Reframing active labor market policy: Experimental evidence of training vouchers for unemployed

In February and March 2021, we launch a field experiment designed in cooperation and implemented by the Public Employment Service (PES) of Lower Austria (Arbeitsmarktservice Niederösterreich (AMS NÖ)). The aim is to increase training and employment among the unemployed by increasing participation in and completion of training programs. The intervention consists of an email newsletter that invites unemployed, who have been registered as unemployed for 3 to 12 months, to voluntarily contact the PES to arrange a consultation on training programs. The newsletters will be sent in three waves on February 9th, February 16th, and March 9th. Individuals are assigned to the waves depending on their duration of unemployment.

Three different treatment arms vary the type of information provided and the (perceived) autonomy that the unemployed have in choosing a training program. The different treatment conditions are as follows:

1. Group: control
2. Group: treatment with newsletter
3. Group: treatment with newsletter, and voucher
4. Group: treatment with newsletter, voucher, and information prime

Group 1 functions as the control group and is not contacted at all.

Group 2 receives a newsletter that includes an invitation to a consultation to discuss potential training programs with the PES' job counselor and provides information about existing financial incentives to start a training program.

In addition, groups 3 and 4 receive a voucher (Figure 2) worth € 15.000,-, which can be redeemed to take part in training programs provided by the PES. Alternatively, the voucher can be redeemed in consultation with the PES for any outside training for up to € 3.000,-. The groups receiving the voucher further obtain a list of typical training programs as part of the newsletter. This should motivate the unemployed in these two groups to already think about their preferred training program before the consultation at the PES. Finally, job counselors are instructed to take serious the voucher received by unemployed. The treatment is designed to increase self-initiative for the unemployed and raise awareness for the financial value of such training programs, thus inducing reciprocity.

Finally, group 4 receives in addition to the voucher an information treatment consisting of a list of occupations with the highest number of job vacancies. This information treatment is intended to counteract a frequently mentioned concern related to asymmetric information in the use of training vouchers: unemployed allegedly do not have enough information to make an informed choice about their optimal training program (Strittmatter, 2016). It will additionally increase (perceived) autonomy as it encourages even more to think about potential course choices before the consultation at the PES.

In general, all groups (including the control group) have access to the same training programs, both provided by the PES as well as outside training. The intervention, thus, consists of the variation in the type of information provided. Additionally, it varies the actual and perceived autonomy that the unemployed have in choosing their courses.

2021-02-10

2021-03-10

Our three primary outcomes are training completion, labor market status, and job quality.

The first one refers to whether or not a course/training program is completed. The second one differentiates between: unemployed, employed, out-of-labor force. We can, thus, differentiate between people exiting unemployment because they found a job and those exiting unemployment, because they dropped out of the labor force. This differentiation is important for drawing conclusions from the findings. Finally, job quality is important, which we proxy by combining earnings and employment stability into an indicator. Employment stability is defined via the duration of the employment spell after unemployment, independent of the specific job. This means, job-to-job transitions still count as employment, but job-to-unemployment transitions do not. We first normalize both variables and then combine them with equal weight to construct the job quality indicator. However, job quality is only observed for those who actually found a job. Therefore, we provide estimates for effects on job quality conditional on finding a job, but also for effects on the probability of having an above-average job quality indicator. The latter can account for this inherent endogeneity (Rothstein and Von Wachter, 2017).

Newsletter read + Clicks;
Training take-up;
Type of training

The following secondary outcomes are not per se desired outcomes, but can be seen as mechanisms leading to the primary outcomes described above. In this context we will look at whether or not the email was read, newsletter clicks, and course take-up. We will differentiate by the type of course to check whether the intervention also changes training choices.

The experimental design is based on a stratified randomized assignment to four different treatment and control goups. The randomization is conducted for each of three waves separately, thus practically stratifying by the three categories of unemployment duration dividing the three waves. In addition, we use the other stratification variables as specified in table 1 to construct strata, i.e. blocks. The treatment assignment is in a next step conducted randomly within these strata.

First analyses of short-term treatment effects are intended to be carried out with outcome data provided by the PES in several rounds in 2021. Longer-term effects are intended to be estimated with data provided by the PES each year until 2026. This will allow us to estimate long-term effects up to at least 5 years after the intervention.

Please refer to the pre-analysis plan for further details.

stratified randomization

individual level

No

-

12,000 unemployed

25% of unemployed in the sample per each of 4 treatment and control arms

PAP_training_vouchers_Austria.pdf