Statistics Canada's monthly Labour Force Survey showed a massive 74,000-job surge in employment in September, the biggest gain in 16 months. It blew away economists' consensus expectation of 20,000, which itself would have been a healthy gain. Unemployment dropped to 6.8 per cent, a post-recession low.
There's a lot to like in these numbers. The gains included a whopping 124,000 new private-sector jobs, and 69,000 full-time jobs. Okay, self-employment slumped by 56,000 jobs, but I think most people would agree that full-time private-sector employment is preferable to uncertain self-employment in terms of evidence of a healthier labour market.
But hold the phone – didn't we see some of these same numbers before, only in reverse? Like, just a month ago?
You may recall that in August, Canada's employment took a surprise step backward, declining 11,000. Most notable in that disappointment were a drop of 112,000 in private-sector jobs, and a surge of 87,000 in self-employment. The numbers sent some distressing signals about the quality of the labour market – suggesting that steady jobs from established employers were disappearing, to be replaced by much less stable freelance and contract employment.
At the time, though, some people (including me, if I may go all told-you-so for a moment) argued that these numbers looked so off-kilter as to suggest a statistical one-off. Remember that this is a survey, and like any survey, it has a statistical margin of error.
Statscan reports the "standard error" for the overall survey at 28,500. That means that statistically speaking, 68 per cent of the time the actual monthly job-change figure will be within a range of 28,500 plus or minus the figure Statscan reports; the other 32 per cent of the time, it will be a figure outside that range.
The standard error on the private-sector employment figure is 38,200, while for self-employment it's 25,900.
What this means is that big numbers in the survey need to be taken with a grain of salt. They may be stretching the bounds of the statistical error. They may just be wrong. Keep in mind that each person surveyed is essentially a stand-in for 200 working-age Canadians; you get a sample that doesn't quite accurately represent the overall population, even by a handful of respondents, and you can get some anomalous numbers pretty quickly.
Given all that, the September numbers certainly strengthen the argument that there was something odd in the August survey, which has now been corrected in the subsequent September poll. Yes, you could argue that the opposite is true, that it's the September survey that's out of whack. But given that the odd numbers that surfaced in August, and reversed in September, it appears more logical that this latest survey is righting the ship.
It's curious that since the recession, every monthly increase of 60,000 jobs or more (there have been six of them) has been either preceded or followed by a decline. In four of those cases, the big gain was both preceded and followed by a decline, certainly raising questions about whether these swings reflected a substantial volume of statistical noise. So perhaps the next survey, for October, will be the tie-breaker, indicating whether there is a true upward trend or whether the September gains might be a statistical spike overstating the labour market's strength.
That said, the weakness in employment data in recent months has been at odds with the acceleration of exports and the Canadian economy in general that should have suggested the right conditions for an improving labour market. The lack of job growth was the head-scratcher.
The huge jump in September almost certainly overstates the actual gains in the month, but a return to a healthier trend certainly makes more sense given the underlying economic conditions. The September number may have plenty of statistical noise in it, but the needle is probably pointed in the right direction.