ارزیابی برنامه کاریابی برای افراد پیر و جوان: تاثیر ناهمگن در طول مدت بیکاری

This paper exploits an area-based pilot experiment to identify average treatment effects on unemployment duration of treated individuals of two active labor market programs implemented in Portugal. We focus on the short-term heterogeneous impact on two subpopulations of unemployed individuals: young (targeted by the Inserjovem program) and old (targeted by the Reage program). We show that the latter program has a small and positive impact (reduction) on unemployment duration of workers finding a job upon participation, whereas the impact of Inserjovem is generally negative (extended durations). These results are robust to a wide variety of constructions of quasi-experimental settings and estimators. The identification of heterogeneous effects showed that the program results were less satisfactory for young workers, for those over 40 and for the less educated. Women also benefited less from the programs. The results seem to improve slightly for young workers in the 2nd semester of implementation, but they deteriorate in the medium term. The lack of wage subsidies in the Portuguese programs may explain the minor impacts obtained, when compared to similar programs.

The effectiveness of active labor market programs (ALMP) in reducing unemployment and speeding transitions into employment has been the subject of a large and growing literature, as reviewed in Heckman et al. (1999) and Kluve (2006). We evaluate a set of ALMP directed at both old and young Portuguese unemployed. In the late 1990s, Portugal developed two initiatives — Reage and Inserjovem — aiming at increasing the employability of the long-term unemployed (the Reage program), and acting earlier on youth unemployment, preventing episodes of long-term unemployment at the beginning of their labor market careers (the Inserjovem program).1 This emphasis on preventive actions on long-term unemployment led us to choose the impact on unemployment duration as the outcome of interest in our evaluation exercise. The most distinctive feature of our study is the possibility of evaluating the impact of ALMP on young and older workers at the same time. In recent years, employment policies have started paying further attention to older individuals (OECD, 2006). The promotion of longer working lives due to issues of population ageing and pension system funding brings about increasing challenges to ALMP. A larger pool of older unemployed requires the attention of public policies, but also the challenges are different from those faced by younger unemployed. Traditionally, the older group faces harder labor market conditions; Addison et al. (2004) report evidence of a non-stationary labor market environment, namely, that the arrival rate of job offers falls sharply with age. Additionally, past experience with ALMP shows that older workers are less motivated to participate in some of the initiatives (e.g. training, a key component in the learning for life approach). By focusing on the differentiate impact across age groups, this paper contributes to this strand of the literature. The relevance of our study is further potentiated by the institutional rigidity of the Portuguese labor market. Despite being applied during a period of low unemployment, the Portuguese labor market is characterized by extremely high employment protection, long unemployment spells and generous unemployment insurance, and a low arrival rate of job offers, even for European standards. Overall, the Portuguese setting constitutes a challenging environment for any ALMP. Thus, in the context of the European Employment Strategy, a rigorous evaluation of the Portuguese experience with ALMP applied to such a diverse group of individuals may be of great relevance for the implementation of similar programs in other countries. Indeed, we find evidence of an heterogeneous impact of the programs. For individuals exiting into employment, those aged 30–40 reduced the duration of their unemployment spells, while for older individuals (over 40) and young cohorts (less than 25) the programs had no impact. The programs under evaluation target all young people (less than 25 years old) before they have been registered for 6 months and all adults before they reach 12 months of unemployment. The early intervention is meant to ensure the timely implementation of responses suitable to each individual's situation. These responses are essentially job-search assistance, including vocational guidance, counseling, monitoring, and training or re-training options. Furthermore, they have a mandatory character, in the sense that failing to comply with the directions of the Employment Office (EO) placement team will result in the loss of subsidies (including unemployment insurance and fee-exemption to access the public health services). The implementation of the Portuguese programs created an almost natural setting for the evaluation exercise. The programs were first introduced in a subset of EOs beginning in June 1998, generating an area-based pilot experiment that we explore in our identification strategy of the programs' impact. Afterwards, they were rolled out sequentially to the other EOs, fully covering the country in January 2001. The pilot EOs were not randomly chosen, but neither was participation based on specific local labor market conditions. Indeed, they were picked for logistical reasons unrelated with the programs' goals in terms of labor market outcomes. We apply a difference-in-differences methodology using the natural treatment and control groups originating from this pilot setting. The treatment group will consist of all registered individuals who participated in the programs in pilot EOs and the counterfactual is drawn from the subset of EOs not implementing the program in the evaluation period. The difference-in-differences methodology is supplemented with a combination of matching methods to generate the difference-in-differences matching estimator, used to eliminate some potential sources of bias Heckman et al. (1997). The goal of this paper is to determine the effects of the programs compared to the outcome of the individual had (s)he continued to search for a job in the absence of the support provided by the programs. We focus on the duration of complete spells of unemployment of individuals leaving the programs during the first 12 months of implementation (a short-term evaluation) and 2 years later (a medium-term evaluation). Given the wide coverage and mandatory nature of the programs, they can be considered as having comprehensive implementation. This raises the possibility of observing indirect effects, such as substitution and equilibrium effects. The first effect operates through a change in the relative price of labor between the treated and untreated individuals and the second through an increase in the supply of labor that lowers wage and increases employment. The evidence collected by Katz (1998) shows that these effects are more likely to arise in the context of wage subsidies, which were not included as a treatment in the Portuguese programs, contrary to what has happened in other countries, for example, the U.K. and Sweden. In our empirical work, we present some evidence on the dimension of possible substitution effects by exploring the regional implementation of the programs. Previous microeconometric studies of ALMP in European countries, taking place at around the same time period, find mixed results. Blundell et al. (2004) find an important “program introduction effect” for the UK; the program effect is much larger in the first quarter than later on. These results are confirmed in the longer-run analysis of De Giorgi (2005) for a sample of young males, using a regression discontinuity approach; he finds no evidence of substitution or general equilibrium effects. Larsson (2003) and Sianesi (2004) find no significant effect in the Swedish programs; the effects are small and positive for the employment rate, but negative on reemployment wages. Still for Sweden, Carling and Richardson (2004) conclude that transition rates improve more for subsidized work experience and training provided by firms than those observed for classroom vocational training. For East Germany, Eichler and Lechner (2002) find that participation in public employment policies implemented after reunification reduces participants' probability of employment. For a massive ALMP implemented in the late 1990s in Switzerland, Gerfin and Lechner (2002)'s evaluation finds a positive impact for a policy involving a wage subsidy, but negative effects for traditional employment programs. An alternative strand of the literature conducts the evaluation of ALMP with duration models, such as in the work of Bonnal et al. (1997) for France, Eberwein et al. (2002) for the US Job Training Partnership Act and van den Berg and Klaauw (2006), who study ALMP applied in the Dutch labor market. The minor impacts on the labor market prospect of the unemployed involved in ALMP are typically confirmed in these alternative studies. Our assessment of the Portuguese programs points to a small reduction in unemployment duration. In the absence of the program, we estimate that unemployment duration of treated individuals would increase by at most 0.4 of a month, which would not represent a large increase in duration given that some workers spend many months unemployed. The results show some degree of heterogeneity for different types of exits and programs. Recipients of treatment in Inserjovem tend to benefit more when moving into training (a reduction in unemployment duration) than into employment (with an increase in unemployment duration), although the impact never exceeds 0.3 of a month. For older workers, Reage program, the results point in the other direction with shorter unemployment spells for individuals placed in a job upon program participation and longer spells if they enter training. Gender, age and schooling seem to play an important role in determining the programs' impact. In transitions into employment, the impact is larger for men in Reage (a reduction of unemployment close to one month). In terms of age, the largest impact is observed for individuals aged between 30 and 40. Workers with a higher degree of education seem to benefit more from the programs, especially those in Reage. When the programs are evaluated for cohorts that were treated after the initial implementation phase, the results point towards similar impacts in the second semester of implementation, but to longer unemployment spells when evaluating the cohorts two years after June 1998; a pattern similar, for example, to Blundell et al. (2004). The results are robust to the choice of treatment and control groups and we were not able to find clear spillover effects from treated to untreated areas in terms of unemployment duration. The modest results of the Portuguese programs in reducing unemployment duration might be explained by the lack of some key treatments, such as wage subsidies that, as claimed in Katz (1998), increase ALMP effectiveness for specific disadvantaged groups. In the context of the “live longer, work longer” debate, the paucity of the results for older workers (aged more than 40) points towards the importance of designing policies specific to these workers and, more generally, to the educationally disadvantaged groups. The paper is organized as follows. The labor market programs are described in Section 2. The evaluation problem, as well as the identification and estimation of average treatment effects are addressed in Section 3. Section 4 presents the data and the results are discussed in Section 5. Finally, concluding remarks are presented in Section 6.

This paper evaluates the short-run impact of Inserjovem and Reage, two ALMP initiatives, introduced in Portugal in the late 1990s. The measure of program effectiveness is the reduction in unemployment duration through program participation. The programs target two distinct age groups: the Inserjovem, for individuals under 25; the Reage, for individuals 25 or over. These programs were mandatory with treatment ranging from job-search assistance to training and vocational or non-vocational guidance. No wage subsidies were included in the possible treatments. The possibility of analyzing the response of young and older workers to such a treatment in the same labor market is one of the main advantages of the Portuguese experience. Identification of the average treatment effect on the treated is achieved by means of an important source of variation generated by the area-based pilot implementation of the programs across the country. This implementation created natural treatment and control groups, drawn from pilot and non-pilot areas, respectively. The positive impact of the programs in reducing unemployment duration is very limited. In fact, through program participation, we find a small reduction on the length of unemployment spells for Reage participants finding a job, and a tiny increase in the spell durations of individuals in Inserjovem. These results are driven by the behavior of men, as women seem to react less to these treatments. Our results point to a more positive impact for individuals aged 30 to 40 and among the better educated. An important lesson to be drawn from this study for the ongoing discussion on ALMP for older and disadvantaged workers is the apparent difficulty of the programs in improving their unemployment experiences. Indeed, the less educated individuals and those over 40 did not benefit at all from the programs, which can be seen as a partial failure of the programs with two of the groups that face the worst prospects in modern labor markets. The impact of the program at later points in time, admittedly after a learning period by EOs and other stakeholders, did not improve the prospects of the unemployed, either. As with job placement, the programs also had a differentiated impact on the other two destination states considered. In particular, the programs reduced unemployment duration for young workers exiting the labor force, but increased it for older workers. A similar result is obtained with exits to training. What can explain the weakness of these impacts when compared with other recently reported in the literature (e.g. Blundell et al., 2004 and De Giorgi, 2005 for the United Kingdom)? A possible driving force can be traced back to the work of Katz (1998), who, in reviewing different ALMP, found that the most successful were those combining wage subsidies with job-search assistance and training. Contrary to the policies implemented, for instance, in the United Kingdom, the first dimension was absent in the Portuguese policy mix. Katz (1998) also showed that such a combination was particularly helpful for more disadvantaged groups, exactly those who did not benefit from the Reage and Inserjovem programs. We looked only at one dimension of the programs' success. There are, however, alternative dimensions for future research paths, namely longer-run effects and post-unemployment job match quality.