Houston's red light camera numbers

The city of Houston has now released its own red light camera study, and to say the least the numbers are interesting.

The number of crashes at Houston intersections with red-light cameras doubled in the first year after their installation, according to a city-financed study released Monday.

But Mayor Bill White argued that the cameras' presence prevented even more collisions and that the study proves the monitoring program is keeping drivers safe.

[...]

The analysis examined crash data at intersections that had a camera monitoring at least one of the four or more traffic signals in an intersection. Most intersections had a camera installed in only one direction, meaning that there were three other signals at that intersection without cameras.

Interestingly, it was those unmonitored points in the intersection that saw the greatest increase in accidents. Where there is a camera, the accidents remained relatively flat or showed only a slight increase.

"Collisions are going up all over the city," said Bob Stein, a Rice University political science professor and one of the report's authors. "But red-light cameras have held back that increase at approaches where they have been installed."

But they supplied no data other than the examination of the non-monitored directions of the 50 intersections to support the conclusion that accidents are up citywide. Stein acknowledged that data from the Houston Police Department shows accidents have declined in the city since 2004, although he said the data is problematic because police officers no longer file reports on every wreck.

If that's the case, then I'm not sure how any study based on accident reports can paint an accurate picture. Right off the bat, it seems you're working with incomplete data. Maybe that means there were more crashes at the monitored approaches than what was studied, but it seems unlikely there would have been that many more to make the spike at unmonitored approaches look less aberrational. About the only thing I get from this is that we'll need to see at least another year's worth of data before we can try to draw any real conclusions about the cameras' effects, if any exist.

The report itself is here (PDF). Cory says it has "some of the most convoluted prose I've ever seen". I actually thought it was the tables that were indecipherable, and as a numbers guy, that's saying something. I have no confidence in my ability to interpret the figures and charts this thing contains, so you're on your own.

Study authors -- who include Robert Dahnke, Benjamin Stevenson and Stein from Rice University's Center for Civic Engagement, and Timothy Lomax from Texas A&M University's Texas Transportation Institute -- plan to analyze insurance industry data in the coming months. They believe that will more definitively prove their supposition that accidents have increased so dramatically. The results of that further research are expected to be completed sometime around August.

I think that's a good idea, and assuming I'll be able to make head or tail of it, I look forward to reading it.

I've been wanting to see a study like this for awhile. I thought it would show some decrease in accidents, as well as a decrease in the severity of the accidents that do get caused (that issue is not explored in any detail in this survey), as that appears to be the usual experience elsewhere. Unless the police reports are greatly understating the situation, I don't see how this result can be taken as evidence for that here. It may well be that the spike in accidents at unmonitored approaches is a one-time fluke, though even if that is the case, it doesn't really enhance a claim that the unmonitored approaches saw any improvements. Maybe we'll see something different next year, and maybe the study of insurance data will clarify the picture, I don't know. All I can say is that this study raises a lot more questions than it answers, and it's far from a triumph for the cameras.

One more thing:

Last week, several attorneys who have opposed the program previously in court filed a lawsuit to force disclosure of a draft copy of the study, attempting to gauge whether the conclusions were somehow changed to suit White's support of the program.

Stein has repeatedly denied this was the case, noting that although some language has been revised, the substance of the study has remained consistent for months and did not change.

As I said before, I see no compelling reason to withhold the earlier versions of the study. Just make it all public and be done with it.

Posted by Charles Kuffner on December 30, 2008 to Planes, Trains, and Automobiles