Less Partisan but No More Competent: Expressive Responding and Fact-Opinion Discernment 

Matthew H. Graham and Omer Yair

In recent years, various studies have examined the ability of citizens in different countries to distinguish between a factual statement, one that can be empirically verified, and an opinion statement. This ability is of great importance in the current information environment, with researchers explaining that without the ability to identify whether the information presented or the statement heard is a fact or an opinion, the public, as well as fact-checkers, will have no basis on which to begin their checks. At the same time, studies have also shown that citizens' political preferences influence the way they distinguish between factual statements and opinions. For example, a 2018 Pew Research Center study showed significant differences – sometimes over 20 percentage points – between Republicans and Democrats in classifying certain statements as “fact” or “opinion,” with respondents more likely to rate statements they agreed with as “fact” and more likely to rate statements they disagreed with as “opinion.”

According to these findings, political rivalry makes it difficult for citizens to agree even on the basic question of whether a statement is an empirically verifiable fact or "just" someone's opinion. On the other hand, a series of studies in recent years have shown evidence of "expressive responding" in surveys, with survey respondents sometimes providing an insincere answer to a survey question, out of a desire to convey some kind of "political-partisan message" (for example, consciously giving an incorrect answer in response to a question about knowledge about the unemployment rate, or about the number of deaths in certain wars [see here, or here]).

In our article, recently published in the Public Opinion Quarterly journal, we examined whether the differences between political groups in classifying different statements as fact or opinion indicate respondents' honest response, or perhaps indicate an expressive, insincere response intended to convey some kind of “partisan message.” To this end, we ran four online surveys with an embedded experiment: two in the United States, in June and July 2021, and two in Israel, in February and October 2022 (10,614 respondents in total in the four surveys).

To examine whether these differences indicate honest or insincere/expressive responses, we drew on literature suggesting that sometimes survey respondents who are asked a certain question incorporate in their answer their position regarding another question, one that was not asked. Imagine a situation where you arrive to eat at a restaurant, where the food is excellent, but the service is abysmal. If, after the meal, you are asked to rate only the food, some people will give their answer to this question while taking into account the poor service, even though they were not asked about it at all; that is, in their response, they have also answered an unasked question. However, if there is a question about the service before the question about the food, there will be no need to answer that "unasked question," and the food evaluation item will be more reliable and accurate.

Accordingly, we hypothesized that sometimes, some respondents would use the question that requires a distinction between a factual statement and an opinion statement to answer a question that is usually not asked: what they think about the statement itself; that is, whether they think the statement is accurate or whether they agree with it. And in our experiments, we examined whether adding to the treatment groups an initial question that measures agreement with a certain statement or whether this statement is accurate, and only then asking respondents to classify the statement as a fact or an opinion, reduces gaps between political opponents in the fact/opinion classification compared to the control group in which respondents were only asked to classify the statement as a fact/opinion.

In experiments in the United States, we examined differences between respondents who identified as Republicans or Democrats, using statements such as "Barack Obama was born in the United States" (factual statement), "The Earth is warmer today than it was 100 years ago" (factual), or "Abortions should be legal in most cases" (opinion). In experiments in Israel, we examined differences between voters of the "Change Bloc" (then in the coalition) and voters of the "Netanyahu Bloc" (then in the opposition), using statements such as "The number of rockets fired from the Gaza Strip into Israel over the past year is the lowest in 15 years" (factual statement), "The Bennett-Lapid government was one of the worst Israeli governments ever" (opinion), or "A law should be enacted to prevent a member of Knesset who has been accused of crimes from being appointed prime minister" (opinion).

In all four experiments, we found that, relative to the control group, allowing the treatment group to answer an “unasked question” reduced disparities between political opponents, with three of the four experiments showing statistically significant differences (ps<.01). More specifically, we found that expressive responding artificially inflates differences between political opponents by about 50-60% in the United States and about 30% in Israel.

However, contrary to our expectations, we did not find a significant improvement in the accuracy of respondents in classifying statements as fact/opinion. We explain this by saying that, on the one hand, the manipulation we conducted (adding said preliminary questions) reduced insincere responses among respondents who knew the correct answer, but chose to answer the wrong answer to convey some kind of "partisan message," which we called "hidden knowledge." But on the other hand, the manipulation also reduced insincere responses among respondents who (expressively) gave the correct answer by mistake, which we called "hidden ignorance." In aggregate, "hidden knowledge" and "hidden ignorance" both reduced the gaps in responses between political opponents, but their contradictory effects on providing the correct answer to the classification question did not lead to a significant improvement in accuracy (and sometimes, as in the Israeli case, even led to less accuracy).

Overall, the findings of the four experiments we conducted in the United States and Israel show that when it comes to distinguishing between a factual statement and an opinion, the public is less biased and partisan – less unduly influenced by its political preferences – than one might think, but also no more accurate or “competent” than one might think. These findings provide further evidence that survey respondents sometimes provide insincere answers to questions about politics, out of a desire to obtain various benefits, such as presenting their own political side in a positive light or presenting political opponents in a negative light. Still, it is worth noting that there are also other studies that have found no evidence of expressive responding in answers to politically “sensitive” questions, and future studies – in Israel and in other countries – will help to understand the extent of this phenomenon.

---
Dr. Matthew H. Graham is a professor of political science at Temple University.

Dr. Omer Yair is a Research Fellow at the Institute for Liberty and Responsibility at the Lauder School of Government, Diplomacy and Strategy at Reichman University.

This post is based on the article:
Graham, Matthew H., and Omer Yair. 2025. "Less Partisan but No More Competent: Expressive Responding and Fact-Opinion Discernment," Public Opinion Quarterly 89(1): 7–30. https://doi.org/10.1093/poq/nfaf008.