Tecumseh’s Curse and the two-term fatigue theory

Tecumseh and William Henry Harrison (W. Ridgway)

From 1952 to 1976, every presidential nominee from an incumbent party of two or more terms lost. This fact is the foundation, and the only foundation, for the two-term fatigue theory, which states that such nominees, all other things being equal, will lose. But this streak involved only four such nominees: Democrat Adlai Stevenson (1952), Republican Richard Nixon (1960), Democrat Hubert Humphrey (1968), and Republican Gerald Ford (1976). The two main parties alternated in power every eight years starting, supposedly, with the enactment of the 22nd Amendment, which prohibited third terms to individual presidents.

By contrast, from 1840 to 1960, every president elected in a census year died in office — William Henry Harrison (1840), Abraham Lincoln (1860), James Garfield (1880), William McKinley (1900), Warren Harding (1920), Franklin Roosevelt (1940), and John Kennedy (1960). This is Tecumseh’s Curse, and it is based on nearly twice as many data points, over five times as long a span. The curse is named for, and was supposedly delivered by, the leader* of the confederacy defeated by Harrison at the Battle of Tippecanoe, and I learned it as a child growing up in the neighborhood of the battleground. The curse is, needless to say, fiction. It involves taking a string of coincidences and imagining them to represent some law of the universe. In the absence of a compelling theory for why such presidents die in office, it has always been recognized by serious people as a string of coincidences. And the myth did not outlive my childhood; I was newly an adult when Ronald Reagan (1980) left office alive in 1989, despite being nearly 78 and having been the target of an assassination attempt. The curse had a sporting chance, but still failed. And for good measure, the next candidate, George W. Bush (2000), was unusually young and fit.

Garbage in, garbage out. We have bails of evidence that incumbency is an advantage running for reelection. The idea that sometimes it is nonetheless a disadvantage would have to depend on a lot of evidence and a compelling theory. But it does not. It depends on exactly four data points, and a theory no more thoughtful than voters get tired of parties after two terms in a way they do not after one, and this only applies to the presidency, and only began when two terms became the constitutional limit for individuals. The constitutional mandate is a particularly-strange element of this theory, since the nominees themselves are not directly affected.

For broader context, here are the presidential elections of the last century where one party had already served two or more terms, with the popular-vote winner highlighted. Because the fatigue theory supposedly affects voter attitudes, not Electoral College geography, Al Gore must be counted as the winner in 2000, and Hillary Clinton in 2016.

The two-term fatigue theory is rather convenient in its timing; the last three elections before the amendment went into force (1940, 1944, 1948) ended in victory for a two-term incumbent party, the Democrats. The new amendment would not have prevented Harry Truman from seeking reelection in 1952, and it did not affect the incumbent presidents in 1968 (Johnson) and 1976 (Ford), both of whom had taken over for an elected president mid-term. For the 22nd Amendment to have played a role in the respective defeats, we must apparently believe that voters instantly internalized the new two-term constitutional limit on individuals as a political norm that should be applied to parties as well.

Moreover, at least three of the four defeats for two-term parties had special circumstances. The incumbent in 1976, Gerald Ford, was never even elected as vice president, and had taken over for a president who resigned under threat of impeachment and removal after exposure of serious abuse of power. The incumbent in 1968, Lyndon Johnson, dropped out of the race because of a war that was growing more unpopular with his electoral base and the electorate at large. And the election of 1960 was extremely close in the popular vote (officially 113,000, or .17% of the vote), with the outcome probably affected by fraud in Illinois and possibly in other states; in other words, the two-term incumbent party in 1960 could reasonably be supposed to have won.

That we could identify special circumstances in all elections, if we wanted, is rather the point. There have not been that many US presidential elections, and even fewer of them have involved two-term incumbent parties since the enactment of the 22nd Amendment. The sample size is tiny, and the evidence for the fatigue theory depends heavily on the four relevant elections from 1952 to 1976.

Since 1976, there have been five elections with a two-term incumbent party (1988, 1992, 2000, 2008, and 2016), and the incumbent party has won the popular vote in three of the five. That means that, even before Hillary’s victory in 2016, two-term incumbent parties had won two of the last four relevant elections. Nearly everyone acknowledges that our electorate has become more polarized and partisan in recent decades, significantly affecting electoral outcomes. This should change our view of the relevance of past elections, and give more weight to the more recent. Political scientists, and their dependent commentators, were nonetheless flogging the two-term fatigue theory as part of the evidence that the Republicans were favored to win the 2016 election.

Perhaps they will just disregard the popular vote by saying the Republicans did win this election, without qualification, just as they did win, as the non-incumbent party, in 2000. For this approach to apply, the two-term fatigue theory can only be supposed to work through an ever-shifting magic in the Electoral College, affecting certain states in some elections, and different states in other elections. For 2016 the Rust Belt states of Wisconsin, Michigan, and Pennsylvania are widely cited. In 2000, Florida was key, while Al Gore could also have won with either New Hampshire (which has voted Democratic in every election since), or his home state of Tennessee. Unlike Bush in 2000, Donald Trump won Pennsylvania, Michigan, Wisconsin, Iowa, and Maine’s second congressional district, and won Florida by enough that it cannot be disputed. Unlike Trump in 2016, Bush won Virginia, New Hampshire, Colorado, and Nevada. Using today’s electoral-vote count and not including Florida, Bush and Trump only won 224 electoral votes in common. Even with Florida, it’s still not enough for a win.

If instead the theory’s proponents claim special circumstances in either popular-vote loss, they must then acknowledge special circumstances in their own supporting evidence as well. In any case, the weakness of any one data point is less important than the paucity of data points, and the fact that they were selected so advantageously. Starting the clock in 1952, in exclusion of a run of three counterpoints, is a patent act of sample hacking. Before Hillary, proponents of the fatigue theory would point to incumbent parties losing six of the last eight times. But before 2016, incumbent parties had actually won five of the last eleven, as well as two of the last four, and six of the last fourteen if we use the arbitrary cutoff of a hundred years. Any one of those alternatives looks essentially like random variation. And that’s accepting the validity of Nixon’s loss and before accounting for Trump’s. Two-term incumbent parties have now won three of the last five affected elections, six of the last twelve, and seven of the last fifteen, by the only relevant measure of national voter sentiment — the popular vote.

Political scientists might point to their models, some of which incorporate the two-term fatigue theory as one element, and claim that it improves the models. That is surely less true as Trump’s popular-vote loss is factored in. But regardless, the inclusion of any data in a model must be justified by theory. If political scientists can’t explain why voters reject two-term incumbent parties, but not individual incumbents (who still have an advantage), and why voters only began to do so with the passage of a constitutional provision that did not affect parties at all, then they have no business including it in their models, any more than they should include the position of the planets in the zodiac. The two-term fatigue theory never made sense. No one would ever propose it today based on the historical data as they now stand, or think to include it in a predictive model. Two-term incumbent parties have no disadvantage. The two-term fatigue theory is dead.

*This sentence originally described Tecumseh as “Harrison’s opponent at the Battle of Tippecanoe”. Tecumseh was actually away at the time of the battle. If local connections mattered for anything, I’d have remembered that.

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store