UNIFORM CRIME REPORTS (police)

 

Of all available statistical series on crime, criminality, victimization, and criminal justice processing of offenders, unquestionably the most widely used (and often misused) is the Uniform Crime Reports (UCR) maintained by the Federal Bureau of Investigation (FBI). The agency’s well-known statistical report, Crime in the United States, published annually in print and online (http://www.fbi.gov/ucr/ucr.htm), presents summary and detailed tabulations of offenses known to the police and cleared by arrest, of persons arrested, and of counts of law enforcement personnel employed throughout the country.

Most attention is focused on Part I crimes—murder and nonnegligent manslaughter, forcible rape, robbery, aggravated assault, burglary, larceny/theft, motor vehicle theft, and arson. Crime counts, rates, and trends for these offenses are presented for the nation as a whole, for geographic divisions, population subgroups based on size, states, metropolitan areas, individual cities, counties, and college campuses. In addition, clearance rates—the percentages of offenses cleared by arrest or some other means—are provided in summary form for national and subnational aggregates.

Part II offenses include an array of violent, property, and status offenses, from simple assault to kidnapping, from drug violations to runaways. Because many of these criminal activities are not discrete countable offenses, the focus is only on the number of persons arrested. In addition, tabulations are generated by age, race, and sex of arrestees.

Thus, it is known offenses that are counted for Part I crimes but known offenders who are counted for Part II crimes.

History of the UCR Program

While federal initiatives to measure the nation’s population, health, and economy are nearly as long-standing as the republic itself, federal involvement in efforts to measure crime dates only as far back as the 1920s. Partially out of concern for refuting media-constructed “crime waves” based on anecdotal evidence (Maltz 1977), the International Association of Chiefs of Police (IACP) recommended forcefully that the U.S. Department of Justice establish a report card on crime. On June 11, 1930, Congress passed legislation directing the Justice Department to undertake this venture, and the attorney general then assigned this responsibility to the FBI.

Initially, the UCR was published monthly but was converted to a quarterly publication in 1932. In 1942, the report was shifted to a semiannual timetable, and later, in 1958, it became the annual publication (Crime in the United States) that exists today (along with semiannual preliminary counts).

From its inception, participation in the UCR program by local police agencies has been strictly voluntary (except for congressional legislation in 1990 mandating that colleges and universities provide crime data). With much to do during the Depression and beyond, the level of participation was limited (initially only 879 cities participated) and grew slowly over the years. Not until the early 1960s did the level of compliance reach the point where 90% of the U.S. population was covered (Maltz 1999).

The first thirty years of crime reporting were not only spotty in terms of population coverage, but there was no attempt to produce a national crime figure. Not until 1958, following the recommendations of the Consultant Committee on Uniform Crime Reporting, chaired by criminologist Peter Lejins (see FBI 1958), did the FBI’s Crime Index emerge.

The committee’s report advanced a long list of recommendations, many of which were immediately adopted (see FBI 1958). For example, manslaughter by negligence was dropped from the Part I group of offenses, and statutory rape was eliminated from forcible rape counts. Larcenies of property valued less than $50, because of their relatively trivial nature, were removed from larceny tabulations, although fifteen years later, in 1973, they were reincorpo-rated into the larceny/theft counts in recognition that a fixed threshold would have changing significance over time.

In 1979, by congressional mandate, arson was added to the list of Part I offenses. Even though arson was fundamentally different in terms of identification (by the fire marshal), the pressure from special interest groups led Congress to insist on this inclusion.

For the next half century, the Crime Index became a closely followed figure, on par with the cost of living index (CPI), the unemployment rate, and other national indicators. Unfortunately, its many limitations were not fully understood at first. But in 2004, after nearly fifty years of usage and recurrent criticism (see, for example, Wolfgang 1963), the Crime Index (specifically the reporting of the total volume and rate of all Part I crimes) was discontinued, even though violent and property aggregates were retained (and could easily be added to derive the previously used crime total). In suspending its use, the FBI hoped to discourage the notion that the crime total was in fact a real index (FBI 2004).

Measuring Crime Incidence

Unlike the efforts of the U.S. Bureau of the Census to conduct a complete enumeration of the population, the FBI’s crime count is hardly considered complete as a picture of the level of crime nationally. Of course, only certain crimes are designated as Part I types and, therefore, included in the crime count. That is, the “Crime Known to Police” tallies are counts of criminal homicides, forcible rapes, robberies, aggravated assaults, burglaries, larcenies, and motor vehicle thefts and arsons, both completed and attempted, ignoring the large array of other criminal activity.

Although participation rates have improved over the years, certain law enforcement agencies do not contribute data to the FBI, either directly or through an intermediate state crime statistics agency. For this reason, the UCR section of the FBI has utilized a variety of imputation strategies for estimating crime levels for agencies that fail to report data for a partial or entire year (Maltz 1999).

No attempt is made, however, to adjust for the most critical and fundamental drawback in completeness stemming from crimes that go unreported to the police. The so-called dark figure of crime (see Biederman and Reiss 1967) varies considerably across offenses and over time, complicating the interpretation of crime measures. While some offenses, such as homicide and motor vehicle theft, have high rates of reporting by victims or other parties, crimes such as forcible rape and larceny are notoriously low in terms of reporting rates. This can be especially problematic when examining trends over time. For example, an increase in the number of rapes in a jurisdiction may reflect either a true rise in sexual assaults or greater willingness of victims to report to the police.

Aside from minor modifications (such as renaming auto theft as motor vehicle theft in 1974) and the temporary dollar threshold for larceny described above, the definition of what constitutes a Part I crime has remained fairly uniform since the designation was introduced in 1958. The application of these definitions is, however, hardly uniform in practice. For example, the number of aggravated assaults jumped in certain jurisdictions when the police started to recognize spousal assault as something much more than a private matter. Also, what constitutes an aggravated assault (designated as Part I) as opposed to other assault (designated as Part II) is far from clear-cut.

Despite all of these complications and caveats, the UCR remains the most widely used measure of crime, far more comprehensive in coverage than alternatives. As a barometer of crime, albeit an incomplete one, it is unmatched.

Combining Crime

For the fifty years it was used, the UCR Crime Index tallied the number of Part I offenses, without any adjustment for type or severity. Even following the recent abolition of the published Crime Index, the tally of violent crimes blends together all reported homicides, forcible rapes, robberies, and aggravated assaults (including attempts), and that of property crimes merges all reported burglaries, larceny/ thefts, and motor vehicle thefts (including attempts). Unfortunately, the disparate levels of these individual offenses make any summary count across offense categories highly suspect.

For example, for 2003 the estimated 1.38 million violent crimes in the United States included 16,503 homicides, 93,433 rapes, 413,402 robberies, and 857,921 aggravated assaults. For the most part, therefore, the tally of violent crime is largely an aggravated assaults indicator, their constituting more than 60% of the violent crime total. By contrast, if the number of homicides had jumped 20%, the violent crime tally would have increased by only one-quarter of 1%. The homicide and rape counts are dwarfed by the robbery and aggravated assault figures. The same problem exists with property crimes, where 70% of the property total is larcenies/thefts.

Another quirk in the counting of crime occurs with complex criminal events involving multiple acts or victims. According to the “hierarchy rule” (FBI 1980), only the most serious element (among the seven Part I crimes in rank order from homicide to motor vehicle theft) of a criminal act is counted. Thus, a carjacking in which the driver is raped is counted only as rape for the purposes of crime statistics, while the carjacking is ignored. Although this may make some intuitive sense, oddities can easily develop. For example, if an offender robs a group of three people on a street corner, it counts as three robberies (one for each victim), but if the robber shoots and kills one of the three victims, it becomes a single homicide (the other two robbery victims would not count by virtue of the crime hierarchy).

In addition to the matter of scale discussed earlier, crimes vary considerably by severity. Thus, crime can increase even when the total level of harm is declining. For example, if a city has four more robberies but two fewer homicides the violent crime count still increases.

In an attempt to address this problem, Sellin and Wolfgang (1964) developed a crime severity scale to weight crime and delinquency by their seriousness. By this protocol, a homicide would count far more than a robbery, and a robbery/murder would count more than just a homicide. Additionally, offenses would carry points according to the dollar value loss. Despite its theoretical appeal, practical limitations in implementing such an approach prevented it from ever being adopted systematically. In addition, Blumstein (1974) found the FBI index and the Sellin-Wolfgang index to be nearly perfectly linearly related, suggesting that the latter would contribute little new information.

Crime Rates

In recent years, Americans have become eager consumers of graphical displays of statistical information, some of which illuminate important facts while others are misleading. This may help to explain why a much criticized feature of the annual crime report—the “crime clock”—keeps on ticking, despite its significant limitations.

Since 1939, the FBI has offered this graphical presentation of the temporal frequency of crime. In 2003, for example, there was a homicide every 31.8 minutes and a rape every 5.6 minutes. It can be rather difficult to interpret or utilize these facts, however. Of course, a murder does not actually occur every 31.8 minutes in any particular town or neighborhood, so there is no cause for someone to breathe easy and feel spared every half hour. More important, since the length of an hour never changes, the acceleration in crime over time (for example, from one homicide every hour in 1961 to one every 31.8 minutes in 2003) confounds real changes in risk with increasing population. The rate of speed with which crime occurs is not really a rate reflecting anything about likelihood or risk.

Clearly, a rate should adjust for population. The UCR presents crime rates as the number of offenses per 100,000 population:

tmp94-1_thumb

Using a 100,000 base, rather than per person or per capita, is simply for numerical convenience. For example, in 2004, there were an estimated 16,503 homicides nationally with a population of nearly 291 million. The per capita rate of homicide would be 0.000057, a rather awkward decimal. Scaling upward by the 100,000 multiplier yields a far more practical figure of 5.7 per 100,000 population.

While the problems outlined above pertaining to the crime volume are challenging, the pitfalls related to population counting are even more severe. Despite cautions printed in the UCR about comparing jurisdictions in terms of crime rates, the appeal in identifying the best and the worst places, and various rankings in between, may be just too great.

It is well known that urbanism can impact upon crime levels. One might expect higher crime rates in a densely populated urban center as compared to a small, rural community. But the effect extends even more in terms of the jurisdictional boundaries of an area. Some cities include within their jurisdictional boundaries large suburban fringe areas that have a moderating effect on the overall crime rate. By contrast, other places are essentially dense urban cores whose suburbs report crime statistics separately.

Jurisdictions differ not only in terms of their demographic makeup and socioeconomic conditions, which impact strongly on crime rates, but they differ in the extent to which the resident population reflects the number of people at risk. Simply put, crime incidents committed by or against nonresidents of a jurisdiction are counted in the numerator, but the nonresident offender or victim is not included in the denominator. As a consequence, cities that have substantial tourism tend to have inflated rates of crime. In a similar respect, part-year residents, including college students, may artificially distort crime rates. Finally, nonresident suburban commuters (and their property) are at risk during the workday, but they are not reflected in a city’s population measure.

In truth, a rate should adjust for the number of eligibles, as is done in other statistical spheres. The birth rate, for example, is typically calculated based on only the count of women in child-bearing ages. Unemployment rates are based on only those who are considered to be part of the labor force. Moreover, mortality rates are typically age adjusted to reflect changing patterns in the risk of death. Unfortunately, crime measures are crude rates, unadjusted for demographic changes or the size of the at-risk population.

There have been persuasive proposals to recast rape rates in terms of the size of the female population and even to calculate auto theft rates in terms of the available stock of registered vehicles (Mosher, Miethe, and Phillips 2002). As compelling as these arguments may be, they have not gathered enough momentum to be implemented as alternative measures of risk.

Finally, like crime rates, clearance rates are also difficult to interpret and may be misleading. These rates are calculated annually by dividing the number of crimes solved (either through arrest or some other means) by the total number of crimes reported, for each type of crime. However, an offense is not always cleared in the same year it was committed (for example, an offense committed and reported in December of one year may not be cleared until January of the following year). In addition, five crimes may sometimes be cleared by a single arrest, while in other cases the arrest of five accomplices may clear only a single crime. And, as Maltz (1999) pointed out, clearance rates can also be misleading because a change in the crime reporting can significantly affect the rates by inflating or deflating the denominator.

Trends with Caution

Every year when the latest crime statistics are released, journalists and reporters tend to focus on the trends: ”Is crime going up or down?” they typically ask. The answers are not always simple or easily interpreted. But the UCR does provide fodder for inquiring journalists, including one-year, five-year, and ten-year changes in crime counts.

One-year changes say little about crime swings. In other words, a one-year change does not a trend make. Unfortunately, civic leaders are often pressed for answers as to why a local crime figure has increased a few percentage points. An increase in crime from one year to the next (or any two points in time for that matter) may, however, reflect an unusually low count in the previous year, an unusually high count in the subsequent year, or part of a real trend. Determining which is the correct interpretation requires a longer, less myopic view of the trend.

To compound the problems related to the overemphasis on one-year changes, the greatest interest and attention is often given to homicide counts, which, because of their low numerical levels particularly in local areas, often fluctuate dramatically from year to year for no predictable reason. For a particular city, the number of murders can vary by some percentage just based on whether certain victims of assault die (thus constituting a homicide) or survive (thus remaining an aggravated assault).

Other Data Series

While ”crimes known to police” is the best known and longest standing initiative, several other data series are collected and disseminated by the FBI’s Uniform Crime Reporting Program. Like the Crime in the United States report, these other data are available online or in printed forms.

In addition to the summary counts of offenses, clearances, and arrests collected from police agencies, detail on each homicide is gathered from the Supplementary Homicide Reports (SHR) form. On an incident-by-incident basis, the month and year, the age, race, and sex of victims and offenders (if known), the weapon, victim-offender relationship, and circumstances are obtained on approximately 92% of all killings in the United States. The SHR data are used to produce detailed tables in the preliminary pages of the Crime in the United States publication. The Bureau of Justice Statistics also maintains a website (http://www.ojp.usdoj.gov/bjs/homicide/homtrnd.htm) with a wide range of geographic and demographic breakdowns of trends and patterns in homicide, overall as well as for various subtypes of homicide (Fox and Zawitz 2005).

Responding to growing concern about hate-inspired crime, the U.S. Congress passed the Hate Crimes Statistics Act on April 23, 1990, directing the attorney general to collect data ”about crimes that manifest evidence of prejudice based on race, religion, sexual orientation, or ethnicity.” For each year since 1992, the FBI has published a separate compendium of national and local data on offenses, assailants, and victims of personal and property crimes (murder and nonnegligent manslaughter, forcible rape, aggravated assault, simple assault, intimidation, robbery, burglary, larceny/theft, motor vehicle theft, arson, and destruction/damage/vandalism and residual categories) where, as a result of investigation, hate is believed to be at least partially the motivation for the offense. Nearly a decade into the program, the reported incidence of hate crime still appears to be surprisingly low (fewer than eight thousand incidents in 2003) perhaps because of the significant difficulties in determining motivation, particularly for unsolved crimes.

Complementing the law enforcement employment data published in Crime in the United States, the FBI also produces an annual report on fatal and nonfatal assaults against police officers. Besides national and local tallies of such victimizations, the annual report also tabulates data on the use of body armor and the circumstances and weapons involved in the attacks upon officers.

By far the most fundamental and significant change in the history of the FBI’s Uniform Crime Reporting Program was the launch of the National Incident-Based Reporting System (NIBRS) (Poggio et al. 1985). In addition to concerns over the lack of severity weighting and the effects of the hierarchy rule, important issues related to the lack of detail in the traditional UCR have been voiced over the years. Indeed, with the exception of homicide (through the SHR), the crime reports offer nothing in terms of the characteristics of victims and offenders and the incident itself. And because they are collected in summary form (specifically monthly tallies), it is not possible (again with the exception of homicide through the SHR) to examine subpat-terns within the data.

NIBRS, on the other hand, is quite rich in detail, soliciting extensive information about the offense, the victim and his or her injuries, the property damaged or stolen, the offender, and subsequent arrest for each incident. Unlike the UCR, NIBRS does not use the hierarchy rule, so all crimes committed in a single incident are reported. And because NIBRS (like the SHR) is an incident-based reporting scheme, it allows an analyst to tabulate or correlate variables in almost any way desired.

The ambitiousness of the NIBRS planners is laudable, yet it has also resulted in slow implementation of the program. Although computerized data entry speeds the process of reporting, the level of detail required can be quite onerous, particularly if an agency experiences a large number of incidents. For this reason, NIBRS has been implemented more for small or rural agencies. However, despite not being representative of the entire nation, the ability of NIBRS to place criminal incidents in a richly detailed context makes it an extremely useful tool for criminologists.

Conclusion

Computerization and the Internet have altered the way that most public and private agencies do business. Technology has and will continue to have an impact on the UCR. More and more crime data will be available for online look-up and manipulation. More significant, as NIBRS continues to take hold across the nation and at some point becomes the standard model of crime data collection, crime reporting will take a very different form. With various software applications for user interface (some of which are already used at the Bureau of Justice Statistics), crime reports can be customized. While this will be a plus in enhancing our ability to understand crime, it remains to be seen whether new issues in data interpretation will surface—surely a worthy topic for future generations of criminologists.

Next post:

Previous post: