The Validity of Robbery-Murder Classifications in Baltimore
Loftin, Colin, Violence and Victims
The most widely used source of nationwide data on homicide in the United States is the Supplementary Homicide Report (SHR) data compiled by the FBI as part of its Uniform Crime Reporting System. This paper describes a study of the reliability of robbery-murder classifications by the SHR for Baltimore, Maryland during 1983. The research is exploratory, but indicates a high level of inconsistency in the data. Of the 42 cases that were classified as robbery murders by the SHR or the replication study, only 20 were classified that way in both studies. The high level of unreliability can be attributed to three major problems: (1) the SHR codes are mutually exclusive, but many homicides could be placed in several of the categories; (2) there are few systematic rules for classifying ambiguously motivated homicides; and (3) the SHR placed too many cases in the "unknown" category. It appears that the reliability of the coding could be increased substantially with relatively minor changes in the procedures currently used.
There are two nationwide sources of data on homicides in the United States: (1) data from death certificates compiled by the National Center for Health Statistics (NCHS) and distributed as part of the vital statistics system, and (2) the Uniform Crime Report data compiled by the Federal Bureau of Investigation (FBI).
Available since the late nineteenth century (Wolfgang & Zahn, 1983), the vital statistics are a continuous series for death registration areas. Because they rely on information extracted from death certificates, these data are naturally victim centered and provide no information about the offender or the circumstances surrounding the offense. They are, nevertheless, extremely valuable and have been a cornerstone of research on U.S. homicide during the past fifty years.
The Uniform Crime Report data, based on reports from local law enforcement agencies, have become especially attractive to researchers in recent years because they contain rich information on the offender and the circumstances surrounding the offense, as well as information about victims. The FBI data provide a continuous series on the incidence of homicide since 1930 (Wolfgang & Zahn, 1983). However, the Supplementary Homicide Reports (SHR), which contains the detailed data on homicides, have only recently been available to researchers. Until 1966 the SHR was not filed by most agencies and information on the offender was not collected before 1976.
The FBI now produces annual public-use tapes extending back to 1966. Access to these data has stimulated basic research on homicide (Jason, Strauss, and Tyler, 1983b; Loftin & Parker, 1985; Parker & Smith, 1979; Smith & Parker, 1980), and gready expanded the possibilities of doing policy motivated research on topics like the regulation of firearms (Pierce & Bowers, 1981; Cook, 1983; McDowall, 1986) and the sentencing of capital offenses (Gross & Mauro, 1984; Paternoster, 1984; Radelet & Pierce, 1985).
Criminologists believe that, relative to other crimes, homicides are accurately reported (Sykes, 1978), but there is little systematic research on the quality of homicide data. The vital statistics data collection system has been studied extensively (Gittelsohn & Royston, 1982), but homicides are a tiny fraction of all certified deaths and therefore most of this research has little direct relevance to the quality of homicide data. There appear to be no studies that directly evaluate the validity of the vital statistics data on homicide and only two studies that deal with specific subsets of the data. Jason, Carpenter, and Tyler (1983a) present evidence of underreporting of the homicides of infants under a year of age and Sherman and Langworthy (1979) found underreporting of homicides by police officers.1
A similar situation exists with the Uniform Crime Report data. In spite of its prominence in recent research, there are no studies of the validity of the SHR data. …