How to Understand the Incumbent Rule

How to Understand the Incumbent Rule

By Sean Trende - October 12, 2010

One of the most cherished rules in political handicapping is the so-called "incumbent rule." The actual "rule," derived from a famous 1989 article by Nick Panagakis, holds that an incumbent under 50 percent in the final poll taken before Election Day will usually go on to lose. But most observers today take a broader view of the "rule," noting that incumbents who fall below 50 percent in polls commissioned even earlier in the cycle frequently go on to lose in the fall.

In a post published yesterday, Nate Silver attempted to throw a bit of cold water on this election analysis rule of thumb. He suggests that the focus on an incumbent's number in the polls is misplaced and that, despite the conventional wisdom, incumbents polling under 50 percent often win re-election.

First, it is important to understand that Silver concedes that incumbents who are both under 50 percent and trailing their opponents are in (quoting Michael Barone) "deep doo-doo." His analysis focuses exclusively on those incumbents who are under 50 percent and leading their opponents.

But by changing the scope of his data set, Silver excludes an awful lot of elections which renders his data set an apple when compared to the orange everyone else is looking at. For example, he shows three Senate incumbents in 2006 who were both under 50 percent and leading their opponent. Two of them went on to win. That 66 percent win rate makes it look like these under-50 incumbents were doing pretty well.

But there were another four Senate incumbents in 2006 who were both below 50 percent and trailing their opponents. All four lost. Overall, 71 percent of the 2006 Senate incumbents who were under 50 percent at this point in the race lost in November.

Likewise, in 2008, five of the six Senate incumbents who were both under 50 percent and leading their opponent won - 80 percent! But overall, 54 percent of the Senate incumbents who were polling under 50 percent lost.

Indeed, Silver's earlier data suggests that the "incumbent rule" is viable even earlier in the cycle than today. In April, he compiled a lengthy list of incumbents' polling data from January to June of 2006, 2008 and 2009. These polls showed that a third of incumbents who polled below 50 percent in polls even at this early time in the cycle went on to lose in the fall, while nearly 2/3 of those below 45 percent lost in November. Only one of the incumbents who was above 50 percent lost: George Allen.

All told, this makes for a pretty good rule of thumb. It helps us winnow out races that are unlikely to be competitive on Election Day and focus more carefully on the incumbents who are really at risk of losing. But it should remain just that - a rule of thumb. The truth is that political science doesn't have much in the way of laws, because we're dealing with extremely complex systems comprised of millions of individual decisions. Instead we have a series of tendencies, some strong (presidents' parties lose seats in midterm elections) and some not so strong (presidents in their sixth year in office have horrendous sixth year elections). And yet far too often, analysts see a poll showing an incumbent under 50 percent and are too eager to shout "Incumbent Rule! She's DONE!"

Careful analysts, by contrast, use adverbs like "usually" or "potentially" when describing the fate of incumbents under 50 percent. Indeed, if you look at the three analysts quoted in Silver's piece, all three, including yours truly, use some kind of qualifier on the rule.

To be more specific, I try to keep six qualifiers in mind whenever dealing with the "incumbent rule." Most of these are just common sense, but I think they're important to remember:

1) The rule is weaker in races where the incumbent is just under 50 percent. An incumbent right at 49 percent is not inherently in deep trouble. After all, an incumbent can usually get some undecideds to vote for them, and it was virtually certain that Mike Huckabee would get the .1 percent of voters he needed to get to 50 percent (Silver lists him at 49.9 percent) But such an incumbent is worth keeping an eye on to see if they fall any further, as 49 percent is the point at which incumbents begin to lose. As I've written before, " [t]he rule of thumb is that any incumbent under 50 percent is in trouble, and any incumbent under 45 percent is deep trouble."

If you look at the 35 Senate and gubernatorial incumbents Silver lists who (a) polled under 50 percent, (b) led their opponent and (c) went on to win in November, you see that almost 60 percent of them were at 48 or 49 percent of the vote.

2) The rule is weaker in races that haven't really engaged yet, where candidates have huge leads. You do need to pay some attention to the spread, especially in House races, which frequently do not engage until late in the cycle. In 2002, Rob Simmons led 46 percent to 24 percent in the polls. Today, Patty Murray leads Dino Rossi 48 percent to 47 percent. Simmons was obviously in a stronger position than Murray is in today, since he only needed to win 13 percent of the undecideds, as opposed to the 40 percent Murray needs to win (it is worth noting that Simmons only managed to win about 27 percent of them, even amidst a GOP mini-tsunami).

3) Beware polls that don't push leaners. You can frequently get an incumbent below 50 percent if you don't ask undecideds for whom they are leaning toward voting. The phenomenon is fairly common in academic polls and campaign polls (Silver rightly didn't include the latter). This is especially important when it occurs in conjunction with #1. Obviously these require some nuance to identify, but pollsters who don't push undecideds are for some reason especially common in Pennsylvania, Rhode Island, and Arizona. Today's Arizona gubernatorial poll from Behavior Research Center is a great example of this phenomenon.

4) The rule is weaker in races where the incumbent is an appointee or is succeeding a retiring governor. The theory behind the incumbent rule is that if an incumbent hasn't convinced voters that he has something to offer after several years in office, he isn't going to do it in the last month of an election. But if the candidate has only been in office for around a year (i.e., Menendez in 2006, Celluci in 1998) this insight becomes less useful. If Senator Michael Bennet of Colorado were performing a few points better in the polls right now, I wouldn't consider his situation nearly as dire.

5) The rule is weaker in races where the challenger is a de facto incumbent. I wrote on this in more detail here, but the incumbent rule is weak late in presidential elections, where both candidates are universally known by the fall. But it can be weak even in Senate and governor's races as well. It was one thing for Jesse Helms to be tied with Governor Jim Hunt in 1984, it was another thing altogether for Jim Webb to be tied with George Allen shortly before the election. This is kind of the inverse of #4 - when both candidates are extremely well-known, the theory behind the rule doesn't really work. This is probably the best thing that Patty Murray in Washington and Martin O'Malley in Maryland have going for them right now.

6) It doesn't apply to races where there is a strong third party challenge for obvious reasons. Among others, the crazy four-way Texas gubernatorial race in '06 really shouldn't be included in the dataset. Rick Perry was at 36 percent in a race and where 26 percent could theoretically win.

This might seem like a lot of caveats, but all political science rules have these. Even the midterm loss rule comes with a caveat that the president has to have an approval rating under 60 percent or so before losses become more likely than not. And most of the caveats above only apply in a small number of cases (appointees, two incumbents running, strong third parties).

At the end of the day, despite Silver's analysis, there just isn't much of a reason to jettison the traditional "incumbent rule" in favor of a "which candidate is leading" rule. Especially in Senate and governor's races - which typically include more robust and better polling than House races - the incumbent rule, properly understood and applied, is a useful guide in forecasting election outcomes correctly.

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.

Sean Trende

Author Archive

Follow Real Clear Politics

Latest On Twitter