Research published Thursday affirms what many Americans know from experience: The tumultuous 2016 election made Thanksgiving dinner awkward or even excruciating for people in politically divided families.

The new study isn't based on surveys, social media comments or personal anecdotes about families cleaved by the election. It comes from the cellphone data of 10 million people in the United States — data that captured their movements when the big holiday arrived just 16 days after Donald Trump was elected president.

Smartphones preserve data on where and when people move around. The new research, published in the journal Science, harvested that data, which was anonymized to protect individual privacy, and compared it with voting patterns at the precinct level.

The results led to a striking conclusion: At Thanksgiving 2016, a significant percentage of people who had traveled in 2015 chose to stay home in 2016 rather than visit relatives with different political views. Among those who did travel for Thanksgiving, they were less likely to linger for the pumpkin pie, according to the study.

The same researchers drew headlines when they published a preliminary version of their paper last year, but this latest iteration of their work incorporates more data from 2015 and makes a more robust argument that the observed shortening of Thanksgiving gatherings a year later was directly linked to political polarization. A key finding is that people traveling from places with very high levels of political advertisements experienced a more extreme holiday-shortening effect.

“Ads are throwing fuel on the fire,” said Keith Chen, a professor of economics at the University of California at Los Angeles and the lead author of the report.

Many Americans avoid talking politics at family gatherings. In a December 2016 Pew Research Center survey, about one-third of respondents said they had political views shared by zero or only a few relatives. Among the people in these politically mismatched families, 6 in 10 said their families keep politics out of their conversations.

The research by Chen and co-author Ryne Rohla doesn't actually detect whether people talked about politics during Thanksgiving 2016. The cellphone data merely reveal that, broadly speaking, people had less opportunity to talk about politics because of the shorter gatherings on average, and the tendency for some people to stay home rather than travel as they did in 2015.

“Nationwide, 34 million hours of cross-partisan Thanksgiving dinner discourse were lost in 2016 owing to partisan effects,” Chen and Rohla, of Washington State University, wrote.

There are some obvious limitations to this study, in which the conclusions rest atop a scaffolding of assumptions. For example, the partisan beliefs of people who traveled or stayed home were assumed to track the partisan leanings of their home voting precincts. To protect privacy, the researchers tracked people only at the precinct or Zip code level, Chen said: “We do not try and identify where someone lives at the level of a street address.”

The researchers also assumed that people exposed to political ads over the course of many months continue to feel the polarizing effects weeks after the election, said Kathleen Hall Jamieson, a professor of communication at the University of Pennsylvania. She said other research shows that advertising effects “decay very rapidly.”

“They’re making very large assumptions at virtually every stage in this process,” she said of the researchers.

In an interview, Chen acknowledged that it is difficult to isolate the cause of the observed change in behavior captured by the cellphone data.

“It was hard to argue causality, in the sense that families that differ in politics also differ on numerous other dimensions,” he said.

“What’s cool about the study is the use of digital data to observe human behavior in the wild. That’s the kind of novel social science here,” said Brendan Nyhan, a professor of government at Dartmouth who was not involved in the new research.

There is abundant evidence of intensifying political polarization in the American public, with party affiliation looming larger as an element of personal identity. Increasingly, people live in places where their neighbors tend to be like-minded. A striking feature of this intensifying tribalism is that Americans have grown more negative in their views of people affiliated with the other major political party.

Since 1994, there has been a near-tripling in the percentage of Democrats and Republicans who say they have a “very unfavorable” view of the other party, according to a report published last year by the Pew Research Center. A June 2016 report from Pew found that 47 percent of Republicans viewed Democrats as more “immoral” than other Americans, while 35 percent of Democrats felt that way about Republicans. Democrats were more likely than Republicans to view members of the other major party as “close-minded” (70 percent to 52 percent).

“Social media is feeding partisan identity and polarization, and it's feeding the notion that people who aren’t like us politically aren’t worthy of being associated with,” Jamieson said. “The candidacies of Trump and Clinton were both encouraging audiences to believe that the other candidate was morally reprehensible. And that’s new.”

Larry Diamond, a political scientist at Stanford University's Hoover Institution, said of the study by Chen and Rohla, “It’s very intriguing, and sounds like a very innovative study, which is certainly telling us something of sobering importance about the way political polarization is affecting American life.”

He added, “There's nothing happening in American politics now to suggest it will be better at Thanksgiving in 2020.”