Network News

X My Profile
View More Activity

Voter rage = 87% reelection rate?

With Gallup registering an all-time high in the percentage of people who say that their congressperson does not deserve to be reelected, it looks like we're headed for a massive wave of anti-incumbent destruction this November. Or does it?

John Sides took the data from 1992 to 2008 and plotted a graph seeing whether that poll question predicts incumbent-reelection rates. Here's how it looks:

housereelection-thumb.png

Sides concludes:

There is a relationship between responses to this item and the reelection rate (and it’s statistically significant, in fact). But the relationship is substantively very small. Perhaps the best evidence is the predicted reelection rate I calculated based on the 1992-2008 data, plugging in the most recent Gallup poll, in which a record 40% declared that their member did not deserve reelection. What is the predicted incumbent reelection rate?

87%.

By Ezra Klein  |  June 8, 2010; 12:00 PM ET
Categories:  2010 Midterms  
Save & Share:  Send E-mail   Facebook   Twitter   Digg   Yahoo Buzz   Del.icio.us   StumbleUpon   Technorati   Google Buzz   Previous: Tom Toles is worth a thousand words
Next: Lunch break

Comments

Has he taken into account that two sitting Senators are already toast? And a third could be today?

Posted by: Calvin_Jones_and_the_13th_Apostle | June 8, 2010 12:18 PM | Report abuse

keep telling yourself the dems are fine!

Posted by: obrier2 | June 8, 2010 12:22 PM | Report abuse

Sadly this counts as an Ezra-fail.

It should be forbidden for anyone with pretensions to wonkiness to use a regression model to perform acts of extrapolation. Only Fox news is permitted to perform such acts.

There are no grounds for assuming that the linear relationship between the predictor and the response will hold up when the proportion saying that their representative deserves reelection exceeds about 33%. In fact, I'd say that the linear relationship looks a little shaky even for the eight datapoints plotted. Logistic regression is probably more appropriate here.

And no, the fact that the slope is a statistically significant is not relevant to these arguments. Besides, the calculation of the slope's significance is based on the assumption of homoskedasticity. Does that plot look homoskedastic to you? (Not that the concept even makes much sense for a sample size of eight, anyway!)

The only value here is a timely reminder that reelection rates are typically very high, even with relatively high levels of voter dissatisfaction.

Posted by: Unwisdom | June 8, 2010 12:23 PM | Report abuse

87% = 56 incumbents lose their seats in the House of Representatives and 13 incumbents lose their seats in the Senate. I would argue that those are not insignificant numbers.

Posted by: js4981 | June 8, 2010 12:51 PM | Report abuse

Are you commenting on the hyperbole of a vocal minority affecting re-election rates? Seems a little confusing.

Posted by: miqcie | June 8, 2010 1:19 PM | Report abuse

1. 87% *is* an insanely low retention rate; mock it all you want, but it's a fact.

2. 87% sounds about right; as another commenter pointed out, there's already 3 senators who lost, and we're not even done with primaries yet!

Posted by: mudlock | June 8, 2010 1:29 PM | Report abuse

"87% = 56 incumbents lose their seats in the House of Representatives and 13 incumbents lose their seats in the Senate. I would argue that those are not insignificant numbers."

Strikes me as not a terribly significant level of turnover -as long as it is weighted roughly evenly by political party- if only 13% of seats end up being truly in play, that's not huge.

In other words, if 13% of the Democratic incumbents are replaced by Republicans, but 13% of Republican incumbents are replaced by Democrats, there is no seismic shift. By how much the Republicans can overcome that anti-incumbent spirit on their own side will determine just how significant the election will be.

Posted by: Patrick_M | June 8, 2010 1:55 PM | Report abuse

"1. 87% *is* an insanely low retention rate; mock it all you want, but it's a fact.

2. 87% sounds about right; as another commenter pointed out, there's already 3 senators who lost, and we're not even done with primaries yet!"


3 Senators? The polls have not closed in Arkansas, but Bill Halter appreciates your optimism.

And obviously in the general election Utah will replace a Republican with another Republican, so that race will have zero impact on the numbers for the two parties in the Senate. It is way too early to predict outcomes in Pennsylvania, or in Arkansas (no matter who comes out ahead there).

Posted by: Patrick_M | June 8, 2010 2:03 PM | Report abuse

Incumbency bears huge advantages of name recognition, and this is reinforced by gerrymandering, seniority, and campaign finance restrictions. The practical range in which the retention rate can vary is thus quite small, making 87% more significant than it otherwise might seem.

Posted by: tomtildrum | June 8, 2010 11:40 PM | Report abuse

The comments to this entry are closed.

 
 
RSS Feed
Subscribe to The Post

© 2010 The Washington Post Company