Convergence of Self-Report and Archival Crash Involvement Data: A Two-Year Longitudinal Follow-Up
Arthur, Winfred, Jr., Bell, Suzanne T., Edwards, Bryan D., Day, Eric Anthony, Tubre, Travis C., Tubre, Amber H., Human Factors
Using a relatively large and representative sample, the present study constructively replicates and extends Arthur et al. (2001) by investigating the convergence between self-report and archival data and, also, the relationships between these criteria and specified predictors using both predictive and postdictive criterion-related validation designs. The value and importance of constructive replication (Hendrick, 1990; Kelly, Chase, & Tucker, 1979), defined as "research that tests the same hypothesized relationships among the same theoretical constructs as a given earlier study but varies the 'operationalization' of those constructs" (Eden, 2002, p. 842), resides in its ability to strengthen confidence in and enhance the credibility of generalizations when the hypothesized relationships are confirmed using different methods. Along these lines, one of the critical goals of this study was to compare the extent to which predictive and postdictive designs yield similar (or different) criterion-related validities.
There is a reasonably large body of research examining the predictive validity of individual differences in motor vehicle crash involvement (for reviews see Arthur, Barrett, & Alexander, 1991; Elander, West, & French, 1993; Guastello, 1993), and in this research self-report accounts of crash involvement have been the most commonly used criterion (Ball & Owsley, 1991). However, although several reasons may explain the prevalence of self-reports (Elander et al., 1993), questions about their veracity and accuracy have been raised (see Elander et al., 1993; Harano, Peck, & McBride, 1975; Loftus, 1993). Consequently, archival data have been considered as an alternative to self-reports.
Self-Report and Archival Data
For several reasons, self-report measures of crash involvement are the most commonly used criterion measures. For instance, they are relatively easy to collect and enable researchers to sample all crash types. However, self-report measures have been criticized for being inherently susceptible to cognitive and communicative biases, such as selective memory and question comprehension (Elander et al., 1993; Harano et al., 1975; Loftus, 1995; Schwarz, 1999). In contrast, archival records of crash involvement--which are usually obtained from state law enforcement agencies and insurance companies--are assumed to be free of the biases associated with self-report measures. However, this type of data carries its own set of disadvantages. For example, it is generally more difficult to obtain. Another major disadvantage of archival driving records is that they are often incomplete because (a) police selectively report crash involvement, (b) moving violation tickets are removed from drivers' records in exchange for driving course credit, (c) the parties involved sometimes choose not to report the crash, or (d) interstate information exchange of data on crashes and moving violations is incomplete (Bums & Wilde, 1995; McGuire, 1973; Smith, 1976).
Given the important trade-offs between using either archival or self-report data in motor vehicle crash research, few studies have investigated the convergence between these two sources of criterion data. A handful of studies have collected crash data via both self-reports and state records (e.g., Dalziel & Job, 1997; Hilakivi et al., 1989; Marottoli, Cooney, & Tinetti, 1997; Owsley, Ball, Sloane, Roenker, & Bruni, 1991), and some have included performance on a driving simulator (e.g., Szlyk, Alexander, Severing, & Fishman, 1992; Szlyk, Fishman, Severing, Alexander, & Viana, 1993; Szlyk, Seiple, & Viana, 1995). However, most of these studies have used very restricted and specialized samples (e.g., older adults or participants with specialized medical conditions, such as retinitis pigmentosa and juvenile macular dystrophies). In addition, the convergence between the criterion sources has typically not been assessed because it was not the primary focus of the study. …