Building · Metric

Talent retention and anchoring in international context

Canada excels at producing talent: it educates globally competitive graduates and selects skilled immigrants effectively. The harder question is whether the economy can anchor that talent — keep it here to build, scale, and lead. Statistics Canada's tax-filer tracking of STEM graduates shows aggregate retention has actually improved. But the gradient runs the wrong way on every axis that matters: the more elite the institution, the more advanced the degree, and the more portable the skill, the less likely the graduate stays.

This is the talent base the rest of the Building pillar rests on — the people who would otherwise found companies, commercialize research, and staff the "missing middle" of firms Canada struggles to grow. The core challenge is not attraction but anchoring: as TD Economics puts it, "Canada is not hollowing out; it is spilling out at the top." The headline metric is graduate retention, measured by whether graduates keep filing Canadian tax returns; a supplementary layer follows the same pattern through skilled-immigrant onward migration and the structural pull factors — compensation, firm scale, and tax — that draw the top tail south.

How this differs from Elite researcher attraction & retention. The two are complementary halves of one talent funnel, measured at opposite ends. That metric reads the inflow: whether Canada still attracts elite foreign researchers (postdoctoral fellows on work permits) and converts them to permanent residents — a gauge of the country's pull on globally mobile scientific talent. This metric reads the outflow: whether the talent Canada has already trained (its own university graduates) and settled (skilled immigrants here) stays to build, rather than leaving for the United States. Different populations, opposite directions, different data (Longitudinal Immigration Database there; graduate tax-filing here) — so the doctoral, researcher, and immigrant figures across the two pages describe distinct groups and are not interchangeable. One asks whether the best are still coming; this asks whether the best we have are staying.

Top-university STEM grads still here (yr 3)
83.5%
vs 91.1% for all other institutions — a 7.6-point gap. Canadian grads, 2015–2020 cohort.
Doctoral grads still here (yr 3)
83.4%
Lowest of any degree level — vs 90.1% master's and 89.1% bachelor's (2015–20). Falls to 76.2% by year 5 for the 2010–14 cohort.
PhD immigrants who leave within 25 yrs
34%
Highest onward-migration rate of any education level, vs 19% overall (TD / Institute for Canadian Citizenship).

The reassuring headline

Read at the surface, the STEM retention numbers look like good news — and were widely reported that way. Across successive graduating cohorts, a rising share of graduates kept filing Canadian tax returns. Among Canadian students, three-year retention edged up from 87% (2010 graduates) to 90% (2018 graduates). The improvement among international students was larger still, from 51% to 65%, tracking the post-graduation work-permit expansion. Taken alone, this is the "brain drain reversed" story.

Share of STEM graduates still filing Canadian tax returns, 3 years after graduation, by graduating cohort (%)
Source: Statistics Canada, Postsecondary Student Information System & T1 Family File (Chart 1 in Choi & Hou, July 2025). "Still filing" = appearance in the T1 Family File; a proxy for active presence in Canada.

The gradient underneath

Disaggregate that average and the reassurance dissolves. Retention is lowest exactly where it hurts most. For the 2015–2020 cohort of Canadian STEM graduates, three years out, 91.1% of graduates from ordinary institutions were still filing — but only 83.5% of graduates from high-ranking universities (Toronto, UBC, Waterloo, McGill, Montréal, Alberta). The same gap opens by degree level — 83.4% of doctoral graduates versus ~90% of bachelor's and master's — and by field, with the most internationally portable disciplines (computer and information science 86.1%, mathematics 85.9%) trailing the STEM average. Combine them and it is starker still: among Canadian computer-science graduates, 77% of those from high-ranking universities stayed, versus 90% from other institutions.

Share of Canadian STEM graduates still filing Canadian tax returns 3 years after graduation, 2015–2020 cohort (%) · red = below the STEM average
Source: Statistics Canada, PSIS & T1FF (Table 1 in Choi & Hou, July 2025). High-ranking universities = top-100 in the Times Higher Education World University Rankings 2025 for engineering, computer science, and physical sciences.

And the gap widens with time

This is not a first-year blip that closes as careers settle. Tracking the 2010–2014 cohort year by year, the top-tier deficit widens: graduates of high-ranking universities and doctoral holders keep dropping off the Canadian tax rolls faster than everyone else. By five years out, nearly one in four doctoral graduates (76.2% filing) and one in five top-university graduates (80.1%) had stopped filing — the cumulative signature of the most capable graduates building their careers elsewhere.

Canadian STEM graduates still filing, by years since graduation, 2010–2014 cohort (%)
Source: Statistics Canada, PSIS & T1FF (Table 1 in Choi & Hou, July 2025). Non-filing may include people who stayed but did not file; the study cannot separate the two, but the cross-group gradient is the signal.
Why the verdict is Weak, not Watch. Aggregate retention is fine and improving — that is real. But the metric exists to flag risk, and the risk is concentrated and accelerating: the graduates leaking out fastest are precisely those with the highest propensity to be builders, founders, and research leaders. Losing the top of the distribution costs far more than the average headcount implies, which is why the page reads the trend as a structural weakness rather than a watch-item.

The same pattern in skilled migration

TD Economics' Canada's Silent Brain Drain (May 2026) extends the picture from new graduates to the whole skilled population, and finds the identical gradient. Onward migration — leaving Canada after settling — rises with education: within 25 years of arrival, 34% of immigrants holding a doctorate have moved on, against 19% overall. The Institute for Canadian Citizenship finds highly educated immigrants are more than twice as likely to leave within five years as lower-skilled ones, concentrated in management, ICT, engineering, and science. University of Waterloo data on the graduate side is blunter still: among Canadian-born students, exit rates at the top of the skill distribution are roughly double those at the bottom.

Onward migration of Canadian immigrants by education level, 1982–2016 cohorts (% who have left, by years since arrival)
Source: TD Economics, Canada's Silent Brain Drain (Chart 1, 21 May 2026), drawing on Statistics Canada longitudinal data. "Onward migration" = emigration after admission to Canada.

Much of this outflow is invisible to conventional brain-drain metrics because it runs through temporary and semi-permanent US channels — TN visas and employer sponsorship — rather than permanent migration. By 2024, foreign-born Canadian citizens made up about 60% of Canadian applicants for US labour certification, up from just over half a decade earlier: Canada increasingly functions as a staging ground that selects skilled immigrants, integrates them, then loses them to larger US ecosystems.

Why they leave: weak anchors, not weak talent

Top graduates earn well by Canadian standards; they leave because the returns to their skills are higher elsewhere. TD identifies three structural anchors that don't hold. First, the "missing middle": Canada leans on fewer than 3,500 large firms for 36% of employment (18% in the US), with a thinner layer of scaling mid-size firms — so there are fewer places to build a frontier career at home. Second, compensation: the median US tech wage runs about 46% above its Canadian equivalent, with more paid in equity. Third, tax: Canada's top marginal rates (above 53% in Ontario, Quebec, and BC) bind at roughly C$218,000, whereas California's top rate doesn't engage until US$1 million and New York City's until US$25 million; Canada also draws 36.7% of government revenue from personal income tax against 29.1% for the rest of the G7.

The same story as IP and startups. This metric rhymes with the Building pillar's intellectual-property and startup findings: Canada generates the raw material — ideas, patents, founders, graduates — but the economic returns are increasingly realized abroad. Talent anchoring is the human-capital face of the same commercialization gap.

Findings

Finding 1

The aggregate is improving — and misleading

Three-year retention rose across cohorts (Canadian STEM grads 87%→90%; international 51%→65%). On its own this reads as a brain-drain reversal, which is how it was first reported.

Finding 2

The best institutions retain the worst

Only 83.5% of Canadian STEM grads from high-ranking universities still filed at year 3, vs 91.1% elsewhere — and just 77% among top-university computer-science grads, vs 90% from other schools.

Finding 3

Higher degrees, lower retention

Doctoral graduates had the lowest retention of any degree level — 83.4% at year 3 (2015–20), falling to 76.2% by year 5 (2010–14). The most portable fields (CS, math) also trail the STEM average.

Finding 4

Skilled migration shows the same gradient

Onward migration rises with education — 34% of PhD-holding immigrants leave within 25 years vs 19% overall. The outflow is selective, US-bound, and largely invisible to permanent-migration stats.

Sources & methodology

Primary sources
Methodology notes
  1. What is measured. Retention is proxied by whether a graduate appears in the T1 Family File (filed an individual tax return, or appeared via a filing spouse / Canada Child Benefit) in a given year after graduation. Non-filers may include people who stayed in Canada but did not file; the study cannot separate stayers-who-didn't-file from leavers. The cross-group gradient (top-tier vs other) is robust to this because the proxy applies uniformly.
  2. Headline = graduate retention; supplement = skilled migration. The page leads with the StatCan graduate series (the cleanest direct measurement) and treats TD's onward-migration and pull-factor analysis as a corroborating second layer — the structure TD itself uses, since it cites the StatCan study.
  3. Cohorts and horizons. StatCan groups graduates into 2010–2014 and 2015–2020 cohorts. Year-5 figures are only available for the older cohort; year-3 for the 2015–2020 cohort uses 2015–2018 graduates. Charts label the cohort and horizon shown; later horizons simply are not yet observable for recent graduates.
  4. High-ranking universities. Defined by StatCan as top-100 in the Times Higher Education World University Rankings 2025 for engineering, computer science, and physical sciences: University of Toronto, UBC, Waterloo, McGill, Université de Montréal, University of Alberta.
  5. No clean G7 comparator. There is no internationally comparable "graduate-still-resident-after-X-years" series, so the page is Canada-structured by design rather than filled with a mismatched cross-section. TD's tax-rate comparison is the nearest cross-country element and is reported in text, not charted, to avoid implying it is the same metric.
  6. TD figures. Onward-migration rates are read from TD Chart 1 (doctorate 11/27/34% at 5/15/25 years; total 5/14/19%). Tax-rate, compensation, and firm-size figures are TD's, drawing on Statistics Canada, the DAIS compensation study, and C.D. Howe; reproduced in the companion workbook with their stated vintages.
  7. Verdict. Weak. Aggregate retention is adequate and improving, but the loss is concentrated among the highest-capability graduates and skilled immigrants — the cohort most likely to found and scale firms — and it deepens over time. The weighting reflects the disproportionate economic cost of losing the top tail, not the average.
  8. Companion workbook. The editable series live in Building_Talent-Retention-and-Anchoring_Data.xlsx: Cohort_Trend, Retention_ByCategory, Retention_Erosion, Onward_Migration, Pull_Factors, Summary, Sources.
Page last reviewed June 2026 · Data current to the 2015–2020 graduate cohorts — the latest published by Statistics Canada