Predicting the Incumbent Party Vote Share in U.S. Presidential Elections
Moghaddam, Masoud, Elich, Hallie, The Cato Journal
Once every four years, it has become an American ritual to have the opportunity to make history and change a major part of the world by electing or reelecting a president. Certainly, presidential election years feature not only the symbolic exercising of a fundamental American right, but also the fruits and labors of an extraordinarily complex political process. However, even the firmly rooted structure of this process, the majority of which is some 240 years old, and the solidarity exuded by "dyed in the wool" Republicans and Democrats, has been transformed into what often appears to be a chaotic frenzy. Clouding the matter even further, recent advancements in information technology have increased mass media attention surrounding campaigns. The publicizing of debates, primaries, polls, and political mudslinging is now embedded in the political system and can be transmitted around the world in the blink of an eye. Unquestionably, the process of electing a president has long involved campaigns, primaries and caucuses, debates, polls, and pundits. Yet, in a modern way, our political preoccupation often appears unnecessarily self--created, mundane, phlegmatic, and to some extent capricious. The impetus of election-year popularity and commercialization are intertwined with an established ritual so profoundly rooted in American history. Indeed, as the modern process unwinds and more often than not meanders, recent history has revealed that, deep--down, more is involved in choosing America's next president than merely flipping a coin. It is abundantly clear that economic, historical, and personal issues influence and form the basis for many swing voters' decisions. Toward that end, in this article, we present a brief summary of the presidential election literature in the United States, describe the empirical model used to predict the 9.008 presidential election, and discuss policy implications of the estimated model.
The Voting Theory
Ray Fair (1978) developed an empirical model to predict the outcome of U.S. presidential elections. The underlying theory behind Fair's vote equation is predominantly that of Anthony Downs (1957), Gerald Kramer (1971), and George Stigler (1973) among others. Accordingly, although well informed voters look back more than a year, voting behavior depends heavily on economic events in the year of the election. Correspondingly, William Niskanen (1979) demonstrated that the Log-it version of the incumbent's popular vote share (I) depicted by In (I/1 - I, where In = natural logarithm) in election years during the 1896-72 period, is determined by economic variables (four-year percentage changes in real per capita net national product, the employment rate, the consumer price index, the stock price index, and the corporate bond rate), fiscal policy tools (four-year percentage changes in the real per capita federal government expenditure or revenue), two political variables (ln [I(t - 4)/1 -I(t - 4)] and a dummy variable depicting whether or not there is an incumbent candidate), along with another dummy variable capturing major wars. In the context of an overparameterized election model (as is the case in other empirical studies), his overall findings tend to suggest that an incumbent president is most likely reelected if economic growth has been preserved, federal spending is under control, and a major war has been avoided. Most interestingly, changes in both real per capita government expenditures and tax income (in two separate regression equations) inversely and significantly determine In (I/1 - I), while the difference between the two is a measure of "debt illusion," (1) By the same token, Fair (1996a, 1996b) contends that voters react to the rate of growth of real per capita GDP in the election year and that they compare the past performance of rival parties. Consequently, they cast their votes for the party that appears to be able to maximize their expected utility (satisfaction). (2)
Fair (1988, 1996a, 1996b, mad 2006) re-estimated his vote equation once every four years in the sample period 1916-2004 using ordinary least squares (OLS), and made predictions about whether the Democratic or Republican nominee would win the presidency. …