The Dynamics of Public Opinion

Mary Layton Atkinson, K. Elizabeth Coggins, James A. Stimson, and Frank R. Baumgartner

Cambridge University Press, Elements Series, Forthcoming 2021

Click here to return to the main page.

A central question in political representation is whether the government responds to the people. To understand that, we need to know what the government is doing, and what the people think of it. In this volume we seek to understand a key question necessary to answer those bigger questions: How does American public opinion move over time? Only if we can answer that question can we assess whether the government responds to the people, whether the people respond to the government, or whether both or neither of those is true. As empirical scholars, we delve deeply into the measurement of public opinion over time and provide the richest report of the long-term dynamics of public opinion so far.

We posit three patterns of change over time in public opinion, depending on the type of issue. Partisan, nonpartisan, and cultural shift issues generate distinct patterns of public opinion response. The largest share of issues are those on which the two parties regularly disagree; these provide clear partisan cues to the public. For these party-cue issues we present a slight variation on the thermostatic theory from Wlezien (1995) and Soroka and Wlezien (2010); our "implied thermostatic model" applies to most issues. A smaller number of issues divide the public along lines unrelated to or different from partisanship, and so partisan control of government provides no relevant clue, and the implied thermostatic model cannot apply. Finally, we note a small but critically important class of issues which capture response to widespread cultural shifts, sometimes occurring over multiple generations.

Each of the three types requires a distinct causal theory to explain movement over time. The implied thermostatic model, like the original, predicts movement in the opposite direction of current partisan control of government. For issues with no party cues we predict stationarity because this issue type lacks the stimulus of parties altering in government which drives the party-cue thermostat. And for cultural shift issues, which we observe to be concentrated on issues of equality, we predict permanent attitude changes which at the macro level produce trending time series. These long-term trends are generated both by individual attitude change and by generational replacement as younger cohorts are socialized into their political attitudes at a time when acceptance of equality norms is greater than would have been true for the generation of their grandparents.

We validate our prior judgments about which issues fit the party cue and no party cue types. Then we develop a statistical test for thermostatic behavior and show that the test is generally met for party cue issues and not for other issue types. Then we pay special attention to cultural shift issues, which though small in number loom very large in political importance. And then we conclude with observations about issues and the party system in America. If we had to pick a single model, it would be the implied thermostat. But with three models of opinion change, we can have a more complete and nuanced understanding of the dynamics of opinion change.


(Last updated, February 25, 2021 )