Direct Indexing vs ETFs: Costs, Tax Losses, and Who It Fits

Direct indexing and an index ETF hold the same companies. The difference is who owns the individual stocks. In an ETF the fund owns them and you own a share of the fund. In direct indexing you own the stocks themselves, in a separately managed account, and a platform keeps the weights close to the index. That one difference is the whole case for direct indexing: losses on individual stocks become yours to harvest, which a fund cannot give you. It is also the whole case against it: more cost, more lots, more paperwork, and a harvest that shrinks as the account ages.

This is a description of how the two structures work and what the published numbers say, not tax or investment advice.

What direct indexing is

An S&P 500 ETF holds roughly 500 stocks at market-cap weights and issues shares against the basket. A direct index of the S&P 500 buys those stocks, or a sample of them, straight into your account. The platform rebalances toward the index, reinvests dividends, and on most platforms runs a tax-loss harvesting program across the individual positions. The output is an account with hundreds of positions and thousands of tax lots over time, all of which belong to you. That ownership creates both the tax opportunity and the administrative weight.

The tax-loss harvesting argument

A mutual fund or ETF nets its internal gains and losses. Net gains may be distributed to shareholders. Net losses cannot; they stay inside the fund, carried forward against the fund's own future gains. A shareholder in an S&P 500 ETF that rose 10 percent in a year has one position, up 10 percent, with no loss to take. The stocks inside the index that fell that year are invisible on the shareholder's return.

In a direct index those falling stocks are separate positions with their own cost basis. A losing position can be sold, the loss realized, and a different stock with similar characteristics bought in its place so the account keeps tracking the index. The tax-loss harvesting guide covers a single harvest. Direct indexing is that mechanic run continuously across hundreds of names, which is the thing the structure does that an ETF cannot.

What the harvested losses are worth

A harvested loss does not produce cash. It produces a deduction, and the deduction is worth the loss times the rate on whatever it offsets.

The order is fixed by statute. Under IRC section 1211(b), capital losses offset capital gains without limit. Whatever remains offsets up to $3,000 of ordinary income a year. Under section 1212(b) the rest carries forward indefinitely. The capital loss carryover post walks through the netting order and why the $3,000 matters less than it looks.

Three consequences follow.

The value depends on having gains. A harvested loss against a short-term gain taxed at 24 percent is worth 24 cents on the dollar. Against a long-term gain at 15 percent, 15 cents. Against nothing, it is worth $3,000 a year of ordinary income and a carryover that waits for a gain that may or may not arrive.

It is a deferral. Each harvest replaces a high-basis lot with a low-basis lot. The gain that was avoided today is embedded in the replacement and comes due when it is sold, unless the holder dies and the basis steps up, or sells in a year with a lower rate. The cost basis post covers how the reset basis is tracked.

The losses run out. This is the part vendor material tends to leave for the footnotes. Harvesting only works on lots that sit below their basis. In a market that goes up over time, lots bought early climb above basis and stay there. Every harvest resets a lot's basis lower, which makes it less likely to ever be at a loss again. After a few years without a drawdown, most of an account's lots are above water and the harvest slows to whatever new deposits and new dips provide. Practitioners call this ossification. Wealthfront's tax-loss harvesting white paper does not use the word, but it notes that harvesting "can add more value to clients who make regular deposits into their accounts compared to ones who make a single deposit," because deposits create new lots with fresh basis.

There is also tracking error. Every harvest sells one stock and buys a substitute, so the account drifts from the index. Platforms manage this within a band, but a direct index will not match the ETF's return exactly.

What the published harvest figures say

The most detailed public dataset comes from Wealthfront's white paper, which reports realized harvesting yields for its automated accounts from October 2012 through December 31, 2025. For the most common risk score, the cross-vintage average annualized yield is 4.00 percent of the portfolio, with individual vintages ranging from 2.00 percent (accounts opened in 2012) to 7.82 percent (accounts opened in 2025). The paper attributes the spread to the market: "there is a strong positive correlation between the annualized harvesting yields realized by the different cohorts and both the volatility and the maximum drawdown realized by the market over the corresponding time period."

Two caveats. That figure measures harvesting across ETFs in a diversified automated account, not a direct index of individual stocks, so it is a reference point for the range rather than a measure of stock-level harvesting. And "yield" is losses harvested as a percent of the portfolio, not tax saved. Wealthfront's S&P 500 Direct page uses an illustration that assumes roughly 4 percent harvested annually, with the usual note that actual results vary. The academic paper most often cited on this topic (Chaudhuri, Burnham and Lo, Financial Analysts Journal, 2020) could not be retrieved at the time of writing, so its figures are not quoted here.

A worked illustration

This is an illustration, not a projection. The assumptions are stated so the arithmetic can be redone with different ones.

A $250,000 taxable account, single filer, 24 percent federal ordinary bracket, 15 percent long-term capital gains rate, no state tax. First-year harvest yield taken as a range of 2 to 8 percent of the account, bracketing the vintage spread in the Wealthfront data above.

Take the 4 percent case, $10,000 of harvested losses, and vary what it offsets.

Against that, the annual cost of the wrapper on $250,000:

In the 4 percent case with long-term gains to absorb the loss, the first year defers $1,500 of tax at a cost of $225 for the cheapest direct index, against $50 for the ETF. In the no-gains case the deferral is $720, with $7,000 of carryover waiting for a future gain. At a 2 percent harvest and no gains, the year defers the same $720, since the $3,000 cap binds either way, at a cost of $225 to $875 depending on the platform.

Two things the tables do not show. The fee recurs every year whether or not there is anything to harvest, and the harvest declines as lots rise above basis. The $10,000 of harvested losses also lowered the account's basis by $10,000, so a $10,000 gain is waiting in the replacements. The net benefit over the holding period is the time value of the deferred tax plus any rate difference at the eventual sale, minus cumulative fees, plus or minus tracking error.

Costs and minimums, checked September 2026

Fees and minimums below were read from the vendor pages linked in the table. Schwab Personalized Indexing and Fidelity Managed FidFolios were also on the list to check, but their product pages could not be retrieved, so they are left out rather than quoted from memory.

The direct index fee is charged on the whole account. The ETF's expense ratio is deducted inside the fund's price. Both are percentages of assets, so at $250,000 the difference between 0.09 and 0.02 percent is $175 a year, and the difference between 0.35 and 0.02 percent is $825.

The wash sale interaction

The wash sale rule is tested at the taxpayer level, across every account. A direct indexing platform screens its own trades, so it will not sell a stock at a loss on Monday and buy it back on Tuesday inside the account it manages. It cannot see anything else.

Where this bites: a 401(k) brokerage window, an employer stock purchase plan, dividend reinvestment in an old brokerage account, or a spouse's account. If the direct index sells a stock at a loss and any of those buys the same stock inside the 61-day window, the loss is disallowed. In a taxable account the disallowed amount moves into the replacement shares' basis and comes back later. Inside an IRA it is gone. The wash sale rule post covers both cases and the arithmetic.

A direct index multiplies the surface area for this. An ETF holder who harvests has one ticker to keep clear across other accounts. A direct index holder has hundreds.

Customization

Owning the constituents means they can be edited. The two common uses are excluding an employer's stock, so that someone with a large RSU position is not adding more of the same company through the index, and applying screens, whether sector, ESG or a single company the holder does not want to own. The ETF overlap post covers the fund holder's version of this problem: the same dozen companies arriving through several funds, invisible on the statement. A direct index can exclude a name; a fund cannot be edited. Each exclusion adds tracking error. A few cost little; a long list turns the account into an active portfolio with an index label.

Where the ETF wins

Simplicity. One position, one price, one line on the 1099-B. A direct index generates hundreds of lots and a long Form 8949, prepared by the platform but reviewed by the holder or their preparer.

No minimum. One share of an S&P 500 ETF is the entry point. Direct indexing platforms set minimums, $5,000 to $50,000 on the pages checked above.

Portability. An ETF transfers in kind to any brokerage in a few days. Leaving a direct indexing platform means transferring hundreds of positions to a brokerage that will not manage them, or liquidating and realizing the deferred gains the harvesting built up.

Fee. Two basis points against nine or more, every year, on the whole balance.

The ETF loses on one axis, the harvest, and that axis is worth the most in the first few years, in volatile markets, for an account with realized gains and a high bracket. It is worth the least for a small account with no gains, in a market that goes up steadily, held by someone in a low bracket.

Where Helm fits

I build Helm Terminal. The free ETF overlap tool shows which companies two funds share and how much of each fund sits in the shared names, which is the fund-holder's version of the concentration question direct indexing solves by editing the constituents. The Pro tax center, $20 a month or $149 a year, lists the lots at a loss in the accounts you connect and screens each against purchases in the prior 30 days across every linked account. That is the manual version of what a direct indexing platform automates, applied to whatever you already hold. The free tax-loss harvesting calculator prices a harvest at a stated bracket and gain level.

Related reading

• How to Find Tax-Loss Harvesting Opportunities
• The Wash Sale Rule: What It Actually Disallows
• ETF Overlap: Why Two Funds Own the Same Companies
• Capital Loss Carryover: How It Works, With the Math
• How to Calculate Cost Basis on Stocks and Funds