Introduction Motor vehicle collisions (MVCs) continue to place an increased burden on both individuals and health care systems. Self-reported and state-recorded police reports are the most common methods for MVC evaluation in epidemiologic studies, with varying degrees of agreement of information when compared in previous studies. The objective of the current study is to address the differences in MVC reporting and provide a more robust measure of the agreement between self-reported and state-recorded MVCs in a community dwelling population of older adults. Methods A three-year prospective study was conducted in a population-based sample of 2000 licensed drivers aged 70 and older. At annual visits, participants were asked to self-report information on any MVC that occurred over the prior year where police were called to the scene. Information on police-reported MVCs was also ascertained from Alabama official state-recorded databases. The kappa coefficient was calculated to determine overall agreement between any self-reported and state-recorded crashes, as well as the raw number of crashes reported. In addition, agreement was stratified by demographics, health status, medication use, functional status (i.e. vision, cognition), and driving habits. Results 1747 participants who completed three years of follow up were involved in 225 state-recorded MVCs and 208 self-reported MVCs yielding overall substantial agreement between any self-report and state-recorded MVC (kappa = 0.64). Cumulative number of self-reported and state-recorded MVCs was also compared, with agreement slightly reduced (kappa = 0.55). The clinical characteristic resulting in the greatest variation in agreement with drivers was impaired contrast sensitivity showing better agreement between self-reported and state-recorded MVCs (kappa = 0.9) than those with non-impaired contrast sensitivity (kappa = 0.6). Conclusion Study results showed substantial agreement between self-reported and state-recorded MVCs for any MVC involvement among the study population. When examining the reporting of the total number of MVCs over the three year period, agreement was reduced to a moderate level. There was consistency in agreement across MVC risk factors except among individuals with contrast sensitivity. These findings have implications for the design and analytic planning of epidemiologic and clinical research focused on MVCs.