Can we ignore spatial dependence when evaluating mergers?

This study explores whether antitrust authorities can use models that ignore spatial dependence in gasoline prices when assessing merger proposals. I estimate two non-spatial and one spatial model and compare merger simulation results based on these models. The identification strategy uses the abrupt change in ownership caused by takeovers of three chains, which generates virtually exogenous shocks in the local markets. The pure non-spatial fixed-effects panel model significantly underestimates the price changes and sometimes even mispredicts their direction. The fixed-effects panel model with added spatially weighted changes of the purchased stations' price level performs better but can still understate the price changes. It also overstates the number of stations that notably change their prices. The SAR model should thus be preferred.