Responsible researcher: Bruno Benevit
Authors: Alberto Abadie, Joshua Angrist and Guido Imbens
Intervention Location: United States
Sample Size: 11,204 adults
Sector: Job market
Variable of Main Interest: Income
Type of Intervention: Training
Methodology: OLS, Quantile Regression, 2SLS, and IV (QTE)
Summary
Competitive inequality in the labor market was a topic of great interest in the USA in the late 1980s, raising concern among public policy makers. Identifying adherence and the effects of these policies on different groups of the population is fundamental, as it provides support for their improvement. In this sense, the objective of this article was to estimate the effect of the Job Training Partnership Act (JTPA) training program on the income of its participants, verifying such effects in different ranges (quantiles) of this outcome variable. Using OLS, Quantile Regression, 2SLS and IV-QTE methods, the authors demonstrated that the JTPA program positively affected the income of its participants in a heterogeneous way in relation to gender and income level.
The distributional consequences of applying subsidized training programs are a topic of great importance for the welfare analysis of their results. Training policies in the United States (US) aimed to mitigate the competitive inequality present in its workforce. According to (LaLonde, 1995), this type of inequality is usually associated with factors such as school dropout, prolonged unemployment, physical or mental disability, low proficiency in the native language and prison history.
Through its Title II component, the JTPA program advocated focusing on participants within the profile associated with such factors. Therefore, identifying the profile of those most affected and likely to accept the program is essential for its effective execution and evaluation.
The identification of the effects of policies for different groups of the population is fundamental for their formulation and improvement, being the focus of analysis in several studies during the 90s and covering programs in the areas of education (LaLonde, 1995), union affiliation (Card , 1996) and minimum wages (DiNardo et al. , 1996).
The JTPA program was created amid the growing concern of public entities between the 1980s and 1990s with the deskilling of the workforce in the USA and the increase in poverty rates observed in the country (LaLonde, 1995). II component of the JTPA provided training services to those selected, with the possibility of compliance or noncompliance with the program.
Title II of the JTPA included a JTPA National Study, providing data from its participants in 16 Service Delivery Areas ( SDAs ). These sites were selected non-randomly, being chosen for their diversity, willingness and ability to implement the experimental project and for the size and composition of the experimental sample they could provide. The article used the database from this evaluation to identify the existence of heterogeneities in the estimates of the effects of JTPA training for different income ranges (quantiles), in addition to also verifying the profile of participants who joined the program.
The JTPA began in October 1983 and was the US federal law that regulated the funding of federal training programs until the late 1990s. The main component of the JTPA was Title II, responsible for training the economically disadvantaged. Title II programs served about 1 million participants per year in the early 1990s, at an annual cost of approximately $1.6 billion. JTPA services were provided in SDAs, distributed in 649 locations around the US.
The program offered training services through community colleges, state employment services, community organizations, and private sector training agencies. The services comprised: (i) training in occupational skills, basic education or both; (ii) on-the-job training and/or job search assistance; (iii) other services that may have included probationary employment and/or a combination of services (i) and (ii).
Title II candidates used to be considered eligible for training if they faced any “barriers to employment.” Such barriers included long-term welfare use, high school dropout, 15 or more recent weeks of unemployment, limited English proficiency, physical or mental disability, reading proficiency below a 7th grade level, and/or an arrest record. Applicants were classified into one of five groups: adult men, adult women, female youth, non-detained male youth, and detained male youth. Among those selected, around 60% chose to participate in the program ( compliers ).
The original JTPA National Study database had 15,981 people. This database has continuous data on earnings for at least 30 months after participating in the program. This article focused on groups of adult men and women with data available on 30 months of their earnings after participation, consisting of two samples: one of 6,102 women, and the other of 5,102 men. The evaluation samples include candidates who applied between November 1987 and September 1989.
Although the selection of SDAs in the sample was not random, candidates were randomly selected for JTPA treatment, where 60% of those selected participated in the program. Individuals who did not receive treatment were generally excluded from JTPA services for 18 months (although they could participate in other programs). Sample data were artificially balanced to maintain a 2/1 treatment control ratio at each site due to distinct selection probabilities arising from the location of the SDAs. Treatment is indicated by a binary bariable that identifies participation in treatment ( compliance ). The instrumental variable is a binary variable that indicates receipt of the random training offer. The outcome variable defined for the evaluation consists of the sum of income over a 30-month interval, which is a precise measure of the program's lasting economic impact on participants' income. The vector of covariates consists of demographic, employment characteristics, and other information collected before randomization.
Least Squares (OLS) and quantile regression methods to present a comparison of the distribution of income between participants and non-participants. The results of both methods do not necessarily represent a causal relationship. - quantile treatment effect ( QTE) methods Unlike the quantile regression method, the QTE method is not limited by the assumption of exogenous treatment selection. The QTE method corrects the possible selection bias in quantile regression by more precisely identifying the group of adherents to the intervention ( compliers ) even in the absence of an observable counterfactual.
Both quantile regression and QTE methods allow establishing estimates of the outcome variable by conditioning the analysis to certain quantiles (points established along the cumulative distribution function of a random variable). In the case of this study, five quantiles were used: (i) two below the median (15% and 25%), (ii) the median (50%), and (iii) two above the median (75% and 85%) . Finally, the study presents estimates of the marginal distribution of the outcome variable of individuals without treatment.
The results found through OLS estimations indicate that the training increased earnings in the 30 months following the intervention by approximately both men and women. The results of the quantile regression estimates show that the results for quantiles below the median are much higher than those above the median, with such differences being more evident for men than for women.
The results of the second analysis show similar effects between the 2SLS estimate and the QTE for the median quantile, although the 2SLS estimator indicates greater precision. The training effects with the QTE estimator are substantially smaller in percentage terms and absolute magnitude compared to the results of quantile regressions. Additionally, no significant effects were identified for men in quantiles below the median. Unlike the results for men, QTE estimates for women demonstrated a significant effect for all quantiles analyzed, showing higher proportional effects in quantiles below the median.
The result found in the marginal distribution of income of individuals without treatment indicates a large discrepancy between the income of the counterfactual group of those who participated in the training and the control group. This suggests that the lack of effect in quantiles below the median may be associated with a selection bias, indicating that the program was not able to effectively select economically less advantaged men.
This article analyzed the impact of subsidized training services offered by the JTPA program on the income of its participants. Using quantile methodologies, the results demonstrated that participants had an increase in their income after joining the program. Women showed increases in earnings at various income levels in the sample, with greater gains observed at the lowest levels of the distribution. The results for men show that the policy only affected men with higher income levels in the distribution.
This evidence indicates that the JTPA was able to increase the income of women who took up the services offered. Considering the objective of reducing competitiveness inequalities, the program design appears not to be efficient in making men with lower income potential want to join the training services.
References
Card, D. (1996), “The Effect of Unions on the Structure of Wages: A Longitudinal Analysis”, Econometrica , Vol. 64 No. 4, p. 957.
DiNardo, J., Fortin, N.M. and Lemieux, T. (1996), “Labor Market Institutions and the Distribution of Wages, 1973-1992: A Semiparametric Approach”, Econometrica , Vol. 64 No. 5, p. 1001.
LaLonde, R.J. (1995), “The Promise of Public Sector-Sponsored Training Programs”, Journal of Economic Perspectives , Vol. 9 No. 2, pp. 149–168.
Check the Institution's registration in the e-MEC System here
© 2025 All Rights Reserved
