Rochester, NY, USA 2014 and 2016 RLC Reports

Official Title: Red Light Camera Effectiveness Evaluation, City of Rochester

Author: SRF Associates (an engineering firm)
 
2014 REPORT:
 
Summary:
This is a 25 page report by SRF Associates(an engineering firm hired to generate this report by the City of Rochester,, NY) on the effectiveness of red-light cameras in the city of Rochester, New York. This report compares the before and after crash and violation statistics of 48 RLC approaches at 32 intersections in that city. The study period of 5 years and 7 months uses several different before and after period lengths from 9 to 33 months of pre-RLC crash data compared to 9 to 33 months of post-RLC data. This report uses a comparison of monitored approaches, not intersections, for it's crash data and analysis. The statistics used in this analysis are from the City of Rochester Police Department and the report did not define what they used as their Distance-From-Intersection inclusion zone for crashes. The "Total crashes" numbers they use only counted crashes that had been categorized as right angle crashes and rear end crashes. There are no data or conclusions in this report that suggest a reduction or increase in fatalities due to RLCs. This report claims that injury crashes decreased 30% and total crashes decreased 26% after the installation of cameras. This report also shows differences in violation rates for two 3-month time periods: the first three months for each camera(which they call the "Before" period), and the last 3 months of the study period in 2014, which they call the "After" period. The report concludes that: "Eight intersections experienced increases in total collisions while 22 experienced decreases in total collisions and two locations remained unchanged." Another conclusion is: "Based upon the results of this study, the red light camera program has reduced the overall number of accidents, the severity of accidents, and related accident costs."
 
Download the 2014 report below
Comments:
This is a brief 25 page report, the data and statistical analysis are not standard or thoroughly explained for a report of this kind. The selection of 32 RLC intersections is large number of intersections for this kind of report. The use of "monitored approaches" instead of complete intersections as the analysis metric is not the standard for this kind of report. In fact, only two other related studies we have found use this method, and it is not explained why they chose this non-standard measurement method. The lack of a defined Distance-From-Intersection inclusion zone for crashes somewhat diminishes the conclusions. The lack of any analysis of traffic volume changes during the 5 year study period somewhat diminishes the results. The lack of any control(or non-camera) intersections in this report to compare the camera intersections to greatly diminishes the results. The before and after RLC date periods are sequential but only 9 months of before and after-RLC data is less than optimal for this type of analysis, and the varying inconsistent time periods used(from 9 months to almost 3 years) somewhat diminish the results. The use of the term "Total crashes" throughout the report is very misleading since only two types of crashes were included in those figures. The use of the term "Intersection crashes" is also very misleading since only crashes indicated as being associated specifically with red light camera approaches were included in the statistics for an intersection's crashes, that could be as few as one approach out of four for an intersection. The grouping of "Right angle crashes" comprises several types of crashes, including "red light running" crashes and the more common "failure to yield" crashes. Yet there is no division of these two primary types of right angle crashes. The violation changes shown in some cases have time periods that are over 5 years apart, and in other cases are only 3 months apart. This completely leaves out any fluctuations in violation rates that occurred during the ignored time periods, and it also ignores any seasonal or traffic volume differences that happened during those time periods. Over one third of the monitored intersections saw an increase in violations, and no explanation is given for those increases. Several intersections saw their crash rates go up, some even doubled, but no explanation is given for this. This report assumes that the only reason for any change in crashes and violations are the presence of red light cameras, even though changes in the road or intersection configurations and changes in signal timing are noted at several intersections for the 5 year study period, they are not mentioned at all in the conclusions. Also, national trends in dropping crash rates during the study period are not mentioned or figured into the results. The data included in this report does not support the conclusions in the report, or the language used to describe those conclusions.
 
Notes:
This is the first report we have seen that includes a red light camera intersection that had zero crashes in the before or after periods. For almost 3 years before a camera was installed at St. Paul Street there was not a single crash, but the city installed a red light camera there anyway. It is hard to make an argument for safety improvements if there hadn't been a single crash at that location.
 
 
 
2016 REPORT:
 
Summary:
This is a 33 page report by SRF Associates(an engineering firm hired to generate this report by the City of Rochester,, NY) on the effectiveness of red-light cameras in the city of Rochester, New York. This report compares the before and after crash and violation statistics of 48 RLC approaches at 32 intersections in that city. The study period of 7 years uses several different before and after period lengths from 47 to 59 months of pre-RLC crash data compared to 15 to 38 months of post-RLC data. This report uses a comparison of monitored approaches, not intersections, for it's crash data and analysis. The statistics used in this analysis are from the City of Rochester Police Department and the report did not define what they used as their Distance-From-Intersection inclusion zone for crashes. The "Total crashes" numbers they use only counted crashes that had been categorized as right angle crashes and rear end crashes. There are no data or conclusions in this report that suggest a reduction or increase in fatalities due to RLCs. This report claims that injury crashes decreased 22% and total crashes decreased 21% after the installation of cameras. This report also shows differences in violation rates for two 3-month time periods: the first three months for each camera(which they call the "Before" period), and the last 3 months outside of the crash study period in 2015, which they call the "After" period. The report concludes that: "Eight intersections experienced increases in total collisions while 23 experienced decreases in total collisions and one location remained unchanged." Another conclusion is: "Based upon the results of this study, the red light camera program has reduced the overall crash rate, the severity of crashes, and related crash costs."
 
Download the 2016 report below
Comments:
This is a brief 33 page report, the data and statistical analysis are not standard or thoroughly explained for a report of this kind. The selection of 32 RLC intersections is large number of intersections for this kind of report. The use of "monitored approaches" instead of complete intersections as the analysis metric is not the standard for this kind of report. In fact, only two other related studies we have found use this method, and it is not explained why they chose this non-standard measurement method. The lack of a defined Distance-From-Intersection inclusion zone for crashes somewhat diminishes the conclusions. The lack of any analysis of traffic volume changes during the 7 year study period somewhat diminishes the results. The lack of any control(or non-camera) intersections in this report to compare the camera intersections to greatly diminishes the results. The before and after RLC date periods are sequential but only 14 months of after-RLC data is less than optimal for this type of analysis, and the varying inconsistent time periods used(from 14 months to over 4 years) somewhat diminish the results. The use of the term "Total crashes" throughout the report is very misleading since only two types of crashes were included in those figures. The use of the term "Intersection crashes" is also very misleading since only crashes indicated as being associated specifically with red light camera approaches were included in the statistics for an intersection's crashes, that could be as few as one approach out of four for an intersection. The grouping of "Right angle crashes" comprises several types of crashes, including "red light running" crashes and the more common "failure to yield" crashes. Yet there is no division of these two primary types of right angle crashes. The violation changes shown in some cases have time periods that are over 5 years apart, and in other cases are only 18 months apart. This completely leaves out any fluctuations in violation rates that occurred during the ignored time periods, and it also ignores any seasonal or traffic volume differences that happened during those time periods. Over one quarter of the monitored intersections saw an increase in violations, and no explanation is given for those increases. Several intersections saw their crash rates go up, some even doubled, but no explanation is given for this. This report assumes that the only reason for any change in crashes and violations are the presence of red light cameras, even though changes in the road or intersection configurations and changes in signal timing are noted at several intersections for the 7 year study period, they are not mentioned at all in the conclusions. Also, national trends in dropping crash rates during the study period are not mentioned or figured into the results. The data included in this report does not support the conclusions in the report, or the language used to describe those conclusions. Another strange part of this report is that the crash data only goes to the end of 2014, but the violation data goes through the end of August 2015, which is far outside of the crash analysis period. This again reduces the validity of this report.
 
Notes:
This is the second report we have seen that includes a red light camera intersection that had zero crashes in the before or after periods(the first being the 2014 version of this report). For almost 5 years before a camera was installed at St. Paul Street there was not a single crash, but the city installed a red light camera there anyway. It is hard to make an argument for safety improvements if there hadn't been a single crash at that location.
 
 
Differences between 2014 and 2016 reports:
 
There are several differences between the 2014 and 2016 reports. The first difference is the change in the study periods. The pre-camera period in the 2016 report starts 22 months earlier than in the 2014 report, and the post-camera period is 6 months later in the 2016 report. Also, even though the 2016 report was produced 16 months after the 2014 report, there are only 6 additional months of crash data supplied. As for crash changes, the total crash rate change was reduced from -26% in the 2014 report to -21% in the 2016 report, and the injury crash rate change was reduced from -30% in the 2014 report to -22% in the 2016 report.
 
Comparing individual intersections shows some other inconsistencies. In the 2014 report, the Mt. Hope @ Elmwood intersection shows a 132% increase in the total crash rate after cameras were installed, but the 2016 report shows that same intersection with an 8% decrease in the total crash rate. So the addition of pre and post camera periods has lead to a 140% swing in the crash rate change at that one intersection. When going further into the numbers of that one intersection, we can see that the first 9 months of the post-camera period had 7 total crashes, while the next 6 months of the same post-camera period had an additional 14 crashes for a total of 21 crashes after the cameras were installed. That's a very significant jump in crashes in a short time period, and there are other intersections like Lake @ Ridgeway that show similar spikes during the after-camera period. None of these spikes is addressed in the report.
 
When looking at other intersections, this pattern is repeated over and over again. So, we decided to go ahead and calculate the crash change using the 2014 pre-camera period and the 2016 post-camera period. But, we ran into another problem. The data shows the West @ Ames intersection having more post-camera injury crashes in the shorter 2014-report than in the 2016-report. The same problem exists on total crashes with N. Goodman @ Clifford, Monroe @ Alexander, Alexander @ Broadway, Clinton @ Norton, State @ Allen, St. Paul @ Upper Falls Blvd, Lyell @ Murray, Lake @ Driving Park, Mt. Read @ Emerson and N. Goodman @ E. Main, intersections' post-camera crash data. The reverse problem exists with the State @ Allen intersection, where the 2014-report shows pre-camera 71 crashes and the longer pre-camera 2016-report shows only 57 crashes. The same problem exists with the State @ Jay, Clinton @ Norton and St. Paul @ Upper Falls Blvd intersections' pre-camera crash data. These changes should be impossible, and they point to clear errors in the data in one or both of the reports that have not been disclosed. So, in calculating the post-camera crash rate, we used the rates from the larger crash numbers where there was a clear data error. The results show that the total crash rate change was reduced to only a 8% decrease, while the injury crash rate change was reduced to only a 6% decrease. Using the lowest crash rates of the two reports for each intersection for the pre-camera period, you get a total crash rate INCREASE of 14% and an injury crash rate INCREASE of 19% after the cameras were installed.
 
Given the clear and very significant data inconsistencies detailed above, nothing in this report can be trusted, and a review of the raw data should be conducted by a third party to figure out what the real crash changes were.
 
 
Ċ
St Pete Driver,
May 24, 2016, 9:35 PM
Ċ
St Pete Driver,
Nov 8, 2014, 5:09 AM
Comments