Easy Access to Pennsylvania's Supplementary Homicide Reports: 1980-2005

The data capturing detailed information on homicide incidents are collected annually by the FBI through its Supplementary Homicide Reporting (SHR) Program. The FBI prepares a data file containing this information and releases the file for public use. SHR data describing incidents occurring since 1980 were obtained by the National Center for Juvenile Justice (NCJJ) from the FBI.

Based on analyses of these data, NCJJ modified the contents of the files by deleting incompatible data and recoding values into more general categories. NCJJ also constructed and added weights to the records so that weighted analyses would yield state-level estimates. A detailed explanation of the procedures used by NCJJ to create the data file that underlies this package can be found below.

While NCJJ has benefited greatly from the work of others, any errors in this analysis and data presentation package are the responsibility of the National Center for Juvenile Justice alone. This package has been made available to the public so that interested users can explore the information compiled in the FBI's Supplementary Homicide Reporting Program. However, the computation and estimation procedures incorporated into this application reflect one of many possible approaches for developing estimates from the raw SHR data.

The SHR data file received by NCJJ captured information on each homicide incident reported by law enforcement. An incident may include one or more victims and one or more offenders. In incidents where there were multiple offenders it had to be assumed that all offenders were involved in the killings of all victims. The incident information included: the name and location of the law enforcement agency reporting the incident; the year of the incident; the age, sex and race of the victim(s); the weapon(s) used; the age, sex, and race of the offender(s) when known; and the offender's relationship to each victim (when known).

From these data two separate files were created. One is a victim-based file in which each record represents one victim with associated information on the weapon used and offender(s) involved. The other file is an offender-based file where each record represents one offender with associated information on the weapon used and victim(s) involved.

To clarify the file structure, a few examples are useful. Incidents with a single victim and a single known offender would generate one record in the victim file and one record in the offender file. If the offender were unknown (i.e., no description was available at the time of reporting either because there were no witnesses, the offender was not apprehended, or the offender information was not reported to the FBI), this incident would generate one record in the victim file with missing offender information and no record in the offender file. The decision was made to create a record in the offender file only when some information was available on the offender(s) because there was no obvious technique for determining how many offenders were involved in the incident and, consequently, how many records to create.

In incidents where there were multiple victims, the offender record contains information on the oldest victim. In incidents where there were multiple offenders, the victim records contain information on the oldest offender.

In summary the resulting victim data file contained a record on every murder victim. (All justifiable and negligent homicide incidents were deleted from the data files.) The offender file, more properly labeled the known offender file, contains a record on each offender for which some information was reported to the FBI. The SHR data are not able to track individuals across incidents; so an offender who was involved in more than one incident will have an offender record for each incident.

The underlying data structure of the raw SHR as compiled by the FBI permits up to eleven victims and eleven offenders to be placed on each record. In those rare instances where a crime results in more than eleven homicides, the victim information is spread over more than one record, with the offender information repeated on each. Without extraordinary knowledge of the incident, an analyst is unable to determine from the data file alone when separate records involve the same offenders.

From these master files, two analysis files were created to reduce the complexity of the data structure so that it could be processed efficiently with this package. The first task was to recode information on the weapon used and the victim-offender relationship into a smaller set of coding categories. The original codes from the SHR file were grouped as follows:

Victim-Offender Relationship

• Family - Husband, Wife, Common-law husband, Common-law wife, Mother, Father, Son, Daughter, Brother, Sister, In-law, Stepfather, Stepmother, Stepson, Stepdaughter, and Other family member.
• Acquaintance - Acquaintance, Boyfriend, Girlfriend, Ex-husband, Ex-wife, Employee, Employer, Friend, Homosexual relation, Neighbor, and Other known individual.
• Stranger - Stranger.
• Unknown - Relationship unknown.

Weapon Used

• Firearm - Firearm, Handgun, Rifle, Shotgun, and Other gun.
• Knife - Knife, Cutting instrument, Icepick, Screwdriver, Axe, etc.
• Blunt object - Hammer, club, etc.
• Personal - Hands, Fists, Feet, and Other personal weapon.
• Other - Poison, Pushed out window, Explosives, Fire, Drugs, Drowning, Strangulation, Asphyxiation, and Other unknown weapon.

The annual variations in the proportion of homicide incidents reported to the FBI could easily confuse any interpretation of the data if the analyses relied solely on reported data. Therefore, annual estimates had to be developed. Each year in its Crime in the United States series the FBI publishes estimates of the number of murder victims in the United States and in each State. These estimates are based on the SHR data and information about the nonreporting law enforcement agencies. Annual national weights were developed for the victim-level data file by dividing the number of estimated murder victims in a particular year (from the Crime in the United States series) by the number of murder victims in the SHR data file for that year. Using this procedure, the annual estimates of the number of murder victims as reported in the Crime in the United States series are replicated by this package. This procedure assumes that the reported data have similar characteristics and relationships to the unreported incidents. There is no way of assessing the validity of this assumption. This same national weight was also applied to each record in the offender-level file. The weighting of both victim and offender records assumes that if non-reporting agencies did report their data (including the pattern of known and unknown offender information), these data would be similar to that in the reported data.

Using the same approach, Pennsylvania-specific weights were developed for each year. These weights were based on the assumption that incidents reported by law enforcement agencies within Pennsylvania were similar to those incidents not reported within the state.

The National Center for Juvenile Justice prepared tables documenting for Pennsylvania the number of homicide victims reported in the annual SHR data files along with the estimated number of homicide victims reported in the FBI's annual Crime in the United States reports. Click here to download this file [.xls format].

The source Supplementary Homicide Reports (SHR) data files include the incidents of the September 11, 2001 terrorist attacks. The records involving the victims of Flight 93 in Somerset County are not included in this application. The FBI prepared a special report on the victims of these attacks in Section V of the Crime in the United States 2001 publication, which is available here.