Predicting future economic growth and the outcome of elections have something in common: both rely on assumptions.
We economists make assumptions about when and by how much the Federal Reserve will increase interest rates or whether the conflict in Iran will lead to continued increases in oil and gasoline prices. Our assumptions affect how fast we expect gross domestic product to grow in the future.
Similarly, with the midterm election around the corner, there are new polls coming out almost every day. The results of political polls vary greatly depending on the assumptions the pollsters make.
Ideally, the sample of poll respondents reflects the demographic makeup of the population. That almost never happens.
So pollsters should weight the sample results of their survey to reflect reality.
For example, an estimated 20.5% of the U.S. population is Hispanic or Latino, according to the U.S. Census Bureau.
If only 10% of the people who answer the survey are Hispanic or Latino, the pollster may need to give those responses greater weight. But determining the appropriate weight for an election poll is more complicated than simply matching the overall population. The pollster must also make assumptions about who is likely to vote. If Hispanic voters are expected to turn out at a different rate than other groups, that expectation should be reflected in the model.
Other demographic characteristics typically considered include age and gender. Other demographics that are more debated in terms of the ability to predict elections are party registration and educational attainment.
However, just having a representative national population is not enough under our election system, especially when predicting winners of the presidential election, since the U.S. uses the Electoral College system to elect the president, rather than the national popular vote.
That is what happened in the 2016 election when Hillary Clinton was expected to win.
The American Association for Public Opinion Research wanted to learn what went wrong when it commissioned a study that was released in mid-2017.
The study found that national polls were generally correct. Collectively, they expected Clinton to win by 3 percentage points. In fact, she won the popular vote by 2 percentage points.
One factor that contributed to polling errors in 2016 was how pollsters weighted respondents by educational attainment. In several key states, college graduates were overrepresented in polls, and some pollsters did not adjust for education. Because voters with different levels of education had significantly different candidate preferences, this contributed to an underestimation of Trump’s support.
As a result, polls underestimated Trump’s support in several key states, including Pennsylvania, Michigan and Wisconsin. Pollsters also understated Trump’s support in 2024.
Another assumption that makes polling difficult is determining who will show up to vote. And that makes predictions even more difficult in a midterm election.
The pollster can ask the person answering the survey about their likelihood of voting. Or, the pollster can use historical voter turnout on the assumption that past behavior is a good predictor of the future.
With all the data from the survey gathered, the pollster’s weighting decisions can have a large influence on the predicted election results.
Do we assume the same percentage of Black voters will vote in 2026 as they did in 2024 or 2022? Will female and Hispanic turnout be higher? How does the likelihood of a person voting change based on a strong economy or still elevated prices at the grocery store?
Based on the assumptions that are made, well-regarded pollsters will predict different outcomes of the midterm elections.
Like economists, pollsters continue to learn from history and adjust their models. I’ve yet to see a perfect economic forecast. I expect the pollsters’ predictions this November will not be perfect either.
Christine Chmura is CEO and Chief Economist at Chmura Economics & Analytics. She can be reached at (804) 649-3640 or receive e-mail at chris.chmura@chmura.com.
October 2026
