Pay Off the Mortgage or Invest Before Retirement? Paired Monte Carlo Results (2026)

Pay off the mortgage or keep investing before retirement?

Key numbers

10,000 quasi-Monte Carlo paths per arm, identical market draws for the two arms of every pair; JPMorgan LTCMA 2026 capital-market assumptions (as of 2025-11-01); all dollars in today’s dollars; pre-tax, no Social Security, no home value. “Success” means the portfolio never reached $0 before 95.

Run the $1M / $50,000 / 65-to-95 plan in the free planner →

Opens the Monte Carlo planner prefilled with the keep-the-mortgage plan — retire at 65, plan to 95, $1,000,000, $50,000 a year, 45/15/40 US equity / international / bonds. To add the payment, create a recurring expense life event of $19,610 a year from 65 to 80 and choose the no-inflation option under Inflation; for the pay-off arm set savings to $800,000 and leave the event out.

What is being compared

Both arms describe the same household on the day of retirement: $1,000,000 of savings, a $200,000 mortgage balance with 15 years of payments left, and $50,000 a year of spending in today’s dollars that has nothing to do with the house. Net worth is identical; the only decision is what to do with the loan.

The pay-it-off arm sends $200,000 to the lender at 65 and retires on $800,000. Spending is the base amount, constant in real terms and raised with 2.5% inflation each year, to 95.

The keep-the-mortgage arm retires on the full $1,000,000 and goes on paying the loan. The payment is the level monthly amount from the amortisation formula — $1,634.17 a month at 5.5%, $19,610 a year — charged for exactly 180 months and then gone. It is a nominal constant: unlike the base spending it does not rise with inflation, so in today’s dollars it shrinks from $19,610 in the first year to about $13,879 in the fifteenth. The engine expresses this directly, as a scheduled expense with no inflation adjustment, so no approximation is involved.

Both arms are run on the same 10,000 market paths, so every statistic below can be taken path by path. The headline for each cell is the paired difference: the keep arm’s wealth at 95 minus the pay-off arm’s on the same path, summarised by its median, its mean, its 10th–90th percentile band and the share of paths on which each arm ends higher. Paths on which both arms end at $0 count as ties. Because the persona’s $50,000 of base spending is already a 6.25% draw on the pay-off arm’s $800,000, many paths in that row are ties at $0, and the paired median can sit inside the tie mass; the share of paths each arm ends higher and the mean difference are the statistics that stay informative there, which is why the break-even below is defined on the share.

The loan

$200,000 balance, 15 years, level monthly payment P = L r / (1 − (1 + r)⁻ⁿ) with r = rate ÷ 12 and n = 180. The generator re-runs each loan month by month with that payment and requires the closing balance to be zero to the cent before anything is written.

Results: all rates and spending levels

Each row is one mortgage rate and one level of base spending; the persona row (5.5%, $50,000) is highlighted. Dollar columns are wealth at 95 in today’s dollars; the last column counts paths in which only one arm reached 95 with money left.

“Paired median Δ” is the median over 10,000 paths of (keep wealth − pay-off wealth) at 95; “Keep higher / pay-off higher” are the shares of those paths with a positive and a negative difference (the remainder are ties, almost all at $0). Base spending is the amount excluding the mortgage; the pay-off arm’s draw is that amount on $800,000, the keep arm’s is that amount plus the payment on $1,000,000 for 15 years.

The break-even return

Without dispersion the arithmetic is one line: paying off the loan earns its rate, guaranteed, so the two arms tie when the portfolio earns that same rate. With dispersion the tie moves, because the arm that keeps the mortgage is exposed to the sequence of returns while the payment is due and the arm that pays it off is not. To find the tie in the simulated world, every cell was re-run with all five of the roster’s expected returns shifted together by −4, −3, −2, −1, +1, +2, +3 percentage points (volatilities and correlations unchanged), the shift at which the share of paths on which each arm ends higher is equal was interpolated between grid points, and that estimate was tightened by 2 further engine runs at the interpolated shift. The table reports the portfolio’s weighted expected nominal return at that shift — 60/40 weights on the roster’s published expected returns, 6.04% unshifted — its real equivalent at 2.5% inflation, and the same crossing for the mean difference and for the success rate.

“Equal shares”: the portfolio expected return at which keeping the mortgage ends higher on exactly as many paths as paying it off. “Closest run” is the sampled shift nearest that tie and the difference between the two shares the engine reported there (a value of 0.00 pp would be an exact tie). A crossing “none in range” means that statistic kept one sign across the whole -4 to +3 pp grid, and the note in brackets says which arm it favoured. For the success-rate crossing the note says which arm has the higher success rate at returns above it. The closed-form reference is the mortgage rate itself.

Three things stand out. First, where a tie exists (2 of 9 cells: 5.5% / $30,000; 6.5% / $30,000) the equal-shares break-even sits −1.24 pp to −0.30 pp from the mortgage rate: the simulated tie is not the textbook tie. Second, in the other 7 cells (4.0% / $50,000; 5.5% / $50,000; 6.5% / $50,000; 4.0% / $40,000; 5.5% / $40,000; 6.5% / $40,000; 4.0% / $30,000) there is no tie inside the grid at all: keeping the loan ended higher on more paths than paying it off at every shift, even with all expected returns cut by 4 points. At those spending levels many paths run out in both arms whatever the return, so the paths that differ are the ones where the extra $200,000 of investable money let the keep arm last longer or finish higher; at $50,000 the pay-off arm’s problem is a plan that draws 6.25% of $800,000, not the mortgage. Third, the three statistics do not agree with each other. The mean difference favours keeping the loan across the whole grid in 7 cells and crosses in 2, and where both the mean and the equal-shares tie exist the mean crosses at the lower return in 2 of 2 cells, by 1.01 to 1.87 pp, because the keep arm’s wins are large and its losses bounded. The success rate runs the other way: paying the loan off has the higher success rate at every shift in 4 cells (6.5% / $40,000; 4.0% / $30,000; 5.5% / $30,000; 6.5% / $30,000), keeping it in 4 (4.0% / $50,000; 5.5% / $50,000; 6.5% / $50,000; 4.0% / $40,000), and the two cross in 1 (5.5% / $40,000 at 4.68%, pay-off higher above it). A household that cares about the average outcome, the typical outcome and the chance of running out gets three different answers to “what return do I need?” from the same simulation, and the answer for the chance of running out depends on how tight the plan already is.

The gap over time

The engine reports each path’s wealth year by year, so the paired difference can be followed through retirement. Until 80 the keep arm holds $200,000 more but pays $19,610 a year for it; after 80 the payment stops and the difference is whatever the extra capital has become. The table shows the median per-path gap (keep minus pay off, today’s dollars) at six ages and the share of paths on which each arm is ahead, for the persona’s spending at the three rates and for the 5.5% rate at the three spending levels.

Median of the per-path difference at each age, with the shares of paths on which the keep arm / the pay-off arm is ahead. The payment ends at 80. 10,000 paths per cell.

What the numbers say, in plain words

A guaranteed return against an expected one

Paying off a 5.5% loan is a 5.5% return with no spread: every dollar sent to the lender saves exactly that much interest, on every path. Keeping the loan and staying invested is a bet that the portfolio’s return beats that number. The roster’s 60/40 expected return is 6.04% nominal — +0.55 pp above the 5.5% rate, +2.05 pp above 4.0%, −0.45 pp above 6.5% — but a 60/40 portfolio delivers that with roughly ten points of annual volatility, and over fifteen years the realised return lands anywhere in a wide band around it. That is why, at 5.5% and $30,000 of base spending, the mean per-path difference is +$58,647 in favour of keeping the loan while it ends higher on 57.0% of paths and lower on 34.2%, and its success rate is −3.0 pp: the wins are big, the losses more frequent than a one-line expected-value calculation suggests, and the failures sit in the tail the mean does not see.

Sequence risk with a level payment

The payment is due whether the market cooperates or not. In the keep arm the retiree draws $69,610 in the first year, 6.96% of the portfolio, and a poor first decade is met with that draw every month; in the pay-off arm the draw is 6.25% of $800,000 with no deadline attached. So the paths on which keeping the mortgage loses are concentrated where returns arrive late, and they show up as failures when the plan had room to fail: at 5.5% and $30,000 of base spending keeping the loan ran out on 305 paths that paying it off survived, and reached 95 with money left on 1 paths that paying it off did not, a success-rate difference of −3.0 pp; at $40,000 keeping the loan ran out on 144 paths that paying it off survived, and reached 95 with money left on 55 paths that paying it off did not (−0.9 pp). At $50,000 the sign flips: keeping the loan ran out on 1 paths that paying it off survived, and reached 95 with money left on 381 paths that paying it off did not (+3.8 pp). The reason is that on a plan drawing 6.25% of $800,000 both arms run out on 66.6% of paths whatever the loan does, and on the paths that are decided one way or the other the arm with $200,000 more of investable money lasts longer — the mortgage’s cost is real but smaller than the value of the liquidity it leaves behind. The payment’s one kindness is that it is nominal: inflation erodes it, so by the fifteenth year it is about $13,879 in today’s dollars against $19,610 in the first, and after 80 it is gone.

Liquidity

The pay-off arm starts with $200,000 less in the portfolio and $200,000 more in the house, and the house cannot be spent. These runs treat that as it is: a path that runs out of savings is a failure even though the retiree still owns a home outright. The keep arm holds the same $200,000 as investable money, which in a bad year can be drawn on and in a good year keeps compounding; that is what its extra wins are made of. The comparison does not credit either arm with borrowing against the house later, and it does not charge the keep arm for the option of paying the loan down early if markets turn — both are outside the numbers.

Where the rate takes the answer

At 4.0% the loan costs less than the roster’s expected return by +2.05 pp; at 6.5% it costs more by +0.45 pp. For the persona’s $50,000 the rate moves the numbers without changing the answer: keeping the loan ended higher on 35.8%, 33.2% and 30.6% of paths and lower on 0.0%, 0.2% and 1.2%, with success-rate differences of +6.2 pp, +3.8 pp and +2.0 pp. With $30,000 of base spending the rate does change the answer: keeping the loan ended higher on 75.0% of paths at 4.0% (lower on 16.2%), 57.0% at 5.5% (lower on 34.2%) and 44.2% at 6.5% (lower on 46.9%), while its success rate trailed paying it off by −1.5 pp, −3.0 pp and −4.6 pp. The break-even table above puts a number on each cell.

What it means

• Compare the rate with the break-even, not with the expected return. Where a tie exists the simulated tie sits −1.24 pp to −0.30 pp from the mortgage rate depending on spending and rate; the roster’s published expected return is a different number from the return a household needs to be better off on more paths than not. At $50,000 of base spending there is no tie in range at all: the plan is too tight for either arm, and the arm with more investable money lasts longer.
• The spread matters as much as the mean. At 5.5% and $30,000 the keep arm is ahead on the mean (+$58,647), ahead on 57.0% of paths, behind on 34.2%, and −3.0 pp on success; the average outcome, the typical outcome and the chance of running out point in different directions.
• A level payment is a sequence-risk exposure — when the plan has room to fail. Its cost shows up as extra failures in bad early decades: 305 paths at $30,000 of base spending and 144 at $40,000, against 1 and 55 rescued. On the persona’s $50,000 plan the sign flips (1 lost, 381 rescued): when both arms fail on two thirds of paths, the arm with more investable money lasts longer.
• The house is not liquid. Paying the loan off moves $200,000 from money that can be spent to money that cannot; these runs count the consequence of that on every path, and nothing else about the house.
• Taxes could move the answer either way. No mortgage-interest deduction (which favours keeping the loan) and no tax on returns or withdrawals (which favours paying it off) are modelled; the sign of the net effect depends on the household.
• The rest of the plan matters as much as the rate. Across the three rates at $50,000 the success-rate difference (keep minus pay off) spans +2.0 pp to +6.2 pp; across the three spending levels at 5.5% it spans −3.0 pp to +3.8 pp, and the share of paths on which paying it off ends higher spans 0.2% to 34.2%.

CC-BY-4.0 — free for any use including republication and journalism, with attribution to QuantCalc Research.

Methodology

Assumptions and limitations

• No tax deduction, and no tax at all. Mortgage interest is not deducted, so a household that itemises would face a lower effective rate than the one modelled — that omission favours paying the loan off. Portfolio returns and withdrawals are not taxed either, which favours keeping it. Neither is modelled; the direction of the net bias depends on the household.
• No home value. The house is the same in both arms, so its price, its upkeep and any change in its value are left out; wealth here is the investment portfolio only. A retiree who pays off the loan still owns the home when the portfolio runs out, and these runs do not value that.
• One pre-tax pool. The simulate contract used here does not distinguish taxable, traditional and Roth money, so the $200,000 is taken from the pool at par with no capital-gains tax on the sale. The engine’s tax-aware mode does keep the three account types but chooses the source of every withdrawal itself, so a pay-from-taxable-only against pay-pro-rata comparison is not something the contract can express; it is not reported.
• A level nominal payment, exactly. The engine supports scheduled expenses with no inflation adjustment, so the payment is modelled as it is: constant in nominal dollars for 180 months. Escrowed taxes and insurance, which do rise, are part of housing costs and not of the loan; they are in neither arm.
• No refinancing, prepayment or borrowing. The keep arm holds the loan to its last payment on every path; the pay-off arm never borrows against the house. A rate that could be refinanced lower, or a plan that pays extra principal in good years, would sit between the two arms.
• Same spending in both arms. Both arms withdraw the same constant-real base amount every month; no spending rule reacts to the portfolio. The persona’s $50,000 is a 6.25% draw on the pay-off arm, which is why both arms in that row run out on many paths; the $40,000 and $30,000 rows show the same loan against more comfortable plans.
• Parametric returns. Returns are lognormal with the stated means, volatilities and correlations; no regime switching, fat tails or historical block resampling. Deeper or longer bear markets than the model draws would widen the gap between the arms in the tail.
• The break-even is a parallel shift. Every asset’s expected return moves by the same amount and nothing else changes; a world in which mortgage rates are 6.5% might also be one with different bond yields, which this does not model.
• Not advice. Educational research on the mechanics of one decision; not financial, tax or legal advice.

Frequently asked questions

Is it better to pay off the mortgage or keep investing before retirement?

It depends on the mortgage rate against the portfolio's return, and on how much dispersion that return carries. For a 65-year-old with $1,000,000 in a 60/40 plan, a $200,000 balance with 15 years left at 5.5% and $50,000 a year of other spending, paying the loan off at 65 reached 95 with money left in 29.6% of 10,000 simulated paths; keeping it and paying $1,634 a month reached 95 in 33.4%. On the same market paths keeping the mortgage ended higher in 33.2% of them and lower in 0.2%, with a per-path median difference of $0 and a mean of +$68,595. At 4.0% the same comparison favoured keeping the loan on 35.8% of paths; at 6.5% on 30.6%.

What return does the portfolio need for keeping the mortgage to break even?

Without dispersion the two arms tie when the portfolio earns the mortgage rate. With it, the tie moves. The study re-ran every cell with all five expected returns shifted together, from -4 to +3 percentage points, and found where the share of paths on which each arm ends higher is equal. At 5.5% and $50,000 of spending there is no such point inside the grid: both arms run out on most paths and keeping the loan ended higher on more of the rest at every shift. A tie exists in 2 of the 9 cells - 4.26% nominal (1.72% real at 2.5% inflation) at 5.5% / $30,000; 6.20% nominal (3.61% real at 2.5% inflation) at 6.5% / $30,000, against the 60/40 roster's 6.04% nominal as published. Across the 2 cells with a tie in range the break-even sits -1.24 pp to -0.30 pp away from the mortgage rate itself.

Why can keeping the mortgage lose on many paths even when the expected return beats the rate?

Because the mortgage payment is a level nominal claim in the first fifteen years, which is exactly when a bad sequence of returns does its damage, while the portfolio return that is supposed to cover it arrives with a wide spread. In the keep arm the retiree draws $69,610 in year one, 6.96% of the portfolio, against 6.25% in the pay-off arm; a poor first decade compounds that difference. The mortgage's guaranteed 5.5% has no spread at all. So the paths on which keeping the mortgage wins are the good-market paths, and the tail is where it loses: at 5.5% and $30,000 of base spending, keeping the loan ran out on 305 paths that paying it off survived, and reached 95 with money left on 1 paths that paying it off did not. When the plan is already so tight that both arms run out on most paths - the $50,000 persona - the picture inverts (keeping the loan ran out on 1 paths that paying it off survived, and reached 95 with money left on 381 paths that paying it off did not), because the arm holding $200,000 more of investable money lasts longer on the paths that are decided either way.

How much does the mortgage rate change the answer?

A lot. At $50,000 of base spending, the share of paths on which keeping the loan ends higher goes from 35.8% at 4.0% to 33.2% at 5.5% and 30.6% at 6.5%; the success-rate difference (keep minus pay off) goes from +6.2 pp to +3.8 pp to +2.0 pp. The monthly payment is $1,479, $1,634 and $1,742 respectively, for total interest of $66,288, $94,150 and $113,599 over the 15 years.

Does the answer depend on how much else the retiree spends?

Yes, through sequence risk. At 5.5%, keeping the mortgage ends higher on 33.2% of paths when base spending is $50,000, 52.7% at $40,000 and 57.0% at $30,000; the success rates are 29.6% vs 33.4%, 61.7% vs 60.8% and 91.1% vs 88.1% (pay off vs keep). The direction of the success-rate effect flips with the tightness of the plan: with $30,000 or $40,000 of base spending the level payment costs success (-3.0 pp and -0.9 pp), while a plan already drawing 6.25% of an $800,000 portfolio runs out on 66.6% of paths in both arms, and on the rest the extra $200,000 of investable money is worth more than the loan costs (+3.8 pp).

What is not in these numbers?

Taxes, on either side: there is no mortgage-interest deduction, which would lower the effective rate for a household that itemises and make keeping the loan look better than it does here, and no tax on the portfolio's returns or withdrawals, which would make keeping it look worse. The home itself is left out because it is the same in both arms; its value, its costs and any change in it cancel in the comparison. The payoff is taken from one pre-tax pool at par, so no capital-gains tax on the sale and no distinction between taxable and tax-deferred money. There is no refinancing, no prepayment and no home-equity borrowing in a bad year. Returns follow the JPMorgan LTCMA 2026 capital-market assumptions the site's planner uses, sampled with quasi-Monte Carlo, 10,000 paths per arm, identical draws for the two arms of every pair, and all dollars are in today's dollars.

Related research

Changelog

• v2026.1 (2026-09-07) — initial release. 9 cells (4.0% / 5.5% / 6.5% mortgage rates, $50,000 / $40,000 / $30,000 base spending) of paired pay-off vs keep arms; 10,000 paired paths per arm; per-path paired statistics at 95 and by year; break-even returns from a 7-point shift grid with 2 refinement steps.

Last updated 2026-09-07. Dataset license: CC-BY-4.0. QuantCalc is an independent retirement-planning research project. The JPMorgan LTCMA 2026 name identifies the published capital-market assumptions the engine uses; QuantCalc is not affiliated with, endorsed by, or sponsored by that firm, and all trademarks belong to their respective owners. Educational research, not financial, tax, or legal advice.

@misc{quantcalc2026payoffthemortgageorinvestbeforeretiremen,
  title  = {Pay Off the Mortgage or Invest Before Retirement? Paired Monte Carlo Results (2026)},
  author = {{QuantCalc Research}},
  year   = {2026},
  url    = {https://quantcalc.app/research/pay-off-mortgage-or-invest-2026/},
  note   = {Accessed <date>}
}