Pursuing Precision in Criminology: Why Ordinal Data Demand Ordinal Methods

Criminological research frequently examines ordinal outcome data, yet the common practice of treating these data as continuous using linear models can obscure patterns, introduce imprecision, and bias estimates. We advocate for the appropriate analysis of ordinal outcomes with ordered regression models. We demonstrate how linear models misrepresent ordinal distributions using simulated data before reanalyzing select results from Pickett, Graham, and Cullen's (2022) study on racial differences in fear of police. Our analysis showcases how ordered regression models, coupled with effective visualization techniques, provide more accurate estimates than linear models—revealing that linear approaches substantially underestimate racial disparities. For instance, cumulative probit models correctly estimate that nearly one-third of Black participants are “very afraid” of police killing them, while linear models underestimate this by nearly two-thirds. By respecting the ordinal nature of data, criminologists can move beyond broad directional statements and quantify effect magnitudes with greater precision, fostering a more robust, evidence-driven discipline capable of addressing its “precision crisis.”