Skip to main content

Remember the awful November job numbers? Looks like they weren’t so awful. It’s a reminder that outliers in Statistics Canada’s Labour Force Survey (LFS) may say more about the LFS’s shortcomings than they do about the job market.

Back in early December, Statistics Canada reported that its monthly LFS showed a loss of a staggering 71,200 jobs in November – the worst month since the Great Recession of 2009. This news was greeted like hearing that your ship had whacked into an iceberg.

Economic commentators lamented a collapse of Canada’s hiring boom. The Conference Board of Canada’s monthly consumer confidence index slumped to its lowest reading in nearly three years. The Conservative opposition in Ottawa warned of a “made-in-Canada recession,” blaming the failing policies of a heartless federal government that was filling Christmas stockings with pink slips.

Except it now appears that November’s job losses were nowhere near as dire as that initial report indicated. That’s according to Statscan’s own subsequent labour data – none of which have garnered anywhere near the same public attention as that initial report.

In late January, Statscan published revised LFS data for the past three years, to incorporate tweaks it had made to its seasonal adjustments for the data. Instantly, that 71,200-job plunge became a considerably more modest 54,400. The bulk of that – about 39,000 – came from public- and private-sector payrolls, with self-employment playing a smaller roll.

A few days later, Statscan’s monthly Survey of Employment, Payrolls and Hours (SEPH) – which has much better statistical accuracy than the LFS – showed a drop of a relatively trivial 12,000 payroll jobs in November. Suddenly that mountain of lost jobs looked more like a molehill.

It’s a reminder that we pay an awful lot of attention every month to a Statscan product that is, frankly, not the agency’s best work. For all its usefulness, the LFS is, by its methodological nature, prone to inaccuracies – sometimes big ones.

The LFS is a monthly telephone survey of about 100,000 people – about one-third of 1 per cent of Canada’s working-age population. This sampling generates an employment estimate that in any given month is accurate to within plus or minus 30,000, with 68-per-cent certainty. Which means that the 54,000 loss in the November report should be read as two-in-three odds of loss of somewhere between 24,000 and 84,000, and a one-in-three chance that the actual change would be outside of that range. Needless to say, “54,000” is an estimate that is at best very rough, and perhaps totally useless. November may have been an outright statistical miss.

The employment figures in the SEPH, by contrast, are derived from a combination of about one million payroll-deduction records filed with the Canada Revenue Agency, and administrative data provided by federal, provincial and territorial governments for their own employees. It is, practically speaking, a very precise measure of payroll employment.

The SEPH’s failing is in its timeliness. It is released nearly two months after the end of the month it is measuring. That’s not nearly quick enough to serve as the go-to gauge of the labour market, which is probably the single most valuable indicator of economic health.

The LFS is the most timely economic indicator in the country; we get it within about a week of the end of the month. But what we gain in timeliness, we sacrifice in accuracy.

Statscan points out that the statistical error in the LFS data shrinks when taken over many months. So the real issue is the broader trend that the past several months of data portray. While that trend is nowhere near as ugly as the November LFS suggested, that headline-grabbing report did focus public attention on a hiring climate that has markedly cooled.

Over the past six months, the LFS employment count has risen by an average of 12,000 jobs a month – less than one-third of the pace for the six months before that. The SEPH tells a similar tale: an average of less than 16,000 in the past six months, compared with 27,000 in the six months prior. Since August, both gauges have been essentially flat.

In particular, they point to one of the key concerns that Bank of Canada officials have expressed: that the global slump in manufacturing exports has washed up on Canada’s shores. Both reports show a significant downward trend in manufacturing employment.

That’s valuable information, and it speaks to where the economy is headed, not just where it has been. This is why labour market analysis – or any economic analysis – is better when using a toolbox of indicators, rather than relying on just one.

November is a reminder that LFS is a statistical hammer. Very handy, but if you try to rely on it for precision tasks, you’re likely to bash your thumb.