The purpose of the survey is to determine and illustrate the collective wisdom of the people, amateurs and professionals alike, who are seriously tracking the upcoming election to project what the balance of power in the House and Senate will be.
Although I consider myself a Democrat and a liberal, and fervently hope that the Democrats will take back one or both of the houses of Congress, I've tried to assemble the survey without bias or prejudice, and I include numbers from sites run by people from across the political spectrum.
From each website I've taken the most comprehensive set of numbers offered, if possible without a "toss-up" category or other caveats. Many of them differentiate between "solid" or "strong", "slightly" or "weak", and "leaning" or "barely" states, but I've combined them all together in order to present numbers which are as comparable as possible. All errors in converting projections from their native format into the one that I use are my own.
I encourage everyone to use the links provided in each report and check each site for the specifics of that site's methodology and presentation.
(For Tradesports, I've used the criteria utilized by Wikipedia, that a "toss-up" is when no candidate has 55% of the vote as polled. For those familiar with my Electoral College Survey of 2004, this is what I called the "Geekmedia conversion" as opposed to the "Fitzgerald conversion," where 50% was the dividing line - i.e. no toss-ups.)
My convention is that the Democrats are listed first and Republicans second. Italics indicate that a site has changed its numbers, or that the site is new to the survey.
Sites which haven't updated in a while, or which have shown no signs of activity, will have their numbers temporarily removed from the list until they're freshened.
Because I began this survey so close to the election, one week is the maximum time I'll keep a "stagnant" site in the survey, but this will tighten up as we get even closer. I also reserve the right to temporarily remove a site if the circumstances surrounding the election change drastically and most other sites reflect that change but the subject site does not.
Temporarily removed sites will be restored as soon as their numbers freshen.
One-time election analysis articles from the news media (as opposed to ongoing features) will be included, but only for a week or less, unless replaced by a new article. I will only include such articles if I can translate their analyses into comparable numbers with some degree of confidence.
As always, if anyone has links for any other sites that regularly track the status of the House and Senate elections, please feel free to send them my way and I'll be glad to add them to the list. I'm also more than happy to hear from the proprietors of any of the sites surveyed here, should they have any complaints, comments, or suggestions for improvements.
Averages
"Mean" is what is colloquially called "average." All items are added up and divided by the number of items.
"Median" is the center point, the middle value in a list. There are as many values larger than the median as there are values that are smaller.
"Mode" is the number in a list which appears the most times. Because the survey includes a relatively small number of entries, the mode tends to be the most volatile and least predictive of the averages, but it's useful in determining if there's some degree of agreement among the sites.
A note on my own numbers
I don't at this time have a projection of my own, but I may add one at some point. If I do, I want to make it clear that the numbers I would publish in the survey under my own name (Ed Fitzgerald) would be the result of my own methodology and are not an "official" result of the survey (although they would probably influenced by it). Also, although I am a partisan, my numbers would certainly not be designed to influence the survey's averages. As mentioned, my interest in doing the survey has to do with determining some collective wisdom about the projected results of the election, and deliberately skewing the results would undermine that goal.
Addendum (10/29)
On 29 October I added an "unfutz automated" entry to the survey. This line uses average numbers (means and median only), calculates the ratio of Democratic to Republic seats, and then applies this ratio to the unassigned seats, which are then added to the previous totals. (The means and median results are averaged and rounded.)
A note on the Independents (10/25)
Many sites project Bernie Sanders in Vermont and Joe Lieberman in Connecticut, who are both running outside the two major parties, to win their races, and some of them list these two projected Independent winners separately from the Democrats. For the purposes of this survey, however, I've included them in the numbers for the Democrats, since I expect that both men will join their caucus. Sanders did so as an Independent member of Congress in the House, and Lieberman has said that he will caucus with his old party if he wins.
I'm fully aware of the concern of many people that Lieberman will not follow through with his promise, and I don't discount that possibility entirely, but I don't think it's likely. Although it's certainly possible that Lieberman could be angry enough at the Democrats to throw in with the other side, I think it's more likely that he'll use any leverage he has to get everything he wants from his old party (i.e. senority and committee chairmanship), since in most circumstances he'd probably get more from them than from the Republicans.
The one scenario in which the possibility of Lieberman defecting seems strongest to me is the one in which the Democrats hold 51 seats (counting Lieberman and Sanders) and the Republicans 49, in which case it would be worth the while of the Republicans to give Lieberman almost anything he wants to entice him to the othe side of the aisle and regain de facto control of the Senate. However, given the current situation, that outcome (51-49) seems somewhat unlikely to occur.
A note on the professionals (10/26)
One of the problems I had with the Electoral College Survey was that its usefulness for its intended purpose -- predicting the outcome of the 2004 Presidential election -- was undermined somewhat by the tendency of the professional election prognosticators (Cook, Sabato, Rothenberg and so on) to be as safe and conservative in their numbers as possible, while, at the same time, hinting at possibly more radical results in their explanatory material, so that (I assume) they could claim credit for these things if they came about. The static nature of their number projections helped give the survey stability, but when push came to shove they were unwilling to make predictions about what would happen, which meant that even just days before the election there was no clear winner projected. (Of course, the 2004 election was astonishingly close as well, which contributed to the difficulty.)
This hesitancy to make bold predictions is understandable -- after all, these are professionals who make their living as political analysts and who have a great deal to lose if they shoot for the moon and miss. Amateur projectors, however, like myself, have much less riding on the outcome (at least as far as reputation is concerned) and are therefore much more willing to take risks.
To try and correct this problem in the current survey, I've decided to, whenever possible, integrate the professional analysts descriptive material into the numbers that I report for them, I hope in as objective a manner as possible.
For example, if Charlie Cook presents numbers which say that the current state of the election in the House is that Democrats will have 207 seats, the Republcians 201 and there are 27 seats that cannot be called (207-201-27, or a 4 seat pickup for the Democrats), but also says in his written material that he expects a 20-35 seat pickup for the Democrats (or that such a pick-up wouldn't be unwarranted in the circumstances, or whatever language he choses to use), I'd adjust the numbers by adding an additional 16 seats to the Democrats and removing them from the toss-ups, for a result of 223-201-11. By using only the lowest number of the predicted range, I hope to enhance the predictive power of the survey without destabilizing it entirely.
