It’s a frustration felt by anyone who’s tried to peer into Canada’s not-so-distant economic past for context to the trends and events of today: when it comes to important historical statistics about Canada’s economy, there’s no there, there.
Crucial gauges of the economy tracked by Statistics Canada, from gross domestic product (GDP) and exports to wages and employment, only go back an annoyingly short period of time. Quarterly GDP extends back only to the early 1980s (while the monthly measure of GDP starts only in the late 1990s). Exports and imports are tracked only to the late 1980s. Average weekly wages, tracked through the survey of employment, hours and payrolls, reach back only until 2001. Any further back, and Canada suffers from a collective state of economic amnesia.
It’s not that StatsCan only started tracking these things in the last two or three decades. While StatsCan was formed in 1971, the Dominion Bureau of Statistics was gathering data on the economy back to the 1920s. But StatsCan has had a long-standing habit of terminating old data series whenever it changes methodologies, rather than revising the older statistics. As such StatsCan’s CANSIM database is littered with “terminated” datasets.
Stephen Gordon, an economics professor at Laval University, has called this StatsCan’s attention deficit disorder, here and here. Now, Gordon has offered a solution. Over the past few months he’s been piecing together the strands of statistical history lurking on StatsCan’s CANSIM website and in other archives and databases. The result is what he calls Project Link, a shared collection of major Canadian economic data sets dating back to the 1950s. (See his post at the Worthwhile Canadian Initiative blog for more info.)
Think of it as Christmas in October for Canadian economists, academics, policy-makers and business journalists.
Here, for example, is a look at real (inflation adjusted) average weekly wages in Canada, with what StatsCan’s current series tells us displayed in orange, and what Gordon found by splicing together earlier terminated series dating back to 1951, in blue. As Gordon notes, suddenly a narrative that was popular during the last election—that Canada has suffered from stagnant wages over the last decade—needs a rethink. The real wage stagnation occurred from the mid-1970s to the early 2000s.
Here also is employment growth back to the early 1950s. I’ve overlaid recessions as determined by C.D. Howe’s Business Cycle Council. The Great Recession of 2008 looks a lot less severe in terms of its impact on employment when compared to earlier downturns, even in the halcyon 1950s.
Lastly, a third example, showing the annual growth of Canada’s economy dating all the way back to the statistical palaeolithic era of 1961. Canada’s growth woes over the last five years are even more apparent in this view.
Gordon says that after complaining so often about the lack of historical and comparable statistics, he realized “it’s better to light a candle than to curse the darkness.” He acknowledges there may be issues with the precision of the data, given the methodology changes made by StatsCan over time, but he believes the growth rates more or less stay the same. “It’s better than nothing and nothing is what we’ve got now,” he says.
It’s hard to overstate how important this type of project is. Because the United States maintains far better detailed and readily available historical statistics, Canadian economists and policy researchers are regularly forced to rely on U.S. data to try to understand what’s happening in this country. The two economies are similar, but not identical, and the performance of wages over the last two decades is an example of that. Likewise, Gordon notes, when the U.S. tumbled into recession in 2007, commentators “could quite confidently tell stories of how things compared to the Great Depression, but we couldn’t do that.”
We still can’t, at least not yet. Gordon says Project Link is a work in progress, and he wants to continue piecing together Canada’s historical economic record. It’s tedious work, which involves scanning microfiche of old publications and manually inputting the data.
In the meantime, unfortunately, StatsCan’s tendency to kill off historical series continues apace.
I see StatsCan ADD strikes again. A new methodology for pipeline stats means 1997 to February 2016 is now only ‘for reference’. pic.twitter.com/MKO3TZHh7E
— Jason Kirby (@jason_kirby) September 28, 2016
Oh hey, StatsCan has new data on federal spending on science & tech.
Handy context going forward, right?
— Jason Kirby (@jason_kirby) May 25, 2016