Interesting statistical work
Just came across economist Alexander Tabarrok's webpage. I found a couple of interesting papers there.
1. He realized that the varying levels of terror alert in Washington DC lead to varying concentrations of police in certain areas, and this could be used to verify whether increased police presence really does reduce crime. What is needed for this is a variance in policing that is unrelated to street crime -- hard to find, since usually police presence comes as a reaction to crime increases. He ran the data, and found that crime dropped around 6.6% on days when police forces were increased. There was an esp. strong effect on auto theft and thefts from autos, outdoor offenses where being seen by an officer is a particularly high risk.
2. He re-ran Lott's famous and controversial study concluding that enactment of "shall issue" carrying concealed weapons laws reduces crime. (I don't know enough about statistics to understand just what he did different, but it certainly sounds like a significant change in method). His conclusion was that such laws indeed decrease violent crime, although they increase property crime. One explanation he suggests is that criminals may abandon armed robbery, still want to make money illegally, and thus turn to burglary and nonviolent theft. Still, a change for the better.