The Accuracy of Supplementary Homicide Report Rates for Large U.S. Cities

• In Event: Measuring Violence and Responses to Violence: Paths to Better Data

Abstract

We investigated factors that influence the precision of homicide rate estimates from the Supplementary Homicide Report (SHR) for large U.S. cities during the five-year period centered on 2000. Homicide rates from the National Vital Statistics System provide a parallel measure of homicide and the logical basis for estimating the reliability and standard error of measurement.

The major factor limiting the quality of estimates is the large amount of undocumented unit-missing data in the SHR. The initial estimate of reliability was relatively low (0.87) and the standard error of measurement more than 3 homicides per 100,000 residents. However, when cities that reported inconsistently were removed, the estimated reliability was much higher (0.99) and the standard error of measurement was 1.2 homicides per 100,000 residents. Rates for specific demographic groups also may be imprecise where the number of residents in those groups is relatively small. For instance, even in completely reporting cities, the standard error of measurement for African American homicide rates was more than 5 homicides per 100,000 residents.

Authors

• David McDowall, University at Albany, SUNY

• Colin Loftin, University at Albany, SUNY

• Karise M. Curtis, University at Albany, SUNY