In their new working paper, “The barriers to female employment: Experimental evidence from Egypt,” Caria, Crepon, Krafft, and Nagy present results of their experiment in Egypt. They offered job matching services (to make jobs easier to find) and childcare subsidies (to make jobs easier to keep) to a sample of mothers with young children. These are women who have very low levels of employment. It is an intriguing study, carefully done and addressing some key facets of low female labor force participation in the region.
The interventions consisted of vouchers for local nurseries and employment services. The latter was done in collaboration with two job matching platforms which connected mothers to a minimum of three open vacancies that were suitable for their skills and their preferences in terms of location, occupation, salary, and working hours. The two interventions were cross-randomized to explore potential complementarities since these mothers might need both to start working. Few of these mothers were working at baseline (11%), while a lot of women reported wanting to work (40% of those with no job) and searching (57% of those unemployed who want to work).
A key finding is that neither intervention, nor the combination, resulted in changes in women’s employment. But take-up is very “limited”. Only one in five women in the job services intervention applied for at least one job; most of the women offered an interview declined it. The 100% subsidy of childcare had low take-up: only 11 percent of households used the free childcare.
In the era of AI, I am limiting my space (and my time) summarizing the paper. If interested for that type of blog, read the ChatGPT-created blog that I generated. Rather, I will highlight three observations.
Pay attention as interventions roll out
Because the team made the effort to track take-up of the childcare subsidy closely after baseline, they were able to spot problematically low take-up early on. (They also anticipated it from the results of their pilot.) As such, they were able to pivot the intervention – shifting to 100% subsidies – in an effort to increase uptake. This is important because simply keeping the 25% and 75% subsidies, and then having very low take-up would obviously leave open many questions about price sensitivity (i.e. “if only they had offered the service for free, then women would have taken it up”). With the revised 100% subsidy, the team also added three additional features to further attract take-up, for a total of 4 subgroups for the childcare subsidy arm; though this did leave me wondering whether the actualized IE is appropriately powered to detect differences between these newly added subgroups.
In a second pivot from their initial plan, the team introduced an additional round of employment services to bump up take-up there in light of high inflation occurring towards the end of the planned period of study. The third round of services sought to boost take-up given the possibility of heightened demand for jobs to protect family income.
Both pivots highlight the value of allow for some flexibility in impact evaluations to maximize learning and relevance, as what seems reasonable on paper proves otherwise in the field.
Norms v. something else, or just more norms?
This paper sets out to unpack what drives low FLFP, by comparing the role of “gender norms” (the view that it is inappropriate for mothers to work as they are primarily caregivers) versus “structural economic barriers” (described as limited demand for female labor, search frictions, and insufficient provision of childcare services). This is the framing in the paper’s introduction.
But arguably, in most (or all) ways, all of these barriers are fundamentally about gender norms. Why do these women turn down jobs further from home, why do they prefer online work or part time work, why do they turn down childcare even when free, why do they say these nurseries are not close enough (though being within 2 km), and why do even women in progressive couples (those where both spouses do not subscribe to gender norms that oppose female employment) have basically no response to these interventions? It is not clear why the mismatch between desired and available non-pecuniary job attributes, shown in the paper, is not fundamentally driven by gender norms (that women want or need to work close to home, not work late, not work with men, etc…).
More on backlash
I blogged a couple of months ago about backlash that women experienced in communities with an intervention to reduce intimate partner violence in Rwanda. I had not expected to find backlash from this intervention in Egypt, and yet, the authors do find evidence of it or at least some backfiring. Among non-progressive couples (those where both spouses oppose women working), the employment services intervention increased intra-household conflict as reported by wives. It also lowered the likelihood that these wives reported expecting their husbands’ attitudes towards female work to improve in the future. As summed up in the paper, and echoing those of the previous blog: “Our results on the possibility of backfire call for caution when designing intervention that try to challenge entrenched norms.”
Join the Conversation