About this Blog
About The Author
Email Me

RealClearPolitics HorseRaceBlog

By Jay Cost

« What's Going On at the NRSC? | HorseRaceBlog Home Page | The Newsweek School of Presidential Psychology »

Is There a Third Term Curse?

I like to think that I have two tasks for this blog. The first is to clarify aspects of our politics that are cloudy. The second - which, I must admit, is nearer and dearer to my heart - is to clarify ways of thinking about our politics that are confused. In other words, I have both substantive and methodological goals for the Horse Race Blog.

Many times, methodological mistakes lead to substantive mistakes. Methodological mistakes can be subtle, and often times they are the product of over-enthusiasm. That is, analysts and pundits are so eager to offer something of substance that they commit some kind of methodological error that, in turn, leads to an erroneous conclusion.

One such error that I often see is a kind of correlation-as-causation fallacy. To be a little cheeky, we might call it the fallacy of the historical curse. I often read pundits who cite historical trends and then use those trends as an argument for why something will happen. For instance, last year, pundits, in their attempts to analyze the congressional elections, were wont to offer a litany of reasons to expect the Republicans to do poorly. One of those reasons was something to the effect of, "Parties of the President always do poorly in their sixth years." While it is essentially true that the party of the President does poorly in the second midterm, it is not a reason, or a cause. It is a tendency, or a correlation. Thus, it cannot be in one's litany of reasons. You cannot cite a tendency as a reason. Correlation is not causation. When you treat historical data that way, it is as if you are implying that there is some kind of curse - that is, an unexplained causal factor that can be explicated only by reference to the pattern that the factor creates. Of course, pundits are not actually enthusiasts of the occult. I imagine they are as "modern" and "naturalistic" as the rest of us. The effect, I am sure, is accidental.

This mistake can take several forms. In one form, the past reasons for the pattern might not be in effect in the current time period. This would be most likely to occur in an instance where your only reason to expect an event is the historical pattern. For instance, suppose that all identifiable metrics about last campaign cycle - except the "sixth year" tendency - favored the Republican Party. How wise would it have been still to predict GOP disaster? Not very. The historical pattern has a cause; if you cannot identify a cause to suspect disaster this year, you should not use just the pattern to predict a Democratic victory - though the pattern might inspire you to dig more deeply to see if potential factors might indeed be influencing the election.

In another form of this mistake, you overstate your case. If you give five reasons to expect something to happen, but one of the reasons is a historical trend that was caused by another of the reasons, then you really only have four reasons to expect something. In this instance, you have over-argued, which is not to say that we should divide your conclusion by some factor. It is simply to say that you have offered a bloated, and therefore imprecise, argument. It might still be the case that, when your argument is brought back down to its appropriate weight, you can still argue what you wish to argue. However, because of the imprecision - we cannot know until we have brought forth a clear version of your argument.

This latter error was on display yesterday in an otherwise excellent article by Frank Donatelli at Politico.com. Donatelli writes,

It is the worst of times for Republicans. President George W. Bush's approval ratings barely top 30 percent. Democrats have opened up as much as a 15-point lead in party identification, a gap not seen since the Nixon-Ford days of the 1970s. Key issues such as immigration and Iraq are causing major fissures in the Republican coalition. The GOP suffered a top to bottom defeat in the 2006 midterm elections, a leading political indicator that a change in party control of the White House will follow in 2008.

History also tells us that 2008 should be a Democratic year. Third terms for the in-party in power are notoriously difficult to win. The only nonincumbents since the middle of the 19th century who achieved this -- Republicans William Howard Taft in 1908, Herbert Hoover in 1928 and George H.W. Bush in 1988 -- all succeeded enormously popular presidents.

Let me say at the outset that I essentially agree with Donatelli's thesis. This is going to be a very tough election for the Republicans. I am not critiquing his argument so as to promote some kind of GOP bullishness. I am, rather, trying to make a somewhat more subtle point - methodological errors can reduce our substantive precision. Let me also say that his article is otherwise a sound analysis of what the GOP needs to do to make itself as competitive as possible. I am not suggesting we throw the baby out with the bathwater. Nevertheless, Donatelli does indeed commit the second form of the fallacy that I have described above.

It should be clear here that Donatelli offers a trend as a cause. Anytime you read, "History tells us," you can be near certain that this is what is coming next. Indeed, in his litany of GOP woes, he mentions: low job approval, low GOP party identification, issues that cut against the Republicans, previous electoral disaster, and history.

[N.B. His preferred phrase, "History tells us," is one of those non-sequiturs that is like fingernails on a chalkboard to my ears. History, speaking either literally or metaphorically, does not tell us much at all, at least not anything that was not self-evident to begin with. Unlike Newtonian physics, history does not present to us clean, crisp, and clear laws of human behavior. History is a messy subject. Historical insights are almost always matters of interpretation and disputation, conjecture and refutation.]

Mr. Donatelli offers a bloated list of reasons for Republicans to be dour. He should not include the historical pattern that he references. This can be seen more clearly if we unpack this historical trend - an action which, I believe, has some utility that goes beyond offering a response to this article. It can help us understand the importance of historical patterns in understanding this upcoming election, and it can offer us some general guidelines on how we should - and should not - use history in our analysis.

Mr. Donatelli writes, "The only nonincumbents since the middle of the 19th century who achieved this -- Republicans William Howard Taft in 1908, Herbert Hoover in 1928 and George H.W. Bush in 1988 -- all succeeded enormously popular presidents." First of all, another major sign of approaching correlation-as-causation is when you read a historical trend that sports an arbitrary cutoff date, as he offers here. What is so special about the middle of the 19th century? That seems a little "suspicious." Another "suspicious" element is that he does not factor in FDR, who of course won three terms (four, as a matter of fact!). His initial argument is about the perils of a party winning a third term. FDR was, of course, a member of the same party as FDR. So, he would have to be counted as an exception. Similarly, he fails to factor in TR, who won his party a third term in 1904. Taft's term was actually the Republican's fourth consecutive term.

Indeed, if we go through the whole history of the Republic, we can appreciate just what a limited pattern this is.


Between 1800 and 1824, the Democratic Republicans won seven consecutive presidential elections.

Between 1828 and 1836, the Democrats won three consecutive presidential elections.

Between 1860 and 1880, the Republicans won six consecutive presidential elections.

Between 1896 and 1908, the Republicans won four consecutive presidential elections.

Between 1920 and 1928, the Republicans won three consecutive presidential elections.

Between 1932 and 1948, the Democrats won five consecutive presidential elections.

Between 1980 and 1988, the Republicans won three consecutive presidential elections.

This would be seven successful attempts to win at least a third consecutive term. How many times has one party or the other failed to win a third consecutive term after having won two? Six: 1860, 1920, 1960, 1968, 1976, 2000. It is interesting to note that in three of these five failed attempts - 1960, 1968, and 2000 - only a fraction of the vote separated the two parties.

Take a closer look at those failed attempts. Three of the five came within 20 years of each other - and they all occurred during a turbulent time in American political life. Many have argued that this period was one of a dealignment that was due to the parties' inability to deal with recurring problems of national scope: race, crime, Vietnam, economic stagnation, Communism, and so on. In other words, it was the presence of tough issues that the party in power had failed to resolve that ultimately damaged their capacity to retain office. Interestingly, issues were the Democrats' problem in 1860, when they had failed to resolve the burgeoning sectional crisis; and 1920, when the nation had grown tired of their idea of America's role in the postwar world.

The historical pattern now makes intuitive sense, which is to say that we have now identified a reason for its existence. Why should we expect a party to lose in its quest for the third term? Our system is very hard to govern over a sustained period of time. We can generally expect that a governing coalition might be worn out by the time the opportunity for a third term presents itself. It has already acted on many of the issues that favor it. So, the issues that linger tend to be issues that hurt it and help the other party. A change seems called for - certain issues need to be addressed and the country no longer trusts the ruling party to address them. So, what happens? The party is saddled with declining partisan identification, declining presidential job approval, and a loss in the next election. This is why parties so often lose elections in attempts for their third go-arounds. Issues induce the pattern.

So, we should be able to see clearly now that Donatelli is "double-dipping." By mentioning issues, he has already accounted for what creates the pattern that he mentions. The pattern itself is thus not another reason to be added to the list. It is not an independent reason - it is, rather, a tendency that one of the reasons explains.

Again, let me stress that my point here is not to engage in Republican boosterism. As you should be able to tell by now, I am in general argument with Mr. Donatelli - 2008 looks to be trouble for the Republicans. I actually might be more bearish about Republican prospects than he because I think there is only a very narrow band of opportunities for the GOP to take up his third recommendation - namely, to separate itself from President Bush. What is more, my intention here is not to single out Mr. Donatelli as being the sole perpetrator of this argumentative fallacy. All in all, I thought his article was well written and well argued. And, anyway, I have seen many analysts make this kind of mistake generally - and I have seen a good number make this particular mistake about third terms.

My point here is simply two-fold: (a) on a methodological level, we should be mindful of committing the fallacy of the historical curse; (b) on a substantive level, the historical fact that parties have recently had trouble acquiring a third term is not, in itself, a reason to be bearish about Republican prospects. The reason to be bearish is the cause that induces the pattern seems to be present in this cycle for the GOP.

-Jay Cost