Prevalence and potential determinants of unregistered work in Germany – Evidence from a double-list experiment

2022-05-01

2023-02-28

The primary outcome in our study is the share of respondents (and certain relevant subgroups of respondents) with experience of unregistered work.

The implementation of list experiments critically relies on the random assignment of respondents into treatment and control group. The control group is confronted with a list of non-sensitive items that represent general situations at the labor market that respondents might have experienced. The treatment group faces the same list plus the sensitive item of interest, here, the experience of having worked without registration at least once in 2021. In both groups, the participants do not have to state which of the items they have experienced already, only how many in total. From the comparison of mean counts and the joint distribution of answers in each group, we can infer the share of respondents who experienced unregistered work in 2021. To increase the robustness of our findings we implement a double list experiment. That is, we run two list experiments with slightly different, but thematically coherent content in the non-sensitive items, the same sensitive item, and a switch of treatment and control groups between experiment 1 and 2.

In the first list experiment we focus on experiences on the German labor market in 2021. The treatment and control groups are confronted with the following items:

(1) Control group: 4-items related to general situations on the labor market
o receiving a higher salary
o not having searched for a job
o having searched for a (new) job
o learned new skills

(2) Treatment group: same 4 items + sensitive item: worked in a job without registration

In the second list experiment we focus on employment-related experiences in 2021. The treatment and control groups are confronted with the following items:
(1) Control group: 4-items related to general situations on the labor market
o Not being employed the whole year
o Worked in job that required skill updating or training
o Worked in a job that was physically or mentally burdensome
o Worked in a job that required (temporarily) working from home

(2) Treatment group: same 4 items + sensitive item: worked in a job without registration

In list experiments, survey participants in both groups (treatment and control group) report one number as outcome. This number is the total count of items they have experienced. In our list experiments, we present them with four non-sensitive control items and one sensitive item. Thus the total counts (outcomes) for the participants in the control group Y0 can range between 0-4, and in the treatment group Y1 between 0-5. The difference in the means between the average counts in the treatment and controls groups identify the share of respondents with experience of unregistered work in both list experiments. If the experimental design is robust, we should observe the same share in both list experiments.

Computer-based randomization implemented in the survey tools.

Individual

Yes

We plan to run the list experiment in three different samples of the population (natives, migrants, refugees). Thus, we will have three clusters.

We aim for roughly 11.000 survey participants in total (5.000 German, 3.000 migrants, 3.000 refugees).

For each cluster, we aim at having roughly the same number of respondents in control and treatment group.

That is: We plan to have 5,500 respondents on both sides.