New research from Club Vita, OPTrust, and Eckler finds that removing gender data from mortality models could distort pension liabilities
New research from Club Vita, OPTrust, and Eckler finds that removing gender data from mortality models could materially distort pension liabilities across Canadian defined benefit plans.
A joint research paper published in August 2026 examines what happens to longevity modelling accuracy when gender data is removed. The findings carry direct implications for plan sponsors, pension administrators, and HR professionals who manage member data.
The paper, The Value in Pension Plans Using Gender Data, draws on Club Vita's Canadian pension dataset. It compares gender-specific mortality models against gender-neutral alternatives across a set of Canadian defined benefit plans.
What gender data loss means for pension funding
Gender has long been a standard actuarial input. Men and women show distinct longevity patterns, and those differences shape the mortality tables that underpin liability calculations.
The research finds that removing gender data reduces model accuracy. Across the plans studied, gender-neutral basic models produced absolute liability changes ranging from around 0% to 6%. Plans with predominantly male members tended to underestimate their liabilities. Plans with predominantly female memberships tended to overestimate theirs.
For the OPSEU Pension Plan, administered by OPTrust in Toronto, gender-neutral assumptions would have understated total liabilities by around 1%. That equates to roughly $270 million.
The research also found that gender-neutral models produced 30% fewer plans within a 95% confidence interval, compared to gender-specific equivalents. Plans that underestimate their liabilities risk setting aside insufficient assets to meet future pension promises.
What this means for plan sponsors and HR teams
OPTrust commissioned the research after members raised concerns about gender data collection.
The paper makes clear that gender data is collected for actuarial purposes, primarily to calculate pension liabilities accurately. It is not used to determine pension entitlements.
For members outside Quebec, gender data is also used to establish the male-to-female liability split needed to derive unisex commuted values. Actuarial standards were updated in early 2026 to distinguish between gender and sex at birth. The updated standards also contemplate non-binary results and the possibility of neither variable being available.
OPTrust currently sees less than 1% of new plan members declining to disclose their sex or gender on enrolment forms. Club Vita's broader Canadian dataset shows a comparable pattern.
Why gender data proxies are not enough
The research tested whether other demographic variables could offset the absence of gender data. Postal code, occupation, pensioner type, retirement health, and spouse age difference all contribute useful information. But none can fully replace gender's predictive power.
Postal code captures socioeconomic patterns linked to longevity and is largely unaffected by gender-mix distortions. Occupation can partially serve as a proxy, particularly for blue-collar plans with a high concentration of one gender. Spouse age difference was also introduced as a new variable in the gender-neutral models.
Even with these additional factors included, gender-neutral full models still produced meaningfully wider variance in liability estimates. The researchers caution against relying on these variables as permanent proxies. Partnership patterns shift over time. Postal codes may become less useful as letter mail declines.
Removing gender data from actuarial inputs would also reduce the industry's ability to track the pension income gap between men and women. Canadian women already retire with roughly 17% less income than men – a gap shaped by earnings differences, labour force participation, and longevity. Accurate gender data remains one tool that supports plans in monitoring that divide.
What plan sponsors should watch
The paper does not recommend that plans ignore evolving privacy norms or member preferences. Some regulatory jurisdictions may move to restrict gender data use in longevity calculations. Courts have also shown limited appetite for arguments that gender data is necessary in this context.
What the research does recommend is continued investment in data quality and industry collaboration. Plans should understand how their demographic composition affects the impact of gender data loss. They should also build governance frameworks that can adapt as member data evolves.
