Critics claim camera study shenanigans

So what else is new?

The Houston Police Department tried to influence the outcome of a controversial city-commissioned study by changing how crashes at intersections with red-light cameras were counted, according to documents included in a lawsuit.

HPD's request was refused by the study's authors, however, who concluded the number of accidents at 50 intersections with the cameras had increased, not decreased as city officials expected, documents say.

Attorneys fighting to end Houston's 2-year-old red-light camera program seized on the documents -- released after an open records lawsuit they filed against the city -- as evidence the study was tainted by a purposefully skewed methodology.

"As in other cities, the red-light camera system in Houston is increasing accidents," said Randall Kallinen, a lawyer who represents ticketed drivers in court. "This is very dangerous for the public, and we must end the red-light camera experiment."

I just want to point out here that by Kallinen's logic, if the next batch of data shows a decrease in accidents at these intersections, it must also be the cameras that caused that decrease. You can't have it both ways.

City officials and Rice University political science Professor Robert Stein, one of the study's main authors, contend the Houston Police Department's requests were part of an ordinary back-and-forth about how best to examine the efficacy of red-light cameras and were not a conspiracy to deliver false data.

[...]

Researchers have studied the impact of such cameras for decades, but the results are mixed and inconclusive, according to an analysis of numerous studies conducted by The Cochrane Collaboration, an international organization that evaluates medical and public health research.

The Cochrane analysis found only five studies that used statistically sound methodology to examine data from the U.S., Singapore and Australia.

The result was that red-light cameras usually reduce the number of fatal crashes but don't necessarily reduce total collisions.

The Houston study's authors and city officials expected that to be the result here. Instead, the review showed crashes doubled at intersections where at least one camera was installed, although the uptick in collisions happened in the approaching lanes without cameras. At the lanes with cameras, the increase was too slight to be statistically significant, the study's authors found.

According to an e-mail included in the lawsuit, an HPD official asked Stein in April to rule out accidents if they occurred more than 100 feet from the intersection. Kallinen also said that documents he obtained indicated the department attempted to rule out crashes that did not involve a red-light violation. Either of those steps would be more likely to lead to results showing the cameras reduced crashes, Kallinen said.

Stein, whose involvement has been criticized because his wife works for White, said the study's other authors rejected HPD's suggested change because they were using what they believed was the best methodology.

Mayoral spokesman Patrick Trahan said the police had legitimate reasons to consider limiting the crashes that way, as they did not want the study to include collisions that had nothing to do with running red lights or the cameras.

Doesn't seem like too unreasonable a request to me, but then I haven't been peddling conspiracy theories about the cameras. Your mileage may vary.

Can I make a simple request here? I know there's another study going on to gather more data about the cameras and the accident rate in Harris County as a whole. How about we make sure this study uses the statistically sound methodology that the Cochrane folks refer to? Maybe we could all even agree beforehand that if such a methodology were to be used, we'd all accept the results, whatever they are. And finally, maybe we could try to get other locations that have the cameras do the exact same kind of study, so we can see if Harris County is getting similar results as they are or not. I mean, it could be the case that we've just done a lousy job of implementation, and if we'd followed the example set by others we'd get better results. Or perhaps we'll learn that there are no better results, or that what we got in the first year was a fluke. All I'm saying is, it can't hurt to have more and better data.

Stein acknowledged that the cameras are not working in Houston as well as he believes they have been shown to work in other cities. The city and critics should be more concerned about why, he said.

"Why are these crashes going up at these intersections?" Stein asked. "Nobody really cares to get at the truth here. Cars are being damaged, people are being injured and a handful of people are dying. ... What I want to know is, why they aren't working in Houston, and what we can do to improve them?"

You know what my suggestion is. And while we're at it, let's please release the draft version of this first study. There's no reason not to, and holding onto it just fans the flames.

Posted by Charles Kuffner on January 30, 2009 to Planes, Trains, and Automobiles