Natural and classical experiments in Swedish labour market policy

Author: Pathric Hägglund, And

Summary of Dissertation series 2006:1

Effects of Changes in the Unemployment Insurance Eligibility Requirements on Job
Duration — Swedish Evidence
This paper investigates the impact of the unemployment insurance (UI)
entrance requirement on employment duration among earlier unemployed in
Sweden. I exploit changes in the rule taking place in 1994 and 1997 to study
behavioural adjustments in the timing of job separation between 1992, 1996,
and 1998. Performing across-year analyses, I find evidence of clustering of job
exits at the time of UI qualification. By using predicted hazard rates for each
week, I calculate an approximate 2.9-week extension in average employment
duration between 1996 and 1998 due to the 5-week prolonging of the entrance
requirement.
Job-search Assistance Using the Internet – Experiences from a Swedish Randomised
Experiment
This paper reports the experience from a randomised experiment offering
voluntary job-search assistance on the Internet to job seekers at Swedish public
employment offices. The purpose is to, i) investigate to what extent the
evaluation design manages to avoid common difficulties in experimental
evaluation, ii) assess the effect of the programme on the employment outcome,
and iii) use the nonbiased experimental results as a benchmark evaluating the
performance of frequent nonexperimental estimators. I find that the evaluation
design successfully circumvents inherent difficulties in the experimental
approach, such as ethical concerns, bureaucratic behaviour and randomisation
bias. However, the voluntariness of the programme caused severe compliance
problems in terms of both no-shows and dropouts. This is accounted for by
analysing the effect of the “intent-to-treat” (the policy parameter of most
interest), which is close to zero. Studying the effects of various doses of actual
treatment, using an nonexperimental instrumental variable model, I fail to
reject the hypothesis of a zero programme effect. Finally, a methodological
comparison suggests that standard nonexperimental techniques succeed in
reproducing the nonbiased experimental results.
Are there Pre-programme Effects of Swedish Active Labour Market Policies – Evidence
from Three Randomised Experiments
This paper takes advantage of unique experimental data from three
demonstration programmes in 2004 to investigate pre-programme incentive
effects of active placement efforts at the employment offices in Sweden. The
exit rate from unemployment between referral to and start of the programme
services is compared between UI eligible experiment and control group
members. The results are mixed. In one of the experiments, targeted towards
a broad group of UI receivers, arranged job-search activities in groups
combined with increased monitoring of job-search efforts generated a 38 per
cent increase in the escape rate from unemployment in the weeks leading up
to programme start. This translates into an almost two-week reduction of the
ongoing UI spell. Referrals to increased monitoring alone did not have the
same effect on exit behaviour. In the other two experiments, targeted towards
youth and highly educated respectively, referrals to active placement efforts
had no effect on the pre-programme outflow.

Effects of Changes in the Unemployment Insurance Eligibility Requirements on Job Duration — Swedish Evidence

This paper investigates the impact of the unemployment insurance (UI)entrance requirement on employment duration among earlier unemployed in Sweden. I exploit changes in the rule taking place in 1994 and 1997 to study behavioural adjustments in the timing of job separation between 1992, 1996, and 1998. Performing across-year analyses, I find evidence of clustering of job exits at the time of UI qualification. By using predicted hazard rates for each week, I calculate an approximate 2.9-week extension in average employment duration between 1996 and 1998 due to the 5-week prolonging of the entrance requirement.

Job-search Assistance Using the Internet – Experiences from a Swedish Randomised Experiment

This paper reports the experience from a randomised experiment offering voluntary job-search assistance on the Internet to job seekers at Swedish public employment offices. The purpose is to, i) investigate to what extent the evaluation design manages to avoid common difficulties in experimental evaluation, ii) assess the effect of the programme on the employment outcome, and iii) use the nonbiased experimental results as a benchmark evaluating the performance of frequent nonexperimental estimators. I find that the evaluation design successfully circumvents inherent difficulties in the experimental approach, such as ethical concerns, bureaucratic behaviour and randomisation bias. However, the voluntariness of the programme caused severe compliance problems in terms of both no-shows and dropouts. This is accounted for by analysing the effect of the “intent-to-treat” (the policy parameter of most interest), which is close to zero. Studying the effects of various doses of actual treatment, using an nonexperimental instrumental variable model, I fail to reject the hypothesis of a zero programme effect. Finally, a methodological comparison suggests that standard nonexperimental techniques succeed in reproducing the nonbiased experimental results.

Are there Pre-programme Effects of Swedish Active Labour Market Policies – Evidence from Three Randomised Experiments

This paper takes advantage of unique experimental data from three demonstration programmes in 2004 to investigate pre-programme incentive effects of active placement efforts at the employment offices in Sweden. The exit rate from unemployment between referral to and start of the programme services is compared between UI eligible experiment and control group members. The results are mixed. In one of the experiments, targeted towards a broad group of UI receivers, arranged job-search activities in groups combined with increased monitoring of job-search efforts generated a 38 percent increase in the escape rate from unemployment in the weeks leading up to programme start. This translates into an almost two-week reduction of the ongoing UI spell. Referrals to increased monitoring alone did not have the same effect on exit behaviour. In the other two experiments, targeted towards youth and highly educated respectively, referrals to active placement efforts had no effect on the pre-programme outflow.