Explaining Presidential Popularity: How Ad Hoc Theorizing, Misplaced Emphasis, and Insufficient Care in Measuring One's Variables Refuted Common Sense and Led Conventional Wisdom Down the Path of Anomalies

Within the last ten years a new conventional wisdom has surfaced in political science which tells us that presidents inexorably become less popular over time. Not much else matters. Neither the economy, nor the Vietnam War, not even Watergate seems to have had much independent effect on presidential popularity once time is taken into account. Before embracing these conclusions we need to reconsider the method that produced them. I argue that previous research too willingly accepted time as an explanatory variable, enshrouding it with theoretical meaning. To preserve its explanatory power alternative, substantive variables were shortchanged in their operational definitions and measurement. In this article I reverse the emphasis. Here, time is rejected as an explanatory variable and is employed only as a diagnostic indicator of the adequacy of the equations. A variety of alternative representations of real-world forces such as the economy and war are tested and some considerably improve the time-series correlation between the environment and presidential popularity. With these substantive variables I propose a simpler, if less glamorous, theory of presidential popularity consisting of two hypotheses: first, popularity is related to real events and conditions, and second, that it responds slowly to environmental change. Popularity is then both experiential and incremental. The findings for Presidents Truman through Nixon support this common-sense view. The Korean War (measured by U.S. casualties), the Vietnam War (measured by the number of bombing missions over North Vietnam and the U.S. war dead), the economy (especially six-month changes in consumer prices), Watergate, international “rally” events, and early term surges of approval all contribute independently to short-term fluctuations in presidential popularity. Moreover, as predicted, popularity appears to be autoregressive even when represented by an instrumental variables surrogate measure to minimize serial correlation. When the equations are specified in this way, time proves to be unnecessary in order to explain trends in presidential popularity.