税务策略 · 投资

税务亏损收割 —— 机制、洗售规则，以及直接指数化 / 130-30 多空策略实际交付什么

Third piece in the Noesis Tax Strategy series (companion to Buy-Borrow-Die and Real estate investment tax strategies). Published 2026-09-03.

翻译进行中

本篇分析的完整正文目前仅有英文版。中文翻译正在整理中，将在后续更新中上线。与此同时，Noesis 的所有界面、术语表查询和其他页面均已完整本地化。

Tax-loss harvesting (TLH) = deliberately selling losing positions to realize capital losses, using those losses to offset capital gains + up to $3,000/year of ordinary income. Unused losses carry forward indefinitely. The wash-sale rule requires that you not buy the "substantially identical" security within 30 days before or after the loss sale.

Noesis does not recommend any specific investment product or platform.

All performance and tax-alpha estimates are ranges based on published academic research and platform disclosures — actual results depend on account size, volatility, gain profile, and how much tax the harvested losses actually offset. Tax-loss harvesting only helps investors who have gains to offset; without them, losses just carry forward. Not tax, legal, or investment advice.

Three implementation tiers

Each with distinct economics. Match tier to account size and gain profile:

Tier	Platform examples	Minimum	Fee	Tax alpha
Robo-advisor auto-TLH	Wealthfront, Betterment	$50k typical	0.25% AUM	0.5–1.0%
Direct indexing	Fidelity FDIS, Schwab Personalized, Vanguard Personalized, Frec, Aperio, Parametric	$100k–$500k	0.15–0.40%	1.0–2.0%
130/30 long-short tax-alpha	Aperio, Parametric, AQR	$1M–$5M	0.60–1.00%	2.0–4.0%

The tax alpha exists but has honest limits: it only helps investors with gains to offset, it's back-loaded (biggest in years 1-3, decays over time), and fees can eat much of the benefit at smaller account sizes. And every dollar of harvested loss reduces the future basis of the replacement position — you defer the tax, you don't erase it. Until you die, at which point step-up in basis wipes out the entire deferred stack — which is why TLH pairs so powerfully with Buy-Borrow-Die.

The mechanic — how tax-loss harvesting works

The core rule (§1211, §1212)

Capital losses in a taxable brokerage account can be used to:

Offset capital gains — dollar-for-dollar, no limit. Realized $50k of losses, realized $50k of gains → $0 net taxable gain.
Offset up to $3,000 of ordinary income per year — for individuals or married-filing-jointly. Once your losses have zeroed out your realized gains, the next $3k reduces your taxable wages.
Carry forward indefinitely — anything above the $3k that's not offset by gains this year rolls to next year. No expiration.

Short-term losses first offset short-term gains; long-term losses first offset long-term gains. Excess short-term losses can then offset long-term gains and vice versa.

The workflow (manual, one-off)

Identify a position in your taxable brokerage account trading below your cost basis
Sell the position, realizing the capital loss
Immediately buy a similar but not "substantially identical" replacement security — so you don't lose market exposure while the wash-sale window runs
31 days later, you can (optionally) sell the replacement and buy back the original position, restoring your original allocation

Common pairs used to preserve market exposure:

Sell SPY, buy IVV or VOO (all track S&P 500)
Sell VTI, buy SCHB (both track total US market)
Sell VOO, buy SPLG
Sell BND, buy AGG (both aggregate US bond indices)

Are these "substantially identical"? The IRS has never definitively defined the term. Industry convention is that funds tracking the same underlying index are substantially identical (SPY vs. IVV vs. VOO), so the safer swap is to a fund tracking a similar but different index. In practice, most robo-advisors and CPAs treat SPY↔IVV as risky and SPY↔VTI (S&P 500 → Total Market) as safe. There's no case law that has resolved this cleanly.

The wash-sale rule (§1091)

Enacted in 1921. Says a capital loss is disallowed if you buy the same or "substantially identical" security within 30 days before or after the loss sale. The disallowed loss isn't erased — it's added to the cost basis of the replacement shares, so you get the deduction eventually when you sell those.

Applies to:

Same brokerage account — obvious
Different brokerage accounts of the same owner
Your IRA and Roth IRA (worst outcome — loss just vanishes, no basis carryover)
Spouse's accounts — treated as your accounts
Wholly-owned entities (LLCs, partnerships you control)

Does NOT apply to:

Cryptocurrency (as of Sept 2026) — IRS treats crypto as property, not securities. Sell BTC at a loss and buy it back next day. Congress has proposed closing this hasn't succeeded yet.
Foreign securities on foreign exchanges (gray area)
Different security types (stock ↔ ETF holding that stock)

Reference: IRS Pub 550 — Investment Income and Expenses · 26 USC §1091

Robo-advisor auto-TLH — the entry-level tier

Since ~2013, robo-advisors have automated tax-loss harvesting for retail accounts. The two dominant US players:

Wealthfront

Auto-TLH on all taxable accounts $50k+
ETF pairs across ~40 asset classes
Published estimate: ~1.16% annualized tax alpha (2011-2020 backtest)
Fee: 0.25% AUM
Realistic net-of-fee benefit: 0.5–0.9%

Betterment

Auto-TLH on all taxable accounts (any size) — no minimum
Similar ETF-pair methodology
Published estimate: ~0.77% annualized alpha
Fee: 0.25% AUM (Digital tier)
Realistic net-of-fee benefit: 0.4–0.7%

What auto-TLH does WELL

Runs daily, catches short-term dips
Zero user effort
Cross-account coordination (no wash-sale surprises)
Includes both long-term and short-term loss harvesting

What auto-TLH does POORLY

Only ~10 major ETFs — 10x fewer harvest ops than direct indexing
Doesn't work well after years — most positions above basis
Small accounts may only see a few hundred dollars annually

Rule of thumb: worth it for taxable accounts $100k+ where the investor isn't going to run TLH manually.

Direct indexing — the middle tier

Instead of holding S&P 500 exposure via a single fund (SPY), the investor owns the individual stocks of the S&P 500 in their own brokerage account. Now the platform can harvest losses at the individual security level — even in a year the S&P 500 is up 15%, roughly a third of its constituents are down, providing many more harvest opportunities.

The dominant platforms (2026)

Platform	Owner	Minimum	Fee	Notes
Fidelity FDIS	Fidelity	$5,000	0.35–0.40%	Lower minimum than most
Schwab Personalized Indexing	Schwab	$100,000	0.35%	Wide index selection, ESG tilt available
Vanguard Personalized Indexing	Vanguard (bought Just Invest 2021)	$250,000	0.20%	Lowest cost of the majors
Frec	Independent	$20,000	0.10%	New entrant; aggressive pricing
Aperio	BlackRock (bought 2021)	$1,000,000	0.15–0.35%	Institutional-grade; historically dominant
Parametric	Morgan Stanley	$250,000	0.15–0.35%	Long history; wide adoption in wealth mgmt

Estimated tax alpha

Academic research (Israelov & Lu 2022, others): 1.0–2.0% annualized tax alpha over long horizons
Sponsor estimates (Parametric, Aperio, Fidelity): 1.0–1.5% typical
Higher for concentrated / high-volatility portfolios — Russell 2000 has more harvest opportunities than S&P 500
Lower after fees — 0.15-0.40% fee vs. maybe 0.03-0.09% for a straight index fund → the incremental fee eats 0.10-0.30% of the alpha

Where DI shines

Appreciated concentrated positions to unwind
Taxable brokerage account $250k+ long-term hold
High tax bracket (37% federal + high state)
Ongoing capital gains to offset

Where DI under-delivers

Small accounts ($100k or less) — fee eats alpha
No gains to offset — losses carry forward
Frequent rebalancing needs — tracking error
Plans to sell everything in the next 5 years

130/30 long-short tax-alpha — the premium tier

The mechanic: the portfolio owns 130% of the target market long AND 30% short (net exposure = 100%). The short positions manufacture ADDITIONAL losses to harvest while keeping full market exposure.

Result: roughly double the tax alpha of long-only direct indexing.

The players

Aperio (BlackRock) — was the original 130/30 tax-alpha provider; still dominant in institutional wealth
Parametric (Morgan Stanley) — offers 130/30 as an overlay
AQR Capital — offers a long-short tax-managed strategy for HNW / institutional accounts
Others appearing in wealth-manager platforms increasingly

Fee reality

Fee: 0.60–1.00% AUM (vs. 0.15-0.40% for long-only DI, 0.03-0.09% for a straight fund)
The incremental fee vs. long-only DI is ~0.50%
The incremental tax alpha vs. long-only DI is ~1.0-2.0%
Net: roughly $50k–$100k additional after-tax return per $1M annually, if the investor has the gains to offset

The cryptocurrency wash-sale gap — legal today, may not be tomorrow

Because the IRS treats cryptocurrency as property (not securities), the wash-sale rule (§1091, which explicitly says "securities") does not apply. This creates a legal loophole:

Sell Bitcoin at a $50k loss
Buy Bitcoin back the same day
Realize the $50k loss for tax purposes
Still own the same Bitcoin position

This is called "tax-loss harvesting on crypto" and is the single biggest legal tax advantage of crypto over stocks in a taxable account.

Political risk: every US Treasury reform proposal since 2021 has proposed closing this loophole. It has been in multiple draft bills. It has not passed. As of Sept 2026 the loophole remains open. If you plan to use it, model your tax planning without relying on the loophole surviving through 2027+.

Reference: IRS Notice 2014-21 — the original guidance treating crypto as property.

The honest failure modes of tax-loss harvesting

1. It only helps if you have gains to offset

TLH generates losses. If you have no capital gains in the current year, the losses offset up to $3k of ordinary income + carry forward indefinitely. If you never realize gains (which is exactly the Buy-Borrow-Die approach), the losses accumulate and never get used.

2. It's back-loaded

The biggest harvesting opportunities are in years 1-3 when many positions are near basis. After 5-10 years, most positions are well above basis and there are fewer opportunities. Estimated alpha decays over the account's life:

Years 1-3	1.5–2.5% annually
Years 4-7	1.0–1.5%
Years 8-15	0.5–1.0%
Years 16+	0.2–0.5%

3. Deferred, not eliminated (unless you die)

Every dollar of loss you harvest reduces the basis of the replacement position. When you eventually sell that replacement, the deferred tax comes due. Over a 30-year hold, deferring $100k of tax at a 5% cost of capital is worth roughly $432k in present-value terms — real, but not the sticker number the harvesting appears to save.

UNLESS: you die holding the replacement position. Step-up in basis at death eliminates the deferred tax entirely. This is why TLH is so powerful when combined with Buy-Borrow-Die — the loss you harvest reduces your income tax now, and the deferred basis-reset liability disappears at death.

4. Tracking error

Any harvest that swaps you into a similar-but-not-identical security introduces tracking error against the target index. Over long horizons this can be as much as 0.1-0.3% annually of return divergence.

5. Fee drag

The incremental fee of the TLH platform vs. a straight index fund is 0.10-0.35% for direct indexing, 0.50-0.90% for 130/30. If the annual tax alpha is 1.0%, a 0.3% incremental fee eats 30% of the benefit before the investor sees any.

6. Not all "harvests" are real

Harvesting a $10k loss and buying a slightly-different security doesn't save $10k in taxes — it saves $10k × your marginal rate on capital gains, assuming you had matching gains to offset. And when you eventually sell the replacement, you owe tax on the deferred gain. The realistic dollar benefit is 10-25% of the harvest amount, not the full amount.

Who this actually helps

Fits well

Taxable brokerage accounts of $250k+
High marginal tax bracket (32%+ federal + high state)
Ongoing capital gains to offset (annual RSU vesting, real estate sales, business income)
Long holding horizon (10+ years, ideally to death)
Willing to pay incremental 0.15-1.00% platform fee

Fits poorly

Wealth mostly in IRAs / 401(k) — no capital gains events, no TLH possible
Low tax bracket — TLH savings proportional to bracket
Small taxable account ($50-100k) — fee eats benefit
Short holding intent — no time for tax alpha to compound
No future gains to offset — losses pile up unused

Warning signs the strategy is being oversold

Advisor pitches TLH without asking about your gain profile
Sponsor projects "3%+ tax alpha" without disclosing after-fee net
Platform's own historical performance claims aren't audited or peer-reviewed
The pitch focuses on the raw harvest amount, not the actual dollar tax benefit