When the dust settles on election night, the story won’t simply be about who won, but who decided the outcome: Which demographic groups shifted the most, stayed home or delivered the decisive votes to the winners?
This year, we’re going to tell the demographic story of the election as we never have before. We will publish a new comprehensive portrait of how each of the key demographic groups in the country voted, called Times Voter Trends. It will break down estimates for the results of U.S. House and Senate elections by age, race, education and gender in every state in the country.
We’ll report our preliminary findings late on election night. Today, we’re retroactively releasing findings from the 2020 and 2024 presidential elections, which offer a baseline for evaluating how voters will shift in the midterm election. You can explore demographic trends in every state, from President Trump’s surge among Hispanics to Kamala Harris’s overlooked resilience among white voters in the Sun Belt.
Times Voter Trends draws on exhaustive data gathered and in some cases generated by The New York Times. It synthesizes the Times/Siena poll, precinct-level election results, voter file data and, of course, estimates from our election night live model, The Needle, into a comprehensive analysis of what happened in the election.
We’ve been making estimates like these internally for a decade. They lurk behind the Times/Siena poll and Needle. We’ve sometimes used the data in articles analyzing voter turnout and demographic trends. Nonetheless, we have never released them comprehensively to the public or reported them so soon after the polls close.
This year, the estimates will be augmented by a new data collection effort: For the first time, in the final days of the election, we will seek to recontact all of the roughly 30,000 people who took part in any Times/Siena poll during the year. We’ll ask them whether they ultimately voted and for whom, trying to catch any late shifts and to better understand the actual turnout. We’ll begin recontacting early voters before the election. We’ll blitz the rest on Election Day.
To make these estimates, we use a statistical model to understand how demographic characteristics drive the preferences of voters. The model allows us to estimate the likelihood that every registered voter in the country voted for the Democrat or Republican or stayed home, using voter file data from L2, a nonpartisan political data firm. We adjust these individual-level estimates to match the actual or estimated result by precinct.
Compared to the exit polls, the Times Voter Trends analysis of the 2024 election shows a larger gender gap, especially among young voters. Similarly the Times analysis shows that President Trump made gains among white voters; the exit polls showed Kamala Harris gaining among them.
While the exit polls have been treated as an authoritative measure of the electorate as voters leave the voting booth, they have had huge problems. In response, the exit polls have transitioned into a kind of hybrid poll, with a substantial pre-election polling component. The transformation of the exit poll — now rebranded as the Voter Poll — seems to have alleviated its previous issues, but the same changes make it much harder to compare the exit poll results over time.
As a result of these shortcomings, several competitors to the exit poll have emerged over the last decade, including studies from Catalist, NORC and Pew Research. Like Times Voter Trends and the rebooted Voter Poll, most of these studies rely on pre-election data collection. The Catalist and Pew studies are worthy of particular note, as they leverage much of the same data as Times Voter Trends. Unfortunately, they are not available until months after the election. Even once they’re published, material differences between the various estimates remain.
None of these methods are perfect. Times Voter Trends is no exception. Like any statistical model, it’s subject to many sources of error. It’s important to be cautious about the results for small subgroups, especially in states in which we did not conduct a Times/Siena poll or where the state’s voter data does not contain party registration or primary vote history. In general, it’s better to look at the preponderance of evidence than any single estimate.
While every analytical approach has limitations, trying to arrive at the best answer matters. The demographic story of the election can reverberate for years, as policymakers, politicians and campaigns devise policies and messages to reward their supporters or win back those they lost. When the numbers are off, the American political system can be less responsive to public opinion. Every additional data point gives us a better shot to get it right.
The post Introducing a New Way to Understand the Election: Times Voter Trends appeared first on New York Times.
