2021-01-01
Measuring returns to vocational education: evidence from France
Todor Tochev
Vocational training is a widely used active labour market policy. On average, OECD countries have spent more than 0.1% of GDP on vocational training measures in 2016 (OECD, 2019). In France, the Plan d’investissement dans les Competences is a five-year plan launched in 2018 that will commit approximately €15 billion to improving and expanding vocational and labour market training. However, the literature evaluating the outcomes of these policies has thus far been constrained by the availability of data. Most work on the subject has used non-experimental matching methods, but whose estimates may not be invariant to assumptions about how individuals are selected into training. Randomized control trials (RCTs) may be more robust to this criticism, but due to practical constraints are often limited in scope to a single programme, a short time frame, or a relatively small scale. Additionally, this also means that very few RCTs investigating vocational training have been conducted in developed countries, and in Europe in particular. In both types of approaches, there is an important and thus far unexplained dispersion in treatment effects estimates (Card et al., 2018; Levy- Yeyati et al., 2019; McKenzie, 2017).
Several explanations have been suggested. First, the quantity of interest is almost always the average treatment effect of the intervention. However, this can mask substantial heterogeneity according to the type of training received. Indeed, the recent literature on returns to education has demonstrated a wide variation in labour market payoffs associated with different fields of education (Altonji and Zhong, 2020; Kirkeboen et al., 2016). A second possibility is that socio-demographic characteristics, such as gender, play an important role in within-training differentials (Kruppe and Lang, 2018). Finally, trainings differ in terms of the types of skills they impart. An important literature investigating inequality and technical change suggests that the specific skills possessed by individuals, as opposed to simply their educational attainment level, also play a role in explaining labour market outcomes (Acemoglu and Autor, 2011).
The proposed research will evaluate the returns to vocational training provided by the Agence pour la formation professionnelle des adultes (Afpa) to job seekers in France. Prior to enrolment in a training, all candidates are required by Afpa to undergo a series of assessments including a set of problem solving and logic tests. Using a fuzzy regression discontinuity design, we will take advantage of the existence of training-specific test score thresholds which form part of the criteria for acceptance into the training in order to estimate the Local Average Treatment Effect (LATE) for individuals with scores near the test thresholds. The outcomes of interest are the probability of being employed, the probability of having a long-term employment contract, the probability of working in the same industry as the training applied for, and wage earnings. The test score data will be combined with administrative data from Afpa and from Pole Emploi (FHS and MMO databases). This will enable exploration of heterogeneity across trainings, as well as across other dimensions such as gender and the skill content of the training.
Membres du projet :
- Todor Tochev, University of Warwick
État du projet : en cours
Ce projet est financé par la chaire Sécurisation des Parcours Professionnels.
Vocational training is a widely used active labour market policy. On average, OECD countries have spent more than 0.1% of GDP on vocational training measures in 2016 (OECD, 2019). In France, the Plan d’investissement dans les Competences is a five-year plan launched in 2018 that will commit approximately €15 billion to improving and expanding vocational and labour market training. However, the literature evaluating the outcomes of these policies has thus far been constrained by the availability of data. Most work on the subject has used non-experimental matching methods, but whose estimates may not be invariant to assumptions about how individuals are selected into training. Randomized control trials (RCTs) may be more robust to this criticism, but due to practical constraints are often limited in scope to a single programme, a short time frame, or a relatively small scale. Additionally, this also means that very few RCTs investigating vocational training have been conducted in developed countries, and in Europe in particular. In both types of approaches, there is an important and thus far unexplained dispersion in treatment effects estimates (Card et al., 2018; Levy- Yeyati et al., 2019; McKenzie, 2017).
Several explanations have been suggested. First, the quantity of interest is almost always the average treatment effect of the intervention. However, this can mask substantial heterogeneity according to the type of training received. Indeed, the recent literature on returns to education has demonstrated a wide variation in labour market payoffs associated with different fields of education (Altonji and Zhong, 2020; Kirkeboen et al., 2016). A second possibility is that socio-demographic characteristics, such as gender, play an important role in within-training differentials (Kruppe and Lang, 2018). Finally, trainings differ in terms of the types of skills they impart. An important literature investigating inequality and technical change suggests that the specific skills possessed by individuals, as opposed to simply their educational attainment level, also play a role in explaining labour market outcomes (Acemoglu and Autor, 2011).
The proposed research will evaluate the returns to vocational training provided by the Agence pour la formation professionnelle des adultes (Afpa) to job seekers in France. Prior to enrolment in a training, all candidates are required by Afpa to undergo a series of assessments including a set of problem solving and logic tests. Using a fuzzy regression discontinuity design, we will take advantage of the existence of training-specific test score thresholds which form part of the criteria for acceptance into the training in order to estimate the Local Average Treatment Effect (LATE) for individuals with scores near the test thresholds. The outcomes of interest are the probability of being employed, the probability of having a long-term employment contract, the probability of working in the same industry as the training applied for, and wage earnings. The test score data will be combined with administrative data from Afpa and from Pole Emploi (FHS and MMO databases). This will enable exploration of heterogeneity across trainings, as well as across other dimensions such as gender and the skill content of the training.
- Todor Tochev, University of Warwick
État du projet : en cours
Ce projet est financé par la chaire Sécurisation des Parcours Professionnels.