Tax Loss Harvesting: Turn Market Losses Into a Tax Asset

An investment that fell from $50,000 to $35,000 gives you nothing on your tax return while you hold it. Sell it and the $15,000 loss becomes real: a deduction that offsets other gains this year or in any future year. The insight that makes harvesting a strategy instead of surrender is that selling does not mean leaving the market. You immediately buy a similar (not identical) investment, so a rebound benefits you just the same, and the tax loss is banked either way.

Two features of the waterfall deserve emphasis. The $3,000 ordinary income offset ($1,500 if married filing separately) is per year, not per loss, and it has not been raised in decades, so a large harvested loss is mostly a gain-offsetting asset, not an income deduction. And the carryforward is indefinite: a $100,000 loss banked in a bad market can sit quietly for years and then erase the gain on a rental property or business sale. Unused losses do die with you, which is a real planning point for older investors sitting on big carryforwards.

Losses do not hit gains at random. Short-term losses first net against short-term gains (taxed at ordinary rates, up to 37% federal plus the 3.8% net investment income tax). Long-term losses first net against long-term gains (taxed at 0%, 15%, or 20% plus NIIT). Only after each pile nets internally do the leftovers cross over to the other side.

The order creates a value hierarchy. A short-term loss that lands on short-term gains saves tax at up to 40.8% for a top-bracket investor; the same dollar of loss absorbed by long-term gains saves at most 23.8%. Practical translations: positions held under a year are the most valuable harvests, and an investor with both kinds of gains should realize the loss whose character matches the expensive gain.

The bottom row is the one the robo-advisor marketing skips, and it gets its own section below. The current 0%, 15%, and 20% bracket breakpoints live on our capital gains tax guide.

That last paragraph is the honest frame for the whole strategy: harvesting is mostly a tax deferral machine with a genuinely free kicker (the rate arbitrage between short-term offsets and long-term payback, and the $3,000 ordinary offset). The deferral becomes permanent forgiveness for shares that get a basis step-up at death or go to charity. For everyone else it is an interest-free loan from the IRS, which is still worth taking.

Buy the same or a substantially identical security within 30 days before or after your loss sale and the loss is disallowed. The window is 61 days counting the sale date, and the "before" half surprises people: shares bought on December 10 can wash a loss you take on December 20, including shares bought automatically by dividend reinvestment or a 401(k) contribution schedule.

A disallowed loss in a taxable account is deferred, not destroyed: it adds to the basis of the replacement shares and comes back when you sell them. The permanent damage happens across account lines. Under Rev. Rul. 2008-5, repurchasing in your IRA within the window disallows the loss with no basis adjustment anywhere: the deduction is simply gone forever. The rule also reaches your spouse's accounts and accounts you control.

One asset class currently sits outside the rule entirely: cryptocurrency, which the IRS classifies as property rather than a security. Crypto losses can be harvested with an immediate rebuy under current law, an asymmetry Congress keeps proposing to close. The full mechanics, the ETF boundary cases, and the 1099-DA reporting angle are on our crypto wash sale rules guide; this page stays in securities land.

The whole harvest depends on the replacement being different enough. The IRS has never drawn a bright line around "substantially identical," but the working consensus is clear at the edges. Rebuying the identical fund, another share class of the same fund, or an ETF that tracks the exact same index as the fund you sold: asking for trouble. Swapping into a fund from a different sponsor that tracks a different index in the same asset class, a total-market fund for a 500-stock large-cap fund, for example: standard practice with near-identical market exposure.

For individual stocks, the clean answers are an industry ETF (sell the bank stock, hold a financials fund for 31 days) or simply waiting out the window in cash for that slice. One stock is never substantially identical to a different company's stock, so rotating within a sector is also fine. After 31 days you can return to the original position if you want it, at the cost of a second round of transactions and any price movement in between.

Direct indexing industrializes all of this: own the index's individual stocks and a manager (or software) harvests losers stock by stock all year, swapping into correlated names to maintain exposure. In the first years it can produce meaningfully more harvested losses than fund-level harvesting, which is why it has become the default pitch to high earners with large taxable accounts. The trade-offs are fees, hundreds of tax lots, tracking error, and the fact that the harvest naturally dries up as the portfolio appreciates. It earns its fee most clearly for investors with big recurring gains to offset or concentrated stock to diversify gradually.

Tech employees paid in RSUs are structurally perfect harvesting candidates: shares arrive every vest at a fresh basis, in a volatile stock, on top of a high salary. Any dip below a recent vest price creates a harvestable loss, and any sale of appreciated older lots creates gains that are often short-term. Losses that offset those short-term gains save tax at 35% or 37% rates, not 15%.

The vest-and-sell math, withholding shortfalls, and lot selection are their own subject: run your numbers through our RSU tax calculator. And for Californians, harvesting is worth more than the federal math alone suggests, since the state taxes capital gains as ordinary income at rates up to 13.3%; see the California capital gains guide.

1. Your gains are already taxed at 0%.

If taxable income keeps your long-term gains inside the 0% bracket (up to $49,450 single, $98,900 married filing jointly for 2026), a harvested loss offsets a tax of zero and lowers your future basis for nothing. Retirees in gap years, sabbatical takers, and early-career investors are usually better off doing the opposite: harvesting gains at 0% to step their basis up, which no wash sale rule restricts.

2. You would repurchase at a much lower basis and higher future rate.

Harvesting trades today's deduction for a bigger gain later. That trade loses when today's benefit lands at 15% and the future sale will land at 23.8% plus a state tax you moved into. Deferral is usually worth it; deferral into a predictably higher rate is not. The break-even is your rate today versus your realistic rate at the eventual sale.

3. The friction eats the alpha.

Bid-ask spreads, market movement during a 31-day exile from a concentrated position, imperfect replacement funds, and your own December attention all cost something. Harvesting $2,000 of losses to save $300 while risking a tracking mismatch on a $500,000 portfolio is activity, not strategy. The move earns its keep on meaningful losses, in meaningful brackets, with clean replacements.

4. Mutual fund distributions are about to undo your work.

Buying a replacement mutual fund in late November can put you in line for its December capital gain distribution, taxable income you did not need to acquire. Check estimated distribution dates before choosing the replacement, or use an ETF wrapper, which rarely distributes gains.

5. You are harvesting to justify a bad portfolio.

The tax tail should not wag the investment dog. If the position deserves selling, sell it and take the loss gladly. If it deserves holding, harvest only when a genuinely equivalent replacement exists. Selling a good investment into a worse one to capture a 15% deduction is how people turn a tax strategy into an investment mistake that costs multiples of the savings.