According to the graphic sidebars in the Detroit Free Press, 273 respondents identified themselves as Democrats -- but only 151 identified themselves as Republicans. The total polled was 675 registered voters who voted in the 2004 presidential race, so 251 must have identified themselves as neither Democrats nor Republicans; we'll call them independents.

Michigan voter registration does not include party affiliation, so these figures represent how people identified themselves to the pollsters. With that in mind, the respondents broke out as follows:

All right; many fewer Republicans than either Democrats or independents. But perhaps this just reflects how Michigan voters actually vote. Does it?

Since there is no state listing of party registrations, our best guide is probably the vote for governor in 2006, which occurred just three months ago, between the Democratic incumbent candidate, Jennifer Granholm, and the Republican challenger, Dick De Vos.

None of the various third-party candidates got a statistically significant percent of the vote; so 2006 was a pretty clean test of Democrats vs. Republicans. Dropping (for the moment) the "independent" respondents and only looking at self-proclaimed Democrats vs. Republicans, we see a huge disparity between the Democratic advantage in the poll and the Democratic advantage in the actual vote.

The next table compares votes by party in the 2006 gubernatorial election to the percent of repondents in this poll calling themselves either a Democrat or a Republican:

Note that while Granholm had only a 14-point lead over De Vos in the November election, self-identified Democrats have a 29-point advantage over self-identified Republicans in this poll.

And one more point: since Granholm was the incumbent, which generally carries a vote advantage when the incumbent is popular (as she is), that means the vote might was probably skewed in favor of the Democrats; more Republicans voted for Granholm than Democrats voted for De Vos. Thus it's likely that a generic ballot ("would you rather see a Democrat or a Republican in the statehouse?") would have found greater support for the Republicans than De Vos got... meaning the gap between the gaps -- the extra advantage to Democrats polled by the Freep in contrast to the way they voted three months ago -- is probably even larger than it appears.

On its face, it appears that this poll wildly oversamples Democrats at the expense of Republicans. But there is one other possibility we must investigate.

Clearly, nearly all "independents" (those identifying as neither Republican nor Democrat) chose to vote either for Republican De Vos or Democrat Granholm. Could GOP-leaning independents account for the huge overbalance in the gap between Republicans and Democrats on the poll? That is, if instead of just dropping the independent respondents out of the equation, we were instead to allocate them between the parties by their responses to poll questions, would that account for the gap between the gaps?

If so, then the 37.2% of independents in the poll would have to have voted 20.5% for De Vos and 16.7% for Granholm, for a split of 55.1% Republican and 44.9% Democratic, in the November election.

But do independents really break in favor of Republicans by 10% in this poll?

Looking at the rerun of the 2004 race, we see that the independents in this poll in fact broke the opposite: 50% to 37% for Kerry. And of the nine head-to-head races in the poll, we only see independents lean GOP by 10 points in only 3 races: Giuliani vs. Clinton, McCain vs. Clinton, and McCain vs. Obama. In three races, the Democrat and the Republican are tied among independents; in one race (McCain vs. Edwards), there is a small advantage, 6%, to McCain; and in Romney vs. Edwards and Romney vs. Obama, there is a huge advantage to the Democrat among independent respondents.

I think it pretty clear that "pushing the independents" cannot account for the difference between the party identification in this poll and the actual gubernatorial vote just three months ago; independents in the poll are as likely to go for Democrats as Republicans.

Thus, I think it pretty clear that this poll did significantly oversample Democrats and "independents" and undersample Republicans... even for Michigan. Its results should be taken with a very big lump of rock salt.