Don’t panic: StatsCan job error’s more awkward than catastrophic
Remember when Statistics Canada said the country added 200 jobs in July?
That was last week: The number-crunching agency really meant 42,000.
The error resulted in an unprecedented recall of StatsCan’s Labour Force Survey numbers, reissued Friday, and prompted consternation among economists, journalists and stats nerds alike. If you couldn’t trust Statistics Canada’s flagship monthly jobs numbers, whom could you trust?
But while the error’s disconcerting it’s likely more ill-timed than portentous: Statistics Canada has said it’s in the midst of a once-a-decade update, and one of the modules used to crunch Labour Force Survey numbers wasn’t properly updated, says Universite Laval economist Stephen Gordon.
“It’s bad luck. It’s human error,” he said. “…It’s just really, really bad timing.”
Bad timing, indeed: The mistake comes amid cuts to StatsCan funding and as the destruction of the long-form census, Canada’s marquee data-gathering tool, continues to reverberate.
But it’s hardly indicative of a statistical body’s death throes, or even a broader decline.
“There may be” a time when people stop trusting Statistics Canada altogether, Gordon said. “But I don’t think we’re there yet.””
This also may be a good time to remember the Labour Force Survey’s hardly iron-clad at the best of times: Its monthly figures are notoriously volatile and have a standard error of more than 20,000 people. To cut that figure in half you’d need to quadruple the number of people the Labour Force Survey covers, to about 200,000 a month. That gets pricey.
“The data are noisy,” Gordon said. “We really shouldn’t be spending a whole lot of energy focusing on changes in one monthly number.”
But we do. And we seldom discuss the jobs numbers’ inherent unreliability because so much depends on them: If you are or ever will be out of a job, the local unemployment rate will determine what you get and for how long. Businesses, analyst forecasts and the Bank of Canada all use these labour numbers. And uncertainty freaks people out.
“It’s a big, expensive project. And if we could think of something better, we’d use it,” Gordon said.
The real crisis Canada faces when it comes to reliable statistics remains the long-form census’s demise, which will keep affecting everyone who uses any population numbers for anything (and that includes public health officials and businesses, as well as researchers, journalists and government analysts).
At the same time, notes Carleton University economics professor and associate dean Frances Woolley, some of the most marginalized populations are also harder for a surveying statistician to reach in traditional ways.
Statistics Canada aims to follow people for six months – first in in-person interviews, then by phone. As the survey’s non-response rate has edged up by several percentage points over the past several years, the highest rate of “slippage” – not being able to reach people by phone you already talked to in person – has gone up the most for people aged 20 to 29, Woolley noted.
“We had this golden age when we could just call people up and ask them questions. … I think we’re in a different world,” she said.
“You’ve got to think outside the box.”
Gordon said many researchers are turning to other sources of data – the Canadian Revenue Agency for income statistics, for example (Global News has done the same) in an effort to avoid using household survey data as a comparator. While some public health bodies have suggested gathering their own data at the local level, where Statscan’s survey is least accurate, doing so can be prohibitively expensive.