What $121,281 of Harvested Losses Was Worth

In November 2014, Mr. Money Mustache moved his last $100,000 into Betterment and started publishing the results. The page is still live, still updating, and still one of the most widely read consumer arguments for automated tax-loss harvesting. Its figures, dated December 2025, report $121,281 of losses harvested and more than $48,000 of tax saved.

The mechanics in that post are correct, and he states the awkward part of them himself. The arithmetic that converts the first number into the second is where it comes apart, and it comes apart in a way that matters well beyond one blogger’s account, because every harvesting service reports that same headline the same way.

The short version

The $121,281 is a real number. Multiplying it by a 40% tax rate is the step that does not follow. Federal law caps the net capital losses that can reach ordinary income at $3,000 a year, so across the account’s eleven tax years at most $33,000 of that harvest could have been deducted at an ordinary rate. The other 73% had to meet capital gains, at capital-gains rates. Whatever was saved is also mostly borrowed, because harvesting cuts cost basis by the same amount and the tax returns when the shares are sold. Run his own figures through one model of that repayment and the outcome ranges from about $30,500 to about $128,700 on the exit alone. His $48,512 is neither the floor nor the ceiling. It is a quantity that cannot be computed from outside his tax returns at all.

The claim

The results page reports that Betterment “has harvested a surprising $121,281 in deductible ‘losses’ on an account with about $500k of taxable money in it.” He values that at a combined federal and state rate: a $121k deduction “has saved me over $48,000 in income taxes already, which I have used to buy still more investments.” Elsewhere on the page the rate is given as “a marginal tax rate of 40% (state+federal).”¹ The implied arithmetic is exact: $121,281 multiplied by 40% is $48,512.

From there the conclusion arrives quickly. If that $48,000 earns a conservative 7%, it produces about $3,300 a year, against a fee he puts at “0.25% of $460,000,” a little over $1,000. His verdict: “Betterment costs less than nothing to use due to TLH alone.”¹ The two balances quoted a paragraph apart, about $500,000 and $460,000, are his; we have used each where he used it.

What the post gets right

He is not confused about the mechanism, and a critique claiming otherwise would be attacking something he did not write. The same page states the cost-basis consequence plainly: “The bottom line is that you save on taxes today but end up with investments which have a lower cost basis. This means you’ll have more taxable gains when you eventually sell them.” He then names the condition that makes the trade a good one, which is selling later in retirement in a lower bracket. ¹ That is a sound description of where harvesting value comes from.

He also discloses that Betterment advertises on the site while stating he was not paid to write the post. ¹ Nothing below turns on that.

Where the accounting stops

Two sentences on the page pull against each other. One is the cost-basis caveat above. The other, a few paragraphs earlier, is that alongside the reinvested savings you “also get to keep the principal you saved from the loss harvesting.”¹ Both hold together only in the case where the shares are never sold, and the rest of the post assumes they will be.

Harvesting a loss lowers the replacement position’s basis by the amount harvested, so those same dollars reappear as taxable gain at the sale. The saving is genuine and worth having. Most of it is a loan whose term the investor controls, and the value is the growth on money not yet handed over, plus any gap between the rate deducted at and the rate repaid at. We worked that general case through in What “Tax Alpha” Actually Means, so this post stays with the specific claim.

The $3,000 ceiling

The 40% rate is the part that does not survive contact with the statute. Section 1211(b) lets capital losses offset capital gains without limit, and then lets net losses beyond that reduce ordinary income by at most $3,000 a year. That $3,000 carries no cost-of-living adjustment in the statute, and the IRS inflation adjustments for 2026 make no change to it, so it is still $3,000 today. ²

The account was funded in November 2014 and the figures run through December 2025. Even if every one of those eleven tax years pressed against the limit, at most $33,000 of the $121,281 could ever have reached ordinary income. The remaining $88,281, 73% of the harvest, had to meet capital gains, and the rate that applies is the rate on the gains it offsets.

What the loss offsets	Amount	Rate	Tax reduced
Ordinary income, at the $3,000 limit, 2015 to 2025	$33,000	40.0%	$13,200
The rest, against long-term capital gains	$88,281	23.8%	$21,011
The rest, against short-term capital gains	$88,281	40.0%	$35,312

Combining the first row with one of the other two puts the current-year benefit between about $34,200 and $48,512, and his published figure sits at the very top of that range. Landing there requires that essentially every dollar beyond the $3,000 annual limit offset short-term capital gains, which are the ones taxed at ordinary rates. He does mention having had stock and real estate gains available to absorb losses. ¹ Real estate held more than a year produces long-term gain, which falls on the middle row, at roughly $14,000 less.

None of this shows that his $48,000 is wrong. He has his returns and we do not. It shows that harvested losses multiplied by a marginal ordinary rate cannot be the method that produced it, and that the same harvest supports answers $14,000 apart depending on facts the post does not report.

Losses that find nothing to offset are not forfeited. They carry forward indefinitely for individuals. They are also worth nothing until something absorbs them, and they do not outlive the taxpayer: a carryover can be used on the final return, including a final joint return with a surviving spouse, and any remainder is gone.³

What survives the sale

Set the rate question aside and grant the $48,512 in full. The basis reduction still has to be repaid at the sale, and what is left over depends on facts the post does not contain. Compounding the saving at 5% a year and repaying the deferred tax on the full $121,281 of basis, the same harvest lands in very different places.

How it ends	What is left	Share of $48,512
Sells in 10 years, rate unchanged at 40%	$30,509	63%
Sells in 10 years at 23.8%	$50,157	103%
Sells in 20 years at 23.8%	$99,853	206%
Donates the shares, or the basis adjusts at death	$128,718	265%

Those four figures are outputs of one model with the assumptions stated above, applied to his reported harvest. They are not readings from his account, and changing the growth rate or the holding period moves all of them. What they show is the shape of the problem. On an identical $121,281 harvest the result moves by a factor of four on the exit alone. In the 20-year rows, $48,512 understates the outcome by half or more. In the first row it overstates it by a third. The losses-harvested number does not tell you which row you are in.

What Betterment says about this now

The 2014 post that launched the experiment cited a specific figure: automated harvesting “could improve the performance of a non-retirement account by about 0.77% annually, which is again several times the fee they charge.” The 0.77% was hyperlinked to Betterment’s own page. ⁴

Betterment no longer publishes that number, or any tax-alpha percentage. Its current tax-loss harvesting methodology, updated August 2026, opens the valuation question this way: tax-loss harvesting “is primarily a tax deferral strategy, and its benefit depends entirely on individual circumstances.” ⁵

The same page carries a worked example that arrives at the point directly. Buy an asset for $100. It falls to $90, so you harvest the $10 and buy a similar asset at $90. Twenty years later the replacement is worth $500 and is liquidated, realizing $410 of gain against its $90 basis. Had the harvest never happened, the gain would have been $400.⁵ The $10 came back.

Betterment also publishes the profiles it considers poorly suited to the service, and the list includes “those in relatively low income tax brackets, and especially those who expect to be subject to higher tax rates in the future,” along with anyone planning a large near-term withdrawal and anyone whose household trades the same ETFs in outside accounts. ⁶

The fee in the post has also moved. Betterment’s current pricing is $5 a month below $24,000, with the 0.25% rate applying at $24,000 or above or with $200 a month of recurring deposits, and reduced tiers above $1 million. ⁷

The counterargument arrived before the experiment did

In August 2014, three months before that first post, Michael Edesess published a critique of robo-advisor tax-alpha marketing in Advisor Perspectives that named the 0.77% figure directly. Running lifetime Monte Carlo simulations with the $3,000 annual limit enforced and losses carried forward, he found the advantage of daily loss monitoring over annual rebalancing to be 14 basis points a year in one setup and 17 in another, against the 77 basis points then advertised.⁸

His diagnosis of the gap describes the same accounting step at issue here. Writing about Wealthfront’s comparable figure, he said its measure of tax alpha “is incomplete, because it takes credit for the losses harvested but conveniently does not debit its tax alpha for the compensating gains that must inevitably be realized eventually (prior to death), as financial advisor Michael Kitces has pointed out.” ⁸

Betterment answered the following month, and the response deserves to be read next to the critique. Daniel Egan and Boris Khentov argued that Edesess had modeled a less sophisticated algorithm than the one Betterment ran and had diverged from their assumptions without explaining why his were preferable, and they pointed to peer-reviewed estimates above his: Arnott, Berkin and Ye at about 50 basis points post-liquidation over 25 years, Smith and Smith at 37 over 40 years. They also conceded the general point about precision. Any such analysis, they wrote, is “hugely dependent on a sizable list of assumptions,” and “varying just one of these assumptions could dramatically alter the estimate.” ⁹

Edesess went further than we would, closing with an argument that harvesting is rent-seeking that works around the purpose of the wash-sale rule. Selling a real position at a real loss and buying a different fund is ordinary use of the code, and we have no quarrel with it. His quantitative point stands without that one.

The harvest decays and the fee does not

One word on the results page carries a great deal of weight. Having computed that the reinvested savings cover the fee, he adds: “Forever.” ¹

Harvesting opportunities shrink as a portfolio appreciates, because a position that has doubled has no loss left to take. Liberman, Krasner, Sosner and Freitas measured this across direct-indexing strategies and found that net realized losses “taper off within the first few years after their inception” and reach a maximum cumulative level of roughly 30% of the capital initially invested. ¹⁰

His $121,281 is an eleven-year cumulative total, and the drawdowns that generated most of it are dated: 2015, 2018, 2020, 2022. A 0.25% fee has no such profile. At the $460,000 balance he cites it is about $1,150 a year, and it scales with the account as the account grows.

The chart does not isolate the fee

The graph on the results page carries four lines: the real Betterment account, his own estimate of the harvesting benefit, an all-US VTI line, and an everything-except-the-US VXUS line. The Betterment account is 90% stocks and 10% bonds, holding both US and international assets.¹

That comparison is informative about markets and quiet about Betterment. A 90/10 global portfolio landing between a 100% US line and a 100% ex-US line is what any 90/10 global portfolio does over a stretch when US stocks beat international ones. The allocation explains the spacing before the fee or the harvesting has a chance to.

An experiment that isolated the service would hold the allocation, deposits, dividend reinvestment, rebalancing rule and tax rates constant, then vary one thing: no harvesting, occasional harvesting done by hand during ordinary portfolio maintenance, and the automated version. The distance between the second and third is what the fee buys. On that question, Khang, Paradise and Dickson concluded that investors with low expected harvesting alpha “will likely prefer to harvest losses as and when they arise, rather than pay fees for a full TLH program that incorporates transaction costs, commission costs, and regulatory compliance costs.” ¹¹

Price your own harvest

The inputs that decide this are the rate the loss offsets today, the rate you expect when you sell, how long you hold, and whether you ever sell at all. Move them and watch the answer swing.

What we recommend

Harvest losses when they show up in a taxable account and a good replacement is easy to find. The saving is real, the deferral has value, and at fund level the whole thing costs a few minutes. We would not pay a percentage of assets for it without first working out what it is worth in your own brackets, and we would stop treating the losses-harvested total as a measure of success. It records how much trading happened. What the trading was worth depends on the three facts below.

Three facts decide the outcome, and none of them appear on a year-end harvesting summary. What rate the loss offsets this year, what rate you expect to pay when you sell, and how long you hold. If your gains are long-term and your future rate resembles your current one, the benefit is the growth on a deferred bill, which is modest over short horizons. If you deduct against short-term gains now and sell in a low-income retirement year, it is substantial. If you expect to donate the shares or hold them until the basis adjusts at death, the bill may never arrive.

On Mr. Money Mustache specifically, he may well be ahead. A long holding period and a lower retirement rate are exactly the conditions under which harvesting pays, and his plan has both. What the post does not support is the arithmetic that reached $48,000, and the conclusion that the service therefore costs less than nothing.

Frequently asked questions

Does tax-loss harvesting save tax or delay it?

Mostly delay, with two exceptions. Selling at a loss and buying a similar replacement cuts your basis by the amount harvested, so the tax returns at the sale. What you keep is the growth on money you have not paid yet, plus any difference between the rate you deducted at and the rate you eventually pay. The exceptions are donating the appreciated shares and holding them until the basis adjusts at death, in which cases the deferred bill is never presented.

Why can I not just multiply my harvested losses by my tax rate?

Because the rate that applies depends on what the losses offset. Capital losses net against capital gains first, by character, and only the net loss left over reduces ordinary income, at up to $3,000 a year. A harvest that meets long-term gains is worth a long-term rate. A harvest with nothing to offset is worth nothing this year and carries forward. Your marginal ordinary rate applies to a smaller slice than the headline implies.

What happens to unused capital loss carryforwards when someone dies?

They can be used on the decedent’s final income tax return, including a final joint return, subject to the usual limits. Anything left over cannot be claimed by the estate, the heirs, or a surviving spouse on later returns. ³ Banking losses indefinitely with no plan to use them is therefore not a free option.

Where does the wash-sale rule genuinely destroy a loss?

Usually it does not. A disallowed wash-sale loss is added to the basis of the replacement security, so it is deferred rather than lost, and the window runs 30 days before and 30 days after the sale. The real trap is buying the replacement inside an IRA or other retirement account, where the loss is permanently disallowed with no basis adjustment anywhere.¹² That includes purchases in a spouse’s accounts.

Is Betterment a bad deal, then?

That is a different question, and this post does not answer it. The service does things beyond harvesting, including automated rebalancing and allocation, which some people value at more than 0.25% and others can replicate in an afternoon. What the results page does not demonstrate is that harvesting alone makes the fee free.

Key takeaways

The losses-harvested total measures activity. The $121,281 on the Mr. Money Mustache results page, dated December 2025, records what was traded. The price of that activity depends on what the losses met and on when the shares are sold.
The $3,000 annual limit caps how much of any harvest reaches ordinary income. Across the eleven tax years of that account, at most $33,000 of the harvest could have been deducted at an ordinary rate. The other 73% had to meet capital gains.
A loss harvested and a tax saved are separated by a repayment. Lower basis means a larger gain later. What you keep is the growth on the deferral plus any rate difference.
Under one model of his reported figures, the exit moves the answer by a factor of four. Selling in ten years at an unchanged rate leaves about $30,500 of his $48,512; never selling leaves about $128,700. Both come from the same harvest.
Betterment now describes the strategy as deferral. Its methodology page calls tax-loss harvesting primarily a tax deferral strategy and publishes no tax-alpha percentage, having retired the 0.77% figure the 2014 post cited.
Harvesting opportunities shrink as an account grows. Realized losses taper within the first few years in the published research, while a percentage fee rises with the balance.

Related guides

What “Tax Alpha” Actually Means, and Why 1% a Year Is the Wrong Number prices the same harvested loss under every exit, and explains where the headline percentages come from.
Tax-Loss Harvesting: How to Turn Portfolio Losses Into Tax Savings covers the mechanics, the netting order and the wash-sale rule.
What Wealthfront’s 9.8% Return Actually Measures takes apart the other number robo-advisors advertise.
There Are Only Four Things You Can Do With a Capital Gain measures what deferral is worth, and which exits reach zero.

Sources

Mr. Money Mustache, “The Betterment Experiment – Results.” Figures on the page are captioned as of December 2025; read September 14, 2026. mrmoneymustache.com
26 U.S.C. § 1211(b), which limits the deduction of net capital losses against other income to $3,000 ($1,500 for a married individual filing separately). The section contains no cost-of-living adjustment, and Rev. Proc. 2025-32, which sets the inflation-adjusted amounts for 2026, makes no adjustment to it. law.cornell.edu, Rev. Proc. 2025-32
Rev. Rul. 74-175, 1974-1 C.B. 52 (a capital loss carryover is personal to the taxpayer who sustained the loss), and IRS Publication 559, Survivors, Executors, and Administrators: capital losses and carryovers “can be deducted only on the decedent’s final income tax return,” and cannot be deducted on the estate’s return. irs.gov/publications/p559
Mr. Money Mustache, “Why I Put My Last $100,000 into Betterment,” November 4, 2014. The 0.77% figure is hyperlinked in the original to Betterment’s tax-loss harvesting page. mrmoneymustache.com
Betterment, “Tax loss harvesting methodology,” updated August 4, 2026. Source of the deferral characterization and the twenty-year $410 against $400 example. betterment.com
Betterment, “Tax Loss Harvesting+ Disclosure,” updated July 31, 2026. Source of the unsuitable-investor list. betterment.com/legal
Betterment pricing, read September 14, 2026. betterment.com/pricing
Edesess, M. (2014). “The Tax Harvesting Mirage.” Advisor Perspectives, August 12, 2014. Monte Carlo simulations over investor lifetimes, top-bracket California rates, with the $3,000 annual limit and carryforwards enforced. advisorperspectives.com
Egan, D. and Khentov, B. (2014). “The Tax Harvesting Oasis: A Response to Michael Edesess.” Advisor Perspectives, September 23, 2014. The studies cited in it are Arnott, Berkin and Ye, “Loss Harvesting: What’s It Worth to the Taxable Investor?” Journal of Wealth Management 3(4), 2001, and Smith and Smith, “Harvesting capital gains and losses,” Financial Services Review 17(4), 2008. advisorperspectives.com
Liberman, J., Krasner, S., Sosner, N., and Freitas, P. (2023). “Beyond Direct Indexing: Dynamic Direct Long-Short Investing.” Journal of Beta Investment Strategies 14(3), 10–41. The taper result is measured on direct-indexing strategies, which harvest at individual-stock level rather than fund level. ssrn.com
Khang, K., Paradise, T., and Dickson, J. M. (2021). “Tax-Loss Harvesting: An Individual Investor’s Perspective.” Financial Analysts Journal 77(4), 128–150. doi.org
Rev. Rul. 2008-5, 2008-3 I.R.B. 271: where a taxpayer sells shares at a loss and acquires substantially identical shares in an individual retirement account within the wash-sale window, the loss is disallowed and the IRA’s basis is not increased. Wash-sale mechanics generally, including the addition of a disallowed loss to the replacement’s basis, are at 26 U.S.C. § 1091 and in IRS Publication 550. irs.gov, law.cornell.edu

Editor’s note

Educational content, not tax advice. The figures attributed to Mr. Money Mustache and to Betterment were read from their own pages on September 14, 2026; his results page carries figures captioned as of December 2025, so it may have moved since. The tables in this post are outputs of the stated model applied to his published numbers, not readings from his tax returns, and they ignore state tax, transaction costs and the specific character of the gains his losses met. Talk to a tax professional about your own situation.