Tax-Loss Harvesting
Tax-loss harvesting sells a holding at a loss deliberately so the loss can offset realised gains and reduce a tax bill. It applies only in a taxable account, and buying back too soon can disallow the loss entirely under wash-sale rules.
Most of investing rewards doing nothing. This is one of the few operations where deliberately realising a loss makes you better off — and it comes with a rule strict enough that doing it carelessly achieves nothing at all.
How it works
A holding worth less than you paid carries an unrealised loss, which is worth nothing until you sell. Selling makes it a realised loss, and a realised loss can generally offset realised gains.
You then buy something similar but not identical, so the portfolio’s exposure is unchanged while the loss has been banked.
The replacement has a lower cost basis, which means a larger gain when it is eventually sold. The benefit is a deferral rather than a cancellation.
A worked example
Take a taxable account with a 10,000 realised gain from earlier in the year.
A holding bought at 40,000 is now worth 32,000 — an 8,000 unrealised loss.
Selling it offsets 8,000 of the gain, leaving 2,000 taxable instead of 10,000.
At a 20% rate that is 1,600 of tax deferred, and the money stays invested and compounds for as long as the deferral lasts.
The replacement holding now has a 32,000 basis, so a future sale at 50,000 produces an 18,000 gain rather than a 10,000 one. Nothing was forgiven; the bill moved.
Why deferral is worth something
1,600 kept for twenty years and compounding is worth substantially more than 1,600 paid today. That is the whole argument, and it is the same arithmetic that makes fees so damaging in reverse.
It is worth most when your marginal rate is high now and expected to be lower later, and least when the reverse is true — in which case harvesting can genuinely make you worse off.
On this site’s arithmetic a 75-basis-point annual drag removes 20.2% of a thirty-year pot. Tax
paid early behaves in a similar way: a smaller base compounding for the same period. The figures are
in research/series-measurements.json.
The rule that decides it
Wash-sale rules disallow a loss if you buy the same or a substantially identical security within a defined window around the sale. The window runs both before and after in several jurisdictions, which surprises people who only look forward.
A disallowed loss is not lost permanently — it is generally added to the basis of the replacement — but the deferral you were trying to create does not happen this year.
The operational risks are the ones nobody plans for: automatic dividend reinvestment inside the window, a purchase in a different account, or a spouse’s account counting as the same person. The wash-sale page covers the mechanics.
Staying invested
The failure that costs most is being out of the market while waiting. A broad fund can be replaced with a different broad fund tracking a different index, which keeps the exposure without being identical.
Sitting in cash for a month to be safe is a market-timing decision you did not intend to make, and on this site’s shared series 54% of 566 ten-bar windows ended higher than they began — so being out is not neutral.
Costs
Every harvest is a sale and a purchase. On this site’s shared series a round trip measures about 2% of the median bar range of 0.493, and it exceeds 10% of the bar on 15 of 576 bars.
A small loss is not worth the transaction and the record-keeping. The exercise has a fixed operational cost, so it makes sense on meaningful positions and not on trivial ones.
The original data
Of the 24,971 unique videos in research/search-study-corpus.jsonl, zero have a title about
tax-loss harvesting. Wash sales appear in 2 videos at a median of 90,257 views and capital gains tax
in 1 at 233,662. The counts come from site/rank_investing.py, which deduplicates by video id.
Zero videos on the technique and 2 on the rule that governs it, at 90,257 views. People search for the trap rather than the operation, which suggests most encounters with this subject start with a disallowed loss rather than with a plan.
The answer to the question on that chart depends on whether you have gains to offset and what rate they are taxed at. With no realised gains the loss offsets very little this year and carries forward instead. And if the position is one you would sell anyway, the tax question is secondary — the harvest is a reason to time a sale you were already going to make, not a reason to make one.
When it fails
The failure is a loss harvested carefully and then disallowed by something automatic. A dividend reinvestment inside the window, a scheduled contribution buying the same fund in another account, or a spouse holding the identical fund elsewhere — none of these are decisions anyone made that month, and any of them can void the whole exercise. The trades happened, the spread was paid, the records got more complicated, and the tax benefit is zero.
The second failure is harvesting into a lower future rate. Deferral can cost more than it saves.
A third is being out of the market during the switch. That is a timing bet.
A fourth is replacing with something substantially identical. It defeats the purpose.
A fifth is harvesting trivial amounts. The costs exceed the benefit.
And a sixth is doing it inside a wrapper. A loss there is worth nothing at all.
Related
Taxable accounts is the only place this works. The wash-sale rule is what decides whether the loss counts. And cost basis is the record the whole operation depends on.
The trap that catches people is the automatic dividend reinvestment they forgot was switched on. A reinvestment inside the window is a purchase, and it can disallow the loss on the sale you just made deliberately. Switching it off before harvesting is a five-second check that saves the whole exercise.
— Michael Whitman
This page is educational, not financial advice. Test every idea on your own charts before risking money.