Safety Evaluation of Red-Light Indicator Lights (RLILs) At Intersections

Chapter 7. Before–After Evaluation Results

This chapter presents the evaluation results, the aggregate results for all intersections, and the results disaggregated by treatment duration, district, indicator type, area type, entering volume, and proportion of volume from the minor road.

Aggregate Analysis

Table 13 provides the estimates of expected number of crashes in the after period without treatment, the observed crashes in the after period, and the estimated CMFs, and their SEs for all crash types considered.

Note: Boldface indicates CMF estimates that are statistically significant at the 95-percent confidence level.

The results in table 13 indicate statistically significant reductions at the 95-percent confidence level for all crash types analyzed except rear-end crashes, for which the negligible increase was statistically insignificant. The crash type with the smallest CMF (which translates to the greatest reduction) was left-turn crashes with a CMF of 0.600. For all crash types combined, a CMF of 0.94 was estimated. The CMFs for fatal and injury, right-angle, disobeyed signal, and nighttime crashes were 0.86, 0.91, 0.71, and 0.89, respectively. An insignificant CMF of 1.02 was estimated for rear-end crashes.

Disaggregate Analysis

The disaggregate analysis sought to identify those conditions under which the treatment was most effective. Because total, fatal and injury, right-angle, and disobeyed signal crashes were the focus of this treatment, these crash types were the focus of the disaggregate analysis. The research team identified several variables as being of interest, including treatment duration, indicator type, level of enforcement, number of indicators, area type, curve presence, major and minor approach traffic volumes, number of lanes, median width, surface width, and posted speed limit. The disaggregate CMFs may be used in prioritizing installation sites, but interpretations should be made with caution. While the research team conducted disaggregate analyses by variables of interest, these characteristics were likely not independent, and the research team does not advise combining disaggregate CMFs. However, based on the disaggregate analysis, one could consider several characteristics qualitatively when prioritizing sites for treatment.

For treatment duration, as shown in table 14 , RLILs became more effective with time. This was evident because the CMFs for total, fatal and injury, and right-angle crashes became smaller as additional time passed after the treatment. The CMFs for total crashes and right-angle crashes were not statistically significant after 1 or 2 years of implementation but became significant after the second year of implementation. While CMFs became smaller over time for most crash types, the CMF for disobeyed signal crashes was significant and stable after the first year of installation.

Note: Boldface indicates CMF estimates that are statistically significant at the 95-percent confidence level.

To assess enforcement and education practices, the research team disaggregated the results by district, as the last two columns of Table 14 show. Across all crash types, the CMFs were smallest for district 1. Local agencies in district 1 responded to the research team regarding the enforcement of the indicator lights. Several counties and cities reported initial advertisements in local newspapers and participation in awareness campaigns. In addition, a few agencies in this district noted that they used the lights and had increased enforcement after their application. No agencies in districts 2 or 5 reported awareness campaigns or increased enforcement. The CMF estimates for districts appear to support these implementation practices (i.e., having publicity and awareness campaigns in combination with some increased enforcements result in smaller CMFs).

For indicator type, as table 15 shows, there was no difference between use of white incandescent indicator lights and blue LED indicator lights for all crash types. For total, fatal and injury, and right-angle crashes, the CMF for white incandescent lights was slightly smaller than the CMF for blue LED lights; however, the difference was not significant at the 95-percent confidence level.

Table 15 also presents the disaggregate results by area type. The results show that RLILs were more effective at rural intersections than urban intersections for total, fatal and injury, and right-angle crashes. These differences were all significant at the 95-percent confidence level. However, although the strategy appeared to be more effective at urban intersections for disobeyed signal crashes, the difference was not statistically significant at the 95-percent confidence level. District 1 did not include any rural sites; all rural sites were located in districts 2 and 5. This implies that the differential effects are likely the result of higher enforcement or awareness campaigns for rural sites.

Note: Boldface indicates CMF estimates that are statistically significant at the 95-percent confidence level.

As shown in table 16, CMFs were significantly smaller for intersections with a total entering volume of less than 40,000 vehicles per day for total, fatal and injury, and right-angle crashes compared with intersections with a higher total entering volume. The strategy was more effective for disobeyed signal crashes at intersections with a higher total entering volume. In all cases, the differences were significant at the 95-percent confidence level. The same trend appears to be true for the proportion of the total entering volume on the minor approach. For intersections with less than 20 percent of the entering volume from the minor road approaches, the CMFs for total, fatal and injury, and right-angle crashes were smaller. However, the difference was not statistically significant at the 95-percent confidence level. For disobeyed signal crashes, the CMF was smaller when more than 20 percent of the entering volume was from the minor road; however, the difference was not significant at the 95-percent confidence level.

Note: Boldface indicates CMF estimates that are statistically significant at the 95-percent confidence level.

Further analysis considered the total number of RLILs present at intersections. There was a positive correlation between total entering volume and number of RLILs, indicating that there were more RLILs at intersections with a higher total entering volume. There was also a positive correlation with the number of RLILs and urban area type and proportion of entering volume from the minor route. In combination, this led to findings that showed fewer RLILs were more effective for total, fatal and injury, and right-angle crashes than most indicators. In addition, there was likely substantial positive correlation between area type and total entering volume, as well as between total entering volume and proportion entering from the minor road. Correlation between area type and entering volume prohibited combining these CMFs for the purpose of crash prediction.

In summary, the disaggregate analysis showed that RLILs were almost immediately effective in reducing disobeyed signal crashes and became more effective over time for all other crash types. In addition, RLILs appeared to be more effective for total, fatal and injury, and right-angle crashes in rural areas at signalized intersections with lower total entering volume and a lower proportion of entering traffic from the minor road. On the other hand, RLILs appeared to be more effective in urban areas at signalized intersections with higher total entering volume and a higher proportion of entering traffic from the minor road. The analysis showed that one should not combine these factors for quantitative analysis, but they could be considered when prioritizing intersections for treatment. The research team found no significant difference in the results between use of white incandescent bulbs and blue LED bulbs; however, the level of enforcement and the level of awareness campaigns conducted appeared to affect the effectiveness.