The long-awaited red light camera study
I have three things to say about the red light camera study that was finally released yesterday.
Red-light cameras have sprouted quickly across Texas in recent years, sparking heated debates about whether they reduce crashes or simply bring easy revenue for the cities that install them.
New data from Texas A&M University's Texas Transportation Institute could help settle the argument.
A statewide study by institute researchers shows that monitored intersections had an overall 30 percent decrease in collisions.
The state-mandated report, released Tuesday by the Texas Department of Transportation, examined data from 56 intersections across the state, including many in Houston, from July 1, 2007, to June 31.
The data and analysis are limited because some cities' cameras went online during the study period and their post-installation data were not complete. But the report states that the cameras could be changing driver behavior.
"While these results cannot conclusively determine that red light cameras are responsible for the overall reduction in crashes ... the presence of the treatment provided some effect on the frequency of crashes at the selected intersections for the limited time period of this analysis," the report states.
The study examined crashes at select intersections from 12 cities that were required to report accidents under a new state law. The data show that right-angle collisions were reduced by 43 percent, while rear-end collisions increased by 5 percent, mirroring the results of other studies across the nation.
The report details a methodology complicated by timing. Some cities did not install cameras until recently, so they reported limited post-installation data. Others, based on exemptions in the law, were not required to report all their pre-installation data.
To compensate, institute researcher Troy Walden "annualized" some cities' crash rates.
The full study can be found here
(65 page PDF) if you're interested. I hope to have enough free time to wade through it.
While I admire Matt Stiles' optimism, I do not believe for a second that this study will settle the argument. The people who don't like the cameras will continue to not like the cameras, and I suspect most of them will continue to believe that it's all about revenue regardless of the impact on safety. If anything, I expect the complaints to become more bitter.
And while I'm encouraged by these results, which frankly are better than I had expected, I don't think the matter is settled, either. I'd like to see a few more years' worth of data, to see if the effects continue or if people eventually go back to old habits, before passing judgment. I'd also like to know if any effects can be measured at other intersections in the camera-enabled cities. It's been suggested that having cameras at some intersections can help reduce collisions at other intersections as well. Let's take a closer look at that and see if there's something to it.
Finally, let's remember that a portion of the fines collected from red light camera violations are supposed to fund trauma centers, but there's been some shenanigans over that. Any time a dedicated revenue stream is created by the Lege, there's the risk that it will get dipped into for other purposes. This money should be used as intended, and if the Lege needs to clarify that intended use, it should take that action next spring.
UPDATE: More from Lose an Eye.
UPDATE: Grits has some substantive criticism.
Posted by Charles Kuffner on December 03, 2008 to Planes, Trains, and Automobiles
1) This "study" is hardly scientific. In fact there is no statistical analysis performed in it, let alone any advanced techniques such as empirical bayes analysis to correct for regression to mean that is typically found in TTI's safety studies. In fact, compared to TTI's typical work, and all the work by Bonneson related to red light running, this "study" is something a high school student could have put together.
2) The "study" acknowledges that it can not determine the safety impact of the camera installations because there is no data related to the myriad of other factors that could be affecting safety and traffic operations.
For instance I know of one location listed in the study where an overpass opened up so traffic volumes at the intersection have plummeted yet crashes went up. That is cause for major concern at that one intersection yet it doesn't play out in this study because only the raw frequency of crashes (and annualized freq per year) is looked at.
3) All the crash data is self reported. With respect to the rear end collisions it is extremely easy to make rear end collisions at an intersection disappear simply by only reporting rear end crashes that actually occur within the legal definition of an intersection versus upstream on the approach to the intersection. There is no set definition for rear end crash on the reporting forms from the state for the cameras so its possible to skew the data in such a way.
4) I would love to see Matt or someone else do some checking using the states CRIS (crash records information system) to determine if the number of crash reports sent by the police to Austin actually match the self reported data.
5) The study itself ackowledges that other factors may bias the data, thus lets do a scientific study and take a look at these factors. For instance, I retimed some signals with red light cameras on them and violations dropped substantially. Why? Longer cycle length resulted in longer green time for the major movement resulting in less congestion and fewer frustration violations from sitting through 2-3 cycles at a light. Also some cities when they installed their cameras, retimed their yellows, installed additional heads, restriped stop bars, installed backplates, and a host of other countermeasures. Lets correct the study for the influences of all these other factors before we tout the success of the cameras.
6) Lets look at actual crash rates for the intersections (crashes per million entering vehicles) to see if there are a) actual safety problems and b) if the intersections actually got safer or not. Crashes can drop while volumes drop and the intersection actually get less safe. In fact, based on the pre-installation crash data and publicly available volume data, the crash rates at some of the intersections is indicative of a safe intersection with no safety problem at all, prior to installation of the cameras.
7) Lets look at violations per 10,000 vehicle cycles as suggested by TTI in previous research. Lets see the impact cameras had on this metric.
The cameras are not solely about revenue, nor are they solely about safety. However; I have seen too many cities already remove non profitable but unsafe locations to think otherwise. I have also seen too many cities install cameras at locations with significant capacity and design issues for them to clamor about safety. Retime your camera controlled intersections every 2-3 years, and set your yellows based on 85th percentile speeds, not the speed limit, set your red clearnace (all approaches red) based on the 15th percentile speeds, and set your cameras to only snap pictures after the thresholds recommended in the research and not at the 0.1 second after onset of red and then I may start to believe it is more about safety than revenue.
I've seen too many things contrary to good engineering practice from the inside of the camera debate to think otherwise.