Crash and injury reduction following installation of roundabouts in the United States.

OBJECTIVES: This study estimated potential reductions in motor vehicle crashes and injuries associated with the use of roundabouts as an alternative to signal and stop sign control at intersections in the United States. METHODS: An empiric Bayes procedure was used to estimate changes in motor vehicle crashes following conversion of 24 intersections from stop sign and traffic signal control to modern roundabouts. RESULTS: There were highly significant reductions of 38% for all crash severities combined and of 76% for all injury crashes. Reductions in the numbers of fatal and incapacitating injury crashes were estimated at about 90%. CONCLUSIONS: Results are consistent with numerous international studies and suggest that roundabout installation should be strongly promoted as an effective safety treatment.

Crash reductions related to traffic signal removal in Philadelphia.

The effect on intersection crashes of converting one-way street intersections in Philadelphia from signal to multiway stop sign control was estimated. Using crash and traffic volume data for a comparison group, regression models were computed to represent the normal crash experience of signal controlled intersections of one-way streets, by impact type, as a function of traffic volume. An empirical Bayesian procedure was used to estimate what would have been the expected number of crashes at the converted intersections had they not been converted. The empirical Bayesian estimates were compared with actual counts of crashes after conversion. Estimates were obtained for different classes of crashes categorized by impact type, day/night condition, and impact severity. Aggregate results indicate that replacing signals by multiway stop signs on one-way streets is associated with a reduction in crashes of approximately 24%, combining all severities, light conditions, and impact types.

A procedure for allocating a safety improvement budget among treatment types.

Highway agencies often have to decide how to allocate a road safety budget among different safety treatments. While some of the traditional resource allocation procedures can provide information to assist in making this decision, they are not tailor-made for this specific problem. In particular, uncertainty in accident frequency estimates can lead to difficulties if the procedure is inherently deterministic. A procedure is presented for using recent safety estimation theory and benefit-cost analysis to accommodate uncertainty in accident frequency estimates while addressing the specific decision-making problem at hand. In the procedure, a site can be treated if the difference between the safety benefit and the treatment cost--the net benefit--is large enough. Recognizing that safety benefits vary across sites and treatment types, the procedure allows a budget to be allocated to different programmes but ensures that the smallest net benefit--the net marginal benefit threshold--is the same for all programmes. The procedure estimates the number and costs of sites for each treatment type for which the net benefit exceeds a specific threshold that is adjusted such that the total treatment cost is in line with the available budget. While this methodology can be useful for agencies responsible for road safety improvements, the data requirements could pose a problem for some jurisdictions. A further difficulty is that the computations required are extensive, particularly in cases where uncertainty to other data items requires the use of sensitivity analysis. To facilitate these computations, a computer programme is available from the authors.

Estimating the accident potential of an Ontario driver.

To run a "demerit point" program, one uses routinely available information about drivers to identify those who are most likely to have an accident in the near future. On the basis of a four-year record for a large sample of Ontario drivers, we have examined several tools for the identification of such drivers and investigated how they perform. Each driver is thought to have an expected number of accidents, m. In a group of drivers with common traits (such as age, gender, record of convictions and accidents) the ms have a mean E(m) and a variance VAR(m). Estimates of E(m) and VAR(m) for all combinations of traits can be obtained within the framework of a multivariate statistical model. The same estimates can then be used to judge how well a model identifies drivers who have a large m. In such a multivariate model it is important to use data about previous accidents and convictions. However, the accuracy with which the m of a driver can be estimated is not improved much by distinguishing between offence type or between accidents as being "at fault" or "not at fault". Without much loss in estimation accuracy, one may attach a weight 1 to a conviction and 2 to an accident. Model performance is described in tangible terms: how many accidents are recorded by the drivers identified by a model, what proportion of identified drivers are "false positives," how many drivers with high m remain unidentified. We conclude that by using a multivariate statistical model one can do substantially better than by using a demerit point scheme in which points are assigned to offenses on the basis of their perceived seriousness. However, even when the best model is used to identify a large group of drivers, many will be false positives.

Relating the effect of safety measures to expected number of accidents.

Traffic engineers have long believed that the effectiveness (percentage reduction in accidents) of a safety measure is greater at locations which had many accidents than at those which had few. That this type of belief has often translated into warrants is evidence of the strength of this belief. Recent research has raised doubts as to whether this phenomenon is real or merely a manifestation of regression-to-the-mean. This paper addresses this issue in the context of an examination of the safety effect of converting 222 intersections in Philadelphia, PA, U.S.A. from two-way to all-way stop control. The results indicate that the conversions were more effective at intersections expected to have many accidents than at the relatively safe intersections. One of the important implications of this finding is that, for measures for which this phenomenon exists, effectiveness cannot be specified as a single accident reduction factor as is currently the practice.