Common Data Visualization Choices Impact Attitudes Towards Refugees and Perceptions of Government Competence

As data journalism has proliferated, audiences increasingly encounter objects such as charts and graphs in media. While these outputs often aim to improve public understanding about political issues, they also involve presentational choices about color, chart types, and annotation that may originate in organizational norms and professional practices. Practical interest in ‘telling stories with data’ as well as developments in digital political communication suggest that visualizations may change viewers’ perceptions and attitudes. But what, if any, specific features may be more responsible for this? While existing work on the efficacy of fact-checking examine factors at the levels of messaging as well as audiences, fewer studies consider visual communication modes involving non-photorealistic elements that nevertheless increasingly accompany journalistic outputs. In response, I present results from a conjoint experiment containing data visualizations about the scale and geographic composition of refugees entering the UK between 2001-20. These treatments—created by a professional designer and following conventions in migration visualizations circulating online to enhance external validity—randomly varied along four dimensions within data journalists’ and editors’ abilities to change: chart type, dominant color, annotative framing, and disclosure of data sources. I fielded the experiment to a nationally-representative British sample (N=3,082) in August 2021. Owing to its timing during the height of the Afghan humanitarian crisis that generated intense concern about how countries, including the UK, would handle asylum-seekers from the country, it provides evidence drawn during periods of high salience—likely when information-based outputs are in greater demand. Exploratory testing revealed that, when compared to the mean response for each outcome, specific design features did indeed cause shifts in attitudes and preferences. Humanitarian framing raised perceptions of how the government was handling the issue of refugees. Conveying the data via an area chart reduced respondents’ willingness to prioritize the most vulnerable refugees for entry. Omitting the data source, displaying an unusual curved chart type, emphasizing gross numbers of admissions, and (perhaps counter-intuitively) using primarily red coloration both lowered respondents’ reported anxiety about refugees and made them less likely to believe that refugees threatened British culture. The results varied by subgroups. Respondents holding lower levels of trust in government institutions (measured pre-treatment) expressed significantly lower levels of anxiety, less agreement with the idea that refugees threatened British culture, and lower approval of the government’s handling of refugee issues, compared to those holding higher levels of trust. Meanwhile, across all visual elements, respondents who self-reported having lower levels of numeracy (again measured pre-treatment) were less likely to prefer policies that prioritized highly vulnerable refugees, compared to those who self-reported higher levels of numeracy. Overall, this study makes three contributions. First, it provides novel experimental evidence about how specific design choices matter for attitudes on a politically salient topic. This extends understanding to the visual realm. Second, by isolating the effects of these discrete choices, this research connects academic interests with design practice. Third, distinguishing among different kinds of respondents opens avenues for further theorization about the conditionality of visual messages’ effectiveness.