The Economics of the 351 Exchange: Optimal Seeding of a Tax-Deferred ETF under Concentration Constraints

Summary

A $10M position with no basis

Consider an investor who holds $10M of one stock with a basis near zero. Selling realizes an $8.5M gain and costs roughly $2M of tax at the 23.8% long-term rate, before any proceeds are reinvested. Keeping the shares leaves her wealth tied to one company. Is there a middle course? Section 351 of the Internal Revenue Code provides one. She contributes the shares in kind to a newly launched exchange-traded fund, receives fund shares carrying the same aggregate basis, and recognizes no gain at the contribution. Sponsors now offer this in volume: a Bloomberg analysis of SEC filings in July 2026 counted 105 ETFs created this way, holding $22.1 billion at launch.

Our paper asks how much of such a position can go in, and what the deferral is worth once it does.

Each contributor is tested on her own basket

Section 351(e) denies tax-free treatment when a transfer to an investment company diversifies the transferors' interests. The regulations supply the screen. A contributor's basket passes if no more than 25% of her basket sits in any one issuer and no more than 50% in five or fewer issuers.

Who is tested? Each contributor, separately (Treas. Reg. §1.351-1(c)(6)(i)). A broad sponsor pool does not count toward her test. The regulation includes an example on this point, in which one transferor brings $10,000 of a single listed stock, fifty others bring $200 each of other securities, and gain is recognized.

We write the 25%/50% test as linear constraints on contribution weights. Choosing the seed basket then becomes a linear program that maximizes deferred gain.

How much of the position fits

To hold a position worth C at the 25% cap, a basket must be worth at least 4C. So she needs at least 3C of other eligible holdings of her own. The share of the concentrated position she can contribute is her eligible ballast divided by three times the position, and never more than all of it. Index-fund shares she already owns count, on a look-through basis. Ballast bought with cash or credit shortly before the seed is excluded; commentators on the structure identify that pattern, which they call stuffing, as the most likely ground for a challenge.

In our worked example the investor has $13.44M of eligible ballast beside the $10M name. The formula gives 44.8%, and the linear program agrees: $4.48M of the position goes into a basket of $17.92M, and the other $5.52M stays outside the fund, to be held, hedged, sold or donated. Deferring the whole position would need $30M of her own ballast. If her other $10M of already diversified assets were held as index-fund shares, the deferrable share would rise to 78.1%.

What surprised us was how little the sponsor pool does for eligibility. It supplies the control group and the scale a fund needs. After the exchange it dilutes her old name. Her deferrable share, though, is fixed before she joins.

A related bound concerns diversification. With one issuer at the 25% cap, a basket has at most sixteen effective names, however finely the rest is spread. In a large pool her old name becomes a small slice of the fund. In a small pool, or one she dominates, sixteen is her ceiling.

Four ways to handle the same stock

How does the seed compare with a sale? We simulated 200,000 paths and scored four structures on after-tax wealth: the 351 seed, an exchange fund, borrowing against the shares, and selling and paying the tax. Each is judged on its mean and on a certainty equivalent at risk aversion of three, with a base step-up probability of 0.35.

Borrowing gives the highest mean, $35.5M at twenty years against $32.3M for the seed. It also gives the lowest certainty equivalent. When the stock falls the loan still has to be repaid, and at thirty years its tenth percentile is -$18.6M. The seed has the highest certainty equivalent at every horizon; at twenty years it is $23.54M, against $21.88M for a sale and $19.79M for the exchange fund. Against the sale, its advantage is that the up-front tax is never paid, so the whole position keeps compounding. Against the exchange fund it is cost and liquidity: 35 basis points all-in against 150, and no seven-year lock.

Future tax rates and the step-up

Congress sets the rate a deferred gain eventually pays, and it can change it. How high would that rate have to climb before deferring was a mistake? We solve for the breakeven in closed form, and it turns out to hinge on today's rate, the growth factor and the chance of a step-up at death, while the size of the position and its gain fraction scale the stakes without moving the crossing point.

Some figures. With no step-up, deferral over ten years survives any future rate below 38.6%; over twenty, below 55.9%; over thirty, below 71.8%. Put the step-up chance at 0.35 and the twenty-year threshold is 86.0%. Stretch that to thirty years and no rate can make deferral lose. Since 1978 the top federal rate has ranged from 15% to 28%.

That dependence on the step-up caught us off guard. Heirs who inherit the fund shares take a fresh basis, so the tax is never paid.

What the model leaves out

The fund's return in the model is a blend of the old name and an index. Tracking error is ignored. So are the anti-abuse doctrines, sequential seeding across several funds and the open questions around the 80% control test, which the legal papers on the structure discuss. Rules may change as well: Treasury officials began discussing guidance on 351 seeding in 2026, and a change in the regulation, or in how in-kind redemptions are taxed, would remove part of what the structure is worth. Nothing here is tax advice for anyone.

Who this is for: CPAs, tax attorneys and researchers who study concentrated stock, and journalists covering the recent 351 ETF launches.