Congressional Elections and the Sixth-Year Myth

Congressional Elections and the Sixth-Year Myth

By Sean Trende - May 1, 2013

This is the first in a two-part series exploring the Democrats’ chances of taking back the House in the 2014 elections and, more broadly, what sort of result we should expect next year. The conclusion is that the Democrats’ odds are pretty slim; the 21 percent chance that the Iowa Electronic Markets currently assigns to that possibility seems rather Pollyannish.

At the same time, Republicans aren’t really in a position to duplicate the kind of gains they saw in 2010. Overall, my expectations would fall somewhere between a pickup of about five seats for the Democrats and a gain of about 15 seats for the Republicans.

To help guide our expectations, we first need to grapple with one of the most common pieces of conventional wisdom regarding this election. It’s the so-called “sixth-year itch.” Different pundits set it forth with varying particulars, but the gist of the argument is as follows: Since FDR, each party has had a horrible outcome in the midterm election occurring in the sixth year it controlled the presidency.

At a superficial level, this argument has some appeal. The following chart sets forth the results for the elections held in the sixth year a party has held the White House since 1932. It shows the percentage of seats lost by the president’s party lost in each such election:

The president’s party took significant blows in 1938, 1958, 1966, 1974, and 2006. Some analysts would also add 1946, when the Democrats lost 22 percent of their House caucus, and which came in amid the fourth Roosevelt/Truman term (a sort of “second sixth year” for the Democrats). The House GOP held its own in 1986, but lost a net of eight Senate seats, while in 1998, the GOP imploded amid its impeachment of President Clinton.

There are three problems here. First, there’s a certain amount of special pleading going on, making excuses for 1986 and 1998 that seem rather post hoc. Second, there’s an awfully small number of observations here for trying to deduce such a sweeping theory.

Third, the cutoff date is fairly arbitrary. In fact, the only reason for using 1938 seems to be that the rule simply doesn’t hold up all that well if you go farther back. Chart 2 takes us back to the end of the Civil War -- a necessary cutoff, because before that calamity, many states held their congressional elections in odd-numbered years, and therefore every election was a midterm to some degree. (A few states continued this into the 1870s, but the practice effectively ended in the 1860s.)

As you can see, the pattern doesn’t hold up as we expand our data. The Republicans were clobbered in 1874, amid the deterioration of the party’s position in the South and the onset of the Long Depression. But 1918 and 1926 were fairly normal midterm elections. In 1902, the president’s party technically gained seats, but only because the size of the House was increased; the Republicans’ share of the House still declined modestly that year.

If we expand our search to all midterm elections for two-term presidents, though, a pattern does stick out:

I’ve highlighted elections where the president’s party lost more than 10 percent of its caucus in the midterm. The pattern is this: Two-term presidents almost always get thumped in one midterm election, but they almost never get thumped twice. The one exception is the Republicans in 1870 and 1874, and that first loss was almost entirely due to the party’s unsustainable strength in the South.

This makes sense when we think about the factors that influence congressional midterms. Presidents rarely have substantial coattails during their re-election efforts, but often have them in their first election. So they tend to enter their first midterms with a large number of members in vulnerable seats. If they lose seats in the first midterm, then they’re pretty well cleared out for the second midterm. If they manage to avoid losing seats the first time, they tend to still be overexposed in their second midterm.

Similarly, it is unusual to make it through an entire eight-year term without hitting a rough patch in the economy. If a downturn happens early in the first term, things will be rough during that midterm election. But given the length of the business cycle, another downturn is unlikely to happen again in the second term (if there is one). At the same time, if a rough patch doesn’t appear in the first term at all, the odds are pretty good it will occur at some point in the second (again, given the business cycle).

All the other things that can go wrong for a president -- war, scandal, overreach -- are likely to happen at some point in an eight-year presidency. But it would be really unusual for them to occur throughout an eight-year presidency.

As a final way of looking at this, let’s try to be a bit more quantitatively rigorous. At this point, there are about as many congressional models as there are stars in the sky, and they suffer from many of the same problems as presidential models (perhaps even more so). But one I liked to use as a rough heuristic device for the 2010 and 2012 elections was borrowed from Gary Jacobson’s “Politics of Congressional Elections.”

It looks at the factors listed above, measuring the percentage of seats lost or gained by the president’s party with respect to the growth of per capita real disposable income in the election year, whether the president’s party held an unusually large number of seats (such that it was overexposed), and the president’s job approval. To increase the number of observations, I included all House elections from 1960 through 2012 (the real disposable income data series only runs back that far), but used a dummy variable for presidential and midterm elections.

We’ll talk about this model next time in some detail. For now, let’s just note that the model explains about three-quarters of the observed results in these elections. Most of the variables are significant, and the “exposure” variable is about 93 percent significant. It isn’t surprising that this one is a bit off, as it’s the most difficult to operationalize.

Anyway, the idea is this: If there really is something unique about sixth-year elections, then an added variable denoting elections occurring in a sixth year -- 1966, 1974, 1986, 1998, and 2006 -- should come back statistically significant as well.

It doesn’t. In fact it isn’t particularly close (for those stats geeks in the audience, the p is 0.56). This suggests that sixth-year elections don’t behave any differently than any other sort of congressional election.

So what does all of this mean for 2014? Well, we know that the election should be governed by traditional factors: the economy, the president’s popularity, how overexposed the party is. The fact that it is a sixth-year election is, by itself, pretty meaningless.

But we should also bear in mind that the fact that the Democrats got walloped in 2010 means that it would be highly unusual for factors to coalesce and create a really bad year again in 2014. This should put a damper on Republican expectations for 2014, and give Democrats some hope that they can make some gains, even if picking up the 17 seats needed to take the House is probably out of reach. 

Sean Trende is senior elections analyst for RealClearPolitics. He is a co-author of the 2014 Almanac of American Politics and author of The Lost Majority. He can be reached at Follow him on Twitter @SeanTrende.

When Republicans Win, Politics Are "Dysfunctional"
Dennis Prager · November 11, 2014
The Other GOP Wave: State Legislatures
David Byler · November 11, 2014
A President Who Is Hearing Things
Richard Benedetto · November 12, 2014
What Happened?
Thomas Sowell · November 11, 2014
Facing Do-or-Die Runoff, Landrieu Goes on Offense
Caitlin Huey-Burns · November 10, 2014

Sean Trende

Author Archive

Follow Real Clear Politics

Latest On Twitter