Facts and Figures in the Bureau of Justice Statistics National Crime Victimization Survey (NCVS)
Victimologists and criminologists have reservations about the accuracy of the official records kept by police departments that form the basis of the FBI’s UCR as well as its NIBRS. Tallies maintained by local law enforcement agencies surely are incom- plete due to victim nonreporting. Also, on occa- sion, these closely watched statistics may be distorted by police officials as a result of political pressures to either downplay or inflate the total number of incidents in their jurisdiction in order to manipulate public opinion about the seriousness of the local crime problem or the effectiveness of their crime-fighting strategies (for example, see Eterno and Silverman, 2012).
Dissatisfaction with official record-keeping practices has led criminologists to collect their own data. The first method used was the self- report survey. Small samples of people were promised anonymity and confidentiality if they would “confess” on questionnaires about the crimes they had committed. This line of inquiry consis- tently revealed greater volumes of illegal acts than were indicated by official statistics in government reports. Self-report surveys confirmed the hypothe- sis that large numbers of people broke the law (especially during their teens and twenties), but most were never investigated, arrested, or con- victed, especially if they were members of middle- or upper-class families. But self-reports about offending did not shed any light on those who were on the receiving end of these illegal acts.
After establishing the usefulness of self-report surveys about offenses, the next logical step for researchers was to query people from all walks of life about any street crimes that may have been committed against them rather than by them. These self-report studies originally were called “victim surveys.” But that label was somewhat misleading because most respondents answered that they were not victims—they had not been harmed by street crimes during the time period in question.
The first national survey about victimization (based on a random sample of 10,000 households) was carried out in 1966 for the President’s Com- mission on Law Enforcement and the Administra- tion of Justice. It immediately confirmed one suspicion: A sizable percentage of individuals in the sample who told interviewers that they had suf- fered losses and/or sustained injuries acknowledged that they had not reported the incident to the police. This proof of the existence of what was termed the “dark figure” (meaning a murky, mys- terious, imprecise number) of unreported crimes further undercut confidence in the accuracy of the FBI’s UCR statistics for all offenses except murder. Confirming the existence of unreported crime also underscored the importance of continuing this alternative way of measuring victimization rates and trends by directly asking members of the gen- eral public about their recent experiences.
V I C T IM I Z AT ION IN T H E UN I T ED S T AT ES : AN OVE RV I E W 73
Copyright 2016 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. WCN 02-200-203
In 1972, the federal government initiated a yearly survey of businesses as well as residents in 26 large cities, but the project was discontinued in 1976. In 1973, the Census Bureau began interview- ing members of a huge, randomly selected, nation- wide, stratified, multistage sample of households (clustered by geographic counties). Until 1992, the undertaking was known as the National Crime Survey (NCS). After some revisions, it was retitled the National Crime Victimization Survey (NCVS).
The Aspects of Victimization That It Measures NCVS respondents answer questions from a survey that runs more than 20 pages. They are interviewed every six months for three years. The questioning begins with a series of screening items, such as, “During the last six months did any- one break into your home?” If the respondent answers yes, follow-up questions are asked to col- lect details about the incident.
When completed, the survey provides a great deal of data about the number of violent and prop- erty crimes committed against the respondents, the extent of any physical injuries or financial losses they sustained, and the location and time of the incidents. It also keeps track of the age, sex, race/ ethnicity, marital status, income level, and place of residence of the people disclosing their misfortunes to survey interviewers. The survey records the vic- tims’ perceptions about the perpetrators (in terms of whether they seemed to have been drinking and if they appeared to be members of a gang), whether they used weapons, and what self-protective measures—if any—the respondents took before, during, and after the attack. Additional questions probe into any prior relationships between them, as well as the reasons why they did or did not report the crime to the police.
The survey is person-centered. It is geared toward uncovering the suffering of individuals 12 years of age or older and the losses experienced by entire households (but not of workplaces, such as burglaries of offices or robberies of banks). The questionnaire focuses on crimes of violence (forc- ible rape, robbery, and aggravated assault) just like the UCR, plus simple assault, but not murder.
It also inquires about two kinds of thefts from indi- viduals (personal larceny with contact like purse snatching and pickpocketing, and without any direct contact), and three types of stealing directed at the common property of households—burglary, larceny, and motor vehicle theft—again, just like the UCR (except that crimes against collectivities like organizations and commercial enterprises are not included in the NCVS but are counted in the UCR). Identity theft is now examined, but the list of possible offenses is far from exhaustive. For example, respondents are not quizzed about instances of kidnapping, swindling, blackmail, extortion, and property damage due to vandalism or arson.
The benefit of survey research is that it elimi- nates the futility of attempting the impossible: inter- viewing every person (of the nearly 265 million people 12 years old or older residing in the entire United States in 2013) to find out how he or she fared in the past year. The combined experiences of just about 160,000 individuals over the age of 11 living in roughly 90,000 households randomly selected to be in the national sample in 2013 can be projected to derive estimates of the total number of people throughout the country who were robbed, raped, or beaten, and of households that suffered burglaries, larcenies, or car thefts (Truman and Langton, 2014).
Shortcomings of the Data At the survey’s incep- tion, the idea of asking people about their recent misfortunes was hailed as a major breakthrough that would provide more accurate statistics than those found in the UCR. But, for a number of reasons, the technique has not turned out to be the fool- proof method for measuring the “actual” amount of interpersonal violence and theft that some victimol- ogists had hoped it would be. (For more extensive critiques of the methodology, see Levine, 1976; Garofalo, 1981; Skogan, 1981b, 1986; Lehnen and Skogan, 1981; Reiss, 1981, 1986; Schneider, 1981; O’Brien, 1985; Mayhew and Hough, 1988; Fattah, 1991; and Lynch and Addington, 2007).
First, the findings of this survey, like any other, are reliable only to the extent that the national
74 CH APT ER 3
Copyright 2016 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. WCN 02-200-203
sample is truly representative of the population of the whole country. If the sample is biased (in terms of factors connected to victimization, such as age, gender, race, class, and geographical location), then the projections made about the experiences of the roughly 250 million people who actually were not questioned in 2013 will be either too high or too low. Because the NCVS is household based, it might fail to fully capture the experiences of tran- sients (such as homeless persons and inmates) or people who wish to keep a low profile (such as “illegal” immigrants or fugitives).
Second, the credibility of what people tell poll- sters is a constant subject of debate and a matter of continuing concern in this survey. Underreporting remains a problem because communication barriers can inhibit respondents from disclosing details about certain crimes committed against them (inci- dents that they also probably refused to bring to the attention of the police). Any systematic suppression of the facts, such as the unwillingness of wives to reveal that their husbands beat them, of teenage girls to divulge that they suffered date rapes, or of young men to admit that they were robbed while trying to buy illicit drugs or a prostitute’s sexual services, will throw off the survey’s projection of the true state of affairs. Furthermore, crimes com- mitted against children under 12 are not probed (so no information is forthcoming about physical and sexual abuse by caretakers, or molestations or kid- nappings by acquaintances or strangers). Memory decay (forgetting about incidents) also results in information losses, especially about minor offenses that did not involve serious injuries or expenses.
But overreporting can occur as well. Some respondents may exaggerate or deliberately lie for a host of personal motives. Experienced detectives filter out any accounts by complainants that do not sound believable. They deem the charges to be “unfounded” and decide that no further investiga- tion is warranted (see Chapter 6). But there is no such quality control over what people tell NCVS interviewers. The police don’t accept all reports of crimes at face value, but pollsters must. “Stolen” objects actually may have been misplaced, and an accidentally shattered window may be mistaken as
evidence of an attempted break-in. Also, no verifi- cation of assertions takes place. If a person in the sample discusses a crime that was supposedly reported to the local police, there is no attempt to double-check to see if the respondent’s recollec- tions coincide with the information in the depart- ment’s case files. Forward telescoping is the tendency to vividly remember traumatic events and therefore believe that a serious crime occurred more recently than it actually did (within the sur- vey’s reference period of “the previous six months”). It contributes to overreporting because respondents think a crime should be counted, when actually it was committed long before and ought to be excluded.
Because being targeted within the previous six months is a relatively rare event, tens of thousands of people must be polled to find a sufficient number of individuals with incidents worthy of discussion to meet the requirements for statistical soundness. For example, about 1,000 people must be interviewed in order to locate just a handful who were recently robbed. Estimates derived from small subsamples (such as robbery victims who are elderly and female) have large margins of error. The NCVS therefore requires a huge sample and becomes very expensive to carry out.
Even with a relatively large number of partici- pants, the findings of the survey can only describe the situation in the nation as a whole. The serious- ness of the crime problem in a particular city, county, or state cannot be accurately determined because the national sample is not large enough to break down into local subgroups of sufficient size for statistical analysis (with a few exceptions). Fur- thermore, the projected absolute number of inci- dents (offenses committed and victims harmed) and the relative rates (victims per 1,000 people per year) are really estimates at the midpoint of a range (what statisticians call a confidence inter- val). Therefore, NCVS rates always must be regarded as approximate, plus or minus a certain correction factor (margin of error) that depends mostly on the size of the entire sample (all respon- dents) and becomes statistically questionable for a very small specific subsample, such as low-income
V I C T IM I Z AT ION IN T H E UN I T ED S T AT ES : AN OVE RV I E W 75
Copyright 2016 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. WCN 02-200-203
young men, living in cities, who were robbed within the last six months.
The NCVS has improved over the years as better ways have been devised to draw representa- tive samples, to determine which incidents coincide with or don’t fit the FBI’s uniform crime definition, and to jog respondents’ memories (Taylor, 1989). In particular, more explicit questions were added about sexual assaults (involving unwanted or coerced sexual contact) that fell short of the legal definition of forcible rape and about instances of domestic violence (simple assaults) (Hoover, 1994; and Kindermann, Lynch, and Cantor, 1997).
Over its four decades, the survey’s questions have been refocused, clarified, and improved. But the accuracy of the NCVS has suffered because of waves of budget cuts. To save money, the sample size has been trimmed repeatedly. Over the dec- ades, expensive “paper and pencil interviews” (PAPI) carried out at people’s homes have been replaced by follow-up phone calls (computer- assisted telephone interviews [CATI]) and mail-in questionnaires. The response rate for individuals who were invited to participate has held steady at about 88 percent; in other words, 12 percent of the people chosen for the study declined to answer questions in 2010 as well as in 2013 (Truman, 2011; and Truman and Langton, 2014).