Aggregation and Dynamics of Survey Responses: The Case of Presidential Approval
In this paper we critique much of the empirical literature on the important political science concept of presidential approval. Much of the recent research on presidential approval has focused on the dynamic nature of approval; arguments have raged about whether presidential approval is integrated, co-integrated, or fractionally integrated. We argue that none of these time-series concepts, imported from an econometrics literature which has fundamentally different types of data than do political scientists, can apply to the presidential approval time series. Instead, we advocate careful use of aggregated approval as a time-series cross-section, or the use of individual-level survey responses. Ultimately most of the important hypotheses political scientists wish to test regarding presidential approval involve individual voters or citizens; thus we argue that using the appropriate data unit is the best methodology.
Additional InformationPrevious versions of this paper were presented at the 1996 Midwest Political Science Association Meetings and at the 1997 Southern California Political Methodology Program Conference at UCLA, and we thank participants in these conferences for their comments. We are grateful for discussions with David Grether. We thank Neal Beck and Simon Jackman for answering our questions about the presidential approval series, and sharing their data. We also thank John Aldrich for making available to us the 1988 Gallup Organization polling data. Katz thanks the John M. Olin Foundation for a faculty fellowship supporting his work.
Submitted - sswp1103.pdf