Which Sleeve Goes Where? Asset Location for Direct-Indexing, Long-Short, and Option-Income Mandates

Summary

The two-asset rule and four sleeves that do not fit it

Most people learn the asset-location rule from a model with one stock and one bond. Ordinary rates tax the bond's interest every year, so the bond goes in the sheltered account. A stock's return mostly waits untaxed until a sale and then pays a lower rate, so the stock stays taxable. Dammon, Spatt and Zhang proved the rule for a long-horizon investor; later authors carried it over to mutual funds and to after-tax accounting.

Equity mandates in practice look different. A direct-indexing account tracks an index and harvests losses lot by lot, producing net capital losses year after year. Its tax-aware long-short extension harvests on both books, produces about three times as much loss, and cannot sit in a retirement account because the shorts need margin. A put-write sleeve sells cash-secured puts. Its collateral earns ordinary interest, and its premiums come back as short-term gains. A long-only factor tilt turns over fifty to a hundred percent a year and realizes gains as it goes. Two of the four produce losses. Nothing in the two-asset rule says where any of them goes.

We write the location problem for these four sleeves across a taxable account, a traditional IRA and a Roth, and solve it in closed form. We describe each sleeve by how much of its return arrives in each tax character; a harvesting sleeve carries a negative short-term entry. Each year the taxable account nets short-term gains and losses across sleeves. We tax a net gain at the short-term rate of 40.8% and value a net loss at some fraction of that rate, and a present-value factor carries the tax on deferred gains to liquidation.

One account after tax

Measure balances after tax and the traditional IRA and the Roth behave as one account for location purposes. Reichenstein made this point years ago. It halves our problem. All that matters is what goes into the taxable account, and we leave the split between the two retirement accounts open.

The tax cost of the taxable account is then convex and piecewise linear in the location, with one kink where the account's net short-term flow is zero. On one side of the kink the household has net short-term gains, and every harvested loss offsets a dollar that would otherwise pay 40.8%. On the other side it has net losses. There a marginal loss is worth whatever the household can eventually do with it. Placing a sleeve in the taxable account costs its character cost plus a regime-dependent price on its net short-term flow.

A floor on the value of a loss

Could the regime dependence reorder the sleeves? We thought it might, and a bound rules it out. A harvested dollar of loss lowers the position's basis, so the household pays recapture when it sells. A loss is never worth less than that recapture, and in the loss-rich regime this pins its price from below. Under the floor, a harvesting sleeve costs at worst what it would cost with harvesting switched off. Table 3 shows it. For the liquidating household in the loss regime, direct indexing costs 124.3 basis points per dollar-year and the long-short sleeve 130.3, the same figures as the no-harvest row. In the gain regime the long-short sleeve comes in at -162.7: it earns more after tax in the taxable account than it would sheltered. Put-write costs 265.2 in the gain regime and 204.2 in the loss regime, the most of the four either way.

Figure 1 draws these marginal costs for three households in both regimes; the unforced sleeves keep one order in every panel. The margin constraint forces the long-short sleeve into the taxable account. After it, the fill order runs direct indexing, factor tilt, put-write, which is what the two-asset rule would have guessed. Against a pro-rata mirror this order is worth 12 basis points a year for a household that will liquidate and 21 for one that will get a step-up in basis. A business owner with outside gains gets 41. Against the growth-to-Roth heuristic we measure 16 to 27, and an exhaustive grid over static locations finds nothing better.

Sizing the long-short sleeve

What the model does move is size. Consider a household that runs direct indexing in its taxable account and has no outside short-term gains. It already holds more losses than it can use, so a sleeve that harvests three times as much changes nothing on the tax side. Each extra harvested dollar is worth only its own recapture, and the long-short sleeve must pay its way through pre-tax alpha. In the simulation, the right-hand panel of Figure 6 shows unused carryforward building year by year under the textbook location; Table 6 puts it at 1.11 per dollar of initial after-tax wealth for the liquidating household.

Outside short-term gains reverse this. For that household, every dollar of wealth moved from direct indexing into the long-short sleeve earns about 184 basis points of wealth per year in tax benefit, until the sleeve's extra losses absorb the outside gains. Figure 3 shows the kink at that matched book. Its size has a one-line closed form.

The simulation

A fixed-seed Monte Carlo over 20 years and 3,000 paths agrees with the flow model. It tracks basis, carries losses forward and charges a liquidation tax at the end. For the liquidating household the textbook location delivers 374.1 basis points of certainty-equivalent growth, against 355.1 for the pro-rata mirror and 341.1 for growth to Roth, and across households the gaps run 19 to 32 basis points over the mirror and 33 to 40 over the heuristic. Harvesting is where the flow model is rough. In practice the losses come early. Twenty-year averages smooth that away.

We expected netting across sleeves to change where the sleeves go. In our results it mostly changes how large the long-short sleeve should be.

Who this is for: CPAs and researchers working on asset location who want to know what a harvested loss is worth once a household has more of them than it can use.