Election 2008: The Latest Polling
Monday, January 28, 2008; 12:00 PM
Washington Post polling director Jon Cohen and polling analyst Jennifer Agiesta were online Monday, Jan. 28 at noon ET to discuss the findings of the findings of the latest polling on the 2008 presidential primaries.
The transcript follows.
Jon Cohen: Good afternoon. A lot to discuss. Bush will give his final State of the Union Address this evening, and someone mentioned an election...
Prescott, Ariz.: The polling was even farther off in South Carolina than it was in New Hampshire wasn't it? Funny, while there was all sorts of hand-wringing about New Hampshire, I haven't heard much about how wrong everyone got South Carolina
Jon Cohen: You're right, by some measures, the late South Carolina polls were more "inaccurate" than those out of New Hampshire. One reason that there's less discussion of the errors is that the pre-election polls had Obama on top, and he won. Perhaps later in the chat we can get back to this, but for more, see Pollster.com, where Mark Blumenthal has started a discussion comparing the polls in those two primaries.
Arlington, Va.: Any information on how the candidates are polling in Virginia? Now that the Feb. 12 primary might actually matter, I'm trying to decide which primary to participate in. My current preference is Obama first, and then Clinton and McCain tied for second. If my vote wouldn't really help Obama much but could help McCain get the Republican nomination, I think I'll have to vote Republican this primary.
Jon Cohen: You bring up a really important point: In many states, voters can choose to participate in the Democratic or the Republican primary. That was true for all voters in South Carolina and for those not registered with either major party in New Hampshire.
I have not seen anything new out of Virginia, but I imagine we'll see a slew of polls soon. However, a couple things to be wary of: Feb. 5 may have a big impact, so looking at data now might not be a great indicator; perhaps more importantly, recall how the New Hampshire pre-election polls that were squeezed in after Iowa had Obama up? McCain and Obama also were battling for independent voters in the Granite State.
Arlington, Va.: Not to be cynical, but is there really an audience who cares what this final State of the Union speech will cover? This president's approval ratings are at record lows, and common political opinion is that he just bullies Congress into caving to what he wants. Should we expect anything other than that tonight?
Jon Cohen: The president is at his career low in our polling (32 percent approve), and he's been stuck in the low 30s for a long time (see full trend here). Just 28 percent approve of the way he's handling the economy, which has become the election's top issue.
On that front, the president and the Democratic leaders in the House did agree on the outlines of an economic stimulus package last week, so it'll be interesting to see how the public responds to new hints of bipartisanship on this important issue.
Washington: Conventional wisdom states that Obama must do better among white voters if he is to win the nomination. Could it also be said that Hillary Clinton needs to do better among men to win the nomination?
Jennifer Agiesta: To win the nomination? Probably not. She's done pretty well among men in the states that have voted so far, with support from 23 percent of men in Iowa, 29 percent in New Hampshire, 43 percent of Nevada men, 51 percent in Michigan and 23 percent of South Carolina men.
And in most of the states with upcoming primaries or caucuses, 2004 exit polls show the Democratic primary electorate to be majority female, while the racial breakdowns within each state vary greatly.
It should be noted that with the exception of Nevada, Obama has not done all that poorly among white voters. In Iowa he won them, with 33 percent support to Clinton's 27 percent; in New Hampshire he ran about even with her, 36 percent to 39 percent; and his 24 percent showing in South Carolina placed him not far behind Clinton at 36 percent. And in our latest national poll, Obama gets the support of 33 percent of white Democrats, while Clinton gets 41 percent.
Check out our Behind the Numbers post on the demographics of upcoming states.
Princeton, N.J.: George Zogby was on C-SPAN a week or so ago, and he said that while 20 years ago a 65 percent response rate was considered reasonably good, today he considers himself lucky to get 16 percent. On the surface that seems pretty bad because you are then polling a self-selected group. Do you guys have good research on how the whole population is represented by those who respond? It would seem hard to get, as the people who do not respond do not respond!
Jon Cohen: John Zogby? It's true that response rates have suffered through the years, but they're only one indicator of survey quality, and methodologically-sound telephone polls continue to produce reliable estimates of public opinion. Here's info on The Post's response rates, at the end of which is a link to this classic from ABC's Gary Langer.
For more on the response rate issue, see the Pew Research Center's report "Polls Face Growing Resistance, But Still Representative."
Fairfax, Va.: Although the GOP as a whole has been hurt in recent years because of the rhetoric they used to fight comprehensive immigration reform, has John McCain's standing among Latinos taken a tumble after he authored a the bill but tried to put that issue in the background now that he's running for president?
Jennifer Agiesta: The Latino vote in the Republican primaries is not typically a large enough group to get a read on in typical pre-election polling, but we should get our first chance to look at this key segment tomorrow in the Florida exit polls.
The impact of immigration on John McCain's support is certainly one area we'll be looking at, as it has been a challenge for him all year. In our latest Post-ABC national polling, McCain has leapt to the front of the GOP pack on the issue. In early December, 18 percent named him the candidate best able to handle the issue; earlier this month, 27 percent considered him tops on it.
Savannah, Ga.: Do you think that a lot of these polling errors are because of significant increases in voter turnout? They mostly seem to be on the Democratic side, which is seeing huge increases in participation. In other words, a lot of people who wouldn't ordinarily be classified as "likely primary voters" are voting, and the pollsters have yet to figure out a way to gauge them?
Jon Cohen: Sure could be. In my view the most probable explanation for the now-infamous errors in New Hampshire was one of modeling. AAPOR, the American Association for Public Opinion Research, has set up an ad hoc committee to look into the New Hampshire polls, so we should know more soon(ish).
Los Angeles: Reporters and pundits delight in analyzing the "tight horse race" or changing poll results. However, political polls cannot predict election results reliably because they don't meet statistical inference prerequisites (most notable are: stable sampling populations or a sampling environments from which multiple samples are drawn over time, and respondents selected randomly with each possible opinion in the population equally able to be sampled).
It's improper to describe election projection results as "a statistical dead heat" or "within the margin of error" because poll designs and error-margin calculation methods aren't based on rigorous statistical estimation models but likely just represent collections of disparate views by a few hundred individuals who are available to respond to poll questions at one point in time. To confirm this, simply publish the methods and assumptions pollsters use to calculate those ubiquitous 3 percent to 6 percent error margins.
Jon Cohen: This is question is packed, but I share your low regard for the term "statistical dead heat"; we try to avoid that.
We can calculate a survey's margin of sampling error, and that's what we report. We only report on "differences" that are statistically and substantively different.
Polls: Do you have any current (post-South Carolina) Super Tuesday polls?
Jennifer Agiesta: There have been some post-South Carolina polls released on the GOP side, but because the vote on the Democratic side was a mere two days ago (though it certainly feels like eons!), none of the current Democratic polling will reflect South Carolina's impact.
USA Today-Gallup is out today with new polls in the two biggest Super Tuesday contests: California and New York. On the GOP side, they find John McCain with a double-digit lead over his nearest competitor in New York -- hometown candidate Rudy Giuliani -- and with a narrower edge over Mitt Romney in California.
Philadelphia: Have there been any noticeable shifts among voters by race in the Democratic Primary?
Jon Cohen: Yes. In our most recent Post-ABC national poll, Obama led Clinton by about 2 to 1 among African Americans. Sen. Clinton had held a narrow advantage in our December poll.
The racial dynamic has been among the interesting aspects to the Democratic contest. Obama beat Clinton by about 4 to 1 among black voters in Nevada and in South Carolina, according to National Election Pool exit polls. In Nevada, where Latinos made up 15 percent of the electorate, about two-thirds voted for Clinton.
At the same time, race does not appear to be as clear a dividing line as it was in 1988, when Jesse Jackson ran for presidency. I put up a post on Behind the Numbers on the differences this morning.
Bowie, Md.: What are the chances Kennedy's endorsement actually hurts Obama? What are the chances most of today's younger voters don't even remember the Kennedys?
Jennifer Agiesta: Nationally pollsters haven't asked about Ted Kennedy in a few years, but a 2005 CNN/USA Today/Gallup poll found the senator from Massachusetts had a net-positive favorable rating: 48 percent said they held a favorable view, 38 percent unfavorable.
Certainly a good chunk of primary voters may not remember (or even have been alive for) the Camelot era, but Ted Kennedy has remained prominent in national Democratic politics, and in that CNN/USA Today/Gallup survey, just 6 percent said they never had heard of him.
Arlington, Va.: So much for polls trying to predict primary results. They totally missed the New Hampshire primary on the Democratic side. In a way, the South Carolina polls were almost as bad. They were predicting about an 8 percent to 10 percent win for Obama and he won by 28 percent -- three to four times the margin predicted. Will pundits put less emphasis on the primary polls from here on out?
Jon Cohen: To the first and last parts of your question, polls are most appropriately used to help us understand why something is happening at a given point in time, not to predict the future. They're not crystal balls, no matter how they're spun.
Part of the problem is an obsession with horse-race coverage; good polls offer much more than that. See this Outlook piece on finding value in pre-election polls (it is there, sometimes).
Fairfax, Va.: Statistics was a long time ago, so when there is a margin of error of like 3 percent and Candidate A has 40 percent of the vote and Candidate B has 35 percent, is that outside the margin of error? Or does the 3 percent refer to each individual statistic, so in my example the range for Candidate A would be 37 percent to 43 percent, while Candidate B would have a range of 32 percent to 38 percent?
Jennifer Agiesta: The short answer: it depends. Margin of sampling error works more like your second scenario, with the plus or minus figure representing a range around each candidate's proportion of the vote.
But that range means that we can say with 95 percent confidence that the real result is somewhere within that range. And of course, there are always other potential sources of error in polling, such as question wording, interviewer effects, my-dinner-is-burning-I-have-to-hang-up-now effects...
There's a more in-depth discussion over at pollster.com.
Pop quiz to follow later.
St. Louis: There was a significant amount of speculation this past weekend on cable television regarding the degree of antagonism between Latinos and African Americans and the political repercussions. Do you know of any polling -- perhaps out of California -- that suggests either group tends to vote against the political figures of the other?
Jon Cohen: None of the recently-released public polls out of California have big enough samples of African American voters to break out those numbers, but new polls from PPIC, Field and Gallup all have Clinton up over Obama by large margins among Latinos.
In winning his Illinois U.S. senate seat in 2004, Obama was backed by 82 percent of Latino voters; 17 percent supported Alan Keyes. That year, 76 percent of Latinos voted for Kerry, 23 percent for President Bush.
Richmond, Va.: Why is immigration an issue that is asked about in all of the GOP polls, but not on the Democratic side? It's not like the Democratic candidates don't have immigration policies.
Jon Cohen: I take it you're referring to the exit polls? Those questionnaires are worked out by the National Election Pool. While we have no idea what went into their decision-making, when we ask open-ended about the election's top issue, Republicans more frequently highlight immigration than do Democrats.
Rochester, N.Y.: This isn't about latest polling per s,e but as one stats geek to another (I'm a statistician) I hope you can answer: Based on what we've seen in SC, how high can Obama push African American turnout in the general? I realize it's a bit of a myth that African Americans vote at a low rate (they actually voted at a 60 percent rate in 2004, just a little off the white rate) but might an Obama candidacy in the general push African American voting rates through the roof to 70 percent or 80 percent?
Jennifer Agiesta: Obama made this very argument himself back in August (and in our latest Post-ABC national poll 43 percent of African Americans said that Barack Obama's opportunity to become the first African American candidate makes them more enthusiastic about his candidacy), saying that if he became the Democratic nominee, it would give the party a chance in states like Mississippi because of the boost in black turnout it would provide.
We took a look at the impact such a huge jump in turnout would have on a general election competition in Mississippi, and found that even with such a huge increase, if everything else remained the same as '04, the Republican candidate still would win. Check out the analysis on The Trail.
Jon Cohen: Sorry we couldn't get to all your great questions and comments; "see" you next time.
Editor's Note: washingtonpost.com moderators retain editorial control over Discussions and choose the most relevant questions for guests and hosts; guests and hosts can decline to answer questions. washingtonpost.com is not responsible for any content posted by third parties.