Experimental Design Details
The randomized trial uses a series of sub-trials to evaluate the impact of the design of the basic compensation and the performance-based compensation. Randomization of participants occur at several important margins, that all allow for analyses of different types of questions. The starting point is a statistical profiling tool used to assign job seekers to KROM. It estimates a profiling score for the job seeker’s proximity to the labor market. According to pre-specified thresholds of the profiling score, job seekers are distributed to different tracks of the KROM program. The different tracks offer different levels of compensation. More specifically, participants who are estimated to be further from the labor market, according to the statistical profiling tool, generate a larger compensation to the private provider than participants closer to the labor market. Participants who are too close to or too far from the labor market are not part of the target population for KROM and are supposed to receive the normal services by the public provider, the PES.
This trial involves randomization of job seekers close to the thresholds, who are randomized between the KROM-tracks. These randomization zones are defined in terms of score thresholds and they also vary along the unemployment duration dimension. Hence, individuals with the same score from the statistical profiling model and the same unemployment duration will randomly receive different services or generate different compensations for the private providers. The KROM design includes several different randomization zones, that all can be used to answer important questions.
Figure 1 illustrates the research design (see supporting material). Each coloured dot is an estimated profiling score (y-axis) for a job seeker profiled at a given duration of unemployment (x-axis, days in unemployment, dagar). The solid lines correspond to the thresholds. The different colours show the five outcomes: green (Not KROM, too close to the labor market), yellow (KROM, track A), orange (KROM, track B), red (KROM, track C), and blue (Not KROM, too far away from the labor market). Around each threshold in Figure 1 there are grey dotted lines, showing the randomization zones. If the job seeker´s score is within a randomization zone, participation in the tracks surrounding the threshold is randomly drawn. This design makes sure that on each side of each threshold there are similar job seekers (in terms of job finding probabilities) who get different treatment in terms of private or public provision or in terms of compensation to private providers.
The difference between the tracks is the compensation they generate to the private providers. The contracted service content is the same between tracks, and providers are in general free to implement the services they find appropriate. In all tracks, compensation consists of (1) a basic compensation at assignment, (2) a performance-based compensation when the participant gains employment or starts education, and (3) a quickness bonus granted if the participant acquires employment or education within 6 months. The maximum compensation per participant varies from SEK 34,820 in track A to 61,760 in track C. In addition, the middle track B consists of two different compensation schemes, where each is implemented in about half of the local labor markets.
The setup of KROM has been similar over time but the thresholds have changed on a few occasions. These changes allow us to identify the effects of private provision and different compensation schemes for a larger range of job seeker characteristics than if the thresholds had remained fixed. Another feature of the program that can be exploited for identification is the gradual roll-out across municipalities. Among other things, this allows for explicit evaluations of displacement effects of job search assistance, which has been an important topic in the recent job search assistance literature.