Accidents more than doubled at the Houston, Texas intersections where red light cameras are installed, according to a study released Monday by Rice University and the Texas Transportation Institute (TTI). This result posed a dilemma for TTI and the city of Houston which had requested the study.

Houston Mayor Bill White was furious when he saw the report's draft text in August. He banned the document from publication and ordered a re-writing of the text that would reflect a more positive result. To accomplish this task, White was able to turn to the study's primary author, Rice University Urban Politics Professor Robert Stein. Stein's wife, Marty, is employed by the city of Houston as a top aide to the mayor. Stein's newly revised report now concludes that "red light cameras are mitigating a general, more severe increase in collisions."

That left the Texas Transportation Institute with its own difficulty. Last month TTI coauthored another study with the Texas Department of Transportation (TxDOT) intended to support photo enforcement on a statewide basis. This report drew its conclusions from an examination of 56 intersections, 31 of which were found in Houston. The TxDOT-TTI study received wide publicity for the reported claim that red light cameras reduced accidents at camera intersections by 30 percent ( view study in 1mb PDF format).

"The TxDOT-TTI results are impossible to square with Houston's results," Houston attorney Randall L. Kallinen told TheNewspaper.

Kallinen and attorney Paul Kubosh have filed a lawsuit to force Houston to disclose the August draft of the Rice-TTI study under state freedom of information laws. Publicity over the lawsuit likely forced the city finally to release the final report during a holiday week. The revised report's data tell a much different story than that presented in the conclusions.

Houston currently tickets motorists at seventy intersections, but the Rice-TTI report examined the first fifty where the most accident data were available. Beginning in September 2006, Houston commissioned American Traffic Solutions to install the machines in groups of ten per month. The Rice-TTI study compared 24 months of pre-installation data to between 13 and 21 months of post-installation data for each of the five groups. According to TheNewspaper's analysis of overall accident data found in the appendices, the average number of monthly collisions went from an average of 15.4 collisions per month in the two years prior to camera enforcement to 58.3 accidents per month in the post-installation period. Although this figure is not reported in the study itself, the general fact is briefly acknowledged.

"The absolute number of collisions at camera-monitored intersection approaches is not decreasing," the study admitted.

To achieve the appearance of success, the study divided red light camera intersections into "non-monitored" approaches -- the directions of travel at the intersection where the red light camera is not looking -- and the "monitored" approaches where ticketing took place. There was a 132 percent increase in collisions at the non-monitored approaches of the intersection where red light cameras were installed and a non-significant 9 percent increase at the monitored approaches. The study treated these increases in both rear end and T-bone collisions as unrelated to the red light camera as long as the accident happened outside of the camera's view.

The study concluded that because the accidents went up at the non-monitored approaches of red light camera intersections, but effectively stayed the same at the monitored approaches, that the red light cameras were responsible for the "benefit" (a smaller increase) at one part of the red light camera intersection, but not the increase in acccidents at the other. This line of thought would suggest that the increased accidents at the non-monitored approaches of red light camera intersections reflected an increase in accidents at the other city intersections that had no red light cameras at all. The study admits this implication is untrue.

"Currently, conclusions on a general increase in collisions across the city are not supportable with available data," the study states.

The Rice-TTI dataset also throws doubt on the conclusions of the TxDOT-TTI study. Increases in Houston collisions documented by Rice-TTI mysteriously became decreases in collisions in the TxDOT-TTI report, as follows:

• Monroe at Gulf Freeway East Service Road: a 913% increase became a 41.7% decrease
• Hollister at Northwest Freeway: a 747% increase became a 60.5% decrease
• FM1960 West at Tomball Parkway: a 307% increase became a 44% decrease
• Richmond at Dunvale: a 103% increase became no change
• South Sam Houston Freeway at Telephone Road: a 164% increase became a 19.3% decrease
• East Freeway North Service Road at Normandy: a 52% increase became a 25% decrease
• North Freeway West Service Road at West Rankin Road: a 18% increase became a 32.7% decrease

In other cases, decreases became more pronounced and increases lessened. To cite just one example, a 217 percent increase in accidents in the red light camera approach of Scott Street at South Loop East North Service Road in the Rice-TTI report became a 50.1 increase in the TxDOT-TTI report. In general, far fewer collisions were reflected in the data used in the TxDOT-TTI report. Since Houston's results comprised more than one-half of the statewide results, it is unclear how TxDOT and TTI could have concluded a statewide reduction in accidents without data manipulation in at least one of the TTI reports.

A full copy of the final Houston report is available in a 400k PDF at the source link below.