Motor vehicle crashes have been identified as a leading cause of death all over the world. To better promote traffic safety and protect lives of the traveling public, transportation agencies develop crash databases to effectively manage reportable motor vehicle crashes. Crash reports from law enforcement authorities usually serve as the primary source for motor vehicle crash data. However, unintentional errors in law enforcement crash reports, errors in state-maintained crash databases, and the migration and rearrangement of data could introduce problems of data consistency in crash databases. One such identified data inconsistency is regarding horizontal curves. It can result in misidentifications of curve-related crashes, which can affect safety analysis using these data. This is significant since many studies have related horizontal curves to crash frequency and severity. To solve this problem, this study proposed a methodological procedure for evaluating (i.e., identifying, classifying, and quantifying) curve-related misclassifications in crash databases. The applicability of the proposed methodological procedure was illustrated through a case study using the Crash Records Information System (CRIS) maintained by the Texas Department of Transportation (TxDOT). The results indicated that transportation agencies could employ the proposed methodological procedure to evaluate inconsistencies associated with curve-related crash data both effectively and efficiently.