Skeptical About a Democratic Takeover

| | Comments (1)

If things go according to the way the polls are reporting, the Democratic party will take over the U.S. House of Representatives in January. The pundits are all saying this and then laughing at a few key Republican politicians who are saying they're not as sure of such an outcome. I've been reading a number of poll-watching blogs every day, and I'm not convinced such an attitude deserves the derision it's getting. I'm not ready to say this is a sure thing at this point, for four reasons.

1. The first piece of information people usually present is the generic polls. The generic polls strongly favor Democrats, but people don't vote for generic representatives in Congress. They vote for their incumbent or an opponent (except in open races, where they still vote for actual people and not generic party candidates). There's a reason Congressional candidates rarely mention their party (and despite Senator Charles Schumer's claim that it's only Republicans that fail to do this, Democrats fail to do it just as much). The best way to get votes from members of the other party is to distance yourselves from the party and emphasize your individuality. Hardcore partisans will of course not be much affected by this. However, independents who are angry with Republicans and marginal Democrats who like their district's Republican representative in general are another story. They often report that they would vote for a Democrat over a Republican but then manage to vote for their particular Republican candidate anyway. So I don't think generic polls tell us as much as some people are taking them to tell us.

2. Every few days another couple seats that looked as if they would turn red to blue have been switching back to red. The movement is clearly back in the direction of the Republicans, with one election projection site even reporting only a three-seat advantage for the Democrats if the election were held today. Other sites are reporting a larger gap still, but this is the site that most accurately predicted the 2004 election results, and he takes into account factors that other sites don't seem to me to be considering (see item #4 below for one of them). The mainstream media (including Fox News) seem unwilling to mention this about the House races (some of them are saying it about the Senate) in any news reports about polling, and almost all of the pundits they give a place to seem not to be aware of this. The governor races do seem to be going leftward in the polls, with two more shifting to blue in the last week, but he movement I'm seeing in House and Senate polls is not by and large leftward. It's rightward. Only a couple of the poll-watching blogs are not acknowledging this, and those are by very partisan Democrats who I think are just in denial. This does not mean that enough seats will move back to red to put a halt to a Democratic takeover, but it's surprising that I'm not hearing much about this anywhere except on blogs.

3. The rapid move toward thinking the House would become Democratically-controlled was right after the Foley scandal surfaced. Around that time people were predicting that large numbers of Republican seats were probably in danger but weren't on the radar simply because no one was polling them. Then several polling firms released polls in a whole bunch of races that no one had been tracking. Sure enough, some of the races everyone had assumed would re-elect Republican incumbents were polling for the Democratic candidate. A number of these races were in or around smaller cities in New York, which are red-leaning purple districts, but several other states were included.

What struck me after looking more carefully was that many of these polls had been conducted by a couple polling firms no one had ever heard of. Even stranger was that some of the races they were polling were coming up with radically different results than what other firms' polls (ones with experience and good reputations) were releasing around the same time, even as much as 20 percentage points different in one case with two polls released within a day of each other. One of these polls was something like 10 percentage points more in favor of the Democratic candidate than a Democratic poll released a couple days earlier had been, and partisan polls are usually at least 3-5 percentage points more in favor of their candidate than other polls are.

All this didn't leave me thinking these polls were all that reliable. If you don't take these couple firms into account, the swing to the Democrats as shown by other polls looks much more marginal than pundits are reporting when they take these polls into account. The election projection sites are including them in their calculations. If they didn't, the control of the House would be much closer, with some of them predicting a Republican-controlled House and others perhaps not. It would be more like the presidential race of 2004, where different projections were coming up with different results for the really close states.

[Note: Today's polls are another example of this. Quite a few races whose last polling had favored Republican were now polling Democratic. One or two were drastic changes. I was really surprised that so many polls could come out on one day that were showing such a sharp turnaround back to Democrats. This didn't make sense. Usually some event leads to such a major change in polling, and there hasn't been one. But then I Iooked at who had released these polls. A whole mess of them were from one of the same suspect firms. Grain of salt. I'll wait to see what other polls from the same districts have to say before taking these polls too seriously.]

[Update 1 Nov 06 (7:22 am): That didn't take long. A poll released yesterday by one of the polling firms I'm skeptical of has Democrat Dan Maffei up by 9 in my district, but a Democratic poll released yesterday has Republican Jim Walsh ahead by 2. Since the partisan polls are usually at least 3 points off in favor of their candidate, that means Walsh is probably ahead by 5. That's a 14 point difference between the two polls released the very same day. The Democratic poll has a smaller sample, but this is a repeated occurrence in the New York polls. Majority Watch releases a poll putting a Democratic candidate way ahead, and then some Democratic pollster releases one within a day that shows the Republican candidate ahead. Something is fishy with these Majority Watch polls, and several of the poll-watching bloggers seem aware of this (e.g. here ) but are not stopping it from affecting their calculations at this point.]

[Update 2 Nov 06: Here's some more discussion of Majority Watch's polls as making assumptions that may well be skewed.]

4. When very close elections have one candidate leading in the polls and the other with lots more money, the one with more money who is behind in the polls typically wins. This is especially the case when that same candidate is in home territory, and these close elections are almost all in red or red-leaning purple territory (with a couple in blue-leaning purple territory such as my own district, but these are often with very popular incumbents, as is the case in my district). Now if there really is a Democratic wave effect, that will count against this, but that seems to be not the case given the direction of most of the polling.

Now these four points are at best an argument against an argument. The argument that the polls show a likely Democratic takeover is somewhat undermined by these factors. That doesn't itself show that Republicans are doing fine, just that the key evidence that Democrats will take over is not as good evidence as it might seem. It still may well be that lots of races not being polled are swinging left. It may even be that some of the polls by these newcomer firms are accurate. But I'm going to be a little skeptical until I see some results by other pollsters in a number of these districts. I don't think there's no reason for Republicans to worry, but I do think there are strong reasons for Democrats not to be as sure of their takeover as everyone seems to be assuming. It might be that everyone is right on this, and we'll find out next week if that's so, but if they're right I don't think it's because all the information is pointing to it. It's more that some indications point to it, and they're ignoring other information that might cause hesitation.


Cold-Hearted Truth has an analysis of the polls compared with election results. It turns out the Majority Watch polls were really unreliable. Only 56% of them were giving them the right result. Their generic polls were way off. But enough of them were right about races no one else was polling that those who factored them in got closer results than what happened when you didn't factor them in (since they got right a number of seats that everyone else was just using 2004's results for, and a number of those seats did flip).

As I was expecting, Jim Walsh seems to have retained his seat, but it was a lot closer than I was expecting.

Leave a comment


    The Parablemen are: , , and .



Books I'm Reading

Fiction I've Finished Recently

Non-Fiction I've Finished Recently

Books I've Been Referring To

I've Been Listening To

Games I've Been Playing

Other Stuff


    thinking blogger
    thinking blogger

    Dr. Seuss Pro

    Search or read the Bible

    Example: John 1 or love one another (ESV)

  • Link Policy
Powered by Movable Type 5.04