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 №7·Week of 2026-09-10 to 2026-09-17·published thursday 07:00 BRT

What the prediction market did

The eight midterm contracts hold USD 20.49M, up USD 1.29M on the week. Both chambers moved toward Democrats, and not in the same way: the Senate changed on six days, the House crossed between the same two values for five.

Senate · Democrats

59.50%

↑ 7.00pp on the week

On USD 2.76M. It sat at 52.50% through September 12, rose to 58.50% on the 15th between 15:00 and 19:30 UTC, and reached 59.50% on the 17th. It is the highest reading this series holds.

Senate · Republicans

40.50%

↓ 8.00pp on the week

On USD 1.83M, the thinnest of the four control books. It is the lowest reading this series holds, below the 41.50% of September 15 and 18.00pp under its own top of 58.50%, from June 27.

House · Democrats

89.50%

↑ 3.00pp on the week

On USD 6.59M, the deepest of the eight books. It crossed between 86.50% and 87.50% five times from September 11 to 16, then rose to 89.50%, matching the series top of August 30.

House · Republicans

11.50%

↓ 2.00pp on the week

On USD 4.81M. Issue No. 6 reported it at 14.50%, the highest that series held; it fell 2.00pp on September 16 and now sits 1.00pp above its floor of 10.50%, from August 30.

The fifth published price, on whether the election happens on schedule, reads 97.65%, up 0.35pp, on USD 0.41M, 0.05pp below the series top of 97.70%, from September 13. Prices are from the certified capture of September 17, 17:37 UTC. 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. Six rounds from five institutes sit inside the thirty-day window, averaging D+5.98 against D+5.17 in Issue No. 6, with the newest fieldwork closing on September 15.

Widest

Big Data Poll / Public Polling Project · D+11

Narrowest

Harvard/Harris · R+2

Spread between institutes

13 points

Big Data Poll polled 2,633 likely voters from September 13 to 15, margin 1.7 points, the only round fielded after September 11. Harvard/Harris polled likely voters through August 30 and is the only round putting Republicans ahead. Between them: CBS News/YouGov at D+8, Quantus Insights at D+6.9, two Economist/YouGov rounds at D+6.

What the press reported

What the fixed-list outlets covered, September 11 to September 17

The week began with the Republican convention and ended with an empty chamber. NPR and the Washington Post wrote on September 11 about empty seats and the takeaways from the convention. On September 16 both the Washington Post and CBS News reported that the House was going home ahead of the midterms even earlier than planned.

The Senate coverage turned competitive. On September 12 Fox News wrote that Democrats are within striking distance in the Senate but should not celebrate yet, and Politico reported growing Republican anxiety over the North Carolina race. On September 14 The Hill carried Joe Manchin endorsing Susan Collins in Maine, and on September 16 CBS News covered Senate Democrats rallying around their Michigan nominee.

Two pieces on September 15 dealt with the mechanics of the majorities. NBC News wrote that the Supreme Court has been pushing back against the president and his party ahead of the elections, and The Hill carried a Decision Desk HQ estimate that Republicans may gain nothing from redistricting. The same outlet reported a Texas Senate poll with the Democrat three points ahead.

Our measurement: The market moved with both. The Democratic Senate price rose from 52.50% to 58.50% on September 15, three days after the first piece and the same day as the second, and reached 59.50% on the 17th; the Democratic House price rose 3.00pp on the week, 2.00pp of it late on September 16. Neither move is attributed to either story: publishing before a price moves is not a cause.

The crossing of the week

The chamber that moved most is the one we cannot check against a poll

The week's largest move, 7.00pp toward Democrats, sits in the one book with no polling number beside it: the national instrument asks a House question, and the Senate is 33 separate state races. We never subtract a market probability from a polling margin, and here there is nothing to subtract from.

One chamber transferred a point, the other created one

The Senate pair summed 101.00% a week ago and 100.00% today: the Democratic side gained what the Republican side lost, and normalizing moves the gain only from 7.00pp to 7.52pp. The House pair went from 100.00% to 101.00%, the book paying more than 100 cents for two outcomes, and normalizing shrinks that gain from 3.00pp to 2.11pp. Same direction, different quality.

The average rose because one poll arrived and five left

Five rounds fielded around August 17 aged out of the thirty-day window on September 17, and one entered: Big Data Poll at D+11, the highest in the set. Without it the average would read D+4.98. The electorate did not change between D+5.17 and D+5.98. The set of rounds we can read did.

How to read this number

Why the Senate number has no poll next to it

Every week this newsletter puts a market price beside a polling average, and this week the largest move has no counterpart. That is not a gap in our collection. The national polling question asks about the reader's own district, so it speaks to the House; the Senate is decided in a third of the states at a time, each with its own candidates and its own state polls. Read 59.50% as what the market says about who controls the chamber, never as a national vote share. If someone offers you the difference between that number and a polling margin, what they are offering has no unit.

Sources

Prediction market, certified capture

Polling, primary sources

Press, outlets cited in sections 4 and 5

Method

Market prices come from a certified capture: two independent readings eight minutes apart, published only when they agree within 0.20 percentage points, contract by contract. Series highs and lows are measured against the full recorded series in our daily database backup, not against the public history route, which caps its window at 90 days and would hide the start of the Senate series. The polling average is the simple mean of every national round fielded in the last thirty days, one round per institute per wave, choosing likely voters over registered voters over adults when a house publishes more than one screen. Press coverage comes from a fixed list of 23 outlets, at most two items per outlet, with no political labelling of any outlet.

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. 7 · The Senate moved seven points in a week, on the chamber no national poll measures | AFOS Analytics