The After-Tax Intermediation Wedge: Distribution-Free Evidence on Replicating versus Allocating Tax-Aware Mandates

Summary

Build or buy, priced after tax

A small investment firm that manages taxable money runs into a choice it cannot avoid. It can build: run a direct-indexing core and, where the account qualifies, a tax-aware long-short sleeve, holding every position inside the account and harvesting losses lot by lot. Or it can buy: send the capital to outside managers, usually hedge funds or commingled vehicles, and accept whatever the fund's K-1 delivers each spring. Buying brings manager skill and outsourced operations. It also brings a second layer of charges and taxable distributions that arrive on the fund's schedule rather than the account holder's. And it gives up harvesting against a specific basis and specific outside gains. Figure 1 draws the decision and the three channels through which the after-tax difference flows.

Practitioners know which direction this trade-off runs. What we wanted was the size. How wide is the after-tax wedge, and how much pre-tax alpha must an outside manager add before buying beats building? Account size and horizon change both answers.

One model, identical gross returns

We hold pre-tax gross returns identical across the two routes. That way nothing in the comparison rewards the in-house route for skill it has not shown. We derive two closed forms: one for after-tax wealth under a constant annual realization fraction, and a second for the share of a marginal point of outside alpha that survives each wrapper. On top of them sits a fixed-seed Monte Carlo, calibrated to Ken French data from 1926–2026, which prices the wedge across accounts from $250k to $25M at tax rates of 40.8%, 23.8% for short-term and long-term gains.

We did not expect what the second closed form showed. At a liquidation exit, the fund wrapper and the SMA wrapper pass a marginal pre-tax dollar through at nearly the same rate, 75.1 versus 76.2 percent. Why? Paying the blended distribution tax each year, with its basis credit, costs about what paying the long-term rate once at exit costs. So the wrapper does its damage almost entirely on the base return. Figure 5 shows the share kept under each wrapper as the realization fraction varies. A step-up at death breaks the equivalence, since the SMA's deferred alpha then escapes tax while the annual distribution tax still clips the fund's.

Over twenty years with a liquidation exit, the in-house route finishes 1.8 to 2.0 times wealthier after tax, and Figure 6 gives the ratio by horizon, tier and exit. In annual terms we measure a wedge of 327 bp for a $250k account, rising to 404 bp for a $25M account. Costs account for 115 to 130 bp of it. That share shrinks in basis-point terms as the account grows. Fund-level tax leakage takes about 166 bp at every size, and no amount of manager skill removes it. Forgone harvesting adds 31 bp for a small direct-indexing account and 116 bp for a large account carrying a 40 percent long-short sleeve.

The hurdle an outside manager must clear

How much non-replicable pre-tax alpha must the outside fund deliver before the allocate route catches up? Under a liquidation exit we find a range from 235 bp for a $250k account at a 5-year horizon to 441 bp for a $25M account at a 25-year horizon. A step-up lifts it to 416 to 722 bp. Deferral explains the jump: a step-up makes deferral permanent, and the in-house route defers far more. Figure 9 plots the full schedule.

Turnover matters a great deal. Take a $5M account at 20 years, where the base case asks for 397 bp; a low-turnover fund brings that down to 289 bp, while a high-turnover fund pushes it up to 528 bp, so two funds with the same skill can face very different bars depending on how often they trade. In no scenario we stressed does the hurdle fall below roughly 260 bp.

Fat tails, intervals and harvesting

A century of equity returns is not Gaussian, so we rerun the simulation under four return processes: an iid Gaussian base, a stationary bootstrap, a two-regime Markov switch and a GARCH-t process with its drift matched to the others. Figure 12 shows the wedge under each with 90 percent split-conformal intervals. Across 2,000 paths the medians move little. Under the bootstrap we get 245 bp for a $250k account and 312 bp for a $25M account, against 233 bp and 298 bp under the Gaussian base. The intervals are wide, yet their lower ends stay above zero in every tier and under every process.

Harvesting is where we hold the claim loosely. For the $250k tier the 90 percent interval on the harvesting channel runs from -9 to 71 bp under the Gaussian process and widens under the others. Below about $1M, the channel is positive on average but it is not an identified claim for any single account. Figure 15 shows the channel per account with its intervals. We say this plainly because the simulation cannot pin that part of the wedge down for a small account.

A last point concerns means and medians. A lognormal terminal multiple has a mean well above its median, and Figure 16 draws the two apart at twenty years. Re-solving the break-even on the median rather than the mean lowers the hurdle by 80 to 90 bp. For a $250k account under the Gaussian process, the median-equalising figure is 258 bp. We take one operating choice from all this: we build the tax-aware core in-house and hold outside allocations as satellites for return streams we cannot replicate, at the hurdles this schedule makes explicit.

Who this is for: Investors with taxable accounts, CPAs weighing fund pass-through against lot-level control, and researchers or journalists who want the after-tax wedge quantified.