Trump Approval Rating

Three books settled on Trump's approval rating for Friday August 21. They settled on three different numbers.

Kalshi's weekly Range resolved on 39.4 to 39.6. Kalshi's own VoteHub binary for the same Friday—Above 38.7%—resolved No. And Polymarket's August 21 book still carries 70.5% on the 38.0–38.4 bucket, with every rung above it quoted at a tenth of a percent.

Two of those are on the same exchange and they are incompatible by more than a full point. Nobody misprinted anything: each contract settles against a different polling average, and Kalshi lists two of them side by side without the difference being visible from the ticker. The lesson is not that one book is wrong—it is that "Trump approval" is not a number, it is a choice of aggregator, and the choice is worth roughly a point and a half.

That gap is stable, not noisy. It reappears this week: Kalshi's August 28 Range puts its mode right back on 39.4–39.6 at 24.5%, while Polymarket's August 28 book prices 39.0 or better at 6%.

The August 28 Range is the anchor, alongside the two books that price the same Friday off different averages. Kalshi's Range settles at 12:00 PM ET; the VoteHub binary asks whether the VoteHub average is above 38.4%, which is where it printed last week.

Note the shape of the Range book: it sums to 109%, so the raw percentages overstate. Normalized, it puts roughly 84% on 39.1 or better and 4% below 38.8.

Volume on the Range is spread evenly across the middle, which is unusual—$1,620 on the 39.4–39.6 mode at 24.5%, $1,161 on 39.1–39.3 at 20.5%, $1,112 on 40.0–40.2 at 11.5%. No single rung dominates, and the whole book turned $7,484.

The two contracts below are the pair that makes the cross-exchange gap concrete: Kalshi's second-most-likely bucket at 23%, and Polymarket's most-likely bucket at 48%—which sits a full point lower. Polymarket's book sums to 146%, so read its levels as relative, not absolute.

The 2026 tails, where the money actually parks. These two Cumulative chains turn almost nothing—$173 and $26 a day—against $174,681 of combined open interest. This is a position book, not a trading book.

The low chain implies a 2026 floor near 38.0, and its deepest single position is $28,088 on Below 38% at exactly 50%. The high chain never crosses 50% at any strike: even Above 43% is only 10.5%, and it holds $12,222.

Both chains still list a 40%-or-below rung as live. Manifold and Polymarket have already closed theirs—their books treat 40% as a level approval has visited and left.

The settled record, kept because it is the cleanest evidence of the split.

Kalshi's Range resolved 39.4 to 39.6—the bucket it had priced at roughly a quarter going in. $28,511 of open interest sat on 38.8–39.0 and lost, more than the $8,735 on the winning rung. On the same day and the same question, Kalshi's VoteHub binary resolved No on Above 38.7%, and Polymarket's book converged on 38.0–38.4.

One Friday. Three settlements. Roughly a point and a half apart.

• Whether Polymarket's August 28 book is right again. It called last Friday's print and Kalshi's Range did not. Two weeks is not a pattern, but a third would start to look like one.
• The Kalshi Range's anchoring. Its mode this week is exactly the bucket that settled last week. If the average has genuinely drifted, that book will be slow.
• The VoteHub threshold. Above 38.4% at 44% is close to a coin flip and settles four hours before the Range does—so it prints first and is the early read on the Range.
• The floor at 38. The low chain has held an implied 38.0 for weeks while the weekly books print 39-and-change on Kalshi's average and 38-and-change on Polymarket's. One of those two is wrong about the level, and the Cumulative chain is quietly siding with Polymarket.
• Nothing here is arbitrage. The three books settle against different sources. The spread is information about measurement, not a trade.