The Bureau of Justice Statistics' National Crime Victimization Survey (NCVS) collects data on personal and household victimization by conducting an ongoing national survey of residential addresses. Begun in 1973, the survey compliments what is known about crime from the FBI's annual compilation of information reported to law enforcement agencies. The survey provides measures for the following types of crimes, including attempts: rape, sexual assault, robbery, assault (including both simple and aggravated), burglary, theft, motor vehicle theft, and vandalism. In addition to estimating the number of victimizations, the NCVS gathers details on each incident insofar as the victim can report them. These include the month, time, and location of the crime; the relationship between victim and offender; characteristics of the offender; self-protective actions taken by the victim during an incident and results of those actions; consequences of the victimization, including any injury or property loss; whether the crime was reported to police and reasons for reporting or not reporting; suspected gang involvement by the offender; and offender use of weapons, drugs, and alcohol.
In general, the NCVS provides good data on juvenile victims of common crimes. One major limitation of the survey for the provision of data on juvenile victims is the exclusion of household members under 12 years of age. Limits on the ability to produce subnational estimates and measurement error that affects subclasses of crime also pose problems for the use of these data.
The NCVS employs a stratified multistage cluster sample. The primary sampling units (PSUs) forming the first stage of the sample were counties, groups of counties, or large metropolitan areas. The PSUs were stratified by combining PSUs with similar characteristics into strata (for non-self-representing (NSR) PSUs) and allocating a single self-representing (SR) PSU (for populous PSUs selected with certainty) to the remaining strata, yielding an initial 1990 design that consisted of 93 SR PSUs and 152 NSR strata, with one PSU per stratum selected with probability proportionate to population size. In October 1996, the number of NSR PSUs was reduced to 110. NCVS continued to use both the 1980- and 1990-based samples through 1997. Beginning in 1998, only the 1990-based sample remained.
In the second stage of sampling, selected PSUs are divided into four nonoverlapping frames (unit, area, permit, and group quarters) from which NCVS independently selects its sample. Clusters of approximately four housing units or housing unit equivalents are selected from each frame. For the unit and group quarters frames, addresses come from the 1990 census. The permit frame ensures that units built after the 1990 census are included in the sample; addresses come from building permit offices. Sample blocks from the 1990 census files are used for the area frame (to account for new construction in jurisdictions not requiring building permits). Addresses are then listed and sampled in the field. A new sample based on addresses drawn from the 2000 census was phased in starting in 2005.
The actual number of households and persons interviewed in the NCVS sample varies slightly from year to year. In 2008, there were 42,093 households and 77,852 individuals age 12 and older interviewed. The household response rate for 2008 was 90.4% and the response rate for persons was 86.2%.
Publicly available NCVS data files from 1992 onward contain "pseudo-strata" identifiers. Software such as WesVar, SAS, or Stata can use the pseudo-strata identifiers to compute NCVS design effects and adjust variances to produce more accurate significance levels and confidence intervals.
DATA COLLECTION PROCEDURES
The sample of housing units is divided into six rotation groups with each group being interviewed every 6 months for 3 years. Within each rotation group six panels are designated, with a different panel interviewed each month during the 6-month period. The initial interview is used to bound the interviews (bounding establishes a time frame to avoid duplication of crimes on subsequent interviews) but is not used to compute the annual estimates; nor is it included in the regular public-use data. However, if a new household moves into a sample unit, the members of that household are interviewed and their unbounded interviews are included in the data.
All members of the household 12 years of age and older are interviewed. Other members of the household are asked to report for those with language or comprehension problems and for 12- and 13-year-olds if the members of the household prefer this approach.
Each respondent is asked a series of screen questions to determine if he or she was victimized during the six month period preceding the first day of the month of interview. Some screen questions concern crime against the household and are asked only of a single household respondent. These questions ask about break-ins or attempts and stolen household items, including motor vehicles and motor vehicle parts. Although not all household members are asked these screen questions, a household incident will still be recorded if it is mentioned by a household member. Another set of screen questions, asked of all respondents, concern specific types of personal crimes, other incidents reported to the police, or any events the respondents thought might be a crime. Whenever a respondent makes a positive reply to a screen question, he or she is asked "how many times" that type of incident occurred during the reference period. At the conclusion of the screen questions, an individual victimization report is completed for each incident mentioned in response to the screen questions.
The Census Bureau has made greater use of telephone interviewing in recent years. Beginning in March 1986, in the computer-assisted telephone interview (CATI) eligible part of the sample, all interviews are done by telephone whenever possible, except for the first interview, which is primarily conducted in person. The interview takes about 1/2 hour. Beginning March 2003, addresses in all segments became eligible for CATI interviewing . CATI cases are interviewed from CATI facilities while the other sample cases are interviewed by the standard NCVS field procedures.
Three basic forms are used to collect the required data for the NCVS: the Control Card, the Basic Screen Questionnaire, and the Crime Incident Report. The Control Card is the basic administrative record for each sample unit. It contains the address of each sample unit and the basic household data, such as the names of all persons living there and their age, race, sex, marital status, education, and the like. Family income, tenure of the unit, and pertinent information about noninterviews are also included on the Control Card. The Control Card serves as a record of visits, telephone calls, interviews, and noninterview reasons. The Control Card information is updated as needed. The Basic Screen Questionnaires are used to record characteristics of all eligible household members, as well as to screen for incidents of crime that were committed against the household and/or household members. Household and individual Screen Questionnaires are designed to elicit information on whether any crimes were committed against the household as a whole or specifically against individual household members. A Crime Incident Report is used to gather detailed information about crimes reported in either the Household or Individual Screen Questions. One Crime Incident Report is completed for each incident of crime reported in answer to screen questions.
The information content of the NCVS has been consistent following the fielding of the redesigned instrument in 1992 and 1993. Changes include additional cues to help respondents recall incidents and report victimizations that they may not define as crimes; and more direct questions on rape, sexual assault, and other sexual crimes (verbal threats of rape or sexual assault, and unwanted sexual contact without force but involving threats or other harm to the victim). These changes have elicited information on more than three times as many sexual crime victimizations than the previous version of the NCVS. The revised questionnaire also aims to measure victimizations by nonstrangers, including domestic violence, with multiple questions and cues on crimes committed by family members, intimates, and acquaintances. More recent additions to the questionnaire pertain to incidents of hate crime and computer crime.
Occasional supplements to the survey, such as the school supplement can substantially increase the volume of available information. The standard instrumentation in NCVS includes data on the demographic characteristics of persons and households, their labor force status, marital status, family composition, length of residence at this address, and educational attainment. There is some information on the jurisdiction in which respondents live, such as the population size and degree of urbanization, and on the neighborhood of residence, e.g. percent in poverty, percent unemployed, and so on. Much more information is collected on the attributes of crime incidents, such as the time of occurrence; location; whether there was forcible entry, threat, assault, or weapons present or used; type of weapon; nature of the assault; age, race, and sex of offenders (if known); degree of injury; amount of loss; type of medical attention; type of property taken; insurance recovery and amount of recovery; reporting to the police and insurance, and the like. NCVS also records the sequence of victim-offender interaction, whether the crime occurred in the jurisdiction of residence, whether the offender appeared to be under the influence of drugs, and the nature of service received from the police.
Interviewers receive extensive training prior to any field work. The quality of interviewing is maintained by periodic direct observation of all interviewers, office edits of completed work, and verification by reinterviewing a sample of completed sample units.
Reinterviews measure how well individual interviewers followed procedures. They also measure errors in coverage of the population resulting from missing units, incorrect address listing, and so forth. The reinterview procedure is used to evaluate the impact of errors or variations in responses.
Data collection in the NCVS is continuous. Reporting is done annually with preliminary reports issued 9 months after the end of the reference year and a final report about 18 months after the end of the year.
NCVS is based on a multistage cluster sample designed to be representative of the noninstitutionalized population of the U.S. Consequently, it provides estimates of victimization that are representative of the nation and for which estimates of sampling error can be made. The publicly released data file contains variables (a pseudostratum code and a sampling error computation unit code) that enable researchers to calculate variances corrected for design effects. The household weight variable allows computation of unbiased estimates.
NCVS data can be aggregated and disaggregated easily. Information can be arrayed at the incident, person, household, segment, and PSU levels. Data can be reported by demographic characteristics of the victim and offender, by jurisdiction type, and by countless other disaggregations.
NCVS data can be aggregated over time to obtain large numbers of even quite rare events. It is possible to construct a longitudinal file of NCVS data, but since the survey employs a rotating panel and not a longitudinal design, there are limits to the utility of the survey as a longitudinal data set.
NCVS METHODOLOGICAL CHANGES IN 2006
A number of methodological changes were implemented in the NCVS in 2006 that impacted the victimization rate estimates for that year. The changes and their impact upon the survey's estimates are described in the "Technical Notes" section of Criminal Victimization, 2006.
Analyses of the 2007 data by BJS and the Census Bureau found that the effects were reversed in 2007, suggesting that the 2006 findings represented a temporary anomaly in the data. NCVS estimates for 2008 are consistent with and comparable to estimates for 2007, 2005, and previous years.
NCVS data are provided in publications and in machine-readable form through the National Archive of Criminal Justice Data (NACJD) of the Inter-University Consortium on Political and Social Research (ICPSR). BJS issues several annual reports using NCVS data including Criminal Victimization, preliminary and final change reports, and the bulletin Households Touched by Crime. In addition to these annual publications, NCVS data have been analyzed for a variety of special reports dealing with specific types of crime (e.g., stranger-to-stranger crime, rape, household burglary, family violence), crimes against special populations (the elderly, teenage victims, Hispanics), and specific features of crimes (weapons use, costs of crime). NCVS data are also routinely made available in various compendia, including the Sourcebook of Criminal Justice Statistics and the annual Statistical Abstract of the United States.
Machine-readable NCVS data are made available through NACJD/ICPSR (see https://www.icpsr.umich.edu/icpsrweb/NACJD/NCVS/). These files are updated annually and include both complete files of NCVS data (both victims and nonvictims) and extract files containing only victims and crime incidents. These files are provided in hierarchical and rectangular formats. Other special-purpose data files developed from NCVS data for special reports are also available for public use. There are no abstract files pertaining exclusively to victimizations involving juveniles. BJS developed an online data analysis tool to promote further access to NCVS data.
A large array of bulletins, special reports, and technical reports are produced routinely by BJS using NCVS data, including:Criminal Victimization
and the accompanying collection of detailed statistical tables present annual estimates of rates and levels of personal and property victimization. The latest edition(s) are available at https://www.bjs.gov/index.cfm?ty=dcdetail&iid=245
.Criminal Victimization in the United States
, a collection of detailed statistical tables, available at https://www.bjs.gov/index.cfm?ty=pbse&sid=58
BJS also prepared several reports describing the redesign of NCVS. These reports are available at https://www.bjs.gov/index.cfm?ty=tp&tid=91
For a comparison between the NCVS and UCR data collection efforts, read The nation's two crime measures
. Available at https://www.bjs.gov/index.cfm?ty=pbdetail&iid=802
For a useful explication and critique of the current NCVS and UCR, Understanding Crime Incidence Statistics: Why the UCR Diverges from the NCS
. (A.D. Biderman and J.P. Lynch. [New York, Springer Verlag, 1991]).
Michael R. Rand
Chief, Victimization Statistics
Bureau of Justice Statistics
Marilyn Monahan (Survey procedures)
Chief, Crime Surveys Branch
Demographic Surveys Division
US Census Bureau