Red-light cameras reduce crashes, Texas study suggests, Paul J. Weber, AP via Chron.com
But a closer look at the survey itself reveals something slightly different.
Yes, during the first year or two there is a (slight, and uneven) decrease in traffic accidents after the installation of RLC's. At 14% (with little controls, this could easily be mostly made up of noise. But, since we're taking the numbers at face value without factoring in noise*....)
You would also be wise to notice that accidents started to INCREASE at almost all intersections after the 2nd year RLC's were in effect. What this suggests is that the "safety" benefits from RLC's are transitory at best, and the intersections will revert to the mean over time as people get more and more oblivious to the RLC's and continue old driving patterns.
What any of this means is uncertain, but it certainly doesn't "suggest" that the installation of RLC's provide any sustained reduction in accidents. If anything, it suggests the opposite. It's just another incomplete data point in the on-going debate over RLC's, which the voters of Houston have rejected in an election the areas former newspaper of record refuses to recognize as valid.
A far better solution could very well be reinforcing already ingrained behavior by lengthening yellow light times. Unfortunately, this option does not drive revenue for the City so it's being rejected out of hand.
*Noise could also relate to reduced traffic volumes due to the economy. The raw number decreases are way too small vs. the general population to be considered statistically significant, especially considering the lack of control intersections (i.e. intersections without RLC's) in the study.