Imputing Unreported Hate Crimes Using Google Search Data

Abstract

Federal law requires the Attorney General to collect data on hate crimes from states and municipalities, but states and localities are under no obligation to cooperate. Their production of such data varies considerably across jurisdictions. This Article addresses the resulting geographically uneven “missing data” problem by imputing unreported hate crimes using Google search rates for racial epithets. It uses two alternative definitions of which jurisdictions more effectively collect hate crime data. The first consists of all states that were not part of the erstwhile Confederacy; the second comprises mostly states with statutory provisions relating to hate crime reporting. We regress hate crime rates for racially-motivated hate crimes with African-American victims on Google searches and other relevant variables over 2004-2015 at the state-year level for each of these classes of benchmark states. Adding Google search rates substantially enhances the capacity of such models to predict hate crime rates among benchmark states. We use predicted values from these regressions to impute hate crime rates that other jurisdictions – that presumptively do not robustly report hate crimes – would counterfactually have reported if their reporting were as robust as that for the benchmark states. The results imply a substantial number of unreported hate crimes, concentrated in particular jurisdictions. It also illustrates the potential value of internet search rates as a source of data on phenomena that are otherwise difficult to measure.

Keywords: Hate crimes; Google search rates; Racial animus; Crime reporting; Prediction

JEL Classification: K42