Skip to main content

Imagine an engineer who underestimates by a third how much structural steel is required to support a 60-storey tower.

Or perhaps a newspaper that stands by two-thirds of its facts.

Well, that is the degree of precision Canadians get from the main gauge of the jobs market – the Labour Force Survey.

On Wednesday, Statistics Canada issued its year-end review of the job scene, using new population data from the 2011 Census to revise its closely watched employment numbers. It turns out the economy created just 121,000 jobs last year, not the more hopeful initially reported tally of 186,000.

As Bank of Montreal chief economist Douglas Porter pointed out, the revision alone is more than three times larger than the 17,000-plus Target Corp. workers poised to lose their jobs across Canada.

It's hard to underplay the significance of the data gap. The jobs market was clearly much weaker than anyone thought, even before the price of oil did its impression of a cliff diver.

But this isn't just about what happened last year. The more fundamental concern is that the Labour Force Survey – arguably the most important economic indicator in the country – is dangerously unreliable and increasingly volatile.

Every day, governments, businesses and individuals make decisions, based on their perception of the health of the economy and the job market. This has implications for monetary policy, government jobs programs and investments by companies. For individuals, it can influence where they put their money, and when to buy or sell a home.

"The Bank of Canada has to look at these numbers to get a gauge of what's going on," pointed out economist Robert Fairholm, partner and economist at the Centre for Spatial Economics. "It's like driving with a blindfold on. The chance of a mishap happening certainly increases."

The massive year-end revision caps a bad year for the survey. In July, a bungled computer system upgrade caused Statscan to mistakenly report that the economy created just 200 jobs. The correct figure should have been 42,000.

And there is far too much about this country that the survey does not tell us. As The Globe and Mail's Joe Friesen reported earlier this month, Statscan doesn't even bother to gather jobless statistics on First Nations Reserves, claiming it's too costly and hard to find people. The result is that half the country's First Nation's people don't appear anywhere in the unemployment figures.

The Labour Force Survey is a poll, and therefore has a significant margin of error. Every month, Statscan sends a questionnaire to 56,000 Canadian households, representing roughly 100,000 people, quizzing them about their employment status. The result is an estimate that the agency acknowledges is "statistically significant at the 68-per-cent confidence level."

Disturbingly, the difference between an economy that is adding jobs and one that's losing jobs is often within the margin of error.

Another major concern is that the survey chronically exaggerates the importance of sparsely populated regions, while under-representing major cities, such as Toronto, Montreal and Vancouver, according to Sam Boshra, an independent Montreal-based economist and statistical analyst.

Toronto, for example, has 17 per cent of the population, but represents less than 6 per cent of the jobs survey sample. That can significantly affect the data when every person in the survey represents about 200 people.

Mr. Boshra points to two other potential problems – a low survey response rate, particularly in the summer months, as well as volatility in the number of people claiming to be self-employed.

Think of the survey as rough-and-ready data. Statscan releases the numbers 10 days after the data is collected.

Statscan does have another way to track employment, but it comes out nearly two months late and it too is prone to large revisions. The Survey of Employment Payrolls and Hours is a combination of a poll of 15,000 employers plus actual payroll data reported to the Canada Revenue Agency.

The U.S. manages to produce monthly surveys of both people and jobs, on the same day, soon after the end of every month – something Statscan insists it can't do.

It hardly matters whether the problem is budget cuts, inadequate resources or lack of will to produce more reliable numbers. Canadians need better jobs data. Period.

Follow Barrie McKenna on Twitter: @barriemckennaOpens in a new window

Report an error

Editorial code of conduct

Tickers mentioned in this story