Crime numbers may mislead

Criminologists argue for considering underreporting and other sources of error

Overconfidence in crime statistics doesn’t pay. In a new study, a team of criminologists makes the case that reported crime rates should acknowledge uncertainty in the data. The research demonstrates that rankings of cities as safer or more dangerous — which can influence tourism and tax spending — can be highly misleading.

“If you look at crime rates from year to year and you see a change, there’s a fundamental ambiguity in whether that change is caused by a real change in crime, a change in reporting or some of both,” says criminologist Robert Brame of the University of North Carolina at Charlotte, a coauthor of the new study. “Our position is we should own that. There’s ambiguity here and we should learn to deal with it.”

To get a sense of that ambiguity, Brame and his colleagues calculated the wiggle room in burglary data for the 10 biggest cities in North Carolina. Based on state population estimates and the residential burglaries reported by police departments, the standard simple calculation suggests that in 2009 Wilmington had a higher residential burglary rate than Charlotte, for example. But when the researchers included known uncertainties in the numbers, Charlotte was too close to Wilmington to discern if one really had a lower or higher burglary rate than the other. Ambiguity in the data also meant that the researchers couldn’t tell if the rate of burglaries in Raleigh and Winston-Salem had dropped or risen from 2009 to 2010. Brame, his University of North Carolina colleague Michael Turner and Raymond Paternoster of the University of Maryland in College Park reported their analysis online April 30 at

A major source of uncertainty lies in how many crimes are actually reported. To investigate this, the researchers used two sources of data: hard numbers of reported crimes and estimates of how many crimes go unreported. The hard numbers came from the FBI’s Uniform Crime Reports, which compiles state data gathered from police departments. For residential burglaries, these numbers are known to be underestimates: If a burglary occurs along with a more serious crime, such as aggravated assault, rape or murder, the incident will be recorded as the higher-ranking offense.

For estimates of unreported burglaries, the researchers turned to the National Crime Victimization Survey of the Bureau of Justice Statistics, which compiles yearly data from interviews with roughly 40,000 households on their experience with crime and whether they reported crimes they encountered.

The NCVS survey data suggest that the national rate at which residential burglaries are reported to the police varies, but has been increasing over time. In 1999, for example, 49.3 percent of victims said the crime had been reported. In 2009, 57.3 percent of burglary victims said the event had been reported, and in 2010, 58.8 percent did.

A major factor that affects these numbers is a community’s perception of and relationship to the law, says sociologist David Kirk of the University of Texas at Austin. Some communities perceive the police as helpful, others see the police as out to get them. This means that more crimes might be reported in a community that trusts its police department, giving the impression that that community actually has more crime.

Using the data on burglaries that go unreported, Brame and his colleagues gave their estimates an upper and lower bound. They also accounted for uncertainty in city size, since population estimates by states often differ from federal numbers. The researchers argue that presenting crime data as an interval offers more meaningful information about whether a particular crime intervention strategy or youth education program is working, even if the answer is, ‘we don’t know.’

Criminologist Richard Rosenfeld of the University of Missouri in St. Louis, who has spent years studying crime trends, says we need more such studies. Not only are the numbers often fuzzier than they appear, but within a city, crime varies greatly across neighborhoods. The best predictors of whether someone is at risk for crime isn’t the city they live in, but factors such as age, gender and whether the person is involved in criminal activity.

More Stories from Science News on Humans