(Not) Thinking About the Future: Maternal Labor Force Participation

To test whether learning about the long-term financial consequences of having a reduced workload impacts mothers’ behaviors, we designed the following intervention material:
1) Informational video containing three main dimensions of objective information about the long-term financial consequences of having a reduced workload that are relevant for teachers.
a. Information on the total earnings lost in the long-term
b. Information on the financial consequences for pensions
c. Information on the financial risk in case of adverse events (such as divorce)
The video also puts the lifetime financial losses in perspective to the short-term childcare costs.
2) Online tool: together with a Swiss bank, we designed an online calculator that allows users to simulate the long-term financial consequences of workload reductions for their individual case. We again provide information about the total earnings loss, the pensions loss, and the comparison with the short-term childcare costs. The tool automatically gathers the individual calculations' data, so we can link them to the individual.

The control group receives a video of similar duration but on an unrelated topic. We use videos created by the national public television as part of their regular programming (https://www.srf.ch/). We randomize the control group into three videos:
- a video containing information about the gender pay gap in Switzerland
- a video with information on proposed tax breaks for families in Switzerland
- a video discussing the current structure of costs of renting or buying housing in Switzerland

In the online baseline survey, participants watch their assigned video. The treatment group receives the online calculator by email directly after completing the baseline. For outcomes in the baseline survey, participants will have received the information in the treatment video only.

2022-11-16

2023-03-31

We have three primary outcomes dimensions:
a) Financial empowerment: women’s financial awareness and the likelihood of considering long-term financial factors.
b) Workload aspirations: teachers’ desired future workload
c) Workload: teachers’ actual workload in the next school year (administrative data)

We will have three families of outcomes where we aggregate groups of questions and construct an index for each family. This will reduce the number of hypotheses tested. We will also separately analyze the impact for the incentivized question where participants are asked to choose between a voucher for a financial consultation or a popular Swiss online store.

We are also interested in the heterogeneity of the background characteristics of respondents and will thus look at outcomes by sub-groups.
In particular, for the financial dimension, we will consider heterogeneity by financial awareness at baseline, the workload at baseline, and marital status.
For the labor force participation dimensions (both desired and actual LFP), we will analyze heterogeneous effects by financial awareness at baseline, gender norms, the workload at baseline, the self-reported difficulty of organizing family life, self-reported flexibility in adjusting their workload, beliefs about the quality of external care, and by mother’s age and age of the youngest child.

We are collaborating with cantonal Departments of Education in Switzerland. In three cantons, we have access to the contact information of female staff aged between 25 and 50 in cantonal employment contracts, including kindergarten, primary, and secondary school teachers, for the school year 2022/23. We invite teachers to our study by email (if available) and by letter (all).

We conduct an individual-level randomization of teachers into the treatment and control groups. Due to the potential presence of spillovers between teachers working in the same school, we add a hold-out control group to this design. For the largest canton (in the following: Canton 1), we thus implement a two-stage randomization in the following way:

1. In the first stage, we randomize 2/3 of the schools into treatment schools and 1/3 into control schools (from now on referred to as the “pure control group”). We stratify the sample by terciles of school size (proxied by the number of female teachers aged 25-50 years working in each school based on the Department of Education contact list), school type (primary or secondary), and type of municipality (rural, semi-urban, city).

2. In the second stage, we randomize teachers in treatment schools at the individual level to treatment or control when participating in the survey. The individual-level randomization occurs during the survey, just before the intervention video starts playing. We assign ½ of the teachers to treatment and the rest to control.

For the rest of the cantons, we only randomize at the individual level (i.e., only the second step) and stratify by canton. Within the control group (both for the pure control and for the control), we randomize three control videos with equal probability. Our main focus for the analysis is the individual-level randomization of teachers into the treatment and control groups.

Not available

Individual-level randomization is performed with Qualtrics. The school-level randomization by computer.

Our main randomization is at the individual level. For Canton 1, we implement a two-stage randomization in the following way: we first randomize schools into treatment and control ("pure control group"), and in a second stage, we randomize at the individual level teachers in treated schools into treatment or control during the survey.

Yes

Canton 1: we invite 9281 teachers to our study. We randomized 330 schools to the treatment group and 165 schools to the pure control group. Our sample's final number of schools will depend on the response rate. The same applies to the final number of teachers who will be randomized to treatment and control within treatment schools.
Canton 2: we invite around 1900 teachers from 86 schools. Again, the final number of teachers randomized to treatment and control will depend on the response rate.
Canton 3: we invite around 548 teachers from 93 schools.

2000 teachers

Individual level treatment assignment in all 3 cantons: In total, we expect around 1000 teachers per treatment arm for the individual level.
(For the school-level randomization in canton 1 to analyze potential spillovers: 330 schools are in the treatment group and 165 schools are in the pure control group.)