David Rosenberg: Canada can’t count to a million — and it’s skewing the data policymakers need

Article content

By Dylan Smith and Atakan Bakiskan

A thoughtful approach to immigration has been a historical strength of Canadian economic policy. The recent post-COVID-19 immigration surge has created a number of well-known challenges, however, as slow-moving housing markets and social services systems adjust to the sudden rise in the country’s population.

Article content

But another, seemingly more basic, problem has arisen for policymakers and analysts alike: Canada’s population has become increasingly difficult to count, and the balance of evidence points toward an undercount of one million temporary residents in the country’s main employment survey. This has eroded the reliability of a host of market-relevant variables involving population counts, from employment figures to housing construction to productivity.

Article content

After raising five-year immigration targets in the aftermath of the sharp drop in immigration in 2020 and 2021, the federal government last week announced immigration targets would be held at the current planned levels through 2026. Over the next three years, Canada will aim to bring in roughly 500,000 permanent residents per year (around 1.3 per cent of the 2022 population). This compares to a pre-COVID-19 target of 310,000, and 250,000 a decade ago.

However, the policy leaves room for a large number of “newcomers” beyond the permanent resident target. Sixty per cent of Canada’s recent population growth over the past four quarters has come from temporary residents. It is this large and growing stock of temporary residents (workers and students on a host of non-permanent visas) that creates the population measurement challenge.

The Labour Force Survey (LFS), a weighted sample of approximately 60,000 households, to take perhaps the most important example, only estimates 500,000 employed temporary workers. Administrative data from Immigration, Refugees and Citizenship Canada (IRCC), however, show around 1.5 million employed temporary residents as of December 2022. That’s a discrepancy of one million workers.

Article content

Article content

Although there are some reasons to think the IRCC data come with a slight upward bias (such as double counting individuals with both study and work permits), there is no question at all that the LFS numbers are heavily biased downward. This owes partly to ambiguous questions in the survey, and partly to a sampling methodology that simply underweights temporary residents due to low response rates (according to recent research from the C.D. Howe Institute).

A similar issue arises between the Canadian census and the quarterly demographic estimates. According to the latest census, there were 925,000 non-permanent residents in 2021. However, the quarterly population data show that number to be 1.38 million, based on a new methodological update that corrects for those who stay in Canada after their permit expires.

It is part of the unfortunate nature of mismeasurement and unreconciled data that we cannot precisely assess the errors in the related statistics, but we can at least attempt to match the best available measurements — with appropriate adjustments — to the analysis at hand. With this in mind, we highlight three areas where we can get a sense of the directionality of potential errors.

Article content

Labour market

There is no reliable estimate of unemployment rates among temporary residents, and an accurate count could influence official unemployment in either direction. More importantly, skill mismatches between vacancies and the available labour force at the industry level cannot be properly assessed without accurate data.

For this purpose, it is worth keeping in mind that mismatches are likely less severe than the LFS implies, meaning that wage pressure is probably lower than it seems at first glance.

Housing and construction

Canada Mortgage and Housing Corp. (CMHC) estimates 3.5 million new units must be built by 2030 to restore affordability in the housing market. But CMHC uses census data, not quarterly demographic estimates to arrive at this number.

Since temporary residents need shelter, too, this is likely an undershoot, and housing construction will need to be even higher than estimated, with upward pressure on prices and rents in the absence of a large enough increase in the stock.

Article content

Productivity

Productivity is measured as real gross domestic product per hour worked. Basic hours worked data come from the LFS (supplemented by the census and various other sources for industry granularity), which, in turn, depends on the number of employed residents. Holding average hours and real GDP constant, adding an extra one million employees would reduce measured productivity in the third quarter by 4.6 per cent.

Given that the employment undercount has built up over time, this means the decline in productivity over the last half-decade is likely even worse than it appears. This has a host of policy implications and suggests the neutral real interest rate is lower than models without such an adjustment would imply, with an obvious positive read-through to lower bond yields.

Dylan Smith is a senior economist and Atakan Bakiskan is a economist at independent research firm Rosenberg Research & Associates Inc. founded by David Rosenberg. To receive more of David Rosenberg’s insights and analysis, you can sign up for a complimentary, one-month trial on the Rosenberg Research website.

Article content