Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
The effect sizes are expected to be low. First, the length of data collection is between 1.5 to 2 years, which leaves a short window for any pregnancies. Second, the control group are also job seekers, thus having a child may not be the first alternative choice for this group. Third, the job may not reduce actual fertility, only delay it. This is why we also include preferred fertility as an outcome variable.
Based on the effect sizes from comparible studies (discussed in the analysis plan attached) and the considerations of a shorter time period and a job-seeking control group, we believe the effect size of actual fertility to be about 1 percentage point (3 percentage points after 18 months) with standard error of 0.02.
On preferred fertility, the time percpective may not be important for the effect size. We therefore expect similar effectsize as Jensen (QJE, 2012), who found that women in the treatment villages reported that they wanted approximately 0.35 (st.error 0.078) fewer children compared to the control group mean of 3.0.
By use of power calculations for the actual and desired fertility with alpha-level 0.05 and sample size 1700, our study is underpowered which a power of
0.09 for finding the true eect size of actual fertility, while it is well powered at 1.00 for nding the true effect size for preferred fertility.