Job guarantee programs eliminate long-term unemployment
Kasy and Lehner 23 (Maximilian Kasy; Professor of Economics at the University of Oxford; Lukas Lehner; an economist at the Institute for New Economic Thinking at the Oxford Martin School (INET Oxford) and the Department of Social Policy and Intervention, University of Oxford; “Employing the unemployed of Marienthal: Evaluation of a guaranteed job program”; Oxford University; 4/21/23; https://maxkasy.github.io/home/files/papers/Jobguarantee_marienthal.pdf) [DTD]
5 Conclusion We conclude by summarizing our evaluation approaches and main findings, before discussing bigger-picture takeaways and avenues for future research. Our evaluation is based on several experimental and non-experimental contrasts, as summarized in Table 4. We use an experimental staggered roll-out design, comparing earlier and later entrants into the program, to identify direct effects of the job guarantee on the treated. We use a synthetic control approach at the municipality level to identify spillover effects of the job guarantee on the untreated, as well as the average total effect of the job guarantee on the labor market. And we compare program participants to observationally similar individuals in control towns, to separate out anticipation effects, and to estimate the long-term effects of the job guarantee. Assignment to the two groups (early and late entrants) in the experimental comparison is based on pairwise matched random assignment. This approach allows us to increase the precision of our estimates by making the two groups observationally as similar as possible. This reduces standard errors relative to conventional random assignment, which is particularly relevant given our small sample size. Both the pairwise matches and the synthetic control weights were pre-registered. This ties our hands and prevents us from cherry-picking results, including for the observational comparisons in our evaluation. Our inference approach is primarily based on randomization inference (permuatation inference). This guarantees finite sample validity without any asymptotic approximations. In Appendix A, we also report conventional confidence intervals, using robust standard errors; the conclusions remain unchanged. Turning to our empirical findings, a first remarkable fact is that everyone offered a job after the 8-week training phase accepted this job. In our experimental comparison, we find large positive effects of the job guarantee on participants’ economic and non-economic well-being. This includes effects on employment, income, and income security, which are expected given the nature of the program. This also includes large positive effects on time structure, activity, social contacts, collective purpose, and social status. These non-econonomic effects of employment have been discussed in the sociological literature, mostly in the context of observational studies, but have received less attention in economics. We do not find effects on physical health and economic preferences, including time and risk preferences, reciprocity, altruism, and trust. The estimated effects persist over time. We further find a large reduction of municipality-level unemployment, which is driven by a near-elimination of long-term unemployment. There appears to be no increase of short-term unemployment. While we were not able to independently verify program costs, it is estimated that the total cost per eligible particpant and year was around EUR 30,000, of which around EUR 20,000 were wages, taxes, and social insurance contributions for participants. These findings have implications for both policy and future research. First, our findings suggest that the job guarantee is a promising policy instrument to reduce long-term unemployment, and to improve the well-being of the unemployed. Crucial for this conclusion was our focus on participant well-being. This contrasts with a focus on market employment as the primary outcome for most existing evaluations of active labor market programs.