We use a randomized field experiment to estimate the effect of having a United States bank account on Mexican migrants’ savings and remittances. With support from the Mexican Consulate and a local bank targeting Hispanic clientele, we randomly assigned assistance in obtaining a matrícula consular card, which we call “treatment.” This consulate-issued identification card is accepted by many U.S. financial institutions for the purpose of establishing identity for new accounts, and has little other use in the small U.S. city where we conduct the experiment. Treated migrants were 38 percentage points more likely to open a U.S. bank account, increased their savings as a share of income by 9 percentage points and decreased their remittances to Mexico as a share of income by 6 percentage points. There is heterogeneity of treatment effects by migrants’ reported degree of control over how their remittances are allocated in Mexico. Among migrants who report having no control (as opposed to shared or sole control), treatment causes a higher take-up of U.S. bank accounts, a larger increase in total savings, a shift away from Mexico savings toward U.S. savings, and an increase in income. These results suggest that extending bank access can raise savings in a low-income minority population. Additionally, they suggest that issues of control alter intrahousehold resource allocations and income, which rejects the unitary model and even the collective model for characterizing decision-making in these Mexican migrant households.