A 9% Mortgage Does Not Lose to 10% Stocks

There is a rule of thumb that circulates whenever someone asks whether to pay down a mortgage or invest the money. It goes: stocks have averaged about 10% a year, so any debt costing less than 10% should be carried rather than repaid. Personal Finance Club states it about as cleanly as anyone has:

“Mathematically speaking, if your interest rate is below 10% and assuming that investing will get an average return of 10%, you’re better off investing.”¹

The 10% is right. The trouble is the word “mathematically,” and you can see why using the same index, over the same century, that produces the 10%.

The claim, with its own caveats

It is worth quoting the surrounding paragraph, because the post is more careful than the sentence that gets extracted from it. Immediately after the claim it adds that “the market is not consistent and it could give you a negative return for multiple years in a row,” and it closes by calling the choice “a decision based on personal preference.”¹ Earlier in the same post it concedes the key economic point in the other direction, that paying off high-rate debt is “essentially giving yourself a guaranteed return on your money.”

The post plainly understands that risk exists. The trouble is one sentence saying “mathematically” sitting beside sentences that concede the maths does not work that way, and the extracted sentence is the one that travels.

One attribution note, since fact-checks should get this right: that post is signed by two of the site’s writers rather than by Jeremy Schneider, who founded Personal Finance Club. The claim belongs to the publication.

Why an average is not a promise

Paying down a loan returns the loan’s rate, untaxed, with certainty, and that is an after-tax return whenever the interest is not being deducted, which for most households since 2018 means always. Buying an index fund returns a distribution whose mean has historically been around 10% nominal and whose annual standard deviation has been about 19%. Comparing the mean of the second to the certainty of the first leaves out the entire reason anyone is paid an equity premium in the first place. We work that framework through in Debt Is a Time Machine and The Risk-Free Rate Is Your Hurdle Rate, so what follows here is the measurement rather than the theory.

The academic treatment builds the risk match into its setup rather than arguing about it. Amromin, Huang and Sialm, examining whether households should prepay a mortgage or contribute to a tax-deferred account, are explicit that the comparison runs against “an asset with similar risk properties.”² Their arbitrage result is real, and it is constructed over risk-matched assets. Swap in equities and it is no longer the same claim.

The test, on 98 years of returns

Take the annual total returns of the S&P 500 from 1928 to 2025. The geometric average is 10.02%, which is where the 10% comes from. Now ask a different question: across every overlapping holding period of a given length, how often did buying and holding actually beat a guaranteed return?

Our calculation on Aswath Damodaran’s annual S&P 500 total return series (NYU Stern), reproducible with the script linked in the Sources. Windows overlap, so they are not independent observations; the pre-1957 index is a reconstruction; taxes, the mortgage interest deduction and the fact that repayment happens on a schedule are not modelled. The equity side is before tax and the guaranteed side is untaxed, so the omission of tax flatters equities.

Against a guaranteed 9%, stocks won 50 of 89 ten-year windows and lost 39. Over twenty years they lost 26 of 79. Over thirty years they lost 3 of 69, all of them windows starting in 1928 or 1929. A strategy that fails a third of the time over two decades is a reasonable bet rather than a mathematical result, and the difference matters most to the person who lands in the third.

The shape of the row is worth a sentence, because the ten-year column is the weakest of the four and it is also the one most people would quote. Five-year windows are lifted by the fat right tail of short-run equity returns; ten-year windows are where the lost decades bite, 1929 to 1939 and 1999 to 2009 among them; by thirty years compounding has smoothed almost everything out. A thirty-year mortgage that is actually held for thirty years, never refinanced and never prepaid, is close to the best case for the rule of thumb. Very few are held that long.

Which average belongs in a compounding calculator

There is a second arithmetic problem hiding underneath. The same 98 years produce an arithmetic mean of 11.85% and a geometric mean of 10.02%. The geometric figure is the one that describes what a buy-and-hold investor actually compounded at; the arithmetic figure is higher because volatility drags terminal wealth below the average of annual returns. Anyone who takes an arithmetic average from a table and puts it into a compounding calculator overstates the result, and the gap widens with volatility. Our house planning number, and why it is lower than either figure, is in What Real Return Should You Assume for Stocks?

Where the rule is right

Read the top row again. At a 3% mortgage, investing beat the guaranteed alternative in 90% of ten-year windows and 99% of twenty-year windows. For the millions of households who refinanced into a 2% or 3% loan, the historical case for investing instead of prepaying is close to overwhelming, and a reader who followed the rule of thumb got the right answer. A fixed-rate loan at those rates is also a short position in a long nominal bond, which is the inflation-hedge argument made in the mortgage-as-inflation-hedge guide; prepaying closes it.

The rule breaks where it matters most, which is at high rates. That is also where people are most anxious and most likely to go looking for a rule. Somebody carrying 9% debt is being told by the version of this claim that circulates that the maths favours investing, when the historical record puts it near a coin flip over a decade.

Run it with your own rate

This calculator models the investing side as a distribution rather than a point estimate, which is the whole disagreement. It starts at a 9% debt rate against a 10% expected return. Its win rate comes out lower than the 56% in the table because it models the decision as a household actually faces it: monthly payments against an amortizing loan rather than a single lump sum, and a 15% tax on the investment gains at the horizon, which the untaxed payoff never owes. Toggle the deduction on and the payoff return falls, so investing wins more often.

What we recommend

Compare the debt rate to a return you could get with matching certainty, which in practice means a Treasury or a high-yield savings rate rather than the stock market. If the after-tax debt rate is above that, repaying is a good risk-free trade. If it is far below, the historical case for investing instead is strong and the table above shows how strong. One more asymmetry favours the prepayer at high rates: a 9% mortgage can be refinanced when rates fall, so the “guaranteed 9%” is available for a shorter horizon than the loan term, and the thirty-year column overstates how long anyone actually holds it.

In the middle, which for most of the last two years has meant mortgage rates around 6% to 7%, treat it as a genuine judgment call rather than a calculation. The record says investing wins most of the time at those rates and not all of the time, and how much you care about the difference depends on your job security, your other liquid assets, and how you would feel in the windows where it did not work. There is no version of this where a single threshold number settles it.

Frequently asked questions

Should I pay off my mortgage or invest?

Compare your after-tax mortgage rate to a safe yield, not to the stock market’s average. Above that safe yield, prepayment is a guaranteed return that is hard to beat on a risk-adjusted basis. Well below it, the historical case for investing is strong. Liquidity matters too: money in a brokerage account can be reached in an emergency and home equity generally cannot.

Is the 10% stock market average wrong?

The number is broadly right as a nominal, US-only, geometric figure over a long history. It is the wrong input for planning, because it is nominal rather than real, US-specific rather than global, and backward-looking. A lower real number is the appropriate planning baseline.

Does the mortgage interest deduction change this?

For most households now, very little, because the standard deduction means they do not itemize. If you do itemize, use your after-tax mortgage rate on the debt side of the comparison, which lowers the hurdle.

What about carrying debt to invest more generally?

The same logic applies with more force, since the borrowing is usually at a higher rate and the borrowing magnifies the bad windows. That case is in Debt Is a Time Machine.

Key takeaways

• A certain return and an expected return are different goods. Comparing the mean of one to the certainty of the other omits the reason the equity premium is paid.
• Over 1928 to 2025, the S&P 500 beat a guaranteed 9% in 56% of ten-year, 67% of twenty-year and 96% of thirty-year overlapping windows. These are outputs of one historical test with overlapping windows rather than a probability forecast, and the thirty-year figure assumes a loan nobody refinances or prepays.
• At low rates the rule of thumb works. Against a guaranteed 3%, stocks won 90% of ten-year windows. The advice is usually right even where the reasoning is not.
• The threshold that matters is a safe yield. A Treasury or savings rate is a much lower bar than the 10% stock average, and it moves with rates.
• Quote the geometric mean. The same 98 years give 10.02% geometric and 11.85% arithmetic, and the higher figure overstates compounding.

Related guides

• Debt Is a Time Machine the guaranteed-versus-expected framework in full, with a crossover calculator.
• Do You Need a Paid-Off Home to Retire? the same question at the point where liquidity starts to dominate.
• What Real Return Should You Assume for Stocks? why our planning baseline is well below 10%.
• Good Heuristics, Bad Theorems the series this belongs to, and the scale we grade claims on; this is the one claim on the scorecard that fails on its own arithmetic.

Author disclosure

Summitward publishes a debt payoff planner, so we have an interest in readers treating this as a decision worth modelling. We have tried to give the rule of thumb its strongest case, which is the 3% row, as prominently as its weakest. Nothing here is personalized financial advice.

Sources

1. Personal Finance Club, “Pay down high interest debt before you start investing,” July 13, 2023, signed by Vivi and Shane. Read September 14, 2026; the page shows no later update. personalfinanceclub.com
2. Amromin, G., Huang, J. and Sialm, C. (2007). “The tradeoff between mortgage prepayments and tax-deferred retirement savings.” Journal of Public Economics 91, 2014–2040. The arbitrage result is constructed over assets with similar risk properties, which is the point relied on here. mccombs.utexas.edu
3. Damodaran, A., “Historical Returns on Stocks, Bonds and Bills: 1928–2025,” NYU Stern School of Business. Annual S&P 500 total returns including dividends, nominal. stern.nyu.edu. The window test and the data file it reads are published at github.com/engineerinvestor/summitward-research.

Editor’s note

Educational content, not financial or tax advice. The percentages in the table are historical frequencies over overlapping windows in a single market, which makes them a description of the US record rather than a probability of any future outcome. They ignore taxes, fees, the mortgage interest deduction, and the schedule on which debt is actually repaid, and they compare a lump sum against a lump sum. Quotations from personalfinanceclub.com were read on September 14, 2026. Talk to a professional about your own situation.