I'm having an issue with analysis of crash history on a road, and I'm a bit rusty on my statistical analysis.

I've got crash data for 2 time periods, 2007-2012 and 2013 and I want to compare the two sets to see if the crash rate has decreased by a statistically significant margin. To analyse the 2007-2012 data set (which has 5 data points, one for each year) should I perform a t-distribution analysis or a normal distribution? And once I've analysed the 2007-2012 data, can I compare that to the 2013 result as a null hypothesis to say that if the 2013 result falls outside the 95% confidence level then the road improvements have had a statistically significant effect on the crash rates?

Also, is it valid to assume that the cash rates will follow a normal distribution?

EDIT - also, i have a bad feeling I may need to use a poisson distribution