Violent crime is undercounted or overcounted, depending on who you ask. Some statistics count gun violence twice, for example, as a criminal getting shot and a police officer doing the shooting.

On the other side, in a paper published in the Canadian Journal of Criminology and Criminal Justice, Wayne State University Ph.D. student Zavin Nazaretian and David M. Merolla, assistant professor of sociology, say that "capping" — which only allows survey respondents to represent a maximum of three incidents per crime type regardless of how many incidents they report — is undercounting violent crime by 87 percent and household crime by 36 percent.

Capping is standard methodological practice used in victim surveys. It has long been used to correct response bias by limiting the number of individual victimization incidents that one person can represent in official rates. It may need to be rethought, they write.

The scholars looked at the 2004 Canadian Victimization Survey and then examined seven of the eight personal or property crime categories, which sampled 23,700 households. Nazaretian and Merolla found substantial differences in how household and violent crimes are affected by capping, with results indicating that violent crime is much more sensitive to capping techniques than household victimization. That pattern is consistent with claims about repeat victimization, and they say this indicates that individuals reporting more than three victimizations likely are providing accurate responses, it is not response bias.

"By capping the numbers, some crime types increase more than others, possibly making the distribution between property crime and violent crime seem more comparable," Nazaretian said. "Perhaps that translates into spending more money on protecting vehicles that could instead be used to support things like homes for battered women."

"We argue that the increase is so much more for violent crimes, it's likely that at least some of these high numbers are real criminal events," Merolla said. "Rather than limiting it to three, we should be working toward developing some sort of methods to better understand when a report is likely to not be verifiable, or when it's likely to be a real crime."

In their paper, the scholars came up with new figures by including all incidences actually reported, leading to much different conclusions for some crimes. For example, Nazaretian and Merolla said the rates barely changed for vehicle theft, but increased for assault by 90 percent, from 51 incidents per 1,000 adults to 96. This would also mean crime throughout modern history is also far higher than official reports, if no one ever exaggerated.

"This shows, even when using its own numbers, that crime, particularly the most egregious types, is possibly much higher than the government reports," Nazaretian said.

The sociologists believe the higher numbers also provide insight into the number of unreported crimes when compared with arrest rates reported by police agencies, and that the cost of crime likely is more than is being reported. The say their report is not intended as a criticism of Statistics Canada, the agency that conducted the survey, but rather as evidence of the need to update a statistical technique for policy reasons and better resource allocation.