The arbitrage band is the range around net asset value inside which an asset backed token trades without anyone correcting it. It is the residue left over after creation and redemption arbitrage has done its work, and its width equals the round trip cost of enforcing the price. Deviation smaller than that cost is not worth trading against, so it stays.
The band is a cost the holder pays, not a tolerance the issuer grants. It is set by the economics of whoever enforces the price, which means an issuer can quote a tight fee schedule and still ship a wide band.
Scroll to see the full diagram
The band is arithmetic, not policy
Add up what a round trip costs the enforcer. The creation fee on the way in. The redemption fee on the way out. Gas and transaction costs on both legs. Custody and settlement friction on the underlying. The cost of capital for however long the position is held. And the market risk carried from order to NAV strike, because primary orders price forward rather than at the level that made the trade look good.
That total is the band. An arbitrageur will not lift a deviation of 8 basis points when the round trip costs 15, so an 8 basis point deviation persists and the holder trading in the secondary market absorbs it. Nothing about this is a malfunction. It is the mechanism working exactly as designed, and the design decides how much it costs.
What the ETF rule contributes to the width
Rule 6c-11 conditions push in the direction of a narrow band without ever naming a number. Creation units transact with authorised participants against a basket, portfolio holdings publish daily, and the fund transacts at NAV.1 Daily holdings publication is the part that most directly narrows the band, because an arbitrageur who cannot price the delivery basket has to charge for the uncertainty.
The Commission frames the whole arrangement as designed to keep market price at or close to NAV per share.2 Close is doing real work in that phrase. The rule sets up the enforcement channel and leaves the width to whatever the enforcement costs, which is why two funds under identical rules can trade with visibly different bands.
Tokenisation compresses the trading spread and leaves issuance alone
The closest primary quantification available comes from the Bank for International Settlements. Studying tokenised government bonds, BIS found a mean bid ask spread of roughly 19 basis points against roughly 30 basis points for comparable conventional bonds, while issuance costs, the gap between what the public pays and what the underwriter pays the issuer, showed no systematic reduction from tokenisation.3
Read that carefully, because it is a bond bid ask study and not a fund premium and discount study. The transferable finding is directional: the part of the band driven by trading friction responds to tokenisation, and the part driven by how the instrument gets originated and distributed does not. An issuer quoting narrower trading spreads as evidence that the whole cost stack compressed is answering a different question than the one asked.
A one way channel produces a one way band
The band is only symmetric if both enforcement legs work. A vehicle that can create units but cannot redeem them has an upper bound and no lower one, so a premium can run as far as demand takes it, because nothing profitable pushes it back. Reverse the asymmetry and a discount can sit open indefinitely.
Asymmetric fees do the same thing more quietly. A redemption fee wider than the creation fee shifts the whole band below NAV, so the token trades under its declared reference on average, by construction. That can be the right call when the underlying is expensive to liquidate. It has to be disclosed, because a holder buying at the reference price is systematically buying into a level the market will trade below.
Widen the band on purpose when the underlying is slow
For a Treasury or money market underlying, the assets liquidate quickly and the band can be genuinely tight. For appraisal driven property or a private credit book, the reference itself updates weekly or monthly, and the enforcer takes real price risk between the order and the strike. Pretending the band is tight in that case does not make it tight. It just moves the loss to whoever trades at the edges.
The observability question matters as much as the width. A holder can only see the band if NAV is published on a comparable cadence to the price. Superstate's USTB publishes both, with net asset value at $11.16 per token against $757,757,779 in total assets on 3 August 2026, on a continuously updating dashboard.4 Read those as point in time figures. Where NAV is published monthly and the token trades continuously, nobody can measure the band at all, which is a disclosure problem before it is a pricing one.
What we ask issuers to compute before they publish a fee schedule
Model the realistic round trip first, then set the fees. Take a participant who actually exists, price their gas, custody, capital and settlement lag, and see what deviation they need before the trade clears their hurdle. That number is the band you are shipping, whatever the marketing says. If it comes out wider than the tracking error you promised, the honest fixes are to cut a fee leg, shorten the settlement window, or widen the promise.
Then write the band into the offering documents as an expected range with the assumptions behind it. In our view an issuer who cannot state their expected band has not modelled their own mechanism, and that gap shows up later as a tracking error nobody budgeted for. The band is not decoration on top of the product. It is part of the price of holding it.
Common questions
What sets the width of an arbitrage band?
The round trip cost of enforcing the price. Creation fee, redemption fee, gas on both legs, custody and settlement friction, cost of capital, and the market risk carried between the order and the NAV strike. Any deviation smaller than that total goes uncorrected because correcting it loses money. The underlying's liquidity is the biggest single driver, since it sets how long capital is tied up.
Is a wide arbitrage band a sign of a broken design?
Not by itself. A wide band on a slow, illiquid underlying is honest engineering, since the enforcer genuinely carries more cost and risk. A wide band on a liquid underlying points at fee calibration, a thin eligible participant set, or a redemption channel that is slower than advertised. The problem is a band wider than the tracking error the issuer promised, whatever the cause.
How is the arbitrage band different from slippage?
Slippage is the price impact of one order against available liquidity at the moment it executes. The band is a standing range around NAV that persists regardless of order size, because it reflects the cost of the primary market round trip rather than secondary market depth. A holder can face both at once: they cross the band to get to market, then pay slippage on top.
Can an issuer set the band to zero?
No. Zero fees on both legs still leave gas, custody friction, capital cost and the risk carried into the forward priced NAV strike, and someone has to earn all of that back. An issuer advertising perfect NAV tracking is describing an outcome nobody is paid to produce. The useful question is whether the expected band is stated, and whether the assumptions behind it are written down.
See RWA Tokenomics Design for how this applies in practice.
Sources
- 17 CFR 270.6c-11, Exchange-traded funds
Electronic Code of Federal Regulations, current
Conditions supporting the arbitrage channel: creation units transacted with authorised participants against a basket, daily portfolio holdings publication, transactions at NAV. - Exchange-Traded Funds, Release Nos. 33-10695; IC-33646 (Rule 6c-11 adopting release)
U.S. Securities and Exchange Commission, 2019
The Commission's framing that the arbitrage mechanism is designed to keep market price at or close to NAV per share, without specifying how close. - Tokenisation of government bonds: assessment and roadmap (BIS Bulletin No. 107)
Bank for International Settlements, 2024
Mean bid ask spread of roughly 19 basis points for tokenised bonds against roughly 30 for comparable conventional bonds, with no systematic reduction in issuance costs. A bond spread study, cited here for band width drivers rather than fund premium and discount. - USTB: Superstate Short Duration US Government Securities Fund, asset page
RWA.xyz, 2026
Net asset value of $11.16 per token against total asset value of $757,757,779, read 2026-08-03. A live product publishing NAV alongside a traded token. Figures update continuously.
Last reviewed 2026-08
Know the terms but not sure how they apply to your project? That is what an engagement is for. We design, document, and stress-test the whole token economy inside the Tokenomics Data Room.
100+ projects advised. Complete tokenomics in 4 to 6 weeks.