Online Public Opinion Polling in U.S. Elections; The Impact of Social Distrust on Public Opinion Polling

The purpose of the technical and STS papers in this project is to describe the current state of public opinion polling in the United States. This encompasses both the various challenges the field of polling is confronting right now, as well as the innovations being developed in response.
For the technical project, the focus is on the relatively recent innovation of online polling methodologies. This includes the circumstances that prompted their development, what sets them apart from traditional polling methodologies, and the merits of their use. For the STS project, the focus is on the rising levels of social distrust in the United States, and the various impacts this cultural shift has had on public opinion polling. These two topics are loosely tied, but share a lot of information that is relevant to them. After all, one indirect result of rising social distrust in the United States is the proliferation of online polls.
As described in both papers, the main reason online polls now make up more than 80% of public opinion polls is because traditional phone-based polls have become much more expensive in the last few decades. One reason is a drastic decrease in response rates for phone-based polls; that is, the percentage of people who, when asked to complete a survey, actually complete it. A 2020 survey from Pew Research Center showed that eighty percent of Americans do not answer cellphone calls from unknown numbers. As such, it takes many more calls for pollsters to collect the same number of responses compared to a few decades ago, making the entire operation cost much more. In addition, the widespread transition from landline phones to cell phones has also made polling harder and more expensive, with one pollster estimating that each cell phone interview costs about three times as much as a comparable landline-based interview. With traditional polls becoming so expensive, many pollsters innovated and switched to significantly cheaper online polling methodologies.
The technical paper delves into the details of online polling methodologies. The main difference between phone-based polls and online polls, besides the medium for interviewing, is that phone-based polls use probability samples, where every member of a target population has an equal chance of being asked to participate, while online polls use non-probability samples, in which there does not exist an equal chance of being asked to participate. In the past, online polls were judged by polling experts to be less accurate and of lower quality than phone-based polls. However, ever since the 2020 U.S. elections, many polling experts found that online polls were no longer worse than phone-based polls. As such, the technical paper concludes that, despite their problems, online polling methodologies are a very viable solution to the rising cost of phone-based polls, and their widespread use should not be a cause for concern. Future studies in this field could involve the difference between online-only methodologies, hybrids of online and phone-based methodologies, and text-messaging based methodologies.
The STS paper discusses the impacts of rising social distrust in the United States on polling. Over the last few decades, various studies have found Americans to be increasingly distrustful of society, government, and even fellow Americans. One impact of this distrust is the aforementioned decrease willingness of Americans to respond to phone calls from strangers, due to suspicions that the calls are scams. As such, this distrust of fellow Americans led to the rising cost of phone-based polls, and thus indirectly led to the widespread use of online polling methodologies.
More significantly, rising social distrust has been theorized by many to be a reason pre-election polls were so inaccurate in 2020. Studies have found that people with low social trust are less likely to answer polls, and that people with low social trust were more likely to vote for Republicans, in particular then-president Donald Trump. As such, a prominent hypothesis from pollsters is that, because the people who answered polls had disproportionately higher levels of social trust, 2020 pre-election polls were biased against Republicans.
Another theory regarding 2020 pre-election polls is that their bias against Republicans stemmed from the COVID-19 pandemic. Since Democrats were more likely to follow pandemic-era rules than Republicans, there were disproportionately more Democrats at home and in a position to answer polls. This would explain why pollsters found significantly higher survey response rates from Democrats after the pandemic than before it. Thus, the theory is that, because the people who answered polls were much more likely to be Democrats during the pandemic, the polls became biased against Republicans. This would be another example of social distrust impacting polls, since a person’s likeliness to follow pandemic-era rules correlates highly with their trust of societal institutions like governmental health agencies.

School of Engineering and Applied Science
Bachelor of Science in Computer Science
Technical Advisors: Daniel G. Graham and Rosanne Vrugtman
STS Advisors: Catherine D. Baritaud and Joshua Earle