APPENDIX C.

Comparison of Pedestrian Fatalities Between the Early and Late 2010s

Our analysis of the FARS database showed that pedestrian fatalities during darkness generally became more prevalent in later years of our study period. Therefore, we also developed cursory binomial logistic regression models to compare factors associated with pedestrian fatalities that occurred between 2015 and 2019 (the late 2010s) with pedestrian fatalities that occurred between 2010 and 2014 (the early 2010s) (Table C1). The first model was based on all pedestrian fatalities (Model A), the second was based on pedestrian fatalities that occurred during darkness (Model B), and the third was based on pedestrian fatalities that occurred during daylight (Model C). Model A includes all fatalities with a known lighting condition, including dawn and dusk (n = 53,318). However, the 1,963 dawn and dusk fatalities are excluded from Model B (n = 39,260) and Model C (n = 12,095).

We excluded 2020 because of the COVID-19 pandemic. We used 2010 to 2014 as the base 5-year period in the models so that parameters indicate the odds of a crash with a particular characteristic occurring between 2015 and 2019. In other words, the estimated parameters in the models show which variables became more prevalent in the latter half of the 2010s (i.e., were significantly associated with the increase in pedestrian fatalities during the decade).

Findings Related to Increase in Pedestrian Fatalities between 2010-2014 and 2015-2019

Relevant to this NCHRP project on pedestrian safety at night, darkness was significantly associated with the increase in pedestrian fatalities. After controlling for all other variables, compared to pedestrian fatalities that occurred in other lighting conditions, pedestrian fatalities during darkness were 18% more likely to occur during 2015 to 2019 than during 2010 to 2014 (Model A).

In addition to fatalities during darkness, we found significant increases in pedestrian fatalities that:

• Occurred on arterial roadways. Compared to local roadways, arterial roadways were between 1.61 and 2.53 times more likely to have pedestrian fatalities occur during 2015 to 2019 than during 2010 to 2014 (odds ratios of 1.61 to 2.53), depending on traffic control.
• Occurred on freeways (odds ratio of 1.78 compared to local roadways).
• Involved pedestrians using drugs (odds ratio of 1.47 compared to pedestrians not using drugs).
• Occurred on one-way roadways (odds ratio of 1.30 compared to two-way roadways).
• Involved pedestrians who were Hispanic/Latino or Black (odds ratios of 1.25 for Hispanic/Latino and 1.20 for Black compared to pedestrians who were White).
• Involved drivers who were aged 25 and older (odds ratios of 1.17 for 25 to 34, 1.10 for 35 to 44, 1.15 for 55 to 64, and 1.20 for 65 years or older compared to drivers aged 15 to 19).
• Involved vehicles turning right or left (odds ratios of 1.20 for turning right and 1.18 for turning left compared to vehicles traveling straight).

• Occurred at marked crosswalks (odds ratio of 1.17 compared to other locations).
• Occurred on shoulders (odds ratio of 1.17 compared to other locations).
• Occurred during rainy weather (odds ratio of 1.13 compared to clear weather).
• Involved hit-and-run drivers (odds ratio of 1.11 compared to non-hit-and-run crashes).
• Involved pedestrians 65 years or older (odds ratio of 1.06 compared to pedestrians aged 17 to 64).
• Occurred on roadways with four or more lanes (odds ratio of 1.05 compared to roadways with fewer lanes).

We also found significant decreases in pedestrian fatalities that:

• Occurred at locations with speed limits of 30 mph and higher (odds ratios of 0.69 for 30 mph, 0.68 for 35 mph, 0.68 for 40 to 45 mph, 0.63 for 50 mph or higher compared to locations with speed limits of 25 mph or less).
• Occurred at intersections (odds ratio of 0.71 compared to non-intersection locations).
• Involved pedestrians younger than age 16 (odds ratio of 0.76 compared to pedestrians aged 17 to 64).
• Involved pedestrians drinking (odds ratio of 0.79 compared to pedestrians not drinking).
• Involved commercial trucks/heavy vehicles (odds ratio of 0.85 compared with cars/sedans).
• Involved reckless driving (odds ratio of 0.85 compared with crashes not involving reckless driving).
• Involved drivers drinking (odds ratio of 0.86 compared to drivers not drinking).
• Occurred in October to December (odds ratio of 0.93 compared with January to March).
• Occurred overnight on Saturday (odds ratio of 0.93 compared with overnight on Monday).

We gained several additional insights by comparing the differences between the 2015-2019 and 2010-2014 time periods by darkness (Model B) versus daylight (Model C). Most of the significant variables in the darkness and daylight models had roughly similar associations with changes in pedestrian fatalities over time. However, we also found:

• There were significant increases in the odds of pedestrian fatalities occurring on arterials (relative to local roadways) over time in both darkness and daylight. However, the increases were larger during darkness for arterial locations with traffic signals (darkness odds ratio 3.04 versus daylight odds ratio 2.01) and no traffic control (darkness odds ratio 1.97 versus daylight odds ratio 1.62). Conversely, the increases were larger during daylight for arterial locations with stop signs (darkness odds ratio 1.22 versus daylight odds ratio 1.76).
• Intersections controlled by stop signs (not specifically along arterials) experienced significant increases in the likelihood of pedestrian fatalities over time during darkness but not during daylight (darkness odds ratio 1.51 versus daylight odds ratio 0.87).
• There were significant increases in the odds of pedestrian fatalities over time during darkness but not during daylight for crashes on roadway shoulders (darkness odds ratio 1.38 versus daylight odds ratio 0.99), on sidewalks (darkness odds ratio 1.44 versus daylight odds ratio 0.92), and in the roadway (darkness odds ratio 1.25 versus daylight odds ratio 1.01).
• The increased likelihood of pedestrian fatalities involving drivers who were aged 65 and higher (relative to drivers aged 15 to 19) over time was larger in darkness than daylight (darkness odds ratio 1.25 versus daylight odds ratio 1.12).
• The reduced likelihood of pedestrian fatalities involving a driver using drugs (compared to not using) over time was larger during daylight than darkness (darkness odds ratio 1.03 versus daylight odds ratio 0.86).
• The reduced likelihood of pedestrian fatalities involving a pedestrian drinking (compared to not drinking) over time was larger during darkness than daylight (darkness odds ratio 0.77 versus daylight odds ratio 0.98).

• The reduced likelihood of pedestrian fatalities involving a pedestrian drinking (compared to not drinking) over time was larger during darkness than daylight (darkness odds ratio 0.84 versus daylight odds ratio 0.91).
• The reduced likelihood of pedestrian fatalities on roadways with speed limits of 30 mph or higher (compared to roadways with speed limits of 25 mph or lower) over time was larger during darkness than during daylight.

Table C1. Binomial Logistic Regression Model Results: Variables Associated with Pedestrian Fatalities in the Late 2010s

1) Days of the week correspond with the 24-hour period from noon for the day listed to 11:59 am the following day (e.g., Friday represents noon on Friday until 11:59 am on Saturday). This is done to capture overnight periods in a single value.

2) Missing values for all location-related variables are controlled by the missing location variable.

3) There are only 10 missing values for work zones in the entire FARS pedestrian fatality dataset, so the estimates for this parameter are very small in this mo

4) Missing values for one-way streets are identical to missing values for the number of lanes, so a missing value indicator variable was not included.

5) The model samples exclude FARS pedestrian fatality records that had unknown values for lighting condition.