We study local government incentives to misreport the information required to implement a formula grant. We focus specifically on population, in theory the easiest variable for the grantor to verify. We analyze the Spanish case and show how a switch from the use of census to registered population data (the latter administered by the municipalities) led to a manipulation of the population numbers used by central government to allocate grants to municipalities. As a result, registers included a proportion of ‘ghost’ citizens, that is, people who presented no trace of actually residing in the municipality. We identify the effects of grants on population over-reporting taking profit of notches in the grant scheme (i.e., one based on weighted population with the weights increasing at specific population thresholds). We document an excess mass of municipalities to the right of the notch threshold and a density hole to the left of it. There are several indications that manipulation (rather than real population responses) is the mechanism at work.