Early cross-sectional studies utilized self-reports of past involvement, while later research efforts focused on estimated future behavior or sometimes both. More recently, deterrence researchers have employed longitudinal designs that have allowed for measures of actual future illegal behavior. However, no perceptual deterrence research to date has employed a panel design to compare deterrence theory findings for future estimates of involvement in illegal behavior at Time 1 with actual future behavior measured at Time 2. The current research will employ a two-wave panel of adults from a large midwestern metropolitan area to compare tests of deterrence theory hypotheses for both of these measures of involvement in the offense of driving while under the influence of alcohol. Findings suggest that the substantive conclusions are similar regardless of the measure of the dependent variable employed: Informal threats of sanctions are more important predictors of involvement in drinking and driving behavior than are formal threats. Given the numerous problems facing longitudinal deterrence research (i.e., costs, sample attrition, and appropriate lag times), similar results from future panel studies might generate a renewed confidence in cross-sectional designs that employ future estimates of involvement in illegal behavior as proxies for actual future involvement in illegal behavior. (Author abstract) 5 tables, appendix, 20 notes, and 50 references.