Thin Files, Thick Lives: Underwriting the Credit-Invisible

A person can pay rent on the first of every month for six years, hold the same job for four, keep a chequing account that never goes negative, and still return nothing a scorecard can read. The record exists. It sits in a bank account, a landlord's ledger and a payroll file.

Thin file underwriting is the practice of deciding on applicants whose credit bureau record is absent or too sparse to score, using evidence collected directly from them instead. Statistics Canada put about 1.1 million Canadian economic families, or 7.2%, in that position, rising to 14.8% among immigrant families in Canada less than two years.

Who is credit-invisible in Canada?

About 1.1 million economic families, or 7.2% of them, according to Statistics Canada's Economic and Social Reports study published on 27 September 2023 and corrected on 29 February 2024. Roughly 26% of those families were immigrants. The rate is highest in the first two years after landing and again after sixty.

That study is the only Canadian estimate of its kind. It pools the 2016 and 2019 Survey of Financial Security into a combined sample of 22,821 observations. The data date is 2016 and 2019. The publication date is 2023.

It is a survey proxy rather than a bureau count. Statistics Canada classifies an economic family as credit invisible when it answers no to every question about holding credit products or reports zero on every balance, and the authors note that some may have no demand for credit rather than no ability to obtain it. Canada has no published figure derived from bureau files themselves.

The shape of the curve is the part most people miss. Statistics Canada's figures run 14.8% for families in Canada under two years, 6.1% at two to four years, 3.5% at ten to nineteen, then back up to 14.1% past sixty. Invisibility clusters at both ends of a life in the country, and immigrant families overall sit at 6.4%, below the 7.5% for Canadian-born families.

What does the newcomer arithmetic actually look like?

Immigration, Refugees and Citizenship Canada set a target of 380,000 permanent resident admissions for 2026, published 5 November 2025, alongside 385,000 new temporary resident arrivals. Statistics Canada counted 2,558,562 non-permanent residents in Canada on 1 April 2026. Each of them started with no Canadian credit record.

Statistics Canada's population estimates of 17 June 2026 recorded 83,149 permanent immigrants admitted in the first quarter of 2026, down 20.2% from 104,210 a year earlier. Even at reduced levels, the annual inflow of people starting a Canadian financial life at zero runs into the hundreds of thousands.

What forms first is a credit card, and often only a credit card. In Statistics Canada's credit visibility study, 93.3% of newly landed immigrant families held a credit card, above the 88.1% recorded for Canadian-born families. Mortgage holding among the newly landed was 12.3% against 54.6% Canadian-born, vehicle loans 14.5% against 42.9%, and its November 2023 summary of the same work put newly landed families 30.6 percentage points less likely than Canadian-born families to hold a line of credit.

A thin file is thin in a specific way. It holds the instrument with the shortest history and the smallest limit, and none of those that demonstrate capacity at size. A model trained on Canadian-born files reads that as inexperience, when the honest reading is recency.

The rest of the arithmetic moves faster than the file does. Statistics Canada reported on 7 April 2026, on data from the third quarter of 2024 and the year to September 2025, that 42.5% of recent working-age immigrants not employed on arrival had found a job or started a business within three months, against 31.3% for the cohort that arrived ten to fifteen years earlier. Its study of 16 June 2026, using 2017 to 2021 data, found economic-class immigrants reaching homeownership rates comparable to Canadian-born residents by their fifth year here, while the Bank of Canada's July 2024 Monetary Policy Report put the general immigrant crossover nearer ten.

Income arrives in three months. A file that can carry a mortgage takes years. A lender asking a newcomer for two years of Canadian credit history is not measuring risk, it is measuring arrival date.

What do young borrowers and cash-economy workers look like in the data?

Canada publishes no age breakdown of credit invisibility drawn from bureau files, and no measure of how many workers are paid in ways that never reach a credit record. The closest Canadian evidence is indirect: Statistics Canada's underground economy series and its gig work measurements, both of which describe income that exists without leaving a trace a lender can pull.

For young borrowers, the honest answer for Canada is that nobody has published it. The most-cited figure is American: the Consumer Financial Protection Bureau's Data Point on credit invisibles, published in May 2015 on records as of December 2010, found 64.5% of 18 and 19 year olds and 20.2% of 20 to 24 year olds in the United States credit invisible, about 10.1 million of 26 million in total. Statistics Canada's credit visibility study reports only that age was negatively associated with invisibility before controls and reversed after them, which is not a rate anyone can underwrite against.

Cash-economy work is measured, just not as a population. Statistics Canada reported on 18 March 2025, on 2023 data, that Canada's underground economy reached $72.4 billion, or 2.5% of GDP, and that $32.7 billion of unreported income went to employees as undeclared wages and tips, equal to 2.2% of economy-wide employee compensation.

Gig work sits alongside it. Statistics Canada's release of 4 March 2024 counted 871,000 people whose main job was gig work in the fourth quarter of 2022 and 468,000 paid directly by a digital platform in the fourth quarter of 2023. Neither was built to answer a lender's question. Each describes someone whose earnings are real and whose pay stub is not.

Three groups in Canada whose evidence sits outside a bureau file, the most recent Canadian measurement available for each, and the date of the underlying data rather than of publication.

Group	Most recent Canadian measurement	Source and publication date	Data date
Families with no credit products	1.1 million economic families, 7.2%; 14.8% among immigrant families in Canada under two years	Statistics Canada, Economic and Social Reports, 27 September 2023, corrected 29 February 2024	2016 and 2019 Survey of Financial Security
People newly arrived in Canada	83,149 permanent immigrants admitted in one quarter; 2,558,562 non-permanent residents in the country	Statistics Canada population estimates, 17 June 2026	1 January to 1 April 2026
Planned future arrivals	380,000 permanent resident admissions targeted for 2026, plus 385,000 new temporary resident arrivals	Immigration, Refugees and Citizenship Canada levels plan, 5 November 2025	Targets for 2026
Workers paid outside recorded channels	$32.7 billion in undeclared wages and tips, 2.2% of employee compensation, within a $72.4 billion underground economy at 2.5% of GDP	Statistics Canada, 18 March 2025	2023
Gig and platform workers	871,000 with gig work as their main job; 468,000 paid directly by a digital platform	Statistics Canada, 4 March 2024	Q4 2022 and Q4 2023
Renter households	5.0 million renter households, 33.1% of all households, growing 21.5% over a decade against 8.4% for owners	Statistics Canada, 2021 Census housing release, 21 September 2022	Census day, 11 May 2021
Young adults with no Canadian bureau history	Not published	No Canadian agency publishes an age breakdown of credit invisibility	Not applicable

Which signals have actually been measured on thin-file populations?

Bank transaction data and rental payment history are the two with published results on populations resembling this one. Grading alternative data categories against each other belongs to a separate piece. The question here is narrower: were the studies everyone cites run on people whose bureau files were thin in the first place?

FinRegLab, a non-profit research organisation, reported on 25 July 2019 across six non-bank lenders that cash-flow metrics standing alone generally performed as well as traditional credit scores, among populations and products like the ones studied, where traditional credit history is not available or reliable. Those lenders served exactly this population, which is what makes the finding relevant here rather than merely favourable. The wider argument about how bank data and bureau data fit together sits in the case for using both sources.

The follow-up study is where the evidence thins, and FinRegLab says so itself. Its July 2025 work modelled 424,546 consumers, and the report states that of 750,266 initial observations only 2,588, or 0.34%, lacked a conventional credit score. The authors write that the sample includes relatively few consumers whose bureau data is so limited that traditional models may struggle to score them, that this restricts their ability to assess impacts on those subpopulations, and that the results would likely have been stronger with more such observations.

The largest and most recent study of cash-flow underwriting contains almost none of the people cash-flow underwriting is meant to serve. The research discloses that in full, so the gap is not a flaw in the work. It comes from where the data comes from: lenders' books hold the applicants who were approved, and approval already selected for having a file.

A fuller grading of alternative data categories against published evidence sits in the alternative data scorecard.

What happened where thin-file underwriting was actually tried?

Two deployments have published outcomes. Fannie Mae added positive rent payment history to its automated underwriting in September 2021 and reports the count of applications it changed. The Urban Institute ran a randomised trial of rent reporting on renters in affordable housing and reported what happened to their files. Both results are real and both are small.

Fannie Mae announced the change on 11 August 2021, effective 18 September 2021, saying then that fewer than 5% of renters have rent payments reported to credit bureaus, and that in a sample of declined applicants 17% could have qualified had rental payment history been considered. That was a projection. The measured outcome, on Fannie Mae's own positive rent payment reporting page, is that as of May 2026 more than 21,000 single-family mortgage applications had improved their automated underwriting recommendation using rent payments in borrower-permissioned bank statements, over September 2021 to May 2026.

Twenty-one thousand mortgages over roughly four and a half years, in a market originating millions a year, is a modest number attached to a real mechanism. It is also the only figure of its kind anyone publishes.

The Urban Institute's randomised trial, published in June 2025 by Brett Theodos and colleagues, enrolled 269 renters in affordable housing across waves in the summers of 2021 and 2022, 141 in treatment and 128 in control. The share of the treatment group with no credit score fell from 16% to 8% while the control group moved from 23% to 25%. Average scores rose 11 points in treatment against 3 in control, an estimated effect of 7 points that was not statistically significant.

File formation is what that trial measures, and file formation is what a credit-invisible applicant lacks. Whether the resulting file predicts repayment better than the rent record it was built from is a different question that no independent study has answered.

What does this change about how an application is built?

Intake design decides which applicants can produce evidence at all. Someone whose income arrives by payroll deposit, whose rent leaves the same account on the first, and who can connect that account in one sitting is legible. Someone asked for two years of Canadian bureau history is not.

Rent is the most widely held piece of evidence in this population. Statistics Canada's 2021 Census housing release of 21 September 2022 counted 5.0 million renter households, 33.1% of all households, growing 21.5% over the decade against 8.4% for owners, and CMHC's mid-year rental market update of 9 June 2026 reported 2025 vacancy rates across the major centres between 2.7% and 5.0%. A third of the country pays rent, and it is recorded almost nowhere a lender can see.

Time in country and time in file move at different speeds. An application that collects only the file collects the slower of the two.

A fallback is part of the design and not an exception to it. Some applicants will not connect an account, some accounts will not connect, and a flow that treats a document upload as failure loses the people it was built for, a pattern catalogued in the field guide to intake abandonment. Anyone who has worked a stips queue knows which files go quiet: the ones where the next step needed something the applicant did not have in the room.

Underwriting the credit-invisible is mostly a collection problem wearing a modelling costume. The signals that work are the ones that arrive complete, and the people who need them can least afford three days assembling paper.

What we couldn't verify

Canada does not publish a bureau-derived count of credit-invisible people. The only Canadian estimate is Statistics Canada's survey proxy on 2016 and 2019 data, and Statistics Canada released a 2023 Survey of Financial Security on 29 October 2024 without repeating the credit visibility analysis on it. The circulating Canadian number describes a country that has since added millions of people.

Canada publishes no age breakdown of credit invisibility, no count of how many workers are paid in cash, and no study of predictive lift from any alternative data source on a Canadian lending population. Every effect size here involving repayment or approvals is American, and no Canadian rent reporting programme has published an outcome.

No published source, Canadian or otherwise, measures how many applications are declined for insufficient credit history rather than for history read and found wanting. Lenders record the decline. The reason sits in a field nobody aggregates.

Common questions

How many people in Canada are credit-invisible?
Statistics Canada estimated about 1.1 million economic families, or 7.2%, in a study published 27 September 2023 and corrected 29 February 2024, using pooled 2016 and 2019 Survey of Financial Security data. It is a survey proxy based on families reporting no credit products, not a count taken from credit bureau files.

Why do newcomers to Canada have thin credit files?
Credit history does not travel across borders, so a file starts at zero on arrival. Statistics Canada found 14.8% of immigrant families in Canada under two years credit invisible, falling to 6.1% at two to four years, and 12.3% of newly landed families holding a mortgage against 54.6% of Canadian-born families.

What is thin file underwriting?
Thin file underwriting is deciding on an applicant whose credit bureau record is absent or too sparse to generate a score, using evidence collected directly from the applicant. The usual inputs are bank transaction history, payroll data and rent payment records, consented to by the applicant at the point of application.

Does rent payment history help applicants without credit scores?
It builds files. The Urban Institute's randomised trial of 269 renters, published June 2025, cut the share of the treatment group with no credit score from 16% to 8% while the control group went from 23% to 25%. The average score effect of 7 points was not statistically significant.

Has anyone measured cash-flow underwriting on people with no credit file?
Only partially. FinRegLab reported in July 2019 that cash-flow metrics standing alone generally performed as well as traditional scores where credit history is unavailable or unreliable. Its July 2025 study disclosed that just 2,588 of 750,266 observations, 0.34%, lacked a conventional score.

---

Carousel's intake collects bank, payroll and income evidence directly from the applicant, which is where a thin file's information actually lives. See how verification fits your flow