequate and reasonable data. Adequacy was defined, in general, as being nationally representative. Reasonable data was defined as having good quality data, appropriately documented, and also that has some minimum set of desired variables. Since these surveys were conducted by the National Statistical Offices of each country, it was necessary to perform some minimum standardization to make the variables reasonable comparable from one country to another, and also from one period to another. A description of the countries, years and name of the surveys can be found in Appendix Table IV. Specific Example Table 1 is an example of these different indicators in a sample of selected countries. Of the nine countries, the strongest labor markets reflected by a low UR (Vietnam and Uganda) also are the strongest according to high ER and low JR. However, Pakistan has the third lowest unemployment rate but the highest jobless rate. Because relatively few Pakistani youth who are not employed are in the labor force, the Pakistan youth UR is artificially low. This reflects a very high home status rate in Pakistan, particularly for women, and so those without jobs are not counted among the unemployed because they are not counted as actively seeking work. Bulgaria also has a relatively high jobless rate, but has a high unemployment rate as well because, in contrast to Pakistan, a high proportion of the jobless are counted as actively seeking employment. In evaluating the pattern of labor market indicators in Table 1, it seems that the unemployment rate provides a very noisy measure of the strength of the youth labor market in developing countries. The employment rate provides a better measure of the economy's ability to generate jobs and the jobless rate is a better measure of how the out- of-school population is faring in the labor market. Among the strongest youth markets according to ER and JR criteria (Vietnam, Poland, Bolivia and Uganda) the UR varies from 0.05 to 0.19. Of the weakest (Bulgaria, Egypt, El Salvador, Pakistan, and Niger) 5 UR varies from 0.10 to 0.39. Consequently, the unemployment rate exaggerates the strength of the youth labor market in El Salvador and Pakistan and understates somewhat the strength of the Bolivia and Poland markets. Table 1: Labor market indicators for youth in selected countries Country UR ER LFPR HR JR SR SWR Vietnam 0.05 0.60 0.64 0.05 0.08 0.39 0.06 Bulgaria 0.39 0.24 0.40 0.19 0.34 0.40 0.02 Poland 0.19 0.45 0.56 0.10 0.21 0.69 0.30 Bolivia 0.14 0.51 0.59 0.09 0.17 0.72 0.21 El Salvador 0.11 0.40 0.46 0.23 0.28 0.38 0.06 Egypt 0.24 0.32 0.42 0.19 0.29 0.41 0.02 Pakistan 0.10 0.40 0.45 0.38 0.43 0.19 0.01 Niger 0.21 0.29 0.37 0.35 0.43 0.32 0.03 Uganda 0.06 0.52 0.55 0.09 0.12 0.37 0.01 One of the confounding factors in measuring the strength of the youth labor market is that many of the youth are in school. The standard measure of schooling intensity is the school enrollment rate (SR), the proportion of the population enrolled in school. However, some of these youth may be also working, and so we propose an alternative measure that examines the share of the youth population that is both enrolled and working (S&LR). At ages 15-24, combining school and work is likely to be particularly important in markets with rapidly expanding opportunities but also a need for human capital in the labor market. In table 1, countries with high enrollment rates also tend to have high rates of combining school and work (Poland and Bolivia). These are countries that appeared weaker according to unemployment rates but stronger according to ER and JR. On the other hand, countries with low ER and high JR tended to have low enrollment rates, suggesting that weak labor market may discourage schooling as well as discouraging labor market participation. The correlation across indicators in a large set of countries. While the specific examples above are useful illustrations of the range of patterns observed for youth time allocations, it is useful to generalize the results across more countries. We can get an idea of the interrelationships and unique features of various labor market indicators from the correlation matrix. If two indicators are measuring the same thing, they should be very highly correlated. As is immediately apparent in Appendix Table II, employment and labor force participation rates, shown in the top panel, are so highly correlated that they can be viewed as measuring the same aspects of the labor market. The correlation is sufficiently high to suggest that variation in youth labor force participation rates across developing countries is due primarily to variation in employment rates. Variation in unemployment 6 rates appears to be more random, as the correlation between labor force participation and unemployment rates is only about one-third the correlation with employment rates. Joblessness and home status also appear to measure similar things. As their definitions include many of the same elements, the high correlation is not surprising. Employment rates are also strongly negatively correlated with unemployment rates. The negative correlation is of similar size with that between the employment and jobless rates. The distinction between the jobless and the unemployment states is that the former includes those not actively seeking work. It seems that weak employment conditions are correlated not just with unemployment but with joblessness more generally, whether or not youth opt to seek work. Schooling is an alternative youth time use. As one would expect if opportunity costs of schooling matter for schooling decisions, the enrollment rate is negatively correlated with the employment rate and positively correlated with the unemployment rate. However, the strongest relationship is the negative correlation between enrollment rates and home status, suggesting that home responsibilities also serve as important sources of opportunity costs for schooling. While some pairs of labor market indicators are highly correlated, half of the correlation coefficients are smaller than .4. For the pairs with the smallest correlation coefficients, that suggests that less than 16% of the variation in one measure can be "explained" by the other1. Differences among young men and women Many of the correlations between labor market indicators are similar between men and women. Most prominent is the high correlation between employment and labor force participation rates for both young men and women. However, there are notable differences in some of the other indicators. · For young women, joblessness and home status are much more highly correlated than for young men, suggesting that variation in female joblessness across countries is more closely tied to variation across countries in female home time. · Labor force participation rates for young women are more highly correlated with home time use and joblessness, while they are more highly correlated with enrollment rates for young men. That suggests that schooling is a more important reason for men not to enter the labor market while home time use is the more important reason form women not to enter the labor market. · High unemployment rates are more closely aligned to high enrollment rates for young women than for young men. On the other hand, high unemployment rates are more closely tied to higher jobless rates for young men. The implication is 1The square of the correlation coefficient can be interpreted as the R2 one would get from regressing one of the pair on the other. A correlation coefficient of 0.4 corresponds to an R2 of 0.16. We use the term "explained" loosely here and do not mean causality but co-movement. 7 that women respond to weaker labor markets by spending more time in school, while young men either continue seeking work or become discouraged and drop out of the labor force. Differences among young urban and rural residents For both urban and rural youth, employment and labor force participation rates are interchangeable labor market indicators. For both, home status and joblessness are also very highly correlated. However, there are notable differences between the geographically differentiated youth. · Unemployment and employment are more closely tied states in rural than in urban markets. Unemployment is also more closely tied to higher enrollment rates and lower labor force participation rates in rural markets. In contrast, urban youth unemployment rates appear relatively uncorrelated with every other measure of labor market success save joblessness for which unemployment is an element. This suggests that variation in youth urban unemployment rates across developing countries is due to idiosyncratic factors that do not affect other measures of urban youth time use. · Unemployment rates are more highly positively correlated with enrollment rates and employment rates are more highly negatively correlated with enrollment rates in rural markets. Urban unemployment rates are virtually uncorrelated with enrollment rates. The correlations suggest that employment is a more important opportunity cost of schooling for rural youth. · Youth enrollment rates are more highly negatively correlated with home status and jobless rates in urban than in rural markets. This surprising finding suggests that it is in urban markets that opportunity costs related to the value of home time or of other non market time use are most important in limiting schooling. The general conclusion is that no single labor market indicator will adequately summarize the state of the youth labor market in a country, but that not all six measures are needed. Because the employment rate and labor force participation rates measure similar things, only one is needed. The employment rate is easier to measure in that it does not require assessing whether the jobless are seeking work. The enrollment rate has significant independent information and is also relatively easily measured. Of the remaining indicators, joblessness is the easiest to measure and it also incorporates much of the information in the home status rate. One might wish to use the unemployment rate rather than the jobless rate because the unemployment rate is most commonly used in international comparisons of labor market strength. Its disadvantage is that it is based on an assessment of active job search among the jobless that requires household surveys and so it can be costly to obtain for budget constrained governments. 8 IV. School Enrollment vs. Labor Force Participation Between the ages of 15 to 24, an important share of youth is still in school. While this share is the highest at age 15, it decreases with age as more youth exit schools. Half of developed country youth aged 15-24 are still in school compared to only 39% of developing country youth. These differences in human capital investments will shape the relative production activities and incomes for developing and developed economies into the future. Figure 2. Median values of the variables by gender, and by sector, and the level of development Age group breakdown 90 60 ) %( geat rceneP 30 0 Unemployment Employment Not in labor force Enrollment Idleness Jobless rates Young developed countries Young developing countries Adult developed countries Adult developing countries Youth: gender breakdown 90 60 ) %( gea entc Per 30 0 Unemployment Employment Not in labor force Enrollment Idleness Jobless rates Male developed countries Male developing countries Female developed countries Female developing countries There are only slight differences in enrollment rates between young men and women in developed countries with women having the advantage. In developing countries, we find a comparably sized gap favoring young males. There is no difference 9 in enrollment rates between urban and rural youth in developed countries. In developing countries, there is a large 15 percentage point gap in enrollment rates favoring urban youth. As young people start leaving school, some join the labor force. An individual is considered part of the labor force if he is either employed or if he is not employed but is actively seeking work. The labor force participation rate is the broadest measure of labor supply--a willingness to provide time to the labor market. Figure 2 shows that the average youth labor force participation rates are very similar between developed and developing countries. However, adult labor force participation rates are higher in developed than developing countries. There is no difference in female youth labor force participation rates between developed and developing countries. Male youth participate in the labor market with greater frequency in developing countries. In developing countries, youth are slightly more engaged in the labor market than are youth in developed countries. The gap between developing and developed country youth LFPR is five percentage points for rural youth and only two percentage points for urban youth.. In both sets of countries, male youth have higher labor force participation rates than female youth. In both sets of countries, youth labor force participation rates are higher in rural than in urban areas. The division between school and labor force participation is not always clearly identified. Many youth combine work and school, some at very early ages. Combining school and work can be detrimental to learning outcomes and could lead to early exit from school. For others, particularly older youth, combining school and work could provide job experience that could facilitate the transition of youth into employment after graduating from school. V. Youth Unemployment The most commonly used measure of labor market difficulty used in developing countries is the unemployment rate. This indicator is typically monitored on a monthly basis in OECD countries, but is only available on a monthly, quarterly or even annual basis for a few developing countries. Our developing country unemployment measures are based on household survey data for various years between 1990 and 2004. In every region the difficulty youth face in entering the labor market is evident in higher unemployment rates for young men and women than for older workers. Youth make up 25 percent of the working-age population worldwide, but 47 percent of the unemployed. The estimated global unemployment rate for youth increased steadily from 11.7 percent in 1993 to 14.4 percent in 2003.2 Across all markets the youth 2ILO (2004). The ILO produces global and regional estimates based on several national surveys and the ILO Global Employment Trends model. 10 unemployment rate is 2-3 times higher than the adult unemployment rate, regardless of the level of aggregate unemployment (figure 3). However, there is considerable cross country variation, not only in the levels of youth unemployment but also in the ratio of the youth to adult unemployment rate. A very high ratio in some countries is a signal of atypical youth difficulties in the transition to work and a source of concern for policymakers. Figure 3: Unemployment is higher for youths than adults 3 : 1 2 : 1 1 : 1 0.4 rate 0.3 0.2 unemployment 0.1 Youth 0 0 0.1 0.2 0.3 0.4 Adult unemployment rate Notes: The 1:1 line represents cases where the estimated incidences among young and adults are identical. The 2:1 (and 3:1) lines represent cases where the estimated incidences among young is twice (three times) as large as adults. The unemployment rates vary widely across regions, gender, urban and rural sectors and with different individual and household characteristics. The availability of household survey data allow for disaggregating across these different dimensions and a deeper assessment of unemployment among youth (see figure 2). Overall unemployment rates for youth and adult age groups are very similar in developed and developing countries. Youth rates are more than double the adult rate regardless of the country's development level, with youth unemployment rates averaging in the mid-teens. Male and female youth unemployment rates are virtually identical in developed and developing countries, averaging about 15 percent. In developed countries, youth unemployment rates are highest in rural areas. In developing countries, the opposite pattern holds with urban unemployment rates more than double the rural rates. According to the ILO, the unemployment rate varies widely across regions, from a low of 7 percent in East Asia to 13.4 percent in industrial economies to a high of 25 11 percent in the Middle East and North Africa.3 In developing countries, survey data indicate that unemployment rate is highest among youth in MENA and LCR. In SSA and SAR the overall unemployment rate for youth remains low particularly due to the large share of youth in rural areas characterized by low unemployment rate. If fact, even in SSA and SAR, the youth unemployment is high in urban areas. Figure 4: Youth unemployment varies with income and educational level Youth average unemployment by quintile 25 rate 20 15 10 Unemployment 5 0 first second third fourth fifth quintile Youth average unemployment by education level etar 30 25 nte 20 m 15 oylp 10 m 5 neU 0 ion lete educat mplete plete mplete ndary inco comp seco no primaryincoprimarycomdary ary secon cond rethan se mo Income quintile As figure 4 shows the unemployment rate among youth in developing countries also varies according to the income level of the households where they reside and according to their educational level. On average, the youth unemployment rate increases 3ILO (2004). 12 with the level of education of youth. Similarly, youth from higher income households are more likely to be found unemployed compared to those from low income households. These patterns are different from what is observed in developed countries. This could be due to the lack of access of low educated and low income youth to income support in developing countries which leaves them with no choice but to find work, as opposed to high income youth who could rely on family support in developing countries for the duration of unemployment. A similar pattern holds when comparing unemployment rates by level of development. Youth unemployment is more widespread in middle income countries than in low income countries, consistent with the hypothesis that the poorest youth cannot afford to be unemployed. The patterns of youth unemployment and employment rates by development level are shown in figure 5. As country incomes increase, the incidence of unemployment among youth also increases, reaching as high as 30 percent in some high income countries. In low income countries, the employment rate for youth is very high, estimated to be higher than 40 percent in low income and low middle income countries. As country incomes rise, youth employment rates come down, also because they devote more time to schooling. Figure 5: Unemployment and Employment across countries with different income level setaR 50% nt 40% meyolp 30% mE dnatn ) 20% (% me 10% oy pl menU 0% Low Income Low Middle Upper Middle High Income h Countries Income Income Countries utoY Countries Countries Unemployment Employment VI. Youth Employment Employment rates have a key measurement advantage over unemployment rates in that the base is the overall population rather than the labor force: the portion of the population that is actively interested in being employed. The labor force reflects both supply and demand factors in that in weak markets, some normally in the labor market will decide to leave the labor force. In stronger markets, some will enter the labor force 13 who normally would not seek employment. This movement into and out of the labor force clouds the interpretation of the unemployment rate because the labor force shifts will understate both bad and good employment conditions. Because the population is fixed (except for in- or out-migration), variation in employment rates are believed to better reflect variation in demand conditions.4 The employment rate for youth is positively correlated with adult employment. As adult employment increase youth employment rate increases (figure 6). However, youth employment is invariably lower than that of older men and women. Some of this difference is due to the large share of youth that remain in school and are not available for work. Adjusting for school enrollment reduces the difference between youths and adults, though the gap persists. These gaps are greatest between younger and older men. The employment rate for young men is always higher than the employment rate for young women, reflecting a stronger attachment to the labor force among males but also reflecting the greater proportion of young women engaged in home production which is excluded from measured employment. Figure 6: Employment differences between youths and adults 100% fm 80% mm m employment f f f mf f ffmmm fm mmm 60% m f fm youth, f f mf mmf m mmmf mm mmmm m m male fm f mm mm f mmm mm mmfm m mm female mf m f 40% f f fff f f f mmfm fmmm f m m mmmf mmmmfmmm mm oyment f fffffff fff mf m mmmm f m f fmf fmmf fm mm ff 20% ff ffffff ff mmmf m empl f f f ffff f ff ff f f fff mmmfmm f f ff Youth 0% 0% 20% 40% 60% 80% 100% Adult employment male, adult employment female Note: The 45° line represents cases where the estimated incidences among young and adults are identical. Observations correspond to country aggregates (f=female, m=male). 4For older groups for whom labor force participation rates are very stable over the business cycle, unemployment rates will better reflect variations in labor demand than will youth unemployment rates. For example, many studies use the unemployment rate for males over 25 as an indicator of the overall strength of the labor market assuming that the adult male labor force participation is nearly constant over time. If labor supply were unstable, the adult male employment rate would be a better measure. 14 Youth employment rates have the same average in developed and developing countries, but adult employment rates average 9 percentage points higher in developed countries. In both developed and developing countries, young males have higher employment rates than young females. However, the male-female gap is much larger in developing countries. Male youth employment rates are higher in developing than in developed countries, while female employment rates are higher in developed than in developing countries. Rural youth have higher employment rates than urban youth in both developed and developing countries. Employment rates for rural youth in developing countries exceed those in developed countries, but urban youth employment rates are higher in developed countries. In developed countries, the most important alternative use of time for those youth not employed is to be enrolled in school. In developing countries, school enrollment is also the most important alternative use of time, but home status is also a prominent use of time for young women. Home status includes informal employment in household production or home enterprise activities that may not be captured in the employment rate. Adding the home status rate to the employment rate eliminates the male-female employment gap in developing countries. Young people's access to productive jobs, which allow them to build skills and experience, also varies enormously between those from rich and poor households (figure 7). Youngsters from poor households are far more likely to be engaged in unpaid family labor than those from rich households, in some cases by a substantial margin. The contrast is particularly striking in Latin America, as in Bolivia, Chile, and Paraguay Figure 7: Incidence of unpaid family work among youth 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% ia ia a ad rdan or u a ela bia ile na ti Jo aziland rag dor uayPer R siaRica icoCh x Me Ta za Za NigergladeshHa ia ia iti nzaEthiomb nia piaiquemb istanbodiaolivonesaragu m B Ind PanamTongazuolom TukeyailandBrazilussta Nic Ecu SalvaPa VeneC Th Argen Mo Ban PakCa SwEl Co Bottom quintile Top quintile Average 15 VII. Home Status Rate Those who are neither employed, nor in school, nor seeking work, are placed in our home category. Some will be working in informal home-based jobs or in unremunerated home production activities and others may be idle. In many countries the proportion of youth who are neither in the labor force nor in school is too large to dismiss as a problem of measurement or a temporary phenomenon. This state barely registers at 3% of developed country youth. It represents 11% of developed country adults, composed largely of housewives and retired. In developing countries, 16% of youth fall into this category. Figure 8 presents a cross-country comparison of young men and women who are neither in school or working compared to the comparable statistic for older men and women in the same country. The 45 degree line would indicate equality in inactivity rates for young and old populations in the country. Points above the 45 degree suggest larger inactivity rates for youth while points below the line indicate higher inactivity rates for adults. · Differences in levels. Female observations [f] almost always lie to the right of the male observations [m], implying a higher incidence of inactivity for females relative to males. Differences in ratios. Most observations for men lie above the 45 degree line, implying that young men are more likely to be seen as inactive than adult men. The opposite pattern holds for women, implying that is the older women who are most likely to be out of school and out of the labor force. Figure 8: The incidence of home status 80% f (%) f 60% f Youth f ff f f f for f f f f f f tes 40% f m m ff f Ra f ff m mm f m m f f f f m f ff ff f f m f m f ff ff f ff Status f f mmfmm mfmm 20% m f f f ff f mmmmff m f f mmm fm m f mffmf f mf f mfmmmmm f mf mm m f ff f mfmm mmmmm mmmm mmmm mmmf Home mfmmmmf f mfm m 0% m 0% 20% 40% 60% 80% Home Status Rate for Adults (%) Note: m = male, f = female Note: The 45° line represents cases where the estimated incidences among young and adults are identical. 16 Some of the high estimates may be attributable to measurement problems, particularly for young women working in their households. In Tanzania, the main reason young women said they were not looking for work was their household responsibilities. For young men, the main reason for inactivity was the lack of market work. The data reveal important differences between males and females in this group-- males are predominately discouraged workers, while females are engaged in non-market activities. The high home status rate in developing countries is not cause for concern to the extent that the youth are engaged in productive non-market activities. This is presumably the case for the majority of the 23% of young women who fall into the category. However, many of the 8% of young men who fall into this category are believed to be truly idle--neither engaged in productive activities nor engaged in human capital production. Truly idle human resources represent lost productive potential for the country as well as a potential supply of discontented youth who may be persuaded to find antisocial uses for their time. VIII. Jobless Rates Youth are considered jobless if they are unemployed or if they are not in the labor force and not in school. This measure has the advantage over the unemployment rate in that it includes discouraged workers--those who have concluded they have no job prospects and have ceased active job search. Jobless rates in developing countries are nearly twice those found in developed countries. Youth jobless rates are similar to youth unemployment rates in developed countries. However, youth jobless rates are nearly twice the unemployment rates in developing countries, suggesting that discouraged workers are a more serious issue for these countries. The gap between jobless and unemployment rates are substantially higher for adults than for youth in both sets of countries. Young women have higher jobless rates than do young men, even though they have virtually identical unemployment rates. The gap is small (3 percentage points) in developed countries, but it is much larger in developing countries. Jobless rates are nearly identical in rural and urban markets in both developing and developed countries. Virtually all developed country jobless youth are considered unemployed (meaning they are still seeking work actively). This is true of urban jobless youth in developing countries as well. However, rural jobless youth in developing countries are generally not seeking work, either because they are engaged in non market work activities or because they are idle. 17 IX. The Effect of Cohort Size and Labor Market Conditions on Youth Labor Market Success We are interested in establishing how youth time allocations vary with respect to country business cycles and the relative size of the youth cohort. Following O'Higgins (2003) and Korenman and Neumark (2000), we propose a specification: (1) ln(LYit) =0 +1 ln(Pit ) +2 lnUit +it Pit Y A A where ln(LYit)is a measure of youth labor market success in country i and year t; ln(Uit ) is A Pit Y the logarithm of the adult (A) unemployment rate in that country; and ln(Pit ) is the A logarithm of the relative size of the youth population to the adult population; and it is an error term. The youth group is taken from ages 15 to 24 and the adult population includes all aged 25-49. The coefficient 1 measures the impact of an increase in the relative size of the youth cohort on youth labor market success. The typical belief is that atypically large youth cohorts create crowding in the labor market, and so youth will experience greater difficulty in the labor market. If this belief is correct, then 1 < 0. However, if labor markets are sufficiently flexible that they can adapt to rising cohort sizes, the impact on youth unemployment would be negligible and so 1 = 0. Some have even argued that large labor market cohorts, particularly educated ones, might induce economic innovation and expansion, in which case 1 > 0.5 Because adults tend to have the most stable employment of all demographic groups, the adult unemployment rate is frequently used as indicator of business cycles. Estimates of 2 that exceed one suggest that a percentage point increase in adult unemployment results in an increase in youth unemployment of more than one percent. When 2 =1, the youth unemployment rate responds proportionally to increase in the adult unemployment rate. When 2 <1, the youth unemployment rate is less sensitive to fluctuations in the adult unemployment rate, and if 2 <0, the youth unemployment rate reacts counter-cyclically to fluctuations in the adult unemployment rate. If the error term in (1) is iid N(0,), all the parameters can be estimated using ordinary least squares. However, there may be country-specific, time invariant effects that are ignored in (1) and that can bias the estimates. Suppose that it =it +i where i is a country-specific fixed effect and itis a purely random effect uncorrelated with the other regressors. If i is correlated with the country's adult unemployment rate or the 5Note that if we use a negative measure of youth labor market outcomes (eg unemployment) rather than a positive measure (eg employment) as the dependent variable in (1), the expected signs will reverse so that 1> 0 if crowding harms youth and 2> 0 if adverse business cycle shocks harm youth. 18 demographic structure, OLS estimates of (1) will be biased. Differencing the data from a second period t' will yield unbiased estimates of 1 and 2 . (2) ln(LYit' ) =0 +1 ln(Pit ) - ln(Pit' ) +2 ln Uit' +it LYit ' Pit Y Pit' Y UitA A A A Specification (2) differences away the fixed effect i . We only have two periods of observations for at most 45 countries compared to as many as 93 countries for which we have a single period. Consequently, the cost of using the differenced specification is the loss of almost 50% of the degrees of freedom. We try multiple positive and negative measures of youth labor market success in assessing how cohort size and business cycle shocks affect the youth labor market. Positive measures include employment rates and labor force participation rates. Negative measures include unemployment and jobless rates. Mixed measures include time use indicators that may remove youth from the labor market including enrollment rates and home status rates. Appendix Table III shows the estimated results for each indicator, and disaggregated by different groups. For example the top panel of the table shows the cross country estimates (the coefficients and t statistics of the relative cohort size, log adult unemployment rate, and the constant term) for youth unemployment rate for the total sample, and by gender, sector, youth educational attainment, household income levels. Three sets of regressions are estimated: the first set showing regressions in levels for all countries; the second set in first differences for the sub-sample of 45 countries where two years of observations were available; and a third set in levels for the same subset of 45 countries. We will discuss each of these indicators in turn. Youth Unemployment Rates It is commonly assumed that youth employment prospects will be significantly disadvantaged by unusually large cohorts of new labor market entrants. There is little evidence supporting that proposition in the regressions over a sample of developing countries reported in column 1 of Table III (labeled rpst). In only one-third of the cases is the estimate of 1 positive and it is never statistically significant at the 5% level. In the remaining cases, the estimates are negative, although also rarely significant. In the second set of estimates where the differenced data is used to remove the possible bias from fixed effects (column headed cpst), we also have no statistically significant evidence of a positive effect of changes in the relative size of the youth cohort on youth unemployment. There is simply no evidence that large youth cohorts cause greater unemployment problems for the young. Youth unemployment in developing countries is certainly pro cyclical. Looking at the estimates for the column headed luAt, it appears that the youth unemployment rate overall and the female and rural youth unemployment rates are roughly proportional to the adult unemployment rate. The male and urban youth unemployment rates react somewhat less than proportionately to the adult unemployment rate. In changes (second 19 set of estimates headed cuAt), the estimates of 2 are less than 1 for all of these groups. Although the estimates are not very precise, they are generally more than two standard deviations away from one. The more interesting results occur when we divide the youth unemployment into education or household income groups. Youth in the lowest education group and in the lowest income quintile respond more than proportionally to the adult unemployment rate. As the youth education level rises, the estimate of 2 gets progressively smaller, implying that more educated youth are less harmed by adverse business cycles. Although the pattern is not perfectly consistent, the tendency is that as household income level rises, the estimate of 2 also decreases, and so it is the poorest youth who are most adversely affected by adverse business cycles. In the differenced specification, the pattern remains for education groups, with the two least educated groups facing more than proportional increases in unemployment. The pattern with respect to income quintiles does not hold up, except that the unemployment of youth in the highest income quintile is the least sensitive to business cycles. Youth Employment Rates In only four cases does an atypically large youth cohort affect youth employment rates, and in three of the cases, the impact is positive. In changes all of these effects disappear. We conclude that there is no evidence that atypically large youth cohorts cause youth employment rates to fall. Negative estimates of 2 are consistent with a negative impact of increased adult unemployment rates on youth employment rates. In the level regressions, every estimate is negative and most are statistically significant. Magnitudes do not vary greatly across genders or across urban and rural markets. There is modest evidence that the adverse impact is greatest for the least educated. In the differenced form equations reported in the second set of estimates, the employment responses for male and female youth and for urban and rural youth are all insignificantly different from zero. Similarly, the apparent negative effect of adult unemployment on youth employment rates also disappears. We conclude that youth employment rates are not sensitive to fluctuations in adult unemployment. Youth Jobless Rates There is more evidence that the crowding from atypically large youth cohorts increase the jobless rate. In the level regressions of Table III, 11 of 16 estimates of 1 are positive and five are statistically significant. However, the effect is fragile with respect to the presence or absence of country fixed effects. In differenced form, 15 of 16 estimates of 1 are negative, suggesting no systematic penalty in the form of youth joblessness resulting from large youth cohorts. Youth joblessness rises with increases in the adult unemployment rate. The estimates of 2 are all positive, ranging narrowly between 0.08 and 0.35. The similarity 20 in estimated effects in the levels specifications suggests uniform joblessness responses across demographic groups. However, in differenced form, the adverse effects are less consistent. The differenced estimates suggest that youth joblessness is insensitive to changes in the adult unemployment rate overall, and for subgroups including only males, only females, only urban youth, or only rural youth. There is more evidence of rising joblessness in response to rising adult unemployment for the less educated groups with the exception of the completely uneducated. Youth in different income quintiles face similar increases in joblessness in response to rising adult unemployment, and so the burden of joblessness is not concentrated among the poorest youth. Labor Force Participation Rates Youth are considered to be in the labor force if they are employed or if they are unemployed (not employed but actively seeking work). This is considered the broadest measure of youth notional labor supply. The level regressions show a mixed response of youth labor supply to cohort size, with some evidence that the least educated increase labor force participation and the most educated decrease labor force participation in response to crowding from an unusually large cohort. However, the different regressions show that increases in relative cohort size raise labor force participation for all demographic groups, although the estimates of 1 are only occasionally significant. The estimates suggest that a 10% growth in the relative size of the youth cohort results in a 3.5% increase in the youth labor force participation rate. The labor supply response is bigger for young males than for young females. The increase is identical for rural and urban youth. The biggest increases are for youth from wealthier families. The pattern by education groups is U-shaped with the biggest increases for the most and least educated, although the coefficient estimates for the least educated are not statistically significant. We conclude that developing country youth in unusually large cohorts respond by increasing labor supply behavior compared to youth in smaller cohorts. The estimates of 2 are more consistent across the levels and differenced regressions. They show that in general, increases in the adult unemployment rate cause a modest reduction in the growth of youth labor force participation. The magnitudes are small--a 10% increase in the growth of adult unemployment rate causes a 1% reduction in the growth of youth labor supply. The effect is uniform across genders and across urban and rural labor markets. There is no apparent pattern of larger or smaller labor supply responses by education level or by household income level. Youth Home Status There is a concern that crowding due to unusually large cohorts or discouragement due to weak labor demand can cause youth to drop out of school and to drop out or never enter the labor market. This state is referred to as idleness. However, idleness is difficult to distinguish from productive but unpaid activities conducted in the home. Nevertheless, variation in youth home status related to cohort size or adult unemployment are likely to be indicators of increased idleness because non-market 21 productive household activities would be expected to be related to longer term social or life cycle plans rather than short term economic circumstance. Reflecting the pattern found for labor supply behavior, the level equations show significant evidence of rising home status in response to large cohorts, but the cross- sectional pattern appears to be driven by country fixed effects. In differences, countries experiencing growth in the relative size of their youth cohorts show decreasing home status rates across every demographic group. The effects are quite large, with the biggest reductions for male and urban youth. The most educated youth and youth from wealthier families demonstrate the greatest declines in home status in response to growth in the size of the youth cohort. Reductions in home status imply either rising labor supply, rising enrollment, or both. Interestingly, home status is virtually insensitive to fluctuations in the adult unemployment rate. Whether home status reflects productive or unproductive uses of youth time, variation in home status appears uncorrelated to short term fluctuations in the strength of labor demand in the economy. Variation in home status across countries does not appear to be a result of short-term variation in the strength of the labor market across countries. Youth Enrollment Rates Estimates of 1are very consistent across the level and differenced enrollment rate equations. Of concern is that youth in unusually large cohorts reduce their time in school. The effect is large--a 10% increase in the relative size of the youth cohort causes a 7% reduction in the youth enrollment rate. The effect is even larger for female and rural youth. There is some evidence in the levels equation that the largest negative effects are for youth from the poorest income quintiles, but none of the effects are statistically significant in the differenced equations. We conclude that youth respond to large cohort size by reducing their investment in schooling and that the effect appears to occur across all demographic groups. The levels regression suggests that countries with large adult unemployment rates induce higher youth school enrollment rates. The effect is found for all demographic groups, but is generally of modest size. However, the effect disappears in the differenced data, suggesting that youth enrollment rates do not respond to increases in the adult unemployment rate. 22 X. Concluding Remarks In this paper we take advantage of a new standardized micro data base to describe the time use pattern of youth in the labor market in a large set of countries. The transition of youth to work is complex and requires the use of multiple indicators to better understand how well youth are fairing and what influences most their transition to work. In almost all countries, youth face some difficulties in transitioning to work. This is reflected in their relatively higher unemployment rates, higher incidence of low paying or unpaid work, and large shares who are neither working nor in school as compared to adult populations. For youth in low income countries unemployment is more concentrated in urban areas and among those who are more educated and come from higher income households. In lot of countries a large share of young girls are found outside the labor market, some engaged in home production. Among the employed, youth are disproportionately found in unpaid work particularly in poor households. Cross country estimates show that changes in the youth relative cohort size is unlikely to have a large effect on how youth are faring in the labor market. The richness of the data allows for different desegregations, yet estimates are not measured very precisely. Controlling for country specific effect further reduce the estimates precisions. In general, changes in the demand conditions, as measured by the adult unemployment rate, have a strong effect on youth unemployment and employment rates. The most affected are the low skilled and low income youth. More empirical research is needed to shed further lights on the determinants of youth transition and to help guide policies to facilitate youth transition to work. 23 Appendix Table I Unemployment Employment Not in labor force Enrollment Idleness Jobless developed co 16.0 36.0 56.0 50.0 3.0 14.0 Young developing co 14.0 36.0 54.0 39.0 16.0 27.0 developed co 7.0 80.0 13.0 2.0 11.0 17.0 Adult developing co 6.0 71.0 23.0 2.0 18.0 25.0 Unemployment Employment Not in labor force Enrollment Idleness Jobless developed co 16.0 38.0 52.0 48.0 2.0 13.0 Male developing co 15.0 43.0 47.0 41.0 8.0 21.0 developed co 15.0 34.0 62.0 51.0 5.0 16.0 Female developing co 15.0 29.0 63.0 37.0 23.0 34.0 Unemployment Employment Not in labor force Enrollment Idleness Jobless developed co 14.0 35.0 57.0 51.0 4.0 15.0 Urban developing co 21.0 30.0 59.0 45.0 17.0 29.0 developed co 18.0 37.0 54.0 51.0 5.0 15.0 Rural developing co 10.0 44.0 49.0 30.0 17.0 26.0 UNEMPLOYMENT Young Adult Female Male Urban Rural EAP 11.0 3.0 10.0 13.0 16.0 11.0 ECA 21.0 11.0 21.0 22.0 22.0 23.0 HIC 14.0 7.0 14.0 13.0 14.0 17.0 LAC 18.0 7.0 22.0 15.0 19.0 11.0 MENA 24.0 5.0 26.0 22.0 35.0 20.0 OHIC 25.0 7.0 26.0 25.0 20.0 21.0 SA 7.0 2.0 12.0 6.0 13.0 5.0 SSA 8.0 3.0 6.0 10.0 23.0 4.0 EMPLOYMENT Young Adult Female Male Urban Rural EAP 39.0 73.0 30.0 48.0 37.0 47.0 ECA 30.0 75.0 28.0 33.0 28.0 32.0 HIC 37.0 80.0 36.0 39.0 37.0 37.0 LAC 41.0 71.0 28.0 56.0 39.0 47.0 MENA 22.0 48.0 14.0 32.0 20.0 26.0 OHIC 27.0 81.0 26.0 29.0 26.0 31.0 SA 40.0 63.0 21.0 57.0 29.0 44.0 SSA 40.0 75.0 42.0 43.0 30.0 54.0 NOT IN LABOR FORCE Young Adult Female Male Urban Rural EAP 50.0 24.0 61.0 43.0 53.0 47.0 ECA 61.0 17.0 65.0 58.0 64.0 57.0 HIC 54.0 13.0 57.0 51.0 54.0 53.0 LAC 49.0 24.0 65.0 36.0 53.0 46.0 MENA 66.0 42.0 85.0 55.0 67.0 68.0 OHIC 63.0 13.0 65.0 62.0 67.0 61.0 SA 55.0 32.0 75.0 40.0 63.0 51.0 SSA 52.0 20.0 55.0 51.0 58.0 43.0 ENROLLMENT Young Adult Female Male Urban Rural EAP 38.0 1.0 37.0 37.0 44.0 32.0 ECA 44.0 1.0 45.0 42.0 53.0 39.0 HIC 46.0 2.0 50.0 43.0 50.0 49.0 LAC 44.0 4.0 44.0 44.0 50.0 31.0 MENA 45.0 1.0 44.0 46.0 49.0 42.0 OHIC 61.0 1.0 62.0 61.0 65.0 56.0 SA 25.0 1.0 19.0 30.0 39.0 23.0 SSA 32.0 2.0 29.0 35.0 41.0 28.0 IDLENESS Young Adult Female Male Urban Rural EAP 21.0 23.0 27.0 9.0 13.0 18.0 ECA 7.0 12.0 10.0 7.0 8.0 12.0 HIC 4.0 11.0 6.0 2.0 4.0 4.0 LAC 16.0 22.0 27.0 7.0 13.0 22.0 MENA 9.0 13.0 26.0 3.0 19.0 19.0 OHIC 2.0 12.0 3.0 1.0 2.0 5.0 SA 26.0 16.0 39.0 10.0 22.0 23.0 SSA 19.0 18.0 24.0 14.0 21.0 17.0 JOBLESS Young Adult Female Male Urban Rural EAP 28.0 26.0 32.0 22.0 22.0 23.0 ECA 24.0 21.0 26.0 22.0 22.0 24.0 HIC 15.0 16.0 17.0 14.0 15.0 16.0 LAC 26.0 28.0 35.0 16.0 23.0 31.0 MENA 29.0 29.0 36.0 22.0 32.0 28.0 OHIC 11.0 18.0 12.0 10.0 9.0 13.0 SA 41.0 16.0 57.0 18.0 36.0 38.0 SSA 28.0 20.0 34.0 20.0 32.0 21.0,. 24 Appendix Table II Young people: total not in labor unemployment employment enrollment idleness jobless force unemployment 1 112 employment -0.6674* 1 112 112 not in labor force 0.3601* -0.9293* 1 112 112 112 enrollment 0.2705* -0.3277* 0.2772* 1 98 98 98 99 idleness -0.146 -0.2487* 0.3747* -0.5670* 1 112 112 112 99 118 jobless 0.4620* -0.5721* 0.4839* -0.4493* 0.7206* 1 112 112 112 98 112 112 Young people: males not in labor unemployment employment enrollment idleness jobless force unemployment 1 112 employment -0.6616* 1 112 112 not in labor force 0.3413* -0.9251* 1 112 112 112 enrollment 0.1806 -0.3465* 0.3444* 1 98 98 98 99 idleness -0.0957 -0.2707* 0.3835* -0.4043* 1 112 112 112 99 117 jobless 0.6065* -0.5949* 0.4429* -0.3239* 0.6332* 1 112 112 112 98 112 112 Young people: females not in labor unemployment employment enrollment idleness jobless force unemployment 1 112 employment -0.6215* 1 112 112 not in labor force 0.3546* -0.9459* 1 112 112 112 enrollment 0.3259* -0.2939* 0.2118* 1 98 98 98 99 idleness -0.112 -0.3657* 0.4875* -0.5910* 1 112 112 112 99 117 jobless 0.3519* -0.6043* 0.5800* -0.4900* 0.8181* 1 112 112 112 98 112 112 Young people: urban not in labor unemployment employment enrollment idleness jobless force unemployment 1 94 employment -0.5784* 1 94 94 not in labor force 0.2015 -0.9085* 1 94 94 94 enrollment 0.0826 -0.2007 0.1927 1 81 81 81 82 idleness -0.1159 -0.2726* 0.3955* -0.5940* 1 94 94 94 82 99 jobless 0.4903* -0.5084* 0.3615* -0.5776* 0.6712* 1 94 94 94 81 94 94 Young people: rural not in labor unemployment employment enrollment idleness jobless force unemployment 1 93 employment -0.7000* 1 93 93 not in labor force 0.4508* -0.9463* 1 93 93 93 enrollment 0.3152* -0.4057* 0.3859* 1 80 80 80 81 idleness -0.1236 -0.2604* 0.3679* -0.4509* 1 93 93 93 81 98 jobless 0.4704* -0.6118* 0.5397* -0.3130* 0.7368* 1 93 93 93 80 93 93 Note: * correlation significant at the 5% level. 25 Appendix Table III Dependent TOTAL SAMPLE SUBSAMPLE OF 45 COUNTRIES variable Sub-group REGRESSIONS IN LEVELS REGRESSIONS IN CHANGES REGRESSIONS IN LEVELS rpst luA_t Constant Obs R2 cpst cuAt Constant Obs R2 rpst luA_t Constant Obs R2 Unemployment total sample luY_t -0.16 -0.88 0.86 19.70 0.40 2.26 93 0.82 cluY_t 0.32 0.83 0.20 1.67 0.14 1.26 45 0.08 luY_t -0.49 -1.84 0.76 11.60 0.00 -0.01 45 0.79 male luY_m -0.19 -1.04 0.80 18.30 0.18 1.04 93 0.80 cluY_m 0.30 0.70 0.20 1.49 0.16 1.26 45 0.06 luY_m -0.59 -1.95 0.70 9.41 -0.29 -0.96 45 0.72 female luY_f -0.11 -0.45 0.99 16.70 0.77 3.25 93 0.76 cluY_f 0.37 0.77 0.24 1.62 0.15 1.06 45 0.07 luY_f -0.38 -1.23 0.82 10.60 0.29 0.92 45 0.75 rural luY_r -0.64 -1.60 0.98 10.50 0.09 0.24 78 0.62 cluY_r 0.47 0.80 0.31 1.74 0.06 0.32 38 0.10 luY_r -1.10 -2.07 0.92 6.88 -0.30 -0.55 41 0.61 urban luY_u 0.41 1.83 0.51 10.00 0.07 0.31 80 0.57 cluY_u 0.12 0.26 0.12 0.83 0.13 0.92 39 0.02 luY_u 0.06 0.19 0.44 5.45 -0.33 -0.99 42 0.44 no education luY_e1 -0.75 -1.35 1.17 9.98 0.61 1.28 70 0.62 cluY_e1 0.51 0.64 1.21 5.67 0.17 1.34 27 0.59 luY_e1 -1.77 -2.70 0.92 7.05 -0.64 -1.15 35 0.67 primary incomplete luY_e2 -0.82 -2.75 0.90 13.90 0.09 0.36 68 0.77 cluY_e2 -2.04 -2.76 1.16 6.66 -0.06 -0.53 32 0.68 luY_e2 -1.29 -2.96 0.79 8.02 -0.58 -1.50 33 0.72 primary complete luY_e3 -0.15 -0.55 0.89 13.10 0.49 1.86 81 0.70 cluY_e3 0.25 0.81 0.93 5.33 -0.05 -0.51 38 0.46 luY_e3 -0.39 -1.02 0.86 9.46 0.25 0.69 42 0.71 secondary incomplete luY_e4 0.03 0.12 0.67 12.80 0.19 0.92 78 0.69 cluY_e4 0.31 1.27 0.94 6.67 -0.04 -0.52 37 0.58 luY_e4 0.14 0.39 0.64 7.14 0.12 0.35 39 0.59 secondary complete luY_e5 0.46 1.72 0.51 8.09 0.14 0.55 85 0.45 cluY_e5 0.34 0.85 0.35 2.80 -0.05 -0.40 41 0.19 luY_e5 0.24 0.62 0.44 4.62 -0.23 -0.59 44 0.34 more than secondary luY_e6 0.49 1.71 0.29 4.19 -0.53 -1.99 82 0.20 cluY_e6 0.45 1.29 0.83 4.18 -0.04 -0.31 40 0.35 luY_e6 -0.04 -0.09 0.20 2.08 -1.05 -2.70 44 0.10 first quintile luY_q1 -1.80 -3.44 1.15 11.20 0.29 0.66 55 0.73 cluY_q1 0.60 1.25 1.13 3.94 -0.07 -0.39 28 0.41 luY_q1 -1.77 -3.81 0.99 8.58 -0.13 -0.28 32 0.78 second quintile luY_q2 -0.77 -1.93 0.98 11.00 0.33 0.88 53 0.73 cluY_q2 0.54 1.14 0.97 3.42 -0.01 -0.08 28 0.35 luY_q2 -0.99 -2.31 0.85 8.04 -0.11 -0.25 32 0.73 third quintile luY_q3 -0.41 -1.01 1.00 11.10 0.51 1.34 52 0.73 cluY_q3 0.27 0.59 1.04 3.81 -0.18 -1.08 26 0.40 luY_q3 -0.82 -1.75 0.82 6.98 -0.19 -0.40 31 0.68 fourth quintile luY_q4 0.19 0.45 0.81 9.03 0.28 0.74 53 0.62 cluY_q4 0.16 0.27 1.17 3.38 -0.02 -0.11 26 0.34 luY_q4 -0.46 -1.07 0.64 5.90 -0.50 -1.12 31 0.59 fifth quintile luY_q5 0.59 1.55 0.57 6.81 -0.19 -0.55 53 0.48 cluY_q5 0.61 1.00 0.74 2.04 0.09 0.40 26 0.20 luY_q5 -0.04 -0.10 0.34 3.01 -1.22 -2.56 31 0.25 Employment total sample leY_t 0.05 0.33 -0.22 -6.08 -1.66 -11.60 93 0.30 cleY_t 0.14 0.76 -0.05 -0.83 -0.01 -0.11 46 0.03 leY_t 0.20 0.92 -0.27 -5.10 -1.65 -7.65 45 0.42 male leY_m 0.13 0.88 -0.18 -5.21 -1.38 -9.81 93 0.25 cleY_m 0.20 1.23 -0.04 -0.81 -0.01 -0.13 46 0.05 leY_m 0.25 1.12 -0.27 -4.88 -1.45 -6.49 45 0.41 female leY_f -0.10 -0.46 -0.25 -5.04 -2.03 -10.40 93 0.22 cleY_f 0.09 0.32 -0.07 -0.82 -0.01 -0.09 46 0.02 leY_f 0.16 0.49 -0.23 -2.86 -1.80 -5.50 45 0.19 rural leY_r 0.08 0.41 -0.24 -5.54 -1.61 -9.14 79 0.30 cleY_r 0.07 0.32 -0.08 -1.29 0.04 0.68 39 0.05 leY_r 0.36 1.32 -0.30 -4.30 -1.58 -5.62 41 0.38 urban leY_u -0.29 -1.47 -0.09 -2.00 -1.65 -9.17 80 0.07 cleY_u 0.44 1.95 -0.02 -0.29 0.00 -0.01 40 0.09 leY_u -0.07 -0.29 -0.19 -3.26 -1.71 -7.30 42 0.22 no education leY_e1 1.33 4.60 -0.22 -3.51 -1.07 -4.43 73 0.33 cleY_e1 0.36 0.44 0.00 -0.02 -0.07 -0.58 35 0.01 leY_e1 0.80 1.96 -0.33 -3.50 -1.53 -4.10 37 0.34 primary incomplete leY_e2 0.39 1.09 -0.31 -3.74 -1.78 -5.42 75 0.19 cleY_e2 0.13 0.50 -0.18 -2.41 -0.05 -1.16 36 0.17 leY_e2 -0.01 -0.03 -0.30 -3.83 -1.71 -5.45 38 0.30 primary complete leY_e3 0.89 2.36 -0.23 -2.61 -1.42 -4.13 82 0.15 cleY_e3 -0.08 -0.32 -0.27 -1.88 0.11 1.24 39 0.09 leY_e3 0.00 0.00 -0.35 -3.22 -1.96 -4.48 41 0.22 secondary incomplete leY_e4 0.42 1.36 -0.13 -1.73 -1.77 -6.09 78 0.07 cleY_e4 0.24 1.05 -0.21 -1.61 -0.07 -0.88 37 0.09 leY_e4 -0.08 -0.19 -0.22 -2.14 -2.03 -4.98 39 0.11 secondary complete leY_e5 -0.40 -1.57 -0.10 -1.70 -1.64 -6.82 87 0.05 cleY_e5 0.37 1.35 -0.20 -2.28 0.06 0.71 43 0.15 leY_e5 -0.43 -1.40 -0.13 -1.75 -1.57 -5.10 44 0.09 more than secondary leY_e6 -0.72 -2.18 -0.02 -0.25 -1.60 -5.14 85 0.06 cleY_e6 0.45 1.56 -0.01 -0.11 0.02 0.28 42 0.06 leY_e6 -0.31 -0.72 0.00 0.02 -1.22 -2.81 44 0.01 first quintile leY_q1 0.71 2.20 -0.25 -3.84 -1.42 -5.20 55 0.28 cleY_q1 -0.30 -0.83 -0.01 -0.03 -0.02 -0.17 29 0.03 leY_q1 0.58 1.88 -0.30 -3.88 -1.46 -4.59 32 0.43 second quintile leY_q2 0.15 0.59 -0.21 -4.21 -1.50 -7.06 54 0.27 cleY_q2 -0.21 -0.70 0.01 0.08 -0.05 -0.45 29 0.02 leY_q2 0.44 1.77 -0.22 -3.60 -1.29 -5.03 32 0.40 third quintile leY_q3 -0.13 -0.50 -0.24 -4.09 -1.71 -6.99 52 0.25 cleY_q3 0.13 0.60 -0.18 -1.41 0.09 1.17 26 0.09 leY_q3 0.36 1.53 -0.22 -3.65 -1.28 -5.24 31 0.40 fourth quintile leY_q4 -0.16 -0.62 -0.22 -3.80 -1.70 -7.19 53 0.22 cleY_q4 0.32 1.39 -0.13 -0.95 0.03 0.38 26 0.10 leY_q4 0.30 1.17 -0.17 -2.74 -1.22 -4.62 31 0.27 fifth quintile leY_q5 -0.31 -1.12 -0.16 -2.60 -1.65 -6.68 53 0.13 cleY_q5 0.46 1.75 -0.08 -0.50 0.02 0.21 26 0.12 leY_q5 0.28 0.93 -0.11 -1.45 -1.09 -3.55 31 0.12 Employment corrected total sample lecY_t -0.25 -2.04 -0.14 -4.86 -1.09 -9.60 93 0.22 clecY_t 0.20 0.87 0.05 0.73 -0.01 -0.16 46 0.03 lecY_t -0.23 -1.25 -0.17 -3.78 -1.11 -6.04 45 0.26 male lecY_m -0.08 -0.77 -0.13 -5.19 -0.83 -8.60 93 0.23 clecY_m 0.30 1.46 0.04 0.67 0.01 0.20 46 0.06 lecY_m -0.11 -0.79 -0.17 -4.91 -0.89 -6.32 45 0.37 female lecY_f -0.46 -2.42 -0.14 -3.21 -1.41 -7.83 93 0.14 clecY_f 0.12 0.36 0.04 0.41 -0.04 -0.49 46 0.01 lecY_f -0.33 -0.97 -0.12 -1.48 -1.25 -3.69 45 0.06 rural lecY_r -0.14 -1.02 -0.15 -4.76 -1.04 -7.96 79 0.23 clecY_r 0.08 0.34 0.00 0.03 0.04 0.51 39 0.00 lecY_r -0.10 -0.50 -0.18 -3.66 -1.08 -5.26 41 0.26 urban lecY_u -0.47 -3.04 -0.06 -1.68 -1.07 -7.45 80 0.12 clecY_u 0.58 2.10 0.07 0.88 -0.01 -0.11 40 0.13 lecY_u -0.41 -1.96 -0.13 -2.41 -1.14 -5.34 42 0.17 no education lecY_e1 0.97 3.02 -0.17 -2.47 -0.97 -3.63 73 0.18 clecY_e1 -0.19 -0.23 0.14 0.71 -0.29 -2.30 35 0.02 lecY_e1 0.73 1.63 -0.31 -3.05 -1.45 -3.55 37 0.28 primary incomplete lecY_e2 0.20 1.07 -0.23 -5.19 -1.22 -7.07 75 0.30 clecY_e2 0.08 0.31 -0.15 -1.95 -0.07 -1.64 36 0.12 lecY_e2 0.21 0.88 -0.25 -4.43 -1.19 -5.48 38 0.38 primary complete lecY_e3 -0.07 -0.46 -0.15 -4.48 -1.09 -8.05 82 0.20 clecY_e3 0.12 0.47 -0.17 -1.18 0.08 0.95 39 0.04 lecY_e3 -0.12 -0.52 -0.21 -3.95 -1.21 -5.74 41 0.29 secondary incomplete lecY_e4 -0.05 -0.31 -0.13 -3.44 -1.05 -6.95 78 0.14 clecY_e4 0.38 2.82 -0.13 -1.68 -0.04 -0.96 37 0.23 lecY_e4 -0.20 -1.00 -0.17 -3.44 -1.16 -5.80 39 0.25 secondary complete lecY_e5 -0.29 -1.94 -0.07 -1.89 -0.98 -6.84 87 0.07 clecY_e5 0.56 2.67 -0.12 -1.89 0.06 0.90 43 0.21 lecY_e5 -0.42 -2.03 -0.14 -2.73 -1.16 -5.65 44 0.19 more than secondary lecY_e6 -0.20 -1.17 -0.07 -1.75 -0.82 -5.17 85 0.04 clecY_e6 0.55 2.72 -0.08 -1.30 -0.05 -0.81 42 0.19 lecY_e6 -0.16 -0.62 -0.10 -1.57 -0.84 -3.28 44 0.06 first quintile lecY_q1 0.46 1.92 -0.17 -3.60 -0.93 -4.68 55 0.25 clecY_q1 -0.15 -0.42 -0.05 -0.25 0.00 -0.01 29 0.01 lecY_q1 0.21 0.80 -0.19 -3.00 -1.01 -3.74 32 0.27 second quintile lecY_q2 -0.04 -0.18 -0.15 -3.69 -1.02 -5.93 54 0.21 clecY_q2 -0.04 -0.14 -0.01 -0.05 -0.01 -0.07 29 0.00 lecY_q2 0.01 0.06 -0.14 -2.62 -0.89 -3.95 32 0.20 third quintile lecY_q3 -0.26 -1.21 -0.17 -3.51 -1.17 -5.73 52 0.21 clecY_q3 0.26 0.89 -0.14 -0.82 0.10 0.93 26 0.05 lecY_q3 -0.11 -0.49 -0.14 -2.55 -0.90 -3.98 31 0.19 fourth quintile lecY_q4 -0.28 -1.30 -0.15 -3.12 -1.10 -5.55 53 0.17 clecY_q4 0.44 1.48 -0.11 -0.66 0.08 0.70 26 0.10 lecY_q4 -0.16 -0.68 -0.12 -2.09 -0.85 -3.55 31 0.14 fifth quintile lecY_q5 -0.40 -1.81 -0.11 -2.32 -1.02 -5.10 53 0.13 clecY_q5 0.62 1.81 -0.06 -0.32 0.07 0.55 26 0.13 lecY_q5 -0.21 -0.81 -0.08 -1.20 -0.72 -2.63 31 0.06 Jobless total sample ljY_t 0.43 2.02 0.24 4.92 -0.48 -2.43 93 0.22 cljY_t -0.62 -1.31 0.11 0.76 0.09 0.64 46 0.05 ljY_t 0.34 0.97 0.31 3.66 -0.38 -1.08 45 0.24 male ljY_m 0.12 0.56 0.31 6.26 -0.71 -3.62 93 0.30 cljY_m -0.76 -1.77 0.13 1.03 0.08 0.66 46 0.09 ljY_m -0.12 -0.37 0.44 5.82 -0.59 -1.89 45 0.47 female ljY_f 0.60 2.47 0.21 3.65 -0.33 -1.44 93 0.16 cljY_f -0.59 -0.98 0.11 0.62 0.12 0.71 46 0.03 ljY_f 0.59 1.47 0.25 2.56 -0.21 -0.52 45 0.15 rural ljY_r 0.08 0.26 0.30 4.15 -0.55 -1.85 79 0.19 cljY_r -0.60 -1.05 0.15 0.88 0.15 0.88 39 0.05 ljY_r -0.01 -0.02 0.41 3.06 -0.31 -0.57 41 0.21 urban ljY_u 0.73 4.39 0.08 2.16 -0.71 -4.55 80 0.22 cljY_u -0.78 -1.61 0.08 0.53 0.00 0.03 40 0.07 ljY_u 0.59 2.11 0.17 2.37 -0.61 -2.15 42 0.18 no education ljY_e1 -0.77 -2.08 0.28 3.47 -0.39 -1.20 79 0.19 cljY_e1 0.07 0.05 0.39 1.10 0.37 1.59 33 0.04 ljY_e1 -1.07 -1.62 0.41 2.81 -0.16 -0.27 40 0.25 primary incomplete ljY_e2 -0.22 -0.46 0.35 3.07 -0.45 -1.02 75 0.13 cljY_e2 -0.34 -0.32 0.72 2.47 0.08 0.49 36 0.17 ljY_e2 -1.02 -1.47 0.36 2.11 -0.79 -1.18 39 0.17 primary complete ljY_e3 0.57 1.46 0.26 2.86 -0.37 -1.05 82 0.11 cljY_e3 -0.94 -1.85 0.63 2.24 -0.06 -0.36 40 0.18 ljY_e3 0.36 0.80 0.25 2.35 -0.46 -1.06 42 0.13 secondary incomplete ljY_e4 0.58 2.05 0.20 2.95 -0.97 -3.65 78 0.14 cljY_e4 -0.96 -2.32 0.48 2.05 -0.08 -0.59 37 0.21 ljY_e4 0.57 1.70 0.22 2.66 -0.85 -2.62 39 0.20 secondary complete ljY_e5 0.12 0.44 0.11 1.59 -0.94 -3.56 87 0.03 cljY_e5 -0.56 -1.45 0.22 1.82 -0.13 -1.14 42 0.12 ljY_e5 0.20 0.48 0.27 2.59 -0.44 -1.03 44 0.14 more than secondary ljY_e6 -0.12 -0.40 0.16 2.25 -1.32 -4.62 85 0.06 cljY_e6 -0.37 -0.96 0.44 1.99 -0.01 -0.11 41 0.11 ljY_e6 -0.03 -0.09 0.30 3.13 -0.88 -2.25 44 0.20 first quintile ljY_q1 -0.38 -0.86 0.30 3.37 -0.66 -1.76 55 0.19 cljY_q1 -0.46 -0.70 0.81 2.05 0.01 0.04 28 0.15 ljY_q1 -0.48 -0.73 0.40 2.49 -0.50 -0.74 32 0.21 second quintile ljY_q2 -0.17 -0.42 0.27 3.27 -0.71 -1.99 54 0.18 cljY_q2 -0.53 -0.80 0.68 1.72 0.02 0.10 28 0.12 ljY_q2 -0.06 -0.11 0.36 2.58 -0.47 -0.81 32 0.19 third quintile ljY_q3 0.44 1.15 0.31 3.63 -0.36 -1.00 52 0.22 cljY_q3 -0.96 -1.41 0.78 1.94 -0.06 -0.23 26 0.19 ljY_q3 0.37 0.72 0.34 2.63 -0.38 -0.72 31 0.20 fourth quintile ljY_q4 0.56 1.40 0.28 3.21 -0.49 -1.35 53 0.18 cljY_q4 -1.39 -1.76 0.94 2.01 -0.05 -0.18 26 0.22 ljY_q4 0.47 0.90 0.27 2.09 -0.63 -1.15 31 0.14 fifth quintile ljY_q5 0.99 2.98 0.18 2.50 -0.67 -2.19 53 0.21 cljY_q5 -1.56 -2.13 0.69 1.58 -0.05 -0.18 26 0.22 ljY_q5 1.00 2.24 0.11 1.02 -1.02 -2.22 31 0.16 Idleness total sample liY_t 0.93 3.49 0.05 0.81 -1.17 -5.10 80 0.14 cliY_t -0.67 -3.04 -0.01 -0.10 -0.14 -1.77 32 0.25 liY_t 1.08 3.07 0.11 1.31 -0.94 -2.78 38 0.23 male liY_m 0.74 2.25 0.03 0.43 -1.86 -6.52 80 0.06 cliY_m -0.82 -2.59 0.17 0.85 -0.06 -0.52 32 0.19 liY_m 0.77 1.93 0.26 2.62 -1.35 -3.53 38 0.22 female liY_f 1.03 3.60 0.05 0.85 -0.82 -3.30 80 0.15 cliY_f -0.64 -3.21 -0.08 -0.66 -0.13 -1.85 32 0.28 liY_f 1.21 3.00 0.06 0.66 -0.67 -1.73 38 0.21 rural liY_r 0.50 1.46 0.13 1.77 -1.10 -3.77 67 0.07 cliY_r -0.63 -2.57 0.11 0.70 -0.19 -1.99 25 0.23 liY_r 0.76 1.90 0.22 2.19 -0.61 -1.56 34 0.21 urban liY_u 1.18 4.73 -0.10 -2.01 -1.52 -7.14 68 0.30 cliY_u -0.95 -3.90 -0.06 -0.34 -0.22 -2.38 26 0.42 liY_u 1.20 3.21 0.02 0.20 -1.24 -3.47 35 0.24 no education liY_e1 -1.14 -3.00 0.18 2.27 -0.88 -2.76 70 0.18 cliY_e1 -0.63 -0.45 0.11 0.34 0.22 0.99 25 0.01 liY_e1 -0.79 -2.34 0.34 4.37 -0.13 -0.42 35 0.45 primary incomplete liY_e2 -0.60 -1.51 0.29 3.03 -0.83 -2.25 67 0.16 cliY_e2 -0.67 -0.84 -0.09 -0.35 0.10 0.69 28 0.03 liY_e2 -0.73 -1.08 0.28 1.67 -0.98 -1.52 36 0.11 primary complete liY_e3 0.61 1.77 0.08 1.03 -1.13 -3.86 71 0.06 cliY_e3 -1.23 -7.94 -0.07 -0.72 -0.15 -2.62 29 0.72 liY_e3 0.91 1.85 0.08 0.68 -1.01 -2.16 37 0.10 secondary incomplete liY_e4 0.90 2.85 0.00 0.06 -1.84 -6.61 70 0.11 cliY_e4 -1.80 -3.80 -0.14 -0.43 -0.17 -0.91 28 0.38 liY_e4 1.03 2.09 0.03 0.26 -1.71 -3.59 36 0.12 secondary complete liY_e5 0.21 0.54 -0.10 -1.20 -2.04 -5.91 71 0.03 cliY_e5 -0.72 -1.65 -0.11 -0.37 -0.27 -1.64 29 0.11 liY_e5 0.33 0.51 0.19 1.16 -1.21 -1.93 36 0.04 more than secondary liY_e6 0.00 0.00 -0.04 -0.35 -2.72 -6.31 71 0.00 cliY_e6 -0.72 -1.30 0.17 0.46 0.04 0.18 29 0.06 liY_e6 0.25 0.31 0.16 0.84 -2.25 -2.95 37 0.02 first quintile liY_q1 0.23 0.50 0.07 0.77 -1.37 -3.52 48 0.02 cliY_q1 -0.27 -0.61 -0.31 -0.95 -0.09 -0.48 19 0.09 liY_q1 0.34 0.61 0.19 1.32 -0.96 -1.65 27 0.08 second quintile liY_q2 0.34 0.77 0.09 1.04 -1.30 -3.46 48 0.04 cliY_q2 -0.27 -0.67 -0.37 -1.21 -0.12 -0.69 19 0.12 liY_q2 0.54 1.07 0.18 1.37 -0.94 -1.76 27 0.11 third quintile liY_q3 1.01 2.45 0.12 1.27 -0.98 -2.53 46 0.14 cliY_q3 -0.73 -2.05 -0.14 -0.51 -0.26 -1.62 17 0.27 liY_q3 1.06 2.17 0.15 1.17 -0.90 -1.75 26 0.20 fourth quintile liY_q4 1.04 2.34 0.08 0.83 -1.18 -2.84 46 0.12 cliY_q4 -0.98 -2.76 -0.05 -0.18 -0.28 -1.77 17 0.37 liY_q4 1.08 2.11 0.12 0.87 -1.13 -2.09 26 0.18 fifth quintile liY_q5 1.56 3.74 0.02 0.19 -1.29 -3.31 46 0.25 cliY_q5 -1.28 -4.00 -0.04 -0.18 -0.32 -2.23 17 0.55 liY_q5 1.67 3.20 0.01 0.05 -1.43 -2.59 26 0.31 Not in labor force total sample lnlY_t 0.21 1.09 0.08 1.80 -0.36 -2.00 93 0.04 clnlY_t -0.35 -2.36 0.08 1.69 0.05 1.15 46 0.16 lnlY_t -0.13 -0.76 0.14 3.30 -0.36 -2.04 45 0.24 male lnlY_m 0.30 1.05 0.06 0.85 -0.54 -2.01 93 0.02 clnlY_m -0.41 -2.73 0.09 1.94 0.08 1.77 46 0.20 lnlY_m -0.36 -1.70 0.17 3.34 -0.58 -2.70 45 0.30 female lnlY_f 0.21 1.16 0.10 2.33 -0.20 -1.17 93 0.06 clnlY_f -0.32 -1.74 0.07 1.35 0.05 0.93 46 0.10 lnlY_f -0.01 -0.05 0.12 2.51 -0.21 -1.08 45 0.14 rural lnlY_r 0.21 0.77 0.11 1.78 -0.40 -1.61 79 0.04 clnlY_r -0.38 -1.87 0.09 1.41 0.09 1.46 39 0.13 lnlY_r -0.29 -1.14 0.18 2.78 -0.42 -1.64 41 0.22 urban lnlY_u 0.37 1.90 -0.03 -0.67 -0.49 -2.72 80 0.06 clnlY_u -0.38 -2.93 0.07 1.75 0.02 0.45 40 0.23 lnlY_u 0.01 0.08 0.05 1.39 -0.46 -3.40 42 0.05 no education lnlY_e1 -0.78 -3.02 0.22 3.77 -0.53 -2.30 79 0.26 clnlY_e1 -0.54 -0.71 0.00 0.00 0.11 0.95 36 0.02 lnlY_e1 -0.55 -1.53 0.24 2.94 -0.36 -1.11 40 0.26 primary incomplete lnlY_e2 0.57 1.10 0.15 1.25 -0.21 -0.42 76 0.03 clnlY_e2 -0.48 -0.86 0.06 0.42 0.08 0.92 37 0.03 lnlY_e2 -0.29 -0.42 0.25 1.46 -0.39 -0.58 39 0.06 primary complete lnlY_e3 0.38 0.91 0.09 0.89 -0.41 -1.04 86 0.02 clnlY_e3 -0.07 -0.19 0.36 3.23 0.02 0.17 41 0.22 lnlY_e3 0.30 0.91 0.28 3.37 0.09 0.27 44 0.22 secondary incomplete lnlY_e4 0.22 0.88 0.03 0.47 -0.27 -1.16 80 0.01 clnlY_e4 -0.28 -1.80 0.21 4.42 -0.02 -0.38 38 0.39 lnlY_e4 -0.04 -0.20 0.08 1.61 -0.25 -1.22 41 0.07 secondary complete lnlY_e5 0.51 2.31 -0.02 -0.34 -0.59 -2.79 87 0.07 clnlY_e5 -0.27 -1.01 0.06 0.69 0.02 0.25 43 0.04 lnlY_e5 0.43 1.48 0.00 -0.04 -0.66 -2.28 44 0.05 more than secondary lnlY_e6 0.61 2.85 -0.01 -0.16 -0.49 -2.38 87 0.09 clnlY_e6 -0.37 -1.88 -0.06 -1.06 -0.07 -1.20 42 0.11 lnlY_e6 0.28 0.99 -0.01 -0.20 -0.67 -2.32 44 0.03 first quintile lnlY_q1 -0.47 -1.44 0.13 1.99 -0.70 -2.55 55 0.11 clnlY_q1 -0.05 -0.13 -0.07 -0.28 0.07 0.46 28 0.00 lnlY_q1 -0.70 -1.69 0.17 1.71 -0.77 -1.81 32 0.20 second quintile lnlY_q2 -0.31 -1.18 0.10 1.91 -0.65 -2.90 54 0.09 clnlY_q2 -0.06 -0.17 -0.05 -0.25 0.11 0.84 28 0.00 lnlY_q2 -0.68 -1.98 0.14 1.65 -0.78 -2.23 32 0.22 third quintile lnlY_q3 0.10 0.36 0.12 1.96 -0.40 -1.54 52 0.07 clnlY_q3 -0.36 -1.15 0.05 0.25 0.04 0.33 26 0.06 lnlY_q3 -0.57 -1.81 0.14 1.75 -0.74 -2.27 31 0.22 fourth quintile lnlY_q4 0.19 0.71 0.10 1.76 -0.37 -1.54 53 0.06 clnlY_q4 -0.45 -1.51 0.01 0.04 0.08 0.76 26 0.09 lnlY_q4 -0.48 -1.57 0.09 1.13 -0.80 -2.53 31 0.14 fifth quintile lnlY_q5 0.36 1.49 0.06 1.12 -0.33 -1.52 53 0.06 clnlY_q5 -0.46 -2.10 -0.02 -0.14 0.07 0.83 26 0.16 lnlY_q5 -0.27 -1.13 0.03 0.44 -0.77 -3.13 31 0.06 Enrollment total sample lsY_t -0.55 -4.06 0.11 3.76 -0.91 -7.82 80 0.30 clsY_t -0.73 -2.24 0.06 0.60 0.07 1.26 31 0.17 lsY_t -0.52 -2.46 0.18 3.48 -0.65 -3.20 38 0.36 male lsY_m -0.33 -2.71 0.07 2.59 -0.88 -8.24 80 0.16 clsY_m -0.56 -1.94 0.07 0.81 0.06 1.18 31 0.15 lsY_m -0.33 -1.72 0.14 3.00 -0.63 -3.43 38 0.27 female lsY_f -0.81 -4.41 0.16 4.06 -0.97 -6.08 80 0.33 clsY_f -0.83 -2.08 0.05 0.42 0.10 1.42 31 0.14 lsY_f -0.73 -2.96 0.23 3.72 -0.66 -2.79 38 0.42 rural lsY_r -0.50 -2.19 0.12 2.44 -1.10 -5.56 67 0.15 clsY_r -1.01 -1.98 0.12 0.89 0.15 1.85 24 0.18 lsY_r -0.74 -2.53 0.20 2.64 -0.93 -3.26 34 0.31 urban lsY_u -0.35 -2.79 0.02 0.75 -0.92 -8.52 68 0.12 clsY_u -0.27 -0.86 0.02 0.24 0.09 1.77 25 0.04 lsY_u -0.26 -1.35 0.08 1.68 -0.66 -3.57 35 0.13 no education lsY_e1 0.10 0.06 0.72 2.69 -1.63 -1.42 32 0.20 clsY_e1 -6.37 -1.02 -1.03 -0.92 -0.87 -1.27 9 0.19 lsY_e1 -2.40 -1.03 0.60 1.64 -3.25 -1.98 15 0.25 primary incomplete lsY_e2 -0.09 -0.17 0.11 0.95 -1.29 -2.90 63 0.02 clsY_e2 -0.71 -1.06 0.09 0.40 -0.03 -0.21 27 0.05 lsY_e2 0.55 0.99 0.29 2.22 -0.58 -1.13 33 0.16 primary complete lsY_e3 -0.22 -0.30 0.12 0.64 -1.44 -2.06 71 0.01 clsY_e3 0.21 0.32 0.18 0.88 0.17 1.40 26 0.03 lsY_e3 0.61 0.47 0.23 0.72 -0.94 -0.77 37 0.02 secondary incomplete lsY_e4 -0.35 -1.24 0.14 2.24 -0.24 -0.97 71 0.09 clsY_e4 -0.53 -1.06 -0.02 -0.11 0.07 0.79 27 0.04 lsY_e4 -0.04 -0.21 0.08 1.81 -0.28 -1.71 36 0.09 secondary complete lsY_e5 0.33 0.84 -0.05 -0.55 -1.17 -3.40 71 0.02 clsY_e5 1.26 0.56 0.74 1.09 0.24 0.60 29 0.05 lsY_e5 0.52 0.81 -0.21 -1.31 -1.67 -2.72 37 0.07 more than secondary lsY_e6 0.31 0.86 -0.03 -0.41 -0.67 -2.07 73 0.01 clsY_e6 -2.92 -2.87 0.08 0.26 -0.30 -1.67 29 0.25 lsY_e6 0.30 0.50 -0.21 -1.45 -1.11 -1.94 37 0.07 first quintile lsY_q1 -1.30 -2.36 0.14 1.24 -1.68 -3.53 48 0.14 clsY_q1 -0.33 -0.17 -0.26 -0.48 -0.04 -0.15 19 0.01 lsY_q1 -1.84 -2.14 0.22 0.98 -1.71 -1.91 27 0.20 second quintile lsY_q2 -0.63 -2.41 0.11 2.05 -1.18 -5.25 48 0.18 clsY_q2 -0.17 -0.22 -0.05 -0.20 0.18 1.40 18 0.00 lsY_q2 -1.06 -2.89 0.17 1.79 -1.18 -3.07 27 0.34 third quintile lsY_q3 -0.44 -1.55 0.11 1.76 -1.02 -3.86 46 0.13 clsY_q3 -0.67 -0.63 0.00 0.01 0.09 0.64 16 0.03 lsY_q3 -1.03 -2.73 0.15 1.50 -1.16 -2.91 26 0.31 fourth quintile lsY_q4 -0.32 -1.24 0.11 1.86 -0.84 -3.49 46 0.11 clsY_q4 -0.50 -0.51 -0.12 -0.54 0.15 1.16 16 0.04 lsY_q4 -0.90 -2.71 0.10 1.18 -1.09 -3.10 26 0.29 fifth quintile lsY_q5 0.09 0.31 0.07 1.18 -0.58 -2.21 46 0.03 clsY_q5 -0.18 -0.27 -0.06 -0.42 0.11 1.20 16 0.02 lsY_q5 -0.55 -2.09 0.07 1.02 -0.79 -2.84 26 0.20 26 Appendix Table IV: List of Surveys used in the Sample ccode Country name year Survey Name AFG Afghanistan 2003 National Risk and Vulnerability Assessment 2003 Inquerito aos Agregados Familiares sobre Despesas e Receitas 1999- AGO Angola 1999 2000 ALB Albania 2002 Living Standard Measurement Survey 2002 ARG Argentina 1994 Encuesta Permanente de Hogares 1994 ARG Argentina 2001 Encuesta Permanente de Hogares 2001 ARM Armenia 1999 Integrated Survey of Living Standards 1999 ARM Armenia 2003 Integrated Survey of Living Standards 2003 AUS Australia 1989 Survey of Income and Housing Costs (SIHC) AUS Australia 1994 Survey of Income and Housing Costs (SIHC) AUT Austria 1994 Austrian Microcensus AUT Austria 2000 Austrian Microcensus AZE Azerbaijan 1995 State Statistical Survey 1995 AZE Azerbaijan 2002 State Statistical Survey 2002 BDI Burundi 1998 Enquete Prioritaire 1998 BEL Belgium 1992 Socio-Economic Panel (SEP) BEL Belgium 2000 Panel Study of Belgian Households (PSBH) / BE ECHP BEN Benin 2003 Indicateurs de Base du Bien-etre BFA Burkina Faso 1994 Enquete Prioritaire 1994 BFA Burkina Faso 2003 Indicateurs de Base du Bien-etre BGD Bangladesh 1991 Household Income-Expenditure Survey 1991 BGD Bangladesh 2000 Household Income-Expenditure Survey 2000 BGR Bulgaria 1995 Integrated Household Survey 1995 BIH Bosnia and Herzegovina 2001 Living Standard Measurement Survey 2001 BLR Belarus 1998 Income and Expenditure Survey 1998 BLR Belarus 2002 Income and Expenditure Survey 2002 BOL Bolivia 1993 Encuesta Integrada de Hogares 1993 BOL Bolivia 2002 Encuesta Continua de Hogares - Condiciones de Vida 2002 BRA Brazil 1995 Pesquisa Nacional per Amostra de Domicilios 1995 BRA Brazil 2001 Pesquisa Nacional per Amostra de Domicilios 2001 BTN Bhutan 2003 Bhutan Living Standards Survey 2003 CAN Canada 1991 Survey of Consumer Finances (SCF) CAN Canada 2000 Survey of Labour and Income Dynamics (SLID) CHE Switzerland 1992 Income Distribution Survey (HINK) CHE Switzerland 2000 Income Distribution Survey (HINK) CHL Chile 1990 Encuesta de Caracterizacion Socioeconomica Nacional 1990 CHL Chile 1996 Encuesta de Caracterizacion Socioeconomica Nacional 1996 CHL Chile 2003 Encuesta de Caracterizacion Socioeconomica Nacional 2003 CIV Côte d'Ivoire 2002 Enquete Niveau de Vie des Menages 2002 CMR Cameroon 1996 Enquete sur les Niveau de Vie 1995 CMR Cameroon 2001 Deuxieme Enquete Camerounaise aupres des Menages 2001 COL Colombia 1995 Encuesta Nacional de Hogares - Fuerza de Trabajo 1995 COL Colombia 2000 Encuesta Nacional de Hogares - Fuerza de Trabajo 2000 CPV Cape Verde 2000 Recesamento Geral da Populacao e Habitaco 2000 CRI Costa Rica 1995 Encuesta de Hogares de Propositos Multiples 1995 CRI Costa Rica 2001 Encuesta de Hogares de Propositos Multiples 2001 DEU Germany 1994 German Social Economic Panel Study (GSOEP) DEU Germany 2000 German Social Economic Panel Study (GSOEP) 27 DJI Djibouti 1996 Enquete Djiboutienne aupres de Menages 1996 DNK Denmark 1987 Income Tax Register DNK Denmark 1992 Income Tax Register DOM Dominican Republic 1997 Encuesta Nacional de Fuerza de Trabajo 1997 DOM Dominican Republic 2004 Encuesta Nacional de Fuerza de Trabajo 2004 ECU Ecuador 1995 Encuesta Periodica de Empleo y Desempleo 1995 ECU Ecuador 2004 Encuesta Periodica de Empleo, Desempleo y Subempleo 2004 EGY Egypt, Arab Rep. 1998 Egypt Labor Market Survey 1998 ESP Spain 1995 Spanish European Community Household Panel / ES ECHP ESP Spain 2000 Spanish European Community Household Panel / ES ECHP EST Estonia 2000 Household Budget Survey 2000 ETH Ethiopia 1995 Welfare Monitoring Survey 1995 ETH Ethiopia 2000 Welfare Monitoring Survey 2000 FIN Finland 1991 Income Distribution Survey (IDS) FIN Finland 2000 Income Distribution Survey (IDS) FJI Fiji 1996 1996 Population Census FRA France 1989 Household Budget Survey FRA France 1994 Household Budget Survey FSM Micronesia, Fed. Sts. 2000 2000 Census of Population and Housing GBR United Kingdom 1994 Family Resources Survey (FRS) GBR United Kingdom 1999 Family Resources Survey (FRS) GEO Georgia 1998 Survey of Georgian Households GEO Georgia 2002 National Survey of Households 2002 GHA Ghana 1991 Ghana Living Standards Survey 3 GHA Ghana 1998 Ghana Living Standards Survey 4 GIN Guinea 1994 Enquete Integrale sur les Conditions de Vie des Menages GMB Gambia, The 1998 1998 Household Poverty Survey GTM Guatemala 1989 Encuesta Nacional Socio Demografica GTM Guatemala 2002 Encuesta Nacional sobre Condiciones de Vida (ENCOVI) GUY Guyana 1992 Living Standards Measurement Survey GUY Guyana 2000 Multiple Indicators Cluster Survey HND Honduras 1995 Encuesta Permanente de Hogares de Propositos Multiples HND Honduras 2003 Encuesta Permanente de Hogares de Propositos Multiples HRV Croatia 2004 Household Budget Survey HTI Haiti 2001 Enquete sur les Conditions de Vie en Haiti HUN Hungary 1998 Household Monitor Survey HUN Hungary 2000 Household Monitor Survey HUN Hungary 2002 Household Monitor Survey IDN Indonesia 1993 National Socioeconomic Survey (SUSENAS) IDN Indonesia 2002 National Socioeconomic Survey (SUSENAS) IND India 1994 National Sample Survey IND India 2000 National Sample Survey IRL Ireland 1992 Living in Ireland Survey / IE ECHP IRL Ireland 1994 Living in Ireland Survey / IE ECHP IRL Ireland 2000 Living in Ireland Survey / IE ECHP IRL Ireland 2001 Living in Ireland Survey / IE ECHP ITA Italy 1991 Survey on Household Income and Wealth (SHIW) ITA Italy 2000 Survey on Household Income and Wealth (SHIW) JAM Jamaica 1996 Jamaica Survey of Living Conditions JAM Jamaica 2002 Jamaica Survey of Living Conditions JOR Jordan 2002 Household Income and Expenditure Survey KAZ Kazakhstan 2002 Household Budget Survey 28 KEN Kenya 1997 Welfare Monitoring Survey III KGZ Kyrgyz Republic 1996 Household Budget Survey KGZ Kyrgyz Republic 2002 Household Budget Survey KHM Cambodia 1997 Cambodia Socio Economic Survey 1997 KHM Cambodia 2004 Household Socio Economic Survey 2003-2004 KIR Kiribati 2000 Population and Housing Census LKA Sri Lanka 1995 Household Income and Expenditure Survey LKA Sri Lanka 2002 Household Income and Expenditure Survey LSO Lesotho 2002 Welfare Indicators Survey LTU Lithuania 2000 Household Budget Survey LUX Luxembourg 1991 Socio Economic Panel (PSELL) LUX Luxembourg 2000 Socio Economic Panel (PSELL) LVA Latvia 2004 Household Budget Survey MDA Moldova 1998 Household Budget Survey MDA Moldova 2002 Household Budget Survey MDG Madagascar 1993 Enquete Permanante aupres des Menages MDG Madagascar 2001 Enquete Prioritaire aupres des Menages MDV Maldives 1998 Household Income and Expenditure Survey MDV Maldives 2002 Household Income and Expenditure Survey MEX Mexico 1994 Encuesta Nacional de Empleo MEX Mexico 2002 Encuesta Nacional de Empleo MHL Marshall Islands 1999 Census of Population and Housing MOZ Mozambique 1996 Inquerito Nacional aos Agregados MRT Mauritania 2000 Enquete Permanantes sur les Conditions de Vie des Menages MWI Malawi 1997 Integrated Household Survey NAM Namibia 1993 Household Income and Expenditure Survey NER Niger 2002 Enquete Emploi NGA Nigeria 2003 Nigeria Living Standards Survey Encuesta Nacional de Hogares sobre Medicion de Nivel de Vida NIC Nicaragua 1993 (EMNV) Encuesta Nacional de Hogares sobre Medicion de Nivel de Vida NIC Nicaragua 2001 (EMNV) NLD Netherlands 1994 Socio-Economic Panel (SEP) NLD Netherlands 1999 Socio-Economic Panel (SEP) NOR Norway 1995 Income Distribution Survey (IF) NOR Norway 2000 Income Distribution Survey (IF) NPL Nepal 1995 Nepal Living Standards Survey PAK Pakistan 1999 Household Income and Expenditure Survey PAK Pakistan 2001 Household Income and Expenditure Survey PAN Panama 1991 Encuesta de Hogares PAN Panama 1995 Encuesta de Hogares PAN Panama 2003 Encuesta de Hogares PER Peru 1994 Encuesta Nacional de Hogares PER Peru 2002 Encuesta Nacional de Hogares PLW Palau 2000 Census of Population and Housing POL Poland 1998 Household Budget Survey POL Poland 2002 Household Budget Survey PRY Paraguay 1995 Encuesta de Hogares PRY Paraguay 2001 Encuesta de Hogares ROM Romania 1994 Romanian Integrated Household Survey (RIHS) ROM Romania 2002 Romanian Integrated Household Survey (RIHS) RUS Russian Federation 1997 Russia Longitudinal Monitoring Survey (RLMS) 29 RUS Russian Federation 2002 Russia Longitudinal Monitoring Survey (RLMS) RWA Rwanda 1997 Enquete Integrale sur les Conditions de Vie des Menages SEN Senegal 1995 Enquete sur les Depenses des Menages SLB Solomon Islands 1999 Population Census SLE Sierra Leone 2003 Sierra Leone Integrated Household Survey SLV El Salvador 1995 Encuesta de Hogares de Propositos Multiples SLV El Salvador 2002 Encuesta de Hogares de Propositos Multiples STP São Tomé and Principe 2000 Enquete sur les Conditions de Vie des Menages SVK Slovak Republic 1992 Slovak Microcensus SVN Slovenia 1999 Household Budget Survey SWE Sweden 1995 Income Distribution Survey (HINK) SWE Sweden 2000 Income Distribution Survey (HINK) SWZ Swaziland 1995 Household Income and Expenditure Survey SWZ Swaziland 2000 Household Income and Expenditure Survey THA Thailand 1990 Socio Economic Survey THA Thailand 1994 Socio Economic Survey THA Thailand 2002 Socio Economic Survey TJK Tajikistan 1999 Tajikistan Living Standards Survey TON Tonga 1996 National Population Census TTO Trinidad and Tobago 1992 Survey of Living Conditions TUR Turkey 2002 Household Income and Expenditure Survey TWN Taiwan, China 1995 Survey of Family Income and Expenditure TWN Taiwan, China 2000 Survey of Family Income and Expenditure TZA Tanzania 1991 Household Budget Survey TZA Tanzania 2000 Household Budget Survey UGA Uganda 2002 National Household Survey UKR Ukraine 1999 Household Budget Survey UKR Ukraine 2003 Household Budget Survey URY Uruguay 1995 Encuesta Continua de Hogares URY Uruguay 2003 Encuesta Continua de Hogares USA United States 1994 Current Population Survey (CPS) USA United States 2000 Current Population Survey (CPS) VEN Venezuela, RB 1995 Encuesta de Hogares por Muestreo VEN Venezuela, RB 2004 Encuesta de Hogares por Muestreo VNM Vietnam 1992 Living Standards Survey VNM Vietnam 2001 Household Living Standards Survey YEM Yemen, Rep. 1998 Demographic and Health Survey YUG Serbia and Montenegro 2005 Household Budget Survey ZAF South Africa 1995 Household Survey ZAF South Africa 2000 Household Income and Expenditure Survey ZMB Zambia 1998 Living Conditions Monitoring Survey ZMB Zambia 2002 Living Conditions Monitoring Survey 30 References ILO. 2004. Global Employment Trends for Youth. Geneva: International Labour Office. Korenman, Sanders, and David Neumark. 2000. "Cohort Crowding and Youth Labor Markets: A Cross-National Analysis." In David G. Blanchflower and Richard B. Freeman, (eds.), Youth Employment and Joblessness in Advanced Countries. Chicago and London: University of Chicago Press for the National Bureau of Economic Research. O'Higgins, Niall. 2003. "Trends in the Youth Labor Market in Developing and Transition Countries." Washington, D.C.: World Bank, Social Protection Discussion Paper Series 0321. 31