2025-09-12
rich descriptive evidence on women’s employment in Uganda and gender differences in employment.
an overview of existing policy efforts (e.g., Skilling Uganda, UWEP, Affirmative Action in higher education, Equal Opportunities Policy, UPE/USE)
only two cross-sectional surveys (2016/17 and 2021) \(\rightarrow\) causal inference is challenging.
but there is variation in timing, geography, eligibility and intensity \(\rightarrow\) policies in Section 2.1 offers ways to construct quasi-experimental comparisons:
Policy Rollouts by Cohorts (Affirmative Action & Education Policies):
The affirmative action in higher education and UPE/USE reforms created different “treated cohorts” of women versus men.
By comparing younger cohorts (who were exposed during school age) with older ones (who were not), it might be possible to examine whether the gender gap in employment, education or formal jobs narrowed more for the “treated” groups.
Geographic Variation in Program Reach (UWEP, Skilling Uganda, UPE/USE): some programs disproportionately benefited certain regions (e.g., UWEP reaching more women in urban Kampala/Wakiso).
construct synthetic “treatment” variables based on exposure: e.g., district-level intensity of UWEP beneficiaries, cohort eligibility for UPE/USE, or whether respondents fall in the affirmative action age group.
use DID logic: (Women in treated group – Men in treated group) vs. (Women in control group – Men in control group) before and after 2016/17.
sector-based analysis: COVID shocks can be leveraged by comparing outcomes for men and women in “face-to-face” service jobs vs. agriculture.
with repeated cross-sections (not panels), causal claims are limited.
BUT careful design using policy heterogeneity + group comparisons + pre/post structure can still yield credible suggestive evidence.
use uneven exposure as a treatment dimension in DiD setup:
Treated vs. Control Groups:
Treated group: Districts with high UWEP penetration (measured by number of female beneficiaries per 1000 women).
Control group: Districts with low or no UWEP penetration.
Before vs. After:
Before: 2016/17 Labour Force Survey.
After: 2021 Labour Force Survey.
Gender Dimension (Double Difference):
Compare women vs. men across treated vs. control districts.
Outcome variables: share of women in self-employment, business ownership, and access to credit.
compute the change over time in women’s entrepreneurial outcomes in treated vs. control districts.
then net out the equivalent change for men (to control for district-level shocks unrelated to UWEP).
\(\rightarrow\) This gives a triple-difference (DDD):
\[ ( \text{women}_{\text{treated,post}}-\text{women}_{\text{treated,pre}} ) - ( \text{women}_{\text{control,post}}-\text{women}_{\text{control,pre}} ) - ( \text{men}_{\text{treated,post}}-\text{men}_{\text{treated,pre}} ) - ( \text{men}_{\text{control,post}}-\text{men}_{\text{control,pre}} ) \] - If the triple-difference is positive, it suggests UWEP had a causal effect in narrowing the gender gap in entrepreneurship outcomes.