Cover Image

Not for Sale

View/Hide Left Panel
Click for next page ( 13

The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement

Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 12
12 as a comparison group for nighttime crashes, based on the traffic volume changes in both the treatment and comparison assumption that only nighttime crashes would be affected by groups since such volumes may be relatively easy to acquire PRPMs. As discussed in Section 2.3, there is evidence that and influential in the evaluation results. PRPMs affect driver behavior during daytime as well, man- ifested by changes in positioning in the lane and significant reductions in lane encroachments, which would be expected Time Trends to impact both head-on and run-off-road crashes. Conse- quently, the use of daytime crashes as a comparison group is Areawide safety changes over time because of many fac- inappropriate. Table 2-5 shows both significant reductions and tors, such as weather conditions, driver demographics, and increases in crash frequency. Indeed, the two largest studies vehicle technology. Reporting levels also directly affect crash show opposing effects--one with 662 treatment locations (13) data. Often either the minimum damage dollar value changes showing a 22-percent reduction in nighttime crashes, and for PDO crashes or the reporting level by police changes. the other with 452 treatment locations (11) showing a 15- to PDO crash data from jurisdictions that have switched from 31-percent increase in nighttime crashes. Re-analysis (15, 14) 100-percent police reporting to self-reporting crash data must of the second study, with its troubling result, continued to be used carefully when accounting for the safety effect of show a statistically significant increase in nighttime crashes PRPMs. Again, most of the previous studies attempted to at some locations. As will be seen in Section 2.3, there are account for the safety effect of PRPMs by using daytime mixed findings with respect to speed and an indication that crashes at the same sites as a comparison group. However, if the installations of PRPMs have a safety effect on daytime speed effects may be site specific. Changes in speed, along crashes and/or the time trends between daytime and night- with the effects of PRPMs on daytime encroachments, may time crashes differ, using daytime crashes at the same sites be factors in the mixed safety effects. as a comparison group will result in errors in the estimate of safety effectiveness. 2.2.2 Methodological Problems in Past Research Regression-to-the-Mean The relative safety at any location is a function of all road- way, environmental, and driver characteristics. A change in If PRPMs were installed at a location experiencing a ran- any of these factors from the before to after period affects domly high number of crashes in the before period, then the safety. In order to derive an accurate estimate of the safety number of crashes in the after period would be expected to effect of PRPM installations, it is important to separate the decrease with or without the installation of PRPMs. This effect of other changes, including the changes described in phenomenon (known as regression-to-the-mean, or RTM) is the following sections. often a factor when study sites are selected based on crash history. Not only could RTM exist for the crash type or loca- tions of interest, but it could also exist in the comparison Changes in Traffic Volumes group, and this existence could exaggerate the positive effects of a measure. For example, Orth-Rogers and Associates (18) Safety directly relates to traffic volumes. As a result, the dif- cite a study by Khan (20) in which 184 sites were selected ference in traffic volumes between the before and after peri- from high-hazard locations having four or more crashes in ods affects the expected difference in the number of crashes 1 year before the installation of PRPMs. At a group of con- between the before and after periods. In most of the previ- trol locations where PRPMs were not installed, it was found ous studies reviewed, traffic volumes were not accounted for that the total number of crashes increased. However, at the explicitly. Daytime crashes have most often been used to treated sites, both nighttime and daytime crashes were reduced. control for changes in safety, on the assumption that these It is clear that, given the site selection criterion, RTM will are unaffected by the PRPM installation. In the treatment- exaggerate the positive effects noted in the Khan study (20) comparison experimental design used by several researchers, and may even explain in entirety the reduction in daytime it was assumed that traffic volume changes are controlled for crashes. However, the increases in the control group may because the percentages of AADT during day and night should also be due to RTM because these locations may have been not change significantly in the before and after time periods untreated because they fell into a group that had fewer than (18). This may be a reasonable way of accounting for traffic the average number of crashes in 1 year. This RTM would volume changes if this assumption is met, providing that the exaggerate the effects of PRPMs even more. changes are small and that the relationship between crashes Only one of the previous studies, Pendleton (16), directly and traffic volume is approximately linear. In the studies accounted for RTM effects. The treatment-comparison exper- reviewed, it was unclear if these provisions were in fact met. imental design can, in principle, use comparison sites to con- It seems reasonable that one should not rely on such assump- trol for RTM, but the treatment and comparison sites need to tions and that one should seek explicit ways of accounting for be matched on the number of crashes. In practice, controlling