Does Question Wording Predict Support for the Affordable Care Act? An Analysis of Polling During the Implementation Period, 2010–2016

ABSTRACT

The Patient Protection and Affordable Care Act (ACA) continues to be the subject of fierce political debate in the United States. Drawing on issue framing theory, together with research on wording effects in survey responding, we tested how common differences in the wording of ACA surveys relate to apparent public support for the law. We report on a content analysis of N = 376 U.S. national opinion surveys fielded during a more than six-year period, beginning 23 March 2010 (when President Obama signed the bill into law) and ending 8 November 2016 (Election Day), and use ordinary least squares (OLS) regression models to predict public support for the law as a function of variation in question wording. We coded questions gauging general sentiment toward the law for differences in issue labeling (e.g., Obamacare, Affordable Care Act), whether or not they referenced particular political entities (e.g., President Obama, Congress) or segments of the public (e.g., You, Your Family), various opinion metrics (e.g., Support, Favor), and different response options (e.g., Repeal, Expand) which we used to model aggregate levels of support. The results revealed several key differences in question wording—for example, generic references to the Healthcare Law were employed much more frequently than Obamacare or Affordable Care Act—a number of which reliably predicted aggregate levels of public support. The discussion considers possible explanations for these patterns and reiterates the value of attending to questionnaire design features when interpreting survey data about politically contentious health policy issues.

Notes

¹ Some questions included more than one descriptor.

² When the dummy variables for polling organization are excluded from the model, two additional results emerge: references to “support” (b = −.04, p < .001) and the dummy variable coding for the 2014 mid-term elections (b = .04, p < .10).

³ We acknowledge that these wording differences, while unrelated to aggregate levels of support, may indeed exert significant and unique effects among certain segments of the public (e.g., Republicans as compared to Democrats), which was not the focus of our efforts here.