What the polls do (and don’t) tell us about Election Day

The Decision Desk HQ Forecast Model illustrates possibilities, not certainties. It serves as a tool to understand the range of potential electoral outcomes. With less than six months until Nov. 5, this explainer kicks off a series that will provide qualitative context to our quantitative projections.

Our latest update finds that if the election were held today, Donald Trump would have a 56 percent chance of defeating Joe Biden. (Note: This does not yet reflect a substantive sample of polling after Trump’s conviction in Manhattan.) In 100 hypothetical elections, Trump would win 56 times. Republicans are also favored to flip the Senate with a 77 percent probability and keep their majority in the House with a 62 percent probability.

Biden still has a 44 percent chance of winning the presidency, mirroring the odds faced by the Kansas City Chiefs against the San Francisco 49ers in this year’s Super Bowl (and in case you forgot, the Chiefs won). Democrats have a 23 percent chance of holding the Senate — roughly the same as flipping a coin and getting heads twice in a row. Democrats also have a 38 percent chance of taking the House, equivalent to the probability of Steph Curry making a three-point shot during the Golden State Warriors’ championship-winning 2021-22 season.

In other words, the GOP shouldn’t start measuring drapes for the Oval Office or drafting their reconciliation bills just yet.

Most of the fundamental factors in our forecast, including the incumbency advantage, campaign fundraising and various macroeconomic indicators, favor Biden. But national and swing-state polls show a clear advantage for Trump. According to Decision Desk HQ and The Hill’s polling averages, Trump leads Biden by 1 percentage point nationally across 727 polls — a significant shift from Biden’s 8.4-point lead in FiveThirtyEight’s final national polling average in 2020.

Our swing state polling averages show Trump maintaining a consistent lead in each of the six states that Biden won by under 3 points in 2020: Arizona, Georgia, Michigan, Nevada, Pennsylvania and Wisconsin. Biden and Trump are statistically tied in Minnesota (Biden +7 in 2020), Virginia (Biden +10) and New Hampshire (Biden +7).

The deeper you look into the polls, the worse it gets for Biden. The president is viewed less favorably than his predecessor, with few voters crediting Biden’s role in the economic upturn. Voters also trust Trump more on key issues like immigration, jobs, the economy and foreign policy.

Perhaps most concerning for Biden: During his 13th quarter in office, from January 20 to April 19, his job approval rating averaged 38.7 percent — the lowest among first-term presidents since Dwight D. Eisenhower in 1956.

Biden’s campaign has dismissed polling as “broken,” cautioning his supporters against relying too heavily on polls surveyed five months before an election. Indeed, the election cycles of 2016, 2020 and 2022 have shown that polls can be subject to systematic errors, potentially skewing results by several percentage points for or against a candidate. These errors tend to run in the same direction across different states and regions, suggesting that current polls are either overestimating or underestimating Biden by a few points across the board.

Nevertheless, it is risky to assume that the polls will shift in one’s own favor. For instance, Trump outperformed polling averages in most swing states in 2016 and 2020, yet Republicans underperformed the polls in most key races for Senate and governor in the 2022 midterms.

This is not to say that we should dismiss the polls entirely. They remain an integral part of our model, and their weight in relation to fundamental factors will increase as we approach election day; historical data shows that they tend to be more accurate closer to November. With that said, if Biden is trailing by similar margins in the polls around Labor Day, he would be considered a significant underdog.

Biden’s current position is particularly weak in the Sun Belt swing states — Arizona, Georgia, Nevada and North Carolina. He has consistently lagged in the polls by 4 points or more in these states, even as he has polled within the margin of error in Michigan, Pennsylvania and Wisconsin for much of the campaign. In 2020, strong minority turnout and elevated support from white college-educated suburban voters secured Biden narrow victories in Arizona (Biden +0.3) and Georgia (Biden +0.2), marking the first time a Democrat has won these states since 1996. But early 2024 polls have consistently shown Trump gaining among Black and Hispanic voters, particularly men. These swings, combined with lower overall turnout relative to 2020, could tip Arizona and Georgia back to Trump.

If Trump’s most efficient path to the White House is to flip these three Sun Belt states along with Wisconsin, Biden’s best shot at a second term runs primarily through the Rust Belt.

Democrats have overperformed electorally in the Rust Belt since the Supreme Court overturned Roe v. Wade in 2022. In the midterms, Democrats swept the competitive gubernatorial races in Pennsylvania, Michigan and Wisconsin, outperforming FiveThirtyEight’s final polling averages by more than 3 points in each race. Meanwhile, Democratic Senate candidates John Fetterman in Pennsylvania and Mandela Barnes in Wisconsin outperformed their final polling averages by 5 and 3 points, respectively.

Sun Belt states are generally more religious and thus more socially conservative. A Pew Research Center survey found that in Michigan, Pennsylvania and Wisconsin, adults favoring the legality of abortion in all or most cases exceed those opposing it by margins of 12, 7, and 8 percentage points, respectively. In contrast, this support is narrower in Arizona and North Carolina, at 3 and 4 points, and it reverses in Georgia, where opponents outnumber supporters by 1 point.

Even if Trump sweeps the four Sun Belt swing states, if Biden manages to hold the three Rust Belt swing states — along with the rest of his less-competitive 2020 states, and one electoral vote from Nebraska —  he would hit exactly 270 electoral votes.

Ryan Gest is a data science and digital content fellow at Decision Desk HQ.