Matt Steinglass doubts the drop in crime over the last few decades has a single reason:
Random controlled trials might very well have found that the broken-windows strategy doesn't prevent crime, "Project Ceasefire" doesn't prevent crime, reducing rates of single motherhood doesn't prevent crime, family planning doesn't prevent crime, banning lead doesn't prevent crime, and so on and so forth; there might have been no statistically significant difference one could isolate for any of these things. And yet it seems extremely likely to me that most or all of these were good things to do! The drop in violent crime probably has to do with all of them. So we probably need to be a bit circumspect about demanding results from our cost-benefit analyses, and go ahead and do things that seem like they probably work.
Noting that crime has dropped almost everywhere, Kevin Drum thinks it's more likely that "there's some single factor underlying the decrease that affected the entire country":
In fact, since drops in violent crime were also recorded in Canada during the past two decades, and elsewhere around the world during other time periods, it's probably some worldwide factor. And on that score, gasoline lead reigns supreme. There's really nothing else that persuasively explains a global rise and fall in violent crime that happens at different times in different countries.
(Chart from (pdf) Understanding international crime trends: The legacy of preschool lead exposure by Rick Nevin)