Tax Loss Harvesting Rules: The Wash Sale Rule and the $3,000 Limit

Three tax loss harvesting rules decide whether a loss you booked actually counts. Under IRC Section 1091 you cannot buy the same or a substantially identical security within 30 days before or after the sale, which is a 61 day window in total. That window spans every account you and your spouse control, not just the one you sold from. And a harvested loss offsets capital gains without limit but only $3,000 of ordinary income a year, $1,500 if married filing separately. Break the first rule and the loss is deferred. Break it into an IRA and the loss is gone permanently.

Those rules exist independently of how you invest, but they bite hardest in a direct indexing account, because holding 200 to 500 individual positions means you are triggering the rules a few hundred times a year instead of twice. This article covers the rules themselves and the traps around them. If you want the economics instead, how much the strategy is really worth and why the benefit fades, that is on direct indexing tax loss harvesting. Educational content only, not investment or tax advice. Confirm anything here with your own CPA.

How does tax loss harvesting work?

The mechanism is simple enough to state in one sentence: you sell something that is worth less than you paid for it, which turns a paper loss into a realized loss the IRS will let you use, and then you buy something similar so you have not stepped out of the market. The complexity is entirely in the word "similar", which is where the wash sale rule lives.

Direct indexing exists because of a quirk in how funds work. An index can rise while a large share of its members fall, and in a year the S&P 500 returns 10% it is normal for well over a hundred of its constituents to end lower than they started. Hold one fund and that dispersion is invisible to you, because the winners and losers net out inside the wrapper. Hold the 500 positions directly and every one of those losers is a lot you can sell. What that is worth in practice is more modest than the marketing suggests: Wealthfront's own whitepaper, backtesting February 2015 to December 2025, estimates an annual after-tax benefit of 0.18% to 0.44% of account value on the US stocks portion.

How does tax loss harvesting work in direct indexing?

The mechanism depends on a fact people underrate: an index can go up while a large share of its members go down. In a year the S&P 500 returns 10%, it is normal for well over a hundred of its constituents to end the year lower than they started. If you hold one fund, that dispersion is invisible to you. If you hold 200 individual positions, every one of those losers is a lot you can sell.

The loop looks like this:

1. Scan for lots trading below cost basis. Managed providers do this daily. Doing it yourself, quarterly is realistic and captures most of the value.
2. Sell the loser and realize the loss. The loss is now on your tax return, not just on your screen.
3. Buy a replacement that is not substantially identical. Usually another company with similar sector and factor exposure, so your index still behaves like the index.
4. Wait out the 30-day window before buying the original back, if you want it back at all.

Step three is where the real skill sits. Sell one large-cap software name and buy another, and your tracking error against the index widens a little. Do that a hundred times and your "index" is drifting somewhere you did not intend. This is the tradeoff nobody advertises: every harvest buys you tax value and sells you a little bit of index fidelity.

What are the tax loss harvesting rules?

Three rules do most of the damage. They are not obscure, but they interact in ways that surprise people.

The wash sale rule. Under IRC Section 1091, if you buy the same or a substantially identical security within 30 days before or after selling at a loss, the loss is disallowed for now and added to the basis of the replacement shares. Note that it is a 61-day window in total, not 30, because it runs both directions from the sale.

It follows you across accounts, including your spouse's. The window is not per-account. If you sell a stock at a loss in your taxable brokerage account and your spouse's account buys it inside that window, the IRS treats it as a wash sale. This is the single most common way DIY harvesters lose deductions they thought they had banked, usually via an automatic reinvestment or a 401(k) contribution nobody was thinking about.

The IRA version is worse than a delay. Ordinarily a disallowed wash-sale loss is only deferred: it rides along in the basis of the replacement shares and you get it back eventually. Under Revenue Ruling 2008-5, if the replacement is bought inside an IRA, the loss is disallowed and your IRA basis is not increased. An IRA has no outside basis to step up, so the deferred loss has nowhere to go. It is gone permanently. The same applies to a purchase in your spouse's IRA.

That last one deserves emphasis because it inverts the usual intuition. A wash sale in a taxable account costs you timing. A wash sale into an IRA costs you the deduction, forever.

What is a harvested loss actually worth?

Losses are not cash. They offset gains, and their value depends entirely on what you have to offset.

The $3,000 cap is the detail that deflates a lot of enthusiasm. If you have no capital gains to offset, a $40,000 pile of harvested losses drips out at $3,000 a year against ordinary income and takes over a decade to use. Meanwhile you have been paying 0.40% a year on the account that generated them. Harvesting is worth real money to someone who realizes gains regularly. It is worth very little to a buy-and-hold investor with no gains, which describes more people than the marketing admits.

The other trap is decay. Harvesting is front-loaded: in year one, plenty of lots sit below cost. By year seven, if the market has done what it usually does, most of your positions are deep in the green and there is almost nothing left to harvest. The fee, of course, does not decay. Studies of tax alpha typically show the benefit concentrated in the early years and thinning steadily after that, which is why "1% to 2% a year" is best read as an average across a cycle rather than a rate you can count on annually.

Is direct indexing tax loss harvesting worth the fee?

Run the subtraction rather than trusting the pitch. Direct indexing costs 0.09% to 0.40% a year at the major providers versus roughly 0.03% for an S&P 500 ETF, so call the premium 0.06% to 0.37%.

Notice what the table says. At a 0.40% provider the harvest roughly pays the fee and no more, so the customization is what you are actually buying. At a 0.09% provider the math works comfortably. The provider you pick matters more than whether direct indexing is a good idea in the abstract, which is why the direct indexing fees and minimums are worth comparing line by line before you commit.

And the whole thing collapses to zero inside an IRA or 401(k). There are no annual taxable gains in a retirement wrapper, so there is nothing to harvest and nothing to offset. If most of your money is in retirement accounts, this technique is not for you, whatever the fee.

Can you do tax loss harvesting yourself?

Yes, and at a modest number of positions it is genuinely manageable. Quarterly, you review lots below basis, sell them, buy a sensible replacement, and write down the 30-day window so you do not trip it. Most brokers show unrealized gain and loss per lot, which is the only data you strictly need. At year end the realized losses appear on your 1099-B and flow onto Schedule D, and if your return is otherwise straightforward you can handle the filing without a CPA as long as your lot records are clean.

What makes it hard is not the selling. It is knowing what to sell and what to buy back so the basket still tracks the thing you wanted to own. That is a design question, and it is answerable before you risk a dollar: decide the universe, decide the weighting rule, decide what an acceptable replacement looks like, and test whether the design holds up. You can backtest the index you have designed against real market history first, then implement it wherever you custody your money.

The honest summary: harvesting is real, it is worth roughly 0.24% to 0.48% a year to someone with gains to offset, it decays over time, and it is worth nothing to a retirement-account investor. If you are in the group it suits, the arithmetic is straightforward and it favors the cheapest provider who will do it properly. If you are not, an index fund at 0.03% is the better product and you should stop reading about this.