Crime Classification Systems: NCVS, NIBRS, and UCR

Lynn Pazzani & George Tita. 21st Century Criminology: A Reference Handbook. Editor: J Mitchell Miller. Sage Publications. 2009.

When asked about their top concerns for their country, Americans have consistently rated crime as a significant concern. Anyone who has been a victim of even a minor crime can attest to the impact it has on his or her life, both emotionally and in terms of time and inconvenience dealing with the aftermath. To reduce crime and the impact it has on individuals and society, criminologists must be able to measure and understand it. Without an organized means of recording information about crimes that have occurred, law enforcement agencies would not have the benefit of social science research in determining who is likely to be victimized so they can provide protection, who is likely to become an offender so they can provide alternatives, and where crime is most likely to occur so they can increase patrol.

In the United States, there are three main systems in place to measure and track crime. The Uniform Crime Reports and the National Incident-Based Reporting System use police report data to determine the scope of crime within particular areas and to provide detail about individual criminal incidents. The National Crime Victimization Survey measures crime on the basis of interviews in which people are asked about their experiences with crime. Each of these systems has been carefully constructed and modified over time to increase detection and improve the amount and depth of information that is collected.

This chapter describes each of these three crime databases. It provides information about how data are collected and discusses what types of crimes might be missed, as well as any issues with the quality or use of the data. It also compares the three systems in terms of their ability to detect crime and in the types of uses to which they may be best suited.

Uniform Crime Reports

In the 1920s, the crime problem in the United States, especially the fear of organized crime being perpetrated by infamous criminals, such as Al Capone and the various New York Mafia families, was seen by officials in government and the U.S. Department of Justice to be a pressing social issue. As a result, in 1929 the Department of Justice implemented a system whereby it could “measure” crime in the United States. This device, the Uniform Crime Reports (UCR), was designed to assess the type and magnitude of crime in U.S. communities. Data from the UCR are used to produce several yearly publications, such as Crime in the United States, Hate Crime Statistics, and Law Enforcement Officers Killed and Assaulted.

Uniform Crime Reports Data Collection

Data for the UCR are based on crimes reported to the police. Each month, every law enforcement agency is asked to submit data about certain crimes to the Federal Bureau of Investigation (FBI). These data are released yearly and allow for the calculation of rates of crime by place and type of crime. The UCR originally asked law enforcement agencies for the number of seven serious offenses that were committed in their district during the month. These crimes, known as Part I or index crimes, include murder, rape, robbery, aggravated assault, burglary, larceny–theft, and the theft of an auto vehicle. An eighth index crime, arson, was added under congressional directive in 1978. The UCR also records the number of 21 less serious crimes, known as Part II crimes: simple assault, curfew offenses and loitering, embezzlement, forgery and counterfeiting, disorderly conduct, driving under the influence, drug offenses, fraud, gambling, liquor offenses, offenses against the family, prostitution, public drunkenness, runaways, sex offenses, stolen property, vandalism, vagrancy, and weapons offenses.

Reporting crime data to the FBI is not mandatory. During the first years of the data collection, 43 states participated, although that does not mean that every police department within that state reported data. More recently, participation has been very high, with police departments accounting for about 93% of the population and 47 states reporting. Statistical techniques allow criminologists to estimate the number of crimes committed in nonreporting districts on the basis of the numbers reported in areas with similar population characteristics. These estimates are part of the calculation of crime rates for the United States as a whole, but nonreporting districts are not listed in more detailed reports.

One particularly useful aspect of the UCR is that because the definitions of particular crimes are uniform across police departments, accurate comparisons of rates of these crimes can be made across different counties and states. The laws determining how to classify a criminal action may not necessarily be the same in every state, so if it were left to individual law enforcement agencies to classify crimes it, might be more difficult to make comparisons across agencies. For example, in some states, rape would be defined only as sexual intercourse against a woman’s will, whereas others might have broader definitions, including other sexual acts or offenses against men. If the rape rates from two states were compared, a state with a very narrow definition of rape would have a much lower rate than a state with a broad definition, even if the actual rates of particular actions were about the same. To prevent this from happening, the UCR produces a handbook with very detailed instructions about how to classify a crime. Even if one law enforcement agency considers an act a rape, if it does not meet the UCR definition—carnal knowledge of a female forcibly and against her will—it would be recorded as the Part II crime of sexual assault. Each criminal incident is only reported in the UCR once, so even if there were several crimes committed by the same person at once only the most serious of those crimes would be listed in the UCR data.

In recent years the government has developed an interest in hate crimes. In addition to reporting crimes to the FBI in the usual manner, law enforcement agencies are asked to provide detailed information about any crimes that might be related to prejudice against the victim based on race, religion, sexual orientation, disability status, ethnicity, or national origin. In cases when a bias crime has been committed, the hierarchy rule does not apply, so all crimes that occurred during the incident will be reported. Crimes against persons and against property can be considered hate crimes. This separate report about hate crimes is used to produce the yearly publication Hate Crimes, which lists the crimes and circumstances surrounding hate crimes in all reporting cities.

In addition to the counts of crime, the UCR also records arrests. Crimes can be reported to the UCR regardless of arrest or prosecuting decisions, but crimes that have associated arrests are also tabulated to create what is called a clearance rate. A crime is considered cleared if there is an arrest, but not necessarily a criminal charge or conviction, associated with the crime. Of course, it is possible for the police to make an arrest when they have not accurately identified the offender, which would mean that crime is not solved, but it would be recorded as cleared. The UCR also records when a crime has been cleared by the arrest of only offenders under the age of 18. Another way that the UCR categorizes crimes that have been reported to the police is on the basis of whether the police determine that a crime has truly occurred. People can report a crime to the police that they have made up or about which they were mistaken, and those reports of crime should not really be counted toward the crime rate. If the police determine that a crime did not really occur, the report is considered unfounded. On the basis of all the ways to classify the crimes that were reported, the UCR reports numbers of crimes in several columns: crimes reported, crimes unfounded, actual offenses, offenses cleared, and clearances involving only juveniles. Clearance rates can vary greatly by state and county, which makes collecting this information very interesting.

The UCR also pays special attention to crimes committed against law enforcement officers and injuries sustained by law enforcement officers. Although a crime committed against a law enforcement officer could be recorded as a crime just like any other, police departments are asked to provide the number of crimes in which officers were victimized while on duty or as a result of their job. A police officer’s house being burglarized would not be recorded in this manner; however, an assault on an officer that occurred during a traffic stop would be recorded in this separate listing. Deaths and injuries that are not a result of a crime are also reported. If a firearm were accidentally discharged, either by the officer or another officer or person, and an injury or death resulted, this would be included in the report to the FBI. On the basis of these reports, the FBI publishes an annual report called Law Enforcement Officers Killed and Assaulted. This publication not only lists the number of these assaults or injuries but also describes weapons used during the incident; the type of injury; whether body armor was used by the officer; and other circumstances surrounding the assault, injury, or murder of law enforcement officers.

Crimes Excluded by the UCR

All crimes that are reported to the police that include an offense that meets the definition of a Part I or Part II crime should be reported to the FBI for inclusion in the UCR. This leads to the obvious exclusion of crimes that are not reported to the police. Although it may seem surprising that a crime victim would not report a crime to the police, there are certain situations in which this decision is quite common. For instance, many physical assaults are not reported, especially when both parties may be acting criminally or when the parties know each other. Burglaries in which items of little value are stolen may not be reported because the victim may not believe it is worth the time and trouble. Rape also has a particularly low reporting rate (estimates range between about 10% and 40%) because the victim may be embarrassed, fear that the police will not take her seriously, does not want to get the offender in trouble with the law, or does not identify the incident as a crime herself. It is possible that the proportion of crimes that are not reported to the police might be different depending on the location, which could make comparisons of crime rates between locations slightly inaccurate.

Crimes that are not reported to the police clearly will not end up being counted in the UCR; however, there are several ways that a crime that is reported may also not be included. First, the UCR does not collect data for all types of crimes. Kidnapping, for instance, occurs very infrequently, so it is not as useful to put resources into detailed tracking of kidnapping rates as it would be for a more commonly occurring crime. The same would apply for the crime of treason. There are also numerous other types of crimes that are so minor or frequently occurring—violating a dog leash law or exceeding the speed limit, for example—that detailed tracking would be a misuse of resources.

When two or more crimes are committed during the same criminal incident, the UCR records only the most serious crime. A rape–homicide would be recorded only as a homicide, and an assault that also resulted in an auto theft would be recorded only as an assault. For the purposes of recording and prosecuting crime within a particular district, all of the crimes that occurred during a single incident would be used.

In addition to just not counting certain crimes, the way some crimes are classified in the UCR compared with most legal systems is similar to excluding some crimes from the report. As discussed earlier, the UCR defines rape as the carnal knowledge of a female by force and against her will. Most states have now defined rape such that sexual attacks against males also are included. In this case, it is not that the UCR does not include a criminal incident as a crime at all, but it would label sexual attacks with male victims as assaults, or possibly even simple assaults, depending on the circumstances of the crime. The UCR considers rape and assault Part I crimes, meaning they are more severe than other types of crimes. A sexual attack against a male that could be legally considered a rape in the district where it occurred could be listed as a simple assault, which is a Part II (i.e., less serious) crime. When computing the rate of serious crime, the exclusion of some sexual attacks against males would lead to lower rates of crime that would be reported if such attacks were considered rapes or assaults as opposed to the less serious crime of simple assault.

National Incident-Based Reporting System

In the late 1970s, many law enforcement agencies felt there was a need for an examination of the UCR system in order to meet the challenges of a changing world and criminal justice system. The data collected, they felt, needed to be enhanced and expanded so it would be more useful in the areas of crime prevention and the study of crime. In March 1988, after much deliberation, the National Incident-Based Reporting System (NIBRS) was created. The NIBRS is an improvement over the original UCR data collection format both in the detail of information about each criminal incident and in that it increased the number of crimes on which information was gathered.

NIBRS Data Collection

NIBRS data are collected on the basis of reports to the police, just as UCR data are collected. For this data set, police are asked to record not only a count of crimes and some basic information but also detailed information about each criminal incident. For the NIBRS, the hierarchy rule that was discussed earlier does not apply. Whereas for the UCR only the most serious crime is recorded, the NIBRS allows for recording up to 10 different crimes per criminal incident. It also allows for information on up to 999 victims, 99 offenders, and 99 arrestees to be collected and associated with a particular criminal incident. Nearly all criminal incidents recorded in NIBRS have 3 or fewer offenses, victims, offenders, or arrestees, which means that sorting through the data for a particular incident is not hugely complicated. In cases where there are many offenses, victims, or offenders, only the first three records are made available for later data analysis. In this case, a hierarchy rule does apply, so that the three most serious criminal offenses are recorded and lesser offenses are excluded.

The NIBRS collects numerous details about each crime, victim, and offender, which, in addition to listing all the crimes and people involved in an incident, increases the depth of information that is available. For each offense, data are collected regarding whether the offense was completed or just attempted, the type of location involved, the time of day when the incident occurred, if force or weapons were used, if the crime may have been motivated by bias, the value of property stolen or destroyed and if property was recovered (if applicable), and whether the incident was cleared by an arrest. For each offender, the age, sex, race/ethnicity, relationship to the victim, and whether the offender was suspected of using drugs or alcohol during or shortly before the incident are recorded. Very similar data are recorded for any arrestees involved in the incident. If a victim is identified in a particular incident, the type of victim is recorded. For NIBRS data collection, a victim can be an individual, business, financial institution, government, law enforcement officer, religious organization, or society in general. If the victim is an individual or law enforcement officer, that person’s age, race/ethnicity, residency status, and any injury sustained as a result of the incident are recorded.

In addition to greater detail about each incident, the NIBRS adds several more crimes to the record database. These crimes are divided into Group A and Group B crimes, similar to how the UCR classifies incidents as Part I or Part II crimes. Group A includes all of the UCR’s Part I crimes as well as most of the Part II crimes, although in many cases, there are slightly different definitions. Group A crimes include arson, assault (aggravated, simple, and intimidation), bribery, burglary and breaking and entering, counterfeiting and forgery, property crimes (destruction, damage, and vandalism), drug offenses, embezzlement, extortion and blackmail, fraud (false pretenses, swindle, confidence game, credit card or ATM fraud, impersonation, welfare fraud, and wire fraud), gambling, homicide (murder, negligent and nonnegligent manslaughter, and justifiable homicide), kidnapping and abduction, larceny and theft, motor vehicle theft, pornography and obscene material offenses, prostitution (including assisting and promoting prostitution), robbery, forcible sex offenses (forcible rape and sodomy, sexual assault with an object, and forcible fondling), nonforcible sex offenses (incest and statutory rape), stolen property, and weapons laws violations. Group B offenses, which are considered less serious and more common, include writing bad checks; curfew, loitering, and vagrancy violations; disorderly conduct; driving under the influence; drunkenness; nonviolent family offenses; liquor law violations; peeping tom; runaway (not a criminal offense but still a police matter); trespassing; and all other offenses. For Group B offenses, only arrest data, not reports, are recorded. For each crime, the NIBRS identifies whether it is an offense against persons, property, or society.

Crimes Excluded by the NIBRS

The NIBRS is an improvement over the UCR because it includes nearly all known offenses. It has the same limitation as the UCR in that crimes that are not reported to the police cannot be recorded. Just like the UCR, some types of crimes are less likely to be reported to the police, so whereas all homicides, for example, will likely be recorded, only a small percentage of rapes will be. The NIBRS also includes many more crimes than the UCR, even within a particular category. The UCR includes forcible rape against a female only in the Part I sex offense category of “rape,” whereas the NIBRS includes rape and sodomy, as well as sexual assault with a foreign object and forced fondling in the Group A sex offense category “forcible sex offenses.”

The Group B category, all other offenses, may make it seem that every crime would be included in the NIBRS. It is important to remember that Group B offenses are recorded only if there is an arrest related to the incident. Thus, an individual could call the police to report a trespasser or a peeping tom, but the incident will be included in the NIBRS only if there is an arrest. A crime may have truly occurred, but without an associated arrest there will be no record of that crime in the NIBRS. Many minor crimes such as these occur frequently, and detection and arrest rates are fairly low, so the majority of these offenses probably would not be recorded. Some minor crimes may also be left out of the database in the few cases where more than three crimes were committed in one incident. Because the hierarchy rule applies, if a criminal incident included murder, rape, robbery, trespassing, and credit card fraud, the trespassing and fraud incidents would be excluded from the data made available for analysis.

Methodological Issues

Collecting data for the NIBRS is a very expensive and resource-intensive process. Because of the resources needed to collect the data, individual cities or law enforcement agencies tend not to be able to send their data to the FBI. Instead, data are processed through each state, which then submits the data to the FBI. Participation in the NIBRS is voluntary and is therefore still relatively low. Whereas agencies reporting to the UCR cover about 95% of the population, agencies reporting to the NIBRS cover only about 20% of the population. Despite the low degree of participation, rates of crime calculated by the UCR and by the NIBRS tend to be very similar. If only agencies with low crime rates had the time and resources to submit data to the NIBRS and high-crime agencies put their resources toward something else, the rates calculated by the NIBRS would be much lower than those calculated by the UCR. This indicates that although there is a low participation rate, the agencies that do participate are fairly good approximation of agencies throughout the United States. Although the rates are similar, the UCR tends to be used when calculating rates of crime rather than NIBRS, because of the greater participation and coverage. The use of a small sample of the population to represent the whole is discussed in greater detail in the “National Crime Victimization Survey” section.

Although the percentage of law enforcement agencies that report to the NIBRS is very low, there is still a lot that can be learned about crime using this system. The depth of the information allows social scientists to learn when and where crime is most likely to occur, who is most likely to be victimized, under what circumstances this victimization occurs, and how likely intoxicating substances are to be involved. It also allows for an examination of some of the outcomes of crime. We are able to determine the circumstances under which a victim is most likely to become injured; what types of crimes, victims, and offenders are more likely to lead to an arrest; and when property is more likely to be recovered. Because so many different types of information are collected, social scientists can do more than just count crimes; they can make predictions about them based on social information. Although a particular state or city may not have reported data to the NIBRS, social scientists can still use the information about crime produced by the NIBRS to help reduce crime in their own jurisdiction. Especially because the crime rates produced by the UCR and the NIBRS are so similar, there is no reason to think that in one state certain types of people are more likely to be victimized whereas in another state there is a completely different type of victim. That information alone can be very useful to law enforcement agencies.

National Crime Victimization Survey

In 1973, another system to attempt to measure crime was put into place by the Department of Justice; this was dubbed the National Crime Survey (NCS). The main difference between the UCR and the NCS was that the UCR measured crime using police report data, whereas the NCS measured crime by surveying households about victimization regardless of whether it had been reported to the police. Although the NCS was a great step toward measuring unreported crime, it underwent some significant revision to increase disclosure of previously undetected crime. Some changes to the survey were introduced in 1986, but the bulk of the changes were completed in the 1992–1994 release of the data. At that time, the survey was renamed the National Crime Victimization Survey (NCVS) to reflect the intended purpose of the survey and to signify the important differences between the old and new versions. The wording of questions was changed to increase reporting of crimes such as simple assault and rape or sexual assault. The rates of those crimes calculated from this new survey are higher than the previous year’s NCS rates, but that is an artifact of increased disclosure of crime rather than an actual increase in victimization.

NCVS Data Collection

Data for the NCVS are collected twice yearly by census officers who visit a small group of U.S. households. During these visits, the census officers interview one or more adults in the household and ask about crimes that have been committed against any member of the household over the age of 12 during the past 6 months. Respondents are asked about rape and sexual attack, robbery, aggravated and simple assaults, and purse-snatching and pocket-picking. This survey also covers property crimes, including burglary, theft, motor vehicle theft, and vandalism. The data gathered from the sample are used to estimate rates of crime overall and in specific states, cities, and types of neighborhoods. These estimates can also be made for each type of crime.

The NCVS collects information about crimes of which household members have been victims, regardless of whether they were reported to the police. When respondents indicate that someone in the household has been the victim of a particular crime, they are asked if the crime was reported to the police. They are also asked for details about the crime, such as whether the victim did anything to protect himself or herself or resist; if he or she was injured and, if so, how severely; what time of day the crime occurred; the type of location (home, outside, etc.) where the crime occurred; and how many people were involved. Details about the offender are also collected if the victim had enough interaction with the perpetrator to answer such questions. In many property crimes, such as vandalism, this would not be possible, because it is unlikely the victim would have seen the crime occur and may not know who is responsible. Details about the offender include race/ethnicity, approximate age, whether the offender may have been intoxicated, and possible motives for committing the crime. All of this information allows social scientists to determine the types of circumstances under which crimes are most likely to occur and aids in the development of effective prevention strategies.

This NCVS also collects other information about the household and the crime that may be useful to social scientists. Respondents are asked to provide information about household income, race/ethnicity and age of household members, level of education, type of household (apartment, house, etc.), household composition, and vehicle ownership. This type of information is known as demographic information. These data allow social scientists to determine what types of people and households might be particularly vulnerable to victimization.

The use of a crime survey that does not rely on police reports is very informative, because not all crimes are reported. Victims’ decisions to report a crime to the police may depend on many factors, such as whether they were doing something illegal themselves or whether the perpetrator was someone they knew and whom they may not want to report to the police. Teenagers in particular may not want to report crimes committed against them for fear of their parents becoming upset. Some types of crimes are also less likely to be reported to the police, such as rape and sexual assault and domestic violence, but these crimes may still be disclosed during administration of the NCVS. Although a victim may not seek out the police to report a crime, if someone comes to him or her and asks about victimization, promising to keep any disclosure confidential and stating that no legal action will be taken as a result of the disclosure, it may not seem as difficult or upsetting to discuss the victimization. Over time, as more has been learned about crime, the NCVS has been redesigned in an attempt to increase the disclosure of crimes that are particularly unlikely to be reported to the police. Specific questions have been added to find out whether the respondent has been victimized by someone he or she knows and to find out if it is possible that certain crimes were motivated by prejudice of some kind (i.e., hate crimes).

Crimes Excluded from the NCVS

Despite constant efforts to improve the survey and increase the disclosure of crime, there are some types of crimes that the NCVS either cannot collect information about or that have a lower chance of being disclosed during the survey. The NCVS is unable to collect data about homicide, for example, because the survey is designed to ask individuals about crime they have experienced themselves. Obviously, a homicide victim would not be able to disclose the crime to the census officers conducting the survey. Respondents would be able to tell the census officers if anyone in their household had been the victim of homicide, but it is important to remember that not everyone has other members of his or her household. All homicides committed against people living alone, or in which all household members were victims, would go undisclosed, while homicides against people with remaining household members would be disclosed. This would lead to a very inaccurate count of homicides. Although NCVS homicide data would be very inaccurate, homicide is one crime for which UCR data can be considered very accurate, so it is not a great loss that it is excluded from the NCVS.

The NCVS also does not ask about crimes committed against children under age 12. The census officers conducting the survey are able to interview children as young as 12 with the permission of a parent or guardian, but children younger than that would probably not be able to provide accurate information about victimization, both because they may not understand it and because they would have a more difficult time disclosing crimes within the 6-month time period being discussed. They might report serious crimes that happened some time ago if the event was very traumatic to them, or they may forget crimes that occurred toward the beginning of the 6-month window if the crimes were not very noteworthy. It would be possible for parents to answer questions about the victimization of their children; however, assaults against children often are committed by parents. It is very unlikely that a parent would disclose this information to a census officer, both out of embarrassment and fear of prosecution. Just as with homicide, data about crimes committed against children would end up being inaccurate and not very useful to social scientists.

Although homicide and crimes against children are intentionally excluded from the NCVS, other crimes are undisclosed for a variety of reasons. Just as not every crime is reported to the police, some people may simply not feel like disclosing their victimization to the census officers conducting the NCVS. They may believe that their victimization was minor and therefore not worth disclosing or that it is embarrassing and therefore difficult to disclose. In addition, respondents can disclose only crimes of which they are aware. For instance, they may have been the victim of a burglary that did not disturb enough of their property for them to notice. Also, when asked about whether other members of the household had been victimized, they are able to disclose only what each household member had told them about. Teenagers especially may not tell their parents about any crimes of which they have been a victim that might indicate they were drinking, out past curfew, or engaging in sexual activity. In cases where household members are not family members and may not be very close with one another, minor or embarrassing victimization may not be discussed. This can cause a respondent to believe that he or she is truthfully reporting that a household member has not been victimized when in fact crimes have occurred about which the respondent simply does not know.

The NCVS asks respondents to report about crimes in which a member of their household was victimized. Of course, crimes can occur in which no individual is victimized. Burglaries of businesses, for example, would not have an individual victim and thus would not be reported in the NCVS. If, in the previous example, there were an employee present who was assaulted during the burglary, that employee would report that he or she was the victim of a crime, but only the assault and not the burglary would be recorded for the purposes of crime reported by the NCVS. Crimes against businesses with no individual victim may be reported to the police, but they would not end up in the crime rate calculated by the NCVS.

Methodological Issues

Rates of crime are generally reported as the number of crimes per 100,000 people over the course of 1 year. To create this rate accurately, the respondents must be able to correctly recall how long ago their victimization occurred and must be able to remember all victimization that occurred within the specified time period. To collect accurate information about the types of crimes that occur, the respondent must also be able to remember details such as the time of day, the number of offenders involved, any injury, and other specific information covered in the NCVS questionnaire. The NCVS asks about crimes that occurred over the last 6 months. This time period was chosen as one that would be short enough to facilitate accurate recall of the number of crimes that occurred and specific details about each but also long enough that resources for data collection would not be wasted. Data collection that occurs in person requires an enormous amount of resources. Census officers need to be paid; travel is expensive; data collection materials are used; and data must be entered, which requires computers and people to perform the data entry. Collecting data twice a year as opposed to three or four times a year saves money without, ideally, sacrificing the quality of the data.

The NCVS is used, among other reasons, to discern rates of crime in specific locations. Of course, not every household in the United States is surveyed. Instead, a smaller sample of households is used to represent the whole country. The technique of sampling to make estimates or predictions about larger groups can be used with great accuracy and is employed in many types of research. For example, predicting election results is done with a sample of voters, not all of the voters. Election polling, when done well, tends to very accurately predict the results of elections. This is a particularly good example because in the case of elections, one can see both the prediction and, shortly afterward, the actual result. If these are very close, as they generally are, we know that social scientists and statisticians are able to use a small sample to fairly accurately represent the views of the whole population.

Sampling leads to accurate results when one has what is called a representative sample. That means that a researcher has in his or her sample approximately the same proportion of people or households that fall into various categories as exists in the larger population. For instance, one might consider factors such as income, education level, number of people in the household, and type of dwelling. If the households in the NCVS sample look very similar to the population on those characteristics, then it is likely that the sample will lead to a good estimate of the true crime rates in the population. In general, the NCVS is carried out in about 50,000 households that would hold approximately 80,000 people. The method is a rolling one, in which the households are kept in the study for 3 years. During any given year, approximately 160,000 individuals over age 12 are surveyed. The NCVS’s use of a nationally representative sample means that rates of crime for certain types of people can be discerned. For instance, rates of crime can be computed for certain age groups or for a particular racial or ethnic group.

Although sampling is done very carefully, there are still a few situations that can lead to slight inaccuracies in the measurement of crime. For instance, if a surveyed individual is a victim of a crime while he or she is far away from home, the victimization will be recorded as though it had occurred in the same area as the respondent’s home. Although the survey does ask about the type of location where the crime occurred—for instance, a park—it does not ask whether it occurred in a park near home or a park in a different state. This would not lead to inaccuracies in the rate of crime for the entire United States, but it can slightly alter the rates for specific places. If this same situation were to occur and be recorded in the UCR, the police department where the crime occurred would be the one reporting the incident, not the department near the victim’s home. Another way the estimates of the crime rate can be slightly inaccurate is if two (or more) people who do not live together are victims of the same crime. If two people walking together are mugged, the police would record this as only one criminal incident. However, because there are now two households that could disclose this victimization in the NCVS, that crime is twice as likely to be included as a crime committed against two people who did live together, or against only one person. These types of situations are not the norm, but it is important to understand that no estimate of criminal victimization is going to be perfect. Slight inaccuracies tend not to make a large difference when data are used for large scale analyses.

Comparison of the UCR, the NIBRS, and the NCVS

The main difference between the UCR and NIBRS compared with the NCVS is the use of police data versus victim self-reports. The UCR and NCVS were designed to complement each other in this way, and the NIBRS was added to give a greater degree of detail to the UCR, similar to the type of detail the NCVS collects. It would be impossible to determine which data source is the best, simply because there are so many different uses for crime data. Depending on the purpose of the request for information, whether it is academic research, or news reporting, or something else, the data set that would be most ideal may differ. News reporting agencies with interest in crimes reported to the police on local level may be best off using the UCR because of the NIBRS’s lower participation rate. Individuals with an interest in specific detail about crime on a national level may be best served by the NIBRS because of the amount of information collected and not having a need to get this detail about a very specific location. Academics who study underreported crimes, such as rape or domestic violence, tend to use the NCVS because of the much higher disclosure of these crimes there as compared with police reports.

Perhaps one of the easiest ways to compare the three data sources is to walk through a few hypothetical examples of crimes committed and how they would be recorded in each of the data sources. For an easy example, assume that a woman who lives alone is robbed at gunpoint and reports this incident to the police. The UCR would report a robbery. If there was an arrest, it would report that the incident had been cleared. The NIBRS would also report the robbery but may also include other offenses, such as a weapons violation. The NIBRS would also collect detailed information about the crime, victim, offender, and whether an arrest was made. The NCVS would report information that is very similar to what is reported in NIBRS, although it would also ask the victim about whether the crime was reported to the police, because the NCVS does not require an incident be reported to the police in order for it to be counted as a crime.

A very simple modification of this scenario shows how easily reporting to the various data sources can change. Now assume this same woman is the victim of the same crimes as just described, but now she is living with her elderly mother instead of living alone. She did not want to worry her mother, so she did not tell her about the crime, although she did report it to the police. Both the UCR and NIBRS would have the crime recorded in exactly the same way as described earlier. The NCVS may actually not report the crime at all. Assume that the robbery victim is not home when the census officers come to administer the NCVS questionnaire, so they interview the mother instead. The victim’s mother is unaware of the incident and, although she intends to be truthful with the census officer, she is unable to report an incident she did not know occurred. Thus, the survey that is often thought to be better at detecting crime because it does not rely on police reports actually misses a crime that was reported to the police.

Another slight modification to the original story can also change which data sources end up recording the crime. This time, the woman recognizes the boy who robbed her from the neighborhood. She knows he is having a hard time right now, and she did not lose anything valuable during the robbery. Therefore, she decides not to report the incident to the police. Under these circumstances, neither the UCR nor the NIBRS would have a record of the crime at all. When the census officer administering the NCVS comes to ask about victimization, the woman is assured that no legal action will be taken as a result of her participation, so she discloses the robbery, and it is recorded in the NCVS.

Each of these three examples is fairly straightforward. If one adds additional offenders or victims, the differences between what is reported in the UCR and in the NIBRS increase. If one changes the crime to one that is more difficult to define, such as a sexual assault of some kind, it is less clear how the incident would be reported in each of the three data sources. An incident of sodomy against a male, for instance, would be recorded as an assault, or perhaps even a simple assault, in the UCR; it would be recorded as a forcible sex offense in the NIBRS; and it would be counted as a rape in the NCVS. Males are even less likely to report sexual assaults than females, and they tend to have even more stigma associated with the event. Even if the victim did report the incident to the police, allowing it to be recorded in some fashion in the UCR and NIBRS, the victim may not want to talk about the incident again and may not disclose it to someone administering the NCVS, or he or she may not have disclosed the assault to household members who may be interviewed in his or her stead.

No data source recording crime is going to capture all the details about every crime that occurs. As participation in the NIBRS increases, efforts to increase the reporting of crimes to police may be the most helpful in increasing knowledge about crime. The redefinition of certain crimes in the UCR may also lead to the capture of more detail using a data source that already has a very high participation rate. The NCVS tends to show higher rates of crime than the other two sources, so we assume it yields more complete disclosure of crimes; however, as shown earlier, it is possible for even crimes that were reported to the police to be excluded from the NCVS. Improvements have been made to both the UCR and NCVS over the last 40 years, and the addition of the NIBRS also serves to increase social scientists’ understanding of crime. As time goes on, each of these data sources will continue to be improved, but there will never be a way to perfectly measure and describe crime.