Book Review By Carmen Daecher

Observational Before - After Studies in Road Safety

by: Ezra Hauer, Pergamon Press, 1997.

Whenever a road improvement is implemented, no matter how small or large, the intent is to improve safety - either by eliminating unsafe elements or improving the quality of traffic flow. Our justification for such improvements is usually considering a short time span of activities (such as accidents), with the assumption that through improvements such accidents can be eliminated.

Hauer, in his book, (Observational Before - After Studies in Road Safety) suggests that this approach is misguided for many reasons. Under laboratory-type conditions, roadway improvements could be made and the actual safety effectiveness of these improvements could be measured. These improvements would then be compared against expected events before such improvements were made. However, Hauer is quick to point out that within the arena of traffic improvements, we cannot have the benefit of laboratory conditions.

Accordingly, Hauer defines as a basic premise in his book that we must predict what would have been the safety effectiveness of a particular roadway element without any change and compare it to an estimate of what will be the safety level with a change or improvement.

Hauer addresses the basic data foundations for a proper approach to this method. Accident frequency rather than accident rate should be used because of its non-linear relationship to traffic volumes. And the trending of these frequencies over extended periods of time rather than for a relatively short three or five year period is advocated. Finally, the proper selection of accidents that should be affected as a result of the road improvements (targeting accidents) and those accidents which will not be affected (comparison accidents) must be clearly defined.

This foundation of data becomes very important as Hauer explains his statistical basis for before/after studies. Important to the statistical treatment for before/after studies is the fact that there are known and unknown variables which may change over time that must be measured or estimated. Current before/after studies do not take into account factors or influences that change over time and affect accident experience. This is largely due to the fact that some factors may not be known, and other factors are not considered in the overall measurement of safety effectiveness of an improvement (e.g. increased traffic because of an installation of a traffic signal). The same safety improvement at different locations could result in different measures of effectiveness. This variability of safety treatment and effect is important towards understanding how probable a mean effect would be for a particular type of treatment at any location.

Hauer advocates estimating the effect of safety treatments through specific estimates of targeted accident frequencies. He also advocates the observation and measurement of untreated entities that are similar to that which will be treated to predict the unrelated safety performance at an improvement location (comparison groups).

Hauer makes a point that using accident data alone as a basis to justify safety improvements causes bias because of an unusually high or low number of accidents as compared to the mean number of accidents. Thus, the Emperical Bayes approach is advocated for precision and for use of unlimited time intervals for data collection in the before period for safety improvements.

As you might expect, this book is replete with algebra and heavy statistics. Hauer attempts to explain in common language that which he also derives through mathematics. However, be aware that this book can be burdensome in places to understand and comprehend.

Simply put, it is Hauer's belief that a proper statistically based approach to evaluating before/after safety effectiveness is the only way to do it properly. He suggests that "common sense" and "intuitive logic" may not be so common or so intuitive. And, as previously discussed, the flaws in using high accident rates and short time trends renders most safety improvement decisions as flawed. While Hauer, for the most part, presents a compelling argument of how to more properly perform before/after assessments of safety improvements, the output of his methods for selecting safety improvements might be hard to accept in the public sector. To suggest what is the equivalent of a paradigm shift in thinking towards selecting safety improvements is a little far reaching. Hauer does not advocate the use of dollars (cost/benefit) in his methods. Or, if he could, it is not discussed to any extent. Simply adding dollar figures in his statistical approach would be very difficult. Yet, cost/benefit has both appeal and foundation for public decisions. Also, the relative weight or sensitivity of unknown variables (e.g. traffic signal installations attracting more traffic than expected), is not adequately considered.

In summary, Hauer's statistical approach is solid and thought provoking. Widespread acceptance and use of his methods is doubtful.

*** Click here to visit the NADME training web site ***