This paper investigates the influence of the size of legislative assemblies on spending and taxation behavior of governments. Using data from local government budgets from 1998 to 2006, I exploit a natural experiment given by a population-based rule that prescribes the size of legislatures in Italian municipalities. The peculiar nature of the data provides a rare opportunity, in political economy literature, to compare identification provided by cross-section and (within) panel data methods. I argue that exploiting the between variation of the observations can be severely misleading, due to possible endogenous sorting around the thresholds, especially in the past. The more reliable panel data approach shows that increasing the size of council members tends to decrease per-capita expenditures, especially in goods and services, of local governments, while no significant effect is found on per-capita revenues from taxation. These findings cast some doubts on whether previous empirical literature has been successful in identifying causal relationship.