Experimental Design
Background information:
The integration management program in our cooperating city is implemented in three consecutive waves. In the first wave (roughly May 2018 to January 2019), the most urgent cases were included, i.e., individuals who needed more intensive assistance for themselves or family members at that point in time. For the other two waves, no special allocation mechanism was foreseen on the part of the city. We therefore agreed with the officials to do the allocation randomly in order to evaluate the effectiveness of the program.
Individual steps:
1. The officials from the integration management program assemble a list of potential participants. That is, they list all refugees in their jurisdiction who have not been part of the program yet and for whom there is no immediate reason or need to put them in a specific wave.
2. We provide a list of randomly generated numbers to the officials. The numbers are matched with the IDs of the potential participants, such that they can be randomly allocated to treatment and control group on the household level and stratified by place of living (16 refugee residences, publicly provided apartments, and private apartments).
3. Treatment group individuals participate in the integration management program between February and October 2019, while control group participants continue to be supervised by social workers or can get help at public counsel offices throughout the same period.
4. We accompany both integration managers and social workers for a couple of days to get first-hand knowledge about the different approaches used (summer 2019).
5. In October 2019, the supervising integration managers collect the final data about the outcomes and background of the treatment group participants.
6. After that, control group participants enter the integration management program (roughly November 2019 to June 2020), while treatment group participants can turn to social workers or public counsel offices for advice, if still needed.
7. At the beginning of wave 3 (November/December 2019), the supervising integration managers collect the information about the control group participants.
8. The officials from the integration management combine the data for treatment and control group, remove any personal and identifying information, and send the resulting data set to us (January 2020).
9. The researchers aim at getting funds for a post-participation survey to obtain information on potential channels of influence at play. To protect the personal information of the participants, the questionnaires will be sent out and collected through the city officials (roughly February/March 2020).
Analysis:
The identification of the treatment effect (i.e., the effect of being supervised by the integration management program) relies heavily on the experimental setup presented above, in particular, the random allocation of participants into the comparison groups. That is, we measure the effect on the three main outcome variables as the difference in the respective means of the two groups, assuming that they are almost identical in all other aspects (observed and unobserved). This is equivalent to running regressions of the respective outcome on a simple treatment indicator. Additionally, we will add potentially relevant control variables to check whether the estimates remain stable, i.e., whether the random allocation of participants to the two comparison groups worked well.
Furthermore, we will examine whether different subgroups react differently to the treatment by separately introducing interaction terms between the treatment variable and indicators for various potentially interesting subgroups, especially women, households with children, individuals with a low formal education, and people living in private apartments.