What if the polls are wrong? How the Democrats’ ‘blue wave’ could become a tsunami or fade into a ripple

Those who were there have not forgotten it. On election night, November 8, 2016, after months in which countless polls had all but crowned Hillary Clinton the clear winner, the states of Pennsylvania, Michigan and Wisconsin, long considered safely Democratic by pollsters, fell one after another to Republican Donald Trump, handing him the presidency.Although Clinton won the popular vote, the polls had missed the mark spectacularly, and the outcome caught much of U.S. society, the media and the political establishment by surprise. In the aftermath came a litany of postmortems and mea culpas from the U.S. polling industry, which admitted it had failed to grasp the mood of the elusive Trump voter.Now, a decade after that statistical debacle, polls are forecasting a so-called “blue wave” in the November 3 midterm elections. In surveys measuring the generic congressional ballot, Democrats lead Republicans by eight points. That is roughly the same margin they held in 2018, when they regained control of the House of Representatives with a net gain of 41 seats. Democrats therefore have plenty of reason to feel optimistic about their chances of winning it back again.Things also look promising for the Democrats in the Senate. Democratic candidates are leading in states that Trump won comfortably in 2024, including Texas, Ohio, Iowa and Alaska. But their advantage in those races is so small that it falls within the margin of error.And on the eve of an election, experts say, it is worth remembering that a poll is only an estimate and that a small statistical miss could turn a potential “blue wave” either into an electoral tsunami that hands Democrats both chambers of Congress or into little more than a ripple that gives them control of the House of Representatives alone.How large are polling errors?In the last five midterm elections, the final polling average missed the actual result by between 0.2 and 3.5 percentage points. An error of two or three points may seem small, but what matters is that polling errors are rarely random, explains Charles Franklin, a prominent pollster who heads the respected Marquette Law School Poll in Wisconsin.When polls miss the mark, they tend to do so in the same direction across the country. In other words, if polls are currently underestimating Republicans in the Senate race in Texas, they are likely doing the same in Ohio and Iowa. That is why a single polling bias can decide several races at once, benefiting the same party. “Most of those battleground states end up all moving in the Republican direction or all moving in the Democratic direction,” explains Franklin.Several recent elections illustrate this cascading effect. In the 2006 and 2014 midterms, polls overestimated Democratic support; in 2010 and 2018, they overestimated Republicans. The same pattern has appeared in presidential elections: polls underestimated Barack Obama in 2012 and Trump in 2016, 2020 and 2024.The chart below shows how those polling misses have played out across recent election cycles.Why does this happen? One reason a “wave” can become a tsunami or fizzle out in the midterms is that local races are closely tied to national political forces, which act like a prevailing current, explains Charles Franklin.“Presidential approval, the state of the economy, how you feel about the war with Iran… those national forces push all those races in one direction,” he says. That is why, if polls misjudge the national mood, the bias tends to be repeated across multiple contests.This time, those “national forces” appear to be hurting Trump’s party. His approval ratings are at historic lows because of inflation, gasoline prices and broader economic concerns. It is even possible, Franklin says, that this anti-Trump sentiment is being overrepresented in polls. He warns that pollsters may be picking up “too many” Democrats who are “happy to take a survey and tell you how much they dislike the president.”What to expect in the House of Representatives?The two chambers of Congress do not respond in the same way to polling swings.In the House of Representatives, the Democratic lead on the generic ballot is so large and so consistent that Republicans could only hold on to their current majority if polling errors exceeded those seen in any of the last nine elections.At this stage, Franklin says, the real question is not whether Democrats will win the House, but by how much. If the polls are overstating Democratic support and the party gains fewer than 15 seats, election night will be “disappointing.” A gain of 15 to 20 seats would be “plausible”; 20 to 25 would be “pretty good”; and a sweep of 30 seats or more would amount to “a great night.”And in the Senate?Here, Republicans begin with an advantage. Of the 65 senators who are not up for reelection this year, 31 are Republicans and, in the event of a 50-50 tie, the deciding vote belongs to Vice President J. D. Vance.Several competitive races are being decided by margins of less than three percentage points, and an error of the kind seen in the 2014 midterms or the 2020 presidential election would be enough for Republicans to retain control of the upper chamber.Again, everything depends on the direction of any polling bias, Franklin says. If the polls have overestimated Democratic support, particularly in traditionally Republican states such as Texas and Kansas, then by the morning of November 4 we could find that Trump’s party has held on to the Senate. If, on the other hand, the polls are underestimating Democrats, they could also capture the upper chamber, potentially ending up with a majority of 52 or 53 seats. Both scenarios, though opposite, are plausible.There are precedents for that as well. In the 2012 presidential election, for example, polls underestimated Obama by more than three percentage points.If Democrats outperform their polling numbers by a similar margin this year, the wave really could become a tsunami. The Democrats would win districts that Trump carried by more than 10 points and put states such as Kansas, South Carolina and Mississippi into play, conservative strongholds that should not, in theory, be giving Republicans headaches. Yet Trump has been forced to campaign in them precisely because of his party’s weak polling numbers there.How to read the polls with cautionNone of these scenarios should be interpreted as a prediction, Franklin says. Recent history suggests that the real question is not whether polls will be wrong, but by how much and in which direction. Polling errors can vary from state to state, and calculations like these do not account for the individual strengths of incumbents or the personal appeal of specific candidates. Faced with the daily flood of polling averages, Franklin advises against treating polls as crystal balls, “unless you happen to know a good witch in your neighborhood,” he jokes. Rather than focusing solely on whether a candidate leads by one or two points, a razor-thin margin, he recommends paying attention to the spread of the data. When polls in the same state point in different directions, with some showing one party ahead and others favoring its rival, the only firm conclusion is uncertainty.His recommendation on the eve of the elections? “Be uncertain.”