Comment by archivator | Hacker News Reader

Comment by archivator | Hacker News Reader

Parent Full thread

I believe this is the study: http://www.library.ca.gov/crb/08/08-007.pdf

Summary of results:

• Neither cameras in Jordan Downs nor Hollywood Boulevard had any significant effect in reducing violent or property crime rates within the target areas;

• The monthly rate of violent crimes fell in both the Jordan Downs and Hollywood target areas; however, the Nickerson Gardens control site and the Hollywood Box matched pair experienced similar reductions and the results were not statistically significant;

• The monthly rate of property crimes decreased in Hollywood, and increased in Jordan Downs, but the results were not statistically significant in either case;

• The evidence on the displacement of crime is mixed; in both locations, some crimes increased at a faster rate in the adjacent areas, indicating that CCTV may displace crime, while other types of crimes decreased relatively more in the buffer areas, though results were not statistically significant;

• CCTV had no statistically significant effect on arrest rates for misdemeanor quality-of-life infractions in Jordan Downs or on Hollywood Boulevard

Thanks. That's not actually it but there was a reference to the SF study and from that it was possible to find the actual report.

Researchers: http://citris-uc.org/news/2009/04/14/citris_helps_city_san_f...

PDF: http://www.ischool.berkeley.edu/files/CITRIS_SF_CSC_Study_Fi...

SF Chronicle TLDR: http://www.sfgate.com/bayarea/article/Crime-cameras-not-capt...

"San Francisco's 68 controversial anti-crime cameras haven't deterred criminals from committing assaults, sex offenses or robberies - and they've only moved homicides down the block, according to a new report from UC Berkeley.

Researchers found that nonviolent thefts dropped by 22 percent within 100 feet of the cameras, but the devices had no effect on burglaries or car theft. And they've had no effect on violent crime.

The cameras have been installed in phases on some of the city's roughest streets since 2005 with large concentrations of them in the Western Addition and Mission District and others in the lower Haight, the Tenderloin and near Coit Tower.

The cameras have contributed to only one arrest nearly two years ago in a city that saw 98 homicides last year, a 12-year high. The video is choppy, and police aren't allowed to watch video in real-time or maneuver the cameras to get a better view of potential crimes."

When page four of the study tells us that "CCTV had no statistically significant effect on arrest rates ..." we learn that the study failed.

Perhaps CCTV reduced arrest rates. Perhaps CCTV didn't reduce arrest rates. The lack of statistical significance means that the people doing the study failed to find out which it was.

Notice that if CCTV is super-duper effective, then it should be fairly easy to get a statistically significant result. This is well known; so the executive summary on page four is an example of weasel-wording. The authors insinuate that CCTV wasn't super-duper effective, heck, it didn't do much of anything. But they weasel out of talking about the power of the study.

If you don't expect CCTV to be super-duper effective, you might still be interested in whether it was a little bit effective. CCTV can be a lot cheaper than foot patrols. It might be disappointing, yet still earn its keep.

We return to the question of the statistical power of the study. A powerful study, that fails to get a statistically significant result, puts a tight upper bound on the usefulness of CCTV. If that is less than what CCTV cost we know to spend our money on something else. A weak study puts a slack upper bound on the usefulness of CCTV. If it fails to get a statistically significant result, then we learn that CCTV isn't super-duper effective, but since neither pro- nor anti- were claiming that, their quarrel continues as before.

I'm giving up, having skimmed to page 15 of the PDF, because, well you can read page 15 yourself and see why I think that it reveals the authors to be completely clueless about statistics.

Reply on news.ycombinator.com