On this page
Every pitch for tax alpha, the industry’s word for return added by tax management, quotes a number. The two questions below are the ones that separate a number that describes your account from a number that describes a fresh account in a good year.
Question one: does the benefit persist past year five?
Ask for the benefit in year six of an account with no new cash. Every published multi-year curve, from AQR’s 2022 and 2023 simulations to a robo-adviser’s live account vintages, shows the benefit crossing below 25 to 30 basis points somewhere between years three and seven once new money stops. The sources that avoid the decay do so by assuming steady contributions, recurring short-term gains, or a charitable exit, all of which describe the investor rather than the product.
A provider quoting “1 to 2% per year” is quoting year one or a flow-assisted case. A good answer sounds like: “About 155 basis points in the first year, 30 a year long-run if you keep contributing, near zero if you do not.”
Question two: is the long/short version right below $5 million?
Tax-aware long/short strategies hold a short book against the long one (130/30 up to 250/150) to generate three to ten times the losses of ordinary harvesting. The pitch is more losses. The problem is that more losses only help if there are gains to absorb them, and the strategy has costs the long-only version does not.
The case for
- Three to ten times the gross losses of long-only harvesting.
- Large short-term gains elsewhere can be sheltered every year.
- AQR's simulation shows a real benefit, before fees.
The case against
- Financing and borrow cost 0.28% to 2.7% a year on top of a 0.50% to 1.30% fee.
- Tracking error of 2% to 7% against the index, by design.
- AQR's own tenth-percentile net-of-cost pre-tax alpha is negative at every leverage level.
- An independent model finds it trails plain sell-and-index by 0.11% a year in a favorable $10M case, turning positive only with a step-up within ten years or a move out of California.
- Whether a diversified long/short book triggers the straddle rules has no ruling.
Minimums tell the same story: $100,000 at one retail entrant, $1 million at most managers, $3 million where the custodian’s margin rules apply. The verdict from the evidence: not a general strategy below about $5 million; conditionally sensible above $3 million for an investor with a specific large gain to absorb each year.
What no provider has
An aged, audited after-tax track record. Harvest yield, the number most providers publish, is a count of losses taken. It is not a return. The strongest evidence for the strategy is a simulation gross of fees; the strongest evidence against is another model. Both are models.