« The Latest Democracy Corps Strategy Memo | Main | Kerry Holds Lead in Two Oregon Polls »

Kerry Leads By 5% , Tied in Pair of WI Polls

John Kerry leads George Bush 48-43 percent of Wisconsin LV's, with 2 percent for Nader, 5 percent unsure and 2 percent other, according to a Wisconsin Public Radio Poll, conducted by St. Norbert's College Survey Center 10/4-13. Kerry's largest margins over Bush included age 18-24 year-old voters 62-39 percent; Independents 48-31 percent; and women 52-41 percent.

Kerry and Bush are tied at 47 percent of Wisconsin LV's in a head-to-head American Research Group Poll, conducted 10/16-19.

Comments

The trends reported in this poll for education and income are quite extraordinary:

"Kerry's support gradually grows from those who do not have a high school diploma (60-40 Bush) to those with a graduate or professional degree (67-32 Kerry). Kerry leads in income categories under $35,000 a year and in the $51,000-$75,000 a year range; Bush leads narrowly in the $36,000-$50,000 category and has wider leads in the higher categories."

I have seen few polls that gave breakdowns by both income and education, but the trends are generally in the same direction although usually weaker. Since education and income are highly correlated, it is very striking to have such opposite trends and when you look at subgroups with the same level of education, the change in Bush/Kerry vote as a function of income must be staggering. And similarly for change in vote as a function of education among people with the same income.

In this circumstance, weighting your sample to give an accurate demographic breakdown by education, but not by income (the latter is nearly impossible for a myriad of practical reasons) will actually INCREASE the Republican bias of the sample created by the oversampling of high-income voters.This bias affects RV samples just as much as LV samples.

I have not studied past poll data on this subject, but my impression was that in past years the overall trend was that Republican vote went up with education level, so that weighting for education allowed you to partly correct for undersampling of low-income voters. But if my recollection is right, this year is different -- meaning that the polls are underpredicting Kerry's vote. (Don't get too excited -- my guess is that the effect is probably not more than one point or so -- but in this election one point is a big deal.)

One conclusion is clear. Polls absolutely should not weight their samples for education this year.

Pew has a new poll out:

"The latest Pew Research Center survey of 1,307 registered voters, conducted Oct. 15-19, finds President George W. Bush and Sen. John Kerry tied at 45%-45% among registered voters, and 47%-47% among likely voters."

And here's the good news:

"There has been little movement in how voters assess the candidates on the issues. But a separate Pew Research Center poll of 803 adults shows that Bush's own approval measures have weakened appreciably. Bush's overall job approval stands at 44%, while solid majorities disapprove of his handling of the situation in Iraq (56% disapprove) and the economy (55%). Even on terrorism, the president's strongest issue, his approval rating stands at 49% the lowest level since the Sept. 11 attacks. Despite this erosion, however, most voters continue to believe that Bush, not Kerry, would do a better job of defending the country from future terrorist attacks (by 53%-35%). Bush also holds a 47%-41% advantage over Kerry as the candidate best able to handle Iraq."

http://people-press.org/reports/display.php3?ReportID=229

A small complaint/request for clarification.

You seem to oscillate between citing LV results and RV results. You have a lot of excellent analysis, but the practice gives the impression of only citing the most favorable side of a poll for Kerry.

Is it merely a matter of polling firms just looking at LVs, or RVs, but not both? If not, perhaps you could cite both results, or, if it's not too simplistic, say why you favor one over the other?

i read in salon (i beleive) that bush has not visitied ohio at all or very much in last three weeks.

can anyone confirm that?

Jason - EMD posts RV's when they are available, LV's when RV's are not available. RV's are better because they are more accurate for predicting outcomes. LV's do gain some value for predicting outcomes as the election gets very close, which will be soon.

Most of the political science research on voting indicates that the likelihood of voting rises with education. Voters with a high school diploma or less tend to break Republican; voters with a college degree or some college education tend to break Democratic; voters with post-graduate education break evenly (those in social sciences breaking heavily Democratic, those in physical sciences breaking heavily Republican).

That has been the general trend since "The American Voter" came out decades ago, although the trend does appear to be weakening some among voters with post-graduate education and college education.