Run the Differences in Differences regression. You should: i) Include gas station fixed effects, ii) weight the regression by p_obs_time; and iii) cluster the standard errors by zip code. If you are using stata, use the "areg" command to add the gas station fixed effects. You will need to extend the usual specification (add terms to the regression equation) to estimate the effects of the spill separately during and after the spill. Interpret the estimated difference in difference treatment effects during and after the spill