What will Democrats' lead be in the congressional generic ballot on September 15, 2026?

As of 2026-09-08, the estimate for Democrats' lead on September 15 is +5.8 percentage points, with a plausible range of +3.0 to +8.5, conditional on using the RealClearPolling national generic-ballot average. The opened average displays Democrats at 48.1% and Republicans at 42.3%; this is a September 8 observation, not a September 15 result. The original question's resolution page could not be verified, so the choice of average remains an explicit assumption. 1

Forecast

• Estimate: +5.8 percentage points (range +3.0 to +8.5).
• Threshold: Democrats lead by at least 3.0 points, 90% probability, plausible probability range 80 to 95.
• As of: 2026-09-08, evidence cutoff and observation date.
• Resolves: 2026-09-15, provisionally using RealClearPolling's national generic congressional vote average at the end of that day, Eastern time.
• Base rate: The observed starting margin is 5.8 points, calculated as 48.1 minus 42.3; no historical seven-day transition sample is claimed. 1
• What would move it: Several comparable new polls above a seven-point lead raise the estimate; repeated readings below three points lower it; confirmation of a different official resolution average requires recalculating it.

Why

1. The target is a published polling average seven days ahead. The estimate therefore holds the observed margin approximately constant rather than adding an unsupported directional trend. The current reading supplies a firmer starting point than an argument about who deserves to win or which party has better campaign messages. 1
2. An independently opened poll supports the scale of the lead. The Economist/YouGov survey fielded August 28 to 31 reports 46% Democratic and 40% Republican support among registered voters in its generic-congressional-vote table. It is one component of the evidence, not an additional independent copy of the aggregate. 2
3. A change in the average need not mean voters changed their minds. Polling organisations use different samples, weighting choices and turnout screens. New polls replacing older ones can move the displayed margin even when each organisation finds little change in its own series. 3

The forecast's range is a judgement about the September 15 number. It is not the sampling margin printed on a single survey and does not describe uncertainty about November's final vote. The lower bound allows a meaningful Republican improvement or a change in the mix of polls. The upper bound allows Democratic improvement or replacement of weaker Democratic readings. Both are possible without a national political realignment.

For a transparent threshold calculation, consider a roughly symmetric predictive distribution centred near 5.8 with a spread of about 2.2 points. A 3.0-point threshold is then about 1.27 standard deviations below the centre, giving approximately a 90% chance of reaching it. This is an illustrative judgement distribution, not an empirically calibrated statistical fit. The +3.0 to +8.5 range covers roughly its central four-fifths, with outcomes beyond both ends still possible.

Evidence for

• The August 28 to 31 YouGov table separates all adults from registered voters. Its all-adult figures are 36% Democratic and 31% Republican, while the registered-voter column gives 46% and 40%. This shows why a headline must specify the population being measured. The registered-voter reading is consistent with a lead near six points. 2
• AAPOR's election resources distinguish probability samples from non-probability samples and emphasise that subgroup results have greater uncertainty than full-sample results. The practical implication is to give the national topline more weight than an apparent surge inside one small demographic group. 4
• AAPOR's reporting guide describes polls as snapshots and explains that uncertainty in a lead differs from uncertainty in one candidate's share. This supports retaining a broad range despite the decimal precision of the displayed average. It also supports resisting a large update based on a single statistically weak move. 5

The affirmative case for the point estimate is deliberately modest. There is an observed Democratic advantage, a recent poll with a similar registered-voter margin, and a short forecast horizon. Nothing in those facts requires a further Democratic gain. The central estimate stays at the measured starting value because no verified evidence here establishes a reliable seven-day drift in either direction.

Evidence against

• Past election polling has shown shared errors across organisations. Pew's overview discusses the unusually large errors in 2020 and warns that close races are difficult to assess precisely. That history weakens any claim that a polling lead guarantees an election outcome, although this question resolves against a poll average rather than actual votes. 6
• A credibility interval for a non-probability survey and a margin of sampling error for a probability sample rest on different assumptions. Combining polls does not make those differences disappear. The forecast therefore does not assign the aggregate a fabricated conventional sampling error. 7
• AAPOR's best-practice guidance asks researchers to disclose sample construction, population, question wording and weighting. A new survey with an unfamiliar method deserves inspection before it changes the forecast. A large sample alone is not enough to establish comparability with earlier polls. 8

There is also a resolution risk that more polling cannot fix. A different average could include different organisations or apply different weights on the same day. Two honest answers could then differ by a point or more. The numeric forecast should travel with its source convention, not as a free-floating prediction about a universally agreed national number.

Assumptions and limits

• The reference supplied with the question is https://www.metaculus.com/questions/45221/. It did not open successfully. This report does not claim that Metaculus names RealClearPolling; that source is a provisional fallback. The forecast must be revised if the original rules specify another provider, time of day or averaging window.
• The aggregate page opened on September 8 showed the displayed figures, but its extracted poll table contained no constituent rows. The full average was not independently reconstructed. The latest individual poll used directly here ended fieldwork on August 31. No survey conducted after September 8 contributes to the forecast.
• Democratic lead means Democratic percentage minus Republican percentage. Undecided voters remain in the denominator used by the source. Renormalising the two parties to 100% would answer a different question and inflate the reported margin.
• This report predicts the September 15 average only. It does not estimate House seats, control of Congress or the November popular vote. District geography and actual turnout would need a separate analysis. A 90% probability of clearing three points is also not a 90% probability of winning an election.

What to watch

• From September 9 through September 15, compare new polls with the same organisations' previous releases. Check fieldwork dates before attributing a movement to a recent event. A poll released later can still describe opinions collected earlier.
• On September 15, preserve the source's end-of-day values and any constituent-poll dates. Confirm the original resolution convention before marking the forecast concluded. If the source is unavailable, leave the outcome pending rather than substituting a different average without disclosure.

Sources

1. 2026 Generic Congressional Vote, RealClearPolling, observed 2026-09-08; live average, https://www.realclearpolling.com/polls/state-of-the-union/generic-congressional-vote
2. The Economist/YouGov Poll, YouGov, fieldwork 2026-08-28 to 2026-08-31, accessed 2026-09-08, https://d3nkl3psvxxpe9.cloudfront.net/documents/econTabReport_yVe1kKt.pdf
3. Polling Accuracy, AAPOR, accessed 2026-09-08; publication day not exposed, https://aapor.org/polling-accuracy/
4. Election Polling Resources, AAPOR, accessed 2026-09-08; living resource, https://aapor.org/election-polling-resources/
5. A Journalist's Guide to Understanding Polls & Surveys, AAPOR, accessed 2026-09-08; publication day not exposed, https://aapor.org/journalist-guide-to-understanding-polls-surveys/
6. Key things to know about election polls in the U.S., Pew Research Center, 2024-08-28, accessed 2026-09-08, https://www.pewresearch.org/short-reads/2024/08/28/key-things-to-know-about-us-election-polling-in-2024/
7. Understanding a credibility interval and how it differs from the margin of sampling error in a public opinion poll, AAPOR, 2012-10-07, accessed 2026-09-08, https://aapor.org/statements/understanding-a-credibility-interval-and-how-it-differs-from-the-margin-of-sampling-error-in-a-public-opinion-poll/
8. Best Practices for Survey Research, AAPOR, accessed 2026-09-08; living guidance, https://aapor.org/standards-and-ethics/best-practices/