Photo by Marcus Reubenstein on Unsplash
The short version
- An ETF is a single tax lot from the holder's perspective: you can only harvest a loss when the whole fund is underwater. Direct indexing holds the underlying stocks separately, so losses can be harvested inside an index that is up on the year.
- The edge is real but front-loaded — harvestable losses are largest in the early years and decay as the account appreciates and embedded gains accumulate. Modeling a fixed "tax alpha" forever overstates the benefit.
- Bottom line: direct indexing suits large taxable accounts with ongoing capital gains to offset and a long horizon; for most investors in tax-advantaged accounts or with modest balances, a low-cost ETF captures the majority of the available benefit at a fraction of the complexity.
The case for direct indexing rests on a single structural fact: when you own an S&P 500 ETF, you own one thing that goes up or down together. When you own the 500 stocks separately, you own 500 things that move independently. In any given year a broad index can finish up double digits while a meaningful minority of its constituents finish down. Direct indexing lets you harvest the losers without selling the index. The question worth answering carefully is not whether that works — it does — but how much it is actually worth, to whom, and for how long.
Context: why the ETF wrapper hides losses from you
Inside an ETF, the fund itself is remarkably tax-efficient. The in-kind creation-and-redemption mechanism lets authorized participants swap low-basis shares out of the fund without triggering a taxable event, which is why broad equity index ETFs so rarely distribute capital gains. That machinery works at the fund level. From the perspective of you, the shareholder, the ETF is a single position with a single cost basis. You realize a loss only by selling shares that are worth less than you paid — meaning the entire fund has to be underwater relative to your purchase price.
Direct indexing dissolves the wrapper. A separately managed account (SMA) buys the individual constituents — sometimes the full index, more often an optimized sample of 150–350 names that tracks the benchmark within a tight tracking-error budget. Each stock carries its own lot-level basis. When one name drops below cost, the manager sells it, books the loss, and buys a highly correlated substitute to keep the portfolio's factor exposures intact and stay clear of the wash-sale rule. The index exposure barely moves; the tax loss is real and bankable. This is the same after-tax logic explored in tax-loss harvesting and the after-tax compounding gap, applied at finer granularity.
The two approaches, side by side
| Dimension | Broad-market ETF | Direct indexing (SMA) |
|---|---|---|
| Tax-lot granularity (holder view) | One lot per purchase | Hundreds of lots, harvested individually |
| Harvest when index is up | No — needs the whole fund underwater | Yes — harvests individual laggards |
| Explicit cost | ~0.03%–0.10% expense ratio | ~0.20%–0.40% management fee |
| Tracking error vs index | Effectively nil | ~0.5%–2% annualized (sampling) |
| Minimum to implement well | One share | Commonly $100k–$250k+ |
| Portability / complexity | Trivial to move, one line | Hundreds of lots; transfers are messy |
| Best account type | Any, including tax-advantaged | Taxable only — the benefit vanishes in an IRA |
Fee figures reflect published expense ratios for broad index ETFs (issuer fact sheets, e.g., Vanguard's VTI at investor.vanguard.com) and the management-fee ranges disclosed by major direct-indexing providers as of mid-2026. The ETF side of this table is the same wrapper analyzed in the three-fund portfolio in 2026 — nothing exotic, just the standard low-cost core.
How large is the harvesting benefit, really?
The academic and practitioner literature on tax-loss harvesting — including work from the AQR and Vanguard research groups — tends to converge on a "tax alpha" of roughly 0.5% to 1.5% per year in the early life of a taxable account, with the higher end requiring high volatility, high dispersion across names, and a high marginal tax rate to absorb the harvested losses. Those are the conditions that maximize the value: more dispersion means more names dip below cost; a higher tax rate means each harvested dollar of loss is worth more.
Two of those three inputs are outside anyone's control. Dispersion is a market regime, not a dial. In a low-volatility tape — the VIX sat at 16.5 in mid-July 2026 (FRED, asof 2026-07-14), a benign reading — fewer constituents fall far enough below basis to be worth harvesting. The realized benefit in a calm year is materially smaller than the benefit in a 2020- or 2022-style year. Any provider quoting a single steady-state "tax alpha" number is quietly averaging across regimes you have not lived through yet.
The harvesting benefit is largest in the years you least expect to need it, and smallest in the account's later years when the balance — and the tax bill it defers — is largest.
The non-obvious part: harvesting yield decays
Here is the effect most marketing material understates. The supply of harvestable losses is not constant — it depletes. In year one, every lot is near its purchase price, so a normal pullback pushes many names below cost and the loss inventory is rich. As the account compounds over a decade, the average lot sits on a large embedded gain. A 10% market drop that would have created a forest of harvestable losses in year one now merely trims unrealized gains on lots that are still comfortably above basis. The account can be down for the year and still offer almost nothing to harvest.
Initially I modeled this as a slow, gentle decline. Then I looked at the lot-level basis distribution more carefully and it is closer to a decay curve: the bulk of lifetime harvesting happens in the first three to five years, after which the account becomes progressively "locked up" in appreciated lots. This has two consequences that matter for planning. First, the benefit is front-loaded, so it should be discounted, not annuitized flat into the future. Second — the genuinely counterintuitive part — a direct-indexing account eventually accumulates a large embedded gain of its own. You have deferred taxes, not eliminated them. Unless those low-basis lots are ultimately donated to charity or passed through an estate with a step-up in basis, the deferred liability comes due. That makes direct indexing as much an estate-and-gifting instrument as a return enhancer, which connects it directly to planning a tax-efficient wealth transfer.
Costs, frictions, and where the edge leaks away
The gross benefit is not the net benefit. Direct indexing carries a management fee an order of magnitude above a broad ETF — call it 0.25% versus 0.03%, a gap of roughly 0.22% (22 basis points) that compounds against you every year, in every regime, harvest or no harvest. Basis points matter; that drag is certain while the harvesting benefit is contingent. There is also tracking error: an optimized sample of a few hundred names will drift from the full index by something like half a percent to two percent annualized, and that drift is symmetric — it can just as easily cost you as help you in a given year. And there is behavioral cost: hundreds of tax lots are far harder to move between custodians, to rebalance cleanly, or to unwind than a single ETF line, which quietly raises the odds of staying put in a mediocre product.
None of this makes direct indexing a poor tool. It makes it a specialized tool whose net advantage depends on a marginal tax rate high enough to value the losses, a taxable balance large enough to spread the fixed complexity over, and — crucially — an ongoing stream of capital gains elsewhere in your financial life that the harvested losses can actually offset. Harvested losses you cannot use are worth only the $3,000-a-year ordinary-income offset and a carryforward; the headline tax alpha assumes you have real gains to shelter.
FAQ
Can I tax-loss harvest with ETFs at all? Yes, but only at the whole-fund level. If your ETF position is underwater you can sell it and buy a similar-but-not-substantially-identical fund to book the loss while staying invested. What you cannot do is harvest individual losing stocks while the fund overall is up — that requires owning the constituents directly.
Does direct indexing make sense inside an IRA or 401(k)? No. Tax-advantaged accounts already defer or exempt taxes, so there is no loss to harvest and no benefit to capture. Direct indexing is a taxable-account strategy exclusively.
How much money do I need? Providers commonly set minimums in the $100,000–$250,000 range because tracking the index well with individual lots requires enough capital to hold a representative sample. Below that, the tracking error and operational overhead tend to swamp the harvesting benefit.
What is the wash-sale risk? Selling a stock at a loss and rebuying it (or a "substantially identical" security) within 30 days disallows the loss. Direct-index managers handle this by buying a correlated substitute rather than the same name, but investors should avoid holding the same stocks — or a fund tracking the same index — in other accounts, which can inadvertently trigger wash-sale disallowance.
Will I actually keep the harvested losses forever? Not exactly. Harvesting defers tax by lowering basis; it does not erase the liability. The deferred gain sits in low-basis lots and eventually comes due unless the lots are donated or receive a step-up at death. The strategy's full value assumes a specific end-of-life plan for those lots.
Key takeaways
- An ETF is one tax lot to its holder; direct indexing is hundreds — that granularity is the entire source of the harvesting edge.
- Realistic tax alpha runs roughly 0.5%–1.5% early on and depends on volatility, dispersion, and your marginal rate — none of which are guaranteed in a calm market.
- The benefit decays as the account appreciates and embedded gains accumulate; it should be discounted, not projected flat.
- A certain 0.22% fee gap and tracking error offset part of the contingent benefit — the net edge is smaller than the gross.
- The strategy fits large taxable accounts with real gains to offset and an estate or charitable plan for the low-basis lots; for most others, a low-cost ETF captures the bulk of what matters.
Editor's read
Direct indexing is a legitimate tool that the industry oversells by quoting steady-state tax alpha as if it were a coupon. The benefit is front-loaded, regime-dependent, and partly clawed back by fees and a deferred-gain liability that eventually surfaces. Where it earns its keep is a specific profile — a large taxable balance, a high marginal rate, ongoing gains to shelter, and a step-up or gifting plan for the terminal lots. Absent all four, the honest comparison is a broad ETF that costs a tenth as much and moves in one line. The editor uses broad-market ETFs in the taxable sleeve and does not run a direct-indexing SMA; the balance and complexity threshold have not made the case.
The editor holds broad-market index ETFs in the taxable core and does not currently use a direct-indexing separately managed account.
Methodology & sources. Expense-ratio and fund-structure figures are from issuer fact sheets (Vanguard, investor.vanguard.com) as pulled 2026-07-16. Direct-indexing fee and minimum ranges reflect major providers' published schedules as of mid-2026. Macro figures — 10-year Treasury 4.58% (asof 2026-07-14), federal funds rate 3.63% (asof 2026-06-01), VIX 16.5 (asof 2026-07-14), CPI year-over-year 3.7% (asof 2026-06-01) — are from FRED. Tax-alpha ranges summarize the published direct-indexing and tax-loss-harvesting research literature; no figure here is a forecast.
This article is for educational purposes and does not constitute personalized financial advice. See our full Disclaimer.