Citation

Fox JA. Homicide Stud. 2004; 8(3): 214-254.

Copyright

(Copyright © 2004, SAGE Publishing)

DOI

10.1177/1088767904265592

PMID

unavailable

Abstract

Although the so-called "dark figure" crime measurement problem has never been a major concern for homicide researchers, the Supplementary Homicide Reports (SHR) as well as other local data series on murder still are plagued by other kinds of missing data issues. Most prominent is missingness in data pertaining to offender characteristics as well as to victim-offender relationship that results from uncleared cases. Ignoring unsolved homicides would, of course, seriously understate calculated rates of offending by particular subgroups of the population, would distort trends over time among these same subgroups, and would bias observed patterns of offending to the extent that the likelihood of missing offender data is associated with offender characteristics. This article presents several approaches for overcoming missing data problems in the 1976-2001 cumulative SHRdata file. First, a weighting procedure is described that uses characteristics of known offenders to serve as proxies for those of unidentified perpetrators. The weighting procedure included in the SHR file archived at ICPSR as well as an enhanced version are both presented and compared. Next, a "hot-deck" imputation strategy is applied to fill in missing offender attributes based on similar cases for which the offender is known. Finally, the matter of imputing victim-offender relationship data is discussed. Because this form of missingness cannot be assumed to occur at random, an ad-hoc procedure for estimating the number of intimate homicides among the pool of unsolved slayings is presented.