_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ t o s -4 l/4 2' POLICY RESEARCH WORKING PAPER 2149 Income Gains to the Poor A workfare program was introduced in response to from Workfare high unemployment in Argentina An ex-post Estimates for Argentinas Xevaluation using matching methods indicates that the Trabajar Program program generated sizable net income gains to generally Jyotsna jalan poor participants Alartin Ravallon The World Bank Development Research Group Poverty and Human Resources July 1999 0 POLICY RESEARCH WORKING PAPER 2149 Summary findings Jalan and Ravallion use propensity-score matching percent of them are in the poorest quintile - reflecting methods to estimate the net income gains to families of the self-targeting feature of the program design. workers participating in an Argentinian workfare Average gains for men and women are similar, but program. The methods they propose are feasible for gains are higher for younger workers. evaluating safety net interventions in settings in which Women's greater participation would not enhance many other methods are not feasible. The average gain is average income gains, and the distribution of gains about half the gross wage. would worsen. Even allowing for forgone income, the distribution of Greater participation by the young would raise average gains is decidedly pro-poor. More than half the gains but would also worsen the distribution. beneficiaries are in the poorest decile nationally and 80 This paper - a product of Poverty and Human Resources, Development Research Group - is part of a larger effort in the group to improve methods for evaluating the poverty impact of Bank-supported programs. Copies of the paper are available free from the World Bank, 1818 H Street NW, Washington, DC 20433. Please contact Patricia Sader, telephone 202-473- 3902, fax 202-522-1153, Internet address psader@worldbank.org. Policy Research Working Papers are also posted on the Web at http ://www.worldbank.org/html/dec/Publications/Workpapers/home.html. The authors may be contacted at jjalan@isid.ac.in or mravallion@worldbank.org. July 1999. (32 pages) The Policy Research Working Paper Series disseminates the finiings 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. Produced by the Policy Research Dissemination Center Income Gains to the Poor from Workfare: Estimates for Argentina's Trabajar Program Jyotsna Jalan and Martin Ravallion' Indian Statistical Institute and World Bank 1 The work reported in this paper is one element of the ex-post evaluation of the World Bank's Social Protection II Project in Argentina. The support of the Bank's Research Committee (under RPO 681-39) is gratefully acknowledged. The paper draws on data provided by the SIEMPRO unit of the Ministry of Social Development, Government of Argentina. The authors are especially grateful to Joon Hee Bang and Liliana Danilovich of SIEMPRO for their help with the data. The authors' thanks also go to staff of the Trabajar project office in the Ministry of Labor, Government of Argentina who provided the necessary data on their program and gave this evaluation their full support. Petra Todd kindly advised us on matching methods. Useful comments were received from Polly Jones, Dominique van de Walle, and seminar participants at the World Bank, the Indian Statistical Institute, Delhi, and the Institute of Fiscal Studies, London. 1. Introduction Workfare programs require that participants must work to obtain benefits.2 They are often turned to in crises such as due to macroeconomic or agro-climatic shocks, in which a large number of poor able-bodied people have become unemployed. Typically, the main aim of workfare is to raise the current incomes of poor families hurt by the crisis. To assess the impact of such a program, we need to measure the income gain conditional on income in the absence of the program. The income gain is the difference between household income with the program and that without it. The "with" data can be collected without great difficulty. But the "without" data are fundamentally unobserved, since an individual cannot be both a participant and a non-participant of the same program. Common practice has been to estimate the gains by the gross wages paid.3 In other words, the unobserved income without the program is taken to be equal to income with the program, minus wages received. This assumption would be a reasonable one if labor supply to a workfare program came only from the unemployed. But that is difficult to accept. Even if a participating worker was unemployed at the time she joined the program, that does not mean that she would have remained unemployed had the program not existed. Even a worker who has been unemployed for some time will typically face a positive probability of finding extra work during a period of search, including self-employment in an informal sector activity. Joining the program will leave less time for search. There are also ways in which behavioral responses help reduce foregone income. There are likely to be effects on time allocation within the household. For example, 2 On the arguments and evidence on this class of interventions see Ravallion (1991, 1999a), Besley and Coate (1992), Lipton and Ravallion (1995), Mukherjee (1997), and Subbarao (1997). 3 See, for example, the various assessments of the cost-effectiveness of workfare programs reviewed in Subbarao et al., (1997). 2 Datt and Ravallion (1994) find that other family members took up the displaced productive activities when someone joined a workfare program in rural India. Such behavioral responses will reduce foregone income, though we can still expect it to be positive. This paper estimates the income gains from a workfare program and how those gains vary with pre-intervention incomes. We study the Trabajar Program instituted by the Government of Argentina, and supported by a World Bank loan and technical assistance. We use propensity-score matching methods (Rosenbaum and Rubin, 1983, 1985; Heckman et al., 1997, 1998) to draw a comparison group to workfare participants from a larger survey of non- participants. A number of features of this setting lend themselves to matching methods. It was possible to assure that the same questionnaire was administered to both the participants and the non-participants, and that both groups were from the same economic environment. The Trabajar participant could be identified in the larger survey.4 Furthermore, using kernel density estimation techniques, we are able to ensure that participants are matched with the non- participants over a common region of the matching variables. Any remaining bias in the matching estimator can thus be attributed to unobserved characteristics. The design of the program can be expected to entail considerable rationing of participation according to observables; the sample of non-participants is very likely to include people who wanted to participate but were unable to do so due to say non-availability of the program. While our application is well suited to matching methods, bias due to unobservables cannot be ruled out by matching alone, since the method is basecl solely on observables. So we also propose and implement a test for any remaining selectivity bias after matching. 4 The researcher may not be able to identify whether an individual participated in the program or not in the larger population sample. In such cases, one can still go ahead with the matching procedure though this adds a "contamination bias" to the impact estimator. In our application this is not an issue. 3 The following section discusses the evaluation problem and our methods. Section 3 describes the Trabajar program. Our data are described in Section 4. Section 5 presents the results, and offers an economic interpretation. Section 6 concludes. 2. Estimating the Income Gains from Workfare In assessing the gains from a workfare program, the workers' earnings are naturally the main focus, and that will be the case here. However, it should be noted that earnings net of foregone income are only one of the potential benefits. There could also be risk benefits from knowing that the program exists. There may well also be benefits from the outputs, depending on (amongst other things) how well targeted the workfare projects are to poor areas.5 We first outline what we see as the model of self-targeting underlying arguments for workfare, pointing to the key role played by foregone incomes. We then describe the matching method we use to estimate foregone incomes. 2.1 The Problem The following rudimentary model has the essential features necessary to characterize the "self-targeting" argument often made in favor of workfare (Ravallion, 1991). The model assumes that foregone income from accepting a workfare job is F(Y), a smoothly increasing function of pre-intervention income Y(scaled to lie between zero and one). Foregone income increases with pre-intervention income due to differences in education, experience and so on that are naturally correlated with both earnings and family income. The workfare program offers a wage W, with F(O)F-'(W)). The schedule of gains is G=W-F(Y) for YF-'(W), yielding post-intervention incomes Y+G. In this simple model, underestimating the foregone income will lead the evaluator to overestimate the impact on poverty. To see why, suppose that, in assessing the gains from the program, we use a biased estimate of foregone income, namely O(A$l) from the population. The second set of data might be the national population census or an annual national household budget survey that has information relevant in the participation decisions of the individuals. Using the two sets of data, we try to match the N program participants with a comparison group of non-participants from the population. The two surveys must include infonnation that helps predict participation in the program. LetXbe the vector of such variables. Ideally, one would match a participant with a non- participant using the entire dimension of X, i.e., a match is only declared if there are two individuals, one in each of the two samples, for whom the value of Xis identical. This is impractical, however, because the dimensicin of X could be very high. Rosenbaum and Rubin (1983) show that matching can be performed conditioning on P(A) alone rather than onX, where P(X) = Prob(D=I1 X) is the probability of participating conditional onX, the "propensity score" of X. If outcomes without the intervention are independent of participation given Xthen they are 7 also independent of participation given P(X). This is a powerful result, since it reduces a potentially high-dimensional matching problem to a single dimensional problem. The propensity score is calculated for each observation in the participant and the comparison-group samples using standard logit models.7 Choice-based sampling methods suggested by Manski and Lerman (1978) can be used to weight the observations given that there is over-sampling of participants. In our case however, we do not know the sampling weights to do the choice-based sample re-weighting. But we can still carry out the matching using the odds ratio pi = P/( 1 -Pi) where Pi is the estimated probability of participation for individual i. Using the propensity score, one constructs matched-pairs on the basis of how close the scores are across the two samples. The nearest neighbor to the i'th participant is defined as the non-participant that minimizes [p(Xj)-p(X) )]2 over allj in the set of non-participants, where p(Xk) is the predicted odds ratio for observation k. In their comparisons of non-experimental methods of evaluating a training program with a benchmark experimental design, Heckman et. al (1997, 1998) find that failure to compare participants and controls at common values of matching variables is the single most important source of bias - considerably more important than the classic econometric problem of selection bias due to differences in unobervables. To ensure that we are matching only over common values of the propensity scores, we estimated the density of the scores for the non-participants at 100 points over the range of scores. We use a biweight kernel density estimator and the optimal bandwidth value suggested by Silverman (1986). Once we estimate the density for the non- ' One could use semi- and non-parametric methods to estimate the propensity scores though Todd (1995) argues that such methods do not make any difference to the impact estimator. Thus for computational simplicity, we use standard parametric likelihood methods to compute the estimated propensity scores. 8 participants, we exclude those non-participEnts for whom the estimated density is equal to zero. We also exclude 2% of the sample from the top and bottom of the non-participant distribution. The mean impact estimator of the program is given by: P NP II Yj E W9jYV() /P (1) j=p i=l where Yjl is the post-intervention household income of participantj, Yyo is the household income of the ith non-participant matched to thejth participant, P is the total number of participants, NP the total number of non-participants and the Wy's are the weights applied in calculating the average income of the matched non-participants. There are several different types of parametric and non-parametric weights that one can use. In this paper we use three different weights and thereby report three different matching estimators. Our first matching estimator is the "nearest neighbor" estimator where we find the closest non-participant match for each participant and the impact estimator is a simple mean over the income difference between the participant and its matched non-participant.8'9 Our second estimator takes the average income of the closest five matched non-participants and compares this to the participant's income. We also report a kernel- weighted estimator where the weight are given by: P Wij = Kuj 1 K,0 (2) j=l where K= K[(P(X,) - .P(Xj )) / a,0 ] (3) E K[(P(Xi ) -P(Xj ))/aNo ] j=1 s The closest match is chosen by the distance metric discussed above. Also we allow for replacement of the non-participants, so a non-participant could be the closest match for more than one participant. 9 and where aNo is the bandwidth parameter, K(.) is the kernel as a function of the difference in the propensity scores of the participants and the non-participants. In our analysis, we have used Silverman's (1986) optimal bandwidth parameter and a biweight kernel function. (The results were very similar using either a rectangular or parzen kernel function.) Lastly, in each of these cases, the associated standard errors of the mean impact estimator are also calculated. We calculated both the parametric and bootstrapped standard errors for the impact estimators. The two were virtually identical. We report the parametric standard errors in the paper. (The bootstrapped standard errors are available from the authors on request.) 2.3 Testingfor Bias due to Unobservables The matching estimate described above will give a biased estimate of the income gains from workfare if there are unobserved variables that jointly influence incomes and workfare participation, conditional on the observed variables in the data used for matching. A natural test for such a bias is look for a partial correlation between incomes and the residuals from the participation model (used to construct the propensity scores) controlling for actual participation. We call this the test for selection bias in the matching estimator. It is a straightforward application of the standard Sargan-Wu-Hausman test. There will, of course, also be heterogeneity in other characteristics relevant to incomes. By performing the test for selection bias on a sample combining the participants and their matched non-participants we will have already eliminated some of this heterogeneity. One can also explicitly introduce a vector of control variables (Z) to give a test equation for income Y of the form: Y1 = a + ±/P + vRi + bZ6 + v, (4) In calculating our mean impact, if the income of the participant is less than the income of the matched non-participant, we treat the impact to be zero rather than the observed negative number. 10 for household i in the combined participant and matched control sample of nearest neighbors, where Ri denotes the residuals from the paiticipation model. Selection bias is indicated if we can reject the null that y=O. In a linear model, identification requires the usual condition that there is at least one variable in X that is not in Z. The non-linearity of the propensity scores in X means that this condition is not essential. 1Iowever, specifics of the program's design (discussed below) will provide a seemingly plausible exclusion restriction allowing identification without relying solely on the non-linearity. We can use this method to test for selection bias only in the nearest neighbor case where there is one matched non-participant for each participant. In the other two cases, it does not appear to be possible to use a regression to replicate the complex weighting of the data on non- participants used in forming the matching estimator. 3. The Trabajar Program Argentina experienced a sharp increase in unemployment in mid 1 990s, reaching 18% in 1996/97. This was clearly hurting the poor; for example, the unemployment rate (on a comparable basis) was 39% amongst the poorest decile in terms of household income per capita in Greater Buenos Aires. Unemployment rates fell steadily as income per person increases. '0 With financial and technical suppoIt from the World Bank, the Government of Argentina introduced the Trabajar II program in May 1997. This is a greatly expanded and reformed version of a previous program, Trabajar I. The program aimed to help in two ways. Firstly, by providing short-term work at relatively low wages, the program aimed to self-select unemployed workers from poor families. Secondly, the scheme tried to locate socially useful projects in poor 10 These data are from the Permanent Household Survey (EPH) for Greater Buenos Aires in May 1996. 11 areas to help repair and develop local infrastructure. This paper only assesses progress against the first objective (on the second see Ravallion, 1999b). The national program budget is allocated across provinces by the center, leaving the provincial governments with considerable power to determine how the moneys are allocated within the province. There is evidence of horizontal inequality in the outcomes of this process, in that equally poor local areas ("departments") obtained very different allocations in expectation from the program depending on which province they belong (Ravallion, 1999b). This decentralized nature of the program is the basis of our identification strategy in testing for selectivity bias. Following Ravallion and Wodon (1998) we use province dummy variables as the instruments. Clearly the province of residence matters to participation given the way program funds are allocated. We then assume that province of residence does not matter to incomes independently of participation. Given that we will include a wide range of local geographic control variables in the income regression this assumption is defensible. The projects are proposed by local governmental and non-governmental organizations who must cover the non-wage costs. The projects have to be viable by a range of criteria, and are given priority according to ex ante assessments of how well targeted they are to poor areas, what benefits they are likely to bring to the local community, and how much the area has already received from the program. Workers cannot join the program unless they are recruited to a project proposal that is accepted on the basis of these criteria. The process of proposing suitable sub-projects is thus key to worker participation in the program. There are other factors. The workers cannot be receiving unemployment benefits or be participating in any other employment or training program. It is unlikely that a temporary employment program such as this would affect residential location, though workers can commute. 12 The wage rate is set at a maximum of $200 per month. This was chosen to be low enough to assure good targeting performance, and to help assure workers would take up regular work when it became available. To help locate the Trabajar wage in the overall distribution of wages we examined earnings of the poorest 10% of households (ranked by total income per person) in Greater Buenos Aires (GBA) in the May 1996 Permanent Household Survey. For this group, the average monthly earnings for the principal job (when this entails at least 35 hours of work per week) in May 1996 was $263.11 (As expected, the poorest decile also received the lowest average wage, and average wages rose monotonically with household income per person.) So the Trabajar wage is clearly at the low end of the earnings distribution. There are two further concerns about the project that the evaluation can throw light on. One concern has been the low level of female participation; only 15% of participating workers in the first six months were female. The key iquestion is why. If it is because women choose not to participate then one would be less concerned than if it arose from impediments to their participation due to discrimination in allocating Trabajar jobs. If there is such a gender bias then there will be unexploited welfare gains from higher female participation. We cannot measure the welfare gain, but we can determine whether the net income gain is higher for women then men, implying an income loss from low female participation. Secondly, while Trabajar I had been targeted to middle aged heads of households, it was decided not to impose this restriction on the new program since it risked adding to the forgone income of participants by constraining their ability to adjust work allocation within the household in response to the program. However, the past practice under Trabajar I may still I l This includes domestic servants. This is ani unusual labor-market group, given that they often have extra income-in-kind. If one excludes them, the figure is $334. 13 have influenced local implementation of the new program. Then one might expect to find that there are unexploited income gains by increasing participation by the young. We will test this. 4. Data Two household surveys are used. One is of program participants and the other is a national sample survey, used to obtain the comparison group. Both surveys were done by the government's statistics office, the Instituto Nacional De Estadistica Y Census (INDEC), using the same questionnaire, the same interviewing teams and at approximately the same time. The national survey is the Encuesta de Desarrollo Social (EDS), a large socio-economic survey done in mid-1997. The EDS sample covers the population residing in localities with 5,000 or more residents. The comparison group is constructed from the EDS. According to the 1991 census, such localities totaled to 420 in Argentina and represented 96% of the urban population and 84% of the total population. 114 localities were sampled. The second data set is a special purpose sample of Trabajar participants done for the purpose of this evaluation. The sample design involved a number of steps. First among all the projects approved between April and June of 1997, 300 projects in localities which were in the EDS sample frame were randomly selected, with an additional 50 projects chosen for replacement purposes. The administrative records on project participants did not include addresses, so Ministry of Labor (MOL) had to obtain these by field work. From these 350 projects, the Labor Ministry could find the addresses of nearly 4,500 participants. However, for various reasons about 1,000 of these were not interviewed. The reasons given by INDEC were that the addresses were found to be outside the sample frame, or they were incomplete, or even non-existent, or that all household members were absent when the interviewer went to interview 14 the household, or that they did not want to respond. In all 3,500 participant households were surveyed. (The number of Trabajar participants during May 1997-January 1998 was 65,321.) We restrict the analysis to households with complete income information, and those who completed all the questions asked of them. Also, we only consider participants who were actually working in a Trabajar project at the time they were surveyed. Since the EDS questionnaire does not ask income-related questions to those below 15 or above 64 years of age, we also had to restrict our attention to the age group 15-64 years for our analysis. We focus on current Trabajar participants in the reference week, fixed at the first week of September 1997, who received wages from the Trabajar program during August 1997. 80% of the Trabajar sample had current participants by this definition.12 With these restrictions, the total number of active participants that we have used is 2,802. 5. Results and Interpretation 5.1 Descriptive Statistics In Table 1, we present selected descriptive statistics for the Trabajar and EDS samples. The Trabajar sample has lower average income, higher average family size, are more likely to have borrowed to meet their basic needs, receive less from informal sources, are more likely to participate in some form of political organlization, and less likely to own various consumer durables (with the exception of a color TVJ, which appears to be a necessity of life in Argentina.) Table 2 gives the percentage distribution of Trabajar participants' families across deciles formed from the EDS with households ranked by income per capita, excluding income from Trabajar. (The poorest decile is split in half.) This is the type of tabulation that is typically made 12 The remaining 20% of the participants are assumed to be beneficiaries of the program that had left work by August 1st, 1997 (i.e. at the start of the survey), or had not yet started the Trabajar job. 15 in assessing such a program. It assumes zero foregone income, so each participating family's pre-intervention income is simply actual income minus wage earnings from the program. Table 2 suggests that a high proportion of the families of participants come from poor families.13 40% of the program participants have a household income per capita which puts them in the poorest 5% of the national population; 60% of participants are drawn from the poorest 10% nationally. By most methods of measuring poverty in Argentina, the poverty rate is about 20%. So 75-85% of the participants are poor by this standard. Such targeting performance is very good by international standards. Does relaxing the assumption of zero foregone income change the results in Table 1? Using the matching methods described above, we will now see whether that assumption is justified, and how much it matters to an assessment of average gains and their incidence. 5.2 Propensity-Score Matching Estimates Table 3 presents the logit regression used to estimate the propensity scores on the basis of which the matching is subsequently done. The results accord well with expectations from the simple averages in Table 1. Trabajar participants are clearly poorer, as indicated by their housing, neighborhood, schooling, and their subjective perceptions of welfare and expected future prospects (relative to their parents). The participation regression suggests that program participants are more likely to be males who are head of households and married. Participants are likely to be longer-term residents of the locality rather than migrants from other areas. The model also predicts that (controlling for other characteristics) Trabajar participants are more 13 We have only considered participants who have earned at least $150 from participating in the Trabajar job while calculating the impact of the project. The minimum wage offered under Trabajar is $200; those reporting less than $150 as their Trabajar earnings imply that these participants are in their last phase of their Trabajar job and or have misreported their income. Since we are interested in the impact on currently active participants in the program, we excluded the observations to get a better albeit a more conservative measure of the impact. 16 likely to be members of political parties and neighborhood associations. This is not surprising given the design of the program, since social and political connections will no doubt influence the likelihood of being recruited into a succ;essful sub-project proposal. However, participation rates in political parties and local groups are still low, even for Trabajar participants (Table 1). Based on Table 3, the mean propensity score for the national sample is 0.075 (with a standard deviation of 0.125). This is of couirse much lower than the mean score for the Trabajar sample, which is of 0.405 (0.266). However, after following the matching method outlined in section 2.2, the comparison group of nearest neighbors drawn from the national sample has a mean score of 0.394 (0.253), very close to that of the Trabajar sample. Tables 4 and 5 give our estimates of average income gains, and their incidence according to pre-intervention incomes. For this purpose we have first estimated the gain for each participating household, by each of the thre e methods described in section 2.2. We then assign each household to a decile using the same decile bounds calibrated from the EDS, but this time the participants are assigned to the decile implied by their estimated pre-intervention income as given by actual income minus the estimated net gain. The nearest neighbor estimate of the average gain is $157, about three-quarters of the Trabajar wage. The nearest five and non-parametric estimator give appreciably lower gains, of around $100. Since the latter estimates use more information and are presumably more robust we will use them in preference to the nearest neighbor estimate. For computational convenience and to circumvent the small sample problem in the sub-group cases, the rest of the paper is based on the "nearest five" rather than the non-parametric estimate. The average gain using the "nearest five" estimator of $103 is about half of the average Trabajar wage. Given that there is sizable foregone income, the crude incidence numbers in Table 1 overestimate how pro-poor the program is, since pre-intervention income is lower than is 17 implied by the net gains. Where this bias is most notable is amongst the poorest 5%; while the non-behavioral incidence analysis suggests that 40% of participant households are in the poorest 5%, the estimate factoring in foregone incomes is much lower at 10%. Nonetheless, over half of the participant households are in the poorest decile nationally even allowing for foregone incomes. Given that the poverty rate in Argentina is widely reckoned to be 20%, our results suggest that four out of five Trabaj ar participants are poor by Argentinean standards. Figure 1 gives mean income gain at each level of pre-intervention income, estimated by a locally-weighted smoothed scatter plot of the data. Gains fall sharply (though not continuously) up to an income of about $200 per person per month (which is about the median of the national distribution), and are roughly constant after that. We will return later to interpret this finding. Figure 2 gives the empirical cumulative distribution functions (CDFs) implied by our results. We give the CDFs for both the Trabajar participants and the national distribution. We also give the counter-factual (pre-intervention) CDF for the Trabajar participants. There is a spike of zero incomes in the national sample, much of which is probably measurement error. If one takes this spike out there is first-order dominance comparing the Trabajar samples and the national samples, with higher poverty in the Trabajar sample at all possible poverty lines. There is automatically first-order dominance of the post-intervention incomes for the Trabajar sample given that we have ruled out negative gains on a priori grounds. The absolute gains are highest for the third decile, but do not vary greatly across the deciles containing participants. The percentage net gains are highest for the poorest, reaching 74% for the poorest 5%. In section 5.3 we will offer an interpretation of these findings. Tables 7-9 report the net wage gains for three different demographic groups: female participants, participants between the ages 15-24 years (typically identified as those who are new entrants into the job market), and workers in the age group 25-64 years. 18 The estimates in Table 7 are not consistent with the existence of income losses due to a gender bias in the program. The net wage gains from the program accruing to female participants are virtually identical to the gains for male participants. However, the distribution of female participation is less pro-poor, as indicated by household income per capita; while over half of the members of participating families are in the poorest decile nationally, this is true of less than 40% of the members of female participants' families. This probably reflects lower wages for women in other work, making the Trabajar wage more attractive to the non-poor. For the younger cohort however, the net gains are significantly higher (comparing Tables 8 and 9). Foregone incomes are lower for the young, probably reflecting their lack of experience in the labor market. Because of this, there wculd be income gains from higher participation by the young. (To the extent that any young participants leave school to join the program, future incomes may suffer.) This suggests that the older workers may well be favored in rationing Trabajar jobs. However, the distribution of gains is more pro-poor for the older workers, with almost 60% coming from the poorest decile. Pushing for higher participation by the young entails a short-term trade-off between average gains and a better distribution. It may also entail a longer-term trade off with future incomes of the young, by reducing schooling. Finally we test whether our impact estimator is biased due to selection on unobservables. For identification, we exclude the province dummy variables from the set of controls in the income regression, as discussed in section 3. The regression coefficient on participation (, in equation 4) was 154.358 (t=5.049) which is very close to the matching estimate for the nearest neighbor case. 14 The coefficient on the residuals from the participation regression (y in equation 14 Of course, if one drops the control variables and the participation residuals then the estimate is identical to that based on the mean differences between the participants and their nearest neighbors. 19 4) was 4.064 but this was not significantly different from zero (t=0.402). Evidently, selection bias on unobservables is not an important concern in our matching estimates. 5.3 Economic Interpretation Although we find that program participation falls off sharply as household income rises, the net gains conditional on participation do not fall amongst the upper half of the income distribution (Figure 1). Since the program wage rate is about the same for all participants, foregone income amongst participants appears to be independent of family income above about $200 per person per month. This may be surprising at first sight. The standard model of self- targeting through work requirements postulates that foregone income tends to be higher for higher income groups (section 2.1). We can offer the following explanation. The Trabajar wage is almost certainly too low to attract a worker out of a regular job. For a worker with such a job, let the foregone income from joining the program befe(Y)> W where (as in section 2.1) Y is the pre-intervention income of the worker's household, Wis the wage rate offered on the Trabajar program and the functionfe is strictly increasing. For an unemployed worker, however, only miscellaneous odd-jobs are available. Anyone can get this work, and it does not earn any more for someone from a well-off family than a poor one. Let this "odd-job" foregone income bef, 1 month -0.4450 -3.182 Born in this locality 0.8215 5.019 ...... in another locality of same province 0.5672 3.373 ....... in another province 0.6523 3.867 Lived habitually in this locality for last 5 years 0.5326 4.876 Affiliated to a health system only through social work -0.6388 -7.750 ....... to a health system through unions and private hospital -0.4694 -3.839 ....... to health system through social work & mutual benefit society -1.0715 -3.291 ....... to health system because he is a worker -1.1213 -6.530 Currently attends an educational establishment for primary/sec school -0.7551 -2.117 Currently a student at tertiary school 0.8775 2.650 Dropped out of school because found syllabus uninteresting -0.5386 -3.656 ..... he/she was finding school difficult 0.6700 3.048 ..... location of school was inconvenient -0.3996 -1.951 Dropped out of school for personal reasons 0.3671 2.100 Taken a course in labor training in the last 3 years 0.4252 5.244 Never a member of a sports association 0.3444 2.826 Regular member of a neighborhood association with some admin. 0.9705 2.482 Responsibilities Regular member of a neighborhood association with no responsibilities 0.8259 2.526 Never a member of union/student association 0.5973 2.413 Member of a political party with some administrative responsibilities 0.7523 1.900 Member of a political party 1.6387 6.020 Occasional member of a political party 1.3609 5.041 Thinks that 20 years hence, economic situation will be the same as parents now 0.3981 5.401 Reason for above is lack of schooling -0.3705 -5.632 Reason for above is economic situation of country -0.7596 -7.291 Thinks that he and his family is very poor 0.5976 6.078 Children born in the last 12 months 0.2281 2.693 Pregnant currently -0.9295 -2.435 Constant -5.6210 -4.390 Log Likelihood -5580 Notes: Only significant coefficients in the logit regression are reported in the above table. For omitted categories and for other variables included in the regression see Addendum (available from the authors). 28 Table 4: Net income gains from the program using different estimators Groups Nearest neighbor Nearest five estimator Non-parametric estimator Full sample 156.770 102.627 91.678 (296.083) (247.433) (230.327) Ventile 1 372.010 108.543 107.862 (409.053) (210.543) (222.831) Ventile 2 132.662 83.351 63.331 (260.851) (200.379) (161.769) Decile2 112.166 119.044 93.506 (230.161) (285.357) (197.679) Decile 3 102.058 136.349 120.430 (176.515) (263.939) (240.703) Decile 4 78.740 82.386 89.295 (248.272) (281.863) (277.294) Decile 5 148.711 107.125 205.050 (434.210) (208.313) (597.605) Decile 6 - 9 80.965 111.229 114.913 (191.337) (278.584) (196.906) Decile 10 No participants in this decile Note: Standard errors in parentheses. Table 5: Persons of participant households using different estimators Groups Nearest neighbor estimator Nearest five estimator Non-parametric estimator Full sample 100.000 100.000 100.000 Ventile 1 21.525 10.207 8.671 Ventile 2 41.278 42.284 39.460 Decile 2 20.732 26.908 27.734 Decile 3 8.084 10.892 13.460 Decile 4 5.403 6.307 7.302 Decile 5 1.842 2.069 1.652 Decile 6 - 9 1.135 1.334 1.722 Decile 10 No participants in this decile 29 Table 6: Net income gains from the program Groups % of Persons of H'hold income Net income Net gain as % of participants in participant of Trabajar gain due to the pre-intervention ventile/decile households participants program income Full sample 100.000 100.000 501.181 102.627 25.926 (364.632) (247.433) Ventile 1 6.070 10.207 299.102 108.543 74.830 (221.119) (210.543) Ventile 2 36.535 42. 284 369.194 83.351 24.746 (265.054) (200.379) Decile 2 26.700 26.908 548.789 119.044 26.566 (353.237) (285.357) Decile 3 12.601 10.892 685.413 136.349 23.056 (358.139) (263.939) Decile 4 11.833 6.307 543.680 82.386 13.483 (441.794) (281.863) Decile 5 3.496 2.069 749.443 107.125 14.975 (384.025) (208.313) Decile 6 - 9 2.766 1.334 879.382 111.229 11.469 (496.091) (278.584) Decile 10 No participants in this decile Notes: These numbers correspond to the nearest five estimator reported in Table 4. Standard errors in parentheses. Table 7: Net income gains for female participants Groups % of Persons of H'hold income Net income Net gain as % of participants in participant of Trabajar gain due to the pre-intervention ventile/decile households participants program income Full sample 100.000 100.000 571.890 103.904 22.818 (382.580) (277.340) Ventile 1 3.289 5.645 351.300 158.240 82.298 (428.177) (409.963) Ventile 2 25.000 31.948 424.370 101.360 30.767 (320.742) (281.681) Decile 2 32.895 34.000 520.800 87.490 18.400 (286.501) (202.641) Decile 3 16.447 15.261 718.660 136.284 21.166 (493.045) (420.507) Decile 4 12.500 8.251 655.579 92.353 14.123 (322.183) (196.851) Decile 5 4.605 2.605 696.143 79.000 12.558 (224.638) (126.926) Decile 6-9 5.263 2.295 963.663 132.663 14.006 (473.150) (248.887) Decile 10 No participants in this decile Notes: These numbers correspond to the nearest five estimator for the sub-group of female participants. Standard errors in parentheses. 30 Table 8: Income gains for those 15-24 years of age Groups % of Persons of H'hold income Net income Net gain as % of participants in participant of Trabajar gain due to the pre-intervention decile households participants program income Full sample 100.000 100.000 618.789 125.241 25.592 (401.990) (255.903) Decile 1 30.214 37.012 434.619 121.500 35.287 (332.660) (261.500) Decile 2 31.567 34.431 636.060 143.657 28.629 (353.555) (272.418) Decile 3 16.234 14.776 738.666 133.560 19.921 (383.006) (275.162) Decile4 11.838 8.313 620.135 73.146 10.559 (378.544) (169.706) Decile 5 10.034 3.618 886.735 152.898 17.400 (422.0520 (262.636) Decile 6 - 9 3.495 1.850 1,069.600 102.142 9.550 (608.221) (176.652) Decile 10 No participants in this decile Notes: These numbers correspond to the nearest five estimator for the sub-group of 15-24 year participants. Standard errors in parentheses. Table 9: Income gains for those 25-64 years of age Groups % of Persons of H'hold income Net income Net gain as % of participants in participant of Trabajar gain due to the pre-intervention decile households participants program income Full sample 100.000 100.000 443.443 85.820 22.241 (328.253) (231.032) Bottom 5% 7.423 13.062 307.386 97.474 38.564 (251.260) (221.489) Next 5% 39.451 45.767 342.499 71.809 22.207 (252.305) (205.962) Decile 2 26.651 24.938 487.939 86.833 21.204 (251.477) (180.047) Decile 3 11.046 8.851 625.097 122.505 25.578 (395.1020 (334.238) Decile 4 10.812 5.046 476.941 74.724 13.594 (410.221) (271.968) Decile 5 2.221 1.343 755.921 123.176 13.996 (561.663) (331.995) Decile 6-9 2.396 0.993 753.736 115.478 15.834 (437.869) (224.021) Decile 10 No participants in this decile Notes: These numbers correspond to the neares. five estimator for the sub-group of 25-64 year participants. Standard errors in parentheses. 31 Figure 1: Mean Income Gain Plotted Against Pre-Intervention Income 150 0 3 125 0 c Q t 100- X C m 75 E 000 0. - 50 Pre-intervention income per capita Figure 2: Empirical Distribution Functions 1_Trabajar sample aja sml (pre-intervention) /; pst-intervention Cainlsml .75- 0 0 100 200 300 400 500 600 700 Income per capita 32 Policy Research Working Paper Series Contact Title Author Date for paper WPS2124 Social Exclusion and Land Robin Mearns May 1999 G. Burnett Administration in Orissa, India Saurabh Sinha 82111 WPS2125 Developing Country Agriculture and Bernard Hoekman May 1999 L. Tabada The New Trade Agenda Kym Anderson 36896 WPS2126 Liberte, Egalite, Fraternite: Monica Das Gupta May 1999 M. Das Gupta Exploring the Role of Governance 31983 In Fertility Decline WPS2127 Lifeboat Ethic versus Corporate Monica Das Gupta May 1999 M. Das Gupta Ethic: Social and Demographic 31983 Implications of Stem and Joint Families WPS2128 Learning Outcomes and School Gladys Lopez Acevedo May 1999 M. Geller Cost-Effectiveness in Mexico: 85155 The PARE Program WPS2129 Agricultural Extension: Generic Gershon Feder May 1999 P. Kokila Challenges and Some Ingredients Anthony Willett 33716 for Solutions Willem Zijp WPS2130 Deep Integration, Nondiscrimination, Bernard Hoekman May 1999 L. Tabada and Euro-Mediterranean Free Trade Denise Eby Konan 36896 WPS2131 Efficiency Wage and Union Effects William F. Maloney May 1999 T. Gomez In Labor Demand and Wage Eduardo Pontual Ribeiro 32127 Structure in Mexico: An Application Of Quantile Analysis WPS2132 A Regime-Switching Approach to Maria Soledad Martinez June 1999 A. Yaptenco Studying Speculative Attacks: Peria 38526 A Focus on European Monetary System Crises WPS2133 Resolution of Corporate Distress: Stijn Claessens June 1999 R. Vo Evidence from East China's Simeon Djankov 33722 Financial Crisis Leora Klapper WPS2134 Interlinkage, Limited Liability, and Kaushik Basu June 1999 M. Mason Strategic Interaction Clive Bell 30809 Pinaki Bose WPS2135 Hungary's Integration into Bartlomiej Kaminski June 1999 L. Tabada European Union Markets: 36896 Production and Trade Restructuring Policy Research Working Paper Series Contact Title Authc,r Date for paper WPS2136 An Empirical Analysis of Comoetition, Scott J. Wallsten June 1999 P. Sintim-Aboagye Privatization, and Regulation in 38526 Telecommunications Markets in Africa and Latin America WPS2137 Globalization and National Andr&; Solimano June 1999 D. Cortijo Development at the End of the 84005 20th Century: Tensions and Challenges WPS2138 Multilateral Disciplines for Bernard Hoekman June 1999 L. Tabada Investment-Related Policies Kamal Saggi 36896 WPS2139 Small States, Small Problerns? Willianm Easterly June 1999 K. Labrie Aart K'aay 31001 WPS2140 Gender Bias in China. the Republic Monica Das Gupta June 1999 M. Das Gupta Of Korea, and India 1920-90: Li Shuzhuo 31983 Effects of War, Famine, and Fertility Decline WPS2141 Capital Flows, Macroeconomic Oya Celasun July 1999 L. Nathaniel Management, and the Financial Cevde: Denizer 89569 System: Turkey, 1989-97 Dong He WPS2142 Adjusting to Trade Policy Reform Steven J. Matusz July 1999 L. Tabada David Tarr 36896 WPS2143 Bank-Based and Market-Based Asli Demirgug-Kunt July 1999 K. Labrie Financial Systems: Cross-Country Ross Levine 31001 Comparisons WPS2144 Aid Dependence Reconsidered Jean-Paul Azam July 1999 H. Siadovich Shantayanan Devarajan 37698 Stepheni A. O'Connell WPS2145 Assessing the Impact of Micro-credit Hassar Zaman July 1999 B. Mekuria on Poverty and Vulnerability in 82756 Bangladesh WPS2146 A New Database on Financial Thorsten Beck July 1999 K. Labrie Development and Structure Ashl Demirguc-Kunt 31001 Ross Levine WPS2147 Developing Country Goals and Constantine Michalopoulos July 1999 L. Tabada Strategies for the Millennium Round 36896 WPS2148 Social Capital, Household Welfare, Christiaan Grootaert July 1999 G. Ochieng And Poverty in Indonesia 31123