Third Time Not the Charm?

In the summer of 2014, with midterm election season still underway, the commentariat was already looking ahead to the 2016 presidential race. Hillary Clinton's nomination seemed inevitable even then, and a slew of articles warned (or hoped) that she might fall victim to the dreaded "third-term curse"—the supposed propensity for American voters to rebel against giving the White House to the same political party for three straight elections. In the pages the New York Times, for example, John Harwood wrote:

For all the strengths she would bring to a 2016 race, Mrs. Clinton would face a significant historical obstacle. American voters have demonstrated their reluctance to award the same political party a third consecutive term in the White House.

Fox News—surprise!—published a similar article just a week later. Under the headline "The historical urge that could hurt Hillary," Howard Kurtz wondered:

For all the talk about Hillary Clinton ... bringing Bill back to the White House, I've long felt that there is a sleeper issue in play for 2016: Will voters want to give Democrats control of the White House for 12 straight years?

The third-term curse is likely to remain a talking point throughout the 2016 campaign, as the tendency to try and identify patterns is hardwired into the human brain (the scientific term for this is 'apophenia'). However, like most (all?) curses, it is nonsense. It is numerology masquerading as political analysis, and the John Harwoods and Howard Kurtzes of the world should know better.

The most obvious problem with predictions of this sort is the small sample size. There have been only 57 presidential elections in American history, only 28 since the start of the 20th century, and only 17 since World War II (generally considered the start of the "modern era" by historians). Even 57 trials is a very small data set—no one would accept that a clinical trial with 57 people is enough to judge the efficacy of a new drug, nor that 57 at bats is enough to evaluate a baseball player's abilities, nor that a warranty is only needed for the first 57 miles of a new car's life. If we drop the sample size to 28 or 17, our conclusions become even more tenuous.

Complicating the small sample size is the fact that each election is distinctive, due to the qualities of the candidates, the state of the economy, whether the nation is at war or not, the likelihood that various demographic or ideological groups will vote or stay home, and so forth. In 2016, the pollsters at Gallup and SurveyUSA and Quinnipiac will lose much sleep (and maybe a bit of sanity) trying to figure out exactly how the election is similar to and how it is different from the 2012 election. Comparing the apple that is 2016 to the orange that is 2008 would be even harder, comparing to the pear that is 2004 or the pineapple that is 2000 would be harder still. And if you were to ask John Zogby to project 2016 based on what he knows of the electorate in 1988 (kumquat?) or 1960 (durian?), he would just laugh.

There are also a pair of issues specific to the third-term hypothesis. The first is that it only works if we start with (or after) the year 1948. Since then, candidates trying for a third term for their party have had 1 win and 5 losses. Prior to that, however, they won 7 of 8. If John Harwood or Howard Kurtz was writing in 1960 rather than 2014, their articles would surely predict a near-certain victory for Richard Nixon. Put another way, believing in the third-term curse requires us to rely on a tiny sample size of six, while at the same time dismissing the first eight trials that happen to prove the opposite.

Now, it may seem justifiable to make 1948 the cutoff: It was the year the 22nd Amendment was passed and it was (as noted) the first post-World War II election. However, this leads to the second issue with this hypothesis: There is no compelling explanation for why the third-term curse exists. With many macro-level developments in politics, we can identify a clear cause and effect. For example, black voters knew very well who freed the slaves, and so were overwhelmingly Republican until the passage of the New Deal of the 1930s and the Great Society and civil rights legislation of the 1960s. These programs were the work of Democratic presidents and they helped black voters enormously. As such, the black community largely switched its allegiance to the Democrats, where it remains today. By contrast, there is no such cause and effect relationship when it comes to third terms, no compelling reason why the 22nd Amendment, or the end of World War II, or the Cold War, or suburbanization, or any of the other events of the 1950s or the 1960s would cause the American public to develop "party weariness" when they previously did not have that tendency.

Similarly, there is also no particularly good explanation of how the tendency works, or would work, when it is in effect. Kurtz takes a stab, with a fairly typical breakdown:

I think much of the country collectively senses that a party in power for two terms may be running low on ideas and has already tapped its most talented folks to run the government. Plus, the party in power accumulates all the baggage of being blamed for everything that has gone wrong over eight years. And even if things are going well—which they're decidedly not at the moment—"time for a change" remains a powerful slogan.

We can practically see Kurtz at his desk as he wrote this, still trying hard to make sense of Barack Obama's victory in 2008. And again, this assessment seems reasonable at first glance, but does not stand up to more careful scrutiny. Note, first of all, that much of the "explanation" actually has little to do with third terms. Errors made by the party that controls the White House, and the question "how is the country doing?" are factors in every presidential election, third term or not. Meanwhile, keep in mind that 35% of the country is going to vote Republican even if the candidate is Barry Goldwater, and 35% is going to vote Democratic even if the candidate is Walter Mondale. These individuals would be delighted to have their party win a third, or fourth, or tenth term in a row. So we are really only talking about a fraction of the electorate—the independents, and the fairly small number of Republicans and Democrats who are willing to cross the aisle. Are we to believe that a large number of these individuals are influenced by some vague sense that "enough is enough," as opposed to more practical concerns like the state of the economy, or their feelings about the candidates?

In short, then, the phrase "third-term curse" should be banished from the lexicon, relegated to the attic along with "Know Nothing," "Billy Beer," and "Ross Perot." Let us replace it with two words that are vastly more instructive when understanding presidential elections: Incumbency and contingency.

Incumbency is the more familiar of the two concepts. When an incumbent appears on the ticket for a presidential election, voters have a very clear cut choice: more of the same, or not. They know (or, least the think they know) exactly what they will be getting if they give the President another four years. This is when the electorate generally speaks most forcefully. Not including the semi-election of George Washington to his first term (he was unopposed), 32 of America's 56 presidential elections (57%) have involved an incumbent, and those 32 have given us all 10 of the biggest Electoral College victories in U.S. history (with the incumbents winning 8 out of 10). If we limit ourselves to the 17 elections since 1960, there have been 11 with an incumbent on the ticket (64%). Those 11 have given us the five biggest electoral victories since World War II (with the incumbents winning 4 out of 5).

In contrast, in elections with no incumbent, you are always buying a pig in a poke. As such, nonincumbent elections are much likelier to be close. Six of the 10 closest contests in Electoral College history and eight of the 10 closest popular votes came in the 43% of elections with no incumbent on the ticket (3 of 5 in both categories among the elections since World War II). Of the six elections in U.S. history decided where the winner and loser were separated in the popular vote by less than 1%, five had no incumbents (that includes two elections since World War II: 1960 and 2000).

Note that these are patterns of a sort, but ones supported with substantive proof and a clear cause and effect relationship. Both common sense and cold, hard evidence support the assertion that nonincumbent elections have a much higher likelihood of being close than incumbent elections. And of the 14 occasions in U.S. history when the American public has had the opportunity to award a third term to a party (six since World War II), an incumbent has been on the ballot only three times: 1904, 1944, and 1976. This means that when we are trying to evaluate the third-term curse (after 1948), or the "third-term charm" (before 1948), we are primarily dealing with elections that are likely to be close. This is where the second concept, contingency, comes in.

Contingency will be a less familiar notion than incumbency, particularly because the term means different things in different professions (i.e., the lawyer's contingency is very different from the philosopher's contingency). The American Historical Association (AHA) explains the historian's version:

To argue that history is contingent is to claim that every historical outcome depends upon a number of prior conditions; that each of these prior conditions depends, in turn, upon still other conditions; and so on. The core insight of contingency is that the world is a magnificently interconnected place. Change a single prior condition, and any historical outcome could have turned out differently. Lee could have won at Gettysburg, Gore might have won in Florida, China might have inaugurated the world's first industrial revolution.

In layman's terms, the AHA is describing something that every sports fan already knows: When a competition is close (the Civil War, the 2000 Presidential election, Asia vs. Europe in the 1700s), the winner will often be decided by small things. These small things are often random (and thus unpredictable) and may well be beyond the control of participants.

Putting it all together, the argument here is that the elections that are supposedly subject to the third-term curse are likely to be close, because they generally involve only nonincumbent candidates. And when an election is close, it is likely to turn on small things that might have resulted in a different outcome. If we examine the post WW II elections that have given rise to the "curse," we can see clear examples of this:

1. The Election of 1960, John F. Kennedy (D) defeats Richard M. Nixon (R):The popular vote does not decide elections, but it is a more precise indicator of closeness than the electoral vote. And by that measure, this was the closest election in American history, with a difference of only 112,827 votes out of 34,220,984 cast—a paper-thin margin of 0.17%. We can point to all manner of small things that might have changed the outcome. For example, in the summer of 1960, someone asked outgoing President Dwight D. Eisenhower, "Can you think of a major contribution that [Vice President] Nixon has made to your administration?" Eisenhower famously replied, "Well, if you give me a week, I might think of one." Eisenhower did not like Nixon, and so may have wanted to derail his candidacy, though the President later claimed he was tired and did not properly hear the question. In any case, if Ike had answered in an even slightly affirmative fashion—for example, "He helped us with Khrushchev," it might well have swung the election for Nixon.

   A few months after that incident, Nixon met JFK in the first-ever debate between two presidential nominees (though we might think that presidential debates are a longstanding tradition, they are actually a made-for-TV event that dates back to 1960). JFK took a nap that afternoon and also made sure to get some sun, such that his complexion was tan. Nixon was sick that day, had not slept very much, and had his characteristic five-o'clock shadow. As the story goes, JFK was overwhelmingly declared the winner by voters who could see him on TV, and Nixon was given the win by radio listeners who could not see the candidates. This is likely overblown—TV owners back then were more likely to be urban and Democratic, those who still used radios were more likely to be rural and Republican—but there is little doubt that the debate helped JFK more than Nixon, and that his more appealing appearance was a part of that. Certainly Nixon thought so, he never again appeared in a debate without having makeup applied first.

   Election night in 1960 brought its share of things that might have gone differently, particularly in the states of Illinois and Texas, which Kennedy won by roughly 9,000 and 46,000 votes, respectively. Had Nixon taken both states, he would have won the White House. 55,000 votes is a tiny number, one that can turn on all manner of trivial factors—weather, gas prices, lines at the polling places (or lack thereof), counting mistakes, a flu outbreak, a power outage, etc. Those totals are also very manageable for those skilled at voting fraud, a charge leveled (probably correctly) at Chicago Mayor Richard J. Daley and Texas Senator and vice-presidential candidate Lyndon B. Johnson in 1960. Kennedy's win was certainly not due to a massive "throw the bums out" view or a rejection of the Republican Party.

2. The Election of 1968, Richard M. Nixon (R) defeats Hubert H. Humphrey (D): In the case of the third-smallest popular vote margin since World War II (511,944 votes of 31,783,783,cast, or 0.70%), the shoe was on Nixon's other foot, as the contingencies broke in his favor. The most familiar, and most unfortunate, was the assassination of Robert F. Kennedy in June of 1968. RFK had emerged as a clear frontrunner for both the Democratic nomination and the presidency, winning the California primary on the night of his assassination. If he had not gone to the Ambassador Hotel that night to celebrate his victory, or he had not exited through the kitchen (a last-minute change of plans), or if Sirhan Sirhan had aimed differently (unlike his brother, RFK was not killed instantly), then Nixon would likely have been a significant underdog in 1968.

   Ultimately replacing Kennedy as the Democratic nominee was Vice President Hubert H. Humphrey, who found himself in a delicate position due to his association with unpopular President Lyndon B. Johnson. Humphrey wanted to distance himself from LBJ, but had to avoid direct criticism, as he needed the President to dial back the war effort in Vietnam. At the same time. Humphrey was also uncertain what to do about Democrat-turned-Dixiecrat George Wallace, who was running as an independent and attracting a lot of support from Democrats who were also racists. It was not until late August, when his campaign appeared hopeless and Nixon was leading by 10-plus points in the polls, that Humphrey finally took action. He subtly distanced himself from LBJ by ceasing to identify himself as Vice President, switching instead to "Democratic nominee." He also began to attack Wallace as appealing to the worst parts of human nature, while calling for a return to the "politics of joy." Thus began a remarkable rise in the polls that was aided by the "Halloween Peace" one week before the election: LBJ's announcement that he had suspended bombing in Vietnam (true) and that a peace agreement was possible (probably false). Humphrey very nearly took the election; had he shifted gears earlier, or if the Halloween Peace had come two or three weeks before the election rather than one, the outcome could easily have been different.

3. The Election of 1976, Jimmy Carter (D) defeats Gerald R. Ford (R): This is the only election since World War II where an incumbent was trying to claim a third term for his party. Of course, the main story here—and the reason for the election's uncharacteristic closeness—was Watergate. Americans were angry about the scandal, and irritated that Nixon was not punished due to Ford's pardon, and they decided to punish the President.

   Watergate is a major historical event, and not truly an example of contingency in action. Nonetheless, the impact of small events was still felt in this election. The most important of these came on June 1, 1975, when Ford arrived in Austria on a state visit. It was a rainy day and, as he walked down the damp steps of Air Force One, he slipped and fell (about six feet). This could happen to anyone, as any male who has worn dress shoes in wet weather can tell you. Nonetheless, NBC's Saturday Night which was then so new that it wasn't even known as Saturday Night Live yet, seized upon the incident. With comedian Chevy Chase taking the role of Ford, the show portrayed the President as a clumsy, bumbling fool who consistently spilled his drinking water, stapled his own fingers, and—of course—fell over nearly everything in his path. 1976 was a close election—the popular margin was 2.06%—even with Watergate. More than a few historians and political scientists have argued that Ford would have survived the pardon of Nixon, but for a rainy day in Austria and a late-night TV show.

4. The Election of 1988, George H. W. Bush (R) defeats Michael Dukakis (D): This election is ignored by believers in a third-term curse, since it is the one time since WW II the curse has not "held." But despite the fact that Bush was the sitting Vice President and that he (unlike Nixon) had the endorsement of a popular incumbent, he only won the popular vote by a margin of 7.72%. Not a close election, exactly, but not a landslide either. Were there small but meaningful events in 1988 that might have added up to 7.27%? Perhaps. The Lee Atwater-produced "Willie Horton" ad argued that Michael Dukakis' attitude towards crime was so liberal that it put white Americans at risk of being raped or murdered by marauding black felons. Though Bush disavowed the clearly racist ad, and it only aired once, it nonetheless hurt Dukakis with urban and working-class white voters. Dukakis also damaged himself, most famously by posing for a photo of himself in an M1 Abrams Tank. The purpose of the photo-op was to communicate that Dukakis was strong on national defense and up to the task of being Commander-in-Chief. The effect, however, was the opposite—the picture made him look small and silly, and significantly undermined voters' confidence. There is no doubt Dukakis was fighting an uphill battle against the perceived protege of Ronald Reagan, but it well within the realm of possibility that that hill might have been climbed.

5. The Election of 2000, George W. Bush (R) defeats Al Gore (D): This is the textbook example of contingency, as the AHA reminds us. If West Palm Beach had used something other than butterfly ballots, if the elderly voters in Palm Beach had cast their votes more carefully, if Ralph Nader had stayed home that year, if Fox News had stayed quiet for another hour on election night, if the Supreme Court had one more Democrat—any or all of these things would have handed the presidency to Al Gore, who took the popular vote by 0.51%.

6. The Election of 2008, Barack Obama (D) defeats John McCain (R): Of the five elections ostensibly affected by the "curse," this is the biggest blowout, with Barack Obama's 7.27% margin of victory barely trailing George H. W. Bush's total in 1988. If we were to pick a turning point, we might choose the day after the Republican convention. On that day, John McCain went with his advisers instead of his instincts and announced Sarah Palin as his running mate. Though she initially generated some excitement, Palin was a terrible choice. She comes from a state whose three electoral votes were already in the Republican column, and she appealed to voters who had few options but to vote for McCain anyhow. Her tea party-ish rhetoric and tendency to embarrass herself ("I don't know what magazines I read") turned off independent voters. Had McCain gone with his preferred choice—conservative Democratic senator Joe Lieberman—he would have reaffirmed his status as a maverick and his willingness to reach across the aisle. He also would have avoided needless embarrassments—if nothing else, Lieberman was a seasoned politician who did not make foolish mistakes. A McCain-Lieberman ticket may have caused the social conservatives to stay home, or may have otherwise lacked the juice to overcome the Obama Hope machine; we will never know. There can be little doubt, however, that McCain sealed his defeat on August 29, 2008.

Here are the data for the "third-term elections" discussed above.

Year	Democratic Nominee	Democratic %	Republican Nominee	Republican %	Winner Margin	Cursed?
1960	John F. Kennedy	49.72%	Richard Nixon	49.55%	0.17%	Yes, but very close
1968	Hubert Humphrey	42.72%	Richard Nixon	43.42%	0.70%	Yes, but very close
1976	Jimmy Carter	50.08%	Gerald Ford	48.02%	2.06%	Yes, but close
1988	Michael Dukakis	45.65%	George H. W. Bush	53.37%	7.72%	No, huge win for incumbents
2000	Al Gore	48.38%	George W. Bush	47.87%	0.51%	Yes, but very close
2008	Barack Obama	52.93%	John McCain	45.66%	7.27%	Yes, but due to other factors

As can be seen, in three of the "cursed elections," the incumbent party lost by less than 1% and in one case (1988) it won. These are hardly evidence of a "throw the bums out" mentality after two terms. In the other two, one of them was a 2% win by Carter. The only one of the six where one can argue "time for a change" was perhaps a factor was 2008, but McCain's poor choice of Veep and the collapsing economy also played a big role. In short, the data do not support any kind of "third-term curse."

In 2016, there will be no Watergate or Vietnam War. Barack Obama is unlikely to cut Hillary Clinton off at the knees by openly disparaging her term as Secretary of State, and no candidate is likely to be assassinated. Lee Atwater is deceased, Sarah Palin's career is over, and butterfly ballots are no longer in use. As such, the six elections here are Red Delicious, Granny Smith, McIntosh, Fuji, Gala, and Pink Pearl apples to 2016's Navel orange. If those apples hold a meaningful lesson, it is this: Presidential elections are not not governed by "curses" or vague feelings like "voter fatigue," but instead by concrete things, large and small, some of which may not reveal themselves until deep within the election season.

-- Zenger