Party, Gender and How Pollsters Handled Indies
“Party, party, party,” Mark DiCamillo, director of the esteemed Field Poll, replied when we asked him back in June if a candidate’s gender or partisan affiliation is more important in a general election.
“If you had to ask just one question that would predict how someone would vote, you’d want to ask their party,” he said.
Despite all the drama some writers, consultants and party activists stirred up around the notion that Meg Whitman could peel away women voters, a crucial part of the Democratic Party base, the results in the governor’s race prove that when push comes to shove, party trumps gender every time.
We’ll walk through the numbers, but there’s one problem: the data we have on the vote by party from the National Election Pool survey by Edison Research is crap.
The NEP survey asked voters leaving the polls to tell them if they think of themselves as Democrats, Republicans, independents or something else. These are not really independent voters as we know them in California – people who “decline to state” a party when they register to vote.
So we have to use data from the most accurate polls – the Field Poll and the USC/LA Times – to understand how actual independents voted when looking at gender and party effects.
Also, understand that Field and USC/LAT used actual voter registration rolls to identify actual voters. And of course, the NEP survey snagged actual voters leaving the polling place. But PPIC asks respondents how they are registered to vote, which means their sample, by party, is a reflection of what respondents say, which may or may not accurately reflect how they are registered. We can demonstrate the hazard in this by comparing the USC/LAT survey which, in addition to using party registration, also asked voters to identify themselves by party.
First, women: According to the NEP exit poll, men voted for Brown over Whitman 51-45% while women voted for Brown 55-39%. That’s a 6-point margin among men and a 16-point margin among women.
These numbers were reflected pretty well in pre-election polling, all of which showed Brown winning: Field had it 46-42% for Brown among men (4 points) and 51-35 (16 points) among women; USC/LAT had it 48-45% among men (3 points) and 55-34% among women (21 points); PPIC had it 41-40% among men (1 point) and 47-32% among women (15 points).
Then, party: But look how much stronger the party vote was.
In the Field Poll, Democrats voted for Brown 77-7% and Republicans voted for Whitman 68-16%; in the USC/LAT poll Democrats were for Brown 81-10 and Republicans were for Whitman 77-15%; PPIC found the Democrats 76-7% for Brown and the Republicans 73-11% for Whitman. These margins were from 69% to 71% among Democrats for Brown and from 52% to 62% among Republicans for Whitman. Much stronger effects than polling found for gender.
The NEP exit poll – in which the vote by party was intensified because of how independents were identified – found it 91-7% among Democrats for Brown and 84-11% among Republicans for Whitman.
DTS versus “independents”: As we all know, in California you may choose not to affiliate with a political party when you register to vote. These are “Decline to State” voters or DTS voters, who comprise about 20% of all registered voters. They are a crucial swing-vote block in California elections and identifying them and tracking their preferences is crucial to understanding how public opinion is moving.
Calbuzz will have more on this tomorrow when we deconstruct some of the myths that already have been spun about this election, including some knowing misstatements about how independents voted here.
But since Mark DiCamillo of the Field Poll and Mark Baldassare of PPIC are doing their Mark and Mark Show at the Sacramento Press Club today, we thought we’d add a few notes to the discussion.
The Field Poll and USC/LAT — like virtually every political pollster hired by any big campaign or interest group — now uses the voter list to develop a sample of actual registered voters. Most pollsters pre-select the list for past voting behavior, only including in the sample people who have voted before, plus newly registered voters. Calbuzz thinks that’s the best practice to identify likely voters. It’s what all the good private pollsters do and what the poll takers for USC/LAT did. Mark DiCamillo of the Field Poll disagrees. He takes a random sample of the voter list and uses past behavior to help select likely voters after interviews are completed. His system works: the Field Poll consistently ranks as one of the most accurate polling operations in the country.
But Mark Baldassare of PPIC asks a scientific random sample of adults: “Some people are registered to vote and others are not. Are you absolutely certain that you are registered to vote in California?” And, “Are you registered as a Democrat, a Republican, another party, or as an independent?” If the person says “independent” he or she is asked, “Do you think of yourself as closer to the Republican Party or Democratic Party?”
That’s the kind of thing you have to do if you’re using random digit dialing instead of working from a voter list. Unfortunately, respondents aren’t always the best sources for knowing how they are registered to vote, no matter how careful the questions are.
The USC/LATimes poll, for example, asked people whether they were Democrats, Republicans or independents and compared their answers to their actual party registration: 22% of those who said they were “independent” were actually registered as Democrats; 34% of “independents” were actually registered Republicans, and just 38% of the self-identified independents were actually DTS voters.
When USC/LAT reported results, they used party registration, not party ID, to describe how people were voting. Which is a good thing, because in their survey, registered DTS voters favored Brown over Whitman 61-24 55-26% (probably too big a margin but at least in the right direction), but self-identified “independents” were favoring Whitman 46-45%.
The NEP exit poll — because it’s a nationwide survey including states with no voter registration — also asked people how they identify themselves. They asked: “No matter how you voted today, do you usually think of yourself as a Democrat, a Republican, an independent or something else.”
Using that definition, the exit poll showed Whitman beating Brown 47-43% among “independents.” It’s exactly the same wrong result PPIC had, when it showed Whitman ahead of Brown 37-36% among “independents.”
Field and USC.LAT did not suffer the same problem because they were polling actual DTS voters. Field had independents 49-34% for Brown and, as we said before, USC/LAT had them 61-24% for Brown.
We can’t say for certain how DTS or independent voters actually split in the governor’s race because we don’t have exit poll data we can rely on. But if you look at the Field Poll as a standard — since they were almost precisely on the numbers everywhere else — and factor in the USC/LAT findings, it’s likely that independents actually voted for Brown over Whitman by about 15 points.
This is bolstered by the fact that self-identified “moderates,” who comprised 40% of the electorate, voted for Brown over Whitman by 60-35%, according to the NEP exit poll. That compared to liberals who went 86-8% for Brown and conservatives who went 78-17% for Whitman.
One other note on the NEP exit poll: it did a lousy job of creating a sample that reflected how Californians are casting their ballots. Final figures are not available yet, but it’s expected that vote-by-mail (VBM) ballots will account for about 50% of the votes cast. But we understand that NEP/Edison only included 600 mail ballots compared to about 3,300 precinct interviews. The margin of error on those 600 mail ballots is huge compared to the rest of the survey and weighting them up would have required some ugly math — not something a reputable pollster would be proud of. The entire survey was obviously weighted to the final unofficial results — 53-42% for Brown. But whether the individual components of that total are accurate is anyone’s guess.
Tomorrow: Myth busters, including Mike Murphy’s bogus argument
my head hurts…