Self-Selection Bias in Driver Performance Studies

Abstract

It is generally understood that there is some bias involved in selection of participants for driver performance studies, but little is known about the extent of this problem. To execute a National Cooperative Highway Research Program project on the effectiveness of traffic control devices, a sample of younger and older drivers was required. The purpose was to gain insight into the bias introduced through participant selection and self-selection. Of interest is whether the drivers who participate in projects for which driving or other testing is required tend to be different than the general population of licensed drivers. Results indicate that, compared with nonparticipants, participants are more active, more likely to travel and drive, less likely to avoid driving in certain circumstances, and less likely to have vision problems. The implication is that project participants represent more highly mobile and confident drivers than would be found in a random sample of the general population. However, project participants also had higher percentages of total accidents and violation points and were involved in more severe accidents than nonparticipants. These problems may be somewhat mitigated, though, by higher driving exposure for participant drivers. Such self-selection bias needs to be considered whenever research like this is undertaken.