Tax-Rate Risk: Pricing and Hedging Legislative Uncertainty in After-Tax Portfolios

Summary

The position hiding inside a deferred gain

Nearly every tax-aware calculation assumes that today's capital-gains rate will still apply on the day the investor finally sells. The statute book disagrees. Since 1978 the top federal long-term rate has moved between 15% and 28% five times, and the scheduled sunset of current law leaves the next move an open question. So what does deferral cost once the future rate is uncertain?

Our answer starts with a change of view. Someone who holds an unrealized gain instead of selling has swapped a tax bill known today for one owed later at a rate nobody knows yet. In cash-flow terms, that is a short position in the future tax rate. Short positions can be priced.

How we price it

We start with two dates, now and a horizon at which the position is sold regardless. The advantage of waiting comes out in closed form. It equals the embedded gain times a difference: today's rate, grown by compounding, minus whatever rate applies at the horizon. Everything the gain earns after today is taxed the same way under either choice, so it cancels and drops out.

Uncertainty enters through the second term alone. A risk-averse investor marks the advantage down, and the markdown is proportional to the variance of the future rate and to the square of the gain. Double the gain and the haircut quadruples.

Table 1 gives deferral value as a percent of the embedded gain. With a constant rate, the benchmark Constantinides worked out, deferral is worth 8.3% of the gain at T = 5y and 120.2% at T = 30y. Mean-reverting rates barely move those figures (8.8% and 120.7%). A rate trending up with δ = 0.5 does more damage, pulling them to 5.6% and 115.9%. Deferral still pays in every scenario we simulate. It pays less when rates are expected to climb, and the loss hurts most at short horizons.

How big is the haircut itself? Table 2 answers in cents per 1 of position, at T = 20y and γ = 3. A gain of 0.25 with rate volatility of 0.03 costs 0.01c. At a gain of 0.75 and volatility of 0.07 the figure is 0.41c. Those are small numbers next to a deferral benefit worth tens of cents, and for most holders the premium stays a second-order correction. Figure 1 plots deferral value against the risk-neutral rate drift, one curve per horizon, and Figure 2 draws the haircut curving upward as the gain grows.

When the classical policy changes

Two dates tell you whether to defer to a fixed horizon. They say nothing about when to sell. For that question we let the applicable rate jump between legislative regimes as a continuous-time Markov chain, and we treat the sale as an optimal-stopping problem whose free boundary separates waiting from selling.

With constant rates the boundary never opens. Nobody in the model sells a gain voluntarily, which is the Constantinides tax-timing policy, and Table 3 reproduces it: at a horizon of 10 the optimal policy and always-defer both come to 1.4837 per 1 of position, while selling at the start gives 1.3942.

Mean reversion changes little from the current regime upward. There is one exception, and it is instructive. If the chain lands in the low-rate regime late in the holding period, selling into that low rate beats waiting, which is why the optimal policy at a horizon of 10 comes to 1.4944 against 1.4912 for always-defer.

Push the risk-neutral drift up far enough and the exercise region opens from the current regime as well. Figure 3 maps it at a drift of 0.6: the markers where selling wins crowd into the low-rate row near the horizon, with a thin strip in the current regime just before liquidation. The trigger is one inequality. Sell early when the risk-neutral expected future rate rises above today's rate grown by the compounding factor.

We expected the reverse going in. A high current rate felt like a reason to get out before things got worse. In the model it is a reason to wait, since a later rate is more likely to be lower, and it is the low current rate that should be used before it disappears.

Scope and claims

We are not arguing against deferral. It remains the better choice in most of the cases we study, and the dollar value of fine timing is small beside the value of deferring at all. The paper adds a way to see deferral as a position with a price, one that is occasionally wrong.

Our rate process is a stylised three-state chain whose band and pace of change follow the historical path of the top US rate. We did not estimate it from data, and every figure comes from simulation under a fixed seed. Readers who fear a different rate path can put in their own states; the direction of the results survives that, the magnitudes do not.

Earlier work on harvesting and deferral, Arnott, Berkin and Ye among it, measures the value of tax-aware trading with the rate schedule held fixed. Constantinides, and later Dammon, Spatt and Zhang, solved the timing problem with a known rate. Here the schedule itself is the risk.

Which number deserves attention? The embedded gain. The benefit of deferral grows with it, and the risk grows faster.

Who this is for: Holders of large low-basis positions, CPAs and tax professionals, quantitative researchers, journalists covering tax policy, and analysts weighing changes to capital-gains rates.