The data set is far too small to draw serious conclusions from yet.... especially the 'increase' in stores that weren't participating (although a 50% drop 'feels' like it should be significant...).
Still, hopefully the effect is real and less people suffer as result. :)
When the sample size is that small, it's possible that there were just one or two dudes doing most of the robberies. If they decided to move, call it quits, or went to jail, that could be an alternative explanation to the drop. I think you'd have to implement the experiment over a region to make it accurate if N is so low.
Still, hopefully the effect is real and less people suffer as result. :)