In the United States, crime rates are measured in one of two ways.

(1) The Uniform Crime Report, administered by Federal Bureau of Investigation, compiles data of crimes reported to local police. The UCR was established in the early 1930s. Because it is based on local information, the UCR permits statistical analysis for local areas. For example, recent reports that crime was down in Allegheny County in the late 1990s are the results of Uniform Crime Report data.

The accuracy of the UCR varies with the likelihood that people will report crimes to the police. The UCR is most accurate for murders, which are almost always reported or noticed, but may be less accurate for crimes that arenít reported as often, such as rape or assault. In addition, for crimes such as theft, the UCR tends to underrepresent the frequency of offenses (because people do not always report minor theft), but overrepresent the value of thefts (because people are more likely to report major thefts). Long-term analyses of UCR data are also questionable because computers have permitted better police record keeping in the last 20 years. Therefore, increases in UCR crime rates partly result from improved recording of crimes by police. Still, the UCR is the standard & most widely cited method of measuring crime in the US.

(2) The National Crime Victimization Survey, begun in 1973, is administered by US Census Bureau. This measure is a representative telephone sampling of approximately 40,000 households to determine how many people were victimized by one of seven crimes in past year. The crimes recorded are rape, robbery, assault, personal theft, household theft, burglary, and motor vehicle theft; the NCVS does not measure murder rates because the victims cannot be surveyed. The NCVS employs a national survey, so it cannot break down data by state or locality. Samuel Walker and most criminologists see the NCVS as more accurate than the UCR; people will acknowledge that they have been victims of a crime to an anonymous survey, even they did not report the crimes to police.

(3) Determining rates of crime (generally, per 1000 people or 100,000 people) requires that the number of offenses (the numerator) be divided by an accurate count of the population (the denominator). The results of the decennial United States Census are conventionally used as the sources of population data.

[Adapted from Steven R. Donziger, ed., The Real War on Crime: The Report of the National Criminal Justice Commission (NY: Harper, 1996).]