In case you have been living under a rock for a while, we are about to get a new president in the United States. Donald Trump not only won the electoral college but also the popular vote – the first time a Republican has done so since 2004. No matter how you put it, this was a significant victory for Trump and his brand of populism, much bigger than what the 2016 election was. 

Many people are asking the same question now: Why? I spent the last few months looking at US polling data, and I believe I can share some insights into that question. To prevent surprises like this in the future, there needs to be a serious change in how we look at American public opinion.

First, let’s figure out what we are dealing with. Donald Trump has won 312 electoral college votes, more than the 270 needed to win and more than the 306 he won last time. He also won every close state (“swing state”), Nevada (Trump + 3.1), Arizona (Trump + 5.7), Georgia (Trump + 2.2), North Carolina (Trump + 3.3), Pennsylvania (Trump + 2), Michigan (Trump + 1.4) and Wisconsin (Trump + 0.8).

Even more astonishingly, Trump won states that were otherwise hard democratic (“deep blue”) to swing heavily toward him. In addition to the 31 states he won, he managed to get 45% or more of the vote in six other states (New Mexico, Minnesota, Virginia, New Jersey, New Hampshire and Maine). Trump won 12 states by more than 60%. Harris only won four by the same margin. Trump won two states by more than 70%. Harris didn’t win a single state by more than 65%. Unsurprisingly, Trump won the popular vote; he made gains (some larger than others) in nearly every city and state.

In hindsight, there were warning signs. We knew that there were more Republican voters in 2024 than there had been since 2004. We knew that non-university voters were moving away from the Democratic Party. We knew that Trump was doing better with Black and Hispanic Voters. We knew that Republican-led issues, like the economy and immigration, were more resonant with voters than Democrat-led issues like abortion and democracy. Above all, we knew that Kamala Harris was nowhere near as popular as Joe Biden was back in 2020. Anyone who paid attention and did an ounce of research knew this would be a good night for Trump and the Republicans, but the question was by how wide of a margin?

Did the pollsters predict this outcome? Perhaps not, and here are some reasons why.  First, smaller sample sizes tend to do much worse than others. Some polling firms still believe that a sample size of a couple of hundred people can accurately capture the mood of a state with millions of potential voters, much less a country with tens of millions of voters. It is an issue, but many polling firms still do this, particularly at the state level or international pollsters.

Second, polls can be improved by weighting the data based on key demographic factors. The best indicator is party identification, as in such a polarized political environment, party support is nearly unanimous (>90% of Republicans vote for the Republican candidate, >90% of Democrats vote for the democratic candidate, etc.). There are also other demographic indicators you can look at, such as race, university education, community type, and religiosity. Extensive research has been done on all these indicators, and the states have publicly collected data. Despite this, polls could only provide the best estimations of the sample area’s demographics. Sometimes, polls from the same pollsters would differ. Without complete information on these indicators, sampling will differ from poll to poll, making comparing polls impossible. 

Finally, some states did little to no polling throughout the election cycle. Out of 56 total presidential contests (50 states + DC + Nebraska congressional districts + Maine Congressional districts), 35 had less than five polls done throughout the election season. That’s 62.5% of races that had a shortage of polling data. However, most were in “safe” states that were definitely going to turn either Republican or Democrat (and most did), but some states were much closer than previously thought (see the states mentioned above). Consequently, pollsters could not tell if any surprises were in store, mainly regarding the popular vote. If we had known, for example, that Trump was outperforming his 2020 numbers in urban states like New York, New Jersey, or California, maybe pollsters would have hesitated to declare that Harris would win the popular vote by a comfortable margin. 

If American (and foreign) pollsters do not make significant improvements in capturing the country’s political preferences, more surprises like this one will be in store. Since polling acts as a window into public opinion, gauging it is vital and not something that could be left up to chance or random guessing. By shedding light on the mishaps of American polling, we can understand what needs to change so that we will not be as surprised next time.

*This article was written by a guest contributor