Young-Jun Kweon

Safety effects of speed limit changes : Use of panel models, including speed, use, and design variables

This work estimates the total safety effects of speed limit changes on high-speed roadways by using traffic detector data and Highway Safety Information System data from 1993 to 1996. To gauge the total effects, the study applies a sequential modeling approach: average speed and speed variance models are first estimated on the basis of roadway design, use, and speed limit information. Then, crash counts (of varying severity) are estimated on the basis of the speed estimates, design, and use variables. The 4 years of data come from 63,937 homogeneous roadway segments along seven Interstates and 143 state highways in Washington State. A random effects negative binomial model was selected among several alternative panel and nonpanel models for count data. Results indicate that the average road segment in the data set can be expected to exhibit lower nonfatal crash rates up to a 55 mph (88 km/h) speed limit. In contrast, fatality rates appear unresponsive to speed limit changes. Fatal and nonfatal rates fall for design reasons, including wider shoulders and more gradual curves, which appear to be key design variables. However, fatal and nonfatal rates move differently when traffic levels rise, with nonfatal rates remaining unchanged and fatal rates falling.

Sidewalk Cross-Slope Design: Analysis of Accessibility for Persons with Disabilities

Current and proposed Americans with Disabilities Act (ADA) guidelines offer no specific guidance on acceptable maximum cross slopes where constraints of reconstruction prohibit meeting the 2% maximum cross-slope requirement for new construction. Two types of sidewalk test-section data across a sample of 50 individuals were collected, were combined with an earlier sample of 17 individuals, and were analyzed, with an emphasis on cross slopes. These tests examined heart-rate changes and user perception of discomfort levels, and they relied on a random-effects model and an ordered-probit model, respectively. Model estimates were used to deduce critical or unacceptable cross slopes for critical conditions and critical populations of persons with disabilities. Predicted values for the most severe or constrained cases ranged from 5.5% to 6% cross slope. These cases included 5% primary slope (main grade) and 45-ft-long sections; the sections were traversed by cane, crutch, or brace and manual wheelchair users who were up to 80 years old. When primary slopes were reduced to 0% in the perception estimates, the critical cross slopes for the critical case rose to 6%. For most other persons with disabilities, the critical cross slopes ranged from 6% to 9% or more. These values substantially exceed the ADA accessibility guidelines’ 2% maximum cross-slope standard for public sidewalks.

Identifying high-crash-risk intersections

Identifying high-crash-risk locations, called hot spots, is an important step in improving roadway safety. Use of the empirical Bayes (EB) method coupled with the use of safety performance functions (SPFs) is considered the state of the practice in identifying such locations. However, application of the EB-SPF method requires considerable resources in preparing data, as well as statistical expertise. As a consequence, many highway agencies still rely on traditional methods that use crash frequency and crash rate to identify locations for potential safety improvements without knowing the accuracy of such methods. This study examined four traditional methods commonly used in identifying potential locations for safety improvements and compared them with the EB-SPF method. The four methods evaluated were crash frequency, crash rate, rate–quality control, and equivalent property damage only. The study was limited to four-leg intersections with either a traffic signal or two-way stop control; 2004 to 2008 data were collected for 1,670 such intersections. The study found that the crash frequency method performed the best of the four in correctly identifying the top 1% of unsafe intersections. However, the method tended to flag top hot spots incorrectly. The rate–quality control method performed the best in identifying the top 5% and 10% of unsafe intersections. The findings are expected to help highway agencies that continue to use the traditional methods choose the most appropriate method so that scarce resources available for safety improvement can be invested effectively.

Differences in traffic violations and at-fault crashes between license suspension and revocation

Upon conviction for particular traffic offenses, drivers can have their licenses revoked. Drivers who receive license revocation have an opportunity to apply for a sentence reduction, and some of those who apply receive a reduced sanction--license suspension. There may be differences between drivers whose license was revoked as originally sentenced and drivers who received the reduced sanction of license suspension with regard to traffic violations and crashes after driving privileges are restored. This study verified the differences during the follow-up periods of 6, 12, and 18 months using analysis of covariance and the t-test with stratified samples based on the police profiles of approximately 154,000 drivers in South Korea. The study found that drivers in the group whose license had been suspended committed traffic violations and caused traffic crashes less often for all time periods than those whose license had been revoked. However, omitted factors such as the attitude of suspended drivers and exposure to traffic violations and crashes (e.g., driving frequency after license reinstatement), are likely to affect the findings; thus, caution should be exercised when the findings are referenced for policy implications.

Potential risk of using General Estimates System: Bicycle safety

Beneficial effects of bicycle helmet use have been reported mostly based on medical or survey data collected from hospitals. This study was to examine the validity of the United States General Estimates System (GES) database familiar to many transportation professionals for a beneficial effect of helmet use in reducing the severity of injury to bicyclists and found potential risk of erroneous conclusions that can be drawn by a narrowly focused study when the GES database is used. Although the focus of the study was on bicycle helmet use, its findings regarding potential risk might be true for any type of traffic safety study using the GES data. A partial proportional odds model reflecting intrinsic ordering of injury severity was mainly used. About 16,000 bicycle-involved traffic crash records occurring in 2003 through 2008 in the United States were extracted from the GES database. Using the 2003-2008 GES data, a beneficial effect of helmet use was found in 2007, yet a detrimental effect in 2004 and no effect in 2003, 2005, 2006, and 2008, which are contrary to the past findings from medical or hospital survey data. It was speculated that these mixed results might be attributable to a possible lack of representation of the GES data for bicycle-involved traffic crashes, which may be supported by the findings, such as the average helmet use rates at the time of the crashes varying from 12% in 2004 to 38% in 2008. This suggests that the GES data may not be a reliable source for studying narrowly focused issues such as the effect of helmet use. A considerable fluctuation over years in basic statistical values (e.g., average) of variables of interest (e.g., helmet use) may be an indication of a possible lack of representation of the GES data. In such a case, caution should be exercised in interpreting and generalizing analysis results.

Examination of Macrolevel Annual Safety Performance Measures for Virginia

Municipal, state, and federal agencies in the United States that are responsible for traffic safety have used crash rates such as fatalities per 100 million vehicle miles traveled (VMT) as traffic safety performance measures. However, the appropriateness of using such rates as performance measures has not been examined empirically, although the rates have been made public. This study examined 20 candidate crash rates (e.g., fatalities per million population and injury crashes per million registered vehicles) for an annual safety performance measure for Virginia by using autoregressive error models and empirical data from 1971 through 2006. The study found that the injury rate per driver and the crash rate per VMT seem appropriate as, respectively, long-term (1971 to 2006) and shorter-term (1995 to 2006) safety performance measures for Virginia. Statistical uncertainty should be considered when these rates are used to measure safety performance.

Analysis of Weigh-in-Motion Data for Truck Weight Grouping in Mechanistic-Empirical Pavement Design Guide

The purpose of this study was to evaluate the Virginia Department of Transportations traffic data plan for implementation of the Mechanistic–Empirical Pavement Design Guide (MEPDG) with weigh-in-motion (WIM) data from 22 sites in Virginia for 2007 and 2008. The evaluation included an assessment of the WIM data for pavement design and for enforcement of overloaded trucks and the appropriateness of the truck weight groups. Grouping the sites on the basis of average equivalent single-axle loads was notably different from the current truck weight groups and grouping results based on traffic characteristics such as truck volume. Thus, further efforts to suggest a better grouping scheme are needed. For calculating monthly traffic factors, an input to the MEPDG, volume data from WIM sites could lead to biased results. Thus, vehicle classification count data are a better source than WIM data for the factors. The enforcement sites were found to carry heavier trucks in terms of average equivalent single-axle loads than the sites installed for pavement data collection. Thus, the concern that truck weights collected at the enforcement sites might be inappropriate for pavement design because of possible avoidance of the sites by overloaded trucks seems unwarranted. However, because of several factors and limitations, a definitive conclusion regarding this result could not be drawn.

Identifying Promising Highway Segments for Safety Improvement Through Speed Management

Speed variation is closely related to the occurrence of traffic crashes. Thus, speed management strategies that reduce speed variation are expected to reduce crash frequency and not only improve safety but also prevent congestion due to crash occurrence. This study developed a modeling approach to identify promising road segments for safety improvement through speed management strategies and to illustrate how to select segments on the basis of model results. With the application of four statistical techniques (generalized additive model, negative binomial model, linear model, and empirical Bayes method) in three sequential steps to data collected on a 190-km section of expressway in South Korea, the study developed empirical models for selecting promising segments for safety improvement by the speed management strategies. This paper presents the five most-promising segments for implementing such strategies.

Stop versus yield on pedestrian-involved fatal crashes in the United States

In an effort to improve pedestrian safety, several states in the United States changed their pedestrian laws by changing the requirement that drivers yield to pedestrians in crosswalks to a requirement that drivers stop for pedestrians in crosswalks. This study examined whether this change had an effect on pedestrian safety in the United States, with its focus on low-speed roads. To examine the association between changes in pedestrian laws and changes in pedestrian-involved fatal crashes, three approaches were employed: before-after analysis, time-series analysis, and cross-sectional analysis. Pedestrian-involved fatal traffic crashes on low-speed roads were extracted from the U.S. national fatal crash database, the Fatality Analysis Reporting System (FARS), from 1980 through 2005. This study found no statistically significant reduction in pedestrian-involved fatal crashes attributable to changes in the laws, yet this finding is not definitive because of study limitations such as the omission of relevant exposure data.

Prediction of Fatality Rates for Comparison Between States

Comparisons using raw fatality rates (per vehicle miles traveled or per population, or both) are likely to lead to biased pictures of relative traffic safety in the states. This study attempts to provide better pictures for a state-by-state comparison by using random intercept models based on the state-level panel data from 2000 through 2004. The data include demographic economic characteristics and roadway characteristics of 46 U.S. states. Predicted fatality rates were computed with the estimated models and fixed factors tuned to the values for Virginia as a reference. New rankings were generated on the basis of predicted rates and compared with raw rankings based on the raw rates. About half of the top 10 safest states in the raw rankings dropped out of the top 10 in the new rankings, which took into account control factors. However, the uncertainty of the estimated random intercepts suggests that the new rankings can be varied considerably; this implies that the rankings should not be used as stable traffic safety yardsticks for comparison between states, although they are better than the raw rankings.