Is it possible to improve women’s agency by providing information about their abilities? Using a lab experiment in the field, I study how perceived abilities and gender stereotypes shape intra-household dynamics. I use an incentivized decision-making game with 525 married couples from 42 rural villages in Bangladesh to investigate whether women are discriminated against because they are perceived to be less skilled than their husband, and whether it is possible to reduce this gender bias within households. During the game, I provide information on women’s abilities and I observe how beliefs and decisions change. The empirical analysis shows that the less capable women are perceived compared to men, the less they are involved in decisionmaking. After the information treatment, husbands with the lowest regard for their wife’s skills are 20 percent more likely to make allocations in her favour. The treatment has a larger impact on younger couples, on men with stronger control preferences and on risk-averse women. This brings further evidence of the inability of spouses to observe each other’s skills. Two weeks after the experiment, women in treated couples report being more involved in household decisions. These results suggest that gender discrimination within households has a statistical component that can be corrected by increasing skills’ observability.