Ricochet is the best place on the internet to discuss the issues of the day, either through commenting on posts or writing your own for our active and dynamic community in a fully moderated environment. In addition, the Ricochet Audio Network offers over 50 original podcasts with new episodes released every day.
Don’t bet your house on the roulette pattern — or your country on the election pattern. Correlation is not causation and past performance does not necessarily predict future results. We all know this, yet it is great fun to prognosticate and to chew the fat over past sports and presidential election seasons. Moving beyond such speculation towards serious analysis requires us to turn to the theory and practice of political science.
In Designing Social Inquiry: Scientific Inference in Qualitative Research, Gary King, Robert Keohane, and Sidney Verba laid out the case for research that is scientific even without large data sets — situations like the small set of presidential elections. Arguing against ad hoc explanations, they laid out the basics of research design: “the research question, the theory, the data, and the use of the data.” (p.13)
A social science theory is a reasoned and precise speculation about the answer to a research question, including a statement about why the proposed answer is correct.
[…] Any intelligent scholar can come up with a “plausible” theory for any set of data after the fact, yet to do so demonstrates nothing about the veracity of the theory. […] Human beings are very good at recognizing patterns but not very good at recognizing nonpatterns. (p.21)
Consider an annual event, which in presidential election years highly correlates with the election outcome.
Discovered by Steve Hirdt of the Elias Sports Bureau, the Redskins Rule that when the Washington Redskins win their last home game during an election year, the incumbent party will retain the White House. If they lose, the challenging party will take the election. (The connection was even cited in 2007 episode of Mad Men.)
Sure, it sounds a little goofy, but here’s the astonishing part about the Redskins Rule: since 1936, it held true for every election year until 2004, when George W. Bush retained the presidency after the Skins lost the last game of the season. The Rule got back on track in 2008, however, when Washington lost its final game and Barack Obama took the White House.
Is this a meaningful pattern? Would you really ask how sporting events significantly influence, let alone determine, the outcome of presidential elections in America? Note that the “theory” applies to precisely one game in one sport. No one who cares about our politics is going to urge political campaign and voting decisions based on the Redskins’ season final home game outcome.
Let’s turn to another possible pattern that is claimed to have predictive power for presidential elections. It has been claimed that “any Republican would have won in 2016.” The basis of this claim is given as follows:
The 22nd Amendment was adopted in 1951, which established a two-term limit on Presidents. The American people have gone one better. They have practically established a two-term limit on the two political parties, a pattern which has persisted in 15 of 17 elections since 1951!
In each election, the question is if the party in power is on their first or second term.
- If the party is in its first term, the American people have historically given that party another four years, with the only exception being in 1980 after Jimmy Carter had been a disaster as President.
- But if the party in power is in its second term, the American people in their wisdom have decided to “throw the bums out”! The only time that any party has gotten a third term was in 1988 when George H.W. Bush was essentially elected to the third term of the greatest president of the Twentieth Century, Ronald Reagan.
Notice that there is no falsifiable theory here. Indeed, the deviations from the perceived pattern are explained by ad hoc claims – “Carter had been a disaster” and “George H.W. Bush was essentially elected to the third term of the greatest president of the Twentieth Century, Ronald Reagan.” We are not invited to consider all presidential elections, just those since the adoption of the 22nd Amendment. There is no explanation for this presumably arbitrary limit.
The 22nd Amendment only formalizes the unwritten constitutional term limit established by President Washington, who would otherwise have been president for life.
Amendment XXII (emphasis added)
No person shall be elected to the office of the President more than twice, and no person who has held the office of President, or acted as President, for more than two years of a term to which some other person was elected President shall be elected to the office of the President more than once. But this article shall not apply to any person holding the office of President when this article was proposed by the Congress, and shall not prevent any person who may be holding the office of President, or acting as President, during the term within which this article becomes operative from holding the office of President or acting as President during the remainder of such term.
This article shall be inoperative unless it shall have been ratified as an amendment to the Constitution by the legislatures of three-fourths of the several states within seven years from the date of its submission to the states by the Congress.
No candidate until Franklin Delano Roosevelt dared run for more than two full terms. Indeed, Theodore Roosevelt, who ascended from vice president to the presidency on the death of McKinley, gave way to Taft after being elected president once. Four years later, TR was the first to stretch the unwritten rule by running for a full second term as a third party candidate. A visual inspection of all presidential elections suggests that election results by party are a nonpattern.
But, maybe there is something to the party incumbency idea. There has been a small literature on presidential incumbency, especially by Alan I. Abramowitz, that shows a significant incumbency effect. He uses three variables, under what he calls the “Time for Change forecasting model.”
[The variables are] the incumbent president’s approval rating at midyear (late June or early July) in the Gallup Poll, the growth rate of real GDP in the second quarter of the election year, and whether the incumbent president’s party has held the White House for one term or more than one term. Using these three predictors, it is possible to forecast the incumbent party’s share of the major party vote with a high degree of accuracy in late July, more than three months before Election Day.
The most basic problem with the model is that even if it worked across all elections, it predicts the political equivalent of Super Bowl yards gained. The game is determined by points scored and the election is determined by Electoral College votes. With the extreme tilt of several Democrat states, models dependent on national level data may be going the way of surveys dependent on land lines.
Beyond not estimating the thing that actually determines presidential election victories, Abramowitz’s model is critically dependent on Gallup Poll data, limiting how many elections can be tested. This creates vulnerability to changing conditions not measured in this elegantly parsimonious model. Indeed, the model treats third party votes as a wash between the two major parties, when that has changed from election to election. He was so concerned about his model in 2012, that he added an extra explanatory variable, POLARIZATION, and then ran new estimates on elections back to 1996.
In October 2016, Abramowitz was trying to explain, in advance, anticipated national popular vote under-perfomance by Donald Trump. He even offered excuses about violated assumptions that do not appear in his earlier work. Yet, he did not include his POLARIZATION variable! He offered all the conventional wisdom and missed both candidates’ campaign decisions in swing states. He thought swing states were important to mention in 2012, but did not take that as a prompt to reconsider his model.
So, there is a theory that fits the data, if you limit cases to 1948 and following, and if you assume that major party popular vote percentages, based as just those votes cast for the two major party candidates, accurately predict Electoral College victory – which the theory’s author does not claim. What other research might help?
The American people do not choose presidents, the American voter does, through the state-by-state aggregating mechanism of the Electoral College. As it happens, academics have carefully studied voting behavior, using scientific polling data, since at least the 1960 publication of The American Voter. It is a very quantitative field supported for decades by the American National Election Studies (ANES). The ANES mission:
[To] inform explanations of election outcomes by providing data that support rich hypothesis testing, maximize methodological excellence, measure many variables, and promote comparisons across people, contexts, and time. The ANES serves this mission by providing researchers with a view of the political world through the eyes of ordinary citizens.
Here is the searchable PDF list of American Election Survey questions. Note that there are 24 questions about congressional incumbency, but only one that can be interpreted as presidential incumbency in the last election, not two elections back. Indeed, the only reference in 2016 to 2008 is in questions about how the economy has done over time. To understand why, see David R. Mayhew’s 2006 article “Incumbency Advantage in U.S. Presidential Elections: The Historical Record” (PDF), showing lack of statistical significance.
So, there is a theory that fits selected data for elections starting in 1948. A prominent scholar suggested over a decade ago that presidential incumbency should be included in voter behavior models and survey instruments. As of the last election, the ANES has not added questions that would get at incumbent presidential party effects. When the scholar with the strongest presidential incumbency model has made adjustments, and offering qualifications to his predictions, over the last two presidential cycles, perhaps you should not bet your country on an incumbency effect.Published in