That's a pretty startling claim of success that flies in the face of meta-research on the topic.
Examining the study itself (pdf) more closely, the topline summary of a 13% reduction based on the cameras looks pretty overstated. When you get down to the bottom line (see the chart at the bottom of page 11), researchers found crime did not go down overall at half the sites.
Researchers divided 18 cameras into 8 different sites because several were placed close to one another. Of those 8 sites, they found no crime reduction at four of them. At two sites, crime went down but increased just outside of camera range (one of the two, said researchers, had a net gain despite crime displacement). At only two of the sites did the cameras reduce crime, supposedly without displacing crime elsewhere.
That's a much more mixed set of results than the topline finding that crime declined 13%. And of course, only analyzing 8 sites for a short period of time, the Philadelphia study is analyzing a pretty small dataset. When the British Home Office looked at camera effectiveness in London (pdf), cameras were already in place much more ubiquitously, and could be analyzed over much longer periods of time.
Over at HowStuffWorks, Cristen Conger had this assessment of the overall body of research on the effectiveness of police cameras that I more or less agree with:
That's a good summary of the current state of research on surveillance cameras' effectiveness. They're useful for some, specific purposes, but in general public spaces are more what Bruce Schneier refers to as "security theater," which is to say, not entirely useless but mostly for show.
So taken as a whole, what do all these numbers mean? The Home Office Research Group conducted another more comprehensive study in 2005, confirming that CCTV networks appear nearly ineffective [source: Gill]. A similar evaluation from 2006 by the U.S. Department of Justice, also questioned the reported success of CCTV systems, finding little evidence that they significantly reduce crime [source: Ratcliffe].
This isn't to say that crime cameras are entirely useless. Evidence consistently points out that cameras reduce auto-related crimes as much as 41 percent [source: Welsh and Farrington]. They are also more helpful with reducing crime in enclosed areas with less foot traffic when combined with other law enforcement efforts. And they're helpful in conducting post-crime investigations [source: Ratcliffe].
Nevertheless, the 2005 Home Office study revealed that the cameras did not produce enough bang for the buck. Federal and state governments have poured millions into the set-up and upkeep of crime cameras, but the Home Office study revealed that they were underutilized and not fully integrated into police strategies [source: Gill].
Over the weekend I happened to listen online to a talk by Schneier in which he reminded his audience not to be entirely dismissive of security theater, but to understand it for what it is and not mistake it for real-world security; that's good advice for policymakers regarding surveillance cameras. There's good research out there (much of it cited above by Conger) about when and how cameras do and don't work to improve safety. Beyond those uses, surveillance cameras in public spaces are more about theatrics than security.