_______________________________________________________

POVERTY and ECONOMIC POLICY (PEP) Research Network Call for Research Proposals _______________________________________________________

Poverty and Employability Effects of Workfare Programs in Argentina

Sandra Fachelli Lucas Ronconi Juan Sanguinetti

First Draft: March 2004

_______________________________________________________ Sandra Fachelli works at the Ministerio de Economia, Argentina. Juan Sanguinetti at the Centro de Estudios para el Desarrollo Institucional (CEDI), Argentina; and Lucas Ronconi at Universidad de San Andres and CEDI, Argentina. This work was carried out with the aid of a grant from Poverty and Economic Policy (PEP) Research Network, financed by the International Development Research Centre (IRC). * We appreciate the valuable comments of Dorothee Boccanfuso, Evan Due, JeanYves Duclos and seminar participants at the Hanoi PEP meeting. Ignacio Franceschelli and Virginia Casazza provided excellent research assistance.

Introduction Argentina suffered a deep economic, social and political crisis during the last few years. The economy shrink by about 11% in 2002, and due to the currency’s depreciation, GDP per capita drop off to US$ 2,850 (down from US$ 8,210 at its peak in 1998). The crisis sharply aggravated the already difficult social situation. During 2002 poverty and unemployment were at their maximum historical level: 55% of argentine households were below the poverty line, and almost 20% of the labor force was unemployed. Unemployment is particularly severe among the least-skilled workers, being higher than 30% (INDEC, October 2002). This extremely negative context also had an impact on the education and health sectors where there is a growing evidence of deterioration in service delivery. The combined effect of all these factors has resulted in an increasingly conflictive social situation with high levels of violence and protests (see Fiszbein 2002). One of the main policies implemented by the government to deal with the crisis was to significantly increase the budget allocated to active labor demand policies. The number of beneficiaries of workfare programs increased from 90,000 at December 2001, to 1,200,000 at October 2002, and to 2,000,000 in 2003 (see table 1). The recent decline in unemployment (form 21.5% in May 2002 to 15.6% in May 2003) and in poverty (from 54.3% to 47.8%) has been presented by the government as evidence of the positive income and employability effects of workfare programs (Ministerio de Trabajo, 2003). Table 1: Long term trends in Poverty, Unemployment, Economic Growth and Workfare programs (1980-2002) Year 1980 1985 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003*

Poverty Rate Unemployment (Buenos Aires) (% of labor force) 8,0 16,0 41,2 26,4 18,7 16,9 17,0 22,6 25,5 25,2 24,9 26,7 28,9 38,3 54,3 47.8

2,0 6,1 7,5 6,5 7,0 9,6 11,5 17,5 17,2 14,9 12,8 14,3 15,1 18,3 17,8 15,6

Workfare programs

Growth of real GDP (%)

Nº Participants (monthly average)

Expenditure (millions of $)

4,5 -2,0 -1,8 10,5 9,9 5,7 5,9 -2,7 5,5 8,1 4,0 -3,4 -0,5 -4,4 -10,9e 5,4e

26.236 33.365 48.909 62.083 126.264 112.076 105.895 85.665 91.806 1.282.000e 1.992.000e

94 118 125 134 299 259 241 162 160 2.312e 3.586e

Source: Own elaboration based on Fachelli (2002), World Bank (2000) and INDEC. Notes: (*) May 2003. e estimated

Allocating more funds to social sectors, and particularly to labor programs, seems to be an adequate policy considering the current difficult situation. However, several questions have been raised in Argentina regarding the fairness and effectiveness of workfare programs. The program has been pointed out as a source of political clientelism and corruption1, and many analysts argued that their employment effects are questionable. In spite of the importance of the topic, most of the arguments are based on anecdotal evidence. There are very few empirical evaluations of the program. Our research objective is to contribute to the debate, providing an econometric evaluation of the poverty and employability effects of workfare programs in Argentina, using the most reliable argentine database: The “Encuesta Permanente de Hogares” (Permanent Household Survey, hereafter EPH). While our focus is on the argentine case, we consider that is relevant to other countries, particularly in Latin America, where active labor policies have been advocated as a way to soften the shocks generated by marketoriented reforms (Heckman et al., 1998; Goldbert L. and C. Giacometti, 1998; Marquez, 1999). The paper is organized in five sections. The second section briefly describes the characteristics of workfare programs in Argentina. The third section presents our research objectives, a review of the empirical evidence and the knowledge gaps. The forth sections presents the methodology and the results of our research (at this stage is only a draft). The fifth section concludes. Brief background of workfare programs in Argentina The “Jefes de Hogar” workfare program was implemented few weeks after president Duhalde took office in February 2002. However, workfare programs in Argentina have been implemented since 1993, and while their names have changed2, they have all the same basic characteristics and objectives3: •

The program is targeted at the least-skilled unemployed workers, preferably if they are head of household. People who receive

1 Ronconi (2001) surveys the main argentine newspapers, and finds that most of the press reports related to workfare programs mention the existence of political clientelism and corruption in the funds allocation process. 2 In 1993 it was called “Programa Intensivo de Trabajo”, from 1995 to 2001 “Programa Trabajar”, and since 2002 “Programa Jefes de Hogar”. Provincial governments also implemented their own workfare program with similar characteristics to the federal ones. In terms of magnitude the most important program was “Barrios Bonaerenses” implemented by the province of Buenos Aires. 3 See the Argentine executive’s decree number 327/1998 for the “Programa Trabajar”, and the executive’s decrees number 165/2002 and 565/2002 for the “Programa Jefes de Hogar”. A detailed description of workfare programs in Argentina is provided in Fachelli (2002) and Ronconi (2001).

• • •

unemployment insurance benefits, a pension or hold an informal job are not allowed to participate4. Participants receive a monthly benefit below the minimum wage5, during a certain period (between three and six months) paid by the government6. During that period participants receive training and have to work (between twenty and forty hours per week7) on communitarian projects at public or non-profit organizations8. The objectives of the program are: To act as a short-term safety net, and to increase employability among the least-skilled unemployed workers.

Research objectives and knowledge gaps Our research objective has three components: 1.- How well targeted is the program? A review of workfare programs in OECD and some developing countries, found that public-service jobs are well targeted at low-income unemployed workers when the wage rates have been set very low (See Dar and Tzannatos, 1999). In Argentina, the benefit is below the minimum wage so we might expect self-targeting. However, several concerns have been raised. Over 50% of total employment is informal and the government has no capacity to detect if the candidate holds an informal job. Hence, it could be the case that some benefits are assigned to people who also hold a job. But a second and potentially more important concern is that, due to lack of sound institutions, benefits are allocated in a political-clientelar basis, and not necessarily to the most needed. Kremenchutzky (1997) and Ministerio de Trabajo (1999) have surveyed a small number of workfare program participants (60 and 159 respectively), and they find few cases (less than 10%) where participants do not meet the requirements to receive the benefit (for example, she/he is well-educated, or already has an informal job). However, the sampling techniques used in both studies are questionable9. Fachelli, Ronconi and Sanguinetti (2002) present ad-hoc evidence showing several cases where the jobs are assigned

4

“Programa Jefes de Hogar” includes the following additional restriction: Candidates, in order to be eligible, have to show proof that their children are attending school and receiving appropriate medical treatment (such as vaccines). 5 The monthly benefit in the “Programa Trabajar” was $200 per month, while the monthly benefit in the “Programa Jefes de Hogar” is $150 per month. The minimum wage in Argentina is between $300 and $350 per month depending on the industry. 6 Both the “Programa Trabajar” and the “Programa Jefes de Hogar” were partially financed by through a World Bank loan. 7 In the “Jefes de Hogar” the work requirement is 20 hours per week, while in the Trabajar was between 30 and 40 hours per week. 8 “Programa Jefes de Hogar” allows a participant to work in private companies provided that the employer pays the payroll tax and some complementary benefit. 9 See Ronconi (2001)

through a political-clientelar basis, or different forms of corruption in the allocation process10. Our first objective is to describe the socio-economic characteristics of participants and non participants, using the Permanent Household Survey, in order to verify if the participants are in fact those who need the program most. We answer several questions, such as: Do participants have any other source of income? How many of them are heads of poor households? Do participants have a low education level? Is there any unemployed and poorly educated worker who does not receive the program11? 2.- Poverty effects A second concern is related to the poverty effect of workfare programs. Even in the case where the program is well targeted, is necessarily to measure the income gain conditional on income in the absence of the program to assess the impact of the program. As mentioned by Subbarao et al. (1997), common practice has been to estimate the gains by the gross wages paid, assuming that the labor supply to the program came only from the unemployed and from people who where out of the labor force. But, even if a participating worker was unemployed at the time she joined the program does not mean that she would have remained unemployed had the program not existed. Ministerio de Trabajo (2003) argues that the “Programa Jefes de Hogar” helped 29.3% of households that were below the indigence line to move out of indigence, and 6.5% of households that were below poverty to become non-poor. Their “estimation” is done assuming that benefits are targeted towards the poorest, and that the income gain of participating in the program is equal to the benefit. For obvious reasons this analysis does not provide much value added. Jalan and Ravallion (1999) estimated the net income gains of workfare programs in Argentina during 1997 constructing the counterfactual from a group of non-participants. They have exploited the cross section characteristic of the Encuesta de Desarrollo Social (EDS), and find average gains of approximately $100 per participant per month (i.e. 50% of the benefit). While this is the first serious attempt to measure the effects of the program, the results may suffer a bias as suggested in Ronconi (2001)12. 10

Just to mention a few examples: In the suburban area of Buenos Aires, participants received 2/3 of the benefit, while the remaining third was hold by a political leader. In La Matanza, funds were distributed by local leaders instead of been assigned directly from the government to the participant as the legal procedure stipulates. Some participants were forced to participate in political manifestations in order to receive the benefits. 11 Regrettably, the EPH does not include any political variable. Hence we are not able to check if in fact corruption and political clientelism characterize funds allocation. However, we consider that providing a reliable estimation of the percentage of participants who do not meet the requirements constitutes an improvement given the poor quality of the existing empirical evidence, and also an input for further studies. 12 Jalan and Ravallion’s (1999) results are based on a sample of 3,500 participants. However, the random sample was composed of approximately 6,000 participants. But they dropped 2,500 observations because the address of the participants could not be found, or because the participant did not want to respond. We consider that the dropped observations have a high probability of including participants who do not meet the program requirements.

In this paper we compute the net income gain of the program13, using a different database (the Permanent Household Survey) and a matching pairs approach (see the methodological section). The data and our empirical approach allow us to estimate the short and medium run poverty effect of the program14,15. 3.- Employability effects Finally, we assess the employability effects of the program. According to Bartik (2001), public service programs significantly increased the long-run earnings of participants in the US, since they provide some work experience and the needed soft skills. Do we observe this effect in the argentine case? How did participants performed in the labor market after the program? Are participants more or less likely to be employed than individuals in the control group after program completion? Does participation affects the odds of getting a formal job16? Do workers receive higher wages due to participation? Or is the workfare program a disguised income transfer? Furthermore, did the program have any negative impact, such as “signaling” or stigma effects on the participants? Or did the program generated dependency among participants? None of these questions have been appropriately answered in Argentina. As far as we are aware, there are no statistically reliable evaluations of the employment effects of workfare programs. Our objective is filling this gap, exploiting the panel characteristic of the Permanent Household Survey and implementing a matching pairs approach to construct the control group. Briefly, our research objective is to analyze how well the program is targeted, and its effectiveness in reducing poverty and increasing employability.

13

Since we do not follow a general equilibrium approach, we ignore indirect effects (such as an increase in income due the increase in aggregate demand generated by the program). These indirect effects are probably small before December 2001 because the number of participants was 1% of the labor force. However, they might have become important after the government increased the number of participants up to 15% of the labor force. 14 We estimate if the direct income gains generated by the program, helped the participants to move out of poverty and indigence. We use the official poverty and indigence lines. The poverty line is calculated based on the 1986/87 income and expenditure survey, and updated using prices indices for its food and non foods components. For 1998, the poverty line is $160 per male adult, per month. To calculate poverty, household composition is converted into male adult equivalents using standard conversion factors. The indigence line, is based on the food consumption portion of the poverty line, and is equal to $69 per male adult, per month. See more details in Annex 2. 15 The long-run poverty effect of the program is much harder to assess. It would be necessarily to measure how the program affects several outcomes. For example, Franceschelli and Ronconi (2002) argue that there exists a causal relation between the introduction of workfare programs and the emergence of the “Piquetero” movement in Argentina. 16 We define formal jobs as those which include unemployment insurance and pension’s benefits.

Methodology and data sources The empirical strategy adopted in this study is the result of the research objectives advanced in the previous discussion and the characteristics of the available data. The analysis is based on the Permanent Household Survey (EPH). The data is collected and processed by the Argentine National Statistical and Census Institute (Instituto Nacional de Estadísticas y Censos, INDEC). The survey has been conducted bi-annually in May and October since 1974, and covers 28 urban centers which represent 70% of total population of the country and 98% of the population living in centers with more than 100,000 inhabitants. The sampling and data collection techniques used by the INDEC ensures the validity and reliability of the information (See Annex 1 for more details)17. The EPH contains information related to occupational, educational and income characteristics, both at the individual and household level. Since October 2000, it includes a specific question that allows determining if the individual participates in a workfare program. Therefore, we are able to verify, at six different points in time, if the participants were in fact those who needed the program most. (October 2000, May 2001, October 2001, May 2002, October 2002 and May 2003). To assess how well targeted the program is, we analyze several variables, mainly years of schooling and household income per capita. We follow INDEC’s definition of poverty and indigence to measure the proportion of beneficiaries below poverty, and the proportion of poor people not receiving the benefit. In order to compute the short run poverty effects of the program, we need to measure the income gain conditional on income in the absence of the program (Heckman, Lalonde, Smith, 1998). The “with” data is provided by the EPH (i.e. we observe the income of participants); but the “without” data (i.e. what would have been the income of participants in the absence of the program) is fundamentally unobserved, since an individual cannot be both a participant and a non-participant at the same time. Following the conventional evaluation literature, we assume there exists a group of individuals comparable to participants except for not having received benefits. We use propensity score matching methods (see Heckman, Ichimura, Todd, 1998) to draw a comparison group to workfare participants form the large number of non-participants available in the EPH18. More specifically: Let Di=1 if individual i participates in the program, and Di=0 if does not participates. Let Xi be a vector of variables that helps predict participation in 17

The EPH can be download form INDEC’s web page www.indec.gov.ar We plan to estimate income gains for 2000 or 2001, since at that time the number of nonparticipants was sufficiently large to ensure that participants are matched with nonparticipants over a common region of matching variables. During 2002, almost every unemployed worker received a benefit, so it might not be possible to obtain a reliable matching pair.

18

the program; and P(X) = Prob(D=1/X) is the probability of participating conditional on X, the “propensity score”. We calculate the propensity score for each individual in the participant and the comparison-group samples using standard logit model. Then we compute the odds ratio pi = Pi / (1-Pi), where Pi is the estimated propensity score for individual i. Finally, we minimize [(p(Xi) – p(Xj)] 2 over all j, in order to obtain the “nearest neighbor” to the ith participant. Then, the mean income gain of the workfare program is given by the firstdifference: GFD = ∑i=1,..P [(Y1i - ∑j=1,..NP (wjiY0ji)] / P where Y1i is the household income of participant i, Y0ji is the household income of the jth non-participant matched to the ith participant, P is total number of participants, NP total number of non participants and wij are the weights applied in calculating the average income of the matched nonparticipants19. Hence, we compute the mean income gain of the program as the difference between the average income of participants and the weighted average income of non-participants. Finally, we compute the percentage of participant households who moved out of poverty due to the income gain G. (See Annex 2 for details regarding how the poverty line is calculated in Argentina). While this strategy does not allow us to claim any long term effects over poverty reduction, at least will assess the importance of workfare programs as short run safety nets. Our third objective is to estimate employability effects. We want to evaluate if participants are more or less likely to find a job, and earn a higher wage after program completion. As in the previous case, the fundamentally unobserved data is the employment performance and salaries of participants “without” treatment. Again, our empirical strategy is to use propensity score matching methods to draw a comparison group to workfare participants form the large number of non-participants available in the EPH20. But unlike the previous case, we need data “before” and “after” treatment. The Permanent Household Survey has a rolling panel structure: once a household is chosen it remains in the sample for four waves (two years). Each period, 25% of the families are replaced. Therefore, we follow participants and the comparison group through time.

19

In this first version of the paper we only compute the nearest neighbor. Hence, the number of non-participants included in the control group (NP) is equal to the number of participants (N) and the weights are all equal to 1. 20 We plan to follow the matched pair technique since the procedure is less arbitrary and program impact measures are easier to interpret (See Dar and Tzannatos, 1999). We consider matching estimates to be reliable since the same questionnaire was administered to both groups, and the EPH contains a wide range of socio-economic characteristics for each individual, allowing to control for a wide range of observable factors.

The difference in difference estimate of the mean income gain of the program before and after treatment is21: GDD = [∑i=1,..P (Y1it – Y1is) - ∑j=1,..P (Y0jt – Y0jt)] / P Y1it is the income of participant i at time t and Y1is the income of participant i at time s. Where t refers to a time period after treatment and s refers to a period before treatment. Hence, GDD measures the difference between the average income of participants after and before treatment relative to the average income of non-participants during the same time period. The sample formula applies to other outcomes such as the rate of unemployment or the labor force participation rate. At this stage an important point should be emphasized: During the period under consideration (from 2000 to 2002), the overall state of the argentine economy suffered major changes. GDP per capita decreased 25%, the unemployment rate went up 3 percentage points, and the share of informal employment increased from 37% to 50%. Under such a crisis, it would be incorrect to attach all the negative changes in participants’ outcomes (probability of being employed and wages) to the workfare program. In other words, the “before and after” estimator should be discarded, or at least taken with extreme caution. However, since we also work with a comparison group of non-participants, and under the reasonable assumption that the crisis had a similar effect over the outcomes of participants and non-participants, we can isolate the workfare programs effects from the economic crisis by computing a difference-in-difference estimator. Briefly, we follow the conventional evaluation literature. The value added of our paper is that we use these standard methods to explore a database and answer several questions that have not been analyzed yet. Preliminary Results 1. - Targeting As mentioned above, since October 2000 the EPH includes a specific question that allows determining if the individual participates in a workfare program. Therefore, we are able to verify, at six different points in time, if the participants were in fact those who needed the program the most. (October 2000, May 2001, October 2001, May 2002, October 2002 and May 2003)22. NOTE: In this first draft of the paper we only analyze how well target to the poor the program was in October 2000 and October 2002. In the next version of the paper we will include the remaining periods. Targeting in October 2000 21

This is the formula when only the nearest neighbor to each participant is selected. The October 2003 survey has not been released yet, but it might be in the next months so we could also include it in the analysis.

22

The EPH includes 561 observations where the person declares that she/he is participating in a public workfare program. These observations represent approximately 0.1 million people. It is important to mention that during October 2000 the number of benefits was a very small share of those who needed support. The number of unemployed people was 1.4 million, and the number of people living in households below the poverty line was 6.9 million. Therefore, we expect to find that a large share of poor and unemployed people did not receive the benefit. But were the scarce benefits allocated properly? Table 2, 3 and 4 present some basic socio-economic characteristics of beneficiaries and their households, and for the rest of the surveyed individuals. Table 2 Characteristics Age Gender (% of female) Head of Household Work experience (in months) Number of members in the household Residence located in a shantytown Lack of access to water, electricity and bathroom Residence ownership (yes=1) # observations

Participants of workfare program 34.6 years 58.1% 38% 31.4 months 4.9 2.5% 5.7%

Non-participants

57% 561

62% 45,647

37.1 years 52% 39% 53.3 months 4.4 1.4% 6.8%

Source: Own elaboration based on EPH (INDEC)

Almost 60% of program participants are female, less than 40% are head of household, and 2.5% of the participants live in a shantytown. The average participant is 2.5 years younger, has less work experience, and a slightly higher probability of living in a shantytown than the average nonparticipant. The differences are not large. Regarding years of schooling we observe that participants are on average less educated than non-participants. But again, the difference is not very large. While 44% of the beneficiaries have 7 or less years of schooling and 18.3% have some college education (i.e. more than 12 years of schooling), the average figures for the group of non-participants is 34% and 25% respectively. Table 3 Maximum education attained Incomplete primary school Completed primary school Incomplete high school Completed high school Incomplete college Completed college

Participants 12.5% 31.5% 21.2% 16.5% 13.0% 5.3%

Non-participants 8.9% 25.1% 21.9% 19.1% 14.6% 10.3%

Source: Own elaboration based on EPH (INDEC). Note: “Completed primary school” means 7 years of schooling. “Completed high school” means 12 years of schooling.

This is preliminary evidence that a significant share of the benefits have not been assigned to the poorest and least skilled workers as established in the normative. The inadequate allocation of benefits becomes more evident when we analyze household income per capita: Only 19.9% (95,000 out of 481,000) of participants are below the indigence line. 40.1% are below the poverty line but above the indigence line, and the remaining 40% are above the poverty line23. On the other hand, only 4.8% of the indigent households have at least one member who participates in the program. Table 4

Beneficiaries Without Benefit Total

Above Poverty line 192 13,843 14,035

Below Poverty line 288 6,577 6,865

Below Poverty line but above Indigence line 193 4,668 4,861

Below Indigence line 95 1,908 2,004

Total

481 20,420 20,901

Source: Own elaboration based on EPH (INDEC)

We also observe that, while most of participants are members of households located in the poorest quintiles, 24.6% of the beneficiaries are members of a household that ranks in the top 50% of the income per capita distribution. Table 5 Distribution of beneficiaries according to income per capita of the household 1st decile (Poorest 10%) 2nd 3rd 4th 5th 6th 7th 8th 9th 10th (Richest 10%)

Beneficiaries 30.6 % 13.8 % 11.1 % 11.8 % 8.2 % 6.7 % 5.8 % 6.0 % 3.6 % 2.5 %

Source: Own elaboration based on EPH (INDEC)

Finally, we found that 25.1% of the beneficiaries declare income higher than $200 per month, which was the workfare program benefit in October 2000. This might be evidence that, contrary to what was established in the normative, these participants are holding a job in addition to the workfare benefit. Summing up, we found evidence that in October 2000 the limited number of benefits were not appropriately distributed. On the one hand, while the average participant was less-skilled and poorer than the average non23

NOTE: In order to properly analyze how well target was the workfare program according to income is necessary to compute the income of participants in the absence of the program. This analysis would be included in the next version of the paper. At this stage we have analyzed the income of participants including the benefit. The other extreme assumption is to analyze the income of participants with out including the benefit. Under this alternative assumption the program appears to be better targeted, but there are still a significant number of participants who are above the poverty line.

participant, the difference was small. On the other hand, many high-skilled workers and members of non-poor households received the benefit, while a large number of low-skilled, unemployed and poor workers did not receive the benefit24. Targeting in October 2002 After the December 2001 political and economic crisis, the new government significantly expanded the number of funds allocated to workfare program (which became to be called “Jefes de Hogar”). The number of benefits increased from 90,000 in December 2001 to 1,200,000 in October 2002. While the increase in the number of benefits was significant, the crisis was so severe that the number of people living in households with income below the poverty line went up from 6.9 million people in October 2000 to 12.3 million people in October 2002. This section analyzes how well targeted were the funds allocated in October 2002. The EPH for October 2002 contains 2,329 observations were the person declares that she/he is participating in a public workfare program. These observations represent almost 1 million people. Tables 6 and 7 present some basic socio-economic characteristics of beneficiaries. Table 6 Characteristics Age Gender (% of female) Head of Household Years of Schooling Incomplete primary school Completed primary school Incomplete high school Completed high school Incomplete college Completed college

Beneficiaries of workfare programs 34.4 years 71.6% 39.3% 15.3% 37.4% 26.1% 13.6% 5.7% 1.9%

Source: Own elaboration based on EPH (INDEC)

Table 6 shows that more than 70% of the benefits were allocated to women. Regarding education, we observe that 52.7% of beneficiaries had 7 or less years of schooling. Compared to the figures for October 2000, we find that the program was better target towards the least skilled in October 2002. However, we also observe that in October 2002 there was still a significant share of beneficiaries (7.6%) who had some college education. The EPH indicates that 61.6% of members of households receiving a benefit are below the indigence line. 33.2% are below the poverty line but above the indigence line and 5.2% are above the poverty line. On the other hand, we find that 41% of people living in a household below the indigence line are receiving a benefit.

24

As a caveat it should be mentioned that, while the INDEC assures the confidentiality of the collected information, it could be the case that some of those individuals who are participating in the program and do not meet the eligibility criteria might not provide accurate information.

Table 7

Beneficiaries Without Benefit Total

Above Poverty line 203 8,813 9,015

Below Poverty line 3,731 8,591 12,322

Below Poverty line but above Indigence line 1,307 5,102 6,409

Below Indigence line 2,423 3,490 5,913

Total

3,934 17,404 21,337

Source: Own elaboration based on EPH (INDEC)

Finally, we observe that 10.4% of the beneficiaries declare an income higher than $150 per month, which was the workfare program benefit in October 2002. As mentioned before, this might be evidence that, contrary to what was established in the normative, these beneficiaries are holding a job in addition to the workfare benefit. Therefore, the evidence shows that the program was better targeted towards the least skilled and poor workers in October 2002 relative to October 2000. However, even during October 2002, the targeting was far from perfect, since many poor workers did not receive a benefit, while approximately 10% of the benefits were assigned to highly-skilled and nonpoor workers. 2.- Poverty and employability effects The first and critical step in our estimation is to find a comparison group of non-participants who has sufficiently similar characteristics to the participants except for not participating in the program. Following common practice in the evaluation literature we use propensity score matching methods to draw a comparison group to workfare participants form the large number of non-participants available in the EPH. We run standard logit models (one for each province) to estimate the propensity score by regressing D (an indicator of participation in the program) on a vector X of individual and household characteristics such as: age, gender, work experience, years of schooling, if the person is head of household, number of members in the household, quality of the residence (if the residence has access to water, electricity and sanitary installations), location of the residence, ownership of residence. Then, we compute the odds ratio pi = Pi / (1-Pi), where Pi is the estimated propensity score for individual i obtained in the logit regressions. Finally, we minimize [(p(Xi) – p(Xj)] 2 over all j, in order to obtain the “nearest neighbor” to the ith participant. Note: in this first version of the paper we only included one “nearest neighbor” in the comparison group for each program participant. While this is a reasonable strategy when a large number of observations are available, it seems to be less appropriate in our case (As explained below, the results for the difference-in-difference estimator for the period October 2000- May 2002 are based on 82 observations, which is too low). Regrettably, we

become aware of this problem too late. Therefore, in this first draft we present the results based on the selection of only one “nearest neighbor”. But for the PEP-June-Conference we will include estimates based on the “3 or 5 nearest neighbors” to each participant. We take October 2000 as the base state. The reason we discarded using May 2002, October 2002 or May 2003 as base state was because they are quite recent and do not allow to analyze the performance of ex-participants several months after treatment. On the other hand, since the workfare program become almost universal by October 2002, it might not be possible to find a reliable matching pair to each participant from the group of nonparticipants, since most of the poor and unemployed people was in the program. Note: The reason we choose October 2000 instead of May 2001 or October 2001 as the base state was arbitrary. However, we realize that it was a bad decision. The October 2000 survey was the first to include the specific questions that allows determining if the person is a workfare program participant. Hence, when we intend to analyze the “before” performance of participants, we are not able to know if those who were participants during October 2000, were participating or not during May 2000. We return to this issue below. For the PEP-June-Conference the paper would include the estimates obtained from taking May 2001 or October 2001 as the base state. In October 2000 the EPH reports 561 individuals who were participating in the six month workfare program. The group of non-participants includes 45,647 observations. From this large group we extracted the 561 nearest neighbor implementing the methodology described above. The following table presents some basic socio-economic characteristics of participants and their respective nearest neighbors. Table 8 Characteristics Age Gender (% of females) Head of Household Work experience (in months) Number of members in the household Residence located in a shantytown Lack of access to water, electricity and bathroom Residence ownership (yes=1) Schooling Incomplete primary school Completed primary school Incomplete high school Completed high school Incomplete college Completed college # of observations Source: Own elaboration based on EPH (INDEC)

Beneficiaries of workfare programs 34.6 years 58.1% 38.0% 31.4 months 4.9 2.5% 5.7%

Control Group

57%

60%

12.5% 31.5% 21.2% 16.5% 13.0% 5.3% 561

12.5% 33.9% 23.7% 12.3% 11.4% 6.2% 561

34.7 years 58.0% 39.0% 32.2 months 4.8 2.5% 5.8%

As we observe, both groups have very similar observable characteristics. The first question we consider is by how much did the income of participants change during the participation in the program due to the program. In other words, which would have been the income of participants during October 2000, if they were not beneficiaries? We find that, on average, participants had an income $40.48 per month higher than the comparison group25. Considering that in October 2000 the workfare program benefit was $200 per month, the net income gain of participating in the program was 20% of the benefit. Our estimated effect is quite smaller than the one computed by Jalan & Ravallion (1999). They estimated a net income gain of $100 per month or 50% of the benefit. Their estimation is for the year 1997 and they used a different database. These two factors might explain the discrepancy in the results. But we suspect that the difference could also be due to a potential bias on the sample used in Jalan & Ravallion. The positive effect on income could be explained by the fact that many participants would have remained unemployed or inactive, and hence without income, in the absence of the program. Specifically, the labor force participation rate was 60.7% and the unemployment rate was 17.5% among the control group. While the participation and employment rate of participants is by definition 100%. However, the effect on income was smaller that the benefit because many participants would have got a job and worked more hours in the absence of the program. The number of hours worked by an average participant was 27 per week, while the average employed nearest neighbor worked 40.5 hours per week. The described first-difference estimator has two major shortcomings. On the one hand, it does not allow measuring “before and after” effects of the program. On the other hand, while the first-difference estimator controls for observed heterogeneity between participants and non-participants, one cannot eliminate latent heterogeneity that could bias the impact estimates of the program using the nearest neighbor as a counterfactual. For example, it could be argued that participants have higher social capital. And it is this higher level of social capital what explains both the higher probability of participating in the program and also the higher level of earnings. Since we do not observe social capital we can control for it. Therefore, the first-difference estimator would inappropriately consider the effect of social capital as part of the program effect. However, if the source of heterogeneity is time-invariant we can eliminate it by computing difference-in-difference estimators. For example, the level of social capital might be quite constant through short periods of time. Hence, we can get rid of the problem by analyzing the performance of participants 25

We also estimated the income gain excluding all those “suspicious” participants and their respective nearest neighbor. The result was basically the same: an income gain of $40.81 per month. (By suspicious we mean those participants who reported income above $300 per month).

and the control group at different points in time. This estimator also allows measuring “before and after” effects. “Before treatment” May 2000 In the EPH for May 2000 we find that 328 people were surveyed, out of the 561 who were workfare program participants in October 2000. However, when we only retain those observations for which the nearest neighbor was also surveyed, we end up with 251 pairs. The average income of the 251 individuals who were participants in October 2000, was $9.3 higher in October 2000 than in May 2000. Considering that during that period Argentina was suffering a recession, the results appears to be quite remarkable. Indeed, the average income of the 251 nearest neighbors went down between May 2000 and October 2000 by $100.4. Therefore, the difference-in-difference estimate for the program is a positive effect of $109.7 per month. Table 9 “Before treatment”

# hours worked per week Unemployment rate Labor force participation rate Works in the public sector # observations

Participants May October 2000 2000 29.4 26.2 11.7% 0% 84.5% 100% 78% 84% 251 251

Control Group May October 2000 2000 37.2 40 18.9% 21.4% 59% 61.1% 20% 20% 251 251

Source: Own elaboration based on EPH (INDEC)

The difference-in-difference for the labor force participation rate is equal to 13.4 percentage points, and for the unemployment rate -14.2 percentage points. As shown in Table x, while the rate of unemployment increased by 2.5 percentage points for the control group between May 2000 and October 2000, it went down by 11.7 percentage points for those who were participants during October 2000. However, there may be a severe problem with these estimates. They are based on the assumption that people who were participating in the sixmonth workfare program during October 2000, were not participating during May 2000. Since, there is a five months gap between May and October, it could be argued that approximately 1/6 of those receiving the benefit in October were also receiving the benefit in May. Regrettably, we cannot control for that problem. The October 2000 survey was the first to include the question that allows us to determine if the person is participating in a workfare program. Moreover, as we argue in the next paragraphs, there is evidence to suspect that while the program’s length is only 6 months, participants continued receiving benefits for a longer period. Therefore, it could be the case that a substantial number of those who received the benefit during October 2000 were also receiving the benefit during May 200026. 26

Another piece of evidence that supports this possibility is the fact that 78% of the October 2000 participants, were working in the public sector during May 2000. And most of workfare projects are implemented by the public sector.

Note: As mentioned before, we became aware of this problem too late. For the PEP-June Conference we would present the results using May of October 2001 as the base state. Therefore, we would be able to determine the participants who did not received the benefits during October 2000, and hence properly measure the “before treatment” situation. “After treatment” May 2001 During May 2001, i.e. 7 months after October 2000, the EPH surveyed 327 out of the 561 individuals who were program participants during October 2000. The average income for this group was $51.8 higher in May 2001 than in October 2000. We find that their labor participation rate went down and unemployment up. This result was expected since the transition from program participation to obtaining a job is obviously not frictionless. However, we are surprised by the fact that only 10% of those who were program participants went out of the labor force and the unemployment rate among ex-participants was only 9.2%. These results are quite surprising considering that the labor force participation rate among the control group was 63.4% and the unemployment rate 15.2%. One of the explanations for the previous result is the fact that 47.1% of those who were participants during October 2000 appear to still be participants during May 2001. This result was unexpected since the normative established that the length of the program is six months. While renewal of benefits was not explicitly prohibited, there was an implicit solidarity objective in the program. The idea was to distribute the scare benefits among as many poor people as possible. Hence, those candidates who did not participate before had preference over those who did participate. In order to compute the before and after estimate we drop all those observations where the individual declares that she/he participated in the workfare program during both October 2000 and May 2001. The number of pairs becomes 129. We find that the average income of ex-participants was $32 higher during May 2001 relative to October 2000, while the average income of their respective nearest neighbor was $59 higher during May 2001 relative to October 2000. Table 10 also shows that the average number of hours per week worked by ex-participants during May 2001 was very similar to the number of hours worked by the control group (37.2 and 37.8 respectively). However, during treatment participants were only working 26.9 hours per week. Hence, income per hour was quite high during treatment. This fact may explain why many participants had an incentive to continue participating in the program. What remains to be explained is why de government continued allocating the scare funds to those that already received the benefits instead of assigning the funds to those needy candidates who never got it.

Table 10 “After treatment”

# hours worked per week Unemployment rate Labor force participation rate Employment rate # observations

Participants May October 2001 2000 37.2 26.9 17.5% 0% 79.8% 100% 65.9% 100% 129 129

Control Group May October 2001 2000 37.8 40.2 15% 22.1% 63.1% 60.6% 54.3% 47.2% 129 129

Source: Own elaboration based on EPH (INDEC)

The difference-in-difference estimate between ex-participants and the control group between May 2000 and May 2001 is $82.727. The average income of those who received treatment was $41.3 higher after treatment – May 2001- than before treatment –May 2000-, while the average income of the control group was $41.4 lower during May 2001 relative to May 2000. The labor force participation rate, the unemployment rate and the rate of employment were higher for the group of ex-participants relative to the control group during May 2001. While 79.8% of ex-participants remained in the labor force during May 2001, the participation rate for the control group was 63.1%. On the other hand, 17.5% of ex-participants were actively looking for a job but could not find it, while the unemployment rate for the control group was 15%. Finally, 65.9% of ex-participants were working while the employment rate for the control group was 54.3%. This analysis provides estimates of how participants performed seven months after completion of the workfare program. Our next step is to analyze the performance of ex-participants during May 2002, a year and a half after program completion. “After treatment” May 2002 The EPH for May 2002 surveyed 119 individuals out of the 561 who received treatment during October 2000. However, not all these individuals are exbeneficiaries as expected: 32.4% of those who were participating in the six month workfare program during October 2000 declare to continue participating one-year-and-a-half later. We drop those observations in order to compare the performance of ex-beneficiaries to their respective nearest neighbor and we end up with 41 pairs of observations. The average income of ex-participants was $7 higher during May 2002 relative to October 2000. While the average income of the control group was $9 higher during May 2002 relative to October 2000. The labor force participation rate was still higher among ex-beneficiaries than their respective nearest neighbors, but the difference was much smaller than the one found for May 2001. During May 2002, 68.3% of exparticipants were actively looking for a job while only 63.4% of the control group. The rate of employment was also higher for the group of ex27

This estimator is reliable under the assumption that those individuals who received treatment during October 2000, were not receiving treatment during May 2000.

beneficiaries (56.1% relative to 48.3%) meaning that participation in the program had a positive effect on the chances of getting a job. Preliminary Comments Besides all the caveats and the methodological aspects that need to be improved, we consider that two preliminary results are worth mentioning: On the one hand, while the average participant was less skilled and poorer that the average non-participant, targeting was far from adequate. On the other hand, we found that many participants received treatment for a longer period contrary to what was established in the normative. We found that one third of those who received treatment during October 2000 were still receiving treatment a year and a half later.

Conclusion To be written

References Bartik T. (2001), Jobs for the Poor. Can Labor Demand Policies Help? Russell Sage Foundation, New York. Dar A. and Z. Tzannatos (1999), “Active Labor Market Programs: A Review of the Evidence from Evaluations”, Working Paper, World Bank. Elías V. and F. Ruiz Nuñez (2000), “An Evaluation of the Impact of the Young Training Program in Argentina”, working paper, Universidad Nacional de Tucumán. Fachelli S., Ronconi L. and Sanguinetti J. (2002), “Política Laboral Activa en Argentina”, mimeo, Centro de Estudios para el Desarrollo Institucional. Fachelli (2002), “Programas de Empleo de Ejecución Provincial”. Dirección de Gastos Sociales Consolidados. Secretaría de Política Económica. Ministerio de Economía. 2002. En imprenta. Franceschelli I. and L. Ronconi (2003), ““Comportamiento político de nuevos grupos: El Movimiento Piquetero y la política pública asistencial”. Segundo Premio de Concurso. Sociedad Argentina de Analisis Politico. Revista SAAP forthcoming. Fiszbein, Giovagnoli and Aduriz (2002), “Argentina’s crisis and its impact on household welfare” Working Paper No, World Bank Office for Argentina, Chile, Paraguay and Uruguay. Goldbert L. and C. Giacometti (1998), Programas de Empleo e Ingresos en América Latina y el Caribe, Banco InterAmericano de Desarrollo y Oficina Internacional del Trabajo, Atenea Impresores-Editores, Perú. Heckman J. and J. Hotz (1989), “Choosing among alternative Nonemperimental methods for estimating the impact of social programs”, Journal of the American Statistical Association, Vol.84, No. 408. Heckman J., H. Ichimura and P. Todd (1998), “Matching as an Econometric Evaluation Estimator”, Review of Economic Studies, 65, 261-294. Heckman J., Lalonde R. and J. Smith (1998), “The Economics and Econometrics of Active Labor Market Programs”, in Ashenfelter and Card (eds) Handbook of Labor Economics, Volume 3A. Amsterdam: North Holland. Jalan J. and Ravallion M. (1999), “Income gains to the Poor from Workfare: Estimates for Argentina’s Trabajar Program”; Working Paper, World Bank. Kremenchutzky (1997), “Evaluación Diagnóstica del Programa Trabajar I”, Crisol Consultores Económicos de Empresas Industriales; SIEMPRO. Marquez (1999), “Unemployment Insurance and Emergency Employment Programs in Latin America and the Caribbean: an Overview” Conference on Social Protection and Poverty, Inter- American Development Bank. Ministerio de Trabajo (1999), “Evaluación del Programa Trabajar III”, Noviembre 1999, Argentina. Ministerio de Trabajo (2003), “Impacto del Plan Jefes y Jefas de Hogar en la Pobreza”, Direccion General de Estudios y Formulacion de Politicas de Empleo, Argentina. Ronconi L. (2001), “Determinates Político-Institucionales de los Programas de Empleo en Argentina”, Working Paper #63, CEDI. World Bank (2000a), “Poor People in a Rich Country: A Poverty Report for Argentina” vol I y II, World Bank. World Bank (200b), “Gestion del riesgo social en Argentina”, World Bank Office for Argentina, Chile, Paraguay and Uruguay.

Annex 1. Encuesta Permanente Hogares – Permanent Household Survey The EPH is a sampling survey, developed 28 in urban agglomerates. The rigorous application of statistical methods ensures the validity and reliability of collected information (selection of sample members is conducted using random selection techniques, data collection methods are uniform, etc). A detailed description of sampling and data collection techniques is available at www.indec.gov.ar. The sample has a wide representation as the following table shows:

Provinces

ratio (%) Total population of province to total Census 91 (A) population

EPH Agglomerates

ration (%) of Total % EPH urban population of sampling population EPH over to total agglomerate agglomera population Census 91 (B) te (B/A) population 2965403 100,0 10,4 7948443 99,7 27,9

Ciudad de Bs As Buenos Aires

2965403 7969324

9,1 24,4

Ciudad de Bs As Partidos del Conurbano

Catamarca Cordoba

264234 2766683

0,8 8,5

Corrientes Chaco Chubut

795594 839677 375189

2,4 2,6 1,1

Entre Ríos

1020257

3,1

Formosa Jujuy La Pampa La Rioja Mendoza Misiones Neuquén Resto de Bs As

398413 512329 259996 220729 1412481 788915 388833 4625650

1,2 1,6 0,8 0,7 4,3 2,4 1,2 14,2

Gran Catamarca Rio Cuarto(*) Gran Córdoba Corrientes Gran Resistencia C. Rivadavia-Rada Tilly Rawson-Trelew (***) Concordia(*) Gran Paraná Formosa Jujuy-Palpala Santa Rosa-Toay La Rioja Gran mendoza Posadas Neuquén-Plottier Bahia Blanca-Cerri Mar del Plata-Batán(*) Gran La Plata San Nicolás de los Arroyos (***) Carmen de Patagones (***)

121815 138853 1175400 258103 292287 127038 97355 116485 207041 147636 219924 80592 103727 773113 210755 183579 265885 519065 642979 119302

46,1 5,0 42,5 32,4 34,8 33,9 25,9 11,4 20,3 37,1 42,9 31,0 47,0 54,7 26,7 47,2 5,7 11,2 13,9 2,6

0,4 0,5 4,1 0,9 1,0 0,4 0,3 0,4 0,7 0,5 0,8 0,3 0,4 2,7 0,7 0,6 0,9 1,8 2,3 0,4

17075

0,4

0,1

Viedma (***) Salta Gran San Juan San Luis-El Chorrillo Rio Gallegos Gran Santa Fé Gran Rosario Villa Constitución (***) Sgo del Estero-La Banda

40398 368659 352691 113074 64640 396991 1117322 41161 261824

8,0 42,6 66,7 39,5 40,4 14,2 39,9 1,5 39,0

0,1 1,3 1,2 0,4 0,2 1,4 3,9 0,1 0,9

Ushuaia-Río Grande Gran Tucumán-Tafí Viejo

67303 652882 20208800

97,0 57,2 61,9

0,2 2,3 71,1

Río Negro (**) Salta San Juan San Luis Santa Cruz Santa Fe

Santiago del Estero Tierra del Fuego Tucumán Total País Total Urbano

506772 866153 528715 286458 159839 2798422

1,6 2,7 1,6 0,9 0,5 8,6

671988

2,1

69369 1142105 32633528 28439499

(*)Aglomerates included in October 1995

Source: EPH-INDEC

0,2 3,5 100,0 87,1

Annex 2.- Indigence and Poverty line methodology The method currently in use by the INDEC to measure poverty and indigence is presented below. Indigence The concept of «indigence level» (or indigence line), IL, aims to assess whether the households earns enough income to purchase a food basket that will satisfy a minimum threshold of energetic and protein needs. Thus, the household that does not meet that threshold or line is considered indigent. The procedure is based on the use of a “Canasta básica de alimentos” -basic food basket- (CBA) of minimum cost, determined as a function of the consumption patterns of a reference population defined according to the results of the 1985-86 Household Expenditure and Income Survey. The procedure also takes into account the prescribed kilocalories and protein requirements for that population (as specified in the «Basic Food Basket for the Equivalent Adult», included below). Once the CBA components have been established, their prices are assigned according to the Consumer Price Index (IPC) for each measurement period. Since human nutritional requirements vary according to age, sex and person’s activity, INDEC adjust for each person’s characteristics, taking as reference the requirements of a male adult aged between 30 and 59 and exerting moderate activity. This reference unit is called the «equivalent adult» and is assigned the value 1. The table of equivalences of energetic requirements for each consumer unit in terms of equivalent adult is the following: Table of equivalences Energetic needs and consumer units by age and sex Greater Buenos Aires Energetic Consumer unit / Sex and age needs (Kcal) Equivalent adult Boys and girls Under 1 year old 880 0.33 1 year old 1170 0.43 2 yrs. Old 1360 50 3 yrs. Old 1500 0.56 4 to 6 yrs. Old 1710 0.63 7 to 9 yrs. Old 1950 0.72 Men 10 to 12 yrs. Old 2230 0.83 13 to 15 yrs. Old 2580 0.96 16 to 17 yrs. Old 2840 1.05 Women 10 to 12 yrs. Old 1980 0.73 13 to 15 yrs. Old 2140 0.79 16 to 17 yrs. Old 2140 0.79 Men 18 to 29 yrs. Old 2860 1.06 30 to 59 yrs. Old 2700 1 60 yrs. old and over 2840 1.05 Women 18 to 29 yrs. Old 2000 0.74 30 to 59 yrs. Old 2000 0.74 60 yrs. old and over 1730 0.64 Note: Extracted from the table by MORALES, Elena, Canasta básica de alimentos, Gran Buenos Aires, Documento de trabajo n°3,INDEC/IPA, 1988.

Each household’s composition in equivalent adults determines a specific CBA value for that household. In September, 2000, the CBA value for an equivalent adult was 62,44 pesos. As a final step, the specific value of each household’s CBA is compared to the household’s total income. If the total income is less than the household’s CBA, the household and its members are considered to be under the indigence level. Poverty The measurement of poverty by the poverty level or «poverty line» (PL) method is based on determining, from the household income reported, whether the household in question is able to satisfy --through the purchase of goods and services-- a set of nutritional and non-nutritional needs considered essential. In order to calculate the poverty level the INDEC determines the CBA value and compound it with the inclusion of non-nutritional goods and services (clothing, transportation, education, health care, etc.) so as to obtain the value of the Total Basic Basket (CBT). For the purpose of compounding the CBA value, the so-called Engel coefficient (EC) is used. The EC is defined as the ratio of food expenditures to total expenditure observed in the reference population in the base year (1985-86). Thus: Engel coefficient = Food expenditures / Total expenditure. In each period, both the numerator and the denominator of the Engel coefficient are updated with the price variations obtained from the CPI. According to the relative price variation, the EC is determined each month for the purpose of measuring poverty. In order to compound the CBA value, in practice its value is multiplied by the reciprocal of the Engel coefficient: CBT = CBA x 1/Engel coefficient. In September, 2000, the reciprocal of the Engel coefficient was 2.42 and the CBA was 62.44 pesos. Thus we have $ 62.44 (CBA) x 2.42 (reciprocal of EC) = $ 151.10 (CBT) for an equivalent adult). As a last step, each household’s CBT value is compared to the household’s total income. If the household’s income is less than the CBT value, the household and its members are considered under the poverty line; otherwise, they will be considered as non-poor.

ANNEX N°2 Basic food basket for the equivalent adult (monthly) Component Grams Specifications Bread 6060 Salt crackers 420 Sweet biscuits 720 Rice 630 Wheat flour 1020 Other flours (corn) 210 Noodles 1290 Potatoes 7050 Sweet potatoes 690 Sugar 1440 Sweets and jams 240 made with milk made with sweet potatoes marmalades Dry legumes 240 Lentils Beans Green peas Vegetables 3930 Chard Onions Lettuce Tomatoes Carrots Pumpkins Canned tomatoes Fruits 4020 Bananas Tangerines Apples Oranges Meats 6270 Short ribs Chuck Minced meat Rump beef Navel and foreshank Bottom round beef Shoulder clod Chicken Eggs 630 Milk 7950 Cheese 270 fresh spread Quartirolo grated Cooking oil 1200 blended Sweet/sweetened beverages 4050 Concentrated (fruit) juices drinks Unsweetened carbonated beverages Table salt Kitchen salt Vinegar Coffee Tea Maté

3450 Soda water 150 90 90 60 60 600

Sources: Documento de trabajo n° 3, Idem n° 8, INDEC / IPA