With the idea of the “dark figure” of crime in mind, the raw numbers of  officially-recorded misconduct would be expected to be lower than self-reported misconduct. Steiner & Wooldredge (2014) found that to be true for assaults & drug offenses, but not for theft.

They also examine if the effects of various predictors differ across the two offense types, and largely find that they do not. For example: the coefficients for the effect of being African American are not (statistically) significantly different across the two data sources. Steiner & Wooldredge (2014) seem pretty convinced of the validity of official measures for the vast majority of examinations.