In this post, I promised I would make my own election predictions — and I will, with one exception.  I am putting my predictions in a separate post because I am unable to make one prediction, and I thought I should explain why.

First, the predictions:  Tomorrow, the Republicans will come close to a sweep.   McDonnell will win in Virginia by 13 points, and the Republicans will make gains in the Virginia legislature.

Christie will win in New Jersey by 4 points.  (It won't be close enough so that the usual, or even more than usual, election fraud can affect the results.)

Mayor Bloomberg, Republican turned independent, will defeat the Democrat in New York city by 10 points.

Conservative/Republican Doug Hoffman will win in New York's 23rd by 7 points.

Democrat John Garamendi will win in California's 10th by 6 points.

Before I come to the prediction where I punt, let mention some well-known facts about election polls:  The bigger the random sample the better.  Polls are more accurate in partisan races.  Off-year races are harder to poll, because turnout can be so variable — and can vary in one-sided ways.  Interviewers can affect the results; even without meaning to be partisan, an interviewer can subtly influence the answers.  (George Gallup, who founded the polling organization that still has his name, used to joke that he had two interviewers who had that effect.  One, a Democrat, consistently got more Democratic responses than the average interviewer.  The other, a Republican, consistently did the opposite.)  Some universities run professional polls; other universities run polls that embarrass amateurs.

The county executive race pits Susan Hutchison against Dow Constantine.  Although the race is formally non-partisan, many (most?) voters know that Hutchison is a Republican and Constantine a Democrat.  (For those unfamiliar with our politics, King county is heavily Democratic, and has become more so in the last decade.)

There have been two recent polls on the race.  Survey USA, polling on 13 October, gave Hutchison a narrow lead.

In an election for King county Executive today, 10/13/09, the day before ballots are available to voters, Susan Hutchison has a slight advantage over Dow Constantine, according to this latest exclusive KING-TV poll conducted by SurveyUSA.  Hutchison, a former local TV news anchor, has polled at 47% in each of SurveyUSA's three identical tracking polls over the past 5 weeks.  Constantine, who chairs the King county Council, has polled at 44%, 44% and today 42%. The contest is fiercely fought.  Both candidates are well positioned.

The second poll, done by a research institute at the University of Washington, the Washington Institute for the Study of Ethnicity, Race & Sexuality (WISER), with phone interviews from 14 to 26 October, gave Constantine a big lead.

The Washington Poll's figures in the county exec's race run counter to findings by two SurveyUSA polls for KING5 news.

The Washington Poll has Constantine at 47 percent, Hutchison with 34 percent with a big 19 percent undecided.

When faced with two such divergent results, I do two things; I look at the polls' records, especially in recent elections, and I look for internal problems in the results.  As it happens, both organizations polled during the primary here, and both organizations had lousy results.  For instance, neither was close on the results of Seattle's mayoralty primary.

I can find problems with both polls.  Survey USA polled only 523 likely voters, and did so more than two weeks before the election.  (They chose that date because it was just before we received our mail ballots.)  That's a small sample, and the date is early enough so that they might have missed late shifts.  And there have been many attacks on Hutchison during these last two weeks.  Even I have seen some of them, and I don't watch much television.

There are more problems in the Washington Poll.  They polled only 724 registered voters, statewide.  Ordinarily, this would give them too small a sample for King county, which has less than 30 percent of the state's population, but they "oversampled" King county, though they don't say by how much.  If King county was half of their sample, then they would be basing their results on just 362 registered voters.  But in this off-year election, I would not expect turnout to be much higher than 50 percent, so their real sample size would be about 200, or even fewer.

And then there is the possibility of interviewer effects.  Interviewers hired by the Washington Institute for the Study of Ethnicity, Race & Sexuality might not be absolutely unbiased, might hint at the answers they want, when conducting interviews.  And that's an especially big worry in an election where gay rights is on the ballot (Referendum 71).

So, no prediction on the Hutchison-Constantine race, because I just don't have good enough data to work with.  (If you twisted my arm, I would say that I think that SurveyUSA will probably be closer than the Washington Poll, but I'm sure you are too nice to twist my arm, just to make me give an answer based on dubious polls.)

But I will make this very safe prediction:  At least one of these two survey organizations is going to be very embarrassed by the results of this contest.  I hope, for the sake of this county, that it is the Washington Poll.

Cross posted at Jim Miller on Politics.

Update: A late Survey USA poll coming out just after I put up this post had results that were close to the results in the Washington Poll.  Survey USA now gives Constantine a 10 point lead, 53-43.  The poll also found that 43 percent (!) of the voters decided who to vote for in the "past couple of weeks" or even in the "past couple of days".  So, it is possible that all three polls I have discussed are close to the truth, and that there was a large shift toward Constantine in the last two weeks of the campaign.

Given Constantine's reactionary* record, and views, I hope that didn't happen and that the earlier Survey USA polls were more accurate.  But I do want to say, before the polls close, that the two polling organizations may have done better polls than I thought when I wrote this post yesterday afternoon.

(*In general, most American "progressives" are now more accurately described as reactionaries.   Typically, they favor 19th century modes of transportation, and 19th century ideas about treating the races differently.  Most want to introduce social welfare measures that would tie citizens closer to the state, just as Bismark did in 19th century Germany.  Often, they have an exaggerated respect for authority figures, and an indifference to individual economic freedom, thinking that ordinary people are not very good at looking after their own interests.)