How many votes ?

by John Q on November 9, 2006

A couple of questions, one substantive and one rhetorical

1. What share of the aggregate popular vote did the two major parties receive in the US House elections ?

2. Why isn’t this reported anywhere (at least anywhere I can see) ?

As regards 2, I know that the aggregate popular vote doesn’t determine anything, but that’s true in all constituency systems and for indirect elections like the US Presidential elections, and the popular vote is generally reported in these cases. Also, I know there were some uncontested seats, but there are usually ways to adjust for this kind of problem.

Update Andrew Gelman writes:

Regarding your blog question on votes, you might be interested in our post-election summary here:

The short story is that the Democrats did much better in 2006 (56% of the average district vote) than the Republicans did in 1994 (when they only received 51.3%). In terms of national voting, the Democrats received much more of a mandate in 2006 than the Republicans did twelve years earlier. Our graph is helpful too, I think, both in showing this pattern and putting it into a longer historical context.

I’ve seen a range of estimates of the Democrats’ share of the two-party vote, from 53 to 57, but I’ve generally been impressed with Gelman and his cobloggers, so I’ll take this as the best estimate.

I still wonder that US national media don’t care about this. Even the exit polls reported by the NYT, which had all sorts of breakdowns, didn’t make it easy to get the aggregate result.

Further update Andrew Gelman has written again to advise that a more detailed recalculation produces an estimate of 54.8 per cent.



abb1 11.09.06 at 3:32 am

According to Brad DeLong it’s 32,100,000 Ds vs. 24,524,000 Rs

21,428,784 – 18,665,605 in 2002 and
37,645,909 – 38,164,089 in 2004

Don’t know if it’s correct, don’t know what the source is.


bad Jim 11.09.06 at 4:06 am

Link to DeLong’s post.

According to someone at Pew who was on with Jim Lehrer tonight, turnout was around 40% and a bare majority of both men and women voted for Democrats. (We are, if nothing else, the party of sexual comity.)

The ‘why’ question is easily answered. Each county counts its own votes by its own method. Not only is the vote not aggregated nationally, it isn’t even aggregated at the state level. The truism that “all politics is local” isn’t actually profound, it describes life on the ground in our screwy federal system.

I was peripherally involved in a city council election (she lost) and the results were slow in coming because the voting machines, or their memory cards, had to be trucked up to the county seat and then had to wait their turn. Semifinal results were available around 2am. As bad as that sounds, it’s actually better than the systems we used to have.


bob mcmanus 11.09.06 at 4:17 am

DeLong’s numbers at the link are for the Senate. Quiggin was asking about the House.


Harald Korneliussen 11.09.06 at 6:03 am

House numbers would say a lot about the effects of gerrymandering, wouldn’t they?


Ginger Yellow 11.09.06 at 6:17 am

The Grauniad has a detailed breakdown of (I think) the combined races, but it’s based on exit polls.According to them, 56% of women voted for Democrats and 51% of men. 51% of white people voted Republican, while 89% of black people and 69% of Latino people voted Democrat. 54% of Christians and 70% of evangelicals voted Republican.

Actually, the NYT has an even more detailed breakdown here.


ed 11.09.06 at 8:43 am

The breakdown of men and women imply that 53% or 54% of the electorate voted Democratic to the House (just average the two figures). This is consistent with the pre-election “generic” polls showing the Democrats at 53%.

I’m trying to find out the aggregate popular vote for past House elections, and though I know the numbers exist, its proving difficult. I think this is the best the Democrats have done in the aggregate House popular vote since at least 1974. No Democratic presidential candidate, of couse, has obtained 53% or better since 1964.

As for gerrymandering, the number of seats match the popular vote shares quite well, as they have in the past. Gerrymandering probably hands the Republicans half a dozen seats, net, but there are a number of Democratic Congressmen representing immigrant heavy districts, that always report very low vote totals. These factors roughly cancel in terms of affecting the vote to seat ratio.


SeanD 11.09.06 at 9:33 am


SeanD 11.09.06 at 9:35 am

Here’s a (partial) answer to just this question- looks like about 53-47 in house races:


arthur 11.09.06 at 9:46 am

House numbers are distorted by the significant number of uncontested seats (about 30 of 435). The nationanl Republican Party this round decided not to put up House candidates who had no chance of winning, mainly in majority Black districts. The theory is that some Black Democrats supoprt their local representative but won’t bother to show up to vote if the House seat is uncontested, resulting in fewer Democratic votes for governor, senator, etc.

The Democrats have done the same thing in the past, but this round the Democrats contested every single House seat, although many of the candidates were not serious.


Ben M 11.09.06 at 9:48 am

There’s an answer over at Scienceblogs:

Basically, the Democrats received 37,662,923 votes, and the Republicans received 33,668,227 votes. The Democrats received 52.8% of votes, whereas the Republicans received 47.2% of votes (keep in mind the caveats above). If you calculate things the way DeLong did, we can call that a 5.6% majority in the House.


Cryptic Ned 11.09.06 at 11:00 am

There were a lot more uncontested Dem candidates than uncontested Repub candidates. Sure, a lot of Repub candidates had no chance of losing, but their opponents still got 30% of the vote just by being Dems. Meanwhile a lot of Dem candidates had no Republican at all anywhere on the ballot.

Also, I think the international press has not taken enough note of this story. He isn’t in a majority Muslim or majority black district, either.


John Emerson 11.09.06 at 11:02 am

Bad Jim: I think that a lot of the changes in election procedures were primarily for the purpose of making TV coverage easier. (The same way a lot of timeouts in US football are for TV too). The one big benefit of voting machines I can see is quick results, making it possible for TV to schedule a 6-hour election extravaganza rather than having the news dribble in for 48 hours.


Martin James 11.09.06 at 11:29 am

One answer to 2. is that the total votes aren’t reported because we wouldn’t want our republic to be mistakenly viewed as a democracy.


Jack 11.09.06 at 12:40 pm

The Office of the Clerk of the House eventually puts out the official tallies, available in pdf files:


theCoach 11.09.06 at 12:45 pm

Just so you are aware, Tyler Cowen is responding to your comments on his post here a while back on his site today.


rekniht 11.09.06 at 1:33 pm

The other thing to consider besides uncontested seats, is that the parties in America are rather geographically diverse. While I think it is safe to say that this election was a nationally referendum on the Bush administration and the war in Iraq, few elections are like this. A democrat from Texas is different than a democrat from New York. As an extreme example, the dixiecrats were a completely seperate part of the democrat party. No statement about the national number of votes for the democrats could say anything useful about how the country felt about issues.

Unless, like this time, there is a single national issue central to the election, I don’t think those numbers will ever be useful in a system where the parties are coalitions of groups rather than the government being a coaltion of parties.


Rich B. 11.09.06 at 2:04 pm

While I think the “effects of gerrymandering” are real, there are actually two different “effects” at work, which tend to cancel each other out.

1. Gerrymandering puts lots of members of the same party together, to ensure a victory. (If Dems won with 52% of the vote, put more Dems in, and they are safer next time with 60% of the votes).

2. Gerrymanding shifts members of the same party into other districts, to maximize districts won. (If Repubs won 2 districts with 65% of the vote, and lost 1 with 45% of the vote, they could move 5% Repubs into the losing district, to give them 3 55% victories.)

While people tend to fear #2 more, it is really #1 that happens more often. Politicians want their party in control, but not at the expense of putting their own seat at risk.

Pennsylvania and New Jersey are good examples of opposites.

New Jersey was gerrymandered for “Safe Seats”, and even in 2006, there was only one race that where the margin of victory (for either party) was less than 10 points. Democrats have 7 (of 13) seats, and will never get more than 8 (but also never have fewer than 6).

Pennsylvania, on the other hand, re-districted after the 2000 census specifically to maximize Republican districts. In 2002 and 2004 it worked well, with more Republicans winning, despite a majority Democrat electorate (but each REpublican won by smaller margins in each district). This year, though, the Democrats picked up five PA seats (to take a 12-7 lead in the state), with a bunch of really close races.

So, you can have safe seats, or you can have more seats, but it’s always a trade off. So, over the long run, it works itself out.


DC 11.09.06 at 5:45 pm

And what about votes for parties other that the Democrats and Republicans?


Alan 11.10.06 at 2:54 am

Fruits and Votes has the beginning of a discussion on the aggregate votes, although the blogger, Matthew Shugart, notes he’s the only US political scientist working with the figures.


Abi 11.10.06 at 4:13 am

Paul Krugman’s first post-election op-ed [Mark Thoma says it’s free access for the next few days] has all the information in one place:

“Tuesday’s election was a truly stunning victory for the Democrats. Candidates planning to caucus with the Democrats took 24 of the 33 Senate seats at stake this year, winning seven million more votes than Republicans. In House races, Democrats received about 53 percent of the two-party vote, giving them a margin more than twice as large as the 2.5-percentage-point lead that Mr. Bush claimed as a “mandate” two years ago — and the margin would have been even bigger if many Democrats hadn’t been running unopposed.”


ed 11.10.06 at 4:56 am

Steve Sailer also tried to answer the same question. Just counting the votes, as reported on the Fox News website, resulted in just over 51% voting Democratic. Since according to exit polls, men voted 51% Democratic, and women voted 56% Democratic, if the overall electorate voted at the 51% figure something is wrong here. However, as Sailer noted, votes for the 30 unopposed (by Republicans) Democratic congressmen were not being reported.

The question is how to factor in those thirty, because they got at least some votes. Sailer calculated 52.8%, which jibes with the figure quoted above, and also with the exit poll data. There were in fact 4 uncontested Republican seats, but obviously they don’t skew things as much as the 30.

The 53% margin is pretty decent. The last president to win by that much was GHW Bush in 1988. I think you have to go back to the 70s to find a party winning by that much in the House. Very few incumbent Democratic congressmen won with less than 60%, while quite a few incumbent Republicans dropped below that figure. Yesterday’s New York Times also has a good breakdown of the exit poll data.

And while the media attention has been focused on the first Muslim represented to Congress, don’t forget that two Buddhists were elected as well, and it also seems that there are more Jewish congressmen than in the past.


abb1 11.10.06 at 9:15 am

Another phenomenon the US media don’t address is a huge anti-Republican swing in the Latino vote in this election. If the Democrats manage to position themselves in opposition to anti-immigration hysteria (it’s not clear to me that they can), then they should be able to ride this wave for decades. It’s the largest and fastest growing minority group.


john henry 11.10.06 at 8:07 pm

The numbers seem low. I aggregate around 56 to 61 million votes cast, depending on which of the above sources – let’s say 61. If we assume turnout of around 40%, it leads to a total electorate of about 150 million. This suggests that just 50% of the entire population is even eligible to vote. And this is cast as the world’s premier democracy! Are there really that many immigrants, children and prisoners? Or am I missing something?

Comments on this entry are closed.