Eyeing the 2026 US polls

The litmus test for survey quality, all over the world, is the ability to forecast the outcome of elections. The next big test for pollsters to watch is the United States midterm election on Nov. 3, with 436 House of Representatives seats (435 for the states and 1 for District of Columbia) and 50 of the 100 Senate seats at stake, or 486 races in all.

If every election race needs a poll, and a poll takes a modest sample size of 625 interviews (4 percent error margin), that’s over 300,000 interviews per shot to cover the US. A single shot won’t tell the story of an entire campaign. Covering it requires polling at the beginning (the baseline), the middle, and close to the end. Middle polls are needed to indicate how well or how poorly the steps taken by the contenders have been working. The final poll becomes the predicted outcome that goes into the pollster’s track record.

Of course, political polling starts even before campaigning. It helps politicians decide whether to run for office or not. It gets conducted on people old enough to vote (A, adults), registered to vote (RV), and/or likely to vote (LV). The basic factor of voting turnout is much lower for Americans than for Filipinos, maybe 50-60 percent for them versus 70-80 percent for us. By the way, election day is not a holiday there, unlike here.

To save on costs, many American polls ask a generic question of whether the respondent is likely to vote for a candidate from the Democratic party (D), Republican party (R), or Other party (O). For example, the Washington Post/ABC News/Ipsos polls average of the last two polls of dozens of pollsters, as of 10/30/26, was: D 47.0 percent, R 45.0 percent, O 8.0 percent. The current consensus from the American polling community is that the Democratic party is likely to win control of the Lower House, and just might win control over the Senate also.

President Donald J. Trump, a Republican, is already the least popular president in polling history, and has been setting a new record-low each week. As of 10/29/26, the daily average of polls scoured by the New York Times (www.nytimes.com) is 61 percent disapprove and 36 percent approve. Thus, the more that Trump insists that the coming election is about him, the worse it is for his party.

Trump’s behavior reveals that the American opinion polls bother him very much. The Gallup’s halting of its time-honored series of presidential approval ratings, suggests that it can’t afford to risk its business with the US government. “Trump says he’ll respect midterm results ‘if they’re honest elections’” (usatoday.com, 10/27/26).” “Trump offers $1M for election fraud convictions week before midterms” (www.newsweek.com, 10/26/26).

I continue to watch Nate Silver (www.natesilver.net), the poll aggregator who founded FiveThirtyEight (the number of US electoral college votes), and now publishes the online newsletter Silver Bulletin. His latest model, FLIPR (“Forecast with Leading Indicators, Polls and (Expert) Ratings”) has this 2026 election forecast, as of 10/29/26: “After today’s update, Democrats remain steady favorites to retake the House, and their Senate odds are also holding steady. Still it’s important to remember that Democrats’ leads in the key Senate battlegrounds–Texas (D +2.5), Ohio (D +3.0), Alaska (D +3.6), and Maine (D +2.4)–are all just a normal-sized polling error away from flipping.”

Election polls have hardly any effect on voters, even though they do influence the candidates, donors, and campaigners in an election. Surveys by Social Weather Stations show that a few voters might flip to apparent favorites, and a few might flip to apparent underdogs, and the net effect is very small. Whatever the effect, it would be no more undemocratic than listening to what one’s neighbors say about their voting intentions.

In any case, the political forces that have tried to disrupt election polls have been foiled by the wise Supreme Court rulings that the constitutional guarantee of freedom of speech includes the freedom of both exit polling (G.R. 133486, Jan. 28, 2000) and preelection polling (G.R. 147571, May 5, 2001).

Personally, I hope for the polls for the coming election to be a great success. It would benefit research on measurement of any subjective aspect of human well-being, including the economic aspects, which is my field. It would be good for democracy in the US and in the world as a whole.