For three weeks, a war over crime statistics has been waged in newspapers across Canada.
On Feb. 9, the Macdonald-Laurier Institute published its study, “Why Canadian Crime Statistics Don’t Add Up: Not the Whole Truth.”. Author Scott Newark, formerly an Alberta Crown Prosecutor, argues Statistics Canada is misrepresenting national crime rates when it claims that crime has decreased over the past decade.
Some editorialists have cheered Newark’s studies, citing them as evidence of a liberal agenda at Statistics Canada.
“A murder is more serious than shoplifting, and if you look at total crime a murder counts once and a shoplifting counts once.”
“The agency’s statisticians know the numbers,” wrote Lorne Gunter in the National Post, referring to statistics about crime committed by parolees. “The only reason I can think for refusing to divulge them is that it doesn’t want to cast doubt on the current fashion for lenient sentences and easy parole.”
But for some in the criminology and legal communities, the study is a dishonest interpretation of the figures. Writing with attorney Edward Greenspan in the Globe and Mail, leading criminologist and statistician Anthony Doob claimed, “Newark’s study is filled with problems.”
The hot debate over Newark’s conclusions mirrors the larger political debate over the Conservative government’s “tough on crime” initiatives. At a time when the government is legislating to send more Canadians to prison, crime statistics make valuable evidence for advocates on both sides of the crime-control debate. Accordingly, both sides have sought to cast doubt on their opponents’ interpretations of available data.
Making sense of statistics
Caught in the middle of this statistical spat are journalists and their audiences, who may have trouble understanding and evaluating statistical claims.
Anthony Doob, the University of Toronto criminologist who attempted to debunk some claims made in the Macdonald-Laurier study, teaches statistical methods to criminology students. According to Doob, there are a few red flags readers should look for when evaluating any statistical claim in a news article.
First, says Doob, readers should always be suspicious of any claim in which “somebody picks two years and says that something has happened between those two years.”
For example, the Macdonald-Laurier study claims that violent crime increased 52 per cent from 1999 to 2009 (from 291,000 incidents to 443,000). Doob points out, however, that these are numbers, not rates based on population growth. Doob’s analysis found violent crime decreased during the decade in question, from 1,440 incidents per 100,000 population to 1,314.
This example also demonstrates another statistical red flag: comparing figures that cannot be compared. Between 1999 and 2008, the Statistics Canada’s definition of “violent crime” expanded to include more offences – driving up the number of reported offences.
Doob also points out that a little extra research can help clarify contentious issues. In the report, Newark criticizes Statistics Canada for excluding drug offences and criminal driving offences from the total crime rate (those numbers are excluded because many police departments specifically target drug and traffic crimes, which could artificially drive the total crime rate up). Doob says this data is readily available online.
“[Newark] makes it into a kind of conspiracy,” says Doob. “It takes about 12 seconds to get that [statistic] off the web.”
The traditional crime rate as compared to the Crime Severity Index, viewed as separate trends.
Some agreement
Gathering and analyzing crime statistics is a daunting task, and Doob agrees with Newark and the Macdonald-Laurier institute that Statistics Canada could improve its methodology and presentation.
Like Newark, Doob says he is “not a fan” of the Crime Severity Index. Introduced in 2008, the index gives statistical weight to individual crimes based on the seriousness of the crime (for example, a murder would carry more weight than an assault). Doob says the index is imperfect, although it makes intuitive sense.
“A murder is more serious than shoplifting, and if you look at total crime a murder counts once and a shoplifting counts once,” says Doob of traditional, non-weighted measures of crime. “That doesn’t represent our feelings about how crime should be described.”
In his report, Newark recommended replacing the Index with “an objectively-determined metric for severe crime, such as defined offences.”
But Doob is unsure how the Crime Severity Index could be improved.
“I’m not critical of Statistics Canada for trying it, in part because there is a problem when you simply count a shoplifting [incident] the same as a murder,” says Doob. “I don’t have a better way to do it.”
