Low-income and Black households are less likely to visit bank branches than high-income and White households, despite the former two groups appearing to rely more on branches as means of bank participation. We assess whether unequal branch access can explain that disparity. We propose a measure of bank branch access based on a gravity model of consumer trips to bank branches, estimated using mobile device geolocation data. Residents have better branch access if branches are closer or have superior qualities that attract more visitors. Because the geolocation data is distorted to protect user privacy, we estimate the gravity model with a new econometric method that adapts the Method of Simulated Moments to handle high-dimensional fixed effects. We find no evidence that low-income communities lack access to bank branches and instead find that lower demand for bank branch products or services explains their lower branch use. But in Black communities, worse access explains their entire drop-off in branch use. For residents of these areas, weaker access is not from having lower quality branches, but from branches being located farther away from them. The results highlight parts of the country that would benefit the most from policies that expand access to banking.