Voluntary unemployment insurance for the self-employed

Treated individuals receive an information brochure. This brochure consists of four pages. The front page consists of a picture, the title ``Unemployment Insurance for Self-Employed'', and the name of the research institute of the FEA. The second and the third page contain detailed information about the VUI. It starts with listing the general advantages, clarifies the time limit for registration, and details the amount of UI benefits and contributions and how one can leave the insurance. The back side of the brochure contains a summary of the different aspects of the VUI. This brochure is complemented with an insurance application form which the self-employed might fill out and either send it to the local office of the FEA or hand it in personally.

2016-02-15

2017-08-15

(1) Decision to enter the VUI;
(2) Duration in VUI;
(3) Duration of self-employment;
(4) Unemployment

Based on administrative data, we observe for every individual in our sample whether he or she entered the VUI. Further, we observe monthly contributions paid by the insured self-employed. This allows us to estimate the effect of the information treatment on participation in VUI over time.

In the administrative data, we only observe periods of subsidized self-employment. These subsidies have a maximum length of 15 months. Our measurement of the duration of self-employment will be based on two surveys. We have conducted a first survey around 20 months after entry into subsidized self-employment. A second survey takes place around 7 and 8 years after the entry into self-employment. The data of this second survey will be available in the fall 2024. In both surveys we ask the interviewees whether they are still self-employed, and if not when the self-employment episode started in 2016/17 has ended.

Based on administrative data, we observe for everyone in our sample whether he or she is registered as unemployed. This information allows us to calculate weekly shares of unemployed individuals in the treatment and control group and to compare them over time.

We will explore effect heterogeneity with respect to the predicted probability of entering the VUI, the predicted probability to leave self-employment, and the predicted probability to enter long episodes of unemployment and to receive means-tested unemployment benefits. The prediction models will be based on a rich set of characteristics observed in administrative data before the date of randomization. Further, we will explore effect heterogeneity of the intervention with respect to the situation on the local labor market.

(1) Dependent employment;
(2) Receipt of UI benefits;
(3) Number of employees;
(4) Investment during self-employment;
(5) Income from self-employment;
(6) Knowledge and subjective evaluation of the VUI

Based on administrative data, we observe for everyone in our sample whether he or she is regularly or marginally employed. Further, we observe the amount of unemployment benefits as well as the type of benefits (unemployment insurance benefits or means-tested benefits). This information allows us to calculate weekly shares of employed individuals and of individuals receiving different types of unemployment benefits in the treatment and control group and to compare them over time.

We do not have direct information in the administrative data about the number of employees hired by the self-employed. Our first measure for this will be based on the two surveys conducted around 20 months and around 7 and 8 years after the start of the subsidized self-employment. For those who participate in these surveys, we will observe the number of individuals employed in their firms at the time of the interview. Second, we will try to explore data from administrative sources. The main idea is to merge data from our sample with administrative data on firms based on the timing these firms appear in the administrative records, the geographical distance between the place of residence of the self-employed and the location of the firm, and the name of the founder and the name of the firm or the contact person in this firm. The use of this data will depend on the quality of this data merge. We will evaluate the quality based on a comparison with the information from the survey data. If the quality is sufficiently high, we will observe the number of employees employed in the firms on June 30 for each calendar year. If this linkage with administrative data based on names of the founder is of sufficient quality, we might also be able to investigate other outcomes including mergers, takeovers, and profits.

Our measurement of the investment into self-employment will be based on the survey we have conducted 20 months after entry into self-employment. In both surveys we ask the interviewees to report their expenditures and revenues related to the self-employment activities. We will use these measures to construct income measures.

Information on knowledge about the VUI and subjective evaluations of the VUI scheme stem from a smaller survey conducted around 5 months after entry into subsidized self-employment among the last cohorts of founders.

We will explore effect heterogeneity with respect to the predicted probability of entering the VUI, the predicted probability to leave self-employment, and the predicted probability to enter long episodes of unemployment and to receive means-tested unemployment benefits. The prediction models will be based on a rich set of characteristics observed in administrative data before the date of randomization. Further, we will explore effect heterogeneity of the intervention with respect to the situation on the local labor market.

Our information treatment focuses on the group of founders who left unemployment and received a start-up subsidy from the FEA. This group can be identified from administrative records. Founders can only enter the VUI within the first three months after entering self-employment. Therefore, the information treatment needed to be sent out quickly after the business was founded.

We randomly selected 50% of each monthly cohort of founders from February 2016 until August 2017 and sent them an information brochure and an insurance application form together with a cover letter. The remaining 50% did not receive any information and serve as a control group.

Not available

Randomization at the individual level, done by a computer program. Randomization has been conducted within cells by gender and region (East and West Germany).

Individual

No

32800 unemployed who entered subsidized self-employment participated in the experiment.

32800 unemployed who entered subsidized self-employment participated in the experiment.

Treatment group: 16400
Control group: 16400

Based on the administrative data, we have a minimum detectable effect of 1.28 percentage points for the probability of participating in the VUI, assuming an average participation rate in the control group of 22%.