Political affiliation moderates subjective interpretations of COVID-19 graphs

Abstract

We examined the relationship between political affiliation, perceptual (percentage, slope) estimates, and subjective judgements of disease prevalence and mortality across three chart types. An online survey (N = 787) exposed separate groups of participants to charts displaying (a) COVID-19 data or (b) COVID-19 data labeled ‘Influenza (Flu)’. Block 1 examined responses to cross-sectional mortality data (bar graphs, treemaps); results revealed that perceptual estimates comparing mortality in two countries were similar across political affiliations and chart types (all ps > .05), while subjective judgements revealed a disease x political party interaction ( p < .05). Although Democrats and Republicans provided similar proportion estimates, Democrats interpreted mortality to be higher than Republicans; Democrats also interpreted mortality to be higher for COVID-19 than Influenza. Block 2 examined responses to time series (line graphs); Democrats and Republicans estimated greater slopes for COVID-19 trend lines than Influenza lines ( p < .001); subjective judgements revealed a disease x political party interaction ( p < .05). Democrats and Republicans indicated similar subjective rates of change for COVID-19 trends, and Democrats indicated lower subjective rates of change for Influenza than in any other condition. Thus, while Democrats and Republicans saw the graphs similarly in terms of percentages and line slopes, their subjective interpretations diverged. While we may see graphs of infectious disease data similarly from a purely mathematical or geometric perspective, our political affiliations may moderate how we subjectively interpret the data.

Graphic resources for health information visualization in the context of Covid-19

“…Among the articles, we highlight the study by Ericson, Albert & Duane (2022), who developed graphs relating the political affiliation of subjects to subjective interpretations of the graphs about Covid-19. The authors demonstrated that while people may view a graph from a purely mathematical or geometric perspective, subjective interpretations of graphs can be tempered by their political affiliations.…”

Politicization of a Pathogen: A Prospective Longitudinal Study of COVID‐19 Responses in a Nationally Representative U.S. Sample

“…In the peer‐reviewed literature, early research during the pandemic suggests that U.S. Republicans were less worried about, and less willing to perform, health‐protective behaviors that prevent the spread of the coronavirus, and this divergence may have been motivated specifically by Trump and Republican elite cues (Allcott et al, 2020; Bruine de Bruin et al, 2020; Conway et al, 2021; Douglas & Sutton, 2022; Engel‐Rebitzer et al, 2022; Gadarian et al, 2021; Gollwitzer et al, 2020; Grossman et al, 2020; Kaushal et al, 2022; Kim & Kwan, 2021; Leventhal et al, 2021; Moore et al, 2021; Rodriguez et al, 2022; Ruisch et al, 2021). However, while each of these prior studies has methodological strengths, this body of research is limited in several important ways: The vast majority of prior studies used either convenience samples (e.g., Conway et al, 2021; Douglas & Sutton, 2022; Engel‐Rebitzer et al, 2022; Ericson et al, 2022; Fridman et al, 2021; Leventhal et al, 2021; Moore et al, 2021; Ruisch et al, 2021) or non‐probability‐based representative samples (e.g., Allcott et al, 2020; Gadarian et al, 2021; Kaushal et al, 2022; Rodriguez et al, 2022), which can lead to sampling bias that limits generalizability (Bradley et al, 2021; Pierce et al, 2020). Most prior research did not compare COVID‐19 outcomes to a nonpoliticized control pathogen to establish polarization (e.g., Allcott et al, 2020; Bruine de Bruin et al, 2020; Conway et al, 2021; Gadarian et al, 2021; Gollwitzer et al, 2020; Kim & Kwan, 2021; Leventhal et al, 2021; Moore et al, 2021; Rodriguez et al, 2022; Ruisch et al, 2021).…”

“…Many prior studies investigated polarized COVID‐19 responses within the first 2 to 3 months of the pandemic only (e.g., Allcott et al, 2020; Bruine de Bruin et al, 2020; Conway et al, 2021; Douglas & Sutton, 2022; Gadarian et al, 2021; Gollwitzer et al, 2020; Rodriguez et al, 2022; Ruisch et al, 2021). Understanding how polarized responses to a novel pathogen play out over a longer period of time is useful for devising long‐term public health and risk‐communications strategies. Another limitation of the prior research is that political identities and COVID‐19 outcomes were concurrently measured during the pandemic (e.g., Allcott et al, 2020; Bruine de Bruin et al, 2020; Conway et al, 2021; Douglas & Sutton, 2022; Ericson et al, 2022; Fridman et al, 2021; Gadarian et al, 2021; Leventhal et al, 2021; Moore et al, 2021; Rodriguez et al, 2022; Ruisch et al, 2021). Consequently, it is possible that partisans' responses to COVID‐19 affected the strength of their political identities. Finally, big data studies (e.g., smartphone mobility data, Twitter data; Engel‐Rebitzer et al, 2022; Green et al, 2020; Jiang et al, 2021) also suffer from biases as their samples are self‐selected, not probability based (Bradley et al, 2021), and are typically not representative of the population.…”

How Do Viewers Synthesize Conflicting Information from Data Visualizations?

“…This behavior also applies to chart comprehension; readers selectively attend to information aligned with their beliefs [26]. For example, Democrats and Republicans differed in their interpretations of mortality rates when the same chart was labeled as COVID-19 or Influenza [27]. Readers tend to impose categorical distinctions (i.e., binary bias) when interpreting visualizations of continuous data, resulting in distorted beliefs [29].…”