Analysis Liberals' claim of 'steady increase' in gun crime rests on a 'drastic' comparison to a low-crime year
Experts says Trudeau government's choice of baseline year makes crime stats look worse than they really are
Ontario Provincial Police officers prepare to bag a firearm after hosting a news conference in Vaughan, Ont., on Thursday, February 23, 2017. (Chris Young/Canadian Press)
Evan Dyer
Senior Reporter
The year 2013 has been enjoying a bit of a renaissance lately in government talking points and statistics.
The alarming crime stats presented by Public Safety Minister Ralph Goodale at the guns and gangs summit earlier this month in Ottawa all took 2013 as their point of comparison. Goodale pointed to a sharp increase in gang shootings since that year.
That memorable year returned again as the Liberal government on Tuesday advanced its proposed firearms legislation, Bill C-71, making the case that the country had seen a sudden increase in firearms offences since … 2013.
"Gun homicides are up by two-thirds" since 2013, Goodale warned — citing what sounds like a shocking explosion of violence.
"Averages and relationships and trends and graphs are not always what they seem," Darrell Huff warned in his 1954 classic How to Lie With Statistics. "There may be more in them than meets the eye, and there may be a good deal less."
The average citizen, hearing how gun crime has soared since 2013, might well conclude that something alarming has been going on.
In fact, it's "a good deal less."
2013: The "most drastic" baseline
What appears to make 2013 attractive as a point of comparison is that any year in the past half century can be made to look alarming by comparing it to 2013.
When presented this way, the graph appears to show a sudden and alarming rise in firearm-related crime in Canada. Statistician Pierre-Jérôme Bergeron is critical of this graph produced by Public Safety Canada. "A good graph should have one story to tell, on one scale," he said. "That is a graph made to confuse." (Statistics Canada/Public Safety Canada)
That's because 2013 saw Canada's lowest rate of criminal homicides in 50 years, and the lowest rate of fatal shootings ever recorded by Statistics Canada.
In 2013, Canadians killed each other at the lowest rate since 1966 — 30 per cent below the average of the previous three decades. Statistics Canada's homicide report for 2013 clearly identifies it as a record-breaking year.
"To be worth much, a report based on sampling must use a representative sample," wrote Huff.
But 2013 does not represent any kind of Canadian norm. Choosing it as a baseline could be seen as an example of what statisticians call "selection bias."
"They obviously picked the one year where it was lowest, so as to maximize the impact, the one year to make the change look most drastic, essentially," said Pierre-Jérôme Bergeron, who teaches statistics at the Department of Mathematics and Statistics at the University of Ottawa.
The McGill-trained statistician, who authored a paper called How to Engage in Pseudoscience With Real Data, said it doesn't look like an honest mistake, either.
"Here, I'm pretty sure they saw 2013 at the bottom, and said, 'We're going to pick that,'" he said. "Just like climate change deniers will say, 'It hasn't warmed since 1998,' but they pick 1998 because it was so hot, one of the highest, and is actually an outlier."
Bergeron noted that while statisticians often look back at recent data in five-year blocks, the Trudeau government has oddly chosen to measure only four years (2013, 2014, 2015 and 2016). Taking the data set back to 2012 would be the more normal practice — but it wouldn't produce the same impression of rocketing gun violence.
University of Ottawa criminologist Holly Johnson specializes in methodologies of measurements of crime and has coordinated surveys of violence and crime for Statistics Canada and the UN.
"I don't know the motivation or reasoning behind it," she said, "but certainly choosing the lowest rate in decades of data would suggest there's a reason for that, trying to make a point of some sort.
"Statistics can be misused, and to take a high point or a low point to try to make a political point is something I think we should all watch for.
"Statistics for rates of population that are based on very rare events, such as firearms homicides, tend to fluctuate, quite a great deal, actually."
The same statistics presented by Statistics Canada in a line graph are considerably less alarming. (Statistics Canada)
Johnson said an objective statistician, looking for a baseline for comparison, typically would look at how a particular year compares to the average of the 10 years preceding it, to gauge how one specific year stacks up against the norm.
"A few years does not a trend make."
Johnson adds: "Certainly I wouldn't call it a crime wave."
This chart shows the long term decline in firearms murders in Canada (light blue line). The unrepresentative year of 2013 is clearly visible as a sharp dip in the line. Since then, firearms murders have ticked back up to about where they were ten years ago. (Statistics Canada)
The crime wave that wasn't
We don't yet know how 2018 will turn out for homicides, but there's a good chance it will be worse than 2013 — because every year since 1966 has been worse than 2013.
But based on Canada's latest-known homicide rate (1.68 per 100,000 in 2016), it's also likely Canada's homicide rate in 2018 will be similar to or lower than it was 10 years before in 2008 (1.83 per 100,000) — or in 1998 (1.85 per 100,000), 1988 (2.15 per 100,000), 1978 (2.76 per 100,000) or, for that matter, 1968 (1.81 per 100,000).
Indeed, the rate today is not far from where it was in 1928 (1.55 per 100,000).
The Trudeau government points out that the sub-category of homicides with firearms has gone up, and this is true … if one takes 2013 as a starting point. Otherwise, not really.
CBC asked Goodale's office why he chose to use 2013 as a baseline.
"Crime rates generally in Canada have been on the decline for more than two decades," said spokesman Scott Bardsley in an email, "but offences involving firearms have become more prevalent, especially since 2013.
"Many communities across the country have been facing a steady increase in gun violence over the past five years. This trend is a break from overall declines in crime over recent decades.
"More broadly, our caucus has heard from Canadians across the country who are concerned about gun violence. Better is always possible."
'Steady increase'
But if, rather than picking the lowest year for comparison, one were to ask how 2016 compared with a decade before, one would find that the rate of firearms homicides remained boringly unchanged. It was 0.69 per 100,000 in 2005, and 0.59 in 2006, and 0.57 in 2007. It dipped during 2013 and was back at 0.62 per 100,000 in the supposedly alarming year of 2016.
The rate of homicides with handguns in Canada in 2016 was 0.36 per 100,000. Ten years ago, it was 0.38 per 100,000.
None of that constitutes a "steady increase."
One might also look back at the long-term trend, and note with satisfaction that the rate at which Canadians are being killed with firearms has seen a historic decline of more than 50 per cent from its high point of 1.26 per 100,000 in 1975.
The rate of killings and crimes committed with non-restricted long guns has declined by more than three-quarters since the 1980s. Robberies involving guns also have declined by more than 75 per cent during the same period.
That is what a statistician might reasonably call "a steady decrease."
So it's not accurate to say that "offences involving firearms have become more prevalent, especially since 2013." It would be more accurate to say offences involving firearms appear to have become more prevalentif we use the abnormal year of 2013 as a baseline.
And that's not hard to do because, as Mark Twain pointed out, "facts are stubborn things, but statistics are pliable."
Liberals' claim of 'steady increase' in gun crime rests on a 'drastic' comparison to a low-crime year - Politics - CBC News
From the CBC, no less!!!!!
Experts says Trudeau government's choice of baseline year makes crime stats look worse than they really are
Evan Dyer
Senior Reporter
The year 2013 has been enjoying a bit of a renaissance lately in government talking points and statistics.
The alarming crime stats presented by Public Safety Minister Ralph Goodale at the guns and gangs summit earlier this month in Ottawa all took 2013 as their point of comparison. Goodale pointed to a sharp increase in gang shootings since that year.
That memorable year returned again as the Liberal government on Tuesday advanced its proposed firearms legislation, Bill C-71, making the case that the country had seen a sudden increase in firearms offences since … 2013.
"Gun homicides are up by two-thirds" since 2013, Goodale warned — citing what sounds like a shocking explosion of violence.
"Averages and relationships and trends and graphs are not always what they seem," Darrell Huff warned in his 1954 classic How to Lie With Statistics. "There may be more in them than meets the eye, and there may be a good deal less."
The average citizen, hearing how gun crime has soared since 2013, might well conclude that something alarming has been going on.
In fact, it's "a good deal less."
2013: The "most drastic" baseline
What appears to make 2013 attractive as a point of comparison is that any year in the past half century can be made to look alarming by comparing it to 2013.
That's because 2013 saw Canada's lowest rate of criminal homicides in 50 years, and the lowest rate of fatal shootings ever recorded by Statistics Canada.
In 2013, Canadians killed each other at the lowest rate since 1966 — 30 per cent below the average of the previous three decades. Statistics Canada's homicide report for 2013 clearly identifies it as a record-breaking year.
"To be worth much, a report based on sampling must use a representative sample," wrote Huff.
But 2013 does not represent any kind of Canadian norm. Choosing it as a baseline could be seen as an example of what statisticians call "selection bias."
"They obviously picked the one year where it was lowest, so as to maximize the impact, the one year to make the change look most drastic, essentially," said Pierre-Jérôme Bergeron, who teaches statistics at the Department of Mathematics and Statistics at the University of Ottawa.
The McGill-trained statistician, who authored a paper called How to Engage in Pseudoscience With Real Data, said it doesn't look like an honest mistake, either.
"Here, I'm pretty sure they saw 2013 at the bottom, and said, 'We're going to pick that,'" he said. "Just like climate change deniers will say, 'It hasn't warmed since 1998,' but they pick 1998 because it was so hot, one of the highest, and is actually an outlier."
Bergeron noted that while statisticians often look back at recent data in five-year blocks, the Trudeau government has oddly chosen to measure only four years (2013, 2014, 2015 and 2016). Taking the data set back to 2012 would be the more normal practice — but it wouldn't produce the same impression of rocketing gun violence.
University of Ottawa criminologist Holly Johnson specializes in methodologies of measurements of crime and has coordinated surveys of violence and crime for Statistics Canada and the UN.
"I don't know the motivation or reasoning behind it," she said, "but certainly choosing the lowest rate in decades of data would suggest there's a reason for that, trying to make a point of some sort.
"Statistics can be misused, and to take a high point or a low point to try to make a political point is something I think we should all watch for.
"Statistics for rates of population that are based on very rare events, such as firearms homicides, tend to fluctuate, quite a great deal, actually."
Johnson said an objective statistician, looking for a baseline for comparison, typically would look at how a particular year compares to the average of the 10 years preceding it, to gauge how one specific year stacks up against the norm.
"A few years does not a trend make."
Johnson adds: "Certainly I wouldn't call it a crime wave."
The crime wave that wasn't
We don't yet know how 2018 will turn out for homicides, but there's a good chance it will be worse than 2013 — because every year since 1966 has been worse than 2013.
But based on Canada's latest-known homicide rate (1.68 per 100,000 in 2016), it's also likely Canada's homicide rate in 2018 will be similar to or lower than it was 10 years before in 2008 (1.83 per 100,000) — or in 1998 (1.85 per 100,000), 1988 (2.15 per 100,000), 1978 (2.76 per 100,000) or, for that matter, 1968 (1.81 per 100,000).
Indeed, the rate today is not far from where it was in 1928 (1.55 per 100,000).
The Trudeau government points out that the sub-category of homicides with firearms has gone up, and this is true … if one takes 2013 as a starting point. Otherwise, not really.
CBC asked Goodale's office why he chose to use 2013 as a baseline.
"Crime rates generally in Canada have been on the decline for more than two decades," said spokesman Scott Bardsley in an email, "but offences involving firearms have become more prevalent, especially since 2013.
"Many communities across the country have been facing a steady increase in gun violence over the past five years. This trend is a break from overall declines in crime over recent decades.
"More broadly, our caucus has heard from Canadians across the country who are concerned about gun violence. Better is always possible."
'Steady increase'
But if, rather than picking the lowest year for comparison, one were to ask how 2016 compared with a decade before, one would find that the rate of firearms homicides remained boringly unchanged. It was 0.69 per 100,000 in 2005, and 0.59 in 2006, and 0.57 in 2007. It dipped during 2013 and was back at 0.62 per 100,000 in the supposedly alarming year of 2016.
The rate of homicides with handguns in Canada in 2016 was 0.36 per 100,000. Ten years ago, it was 0.38 per 100,000.
None of that constitutes a "steady increase."
One might also look back at the long-term trend, and note with satisfaction that the rate at which Canadians are being killed with firearms has seen a historic decline of more than 50 per cent from its high point of 1.26 per 100,000 in 1975.
The rate of killings and crimes committed with non-restricted long guns has declined by more than three-quarters since the 1980s. Robberies involving guns also have declined by more than 75 per cent during the same period.
That is what a statistician might reasonably call "a steady decrease."
So it's not accurate to say that "offences involving firearms have become more prevalent, especially since 2013." It would be more accurate to say offences involving firearms appear to have become more prevalentif we use the abnormal year of 2013 as a baseline.
And that's not hard to do because, as Mark Twain pointed out, "facts are stubborn things, but statistics are pliable."
Liberals' claim of 'steady increase' in gun crime rests on a 'drastic' comparison to a low-crime year - Politics - CBC News
From the CBC, no less!!!!!
