Predicting the Election: Enough Already?

Predicting the Election: Enough Already? 

By this point in the season I'm suffering from election fatigue. As the contests drags on, I'm probably not the only person wondering, "wouldn't it be easier if we could predict the winner in advance?"

On August 28th, 2012 U.S. News and World Report's Economic Model Looks Ahead: See's Victory for Romney claimed "Kenneth Bickers and Michael Berry have developed an economic and political forecasting model to analyze economic data from the 50 states and D.C. going back to 1980 in order to predict the presidential campaign winner. And, they say, their model predicts that Romney will win." Last year USNWR provided the contrasting prediction of Allan Lichtman, the American University professor whose "12 keys" election formula has correctly called every president since Ronald Reagan’s 1984 re-election", assured Barack Obama's re-election is in the bag. These prognosticating political pundits boast of having predicted 7 consecutive elections.  This November only one perfect predictor will remain.

So who will win? Litchman or Bickers and Berry? Obama or Romney?


The predictive models, are nothing better than educated guesses. Any attempt to divine the presidential winner suffers from a simple error of statistical inference: small sample size. There have only been 44 presidents. Not a large sample, but big enough. Except that the elections of Washington, Jefferson, J.Q. Adams, Abraham Lincoln, or Rutherford B. Hayes do not much resemble the modern presidential contest in a number of ways. 

If we assume that the "modern" presidency and "modern" presidential politics starts in the mid-20th century, then we are left with a radically reduced sample size. 13 presidents and 20 elections:  Roosevelt (4), Truman (1), Eisenhower (2), Kennedy (1), Johnson (1) , Nixon (1) , Ford (1) , Carter (1) , Reagan (2) , Bush I (1) , Clinton (2), Bush II (2), Obama (1).  

That leaves us 20 observations of the dependent variable. But in terms of predictive power, we actually have much less to work with for assessing the variation. The Key metric is the degrees of freedom. As a Tuft's University Bio-statistician, Gerald E. Dallal describes: degrees of freedom as a way of keeping score. A data set contains a number of observations, say, n. They constitute n individual pieces of information. These pieces of information can be used either to estimate parameters or variability. In general, each item being estimated costs one degree of freedom. The remaining degrees of freedom are used to estimate variability."

A general rule of thumb is that at least 20 degrees of freedom are necessary for any meaningful statistical inference. Each variable in the estimating equation costs "one degree of freedom" and the need for a least one unit of variation cost "one degree of freedom." Such that the final number of degrees of freedom = N - P -1. 

In Litchman's case  N = 20, and P = 12. 20 - 12 -1 = 7 degrees of freedom. 7 < 20. Litchman's inference lacks statistical power.  Bickers and Berry's model logically suffers the same. Neither predictive model is more than an educated guess.

 The pollsters at least have an established survey technique., yet inscrutable polls are no less conflicted than the prognosticators.  Each poll is subject to its own assumptions and biases, and poll numbers of love have a less than perfect track record of predicting elections. The Crystal Ball's  most recent estimation prudently advocates three potential outcomes: [1] Obama victory [2] Romney Victory [3] Tie 269-269 in the Electoral College. Perhaps the "still to close to call it" stance is the only prudent position. 

Physicists predicted the trajectory of a rocket with enough accuracy to land a robot on Mars.   The enlightenment conquest of nature, and the submission of the world to a rational scientific order has produced wondrous technological and scientific advancements. Political scientist can offer no commensurate achievement. We are far better at predicting the world around us, than at predicting ourselves. Recalcitrant, stubborn, contrary, human nature holds out.  We continue to confound ourselves. As Montainge once said, "the only thing certain is that nothing is certain." Or as Larry Sabato of the University of Virginia rightly warns, "He who lives by the Crystal Ball ends up eating ground glass!"

Want to know the future. Wait. It will be here soon enough. 

- John Louis
Graduate Fellow
Clough Center for the Study of Constitutional Democracy