Illustration: Rebecca Zisser/Axios
Heightened partisan divisions associated with the 2016 presidential election resulted in a shortening of Americans' Thanksgiving dinners that year, by 30–50 minutes, a new study in Science found.
Why it matters: The research, using precinct-level polling information and a vast database of anonymized cellphone location records, suggests that Americans are less willing to socialize with family members who hold opposing political viewpoints. However, at least one outside expert is raising the alarm on using that level of data for non-government surveillance.
"I have not come across a paper that relied on such precise cellphone data. This strikes me as novel and innovative. It also makes me feel a little uneasy."— Jeremy Frimer, political scientist, University of Winnipeg
The data: The study examines two databases.
- The first is via the company SafeGraph, which looks at anonymized smartphone-location data from more than 10 million Americans, allowing researchers to observe actual travel at precise time and geographic levels. The study examined a database of 21 billion pings from November 2016, as well as 4.5 billion pings from November of the prior year.
- The second is a precinct-level database for presidential voting for the 2016 election, as well as 2015 voting information. That data encompasses 172,098 precincts across 99.9% of counties nationally, the paper states.
What they did: The cellphone data gave researchers Keith Chen of UCLA and Ryne Rohla of Washington State University an idea of which precinct a voter was in, by inferring each smartphone user's home based on the pings between 1am and 4am local time.
- They used the precinct-level voting information, along with Census data, to infer how each anonymous phone number likely voted.
- They incorporated Thanksgiving Day smartphone pings between 1pm and 5pm during both 2015 and 2016 for about 10,000 phone numbers (i.e. voters) that showed up in both years' data.
- The scientists also looked at the effects of the barrage of political advertising that people living in swing states were subjected to during the 2016 election, compared to 2015.
What they found: They found that political advertising likely more than tripled the "Thanksgiving effect" in 2016.
- Thanksgiving dinners attended by people who were likely to be of opposing parties were cut short by 30 to 50 minutes. Nationwide, 34 million hours of cross-partisan Thanksgiving dinner discussion was lost in 2016, the researchers found.
- The Thanksgiving effect was not there in 2015, although the sample size that year was smaller. “That’s not, you know, bulletproof but that’s kind of interesting,” Chen told Axios.
- Cross-partisan Thanksgiving dinners were cut especially short in areas where people were exposed to high volumes of political ads, such as Orlando, Florida, which is a swing part of a swing state.
Privacy concerns: Frimer, who was not part of this study, said the use of the precision cellphone data raises some alarm bells, particularly in the wake of the debate regarding the 2016 campaign political ad targeting methods that were used. He said:
"Using cellphone data for national security purposes (with judicial oversight) seems like a reasonable compromise between needs for privacy and safety. Whether allowing this level of surveillance for commercial/scientific purposes is another matter and worthy of reflection and discussion. Personally, I feel uncomfortable with allowing companies to sell and distribute such personal data for commercial/scientific purposes."
Yes, but: Frimer said the analysis in the paper is "an extremely impressive demonstration of a well-established psychological phenomenon called the selective exposure," or confirmation bias, although he questioned some of its findings.
Study limitations: The study implies that there's real personal damage being done to Americans due to the politically charged atmosphere. “It has personal costs," Chen said. However, the research is only based on two years of data, and the U.S. has been through periods of intense political divisions before.
What's next: The study's authors plan to track Thanksgiving data for 2018 and beyond to see if the partisan divisions harden, or if people get used to this level of political tension and the effects diminish.
“This particular blend of partisanship that is emerging now…I suspect that people are going to acclimate to that," Chen said.