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 №8·Week of 2026-09-17 to 2026-09-24·published thursday 07:00 BRT

What the prediction market did

The eight midterm contracts hold USD 22.10M at the September 24 close, up USD 1.58M on the week. Both pairs summed exactly 100.00% at each end, so the 3.00pp each chamber moved toward Democrats survives normalization.

Senate · Democrats

62.50%

↑ 3.00pp on the week

On USD 3.12M. It closed the week before at 59.50%, fell on no day until the last, and reached 66.50% on September 23, the highest this series holds since April 14, before giving back 4.00pp in a day.

Senate · Republicans

37.50%

↓ 3.00pp on the week

On USD 2.12M, the thinnest control book. It fell to 34.50% on September 22, the lowest this series holds, and recovered 3.00pp by the close.

House · Democrats

92.50%

↑ 3.00pp on the week

On USD 7.05M, the deepest of the eight books. It did the whole move on September 20, touching 93.50% that morning, and has closed at the same number five days running.

House · Republicans

7.50%

↓ 3.00pp on the week

On USD 5.19M. It reached 6.50% on September 20 at 11:30 UTC, the lowest this series holds, and closed 1.00pp above that floor.

The fifth published price, on the election happening on schedule, reads 98.10%, up 0.40pp, on USD 0.42M, 0.30pp under the series top of 98.40% from September 19. No double reading was taken on September 24: the four control closes were corroborated the next morning by a certified capture that returned all four to the tick.

What the polls did

The disagreement comes before the average, because it is the finding. Thirty-eight rounds from twenty-eight institutes sat inside the thirty-day window on September 24, averaging D+7.98 against D+5.98 in Issue No. 7, with the newest fieldwork closing on September 21.

Widest

The Economist/YouGov · D+14

Narrowest

Harvard/Harris · R+2

Spread between institutes

16 points

The top of the range is the September 18 to 21 wave of The Economist/YouGov, 975 likely voters. We measured that wave and the two before it as a different instrument: the ballot names the actual candidates running in the respondent's district, not the generic question. The index carries all three, and dropping them moves the same window to D+7.71 and the top to D+13. The floor is the Harvard/Harris likely-voter screen through August 30, the only round with Republicans ahead.

What the press reported

What the fixed-list outlets covered, September 18 to September 24

Republican Senate candidates spent the week distancing themselves from the president over the Iran war. On September 21 NBC News reported an Iowa Senate candidate calling for the war to end after voting to let it continue, and The Hill carried her saying Iowans should not foot the bill. On September 22 The New York Times and the Washington Post both wrote that Republicans in tight Senate races wanted it ended, and Reuters covered a war powers vote two days later.

The money and the ratings pointed opposite ways. On September 21 The New York Times reported Republicans entering the final stretch with a significant financial edge and CNN reported Democrats privately alarmed at it. Two days later the Washington Post wrote that the Republican hold on South Carolina is slipping and The Hill reported the Cook Political Report shifting Kansas and South Carolina toward Democrats.

Three pieces dealt with whether the votes will be counted as expected: Axios on September 19 on the Postal Service slowing the mail to cut costs, the Washington Post and Fox News on September 23 on the Missouri congressional map returning to the Supreme Court for a third time, and the Washington Post on September 24 on a Texas voter-registration backlog.

Our measurement: On direction we agree: of the forty rounds in the September 24 window, thirty-nine put Democrats ahead and one puts Republicans ahead by two. What a single poll cannot show is the width, and those rounds run from R+2 to D+14, a spread of 16 points.

The crossing of the week

Three series records were set this week, and the week closed at none of them

House Democrats touched 93.50% on September 20, the highest that book holds since July 29, and Senate Democrats 66.50% on September 23, the highest since April 14. Senate Republicans touched 34.50% on September 22, their lowest. The closes were 92.50%, 62.50% and 37.50%: the net is 3.00pp in both chambers, and the path was larger both ways.

The average rose two points and the electorate may not have moved at all

The published average went from D+5.98 to D+7.98. But the September 17 figure rested on six rounds, and the same window, recomputed with everything the index now holds, reads D+7.71 across 45. Thirty-nine rounds whose fieldwork falls inside a window that had already closed reached the index afterwards. On the same footing, the week moved 0.35pp.

The coverage pointed one way all week and the Senate price went both ways

From September 18 to 24 the Senate coverage on our fixed list ran one way: a seat slipping, two races reclassified, candidates breaking with the president. The Senate price rose 7.00pp through September 23 and gave back 4.00pp in a day. We do not read a price move out of a headline, and we never subtract a market probability from a polling margin.

How to read this number

Why an average dated to a day is not final on that day

A round exists in the world before it exists in the index we read, and it can arrive weeks later while still belonging to a thirty-day window that has already closed. That is why the September 17 window has been revised from six rounds to 45 since we last published it. Nothing about the electorate changed in between; what changed is how much of it we could see. Read the date on a polling average as when it was computed, never as when the fieldwork ended.

Sources

Prediction market, September 24 closes

Polling, primary sources

Press, outlets cited in sections 4 and 5

Method

Market prices in this issue are September 24 closes, the last point recorded that day in our daily database backup. The House series opens on July 29, 2026 and the Senate on April 14, 2026, so every superlative here means since those dates, and weekly deltas are close to close, which is why the House Republican move reads 3.00pp against the 11.50% certified reading published in Issue No. 7. The two-reading certification we normally publish did not run on September 24; the four control closes were corroborated the following morning by a certified capture, two independent readings eight minutes apart agreeing within 0.20 percentage points, which returned the same four values. Series highs and lows are measured against the full recorded series in that 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. 8 · Three series records were set this week and the week closed at none of them | AFOS Analytics