Appendix A: Data Sources and Limitations

The type and quality of data used for estimating RV influences the amount of revictimization that can be detected. This, in turn, may mask key elements of revictimization and limit how we can determine the impact of reducing revictimization.

Police Data

Recorded crime data are most often used to detect repeat victimization.

† Citizen-initiated calls for service (911 calls) are valuable for identifying repeat victimization, particularly by address. While call data have limited variables and are not subject to the same level of verification as incident reports, they can provide important insight into related problems, such as 911 hang-ups and domestic violence, or groups of problems, such as those occurring at schools or bars.

However, the actual amount of repeat victimization is often masked by common problems, including:

• Underreporting. Much victimization is unreported but reporting varies from one offense to another. Virtually all bank robberies, vehicle crashes, and most vehicle thefts and commercial robberies are reported. On the other hand, most domestic violence, much theft, and much vandalism are unreported. Underreporting will limit the ability to detect repeat victimization.
• Incomplete or inaccurate crime reports. Offense reports completed by responding officers are rarely used for analysis, and often contain incomplete or inaccurate information about offenses, including poor address or victim information, or missing information. Analysis of repeat victimization will likely provide an opportunity to involve line personnel in improving the quality of their initial reports; this may contribute to reductions in their workload and provide opportunities to participate in the most effective crime prevention strategies.
• Time period covered by the data. A single year of offense data is customarily used for crime analysis but provides a weak basis for detecting repeat victimization. Undercounting occurs when offenses straddle calendar years, with an initial offense occurring late in one calendar year and a recurrence early in the next year. The problem is more common when the time course between offenses is longer. While this data may be all that is available to police, truncating data by calendar year substantially masks repeat offenses.
• Data accuracy. Police data are fraught with errors that reduce the amount of repeat victimization that can be detected. Some errors reflect data entry problems or errors on reports such as misspelled victim or street names, or erroneous street numbers. Basic inconsistencies in data, such as the use of a street suffix or direction (such as north or south), may reduce data quality. Accuracy of data is also affected by victim reporting-victims may delay reporting, may fail to recall key elements such as all types of property taken, may fail to report suspect information, or may provide false information to police such as claiming that properties were secured.
• Data precision. Even in offense reports, police data often fail to record all the elements of a unique address. Street number, apartment building or building name, floor, and apartment or suite number are often necessary for accurately detecting revictimization at addresses. For large properties, such as parks, parking lots, or sporting venues; or in locations without an address, such as pedestrian or bicycle paths, along creeks or in canyons, or within alleys; very precise address information will be unavailable. Some properties, such as schools or businesses, may have multiple addresses, including different streets. Police data may fail to distinguish between types of residential properties, such as apartments or single-family dwellings, or a type of commercial enterprise.
• Imprecise victim names. Detecting revictimization of individuals involves matching recorded victim names. But victims may alter their name, move after being victimized, be victimized in different locations, and households may contain individuals with different names who report separate incidents. When a commercial location is victimized, recording victim names should be standardized to consistently distinguish a corporate name from a store name or acronym.
• Hierarchical crime classification and recording inconsistencies. Hierarchical classification of offenses may mask revictimization for some offense types. For example, the occurrence of a burglary at the address where domestic violence has occurred may result in undercounting burglary or domestic violence revictimization. Some offenses are not consistently classified-an attempted break-in of a motor vehicle may be reported as vandalism.
• Victim groups. "Virtual" repeat victims may comprise groups, for example, specific restaurant chains, a specific chain of convenience stores or gasoline stations, college students, or surface parking lots.
• Boundary issues. Repeat victimization may occur across territorial boundaries, particularly for victim groups. For example, a specific chain of fast food restaurants may be robbed but offenses occur in different jurisdictions or police districts, and therefore are not easily detected. In areas where offender mobility is high, RV may be masked.

Victim Surveys

Victim surveys improve upon police data because the data address a common problem in police data-underreporting. Since victimization for any particular offense is usually quite low, random surveys are generally not cost effective. Police, however, can use two basic methods to survey for repeat victimization:

Victim follow-up surveys. Offense reports (or call histories) are used to construct a sampling frame; victims are then surveyed some time after their initial victimization to determine if they have been revictimized, and whether or not they called the police. The time period may be 30 days, or three, six, or 12 months-depending upon the time course for the offense type being examined.

Modified offense reports. A simple way for police to incorporate revictimization surveys into routine police work is to modify offense reports and/or the initial investigation of responding officers to include a few basic questions about victimization such as:

• Have you previously been victimized?
• Were you a victim of the same or a different type of offense?
• Were you victimized at this address or another address?
• How many times were you victimized?
• When did the victimization occur?
• Did you report the victimization to the police?

Questions about victimization experience should be used to identify high-risk victims and may also shed light on the development of the most effective responses. Questions for victims should relate to the specific problem being examined.

In some cases, police may want to carry out larger surveys. An entire population can be surveyed-such as all convenience store managers, budget motel managers, women on small college campuses, public school principals, or homebuilders. Some surveys may be observational, such as crime prevention surveys that assess environmental features such as lighting or building layouts.

In general, methodological issues must be considered in victim surveys:

• Continuous data. Unlike police data, which are collected continuously over time, data obtained from victim surveys are usually discrete, and data collection occurs at a single point in time. Ideally, surveys of victims should follow victims over time to detect if and when further victimizations occur.
• Attrition. Victim surveys will be subject to attrition as some victims move or refuse to be surveyed. Attrition may be more likely for individuals who have been revictimized, thus distorting the amount of revictimization detected. Conversely, some studies suggest that victims who are pleased with the police response to an offense are more likely to call the police again.
• Accuracy. Victims may be unable to recall incidents accurately- particularly dates of incidents-or may be reluctant to discuss victimization, particularly incidents such as domestic violence or sexual assault. Surveys relating to such problems must be sensitive to the presence of an offender or others who may inhibit honesty. Victims may not consider some types of victimization crime or may not consider themselves victims.

For guidance on developing surveys, refer to A Police Guide for Surveying Citizens and their Environments, or Conducting Community Surveys. Both are listed in the recommended readings. There are also specialized survey instruments, such as those used for conducting follow-up surveys with women victimized by domestic violence. For example, www.vaw.umn.edu/ contains links to validated instruments for such surveys that measure the amount of conflict that a victim has experienced. While many of these survey instruments are copyrighted and too in-depth for police use, they provide ideas about reliable questions can be incorporated into any followup survey.

Other Data Sources and Methods

While there are other existing sources of data about revictimization, data may need to be collected to document repeat victimization.

Important data may be collected through environmental observations. Properties that are vulnerable to graffiti such as vandal-prone walls in urban areas, pedestrian tunnels, or transportation corridors should be monitored to improve the amount and accuracy of information about offenses. Such observations may be daily or weekly, or reflect periods of vulnerability, such as following school holidays. (For more detailed guidance on environmental surveys, see A Police Guide to Surveying Citizens and Their Environment. This publication is listed in Recommended Readings at the end of this monograph.)

Interviews with apprehended or experienced offenders may improve understanding of repeat victimization. Although these interviews won't provide an empirical measure about the amount of repeat victimization, insights from offenders may provide understanding of target selection and an opportunity to prevent recurring victimization. (For more detailed guidance on collecting information from offenders, see the companion guide to this series, Using Offender Interviews to Inform Police Problem Solving.)

Other data sources that have been used to detect repeat victimization include medical records, such as hospital admissions or treatment in emergency rooms, admissions to domestic violence shelters, and inventory systems.