Statistics Canada's dismal July jobs report made the country's employment situation look far worse than it actually is, prompting economists to question how the agency counts the country's 17 million workers.
Statscan's chronic inability to accurately track teachers in the summer months, rather than a faltering economy, largely explains why Canada's job creation machine abruptly stalled in July.
And that's left people wondering if they can trust what is probably the most closely tracked barometer of the economy - how many Canadians are actually working every month.
In a report Monday, Scotiabank economist Derek Holt suggested investors should "throw out the jobs report."
Even Statscan officials acknowledge the monthly jobs reports aren't a particularly good barometer of anything because they often obscure longer-term trends.
It turns out the drop in the number of teachers was entirely predictable, and almost certainly temporary. It has happened every summer for the past several years, and the lost jobs miraculously reappear in September and October when teachers go back to work. Each July since 2007, Canada has lost an average of nearly 50,000 teaching jobs in July.
"It's not that teacher jobs suddenly became as risky as Wall Street investment banker jobs," Mr. Holt said in an interview. "Don't read too much into the huge swing."
The economy unexpectedly lost 9,300 jobs in July, a drop caused almost entirely by the largest decline in education workers in more than three decades - 65,000. Statscan's definition of educational service workers includes everything from college professors and school teachers to driving and flight instructors. It doesn't include custodial staff or other people who might work in schools. But high school and elementary teachers make up the bulk of the category.
The report was so bad it has raised doubts whether the Bank of Canada can afford to continue raising its key interest rate in the months ahead.
But economists warn investors not to put too much stock in one month's data.
Unfortunately, there's not much Statistics Canada can do about the statistical aberration, given the limitations of its monthly Labour Force Survey.
In recent years, school boards and colleges have shifted thousands of employees onto contracts that typically run from September to June, rather than a full 12 months, as they did in the past. Technically, that leaves many school employees without a job in the summer months, even though virtually all of them will be under contract again when school returns in the fall.
Statistics Canada, like data-collection agencies in other countries, smoothes out these predictable seasonal variations by removing known aberrations from the raw data. They make adjustments for the weather, for the calendar and for predictable seasonal patterns.
But that isn't possible in this case, without changing the survey itself, because the seasonal pattern has changed so dramatically for one category of workers.
As Statscan official Vincent Ferrao pointed out, "the raw numbers still show a decline." Investors, economists and other consumers of its data should never put too much emphasis on monthly numbers, Mr. Ferrao recommended.
"We like to look at the longer term trend," he said.
In a recent research paper, Statscan acknowledged seasonally adjusted numbers often aren't a good reflection of underlying trends - in jobs, in retails sales or housing starts. That's because they can't filter out so-called "irregular" components.
"Seasonal adjustment removes only predictable ups and downs in the data," the agency said in the report. "Analysts still face the challenge of identifying the unexpected ups and downs from the irregular component, comprised of random events and emerging seasonal patterns."
Toronto-Dominion Bank economist Francis Fong praised Statscan as "one of the greatest statistics organizations in the world." But he said there's clearly been "a structural break in the seasonal pattern" affecting teachers, and the agency is apparently "missing something" with its seasonal adjustments.
He urged the agency to "adjust" its adjustment when it comes to counting teachers. Mr. Fong said the jobs survey is simply too important to people's understanding of the economy to get wrong.
Mr. Fong traces the data reporting problem to 2004, soon after Ontario eliminated Grade 13. Education employment, which had been historically stable, suddenly jumped up by 10 per cent or nearly 100,000 people over an 18-month span. He said he "can't figure out why," beyond the obvious explanation that colleges ramped up staff to cope with the one-time double-cohort effect and then never let them go.
The bottom line is that the July report makes Canada's job situation look worse than it really is. No, the economy isn't likely to continue generating jobs at a rate of 50,000 per month as it did coming out of the recession.
But Mr. Fong said steady monthly job gains of 15,000 to 20,000 are attainable.
Mr. Holt pointed out that just as Canada's job-creation machine of earlier this year wasn't a reliable indicator of what to expect in the months ahead, neither is July's down trend. He called the month's data "a long expected give back."
The test will come in August, September and October when all those teachers should be happily back on the payroll.