# An Examination of Supplementary Homicide Reports and Fatal Encounters Data on Police-Caused Homicides

## TL;DR

This paper examines police-caused homicides using official and independent data sources, comparing their accuracy and reliability over time.

## Contribution

The study introduces multiple imputation to improve estimates of police-caused homicides from 1980 to 2010.

## Key findings

- Fatal Encounters data underreports police-caused homicides in the early 2000s compared to the early 2010s.
- Imputed estimates from Supplementary Homicide Reports provide more reliable national trends from 1980 to 2010.
- Comparing imputed and non-imputed data reveals consistent patterns in police-caused homicides over time.

## Abstract

The two official sources of police-caused homicides, National Vital Statistics System and FBI Supplementary Homicide Reports, have been widely criticized for underreporting and measurement error. Independent data collections emerged in the 2010s after several high-profile incidents of the police killing unarmed citizens. These collections include Fatal Encounters (FE), Mapping Police Violence, The Guardian, and The Washington Post. The validity of these datasets is unknown, and none of them extend before the year 2000. Researchers interested in studying police-caused homicides before 2000 are left with the two official sources with known data problems. This study performs multiple imputation on Supplementary Homicide Reports data to obtain national estimates from 1980 to 2010. We present a comparison and validity check with FE to our imputed estimates and find FE data to be underreported in the early 2000s compared to the early 2010s. We also perform a statistical analysis from 1980 to 2010 on police-caused homicides using non-imputed and imputed values and compare results.