Youth in the Labor Market and the Transition from School to Work in Tanzania

SP DISCUSSION PAPER NO. 0606 Youth in the Labor Market and the Transition from School to Work in Tanzania Florence Kondylis and Marco Manacorda July 2006

Youth in the Labor Market and the Transition from School to Work in Tanzania Florence Kondylis & Marco Manacorda July 2006 1

Youth in the Labor Market and the Transition from School to Work in Tanzania 1 Florence Kondylis* and Marco Manacorda** 1. Introduction Tanzania, together with many other Sub-Saharan countries (see Guarcello and others 2005), suffers from a severe youth unemployment and inactivity problem in urban areas (Mjema 1997). Despite sustained growth in the second half of the last decade, during the 1990s labor market outcomes have further deteriorated (Government of Tanzania 2003). Although unemployment is by no means a problem unique to youths in Sub-Saharan Africa, the problem there is compounded by disappointing education outcomes that make the prospects of youths appear rather dim and by the circumstance that work is often the only asset for a large part of the population while no publicly provided insurance mechanism against the risk of unemployment is in place. Despite unemployment being largely an urban phenomenon in Tanzania, labor market outcomes of rural youths are not much rosier. Although rural children transition at very early ages into work (with no or little schooling or sometimes in combination with school) (see for example Beegle and Burke 2004; Beegle and others 2004.most end up in low productivity jobs on the household farm. This is a possibly a major reason behind increasing migration from the countryside to urban areas (U.S. Census Bureau 1995), even in the face of poor and deteriorating urban labor market prospects. Urbanization reflects wider demographic trends: between 1957 and 2002 population grew fourfold, and this trend is not expected to end any 1 We are grateful to Jean Fares, Marito Garcia, and seminar participants at the workshop on Youth in Africa's Labor Market, Washington, D.C., for many helpful comments and suggestions. All remaining errors are ours. Corresponding author: Marco Manacorda: Centre For Economic Performance, London School of Economics, Houghton Street, WC2A 2AE London, United Kingdom. Email: m.manacorda@lse.ac.uk. *Department of Economics, RHUL and CEP- London School of Economics, UK **Department of Economics, QMUL and CEP London School of Economics, UK 1

time soon, 2 casting serious doubts on the possibility that the youth joblessness problem will disappear in the meantime. Although we are not the first to document the level of youth joblessness in Tanzania (Mjema 1997; Government of Tanzania 2003; LO/FTF 2003), our paper aims to shed some additional light on this phenomenon. First, we provide evidence on different dimensions of youths' labor market performance. For this exercise we can rely on micro data from the Tanzanian Integrated Labor Force Survey (ILFS) of 2000/01, a rather large household survey (approximately 11,000 households) that provides a rich array of information on employment, job search, schooling, training, and migration, together with basic information on individuals' and their households' characteristics. Second we attempt to uncover the determinants of youths' labor market outcomes and to tease out significant predictors of labor market success and failure using simple regression tools. The structure of the paper is as follows. Section 1 presents an overview of the youth unemployment problem. Section 2 presents the data. Section 3 presents detailed descriptive statistics on youths' labor market performance. Section 4 presents the regression results. Section 5 concludes. 2. Youth in the labor market By now an extensive literature analyzes youths' labor market outcomes and their transition into adulthood in Organisation for Economic Co-operation and Development (OECD) countries and especially the United States (for all, see OECD 1996, 1998, 1999, 2000 and Ryan 2001). As discussed in Rees (1986), youths typically display lower labor market attachment and lower employment rates than older workers. Some of them are still engaged in full time education or combine education with work; others devote time to job searching or move from one job to another as part of their investment in human capital or as a process of mutual information gathering with employers. From this perspective youth joblessness reflects (a potentially efficient) mechanism of allocating workers to jobs. Lack of dependents and the possibility of relying on parental support often make joblessness a less painful alternative for young workers and less of a problem from the perspective of the social 2 See http://www.tanzania.go.tz/ppu/demografic_situation.html. 2

planner. Lower wages associated with lower experience levels or stronger preferences for leisure also potentially imply lower disutility of being out of employment for young workers. Not only do youths display higher rates of joblessness and unemployment than adults due to frictional reasons at any given point in time, but they also appear to be more sensitive to the state of the economic cycle. The youth unemployment problem in most developed countries (and in particular in OECD countries following the oil shocks of the 1970s) is largely attributed to the weakness of the economy and overall lack of labor demand (Rees 1986; Freeman and Wise 1982; Blanchflower and Freeman 2000a, 2000b; ILO 2000; Card and Lemieux 2000). Disadvantaged youths in particular appear to bear a disproportionate share of the cost of economic downturns or weak labor demand in their area of residence (Freeman 1991; Freeman and Rodgers 1999). Reasons for the extreme vulnerability of youths in the labor market to economic downturns have largely to do with their lower level of labor market skills (experience and sometimes education) and lower labor market attachment (including lower job search), employment protection legislation, and hiring and firing rules that often penalize recently hired workers. If aggregate demand matters, aggregate supply does too. At given labor demand, a rise in the proportion of youths in the labor market seems to disproportionately affect the youths themselves, consistent with a world where youths and adults are only imperfect substitutes in production. Excess supply relative to demand affects wages, employment, or both (Welch 1979; Card and Lemieux 2001; Koreman and Neumark 2000). As young workers see their employment prospects deteriorate, not only do they tend to work less, but they respond with adjustment at different margins, including a higher probability of staying in school, residing with parents (Card and Lemieux 2000), or even committing crime (Freeman 1996, 1999). Much less is known about the behavior of youths in developing countries. Rozenzweig (1988) and O'Higgins (2003) both show that higher unemployment and joblessness rates among youths are widespread in many developing countries. Work by Guarcello and others (2005) also document very high inactivity rates among youths in 13 Sub-Saharan African countries. Other researchers present similar pictures from other parts of the world (for example, Rama 2003 for Sri Lanka). 3

A commonly held view about urban labor markets of developing countries is that (youth) joblessness is a luxury accessible only to those from more advantaged backgrounds, often proxied by their education. Unemployment is often regarded as an option pursued by youths queuing for a job in the public sector or waiting to fill a vacancy in the formal private sector. In the presence of widespread poverty and in the absence of public provision of welfare nonemployment is just a nonviable option for the poor, who will have no other option but making ends meet through informal and causal work. From this perspective the youth unemployment problem should not per se be a source of major policy concern since this is by enlarge a voluntary phenomenon, In the rest of this paper we document youths' labor market outcomes in Tanzania and we explicitly attempt to document what role if any market forces play in shaping these outcomes and how individuals respond to changing economic incentives. We argue in particular that youth joblessness is by no means a voluntary phenomenon in Tanzania. In the last part of the paper we summarize these findings with an eye to the potential role of the policymaker. 3. Data In this section we present basic descriptive evidence on school attendance and labor market performance of teenagers and youths in Tanzania. For the purpose of this exercise we use micro data from the 2000/01 ILFS. The ILFS is a rather large sample survey (43,558 individual observations in 11,158 households) collecting a rich array of information on several features of individuals' work activity, schooling and (off the) job-search together with information on a number of individual and household characteristics. In the rest we present evidence on individuals aged 15 19 (teenagers) and 20 24 (youths) relative to individuals aged 35 49 (prime-age individuals). We present separate results for men and women and for the main geographical areas of the country: Dar Es Salaam, other urban areas, and rural areas. We refer sometimes to Dar Es Salaam plus other urban areas as urban areas. Dar Es Salaam, other urban areas, and rural areas account for 15%, 17%, and 69% respectively of the observations in the sample and for 7%, 16%, and 77% of the weighted sample. Teenagers and youths account for a slightly higher share of the population in urban areas, accounting for 8% of the weighted sample in Dar Es Salaam, 17% 4

in other urban areas, and 75% in rural areas, possibly suggesting an increasing trend toward rural-urban migration. For men we have 1,008, 996, and 3,971 observations in Dar Es Salaam, other urban areas, and rural areas respectively. For women these figures are 1,162, 1,221, and 3,982. Because we rely on a single (albeit very rich) cross section of data, in order to uncover the determinants of youths' labor market outcomes we exploit differences between different groups of workers within Tanzania (differences between young and old workers or across groups of young workers with different observable characteristics such as region of residence, gender, or age). While this strategy has the advantage of generating sufficient variation in the data to credibly identify the impact of the variables of interests, this also implies that we remain largely agnostic on the macro-determinants (that is, those common to everybody in the labor market) of the state of the labor market in Tanzania. 4. Descriptive evidence We start by presenting data on individuals' labor force status and school attendance. We move on to examine the characteristics of those working and then concentrate on those out of work. Labor force status and schooling choices Table s 1A reports basic labor supply indicators for men. In all tables the data are weighted by sampling weights. Column 1 reports information on the proportion of individuals in school. The data illustrate that school attendance is on the order of 58% for male teenagers in Dar Es Salaam and tends to fall as one moves to other urban areas and then to rural areas, where school attendance is on the order of 39%. A similar pattern can be identified for male youths, with around 14% of them in school in Dar Es Salaam and only 2% in rural areas. Among individuals out of school, some drop out at an earlier age while others never attend. Column 2 reports information on those who never attended school. School attendance at one point in an individual's life is almost universal in Dar Es Salaam (on the order of 97% for men. A similar picture emerges in other urban areas where the proportion of males who never attended school is on the order of 4% irrespective of age. School attendance though is far from universal in rural areas: 15% of male teenagers and youths have never 5

attended school. This proportion rises to 19% for prime-age men, suggesting an improvement in education outcomes across subsequent cohorts of men. Table 1A: Labor Force Status and Schooling MALES (1) (2) (3) (4) (5) Age Group School Never attended Work Work & School No work & no school Dar es Salaam Teens 0.581 0.026 0.206 0.036 0.249 Youth 0.142 0.042 0.472 0.020 0.407 Prime-age 0.000 0.021 0.974 0.000 0.026 Urban Teens 0.508 0.040 0.435 0.122 0.178 Youth 0.078 0.038 0.760 0.016 0.177 Prime-age 0.000 0.042 0.949 0.000 0.051 Rural Teens 0.391 0.146 0.761 0.219 0.067 Youth 0.023 0.154 0.922 0.009 0.065 Prime-age 0.000 0.192 0.968 0.000 0.032 Patterns of work participation in column 3 are, to a large extent and in all areas, the mirror image of patterns of school attendance. Work here refers to any work activity in the week prior to the survey. The data include people with a job but temporarily absent from it. While around 20% of male teenagers are working in urban areas, the corresponding proportion is 43% in other urban areas and 76% in rural areas. Similar patterns can be identified for youths, with an employment-to-population ratio that increases from 47% in Dar Es Salaam to 76% in other urban areas. By contrast, the majority of male youths living in rural areas work, with an employment-to-population ratio of 92%. Teenager and youth participation rates are always below prime-age men's, which is on the order of 95 97% with little variation across areas. Column 4 reports the proportion of young men combining work and school. One can clearly see that part-time work and school is essentially a phenomenon affecting teenagers in all areas. In other urban areas there is a small proportion of individuals doing both activities (less than 4%), but this proportion rises to 22% in rural areas. This is probably due to the 6

circumstance that rural teenagers are able to provide their work services on the household farm, without the need for a lengthy job search or formal contractual arrangements. In addition, lower household income in these areas makes these individuals potentially more likely to work even while still in school, while the lack of substantial alternative work opportunities other than on the household farm makes the return to search quite low. Column 5 analyses the proportion of individuals who are neither at work nor at school (sometimes defined as jobless; see Ryan 2001). 3 This column provides a first illustration of the problems that young individuals face in Tanzania s labor market. Around 25% of teenagers and 40% of youths are neither at school nor at work in Dar Es Salaam. The corresponding proportions in other urban areas are 18% for both teenagers and youths. In rural areas, joblessness is lower and on the order of 7% for both age groups. Table 1B: Labor Force Status and Schooling FEMALES (1) (2) (3) (4) (5) Age Group School Never attended Work Work & School No work & no school Dar es Salaam Teens 0.442 0.036 0.265 0.027 0.319 Youth 0.054 0.043 0.374 0.009 0.580 Prime-age 0.000 0.120 0.693 0.000 0.307 Urban Teens 0.368 0.061 0.445 0.089 0.276 Youth 0.025 0.050 0.672 0.000 0.304 Prime-age 0.000 0.157 0.892 0.000 0.108 Rural Teens 0.343 0.186 0.762 0.169 0.065 Youth 0.012 0.213 0.924 0.007 0.071 Prime-age 0.000 0.474 0.951 0.000 0.049 Note: The table reports the proportion of teenagers (aged 15 19), youths (aged 20 24), and prime-age individuals (aged 35 49) who report being enrolled in school (column 1), never having attended school (column 2), in work (column 3), combining work and school (column 4) and neither in work nor in school (column 5) in Dar Es Salaam, other urban areas, and rural areas. All data are weighted by sampling weights. Source: IFLS 2000/01 micro data. 3 One can check that the sum of those in work (column 1), plus those in school (column 3), plus those neither in school nor in work (column 4) minus those combining work with school (column 5) adds up to one. 7

Looking at women's labor force status (table 1B), notable differences between genders emerge. Women are less likely to be in school relative to men of the same age. This is particularly evident in urban areas: in Dar Es Salaam the proportions of female teenagers and youths in school are 44% and 5% respectively (that is, 14 percentage points and 9 percentage points respectively less than boys of the same age). In other urban areas, the proportions of female teenagers and youths in school are 37% and 2% respectively (12 percentage points and 5 percentage points respectively less than boys of the same age). In rural areas, where boys' school attendance is lower, differences between girls and boys are less evident, with a proportion of female teenagers in school of 34% and a proportion of female youths in school of 1% (that is, 5 percentage points and 1 percentage point respectively less than boys of similar age). Column 2 investigates whether these differences are due to girls being less likely to enroll in school in the first places. The proportion of teenage and young girls who never attended school is around 4% in Dar Es Salaam and between 5% and 6% in other urban areas, hence exhibiting little difference with respect to boys. This suggests that girls are on average less likely to remain in school that boys are. The proportion of teenage and young girls who never attended school is much higher in rural areas: 19% and 21% respectively, or between 4 percentage points and 5 percentage points more than men. Although girls appear to do worse than boys in terms of school attendance, a comparison with older individuals shows that recent cohorts of women have experienced a remarkable progress relative to men in both rural and urban areas. The proportion of prime-age women who never attended school is 12% in urban areas (10 percentage points more than men), 16% in other urban areas (12 percentage points more than men), and 47% in rural areas (27 percentage points more than men) As with men, female employment ratios increase with age in all areas, and they are at their lowest in Dar Es Salaam and at their highest in rural areas. As illustrated in column 3, the proportion of female teenagers at work is 27% in Dar Es Salaam, 45% in other urban areas, and 76% in rural areas. The corresponding proportions among female youths are 37%, 67%, and 92%. In general, teenage girls are more likely to be working than teenage boys: this differences range from 6 percentage points in Dar Es Salaam to 1 8

percentage point in other urban areas. Differences are statistically significant. 4 No differences emerge in rural areas. The pattern is reversed among youths, as young girls are less likely to be in work than young boys. Here differences range from 10 percentage points lower in Dar Es Salaam to 9 percentage points lower in other urban areas. Again, no differences emerge between girls and boys in rural areas. One potential explanation for this pattern is that girls in urban areas drop out of school earlier than boys and enter the labor market earlier. However, as they age, some of them tend to withdraw from the labor market, as they get gradually absorbed by childrearing and other domestic activities, while potentially a smaller proportion of school leavers enter the labor market. This is confirmed by an analysis of employment to population ratios among prime-age women that shows a negative female-male gap. The differences in the employment-topopulation ratios between prime-age women and men are 28 percentage points lower in Dar Es Salaam and 5 percentage points lower in other urban areas. Differences in rural areas are on the order of only 1 percentage point. Column 4 shows that girls are also less likely than boys to combine work and education. This is a largely a reflection of the fact that fewer women are in school full time. If one standardizes the proportion of those combining work and school (in column 4) to the proportion in school (in column 1), results are very similar for men and women, so that, conditional on being in school, the probability of work is similar for boys and girls. Finally, column 5 reports the proportion of women neither at school nor at work. As one could have expected from the data presented in the previous columns in urban areas, girls are more likely to be jobless than boys. This likely partly reflects lower labor supply of women together with potentially lower demand for their work services. As with men, it appears that young girls aged 20 24 (youths) are at greater risk of being neither in school nor in work. For example, the proportion of jobless women rises from 32% for teenagers in Dar Es Salaam to 58% for youths and falls to 30% for prime-age women. The corresponding proportions in other urban areas are 28%, 30%, and 10%. There are no 4 One can easily check that with around 1,000 observations (N), the standard deviation of a proportion (p) is at most 0.015 (this is the square root of (p)(1-p)/n for p=0.5, that is, at its maximum). This suggests that most of the differences between boys and girls in this and the remaining tables are statistically significant since they are outside the other gender group confidence interval. 9

discernible differences in the prevalence of joblessness between boys ands girls in rural areas. In sum, there is evidence that a non-negligible proportion of the population drops out of school and starts to work at a rather early age, especially in rural areas. In general, girls drop out earlier and enter the labor market sooner than boys. However, as an increasing proportion of individuals drops out of school, the chances of finding a job tend to fall in urban areas. Whereas, most men eventually appear to get absorbed into the labor market, a large proportion of women remains out of the labor market; especially in Dar Es Salaam, possibly devoting their time to home production. In rural areas the data suggest a smoother transition, with a large proportion of individuals transitioning into work at early ages. This is true for both men and women. It is important, however, to emphasize that this smoother transition in rural areas might be the result of individuals being required to work at early ages in order to guarantee their household's and their own survival, together with lower returns to education and job search. Rural jobs are likely to provide only subsistence for many individuals. In this sense, such quicker transitions are possibly associated with worse lifetime outcomes for individuals in rural areas than for those in urban areas. Employment In this section we concentrate on the characteristics of those in employment. Tables 2A and 2B report the distribution of work by occupation, together with some information on hours of work and underemployment. Columns 1 to 5 report the proportion in work split into five categories: those in paid employment (employees), self-employed (split between those with and without employees), those performing unpaid work in the family nonagricultural business (typically shops), and those working on their own farm. Work for pay includes payment both in cash and in kind. The data refer to the individual's main occupation in the week prior to the survey. Boys are in general more likely to perform work on the household farm or business and less likely to be in paid employment or to run their own business than are prime-age men. For example, among teenagers the proportion of employees is 41%, 15%, and 4% in Dar Es Salaam, other urban areas, and rural areas respectively. For 10

prime-age men these proportions are 55%, 37%, and 9%. Similarly, the proportion working in the family business (columns 4 and 5 together) in the three areas is 33%, 68%, and 94% for teenagers and 5%, 29%, and 86% for prime age men. Self-employment (columns 2 and 3 together) interests respectively 27%, 17%, and 2% of teenagers and 40%, 34%, and 5% of prime-age men. One possible interpretation of these figures is that paid employment might require a lengthy job search, and access to self-employment might require either capital or access to credit, with both these conditions being probably harder to fulfill for younger individuals. 5 Column 6 presents data on total hours of work in all jobs. In general workers in urban areas tend to work more hours. Prime-age men work on average 58 hours a week in Dar Es Salaam and 62 hours in other urban areas. In rural areas average hours of work are lower, on the order of 54 hours. Both teens and youths tend to work less then prime-age men, but patterns across areas largely reflect those of prime-age men's. Average number of hours of work among teenagers is on the order of 53, 44, and 43 respectively in Dar Es Salaam, other urban areas, and rural areas respectively. For youths, these numbers are 58, 58, and 52 respectively. 6 Data on hours of work include all jobs held by individuals. A non-negligible proportion of individuals in Tanzania hold at least two jobs. Column 7 reports this proportion. Multiple job holding is particularly widespread in other urban areas and in rural areas and is more common among prime-age men than among teenagers and youths. For example only 2% of youth in employment have a second job in Dar Es Salaam compared with around 8% of prime-age men. In rural areas these figures are 18% and 5 In addition a compositional effect is likely to be at work. This is because, as the labor force ages, an increasing proportion of it is composed by individuals with higher education. These trends potentially reflect the circumstance that more educated individuals are more likely to enter into paid employment or to start their own business (especially with employees) than less educated individuals. Regression results (not reported) show that conditional on education the probability of being an employee does not grow with age. However, for selfemployment with employees there is still a pronounced age growth, even conditioned on education. This suggests that compositional effects are important in explaining the growth in dependent employment over the life cycle but not the growth in self employment with employees. 6 Because some teenagers tend to combine work and school, one might think that a more appropriate comparison is between young and prime-age individuals out of school. Effectively, when one restricts to this sample, the data (not reported in the table) show that teenagers tend to work less than older workers in both rural and other urban areas, but more than older workers in Dar Es Salaam. 11

24%. Overall it appears that young individuals work fewer hours than prime-age men and are less likely to hold more than one job. One might wonder whether these differences in hours worked across different age groups reflect differences in the supply or the demand for labor. In what consists of an admittedly imperfect measure of the imbalance between the demand and the supply of hours of work across age groups, we report an indicator of underemployment in column 8. This measures the proportion of individuals who work fewer than 40 hours a week and declare a desire to work more hours. 7 It is interesting to observe that this proportion is always the highest among young individuals. For instance, in Dar Es Salaam 7% of teenagers and 4% of youths declare being underemployed. For prime-age men this proportion is only 1%. In rural areas the corresponding proportions are respectively of 5%, 6%, and 3%. 7 Unfortunately, the questionnaire does not ask whether this refers to a desire of working more hours at the same wage. 12

Table 2A: Job Characteristics MALES (1) (2) (3) (4) (5) (6) (7) (8) Occupation Self-employed, Own Usual Multiple no employees farm hours jobs Age Group Employee Self-employed with employees Unpaid family worker Underemployed Dar es Salaam Teens 40.76 0.00 27.20 22.75 9.29 53.061 0.003 0.066 Youth 47.38 1.72 38.36 6.65 5.88 58.077 0.021 0.042 Prime-age 54.46 10.32 30.00 0.00 5.22 57.808 0.078 0.010 Urban Teens 15.05 0.21 16.74 5.24 62.76 44.137 0.160 0.028 Youth 21.77 3.06 22.49 3.82 48.87 58.180 0.104 0.037 Prime-age 36.94 5.54 28.48 0.38 28.66 62.288 0.211 0.021 Rural Teens 3.63 0.04 2.32 3.47 90.54 43.116 0.125 0.048 Youth 5.62 0.34 4.63 0.83 88.57 52.876 0.183 0.057 Prime-age 8.95 0.99 4.44 0.25 85.37 54.701 0.240 0.033 Note: The table reports the characteristics of those in work. Columns 1 5 report the occupation held (employee, self-employed with employees, selfemployed with no employees, unpaid family worker, and working on own farm). Column 6 reports usual hours of work among those reporting positive hours. Column 7 reports the proportion holding more than one job. Column 8 reports the proportion working less than 40 hours a week and wishing to work more hours. See also notes to table 1A. 13

Age Group Employee Selfemployed with employees Table 2B: Job Characteristics FEMALES (1) (2) (3) (4) (5) (6) (7) (8) Occupation Self-employed, Own Usual Multiple no employees farm hours jobs Unpaid family worker Underemployed Dar es Salaam Teens 52.06 0.42 26.12 15.24 6.16 56.324 0.036 0.071 Youth 36.38 2.77 53.82 4.37 2.66 54.724 0.025 0.136 Prime-age 27.22 3.27 55.79 0.39 13.34 50.425 0.070 0.104 Urban Teens 16.77 1.27 11.72 20.86 49.39 43.443 0.187 0.122 Youth 19.08 1.08 35.18 6.51 38.14 48.976 0.150 0.078 Prime-age 14.12 5.85 28.40 2.88 48.77 52.829 0.263 0.095 Rural Teens 1.86 0.00 1.91 5.16 91.06 41.728 0.136 0.080 Youth 1.32 0.28 1.94 1.38 95.08 46.943 0.149 0.109 Prime-age 2.29 0.19 3.03 0.78 93.70 48.642 0.199 0.082 Note: See notes to table 2A. 14

Table 2B reports the employment characteristics of women. Teenage girls appear to be more likely to work as employees than teenage boys in urban areas (52% and 17% in Dar Es Salaam and other urban areas respectively, versus 41% and 15% for boys) and less likely to work in the family enterprise. Changes in the distribution of women's employment over the life cycle appear rather different from men's. As with men, the proportion in nonagricultural self-employment rises with age in each area (from 27% to 59% in Dar Es Salaam, from 13% to 36% in other urban areas, and from 2% to 3% in rural areas), and the proportion engaged in unpaid nonagricultural family work falls (from 15% to less than 1% in Dar Es Salaam, from 21% to 3% in other urban areas, and from 5% to 1% in rural areas). However, in contrast to men, the proportion in salaried employment falls (from 52% to 37% in Dar Es Salaam and from 17% to 14% in other urban areas) or stays constant (at 2% in rural areas), while engagement in the household farm rises. In urban areas, prime-age working women are less likely to be in paid employment or to be self-employed with employees than men are and more likely to be self-employed with no employees or to work in the family enterprise than prime-age men are. In rural areas most working women tend to engage in work on the household farm. These women account for 94% of working women in rural areas (compared with 85% of working men). These patterns might reflect different opportunities in the access to salaried employment for women compared with men, possibly due to their lower labor market characteristics (for example, education) or as a result of gender discrimination. The need to take care of their children and families might also make salaried dependent employment a less attractive opportunity for women in Tanzania. Information on hours shows that on average women tend to work fewer hours than men. Differences in average hours of work between prime-age women vary between minus 8 in urban areas to minus 5 in rural areas. The same does not apply for teenage girls: the average differences in hours of work among teenage girls and boys range from a positive value of 3 in Dar Es Salaam to a negative value of 1 elsewhere. This is consistent with the notion that women tend to engage increasingly in household chores, although an alternative explanation may be that women who start to work at very early ages are those with the higher marginal utility of consumption relative to leisure, that is, those from poorer backgrounds or whose leisure parents value less, hence those who provide more hours in the market. 15

Underemployment is larger for females than for males. For example, 14% of female youth in Dar Es Salaam declare being underemployed compared with only 4% of men. In sum, the career profiles of men and women appear rather different. As they become older, urban men tend increasingly to move away from work in the household enterprise or farm toward salaried employment and self-employment. At the same time these individuals tend to work more hours. This is a combination of true job changes, together with the fact that those who leave school later tend to be more likely to engage in salaried and self-employment and to work more hours (plus possibly differences across cohorts). A large majority of working men in rural areas are engaged in work in the household farm, although movements toward salaried employment and self-employment that are qualitatively similar to those in urban areas are observed here. As they grow older, women in urban areas increasingly work either as self-employed with no employees or in the family farm. In part, this might reflect a movement away from salaried employment due to the need for more flexible working arrangements in order to attend to domestic duties. In rural areas, women's participation is higher at any age, and there is little indication that rural women withdraw from the labor market. Almost all these women work on the household farm. Women tend to work fewer hours than men, but they are also more likely to declare being underemployed. In this sense, lower labor market attachment on the part of women does not seem to be completely ascribable to their lower labor supply. There is evidence that women in Tanzania find it particularly hard to access the labor market in urban areas, probably because of a combination of discrimination and lower market skills. Inactivity and unemployment So far we have concentrated on the characteristics of those in work. Tables 3A and 3B report instead of information on the characteristics of those out of work and school. We have already documented the levels of joblessness (that is, those out of both work and school) in tables 1A and 1B. The first column of table 3A reports the proportion of active men, according to the International Labour Organization (ILO) definition. This includes those working plus the strictly unemployed, which includes individuals who are out of work but are willing to take a job if offered one and who have taken some steps in the last week to look for work. Not surprising, activity rates increase with age and are lowest in Dar Es 16

Salaam and highest in rural areas among teenagers and youths, paralleling patterns of employment. Activity rates among teenagers increase from 39% in Dar Es Salaam to 77% in rural areas. Participation rates for youths vary between 79% in Dar Es Salaam and 93% in rural areas. For prime-age men, participation is almost universal, with the proportion of inactive individuals varying between 1% in Dar Es Salaam to 3% in rural and other urban areas. Unemployment rates the measure generally used to ascertain joblessness in developed countries are reported in column 2. These are obtained as the ratio between the number of (strictly) unemployed individuals and the number of active individuals, according to the ILO definition. Unemployment rates are remarkably high among teenagers and youth in urban areas. There is virtually no unemployment in rural areas. For teenagers these rates range from 47% in Dar Es Salaam to 13% in other urban areas. Similar patterns emerge among youths: unemployment rates are on the order of 40% in Dar Es Salaam and 11% in other urban areas. Interestingly, unemployment rates are virtually zero among prime-age men. Male unemployment in Tanzania is hence primarily an urban phenomenon disproportionately affecting young workers. 17

Table 3A: Unemployment and Inactivity MALES (1) (2) (3) (4) (5) (6) (7) (8) Active (ILO) Unemployment rate (ILO) Unemploymentto-population ratio Available & no search Not available Long-term unemployment Very longterm unemployment Unemploy ment duration Dar es Salaam Teens 0.386 0.465 0.179 0.027 0.054 0.487 0.265 3.596 Youth 0.790 0.403 0.318 0.020 0.073 0.798 0.488 9.131 Prime-age 0.989 0.015 0.015 0.004 0.007 0.599 0.499 3.392 Urban Teens 0.498 0.126 0.063 0.035 0.082 0.701 0.525 5.025 Youth 0.850 0.106 0.090 0.016 0.077 0.737 0.646 5.660 Prime-age 0.965 0.017 0.016 0.005 0.030 0.493 0.367 1.974 Rural Teens 0.773 0.016 0.012 0.025 0.037 0.245 0.208 1.776 Youth 0.936 0.015 0.014 0.019 0.032 0.265 0.156 1.730 Prime-age 0.974 0.007 0.007 0.007 0.018 0.334 0.237 1.677 Note: The table reports the characteristics of those out of work. Columns 1 reports the proportion active (in work or strictly unemployed). Column 2 reports the proportion strictly unemployed conditional on being in the labor force. Column 3 reports the proportion unemployed as a percentage of the population in that age group. Column 4 reports the proportion of those available to accept a job if offered one but not looking for work (as a proportion of the population). Column 5 reports the proportion unavailable to take a job if offered one (as a proportion of the population). Columns 6 and 7 report the proportion in long-term unemployment (one year or more) or very long-term unemployment (two years or more). Column 8 provides an estimate of in unemployment duration. See text for details. See also notes to table 1A. 18

Table 3B: Unemployment and Inactivity FEMALES (1) (2) (3) (4) (5) (6) (7) (8) Active (ILO) Unemployment rate (ILO) Unemployment -to-population ratio Available & no search Not available Long-term unemployment Very long-term unemployment Unemploy ment duration Dar es Salaam Teens 0.466 0.431 0.201 0.061 0.070 0.599 0.373 4.715 Youth 0.679 0.449 0.305 0.091 0.185 0.619 0.432 5.894 Prime-age 0.778 0.109 0.085 0.056 0.166 0.498 0.445 2.805 Urban Teens 0.472 0.057 0.027 0.118 0.138 0.663 0.563 3.514 Youth 0.750 0.104 0.078 0.083 0.143 0.784 0.634 5.430 Prime-age 0.906 0.015 0.014 0.041 0.053 0.838 0.616 9.271 Rural Teens 0.767 0.007 0.005 0.031 0.038 0.266 0.222 1.634 Youth 0.935 0.012 0.011 0.021 0.039 0.310 0.190 1.579 Prime-age 0.954 0.004 0.003 0.007 0.039 0.224 0.189 1.427 Note: See notes to table 3A. 19

One might wonder whether these data provide a good indication of the extent of joblessness among different age groups. To the extent that fewer youths are active (recall that active individuals do not include those in school), one might suspect that this mechanically inflates the prevalence of unemployment among this group. In particular, if those out of school are those with a lower probability of finding a job, these figures overestimate the extent of unemployment for a random individual in that age group. By contrast, if those who have no job opportunities stay in school or declare themselves inactive, this leads to the opposite bias. An alternative measure of joblessness relates the number of unemployed to the entire population, abstracting from whether these individuals are in school or not (or even active or not). The unemployment-to-population ratio is reported in column 3. Interestingly, even if one standardized unemployment to a much larger population at risk (that is, the entire population in that age group), unemployment prevalence is still higher among youths and teenagers. Figures for prime-age individuals are virtually identical to the ones in column 2, given that, as shown in column 1, essentially all prime-age men participate in the labor market. Although we have so far concentrated on unemployment, it is important to realize that not all of those out of work (or school) are strictly unemployed. Columns 4 and 5 report the proportion of individuals who are available to take a job if offered one and have not looked for work in the week preceding the survey and those who self-declare unavailable respectively. These are groups with increasingly lower labor market attachment. The last group includes truly idle individuals. The proportion of available individuals who declare not having searched over the previous week is rather small at all ages and in all areas. This labor market status is most prevalent among teenagers and on the order of 3% of the population in all areas. Figures on inactivity show, rather worryingly, that 5% of teenagers and 7% of youths in Dar Es Salaam are idle. The proportions in other urban areas are 8% and 7% respectively. In rural areas 4% of teenagers and 3% of youths are inactive. These figures make the extent of joblessness even more worrying among teenagers and perhaps suggest that unemployment ratios might understate the extent of the problem. 20

The figures above show that urban teenagers and youths are at very high risk of being unemployed. However these data are unable to tell us whether these individuals circle in and out of unemployment or whether they get stuck in unemployment for long periods of time. Columns 5 and 6 report the proportion of long-term unemployed among those available to take a job (whether searching or not). Column 5 reports the proportion of those with at least one year of job search. Column 6 shows the proportion of those with at least two years of job search. These are sometimes labeled very long-term unemployed. In urban areas, where unemployment is more likely to occur, long-term unemployed accounts for about 50% or more of the overall unemployment pool. What is remarkable is that long-term unemployment is particularly widespread among youths. This is a major difference with respect to European countries in the 1990s, where long-term unemployment is thought to have been a problem largely for older individuals (Machin and Manning 1999). More than 70% of unemployed youths in urban areas are long-term unemployed. This contrasts with a proportion of longterm unemployed on the order of 37 50% for prime-age men. In rural areas, where unemployment is almost nonexistent, the data show that the few who declare being unemployed transition quite rapidly through this state. The proportion of long-term unemployment varies between 16% and 24% depending on age among rural men. Column 7 reports the average duration of unemployment as estimated from data on inflows and unemployment prevalence. 8 Average unemployment duration is remarkably high in urban areas, especially for youths, consistent with the observation that these individuals have a disproportionate risk of being long-term unemployed. Average duration in Dar Es Salaam varies between 3.6 years for teenagers and 9.1 years for youths. In other urban areas these figures are respectively 5 and 5.7. Notice that unemployment duration is also higher in Dar Es Salaam for prime-age men (3.4 years compared with 2 years in other urban areas) but still much lower than for teenagers and youths. 8 Notice that cross-sectional data including these typically report information on unemployment duration for the unemployed only. Hence the duration of unemployment spells is right-censored. The simple average of these durations will tend to underestimate and potentially by a large amount actual unemployment duration. To derive duration here we use the following identity that holds in steady state u=i/(i+h), where u is the unemployment rate, i is the inflow rate, and h is the outflow rate (Machin and Manning 1999). In steady state average duration equals the reciprocal of the outflow rate (that is, 1/h). To obtain these figures, we have computed unemployment rates as the proportion of individuals available (that is, strict and nonstrict unemployed) over the population in work or school or available. We compute inflow rates as the number of individuals with at most three months of unemployment standardized to the sum of those in work and school. 21

The question naturally arises as to why so many young individuals are inactive or not looking for a job. In table 4A we report the subjective reasons provided by these individuals for not looking or for not being available tout court. As columns 1 to 6 show, a large proportion of individuals answers that they are not looking for a job due to the poor expectation of finding one. This proportion is particularly high for teenagers and youths in urban areas. For example, in Dar Es Salaam 66% of individuals report this as the main reason for not looking. This compares with 21% among youths (and zero among prime-age men). A relevant share of those not looking reports to be waiting for a reply to a job application or waiting for a job to start. No systematic patterns can be detected through areas or age groups. Only in urban areas is there a large proportion of inactive youths and teenagers (around 30%) declaring not to be looking due to their involvement in home duties. Columns 7 9 report the reasons provided by those not available. Between 20% and 30% of these inactive teenagers report being involved in household chores. The relative importance of this explanation falls as individuals age. Rather interestingly, between 11% and 46% of teenagers, depending on the area, report being inactive due to sickness or disability. This accounts for an even greater proportion of inactive youths. These worrisome figures are most likely the result of the HIV/AIDS epidemic. Although there is no way to ascertain this with the data at hand, evidence from other sources suggests that youths are at the highest risk of contracting HIV/AIDS in Tanzania, one of the countries with the highest prevalence in Sub- Saharan Africa. 9 The residual category other accounts for a large share of the inactive. Overall it appears that inactivity hides some productive employment in the household, leading to an overestimate of the true extent of joblessness among young individuals. It is unclear, though, whether the category other includes individuals who have stopped being available due to poor labor market prospects. 9 Recent estimates indicate HIV prevalence of 8.8% among the population aged 14 49 (UNDP 2005), above the average of Sub-Saharan Africa (7.3%). Infection rates for young adults in the age groups 20 24, 25 29, and 30 34 range from 5.9% to 7.9% among males, and from 9.3% to 10.1% among females, who are affected at earlier ages than males (Tanzania Commission for AIDS, www.tanzania.go.tz/ hiv_aids.html#prevalence%20of%20hiv%20infection). 22

Thought would not find Table 4A: Reasons for not Looking or Inactivity MALES (1) (2) (3) (4) (5) (6) (7) (8) (9) Reason not look Reason not available Waiting for Offseason Household Temporarily ill Other Household Sick Other job or reply duties duties Dar es Salaam Teens 64.48 13.58 0.00 3.16 3.60 15.18 28.03 11.26 60.71 Youth 25.55 32.48 0.00 2.68 0.00 39.30 18.71 24.71 56.57 Prime-age 0.00 100.00 0.00 0.00 0.00 0.00 0.00 100.00 0.00 Urban Teens 28.28 48.83 0.00 7.53 1.48 13.88 25.83 31.66 42.51 Youth 51.83 21.84 0.00 18.02 0.00 8.31 16.85 15.74 67.41 Prime-age 16.20 7.01 38.32 0.00 0.00 38.47 12.56 83.41 4.03 Rural Teens 17.43 22.11 8.25 30.70 1.32 20.18 22.67 45.91 31.42 Youth 19.92 27.88 8.68 37.59 0.00 5.93 10.49 58.62 30.89 Prime-age 33.80 39.35 17.72 9.13 0.00 0.00 3.17 92.11 4.73 Note: The table reports the characteristics of those out of work who report not looking for a job (columns 1 6) or those unavailable to take a job if offered one (columns 7 9). In each column the table reports the distribution of the main self-reported reason for not looking or being unavailable. See also notes to table 1A. 23

For women, several interesting patterns emerge. As column 1 of Table 3B shows, activity rates are higher among teenage women than teenage men in Dar Es Salaam: here the activity rate for women is 47% (8 percentage points higher than for men). In rural areas there is no substantial difference between teenage boys and girls, with a teenage female activity rate on the order of 77% (the same as men's). In other urban areas girls are slightly less likely to be active than men (47% compared with an activity rate for teenager men of 50%). Although, women s activity rates increase with age, the rise is less pronounced than for men, and hence men overtake women by the time they reach prime age. Participation rates for prime-age women are on the order of 95% in rural areas (2 percentage points less than men) and 78% in Dar Es Salaam (20 percentage points less than men). Although activity rates are lower for women than for men, a higher or equal proportion of active women are unemployed at least in Dar Es Salaam, as shown in column 2. In Dar Es Salaam unemployment rates are 43% for female teenagers (3 percentage points less than men), 45% for female youths (5 percentage points more than men), and 11% among prime-age women (10 percentage points more than men). In other urban areas women are less likely to be unemployed than men when in their teens (6% compared with 13%), but women are equally likely to be unemployed when they are in between the ages of 20 and 24 (10% compared with 11%). As with men, there is no unemployment in rural areas. Urban unemployment hence is a similar problem for young males and females. In addition, prime-age women are also at high risk of unemployment, at least in Dar Es Salaam. An analysis of columns 4 and 5 shows that a much higher proportion of women than men are available but not looking or unavailable. For example, in Dar Es Salaam the proportion of female youths available but not searching is 9% (compared with 2% for men), and the proportion of idle youths is 18% (compared with 7% for men). Idleness rates rise from teenagers to youth and then stay constant (at around 17%) in urban areas. By contrast, this rate increases and then decreases in other urban areas. Most likely, women here re-enter the labor market as they are relieved from major childrearing responsibilities. Idleness is low and stable in rural areas. When one moves to analyzing long-term unemployment and unemployment duration, it turns out that young women display generally shorter durations than young men. This is 24