During the summer of 1987, color coded mailback questionnaires that identified belted and unbelted North Carolina drivers were handed out at the 72 sites that constitute the probability sample for determining the belt use rate in North Carolina. By obtaining identifying information ostensibly to determine the winner of a $500 prize from among the respondents, accident and violation records from the North Carolina driver history file were linked to the belted and unbelted respondents. Analyses found that non-belt users are overrepresented in accidents and violations.