Mette Rasmussen forsvarer sin ph.d.-afhandling

Mette Rasmussen forsvarer sin ph.d.-afhandling:"Empirical Essays on the Effectiveness of Active Labor Market Policies: Caseworkers, Vocational Training & Statistical Profiling"

Kandidat

Mette Rasmussen

Titel: "Empirical Essays on the Effectiveness of Active Labor Market Policies: Caseworkers, Vocational Training & Statistical Profiling"

Tid og sted: 4. juni 2021 kl. 16:00 i lokale CSS 26.2.21.
Link til at logge på overværelse af forsvaret pr. Zoom følger her: https://ucph-ku.zoom.us/j/9536476672

En elektronisk kopi af afhandlingen kan fås ved henvendelse til: charlotte.jespersen@econ.ku.dk

Bedømmelsesudvalg

  • Professor Søren Leth-Petersen, Økonomisk Institut, Københavns Universitet, Danmark (formand)
  • Professor Sally Sadoff, Rady School of Management, UC San Diego, USA
  • Professor Bas van der Klaauw,  VU University Amsterdam, Holland

Abstract

This PhD dissertation consists of three self-contained chapters that all study how different types of active labor market policies can help unemployed jobseekers transition out of unemployment. All three chapters rely on novel micro data provided by the Danish Agency for Labor Market and Recruitment.


The first chapter investigates the importance of caseworker quality for jobseeker transitions out of unemployment. I combine a novel data set on meetings between jobseekers and caseworkers with rich Danish administrative data. To identify the individual impacts of caseworkers, I exploit that many jobcenters assign caseworkers to jobseekers based on their birthday. I verify that this effectively corresponds to a quasi-random assignment mechanism. The chapter offers three sets of results. First, I find that variation in caseworker quality is substantial and can explain about 6% of the heterogeneity in unemployment spells within a jobcenter and year. Assignment to a caseworker, who is one standard deviation above the mean, reduces the unemployment spell by about one week. Second, I find no smoking gun suggesting that this is at the expense of subsequent labor market performance. Rather, the caseworkers that reduce unemployment spell length place the jobseekers in jobs that are of similar quality, in terms of wages, hours and stability. As a result, these jobseekers have on average accumulated an additional 7,500 DKK (1,200 USD) and 35 working hours after two years. Third, I show that the variation in caseworker quality not necessarily is driven by unobserved personality traits only. Namely, I find that the high quality caseworkers tend to be more 'pro-active': They meet earlier and more frequently with the jobseekers, assign them earlier to training and tend to increase their use of network and unsolicited job search. Overall, this paper suggests that it would be Pareto improving to teach all caseworkers these pro-active strategies.

The second chapter (co-authored with Anders Humlum) investigates whether vocational training of jobseekers can help them re-attach to the labor market. To investigate this, we combine the novel data on caseworkers from the first chapter with new data on caseworker assignments to training. Specifically, we estimate the effectiveness of vocational training using a caseworker leniency instrument. This instrument exploits that i) jobseekers are quasi-randomly assigned to caseworkers, and that ii) caseworkers differ in their propensities to assign jobseekers to vocational training. Using our caseworker leniency instrument, we cannot reject that training courses on average have zero impacts on labor market outcomes after one year. In contrast, OLS regressions show strong negative correlations between training and employment, indicating that it is jobseekers with adverse job prospects who select into training. To investigate whether vocational training is more beneficial for workers who are exposed to rapid structural change, we zoom in on jobseekers whose previous jobs were in manufacturing. Although the estimates become more noisy, we find economically significant long-run benefits to vocational training for former manufacturing workers. This suggests that there is large heterogeneity in the benefits of training, which potentially could be reaped by better targeting of courses to workers.

The third chapter (co-authored with Nikolaj Harmon and Robert Mahlstedt) evaluates the effect of using statistical profiling tools to inform workers about their individual risk of long-term unemployment. In Denmark, a Machine Learning algorithm informs both newly unemployed jobseekers and their caseworkers of whether they belong to a group with a high risk of remaining unemployed for more than six months. Leveraging age discontinuities in the Machine Learning algorithm, we use a regression discontinuity design to estimate the effect of being reported as being in high risk. We estimate that jobseekers marginally flagged as high-risk are between 5-14% less likely to be unemployed after 6 months. After 12 months however this difference has disappeared. Unfortunately, standard validity checks suggest that the identifying assumptions of our regression discontinuity design may not hold. While our results thus points to statistical prediction tools as a promising way to speed up unemployment exit, more evaluation is necessary to reach any firm conclusion.