The United States has three main sources of information about patterns and trends in crime:
The FBI's Uniform Crime Reports (UCR) provides data on the total number of crimes that citizens report to the police, compiled from local law enforcement reports to the FBI. The UCR data have been collected going back to 1930, although the data quality has improved greatly over time. The UCR lets us measure crime counts and rates at the national, state, and local level, although the quality of these data declines for smaller geographic units. (A rate is simply a count divided by the number of people living in the county, usually expressed as a rate per 1,000, 10,000 or 100,000 people, depending on the frequency of the criminal event being examined). The UCR also records the number of arrests made by police departments for different offenses, including some limited information about arrestees such as race and juvenile versus adult status. For homicides, the FBI provides more detailed information about victims, offenders and crime events, which are compiled through what is known as the Supplemental Homicide Reports (SHR). Public-use versions of UCR and SHR data and documentation can be obtained from the University of Michigan's Inter-University Consortium on Political and Social Research. There is also a user-friendly web tool that enables users to calculate some simple cross-tabulations with the SHR.
The National Crime Victimization Survey (NCVS) was launched in 1973 by the U.S. Department of Justice's Bureau of Justice Statistics (BJS) to address the problem with the UCR that not all crimes are reported to the police. The NCVS conducts surveys with a nationally representative sample of 73,000 households every six months, achieving a remarkably high response rate by any standard. BJS publishes a number of extremely useful statistical reports that summarize key findings from the NCVS survey and are publicly available.
The Vital Statistics (VS) system compiles death certificates throughout the U.S. each year and allows researchers and policy analysts to generate quite accurate homicide counts and rates down to the county level. The VS only captures homicide, but homicide is widely thought by criminologists to be the most accurately recorded of all crimes. (It is worth noting that the SHR and VS still do disagree slightly on homicide counts). Homicide trends over time also tend to be highly correlated with trends in other key violent crimes, such as aggravated assault or robbery. The death certificates compiled by the VS include some basic characteristics of the decedent including age, race, marital status, and educational attainment. The CDC provides several excellent on-line tools (WISQARS and WONDER) that enable users to easily tabulate homicide and other injury statistics by year and different population sub-groups. In selected years (1966-8, 1986, and 1993) the National Center for Health Statistics has also conducted the National Mortality Followback Survey (NMFS), which comes from surveying next-of-kin to obtain more detail about decedents and the circumstances of their deaths. The VS and NMFS data are available from ICPSR.
In addition the federal government sponsors a number of ongoing national population surveys that ask about people's involvement in a variety of behaviors that are risk factors for crime offending or victimization. These surveys include Monitoring the Future, the Behavioral Risk Factor Surveillance System, and the Youth Risk Behavior Surveillance System.