Guest column: Analysis of traffic camera information

By Dale Gedcke

The referenced documents were downloaded from the Oak Ridge City Council posting in advance of the (Feb. 24) work session. The numbers quoted below were calculated from the data supplied in the referenced report.

Accident Rates During 2013

Using the total crashes listed in the report, and the number of detected vehicles through the pertinent camera locations from both directions, the accident rates during 2013 were:

• Oak Ridge Turnpike at Civic Center: 2.58 accidents per million vehicles
• North Illinois Avenue at Robertsville Road: 0.90 accidents per million vehicles
• Oak Ridge Turnpike at Lafayette Drive/New York Avenue: 1.8 accidents per million vehicles
• Robertsville Road at Iroquois Avenue/Willow Brook School: 0.00 accidents per million vehicles
• Total of all four locations: 1.68 accidents per million vehicles

One way to gauge these numbers is to compare them to the average of 18 intersections in the 2004 Greensboro study, which was 1.1 accidents per million vehicles (before RLC installation), and the 2007 Garber et al. Virginia report with an average of 0.84 accidents per million vehicles over 23 intersections. The Oak Ridge accident rates are slightly higher than the typical rates in the Greensboro and Garber reports.

Note that the NHTSA report shows that the fatal accident rate for the entire United States has declined by 23.6 percent since 2005 and by 22.5 percent since 2007. Consequently, one would expect the 2013 Oak Ridge accident rates to be lower than the 2004 Greensboro and 2007 Garber surveys. Oak Ridge is running a little high, even after traffic cameras have been in operation for about four years.

Change in Accident Rates After Camera Installation

From the data supplied for the work session, one can calculate the average number of accidents per year before (March 2005 to March 2009) and after (April 2009 to September 2013) camera installation.

The data was normalized for the number of months in each set of data, i.e., 49 months and 53 months, respectively. The results are:

Accident Rates Before Versus After Camera Installation

Because the probability of an accident is extremely low (less than 2.6 accidents per million vehicles through the location), the statistical uncertainty (sigma) in the reported number of accidents is simply the square root of the reported number of accidents. When combined in a comparison ratio, that individual uncertainty results in the percent uncertainty tabulated in the fifth column of the table. If the percent change in the number of accidents per year is more than twice the sigma in the fifth column, the change is statistically significant. In other words, above the factor of two, one can be 95 percent confident that there is a systematic change in the accident rate caused by some forcing factor. If the change is less than two times sigma, the change is likely due to random chance. The ratio of the change to sigma is tabulated in the sixth column of the table.

The results for the camera on the Oak Ridge Turnpike at Lafayette Drive and New York Avenue meets the significance test for a systematic change. However, there are factors other than the cameras that could be responsible for that change. Weather is an obvious example. Changes in traffic volume could also affect the outcome. One way to control for those changes would be to compare the accident rates at the camera locations to accident rates for non-camera locations in all of Oak Ridge. Such normalizing information is not available in the work session report.

Lacking the controlling information from the Oak Ridge area, one can compare the decreases in the above table to the data in the referenced NTHSA report. Compared to the total number of fatal accidents in the USA for 2005 through 2008, the results for 2009 through 2012 are down by 19.3 percent. If that bias is subtracted from the largest change in the above table, the net change at the Oak Ridge Turnpike and Lafayette/NY becomes -18.3 percent. That reduced change amounts to 1.7 times the calculated sigma. Consequently, the net change is no longer a statistically significant indicator that the cameras are solely responsible for reducing the accident rate.

The result is not surprising. All of the comprehensive and sound statistical studies of traffic cameras have yielded the same result. The cameras cause a change in accident rates that is so small that it is difficult to detect it with any statistical certainty.

The Effect of the Cameras on Compliance

Where past studies have demonstrated a significant effect caused by the installation of traffic cameras is in compliance with speed limits and stopping on red. In that regard, the documentation provided for the work session offers some interesting information. The 2009 to 2013 trends in the speeds of detected vehicles shows that the average speeds have dropped by a rather small amount in the school zones. The Turnpike at Oak Ridge High School westbound school zone showed the largest decrease at -24.55 percent, followed by the Turnpike at ORHS eastbound school zone with a 13.48 percent decrease. All the rest ranged from -11.81 percent to -1.67 percent. Thus, there is compliance improvement on the Turnpike at ORHS, but little benefit in the Robertsville school zone.

Other than the school zone table, the report does not provide an historical compliance trend on speeds and red-light infractions. However, the consolidated table for all of 2013 indicates that the red light infractions amount to 112 cases per million vehicles through the intersections, and the speeding violations are 1,288 per million vehicles. These are pretty miniscule infraction rates, 0.011 percent and 0.129 percent respectively. It may imply that the violations have reached the bottom where random distraction of the driver limits further improvement.

The speeds for red-light violators indicate that the average speeds are within the posted speed limits. That implies that the violators are unable to successfully estimate and manage the amber interval. There may be little that can be done to further reduce these violations.

For the speeding violations, the table does not offer data for the average speed of detected vehicles to compare to the average speed of ticketed drivers. There is not a lot of useful information that can be gleaned from average speed above the speed limit. This data is probably biased by the speed threshold set for triggering the cameras and issuing a ticket.

From an archived source, information was available on tickets issued for the 12-month period for July 2009 through June 2010. The tickets issued for red-light running amounted to 5,222. For speeding, the total tickets issued were 47,393. The information for the work session documents the numbers during 2013 as 1,449 for red-light running, and 23,889 for speed violations. Thus, the speeding citations are down by a factor of 1.98, and the red-light infractions are down by a factor of 3.6. This improvement in compliance is consistent with what has been found in all the major traffic camera studies. Habitual drivers in the area soon learn where the cameras are, and figure out how to avoid camera tickets. Those who are least aware of camera locations are the 68 percent to 84 percent of the ticketed drivers who are non-residents of Oak Ridge.

Note that a significant portion of the reduction in tickets for red-light running was a result of a Tennessee law that forbids ticketing right turns on red. Prior to that law, tickets for right turns on red accounted for a large fraction of the red-light violations.

The Bottom Line

The data on reduced accident rates is too minor and nebulous to conclude that the cameras have generated a significant reduction in the accident rates. The accident rates after camera installation are still at the high end of the range that one typically finds for intersections without cameras.

As predicted back in 2008, the compliance through the cameras has reduced ticketing by a factor of two for speeding and a factor of 3.6 for red-light running. That improvement represents a decrease in ticket revenue for the City of Oak Ridge and for Redflex.

Obviously, reduction of the accident rates is irrelevant to the decision regarding whether or not to renew the traffic contract with Redflex. The cameras are demonstrably ineffective in significantly reducing accident rates. However, it may be tempting to retain cameras in the school zones for a more convenient means of maintaining statutory speed compliance.

Ultimately, the financial factors have to be addressed. Does the reduced camera revenue still offset the costs the City of Oak Ridge bears in court costs and staffing to monitor and manage ticket issuance? Is Redflex willing to siphon off less revenue now that the installation costs have been covered, leaving a greater share of the revenue for the City of Oak Ridge? Would an alternate traffic camera company be willing to offer a better financial arrangement?

Finally, one needs to consider the impact the program has on the 68 percent to 84 percent of the ticket recipients who are not residents of Oak Ridge. They must be pretty annoyed with the City of Oak Ridge. If we want to attract more retail business from non-residents, in competition with Farragut and Knoxville, is the traffic camera program the best way to achieve that competitive advantage?

Dale Gedcke is an engineer and marketing and technical consultant who lives in Oak Ridge.