Liz, try a non-parametric comparison [Chi squared] of accidents after 9 PM now and several years ago. This comparison can be used to tie accidents per miles driven to expected accidents. I have no idea what the results would be but they might be interesting.