Overtaking the odds: Do passing zones make rural roads safer?

UNIVERSITY PARK, Pa. — The frustration of getting stuck behind a slow vehicle on a remote road is all too familiar to drivers in Pennsylvania, where rural roads make up about 60% of highways. One of the roadway features that addresses this issue are passing zones — dotted sections of two-lane roadways that allow for vehicles to cross into the opposing lane of traffic to pass other vehicles — but are they safe? 

“There’s been some studies completed on the safety impact of dedicated passing lanes, but none for passing zones,” said Eric Donnell, senior associate dean of the Penn State College of Engineering and professor of civil and environmental engineering, who recently co-authored a paper on the topic. “The more that we looked through state and federal traffic safety resources like Highway Safety Manual, the more we noticed there was a lack of data, which really drove us to pursue this research.” 

Donnell and his collaborators at Penn State used a large data set of Pennsylvania Department of Transportation (PennDOT)-provided crash statistics and roadway features taken from all state-owned, rural roads to compare the safety performance of roadway segments with passing zones to stretches without marked passing zones. Their computational analysis revealed 11% fewer total crashes, and 12% fewer crashes leading to significant injury or fatalities on sections of rural highway marked as passing zones compared to sections of road marked as no passing. The study, available online now, is set to publish in the September issue of Accident Analysis and Prevention. 

According to Vikash Gayah, director of the Larson Transportation Institute, professor of civil and environmental engineering and co-author on the paper, roadway experts already suspected that passing zones had a significant impact on road safety, but did not have a system to accurately quantify their impact. 

“In previous work, these passing zones continually came up as areas that experienced fewer reported crashes than other sections of road,” Gayah said. “We knew where all the passing zones were and knew that they were associated with fewer crashes, but this was an opportunity to really dig into the data.” 

The team modeled the data using two approaches. The first method, known as a heterogeneity or random parameters model, is a traditional way of processing data that uses randomized parameters to account for variation across observations in data. The other method uses a propensity scores-potential outcomes (PSPO) framework. A propensity score is used to determine how likely a passing zone exists within a road segment based on a set of explanatory variables, like road width, annual daily traffic flow, horizontal curvature, posted speed limit and more.  

Road segments with and without passing zones are then matched together based on the similarity of their propensity scores. The safety performance of roads with and without passing zones were then estimated using a statistical model, allowing the researchers to develop a crash modification factor (CMF), or a numeric value that quantifies the safety impact of a road feature, for passing zones. Finally, they compared the CMFs across models to check for consistency. 

According to Donnell, the data was almost overwhelming at first. The team initially had over 55,000 year-miles, a unit of distance that quantifies the total distance of roadway traveled over an analysis period. 

“We had well over 100,000 data points to estimate the heterogeneity models,” Donnell said. “The PSPO method uses computer algorithms to match data, which reduced the sample to about 63,000 observations, helping us find connections between crash frequency and road features.” 

The researchers applied the CMF to three different categories of vehicle crashes — the total number of reported crashes on the road, fatal and injury-causing crashes and head-on/sideswipe crashes, which are the most common crash types when passing. Their models showed 11.2%, 12.2% and 10.6% fewer crashes, respectively, for each category on roads with passing zones compared to roads without passing zones. According to Gayah, while passing zones may play a factor in decreasing crash risk, there's a host of related factors that may also influence these findings. 

“Our best interpretation is that these areas are usually wider, straighter, flatter and have more visibility, lowering the crash risk,” Gayah said, explaining the characteristics required to establish a passing zone. “Additionally, the types of passing maneuvers people undergo in areas not marked as passing zones are usually much riskier, increasing their likelihood of a crash.”  

The team also found that traffic, curvature and intersection frequency is associated with higher crash frequencies, while wider shoulders and rumble strips are associated with fewer crashes.  

“These statistical models and tools we created give us a better collective understanding of the impacts designating an area as a passing zone can have on a roadway’s safety,” Donnell said. “We can really start to understand the safety trade-offs when making decisions about road features.” 

According to Donnell, the next step of this research is to implement additional variables into their PSPO framework models, including passing zone length and type — single directional, where only one lane has the ability to pass, or bidirectional, where both lanes of traffic can pass.  

The third co-author of this paper is Prakash Poudel, a doctoral candidate studying civil engineering under Donnell at Penn State. This research was supported by PennDOT.