From Informational Beliefs to Complex Political Attitudes: A Bayesian Analysis of American Public Opinion on the Affordable Care Act

Abstract

It has long been assumed that the formation of individuals’ attitudes toward a complex attitude object should reflect an integration process whereby individuals evaluate different constituent parts of the attitude object based on the information they have about these components and synthesize these evaluations to render a summary judgment. To what extent this expectation is true, however, remains a question to be answered. Leveraging a nationally representative survey on American public attitudes toward the Patient Protection and Affordable Care Act of 2010, this project investigated how a multitude of informational beliefs about what the law is comprised of, the level of certainty with which individuals hold these beliefs, and evaluations of corresponding provisions of the law, collectively shaped Americans’ overall attitude toward this important and highly contentious piece of legislation. Using innovative Bayesian modeling techniques with Markov Chain Monte Carlo methods, this project successfully established a model where beliefs and evaluations associated with different attributes of an attitude object were presumed to have structurally similar relations with the overall attitude, and the variations in the effects of distinct beliefs and evaluations were determined by varying importance levels of different attributes of the attitude object. Furthermore, through a series of counterfactual simulations leveraging Bayesian linear regression and logistic regression models, this project identified the pivotal roles of belief accuracy and belief certainty in the process of political attitude formation. Possessing correct beliefs about what the law would and would not do improved the law’s overall favorability. Holding correct beliefs with full certainty could even transform a slightly negative overall attitude toward the law to a slightly positive one. However, the counterfactual simulations also revealed that the increase in belief certainty resulted in a tremendous increase in attitude extremity. The size of the “neutral” group diminished while the group of those who were expected to express strong preferences (either favor or oppose) expanded enormously. The improvement in individuals’ state of information (i.e., belief accuracy and belief certainty) led to both more favorable and more polarized overall attitudes toward the law. On this highly contentious public policy where there is persistent partisan divide, only a small portion of the partisan attitudinal gap could be accounted for by the state of information about and evaluations of distinct components of the law, while a large portion of the partisan discrepancy might reflect some more profound and inherent partisan considerations.