Center for Problem-Oriented Policing

Appendix B: Improving Data Integrity

The extent and the importance of data limitations vary from one problem to another. Data source and quality are seldom an issue for some problems-data on bank robberies are highly accurate and reliable for detecting revictimization. Poor data mask revictimization in other problems, such as burglaries and domestic violence. While data quality can often be improved, a decision to undertake such effort should consider the following factors:

  • Extent of data limitations for problem being examined
  • Presumed value of improving data quality, including short-term and long-term benefits
  • Availability of alternate data or methods to improve data quality
  • Resources necessary to improve data quality

    Depending on the problem being examined, more reliable data may be necessary. For example, if data suggest that private property owners should adopt costly preventive measures, extremely reliable data may be necessary to educate, encourage, convince, or coerce them into doing so.

    An important reason to carefully document the extent of repeat victimization is to provide a foundation for a response-including getting buy-in from others who may help reduce victimization. Depending upon the type of problem being examined, the integrity of data can be easily improved:

  • Collecting additional data. As discussed in Appendix A, offense reports can easily be modified so that victims are asked a short series of questions about prior victimization.

    Offense reports can also be modified to record all victimizations, regardless of whether charges were filed, to identify repeat offenses that may be masked by hierarchy rules. This is consistent with NIBRS (National Incident-Based Reporting System) procedures and will provide more complete information about victimization.

  • Creating data layers. In most jurisdictions, common types of locations- bars and nightclubs, budget motels, schools, banks, movie theatres, convenience stores, gasoline stations-will routinely generate similar types of crimes. For example, assaults will occur at the bars, vandalism at schools, robberies at banks, beer runs at convenience stores and so on. Police should integrate location identifiers into their records management systems so offenses among these location groups can be routinely monitored.
  • Combining data sources. For some problems, victim interviews will provide better information about the impact of responses to problems such as domestic violence than police records. Crime reports may underestimate or overestimate the impact of responses-increased reporting does not demonstrate increased victimization and reduced reporting does not demonstrate reduced victimization. Conducting follow-up surveys with victims who previously reported an offense to the police provides more reliable measures of revictimization. Other data sources-such as calls for service, offense reports, arrests, property recovered, and warrants-can also be combined to create more comprehensive databases, for example, facilitating searches by both name, address and unique identifiers when available.
  • Increasing data precision. Incident reporting forms and police records can be revised to improve specific location information. Whenever possible, police should avoid recording offense locations as intersections or hundred-block addresses. Each unit of an address should be routinely recorded-including building name or number, floor, office suite, or apartment number. For properties that have multiple addresses (such as an apartment complex that has addresses on more than one street), records management systems should be modified to link addresses. An alternative is to use mapping to reveal near-repeat offenses that are recorded on different streets. For offenses that occur in large public or private spaces such as parks or parking lots, police can use global positioning system (GPS) equipment to record precise coordinates for offenses.
  • Improving data quality. Simple procedures, such as mapping offenses, can identify "match rates," or the proportion of offenses that can be linked to valid addresses. Consistent recording of offense types is very important. The distinctions between some offenses may seem trivial but have implications affecting revictimization. For example, if appliances are stolen from a house under construction, this might be classified as larceny or theft from a construction site, residential burglary, or commercial burglary. Offense classifications and recording practices should be monitored to ensure consistent classification and recording methods.
  • Data sharing. To detect revictimization across boundaries and jurisdictions, police agencies should routinely share data about victims. This is practical for counties or neighboring jurisdictions that participate in regional information networks. Individual victims may (unsuccessfully) relocate to avoid revictimization and revictimization of commercial properties. Virtual victims may be especially common across boundaries.