My attempt below in “Code blue in Minnesota” to calculate the party identification of the respondents in today’s Star Tribune Minnesota Poll is obviously flawed, even if my math is right (which it may not be). My effort stupidly assumes that the Minnesota Poll assigns an equal weight to all respondents. On the contrary, however, as I note in “The trouble with the Star Tribune Poll,” the poll weighs responses under a secret formula. The poll published today does not disclose its internals. Here’s the Star Tribune’s explanation of how today’s poll was conducted:
The results are based on a Star Tribune Minnesota Poll conducted statewide by telephone Oct. 31 to Nov. 3 with a representative sample of 1,027 randomly selected Minnesota adults. Results were adjusted to account for people’s different probabilities of voting; thus they reflect the views of likely voters. Respondents who are more likely to vote were counted more heavily than those who are less likely to vote. The adjustments were made based on answers to questions about voter registration status, voting history, interest in the election and self-described probability of voting.
For some reason, today’s explanation leaves out the statement that “all adults (except for some special populations such as felons, aliens and certain populations in group quarters), are likely voters.” The Star Tribune Minnesota Poll today stands in the ignoble tradition to which I allude in the NRO column. I stand by my prediction that Governor Pawlenty will do at least five percentage points better on Tuesday than the poll today shows.