This is a very tricky problem of statistical analysis. Here is a BBC article presenting possible hypotheses;
Now here is the Economist article criticising the work done by Levitt and presented in Freakonomics
What is the actual most likely cause? For me it is probably everything, although at different amounts and at different times and in different locations. There are never any simple answers - only poor models. The world about us is not subject to simplistic assumptions but our ability to use useful fudges to make predictions is a very useful tool indeed.