We've all watched as Bush's approval ratings hover in poisonous territory for almost a year, and wondered how much it mattered. After all, Bush is not on the ballot this year. Who cares about him?
Last week I showed that as Bush's approval ratings have gone down, so has the net favorability of the entire Republican party. This time, we'll see just exactly how Bush's approval affects individual House contests. To continue with the storm metaphor, looking at numbers only (not campaign funds, operations, and viability of opponents) the Republicans are extremely, extremely vulnerable. Hurricane Bush is the worst thing in years for the Republican party, and the forecast is grim.
Click to enlarge.
Below, an explanation of the forecast, including discussions on Bush's approval in districts, voter motivation advantages, and exactly how Bush's approval relates to the performance of Republican candidates. And lots of graphs, of course. (Cross posted at MyDD.)
We start by trying to find an estimate of Bush's approval ratings in individual congressional districts. (For an explanation of the forecast, skip to the end!) Fortunately, we have
SurveyUSA's 50 state polls to work with, and Bush's approval has been relatively steady since June, with the exception of a brief bump for September 11. We can plot approval level for states versus the percent of the vote Bush received in 2004 and find a very nice correlation:
Click to enlarge.
The dark green line is the regression; the two light green lines show the upper and lower limits of Bush's approval given a certain percent vote in 2004. The estimate of Bush's approval has a range of plus or minus about 6 points. We would expect measured approval outside of this range 5% of the time. Previously, I mapped approval ratings by district that serve effectively as upper and lower bounds, so see this diary for district-level maps representative of the high end of Bush's current approval, and this diary for the low end.
Over the last several months, we've had a gold mine of data from Survey USA and Majority Watch from congressional districts, including approval ratings of Bush. So now we can check the accuracy of the predictions:
Click to enlarge.
The lines are the predictions from the statewide data. The observed values for House districts fall nicely in the predicted range, although SurveyUSA's results are generally on the higher side of the range. This is not surprising, considering these results are for likely voters, but the regression was developed for all adults. We might have predicted all observed approval ratings should be higher than what we see, as Republicans generally gain an advantage among likely voters, but not so much this year.
So now we know we can predict Bush's approval ratings among likely voters in individual congressional districts with a fair degree of accuracy. Let's see how these approval ratings compare to Republican support in House district polls for Republican incumbents, open seats, and Democratic incumbents. Only non-partisan polls where 90% or more of respondents had a candidate preference were included. Most of these by far were from SurveyUSA or Majority Watch.
Click to enlarge. Top line is Bush's 2004 vote, bottom line is estimated approval rating.
An interesting pattern can be seen (click here for all three together). For Republican incumbents and open seats, Bush's vote is a good estimate for the maximum support the Republican candidate can achieve, and Bush's approval is a good estimate for the minimum. This is not true for Democratic incumbents, where we see an incumbency advantage. We do not see any advantage whatsoever to being a Republican incumbent, in line with Democracy Corps' conclusions. It's a throw the bums out year, but Republican bums only.
Partisan polls also fall within the same range; they may be biased but not outrageously so. Many of the polls where fewer than 90% of respondents express a preference are well below the range, however, as might be expected. All data on one graph can be seen here.
From these charts, we can reasonably conclude that in these districts, almost all people who approve of Bush will vote for Republicans, and almost nobody who did not vote for Bush will vote for Republicans.
Survey USA and Majority Watch allow us to poke through their data, so we can check one of these conclusions. In districts with a Republican incumbent, on average 86% of those who approve of Bush support the Republican (median 87%). This number is 87% for open seats, and 75% for seats with a Democratic incumbent. Again, we see an advantage for Democratic incumbents. Of those who disapprove of Bush, the respective numbers are 78%, 81%, and 85% who support the Democratic candidate. Support of Bush is often a better indicator of voting preference than political affiliation.
So if people change their minds and now disapprove of Bush, will they vote for the Democrat? In many cases, yes. Assume everybody who voted for Bush in 2004 approved of Bush. Since then, about a third have changed their minds. If they still supported their local Republican candidate, all the points on the graphs above would fall along the line labeled "2004 Bush vote." This is clearly not the case. In many districts, large portions of those who changed their minds about Bush are going to vote for Democrats.
This brings up another advantage Democratic candidates have: enthusiasm of their supporters. Majority Watch measures voter enthusiasm as well. In most races polled, those supporting Democrats are more enthusiastic than those supporting Republicans. Note that this is not the same as the enthusiasm of those affiliated with each party! The chart below shows how skewed enthusiasm is toward Democrats.
Click to enlarge.
There is a problem, however, and that is selection bias: the races polled and graphed above were the races thought to be competitive already. Making a race competitive, driving the numbers towards the Bush approval ratings, is up to Democratic campaigns. At one extreme, Republicans without an opponent will get 100% of the vote not matter what. At the other extreme, good old fashioned hard work will bring the Republican down to Bush's level.
And this brings us back to the first graph: the Republican Weather Forecast.
Click to enlarge.
This chart shows just how many races could be competitive under the current conditions, given a well-executed Democratic campaign, or lack of a decent Republican candidate (ID-01, anybody?). Remember, Bush's approval is approximately the minimum support for the Republican candidate.
Total campaign devastation - Bush didn't even get 50% of the vote in these districts.
Hurricane warning - Bush barely topped 50% in 2004.
Hurricane watch - Bush's approval is around 35-45%.
Tropical storm warning - Maximum estimate of Bush's approval is 50%.
Flood warning - Bush's approval is likely below 50%.
Flood watch - Bush's approval may be below 50%.
Heavy rain - These few Republicans are truly safe.
The professional political forecasters have already picked out their slate of vulnerable Republican seats, and it is a very large number but not gigantic. Clearly though, as new information becomes available, they upgrade the competitive seats. And many districts have no polls. So we may have a few pleasant surprises. The chart below shows the progress of the Cook Political Report this election season:
This diary simply shows the potential of seats that currently aren't getting a lot of attention. If there's an electoral hurricane, or tidal wave, or any other massive metaphor, a fair number of seats not on anybody's toss-up list could flip to the Democrats. Bush is hurting both the entire Republican party and the chances of individual Republican candidates.