Look Before You Leap (to Conclusions): Quebec’s Labour Market May Not Be as Weak as the LFS Suggests

Quebec’s labour market stabilized in October. Employment rose by 7,500 following sharp declines in August and September, while the unemployment rate fell back to 5.7% from 6.0%. However, the increase was almost entirely concentrated in part‑time employment. As we explain in our analysis External link. published this morning, the overall picture therefore remains mixed. More importantly, the latest data do not fundamentally change the picture that has emerged in recent months or the question we explore here: Why is the Labour Force Survey (LFS) painting a much weaker picture of Quebec’s labour market than several other indicators?

October’s mixed results are only the latest chapter in what has been an unusually difficult year to read for Quebec’s labour market. According to the LFS, Quebec lost 54,100 jobs between August 2025 and August 2026, putting it dead last among the provinces (table). But the Survey of Employment, Payrolls and Hours (SEPH) tells an entirely different story, with payroll employment increasing by 5,900 over the same period.

Of course, analysts are going to analyze, politicians are going to politicize and journalists are going to chase a good headline. Who can blame them? We’re also regularly contacted by the media for our take on the LFS results. The survey has a lot to offer. It gives us the most up‑to‑date snapshot of the labour market and allows for detailed analysis across a range of sociodemographic characteristics. But there’s a trade‑off for the speed with which it delivers this information. Its estimates come with some uncertainty, and month‑to‑month changes can be volatile. And as we saw in September with education employment in Quebec External link., shifts at the national level can largely be driven by the change in one sector in one province. So it’s worth taking a second look before drawing conclusions.

This caution is especially warranted in times of significant demographic change. In its technical note External link. published this summer, Statistics Canada reminded us that the LFS accounts for changes in the number of non‑permanent residents (NPRs) using 12‑month moving averages. This method keeps estimates more stable but means that rapid demographic changes are reflected with a lag, which can contribute to temporary divergences between the LFS and SEPH. Statistics Canada also recommends looking at trends over several months, instead of reading too much into any single release.

Like all sample surveys, the LFS needs to keep pace with major changes in the population to remain representative.¹ Its rotating sample must therefore continually adjust. When the signal it gives is surprising, it’s wise to see what other indicators are saying. The SEPH provides a useful point of comparison here. Because it relies largely on administrative payroll records, it’s much less prone to sampling error than the LFS.

Two Surveys, Two Stories

The fact that these two surveys produce different results isn’t in itself a problem. They don’t measure exactly the same thing. The SEPH looks at payroll employment, while the LFS includes self‑employment. Some divergence is inevitable. Even so, the two surveys have essentially told the same story in Quebec for years, even through the extreme shifts caused by the pandemic. What’s striking today is how widely the two diverge (graph 1). That said, this gap may narrow somewhat as the latest data are revised, so we should remain cautious. Furthermore, the two surveys don’t always diverge in the same fashion. In Ontario, for example, the LFS shows much stronger employment growth than the SEPH. This was also the case for Quebec in 2018 and 2019. When two weather apps disagree, we always have the option of looking out the window. Let’s do the same here and look for clues elsewhere.

One of the first places we can turn to for corroborating evidence is public finances. Tax revenue is obviously not a direct measure of labour market conditions, since it is also influenced by income levels and the tax system. Still, it can provide a useful cross‑check. Over the last 12 months, tax revenue has been more resilient than we’d expect given the drop in hours worked reported by the LFS (graph 2, left). The evolution of hours worked and hourly wages also aligns more closely with the relatively stronger picture painted by the SEPH than with a sharp deterioration in the labour market.

Such pronounced weakness in the labour market should also have left a clearer mark on businesses. But mass layoff notices don’t signal a widespread deterioration in 2026 (graph 3 on page 3). Levels so far this year are broadly comparable to those in 2023 and 2024, while the spikes in 2025 were largely driven by a few exceptional events, including the closure of Amazon warehouses and the liquidation of Hudson’s Bay.

These notices obviously do not cover all types of job loss and reflect intended layoffs rather than layoffs actually carried out. When combined with the SEPH and tax figures, though, they point in the same direction. Quebec’s labour market has indeed slowed, but it’s not weakening as sharply as the LFS would suggest.

No Need to Throw the LFS Out with the Bathwater

The Labour Force Survey is still essential. It provides timely data and captures dimensions that other surveys do not, including duration of unemployment and involuntary part‑time work. It also allows the data to be broken down by a wide range of sociodemographic characteristics, including age and immigrant status in Canada. The LFS release also moves financial markets, sometimes more than the underlying data seem to justify. Markets react quickly to the headline number, but the movement behind it can sometimes be heavily concentrated in one province, one sector or one group of workers. This is one of the drawbacks of a statistic that is released so quickly and followed so closely.

Because the LFS carries so much weight, it is all the more important to take a step back when monthly movements appear inconsistent with the rest of the available data. No single indicator allows us to conclude that the LFS is “wrong.” Together, though, they suggest we should not take the LFS’s particularly weak signal at face value.

The same caution applies when the numbers are good. One strong month isn’t enough to conclude that a sustained recovery is underway. When a national result can be shifted by fluctuations in a single sector within one province, analysts, politicians, journalists and investors would be well advised to look beyond the headline figure and cross‑check other sources before drawing conclusions.

The same goes for monetary policy. For the Bank of Canada, one striking LFS print matters less than what the data collectively say about the degree of excess capacity in the economy. A labour market that is actually contracting tells a very different story than one that is simply growing slowly amid demographic stagnation.

¹ The challenge of ensuring representativeness is not unique to economic statistics. Election pollsters, for example, have had to adapt their methods in response to non-response bias External link.. Nate Cohn of the New York Times, in particular, has shown how weighting decisions can lead to multiple interpretations of the same race External link.. And indeed, the results of the 2026 midterm elections suggest that some of these methods may have overcorrected in the other direction. More time is needed before we can issue a final verdict.