Official Title: Chicago Red Light Camera Enforcement: Best Practices and Program Road Map
Authors: HANI S. MAHMASSANI, JOSEPH L. SCHOFER, BRETON L. JOHNSON, OMER VERBAS, AMR ELFAR, ARCHAK MITTAL and MARIJA OSTOJICThis is a 104-page 2017 analysis by the NORTHWESTERN UNIVERSITY TRANSPORTATION CENTER that compares the injury crash rate at 85 of the Red-Light camera intersections(340 approaches) in the city of Chicago compared to 59 non-RLC intersections(236 approaches) in the city. The analysis uses 36 months of pre-RLC injury crash data compared to 36 months of post-RLC data with a 24 month gap in the data between the before and after time periods. The injury crash statistics used in this analysis are from the Chicago Department of Transportation(CDT), which uses an distance from intersection definition of 25-50 feet from the intersection, although they did mention that they only used injury crashes that were coded as "intersection-related" in IDOT data fields within the data they received from CDT. There is no data in this report that suggest a reduction or increase in fatalities due to RLCs, although they did collect data on fatalities and they did suggest fatalities are reduced, without showing data to prove that suggestion. This analysis does not go over total crashes at all. The results of their data analysis show that while angle and turn injury crashes in the city are down 19%, the rear-end injury crashes at RLC-equipped intersections are up 14%.
Download this study below (2,872k)
This is a very thorough statistical analysis of only the injury crash rates at selected intersections within the City of Chicago, although there are several significant issues with this report.
First, the 2 year gap in the data collection is not a standard practice in a study like this and no reason is given for ignoring this data. This very large gap in the data greatly reduces the significance of the conclusions. Related to this, they spend a great amount of space in this report discussing how they will account for regression to the mean, yet they chose relatively short data periods. They could have easily used two additional years of before and after data, as well as removing the 2 year gap which would have enabled them to be able to have up to 6 years of both before and after crash data. They provide no reasons for their very limited data range choices.
Second, this study ignores Property-Damage-Only(PDO) crashes completely, in spite of frequently using the phrase "Total Crashes" they never explicitly mention in any of their conclusion sections that it really only means total injury-only crashes.
Third, the "Documentation Research" ignores results in those studies that do not fit their pre-conceived conclusions. For example, they mention the 2007 Virginia Transportation Research Council report but they only state that there was an increase in rear-end crashes and a decrease in "red light running" crashes while completely ignoring that total crashes went up 23% and injury crashes went up 17% in that study. The 2005 University of Arizona study they mention shows similar numbers for increases in total and injury crashes within Phoenix as well, yet those results are not mentioned either. Also, almost all of the studies cited analyzed true total crashes, including PDO crashes, while this report did not, making for an invalid comparison of this report to the studies cited in the literature review. Overall, the literature review is inconsistent and their summary of the literature review is not accurate given the conclusions of the studies cited.
Fourth, they appear to equate "red light running" crashes with "angle" crashes, while the two crash types are not the same since angle crashes contain a very significant proportion of failure-to-yield crashes that do not involve red light running.
Fifth, they collected crash data from the Chicago Department of Transportation(CDT), which they say originally came from the Illinois Department of Transportation(IDOT). They never state why they didn't just get the crash data directly from IDOT. This is significant because the 2010 University of Illinois at Chicago study showed that there were very significant differences between the IDOT and CDT crash data, because the CDT data excluded any crashes outside of 25-50 feet from the intersection while the IDOT data allowed for crashes up to 150-300 feet from the intersection.
Sixth, they do not include any of the raw crash data on a per-intersection basis, meaning that there is no way to validate their calculations or replicate the analysis because only the totals and summaries are shown. In fact, they don't even provide lists of the 85 RLC intersections and 59 non-RLC intersections that they chose to perform the crash change analysis on.
All of these issues put together greatly degrade the significance of this report.