Fact Polarization
Both casual observation and contemporary scholarship depict Democrats and Republicans in the United States as warring groups with clashing worldviews and mutual animosity (Iyengar and Westwood, 2015, Hetherington and Weiler, 2018). These differences extend to basic questions of fact. Partisans express sharply divided beliefs about economic conditions (Bartels, 2002; Jerit and Barabas, 2012), income inequality (Kuziemko et al., 2015), conflict (Gaines et al., 2007), and scientific issues including COVID-19 (Barrios and Hochberg, 2020; Druckman et al., 2021) and climate change (McCright and Dunlap, 2011).
The explanation for such “fact polarization” remains a matter of debate. Bayesian learning—the rational standard for updating beliefs—describes a rigorous way in which learning from observed evidence depends on the information structure. When exposed to similar evidence, beliefs of different people should converge (Becker et al., 2017). Fact polarization defies this expectation, and some argue that motivated reasoning—automatic rejection of information that challenges one’s beliefs—is responsible (e.g., Kahan, 2016; Taber and Lodge, 2006), resonating with the idea that partisan identity serves as a “perceptual screen” through which the world is viewed (Bartels, 2002; Berelson et al, 1954; Campbell, 1960; Zaller, 1992).
Others challenge this account of fact polarization (Gerber and Green, 1999; Bullock, 2009; Druckman and McGrath, 2019; Little, 2021; Coppock, 2022). These scholars argue that empirical patterns of belief divergence are consistent with rational, Bayesian learning when there are differences in perceived credibility of information sources, or second-order beliefs.
The two explanations are observationally equivalent based on existing evidence. Which explanation is correct matters not just for reasons of basic science, but also because any attempt to dispel factually inaccurate beliefs will depend on the underlying cause of the inaccuracy (Druckman and McGrath, 2019).
To investigate these accounts of fact polarization, a new study by Burdea et al. (2023) uses a novel research design that directly measures people’s beliefs regarding the credibility of in-group and out-group information sources. To do so, the authors asked participants “How many Republicans [Democrats] (out of 100) CORRECTLY determined whether the following statement was true or false?”. The statements involved various politically and economically relevant facts such as: “The difference in median household incomes between white and black Americans has increased between 1970 to 2018.” [True], “Under half of all state prisoners in the United States were convicted of violent crimes.” [False]. Importantly, participants are asked about the credibility of both in-group and out-group partisans which enables the authors to take the difference as a measure of bias in second-order beliefs. Moreover, before running this study, the authors elicited the true credibility of 100 Democrats and 100 Republicans and selected those statements on which there were no differences in accuracy between the two groups.
The results of this study show strong evidence of partisan bias in source credibility estimates: participants think that their in-group is approximately 9 percentage points more credible than the out-group.
At first glance, this finding suggests that the Bayesian account of fact polarization may be supported. However, the partisan credibility gap estimates could themselves stem from non-Bayesian channels. In assessing source credibility, people may be blinded by social identities and attachments (Tajfel and Turner, 1979; Tajfel, 1981; Huddy et al., 2015). They may use partisan identities as affectively charged heuristics to determine whether to trust sources, as in bounded rationality models (Simon, 1985; Kahneman, 1982; Bendor, 2010). In particular, citizens may use subjective feelings or emotions as a substitute for information or memory-based accuracy judgments. Consequently, strong positive feelings toward the in-party should lead to an upward bias in second-order beliefs (overestimating in-party accuracy), while strong negative feelings toward the out-party should lead to a downward bias (underestimating out-party accuracy). Alternatively, partisans may simply know less about out-party members’ beliefs and their sources. A rational model may well predict that the more an individual knows about a party, the more accurate (less biased) the second-order beliefs will be. All of these mechanisms may function simultaneously, varying by situations and for people with different traits.
The study of Burdea et al. (2023) also investigated how the partisan gap in source credibility may be affected by these different channels. The principal finding is that out-group affect—the main component of affective polarization (Iyengar et al., 2012)— is the most important predictor of source credibility bias. People with the coldest feelings toward the out-party are those with the largest biases in second-order beliefs. This finding is consistent with the hypothesized mechanism of using affect as a heuristic—which is also one of the postulated bases for motivated reasoning (Taber, 2006).
However, the authors find little consistent evidence in support of the mechanisms based on social identity or a rational model (i.e., knowledge-based). Finally, although Democrats evince more bias in the second-order beliefs than Republicans, there is no evidence of asymmetry across parties in the mechanisms.
These results seemingly cast doubt on the practical importance and relevance of the differences between the motivated reasoning and rational learning explanations for fact polarization. They highlight that social learning is a multi-dimensional problem, influenced not only by what is being said but also by who says it. Importantly, this influence is moderated by feelings and emotions that seem to raise the “perceptual screen” through which the information from different sources is processed.
This text is jointly published by "Researching Misunderstandings" and BSE Insights.
Authors: Valeria Burdea, William Minozzi, Jonathan Woon
References:
J Jerit, J Barabas, Partisan perceptual bias and the information environment. The J. Polit. 74, 672–684 (2012).
JM Barrios, Y Hochberg, Risk perception through the lens of politics in the time of the covid-19 pandemic, (National Bureau of Economic Research), Technical report (2020).
LM Bartels, Beyond the running tally: Partisan bias in political perceptions. Polit. Behavior 24, 117–150 (2002).
J Becker, D Brackbill, D Centola, Network dynamics of social influence in the wisdom of crowds. Proc. National Academy Sciences 114, E5070–E5076 (2017).
J Bendor, Bounded rationality and politics. (University of California Press), (2010).
BR Berelson, PF Lazarsfeld, WN McPhee, Voting: A study of opinion formation in a presidential campaign. (University of Chicago Press), (1954).
JG Bullock, Partisan bias and the Bayesian ideal in the study of public opinion. The J. Polit. 71, 1109–1124 (2009).
V Burdea, W Minozzi, J Woon, Fact Polarization, Motivated Reasoning, and Source Credibility. Working Paper, (2023).
A Campbell, PE Converse, WE Miller, DE Stokes, The American Voter. (University of Chicago Press), (1960).
A Coppock, Persuasion in Parallel: How Information Changes Minds about Politics. (University of Chicago Press, Chicago), (2022).
JN Druckman, S Klar, Y Krupnikov, M Levendusky, JB Ryan, How affective polarization shapes americans’ political beliefs: A study of response to the covid-19 pandemic. J. Exp. Polit. Sci. 8, 223–234 (2021).
JN Druckman, MC McGrath, The evidence for motivated reasoning in climate change preference formation. Nat. Clim. Chang. 9, 111–119 (2019).
BJ Gaines, JH Kuklinski, PJ Quirk, B Peyton, J Verkuilen, Same facts, different interpretations: Partisan motivation and opinion on Iraq. The J. Polit. 69, 957–974 (2007).
A Gerber, D Green, Misperceptions about perceptual bias. Annu. Rev. Polit. Sci. 2, 189–210 (1999).
M Hetherington, J Weiler, Prius or pickup?: How the answers to four simple questions explain America’s great divide. (Houghton Mifflin), (2018).
L Huddy, L Mason, L Aarøe, Expressive partisanship: Campaign involvement, political emotion, and partisan identity. Am. Polit. Sci. Rev. 109, 1–17 (2015).
S Iyengar, G Sood, Y Lelkes, Affect, not ideology: a social identity perspective on polarization. Public Opinion Quarterly 76, 405–431 (2012).
S Iyengar, SJ Westwood, Fear and loathing across party lines: New evidence on group polarization. Am. J. Polit. Sci. 59, 690–707 (2015).
DM Kahan, The Politically Motivated Reasoning Paradigm, Part 1: What Politically Motivated Reasoning Is and How to Measure It. (John Wiley Sons, Ltd), pp. 1–16 (2016).
D Kahneman, SP Slovic, P Slovic, A Tversky, Judgment under uncertainty: Heuristics and biases. (Cambridge University Press), (1982).
I Kuziemko, MI Norton, E Saez, S Stantcheva, How elastic are preferences for redistribution? evidence from randomized survey experiments. Am. Econ. Rev. 105, 1478–1508 (2015).
AT Little, Directional motives and different priors are observationally equivalent. Unpubl. Manuscript pp. 1–31 (2021).
AM McCright, RE Dunlap, The politicization of climate change and polarization in the American public’s views of global warming, 2001–2010. The Sociol. Q. 52, 155–194 (2011).
HA Simon, Human nature in politics: The dialogue of psychology with political science. Am. Political Science Review 79, 293–304 (1985).
CS Taber, M Lodge, Motivated skepticism in the evaluation of political beliefs. Am. Journal Political Science 50, 755–769 (2006).
H Tajfel, Human groups and social categories. (Cambridge University Press), (1981).
H Tajfel, JC Turner, An integrative theory of intergroup conflict. Organ. Identity: A reader 56, 9780203505984–16 (1979).
JR Zaller, The Nature and Origins of Mass Opinion. (Cambridge University Press), (1992).