Chapter 5: Tax-loss harvesting without the foot-guns
In a rough October, Maya sold an international fund that was $14,000 underwater, moved the money into a similar fund the same afternoon, and stayed fully invested through the recovery. Total time spent: twenty minutes. Tax saved on her 2026 return: about $4,900. That move is tax-loss harvesting: selling an investment below your cost basis on purpose, so the paper loss becomes a real deduction, while replacement shares keep your portfolio intact. The strategy is genuinely good and genuinely booby-trapped, and this chapter covers both halves with equal care.
What a harvested loss is worth
A capital loss never reduces your tax bill directly. It works through a strict netting sequence, and the order decides the value, because different gains carry different rates.
- Short-term losses net against short-term gains first. These are the precious ones: short-term gains are taxed at full bracket rates, up to 40.8%.
- Long-term losses net against long-term gains. These offsets save tax at the gentler 0/15/20% rates.
- The survivors cross-net. A leftover loss on either side then absorbs whatever gains remain on the other.
- Up to $3,000 of any remaining loss deducts against ordinary income, wages included, each year.
- Everything past that carries forward indefinitely, repeating this sequence every year until it is used up.
The carryforward means a big loss is never wasted, just rationed. Harvest $40,000 in a crash with no gains to offset, and you deduct $3,000 a year against wages until future gains soak up the rest. One somber footnote: carryforwards die with you, which matters in the endgame section below.
The calculator above prices a harvest for your own situation: enter your gains, the loss, and your bracket, and it walks the netting order and shows the dollar value, this year and carried forward.
Maya's harvest, traced
Maya's 2026 income is $220,000 of salary plus $120,000 of vested RSUs, which puts her marginal rate at 35% and her MAGI far past the $200,000 line where the 3.8% NIIT starts. During the year she sold some vested RSU shares eight months after vest for a $9,000 short-term gain, and a rebalancing sale produced a $2,000 long-term gain. In October she sold the international fund she bought in February for a $14,000 short-term loss and immediately bought a different international index fund.
Run her loss through the sequence:
| Step | Arithmetic | Result |
|---|---|---|
| Short vs short | $9,000 gain minus $14,000 loss | $5,000 short-term loss left |
| Long vs long | $2,000 gain, no long-term losses | $2,000 long-term gain left |
| Cross-net | $2,000 gain minus $5,000 loss | $3,000 net loss left |
| Ordinary income | $3,000 allowed against wages | fully used |
| Carryforward | nothing remains | $0 |
Now price each line. The erased $9,000 short-term gain would have been taxed at 35% plus 3.8% NIIT, so 38.8%: $3,492 saved. The erased $2,000 long-term gain was headed for 15% plus 3.8%: $376. The $3,000 against salary saves 35%: $1,050. Total: $4,918, about 35 cents per dollar of loss, from a fund swap that left her portfolio's risk unchanged.
What the loss is honestly worth: deferral plus arbitrage
Harvesting is not free money, and the brochures that imply otherwise skip a step. When Maya sold at $46,000, her replacement shares start with a $46,000 basis instead of her original $60,000. If the new fund recovers and she eventually sells, her gain is $14,000 larger than it would have been. The harvest moved tax from this year to a future year. On its own that deferral is worth something, since money saved now compounds for years before the bill returns.
The bigger prize is rate arbitrage. Maya's loss offset income taxed at 38.8% and 35% today, while the repayment, that extra $14,000 of future gain, will likely be long-term and taxed at 23.8% or less. She saves at a high rate and repays at a low one, pocketing the spread on top of the deferral. And sometimes the repayment never comes due: shares donated to charity, shares held until the basis step-up at death, or shares sold in a future 0% year settle the deferred bill at a rate of zero.
The honest summary: a harvest is worth the most when your current rate is high and the loss can erase short-term gains. It is worth the least when your rate is low, which is the "when not to harvest" section below.
The wash-sale rule, precisely
The one rule that can vaporize all of this is the wash sale. Claim a loss, and if you buy the same or a substantially identical security within 30 days before or 30 days after the sale, the loss is disallowed. Count the window carefully: 30 days back, the sale day itself, and 30 days forward, a 61-day minefield. Sell on December 12 and the danger zone runs from November 12 through January 11. The "before" half surprises people: shares you bought in late November can disqualify a loss you take in mid-December, and an automatic dividend reinvestment landing inside the window washes a slice of the loss all by itself. Pause auto-reinvest in any position you might harvest.
Three details carry most of the pain.
Substantially identical is about the thing, not the wrapper. The same fund at a different broker, or a different share class of the same fund, is identical. A different fund tracking a meaningfully different index is not. The gray zone is two funds from different companies tracking the same index; the IRS has never ruled, and careful harvesters simply pick a different index.
The window spans every account you and your spouse touch. Selling in your brokerage account and rebuying in your spouse's, or in your 401(k) through a fund election, still washes the loss.
The IRA version is the one true disaster. In a normal wash sale the disallowed loss is not destroyed: it is added to the basis of the replacement shares, so you get it back when you finally sell them. But if the replacement purchase happens in your IRA or Roth IRA, the IRS has ruled there is no basis adjustment at all. The loss is not deferred. It is gone forever. A $14,000 loss erased by an IRA autobuy is a five-figure unforced error, which is the second reason to switch off automatic investments during harvest season.
The replacement playbook
The whole craft of harvesting is staying invested without buying back the identical thing. Sell and sit in cash for 31 days and you have replaced tax risk with market risk; the market's best days have a habit of landing in exactly such windows.
Swap pairs that keep your portfolio's shape
S&P 500 index fund out, total US market fund in. Developed-international fund out, broader all-world ex-US fund in. A single stock out, a diversified sector ETF in. Each pair moves together closely enough that 31 days inside the replacement costs little, and after the window closes you may switch back or simply keep the new fund. When in doubt, choose a different index, not just a different logo.
When not to harvest
A loss is a coupon whose value is your tax rate, so there are years when the coupon is worth almost nothing.
- In the 0% gains zone, harvest gains instead. Carlos and Elena, with $59,100 of 0% headroom in Chapter 4, would waste a harvested loss: it would offset gains the IRS was not going to tax anyway, and the $3,000 against ordinary income saves only 12 cents on the dollar in their bracket. Their move is the mirror image, realizing gains at 0%.
- In a temporarily low-bracket year, the arbitrage runs backward. Deduct a loss at 12% this year and repay it as a 15% gain later, and you have manufactured a small loss. Save the harvest for high-income years.
- Late in life, think about the step-up. Appreciated shares held to the end escape gains tax entirely, so a low-bracket retiree gains little by churning basis. The mirror trap matters more: basis steps down too, so a loss position held at death is destroyed forever. Elderly investors should generally realize their losses and sit on their gains.
- Tiny losses are not worth the bookkeeping. A few hundred dollars of loss buys little and adds wash-sale tracking across every account; most robo-advisors and brokers set minimums for a reason.
Harvest when your bracket is high and the loss can erase short-term gains; replace with something similar but not identical; and let nothing, including an automatic reinvestment or an IRA purchase, buy the sold security back within the 61-day window. In a 0% year, skip the harvest and realize gains instead.
Where people go wrong
- Rebuying the same fund inside 30 days because the market turned. The loss is disallowed and parked in the new basis, undoing the work.
- Forgetting the 30 days before the sale. Last month's purchases, including reinvested dividends, count against this month's harvest.
- Letting an IRA or a spouse's account wash the loss. Across accounts the rule still applies, and through an IRA the loss is permanently destroyed.
- Harvesting in a 0% gains year. The same trade that saves Maya $4,918 saves a retiree in the 0% zone nearly nothing.
- Selling at a loss without a replacement plan. The point is to keep the portfolio and capture the deduction; cash on the sidelines is a different, worse trade.
Key takeaways
- Losses follow a strict order: short against short, long against long, cross-net, then $3,000 a year against ordinary income, with the rest carried forward indefinitely.
- Maya's $14,000 harvest erased a $9,000 short-term gain at 38.8%, a $2,000 long-term gain at 18.8%, and $3,000 of salary at 35%: $4,918 saved.
- A harvest is a deferral (the replacement's basis drops by the loss) plus a rate arbitrage (deduct at 35–38.8% now, repay at 23.8% or less later, sometimes never).
- The wash-sale window is 61 days, covers spouses and IRAs, and a wash through an IRA destroys the loss forever.
- Skip harvesting in low brackets and the 0% gains zone, and remember that basis steps down at death: late in life, realize losses, keep gains.
Sources: IRS: Publication 550, Investment Income · IRS: 2026 Tax Inflation Adjustments · IRS: Publication 590-A · Finvest Personal Finance Guide