How Policy Information Enlightens Public Opinion in the Real World

• In Event: Policy Attitudes and Preference Formation

Abstract

A well-functioning democracy is a tall order and requires a lot from its citizens. Not only do citizens need to have some knowledge of facts about policy but more importantly, they should also use the policy information they learn to enlighten their opinions on political issues (Gilens 2001; Hochschild and Einstein 2015). For several decades, political scientists have studied the role of policy information in public opinion formation, but under the constraint of an inherent obstacle: In the real world, political issues become salient very suddenly meaning that survey research typically arrive too late in the field to identify which people actually learn policy information related to those issues, as this requires measures of their prior knowledge. Consequently, political scientists have mainly relied on experimental designs to examine the impact of policy information on citizens’ issue opinions (e.g., Bullock 2011; Boudreau and MacKenzie 2014; Druckman, Peterson, and Slothuus 2013; Cohen 2003; Rahn 1993). These experiments expose random groups of respondents to certain information about the immediate and likely consequences of adopting certain policy proposals and subsequently measure the attitudinal response. While this work has vastly improved our understanding of the impact of policy information on public opinion, the experiments – by design – face two limitations: First, to obtain credible and strong manipulations of policy information, researchers must either use fictional policy information or ask respondents to engage with information on issues that are low-salient (Boudreau and Mackenzie 2014, 49). Either way, subjects receive information that they know will not affect them or which they find unimportant. Second, citizens are exposed to policy information in a sterile survey setting without the presence of other competing forces of influence making participants more attentive to the stimuli than they are in their everyday life engaging with politics (see Slothuus and Bisgaard 2021; Barabas and Jerit 2010). For these reasons, most of what we know about the influence of policy information derives from context that are very different from how citizens in reality are confronted by and engage with such information. Ultimately, the important question is whether citizens change their opinions when learning about tangible policy information on important issues in the real world. However, the existing experiments cannot provide the answer. In this article, I advance the current literature by examining whether citizens who learn about the actual consequences of adopting certain policies become more inclined to change their policy opinions in the real world. I take advantage of original and existing panel survey data from four European Union referendums across different countries and political issues. These referendums contexts are well-suited for the purpose of the study as certain policy information is often salient during the campaigns, while voters have to act on their opinion towards a narrow and novel issue. Specifically, I exploit the individual level panel data to compare the policy opinion change from an early to a late campaign stage among citizens who learned policy information during the campaign relative to citizens who did not (a difference-in-differences design). In each of the four referendums, I find that citizens who learned policy information became substantially more inclined to change their stance on the referendum issue in direction of what they learned compared to similar citizens who did not grasp the information. As my design relies on non-random variation in whether citizens learned certain policy information, I explicitly address potential violation to the identifying assumption of parallel trends in several ways, and I analyse the sensitivity of the findings to various violations of it. While the internal validity of the design is not perfect, the similar and substantial results across the different model specifications support a causal interpretation of the results. Interestingly, the magnitude of the opinion change is similar to the findings of the extant experimental studies suggesting that such effect sizes of policy information do generalize to a real-world setting. In short, my results suggest that citizens actively use information about the consequences of adopting certain policy proposals to enlighten their opinions on these issues in the real world.

Author

• Jannik Fenger, Aarhus University