Understanding U.S. Presidential Elections

David R. Mayhew has a knack for bringing impressive data analysis to bear on American politics, tackling in his recent book, Partisan Balance: Why Political Parties Don’t Kill the US Constitutional System, how political parties, rather than fostering the  “gridlock” and “polarization” we often lament, have actually been the key to the continued vibrancy of our democratic political system. Writing for Election 101, he offers a characteristically thoughtful take on how U.S. presidential elections work, the idea of “victory patterns” over time, what kinds of events cause party domination, and the role of randomness. Read his piece after the jump.


Understanding U.S. Presidential Elections

David R. Mayhew


To offer fairness, an election system of two parties should exhibit at least two arithmetic properties.  One is equality.  Over a long span of time, the parties should fare roughly equally in elections.  The other is changeability.  Victories should go back and forth.  The equality standard would be satisfied if one party won every contest for a century, then the other party for the next century, but the changeability standard would not be.

By these standards, the U.S. system does rather well.  Let’s take equality first.  Starting in 1828, when mass participation of voters began, the country has had forty-six presidential elections.  The Democrats’ median share of the two-party popular vote has been 50.4%, their mean share 49.5%.  But what about the Electoral College?  The Democrats have taken the White House twenty-one times, the Republicans (or the Whigs before them) twenty-five times.  Three times—in 1876, 1888, and 2000—the Republicans did that by winning the electoral vote but not the popular vote.  On the other side, a slight counterfactual curb on the intimidation of African-American voters in the South might have brought an opposite result in 1884.  Surprisingly, there exists a respectable albeit mind-boggling case, which hinges on how to count the Alabama vote, that Nixon beat Kennedy in the national popular vote in 1960.  All this adds up to something like a wash.  Since World War II, using one plausible metric, the Republicans have performed a shade better in the Electoral College than in the popular vote eight times (certainly in 2000), and the Democrats eight times (including in 2004 and 2008).

As for changeability, that raises the question of victory patterns across time.  One popular theory sorts U.S. history into long “party realignment eras” bounded by “critical elections.”  Examples are the Jacksonian era, the McKinley era, and the New Deal era.  During eras like these, which are said to exhibit a signature dynamic, one party dominates the other for a long time—a common view is thirty years or so.  In this view, change is a rare thing.  Just recently, in the excitement of the 2008 election, it was forecast that the new Obama coalition would go on, if not forever, for a long time.  A new party era was being born.

It is wise to steer clear of such era theories.  For one thing, consider the explanatory default of randomness.  Even if party victories were distributed across a long history randomly, there would be clumps.  That is the way randomness works.  In coin flips, five heads in a row come up surprisingly often.  There is a relevant study by Daniel  J. Gans:  In the sequence of presidential elections from 1856 through 1980, the distribution of victory “runs” by party (Carter, for example, was a run of one for the Democrats; Reagan and G.H.W. Bush a run of three for the Republicans) did not different significantly  from what you would expect to get in runs of heads and tails through coin flips.   Also, a party’s performance in any one election supplies virtually no predictive information on how it will perform in the next election.

For another thing, there are plenty of obvious reasons why voters might make up their minds every four years, not every thirty years.  Events intrude.  Economic crises occur as in 1932 and 2008.  Wars carry weight as in 1952 and 1968.  We may see a “thermostatic pattern” in which voters blanch pretty quickly at a victorious party that pushes policy too far to the left, react by electing the other party which pushes too far to the right, then lurch back the other way, and so on.  The last two decades have seen a lot of this.

A fresh choice every four years.  Lots of changeability.  That is the basic story of U.S. presidential elections.  But there is one fly in the statistical ointment, one impairment to changeability.  In any election, does a party field  an incumbent president as its candidate?  Does a party get a leg up in an election that way?  The answer is yes.  Here is the record.  Going all the way back to 1788 this time, the country has held fifty-six presidential elections through 2008.  For fifty-four of them (all but 1788 and 1824), the question can be usefully posed:  Did the party holding the White House keep it?  In the twenty-three of those cases where no incumbent president figured in a November contest, the party holding the White House kept it eleven times but lost it twelve times.  On average, open-seat presidential elections are a jump ball.

But elections clogged up by an incumbent candidate are not a jump ball.  There are thirty-one instances.  The incumbent candidates have won twenty-one of them, lost ten of them.  This is roughly a two-to-one advantage.  To be sure, incumbents can get slammed out of office as were Hoover in 1932, Carter in 1980, and George H.B. Bush in 1992.  But generally they win.  The likely reasons for this are several.   Voters can be risk-averse.   Incumbents learned how to run a campaign last time and will have no trouble raising vast money for one this time.  They gain experience in office.  As presidents, they can act strategically by making appealing speeches, shooting off pleasing executive orders in September and October, revving up the economy, delaying root-canal questions until post-election Decembers, and so on.  Many are the ways.  (Strategic action by parties or candidates figuring out whom or whether to run can also enter into this discussion, but in practice it probably doesn’t make a big dent at the presidential level.)

Thus, a party can gain a temporary advantage by running an incumbent.  The Democrats hold the all-time record with their four incumbent candidacies in a row in 1936, 1940, 1944, 1948—the last three during national security crises.  But over time the party advantage averages out.  And especially with the presidency constitutionally term-limited since 1951, the possibility of running incumbents doesn’t impair party changeability very much.  Open-seat elections, as in 2008, can offer a paradigm of freshness.

In sum, we see in the U.S. system a pretty good approximation of equality and changeability.

David R. Mayhew is Sterling Professor of Political Science at Yale University. His books include Congress: The Electoral Connection, Divided We Govern, and Electoral Realignments.



  1. I think the “changeability” will not occur in this election because it seems that the chance of the current party is higher than its competitor.