AFOS Weekly · USA-2026 midterms Political Risk Weekly

AFOS Analytics

Real-time political pricing
Prediction market × polls × press
no smoothed averages

Harvard Dataverse · DOI 10.7910/DVN/XRUT8U
United States·Issue №6·Week of 2026-09-03 to 2026-09-10·published thursday 07:00 BRT

What the prediction market did

The eight midterm contracts hold USD 19.20M, up USD 0.62M on the week. The chambers did not move alike: the House changed on five separate days, two of them on one side only, while the Senate changed once, on September 8.

House · Democrats

86.50%

↓ 3.00pp on the week

On USD 6.04M, the deepest of the eight books. It fell on September 4, 5 and 7, took a point back on the 9th and gave it away on the 10th, ending 1.00pp above the series floor of 85.50%.

House · Republicans

14.50%

↑ 3.00pp on the week

On USD 4.48M. It rose on September 5, 7 and 10, and 14.50% is the highest reading this series holds, above the 13.50% of July 29 it had matched since September 7.

Senate · Democrats

52.50%

↑ 1.00pp on the week

On USD 2.50M. The whole move happened on September 8 and the price has not changed since. It sits 5.00pp below the series top of 57.50%, from April 16.

Senate · Republicans

48.50%

↓ 1.00pp on the week

On USD 1.74M, the thinnest of the four control books, 10.00pp below its own top of 58.50%, from June 27.

The fifth published price, on whether the election happens on schedule, reads 97.25%, up 0.45pp, on USD 0.40M, and it is the highest this series holds. The House series opens on July 29, 2026 and the Senate on April 14, 2026, so superlatives here mean since those dates.

What the polls did

The disagreement comes before the average, because it is the finding. Eleven rounds from nine institutes sit inside the thirty-day window, averaging D+5.17. The newest fieldwork closed on August 31, ten days ago.

Widest

Emerson College · D+8

Narrowest

Harvard/Harris · R+2

Spread between institutes

10 points

Emerson polled 1,000 likely voters through August 17. Harvard/Harris polled likely voters through August 30 and is the only round in the window putting Republicans ahead; the same wave among registered voters reads D+2, one survey with two screens and opposite signs. New here is Quantus Insights, 1,200 likely voters through August 28 at D+6.9, read from the pollster's own tables.

What the press reported

What the fixed-list outlets covered, September 3 to September 10

One story ran through the week, about maps rather than candidates. On September 3 the Missouri Supreme Court paused the new Republican map: Politico wrote that it likely nets Democrats one House seat, Roll Call paired it with a Maryland ruling, the Washington Post followed on September 4. On September 10 the Supreme Court refused to let Missouri use it, covered within hours by NBC News, Fox News and The Hill.

The Senate coverage pointed one way. ABC News reported on September 8 that Republicans once confident of holding the chamber now face a wider fight, and NewsNation carried Decision Desk HQ the same day calling the Senate a pure toss-up. Reuters had listed on September 7 the nine races that decide it.

The Republican convention in Dallas closed the week, with The Guardian on September 9 on the president center stage amid party concern, after Axios and the New York Times on September 5 and 6 on a campaign built around him. Underneath: Gallup in the New York Times on September 5, finding more Americans identifying as Democrats, and oil above USD 100 a barrel in the Washington Post on September 9.

One outlet wrote about the instrument itself: the Washington Post, on September 10, a guide to reading midterm polls.

Our measurement: The market moved with the first claim and against the second. The Republican Senate price fell 1.00pp on September 8, the day both outlets published, and stands 10.00pp below its June top. On the House, where the courts moved a seat toward Democrats, the Democratic price fell 3.00pp.

The crossing of the week

The clearest news of the week and the price went opposite ways

The seat is the unit the House is won in, and two courts moved one toward Democrats between September 3 and September 10. Over those days the Democratic price fell from 89.50% to 86.50%. Both facts are measured; the link between them is not, and one seat is small against a 218-seat majority. A reader expecting the rulings to show as a higher price will not find that here.

The House did both kinds of move in the same week

Issue No. 5 separated a point transferred from a point added. The House did both. The Democratic side fell alone on September 4, closing the pair from 101.00% to 100.00%, and rose alone on the 9th, reopening it to 101.00%. On the 5th, 7th and 10th each point came off one side and onto the other, pair unchanged. The Senate did only the clean kind.

A round that already existed entered our average this week

Quantus Insights published a national round fielded through August 28 and the index we read never received it. It sat inside our thirty-day window all along and outside our average until today. It carries D+6.9 and moved the average from D+5.00 to D+5.17. Nothing changed in the electorate between those two numbers. What changed is what we could read.

How to read this number

A thirty-day average moves on its own, and this week it moved 0.90pp

Our number is the simple mean of every national round fielded in the last thirty days. The window slides forward daily, so a round inside it yesterday is outside today without one more voter being asked anything. Six left this week, among them John Zogby Strategies at D+11, the highest number in last week's edition. Losing the top of a range pulls the mean down even when every remaining poll says what it said before. When an average moves, the first question is not who changed their mind. It is whether anyone was asked.

Sources

Prediction market, certified capture

Polling, primary sources

Press, outlets cited in sections 4 and 5

Method

AFOS reads three signals and does not merge them: prediction market prices, published polls, and press coverage. The house average of the generic ballot is a simple mean, unweighted, over a rolling thirty-day window, and when one institute publishes more than one cut of the same round only one enters, by the hierarchy likely voter, registered voter, adults. Rounds that a pollster has published and the index has not yet received are read from the pollster's own tables and carry a marker saying so; when the index catches up, the indexed row replaces ours. Market prices come from a certified capture, two readings eight minutes apart that must agree inside 0.20pp before anything is published. Band markets are published only when the bands sum between 95% and 105%; the popular-vote-margin market is currently above 150% and is collected daily for the series without being published as a reading. This brief never subtracts a market probability from a polling lead, because they are different quantities. Open source, Apache 2.0.

AFOS
AFOS Analytics
Global Political Risk Intelligence · GLOBAL BY DESIGN
www.afos-analytics.com · contact@afos-analytics.com · github.com/AFOS-Analytics

Get it by email

AFOS Weekly follows the US midterms every Thursday, written for voters: where the prediction market, the polls and the press disagree, and what each one measures.

No spam. Unsubscribe anytime.

After signing up you choose whether to receive it in English, Portuguese or Spanish.

AFOS Weekly · Issue No. 6 · Republicans reached the top of their House series in the week the courts went the other way | AFOS Analytics