An excellent example of the latter may be found in a Houston Chronicle story published yesterday revealing that traffic accidents in Houston increased after the installation of red light cameras ("Was red light camera study flawed?," Jan 1).
While not long ago, the Chronicle touted the results of a much less rigorous study uncritically, in this case, where the results do not support red light cameras, the whole story was centered around claiming the research methodology was flawed. Writes Bradley Olson:
This is utter gobbledy gook, not a legitimate statistical analysis! Red light cameras create their OWN control group when researchers monitor accidents at the intersection BEFORE and AFTER enforcement begins.
Because red-light cameras are known to have a spillover effect — meaning that they have been shown to impact the number of accidents at intersections where there are no cameras — robust examinations of camera programs always compare crash data with that in other cities.
It's what statisticians call a control group. Unless the study authors compare crashes at the 50 intersections where red-light cameras have been installed with other intersections in which they have not been — preferably in other cities — no conclusions can be drawn from it.
That was the flaw I identified in the Texas Department of Transportation study published in December - in the vast majority of intersections they studied, data was not gathered before cameras were installed to compare them. In the Houston study, they had that data going back several years, so IMO the results are actually much more robust and probative than TXDOT's.
Olson writes that, "One specialist from a renowned traffic research organization who reviewed the study for the Houston Chronicle said the methodology was "flawed" and has serious "limitations." But it turns out the "renowned" research group was the Insurance Institute for Highway Safety, which is an arm of the insurance industry, not a neutral source of research.
The reporter also claims, bizarrely, that, "At a ratio of 10 to 1, study after study on the effect of red-light cameras ... have found that they drastically reduce crashes." That's simply false on its face, which further leads me to think Olson's sources were feeding him a line of bull. As I wrote in reaction to the TXDOT study:
in Lubbock red light cameras were discontinued after accidents overall increased 52% at intersections with cameras. Similarly, the state of Virginia eliminated their use after studies in every city using the devices found the number of accidents increased. In other jurisdictions, studies have found reductions in right-angle accidents but nearly equal increases in rear-end collisions, including in injury accidents.The Houston figures jibe closely with those results, so I don't see why Olson or Mayor White are so surprised that accidents increased in Houston, too. That's what happens when officials prioritize revenue generation over public safety.
Here are the bottom line results:
See more from BlogHouston and Off the Kuff.