Women in poor countries exercise little agency. I investigate whether agency is constrained by women's beliefs in their general ability to reach goals, beliefs referred to as generalized self-efficacy (GSE). I study agency in decisions about women's labor supply in India, a setting where women's employment is low, women have little say over their labor supply, and many women are interested in working. My experiment offered women a psychosocial intervention to raise GSE. I cross-randomized a video promotion of women's work for women's family members. The GSE intervention produced a persistent increase in GSE. The promotion made family members see more financial value in women working, but reduced women's interest in working. Effects on women's employment in the short-run are consistent with GSE treatment leading women to advocate in their households for their preferred outcome; GSE treatment had a positive effect when the promotion was not given but a negative effect when it was. There are no effects on long-run employment, perhaps because household chores made women's work unsustainable. I do find effects of GSE treatment on another economic outcome - saving - did persist. Taken together, my results suggest a key constraint to women's agency is women's own sense of agency.
We outline a model of household decision-making that has support in recent empirical research and present results from an experiment that tests it. The key feature of the model is that households choose whether or not to bargain. Bargaining is costly, and information about the household's choice set may be asymmetric. In this model, spouses may withhold information to manipulate the choice set and may avoid bargaining to prevent certain outcomes from being realized. We test the model in the context of female labor supply in India. Spousal preferences are misaligned: wives are significantly more supportive of women's employment than their husbands. The model predicts spouses may withhold information or avoid bargaining to manipulate decisions about female labor supply. We experimentally vary enforcement of common knowledge and enforcement of bargaining. We randomize whether husbands or wives are given information about a women's job opportunity and an enrollment ticket. We cross-randomize whether non-targeted spouses are not informed, informed separately, or informed at the same time as their targeted spouses. In the third condition, we explicitly encourage discussion with the view of enforcing bargaining. Surprisingly, we find that husbands do not withhold information and that discussion significantly decreases enrollment. Our results contradict the standard predictions outlined in the model.