HELOC interest, and what deductibility turns on in Canada

Everything below is Canadian. The unqualified version of that question returns mostly American answers — US banks, US states, and a mortgage-interest deduction Canada has never had — so the rules here are cited to their sources: the Income Tax Act on the Justice Laws site, Supreme Court judgments on CanLII, and the CRA’s own published interpretation of both.

The four things paragraph 20(1)(c) asks

Paragraph 20(1)(c) of the Income Tax Act carries the general rule, and it asks four things at once:

1. An amount paid in the year, or payable in respect of the year — whichever matches the method you regularly follow in computing income.
2. A legal obligation to pay interest. A borrower-lender relationship has to exist, and the liability has to be absolute rather than contingent on some future event.
3. Borrowed money used for the purpose of earning income from a business or property. Money borrowed to acquire property whose income would be exempt, or to buy a life insurance policy, is carved out by name.
4. A reasonable amount. The claim is the lesser of what was actually charged and what is reasonable, judged against prevailing market rates for debts of similar term and credit risk.

The first, second and fourth are settled by the loan agreement and the statement. The third decides real cases, and it is the only one a household can break years after the borrowing without touching the loan at all.

Use, not security: the direct-use test

The direct-use test is the Canadian rule that decides whether interest qualifies for deduction: what matters is the use the borrowed money was actually put to, not the security pledged or the account it passed through. The Supreme Court applied it in Singleton v. Canada and in Ludco Enterprises v. Canada.

That the security does not decide it is the point everyone finds counter-intuitive, and it cuts both ways: a line registered against a home is not disqualified by the house, and an unsecured loan is not qualified by the absence of one. In Bronfman Trust v. The Queen the Supreme Court put the burden where it has stayed since: “The text of the Act requires tracing the use of borrowed funds to a specific eligible use,” and “the onus is on the taxpayer to trace the borrowed funds to an identifiable use which triggers the deduction.”

Two later judgments sharpened it. Shell Canada Ltd. v. Canada asked whether “a direct link can be drawn between the borrowed money and an eligible use.” Singleton v. Canada held that a sequence of transactions is read transaction by transaction rather than collapsed into one: “it is an error to treat this as one simultaneous transaction. In order to give effect to the legal relationships, the transactions must be viewed independently.” A lawyer withdrew capital from his firm to buy a house and borrowed to replace the capital; the order of the steps was the case.

The companion judgment, Ludco Enterprises Ltd. v. Canada, answered what “income” means: “income generally, that is, an amount that would come into income for taxation purposes, not just net income,” and courts “should not be concerned with the sufficiency of the income expected or received.” A distribution that never covers the interest still meets the purpose test. An expectation of capital gains alone does not, because a capital gain is not income under that paragraph.

The CRA’s reading of all of it is Income Tax Folio S3-F6-C1, Interest Deductibility, and it is worth reading rather than reading about: ¶1.32 states the tracing onus, and ¶1.34 names the technique of paying eligible expenses from a segregated borrowing account as cash damming — the CRA’s own word for it, in an example of tracing done the easy way.

Current use, not first use

A qualifying use at the outset is not a qualifying use forever. Folio ¶1.35 puts it plainly: the relevant use is the current one, not the original, and the link has to be re-established whenever the property behind the borrowing changes.

Where one investment is sold and the whole of the proceeds buys another, the borrowing follows the replacement — the finding in Tennant v. M.N.R., and folio ¶1.36. Where the proceeds buy several replacements, the outstanding borrowing is allocated across them dollar for dollar, and pro rata where the replacements are worth less than the debt (¶1.38, ¶1.39).

Where the use disappears altogether, the disappearing-source rules in section 20.1 can carry the interest forward. The folio’s own example at ¶1.41: $100,000 is borrowed to buy an income-earning property, the property is sold for $60,000, and the $60,000 goes against the loan. If the conditions in section 20.1 are met, the remaining $40,000 is deemed to be used for the purpose of earning income, and the interest on it continues to be claimable even though the asset that justified it is gone.

Where a qualifying use ends

Four ordinary situations end a qualifying use, and none of them is exotic:

• Registered accounts. Subsection 18(11) prohibits a claim for interest on money borrowed to contribute to an RRSP, a TFSA, a RESP, an RDSP or a registered pension plan, by name and whatever the borrowing was secured against.
• A holding bought for the gain alone. Per Ludco and folio ¶1.27, the purpose has to be earning income; an expectation of capital appreciation on its own is a different thing.
• A return of capital. Money handed back to the investor is money no longer standing behind an income-producing use. Folio ¶1.40 works it through: where the returned amount is put to an ineligible use, the interest on the matching slice of the loan stops qualifying.
• Personal spending on the same facility. One convenient draw, and the balance carries two purposes at once — and a repayment cannot afterwards be aimed at the personal half (folio ¶1.43). What happens to a line that carries both is a page of its own, because it is the failure this segment meets most.

Interest you did not pay in cash is a different paragraph

This one is missed almost universally in what is written about the Smith Manoeuvre, and it is the reason every page on this site says to pay the line’s interest from cash.

Interest that is left to accumulate onto the balance becomes interest on interest, and folio ¶1.81 is unambiguous about where that lands: compound interest is claimable only under paragraph 20(1)(d), and only in the year it is actually paid. Accrual does not reach it. Folio ¶1.83 adds the harder case — accrued interest added to the principal of an existing loan is not a payment at all, and part of what the enlarged loan charges afterwards is compound interest carrying that same cash-basis restriction.

So a month of capitalised interest does two things: it moves an amount out of the paragraph the rest of the claim sits in, and it breaks the arithmetic identity between the balance and the sum of the draws that a tracing record depends on.

What the record actually has to show

Every rule above is a question of fact, and a fact is something you can still evidence in year fourteen. The onus in Bronfman Trust is not an onus to have been right; it is an onus to trace — to connect a specific borrowed dollar to a specific eligible use, with documents from the time.

That is a records problem, not a tax-planning problem, and it is the one a spreadsheet structurally cannot solve: a cell asserts what was typed into it, while a claim rests on the credit-line statement, the trade confirmation, and an account history with nothing in between that breaks the connection. What ACRU traces is that trail, as it happens, which is a description of record state rather than a characterisation of anyone’s tax position. It is also the record an accountant is handed at year end when the question finally gets asked out loud. If you want the projection side instead, the Smith Manoeuvre calculator prints the assumption set behind every figure it shows.