The Tax Alpha That Decays

A holder with five million dollars in a taxable brokerage account, asked to evaluate a direct-indexing pitch, will hear a number that lands cleanly. The pitch is built around it. The number is most often phrased as a range — half a percent to one percent in after-tax improvement, per year, for as long as the sleeve runs — and it is offered without serious qualification, as though the harvesting that produces it were a feature of the asset class rather than a feature of the early years of the structure that runs over it. The number is, in the early years, broadly accurate. It is not, over the sleeve's life, the steady-state number it is sold as.

This dispatch is about the gap between those two facts. It is about what direct indexing actually does for a portfolio at five million dollars, what the alpha figure means once the curve under it is drawn, what the threshold at five million is actually measuring, and what the holder is being asked to plan for that the pitch generally does not name. The question is not whether direct indexing works. The question is what work the holder is asking it to do, and what the holder accepts the structure to look like once that work is finished.

What the sleeve actually produces

A direct-indexing program replicates a chosen index — most commonly the S&P 500, the MSCI ACWI, or a domestic large-cap variant — by purchasing a representative sample of the underlying constituents. Sample sizes vary by provider and by the holder's tracking-error tolerance; the range is typically two hundred to four hundred individual securities, weighted to approximate the index without buying every member. The position is held in a separately managed account in the holder's name, distinct from a fund wrapper, which is the structural feature on which everything else depends. Each individual purchase is tracked as its own tax lot. The granularity that fund holders cannot see and cannot act on is, for the sleeve, the operative unit.

Through the year, the sleeve's harvesting logic monitors the cost basis of every lot it owns relative to market. When a lot moves underwater, the logic evaluates whether realizing the loss is worth the trade — accounting for the wash-sale constraint, the substitute name available, the tracking-error implication, and the bid-ask cost. When the evaluation favors the trade, the lot is sold, the loss is realized, and a correlated but not substantially identical substitute is purchased to maintain the sleeve's market exposure during the thirty-one-day wash-sale window. After the window, the original name can be restored, often at a basis higher than the lot the sleeve sold, which is the mechanism by which the average basis of the sleeve rises over time.

The cadence is operational, not annual. Some sleeves harvest monthly, some weekly, some on a continuous evaluation against price thresholds. The capacity of the sleeve to generate harvestable losses in a given year depends on the underlying market's dispersion — how widely individual constituents move relative to the index aggregate — and on the sleeve's harvest cadence. A year in which the index closes up six percent might produce gross harvested losses of two to four percent of the sleeve's value in early years, depending on dispersion and on the calendar of underlying moves; a less dispersed year, or a sleeve later in its life, produces meaningfully less.

The decay curve

The number worth holding next to the pitch is the curve that sits behind it. Year one harvested losses tend to run in the low single-digit percentages of the sleeve's value, sometimes higher in dispersed markets, sometimes lower in concentrated rallies. The translation from harvested losses to after-tax alpha depends on the holder's marginal capital-gains rate and on whether the harvested losses are being applied against same-character gains the holder would otherwise be realizing at full cost. At the federal twenty-percent long-term capital-gains rate plus the three-point-eight-percent net-investment-income tax, an applied long-term loss is worth roughly twenty-four cents to the holder for each dollar harvested. A sleeve generating three percent in harvested losses against a fully usable budget produces, on this arithmetic, after-tax alpha on the order of seventy basis points for that year.

The mechanism that produced the year-one figure is also the mechanism that erodes it. Each loss harvested replaces a low-basis lot with one closer to current market value. Over time, the lots that remain in the sleeve have an average basis that approaches the sleeve's market value. The pool of underwater names available for the sleeve to harvest, year by year, shrinks. The decay is not linear and is not symmetric across market regimes; a sharp drawdown can refresh the harvesting pool, and a long uninterrupted rally can dry it down faster than the average. The general shape, across providers' published studies and across academic estimates, is for gross harvested losses to fall from the low single-digit percentages of year one toward fractions of a percent by years six through eight, with the precise trajectory governed by the path the market actually takes.

The implication for the way the alpha number ought to be carried in the holder's planning is structural. The honest framing is a curve, not a flat number. The portion of the pitch that converts the curve into a flat figure does so by averaging across an assumed market path, often over a twenty- or thirty-year horizon, and reporting the average as if it were a sustainable annuity. The average is not wrong, on its own terms. It is also not the sequence in which the alpha actually arrives. A holder evaluating the structure should know that the alpha is front-loaded, that the sleeve at year eight is a different instrument than the sleeve at year one, and that the planning question accompanying the decay is real and arrives whether the holder is ready for it or not.

What five million dollars actually buys

The five-million-dollar figure that providers cite as the minimum for a useful direct-indexing program is not a precise threshold. It is the rough level at which several costs and capacities cross into one another.

The first is the fixed operational cost of running a separately managed account. The sleeve requires reporting, lot-level tax preparation support, and the back-office machinery that distinguishes a separately managed account from a fund holding. These costs are largely fixed at the sleeve level rather than proportional to the holder's assets. At a smaller sleeve, the fixed costs are too large a fraction of the harvested-loss value to make the structure pay; at five million dollars and above, the fixed costs become a smaller fraction of the benefit, and the math begins to clear.

The second is the advisory fee that runs alongside the sleeve. Direct-indexing fees, charged by the provider or by an overlay advisor, typically run in the range of fifteen to forty basis points on the sleeve's assets per year. Against an early-year after-tax alpha estimate of fifty to one hundred basis points, the fee is a real fraction of the gross benefit. The net figure that the holder actually receives is the harvested-loss benefit minus the fee, minus the implementation drag from trading and tracking error, minus the marginal time the holder or the holder's existing advisor spends coordinating the sleeve with the rest of the household. A holder accustomed to evaluating fund costs in single basis points should hold the direct-indexing fee against the actual after-tax delta, not against the pitch's headline alpha.

The third is the absorption capacity the sleeve can offer the rest of the household. A holder with a concentrated position somewhere else in the portfolio — restricted-stock vests rolling forward, a founder's stake whose unlock is in some sense already being planned, a pre-OBBBA legacy holding whose embedded gain has accumulated over decades — has a use for harvested losses that a holder without one does not. The sleeve becomes meaningfully more valuable when the harvested losses it generates can be applied against gains the household would otherwise realize at full cost. Below five million dollars, the sleeve has too little capacity to do much of this work; at and above, the capacity is large enough that the absorption job, rather than the standalone harvesting, becomes the case that justifies the structure.

The threshold is the place where these three pieces converge. The number is not magic. It is the place where the structure stops being a less efficient version of a tax-managed fund and starts being something genuinely different.

The two jobs the sleeve can do

A holder approving a direct-indexing program is, often without explicitly choosing, asking the sleeve to do one of two jobs.

The first is to produce harvested losses that the holder banks as carryforwards, applied gradually against gains that arrive over time in the household's ordinary course — small rebalances, fund-level capital-gain distributions, the occasional opportunistic sale. The sleeve operates as a perpetual harvest engine; its losses are saved and spent at the pace of the household's modest realization flow. For a holder without a concentrated position, without a near-term liquidity event, and without a defined gain to absorb, this is what the sleeve is doing. The job is real but limited. The decay curve matters because the engine is the whole point; once the harvest thins, the engine is producing little.

The second job is absorption. A holder approaching a known gain event — an IPO lockup expiration, a sale of a privately held interest, an exchange-fund contribution under §721 that the holder is staging into rather than at once, a multi-year diversification program for a concentrated stake — runs the direct-indexing sleeve as a paired structure whose harvested losses absorb the gain on the other side. This is the role the sleeve plays in the three-tranche framework laid out in The Lot You Sell First. The sleeve is not a perpetual engine in this configuration; it is a budget, sized to the apportionment of the gain the household plans to realize, and run for the duration of the plan. The decay curve still applies, but the curve is being compared to a specific dollar value of gain to absorb, and the question is not whether the alpha lasts forever — it is whether the sleeve's available losses, over the planning horizon, are sized to the budget the holder needs.

The honest answer to "is direct indexing worth it" depends on which job is being asked. For the first job, the sleeve is worth it conditionally, in the early years, for holders large enough that the operational costs clear. For the second, it is more often worth it, and worth it for longer, because the comparison is against a specific gain that would otherwise be realized at full cost.

A question worth asking

Is the sleeve being asked to produce harvest, or to absorb a gain the household already owns?

What happens when the sleeve hardens

A direct-indexing sleeve four, six, or eight years into its life looks different than one in its first year. Its average basis has risen close to market. Its harvest engine has slowed. Its assets are tracked across hundreds of individual lots, each with its own basis and holding period, distributed in shapes that reflect the path the market took during the sleeve's harvesting life. Liquidating the sleeve outright would realize a meaningful portion of the embedded gain — by some structural irony, a similar gain to the one the sleeve was designed to harvest against.

The set of paths from the hardened sleeve is not a failure of planning; it is the set of structural choices the sleeve was always heading toward.

The hardened sleeve can be left in place as a low-cost index tracker, with its harvest function essentially complete, and held for the embedded-gain forgiveness that the basis step-up at death provides under §1014 — the long-tail equivalent of the held tranche in The Lot You Sell First. The harvesting was the work of the sleeve's first phase; the holding is the work of its second.

The sleeve can be contributed in kind to an exchange fund under §721, in part or in whole, where it joins a pool of other contributed positions and trades the sleeve's hundreds of lots for a single limited-partnership interest with a seven-year lockup and the deferral of immediate recognition. The exchange-fund mechanics are treated at length in The Lockup You Sign For. For a holder whose hardened sleeve has too much embedded gain to liquidate but whose situation has shifted such that the sleeve's continued harvesting is no longer the highest use, the §721 path is a real option, with its own costs and its own constraints.

The sleeve can be donated, in whole or in part, to charity outright or through a donor-advised fund. Appreciated lots are gifted at fair market value, with the holder receiving a deduction at that value and the charity realizing the proceeds free of capital-gains tax. For a holder with sustained philanthropic intent, the hardened sleeve is unusually well-suited to charitable transfer: its lots are pre-segregated, their bases are tracked, and the highest-gain lots — which are precisely the lots the holder would most prefer to clear from the position — are the most efficient ones to gift. The donor-advised fund extends the timing flexibility, separating the deduction year from the grant year.

The sleeve can be transitioned over multiple years against a defined harvested-loss budget — either from a successor sleeve, from concentrated-position losses on the other side of the household, or from intentionally realized losses elsewhere — in a slow unwinding whose pace matches the household's available absorption capacity. This is the closest analog to the multi-year diversification plan that follows an IPO lockup, with the sleeve playing the role the concentrated position plays in that picture.

These paths are not mutually exclusive. A common shape involves the household using the sleeve's hardened lots in combination — gifting the highest-gain lots, transitioning the middle, holding the lowest-gain pieces for the step-up. The shape worth planning is the one that fits the household's profile. The shape worth avoiding is no plan at all, in which the sleeve continues to run quietly past its productive life, with its harvest curve flat against zero, accumulating advisory fees against benefit it is no longer producing.

What this dispatch doesn't pull

Several adjacent threads run alongside the direct-indexing decision and are addressed in separate pieces or are their own subjects.

The mechanics of harvested losses, including the wash-sale interaction across the holder's IRAs that produces permanent loss destruction under Rev. Rul. 2008-5, are treated in The Free Lunch Sitting in Your Losers. The §721 contribution route into an exchange fund, including the seven-year lockup, the qualified-purchaser threshold, and the twenty-percent illiquid sleeve common to most current fund structures, is treated in The Lockup You Sign For. The diagnostic for whether the household's broader position has become the household's plan — and for which of the apportionment frames the holder sits closest to — is treated in When the Position Becomes the Plan.

Other adjacent subjects sit outside this dispatch. The IRC §475 mark-to-market election, which trader-status holders sometimes use to recharacterize gains and losses as ordinary, is its own subject and does not generally apply to direct-indexing holders. Comparative analysis of specific providers — Parametric, Aperio, Cache, the larger custodians' in-house programs, and the more recent AI-driven entrants — is deliberately outside the scope of this publication's editorial frame; the publication treats mechanics, not endorsements, and the choice of provider should be made on operational fit, sleeve quality, and fee, after the structural decision in this dispatch is made. The interaction of direct indexing with state tax rules, particularly the California Franchise Tax Board's treatment of in-state realization, is its own subject and is treated in adjacent Pillar C work. Each of these can, for a particular holder, dominate the picture this dispatch lays out. The dispatch is the framing layer beneath them.

Where this lands

The honest case for direct indexing at five million dollars is a conditional one. The sleeve produces real after-tax alpha in its early years, on the order of fifty to one hundred basis points in usable benefit for a holder large enough that the operational costs clear and the sleeve's absorption capacity has somewhere to land. The alpha decays, and the sleeve eventually hardens into a position whose harvest engine has done most of what it can do. The hardened sleeve is not a problem if it has been planned for; it is the structure the holder's first-phase decision has produced, and its second-phase paths — the step-up hold, the §721 contribution, the charitable transfer, the multi-year transition — are the choices the planning conversation should already have been having.

The structure is worth it when the holder has a defined job for the harvested losses, when the household's other assets supply gains for the sleeve to absorb, when the advisory fee and the tracking-error tolerance are sized to the actual after-tax delta rather than to the headline alpha, and when the holder has accepted, before signing, the shape the sleeve will be in some years from now. The structure is less worth it, and sometimes not worth it, when the pitch is taken on its surface — a perpetual half to one percent, indefinitely — without the curve that sits behind that number being drawn, and without the decision that follows the curve being made.

The number on the pitch is not the plan. The curve under it is, and so is the structure the curve resolves into. A holder evaluating direct indexing at five million dollars is being offered an instrument with two phases. The first phase produces the alpha the pitch describes. The second phase is the structure that follows it. The holder approving the first without contemplating the second has approved half a decision. The work of the evaluation is in completing the other half before the sleeve does it on the holder's behalf.