Charles Franklin, Mark Blumenthal and Nate Silver
Co-Developers, Pollster.com; Founder, FiveThirtyEight.com
Wednesday, October 1, 2008 12:00 PM
Poll-crunchers Charles Franklin and Mark Blumenthal of Pollster.com and Nate Silver of FiveThirtyEight.com were online Wednesday, Oct. 1 at noon ET to discuss national and state-by-state polling in the 2008 presidential and Senate races, and how they aggregate and weight those numbers to try to determine public opinion.
The transcript follows.
Franklin is a professor of political science at the University of Wisconsin, and the creator of the Political Arithmetik blog.
Silver is a writer, analyst and partner at Baseball Prospectus and creator of the PECOTA algorithm for projecting future team and player performance.
Nate Silver: This is Nate from FiveThirtyEight.com. I'm really excited to have the Pollster.com guys with me, and happy that you could join us all today.
Charles Franklin: And I'm Charles Franklin from Pollster.com. Very glad to be here with Nate.
Oakland, Calif.: Mr. Silver, what is the difference between the trend and projection lines on your website, and why do they sometimes disagree?
Nate Silver: Hi, Oakland. The difference is that the trend line lists what we think would happen if the election were held today. The projection line is what we think will happen on Nov. 4. The difference is that most elections tighten some down the stretch, and so that assumption is built in to the model. We think that Obama would win by five or six points if you had the election tomorrow, but the best estimate is more like four points once all is said and done.
Fort Myers, Fla.: As of last night, John McCain was five to six points down in most tracking polls. Has any candidate overcome a deficit of that size just a month before the election to win? If so, what were the circumstances?
Charles Franklin: In 2000 Bush changed a two-point Gore advantage into a four-point Bush lead in three weeks, but that came in August/September. The past two elections have been relatively stable in the last three weeks of the campaign.
Ann Arbor, Mich.: Nate, most of the latest polls in Michigan, including Ann Selzer's, have shown Obama up by five-point-or-more margins. You had discussed some of Obama's potential problems in a New Republic piece and on your site. I wonder, what's your sense of why Obama appears to, if anything, be overperforming in Michigan, rather than suffering from some of the racial/Kilpatrick scandal/etc. problems? Is it the economy and that's it?
Nate Silver: It may be that Kilpatrick left the scene just in the nick of time to avoid causing real problems for Obama. But, yeah, the economy would have to be the most significant factor, as Michigan's unemployment rate has been 1970s-like for quite a while. I don't expect the Obama campaign to start counting its chickens in Michigan any time soon, though. They are aware of some of the difficulties that I outlined in the New Republic piece.
Chicago: How does this campaign at this stage compare to Kerry-Bush campaign at a similar stage in 2004? Nate, have you tried to apply your models to the existing data from that campaign to see how well they perform?
Charles Franklin: I did a comparison a while back, but it needs updating to be current. Right now Obama is about three points ahead of Gore at the same time and nine points ahead of Kerry.
Princeton, N.J.: The election is clearly turning in Obama's favor, with states like North Carolina, Indiana and Missouri coming into play. Is this a vindication of the Obama campaign's 50-state strategy? Or do you think that if Obama wins North Carolina or Indiana he has already won the election, and still essentially is wasting resources in these states?
Charles Franklin: If Obama wins those states, he almost certainly has done well elsewhere as well and should win. But the race is pretty tight, and North Carolina and Indiana are tough states, even though the North Carolina trend is consistently in Obama's direction.
Nate Silver: North Carolina has been moving so sharply to Obama that I don't see how they possibly could pull out now. That state has a huge retail-banking sector, and has been hit particularly hard by the economy. But also, we may be seeing the dividends of their having been on the ground in that state essentially since May, whereas the McCain campaign just pulled in last week.
Indiana is an odd place to call a battleground state, but the Obama folks feel like they're on home turf. They overperformed there during the primaries, and so they're likely to stay for the time being.
Riverdale Park, Md.: Which pollsters are being the most creative at tackling the problem of the under-35s being "cell-phone only," and thus hard to reach?
Charles Franklin: A number of pollsters have been including cell-only folks for most of the year, including Gallup, Pew, CBS/New York Times, NBC/Wall Street Journal and I think Time. Field in California does as well. It is going to be increasingly common.
Toluca Lake, Calif.: Of the traditional three mega-battlegrounds -- Florida, Ohio and Pennsylvania -- Barack Obama only has polled well consistently in Pennsylvania, yet now it appears he is making headway in the other two as well. Which do you think presents the more attractive target for Obama and the Democrats -- heavily unionized Ohio, or foreclosure-battered Florida?
Nate Silver: At a gut level, I tend to like Florida a little better for Obama, as it's somewhat more virgin territory, whereas in Ohio they sort of swung and missed during the primaries. But the early voting window in Ohio could be a big equalizer.
Cambridge, Mass.: What do you think about weighting by party identification?
Charles Franklin: Weighting by party identification depends on "knowing" the PID distribution. but it has been shifting since 2005. One pollster, GW-Battleground tracking poll by Tarrance and Lake just shifted away from using PID weights.
Nate Silver: Weighting by party ID can be a little bit like putting lipstick on a pick. It pretties a poll up and certainly reduces the fluctuations from survey to survey. My guess is that it will make a poll more accurate on average, but it sometimes can be a cover for poor methodology or small sample sizes. If you're going to weight by party ID, I think Rasmussen (which weights based on the thousands of interviews they do each week) takes a pretty good approach. Just "guessing" at what the party ID composition should be isn't sound, in my opinion.
Wooster, Ohio: I've heard that some inside the Obama campaign are, on the basis of their internal poll figures, predicting an electoral-college landslide for the Democratic candidate. Do campaign "internals" differ significantly from other polls? If so, how and why?
Mark Blumenthal: "Internals" in this context can mean a lot of things -- both the current "horse-race" numbers as measured by the campaign, as well as projections based on additional information. Campaigns do their own polling, and their measurements of vote preference should be similar to what we're seeing in the public polls.
Washington: We've been hearing a lot about the presidential polls, but not so much about what polling is showing in the fight for the Senate. Any projections on what the makeup will be after Election Day?
Charles Franklin: At the moment at Pollster we have Oregon and Alaska as tossups with North Carolina leaning Democratic (Dole is the Republican incumbent there, so that would be a pickup). New Hampshire also leans Democratic, as does Colorado. Kentucky still leans Republican. Bottom line is 42 Republicans, 54 Democrats and two independents, and two tossups.
Nate Silver: That sounds about right to me, but a couple of races that looked out of reach may be coming back into play. There was polling out in Georgia yesterday that showed that race tightening pretty dramatically, and likewise in Kentucky. My guess is that there will be a fairly large anti-incumbent surge, which could be good for the Democrats in the Senate, but might have ambiguous effects in the House.
St. Louis: There's been a lot of recent talk relating to the cell phone problem in polling, but I wonder if things have changed all that much since 2004. Why do analysts believe that cell-phone polling would more dramatically affect the numbers in this race when polling from 2004 did not show a Republican bias?
Charles Franklin: I think the problem is significantly exaggerated. There has been a lot of research on this, and the bias of leaving out cells has been small. After all, they may be mostly young, but not many people buy a cell-only phone plan based on the vote choice -- hence not much political bias. And the young are weighted up in the end anyway. But a new Pew research report says that bias may be a bit more now than last year, so expect much more cell inclusion coming up.
Chicago: How can polls be weighted for turn-out? Instinctively it feels like Obama will have massive youth turnout, which doesn't seem to be reflected in the polls sampling/weighting. How do polls get from "registered voters" to "likely to vote"?
Charles Franklin: Turnout is a big question mark this year. Most pollsters allow respondents to say how certain they are to vote, so that would be picked up this year, but turnout among young and inexperienced voters is notoriously hard to increase as much as campaigns want.
Nate Silver: Right. Most pollsters are going to be screening a certain percentage of young votes out with their likely voter models, which is probably the "right" thing to do. But I'd guess that some of them also are overcompensating. Youth voter participation increased by something like 50 percent as a share of the Democratic primary electorate, so I certainly do not think that pollsters should be taking the 2004 turnout ratios as gospel.
Shorewood, Wis.: You both do amazing work. Could you give us a sneak preview of kind of capabilities an innovations that your foresee for FiveThirtyEight and Pollster.com over the next few election cycles? I noted with great interest the ability to customize graphs on Pollster. What else in in store?
Nate Silver: I'm looking forward to covering House races in 2010, where I think our data-mining capabilities really should help us, given that there are so many races to cover at once ... we haven't had time to do that this year. But also, we hope to get into a lot of metrics regarding the Congress -- improving, for instance, on the traditional liberal/conservative Senator rankings.
Charles Franklin: I'm just hoping to survive 2008!
I think we can make some real progress on congressional and other state-level races. Also, with a little breathing room we can add new features, like our new user-customizable charts that we put out last week. But there is always lots more we'd like to add.
Mark Blumenthal: We've been so focused on 2008 (with a few items still left on the to-do list) that we haven't given ourselves permission to think much past November. But one thing I'd definitely like to do more of is allow users to save their customized charts, and even better, create their own overall projections based on their tinkering with individual races.
New York: The polls for Indiana are fascinating, but is the state more Missouri or Ohio in terms of its likelihood of ultimately breaking for Obama?
Charles Franklin: More Missouri.
Nate Silver: I'm a bit more optimistic about Indiana from Obama's point of view, simply because it's hard to quantify the effect of Obama having some 30 field offices in the state whereas McCain has none.
Marion, Ind.: Is there any history of the vice presidential debates moving the polls?
Charles Franklin: No -- though post-debate polls sometimes have found clear winners and losers in the vice presidential debates, there isn't much evidence that the national polls at least move much.
Montreal: Could you discuss a little what you feel the effect of having Palin on the Republican ticket has been? Has there been a bounce, or steady decline, or just a flat shift in support as a result?
Nate Silver: The one area where Palin seems to have helped McCain is in the *rural* west, by which I mean states like Montana, North Dakota and (of course) Alaska, all states where Obama had hoped to compete before. States like Colorado and Nevada, by the way, should *not* be described as rural, since most of their population lives in cities.
But I don't think she's been a help to McCain most anywhere else on the map. For a little bit, it looked like she was helping him win over some pro-life moderate Catholics in Pennsylvania and perhaps Wisconsin, but the Pennsylvania polls are now swinging back to Obama pretty rapidly.
Charles Franklin: She seems to have moved white women in the Republican direction pretty strongly after the Republican National Convention, but that now has come back by half, to about a 6-8 point McCain lead. Interestingly white men didn't react to Palin but have now moved towards Obama since the financial crisis arose.
Alexandria, Va. -- for Nate: I am a complete FiveThirtyEight.com junkie. I cannot log off. But I don't understand how you calculate the state "win" statistic. Is that an internal likelihood-of-your-likelihood figure?
Nate Silver: It's simply the percentage chance that we think a candidate has of winning a state on Election Day, as determined by the 10,000 simulations that we run each day.
Ashland, Ore.: Do you see any evidence that voters actually have made up their mind, and that they aren't listening to what the candidates say or even watching the news? Are there fewer undecided voters now, and will they stay that way till Election Day?
Charles Franklin: Undecided is running strikingly similar to 2000. 2004 was less undecided but with an incumbent running you'd expect that.
Nate Silver: Indeed, 2004 had an exceptionally low number of undecideds, as Bush so significantly had polarized the electorate before. The number of undecideds in this election had been fairly high before, but went down by a few points after the conventions, and now seems to be fairly normal.
Fairfield, Iowa: To what extent are polls bilingual (or multilingual)?
Charles Franklin: A few offer Spanish as an option, but that is more common with expensive "academic" or government research, and not so common among media polling. However, in states with large Spanish speaking populations they usually do include both.
Richmond, Va.: It is quite extraordinary how much we depend on the polls -- even in the face of so many of them being wrong so much of the times. Here is a perfect example: According to Real Clear Politics, the latest ABC/Washington Post poll shows McCain picking up 5 points after the debate, but only that same poll had him very far down pre-debate. On the other hand, every other poll out since the debate (including yesterday), had Obama picking up a substantial lead post-debate What's an ordinary person (who looks to the polls every day) to do?
Charles Franklin: Read fivethirtyeight.com and pollster.com! <;-)
Seriously, I think that's what you are getting by seeing the polls all in context of one another rather than one at a time. The relationship of one poll to all the others is crucial and some media outlets are getting better about showing that.
Nate Silver: In this particular instance -- and this is no slight against The Washington Post's polling, which we have rated as the best of the major media pollsters -- you're mostly seeing regression to the mean after what had looked like an outlier before. If there starts to be a momentum shift back toward John McCain, you will see it in plenty of polls at once -- it won't be hard to find.
Dumb Question: What does 538 signify?
Nate Silver: It's the number of electors in the Electoral College.
Philadelphia: Can you comment on how much of the recent polling movement (favoring Obama) is about undecideds committing to one candidate over another, versus a relative softening of McCain's support? Nate, you already posted about Obama shoring up previously soft support (Clinton Democrats, possibly). I'm really more interesting in getting a sense of whether McCain is losing undecideds, or if his support is softening.
Charles Franklin: Undecideds slowly have been going down, but they aren't moving enough to account for the pro-McCain swing after the Republican National Convention or the pro-Obama swing we currently are seeing.
New York: Normally when closing into an election, party loyalty numbers increase. Do you think this is a matter of people changing who they are supporting, or rather changing what party they describe themselves as belonging to?
Charles Franklin: All the research says people's self-identification as a partisan is more durable than their vote choice, so it mostly should be partisans coming home rather than converting. But since 2004 we've seen a slow but steady shift away from the Republican Party, dropping from the low 30s to the mid-20s nationally and in a number of states. The key question is whether those former Republicans will vote like independents or be drawn back to their old party in the voting booth. That kind of instability of partisanship is possible, I think, for recent Republicans now saying they are independent.
Fairfax, Va.: Will Sen. Obama's beloved White Sox have a decent chance to make it to the World Series?
Nate Silver: Their pitching is underrated, but of the eight playoff teams, I probably wouldn't have them ranked higher than seventh in terms of their likelihood of winning it all. It's a tough field this year.
Aberdeen, U.K.: As pollster.com is partially owned by YouGov, do you take a broader interest in races abroad, such as in the U.K.? What in general do you think about the professional standards of polling companies, and do you think there is need for improving disclosure, validation and reliability, either in the National Council on Public Polls or alternatively the British Polling Council?
Charles Franklin: At Pollster we've focused almost exclusively on the U.S., which is what we know best. My partner Mark Blumenthal wrote extensively on disclosure in the spring and it is clear that this is an area many pollsters could improve upon. Of course when they disclose they get slammed, so one can understand the hesitation.
Mark Blumenthal: The bottom line is that I think pollsters can do much better in disclosing some basic details about their work, and I wrote an op-ed in another newspaper that explains in more detail. You might also want to browse some of what we've written on the topic of disclosure at Pollster.
But the bottom line is that when pollsters disclose some of the basics, as the GWU/Battleground pollsters have done, it allows commentators like Nate and me to ask pointed questions about the polls and how they were conducted. That helps everyone be better informed.
Chicago: Hey Charles, is there a formula you use for determining the date ranges for your national poll average?
Charles Franklin: We estimate our trends using a statistical technique called "local regression" rather than a moving average, so the date range is not really an issue for us. What matters is the degree of smoothing we do, and we frequently offer alternatives to that so you can see if a more sensitive (less smooth) estimate is different. Our new interactive charts let you adjust that yourself, and also decide which pollsters to include or exclude as well.
Seattle: Nate - -love your site and I understand your position, as discussed in yesterday's chat, that the whole point of polls is to predict the election, not what would happen today. But, are the potential effects of developments between now and then (debate gaffes, questions of security, whatever) likely to have more effect on changes between now and the election than your hypothesized trends? In short, although I love all the math, aren't the "unknowns" far greater in magnitude than the hypothesized "knowns"?
Nate Silver: We don't know exactly what the unknowns are going to be, but we do know that there are going to be unknowns, and we can account for that by essentially widening the margins of error on our estimates the further out that we are from Election Day. We have a pretty good idea, for instance, of how meaningful a five-point national lead is 35 days before the election, because we can look at how much polls moved during the final 35 days in past elections.
Chicago: Are there any swing states in particular that you feel might be more prone to a Bradley effect? McCain was in Iowa the other day ... is that a sign of something they may be seeing?
Charles Franklin: I'm quite puzzled why he went to Iowa. The Big Ten Poll, which I co-direct, is the only recent poll there showing it close -- everything else has a large lead, including Pollster's trend estimate.
Tirana, Albania: Hi Nate. Big fan of your site. As an American living overseas, I'm wondering whether any polling is done of this fairly large "community" of voters?
Nate Silver: The expat vote almost certainly is not getting polled. I've seen no attempt to quantify its effect; my hunch is that it might be worth a tenth of a point or so for Obama.
Durham, N.C.: If a poll says 700 likely/registered voters were polled, how many typically actually were called before they screen that number down to 700?
Charles Franklin: The "incidence" of registered voters is about 85 percent -- meaning of the numbers who will answer your questions, about 85 percent are registered. The response rate of completed interviews divided by total attempts to get a respondent is much lower -- in the 20s for most polling now, though that varies with how many callbacks you do and how long you stay in the field.
Durham, N.C.: Will Internet polling ever catch on, or is it too inherently biased to ever work accurately?
Nate Silver: I know of at least one very strong pollster who believes that Internet-based polling is the future. It might have to be the future, with people getting harder and harder to reach by phone. But right now it seems to be in a fairly primitive state. Zogby's approach, which relies on people volunteering to fill out survey forms, has such problems with selection bias that I don't think it's workable. But Economist/YouGov, which recruits and selects a panel ahead of time, has had pretty good success in the U.K. and has an interesting approach.
Mark Blumenthal: Okay, first, interests disclosed: Pollster.com is owned by YouGov/Polimetrix, the people who do the aforementioned Economist poll.
Second, the big problem is in drawing a true random sample of voters online. Right now there is no obvious way to do that. On the other hand, as Nate suggests, our ability to interview true random samples via telephone gets tougher and tougher every day. So I agree, eventually, we may have to.
Stratford, Conn.: It's interesting to see the differences between the ways different news organizations are projecting the outcomes of the individual state contests -- particularly in the number of states characterized as toss-ups. At what point would you expect those projections to start to look more alike?
Charles Franklin: Part of it is methodology -- ours is based on polls alone but Nate incorporates much more information about the states. I'd expect pretty good agreement, especially in this last month of the campaign, but because we (and other media) have different approaches we may still disagree even on Nov. 3.
Nate Silver: Charles has explained the differences between our two sites pretty well. Now, if you're comparing what we might do or what Charles might do to the maps you see from the major media, there may be a couple of differences: The major media organizations sometimes have access to internal polls, which can influence the way they're characterizing a state. But also, there is probably some "bias" in the mainstream media to characterize states as toss-ups when they really aren't, as I think they tend to be wary of calling a state wrong and having to walk it back later. They're overly cautious, in other words.
Naperville, Ill.: Recently, I heard Chuck Todd say that until about mid-October he prefers polls of "registered voters" instead of "likely voter" polls. How big is the difference, if any, and do you weight pollsters based on likely/registered? Thanks.
Charles Franklin: I looked at the RV LV comparison here. There are some differences -- not huge, but enough to start a bar fight.
Orange County, Calif.: I'm a certified public accountant and love math and numbers. What are your math backgrounds?
Charles Franklin: Applied stats in political science. I teach graduate methodology for social scientists in my day job. But you always should take one more stats class!
Nate Silver: My BA is in economics, so I'm definitely approaching this more from the applied side as well, and trying to take a problem-solving approach.
Gainesville, Fla.: Is any effort made to poll U.S. servicemen currently serving abroad? I don't know much of an impact their numbers would have on the election, but they certainly would be interesting to see in their own right.
Charles Franklin: Not much. MilitaryTimes.com has done some surveys of servicemen, but based on subscribers rather than random samples. As a percent of the U.S. population they are too small to be worth the effort, though absentee votes in some states can make a difference. Still, no sample frame I know of exists for doing this.
State College, Pa.: To what extent do you think polls influence how people ultimately vote? If one of the candidates has a huge lead in the polls come Election Day, is there an "I'm going to go with the winner" effect?
Charles Franklin: If it was a big effect, Clinton could not have won New Hampshire this year, and Truman never would have won in 1948.
Nate Silver: Indeed, the conventional wisdom holds that the opposite tends to be true -- that supporters of a candidate with a large lead can sometimes become complacent -- but I haven't seen any work on this.
Arlington, Texas: First, just wanted to say that I love both of your Web sites, but the idea of keeping track of House races got my interest piqued. It seems like the only races that get less polled than House races are the state legislature races, and so mapping them out seems less like math and more like fancy, if you'll forgive me (I won't pretend to understand the statistical analysis jargon y'all throw out there occasionally). Are the surges in enthusiasm around Obama on the one hand and Palin on the other having an effect on polling frequency, or is this just part of the maturation of online pollsters in general?
Mark Blumenthal: I think the surge in enthusiasm for this campaign absolutely has helped increased the frequency of polls you're seeing on the news and online.
I wanted to chip in two thoughts: One is that we will have been entering and charting data for House races on Pollster -- it's just a little hard to find (follow the map for "all other races"). By week's end, however, we should have a Congressional District map in place.
Second, in case anyone is wondering about those "customizable charts" Charles keeps talking about, we posted information and video tutorial on how to use them here.
Manhasset, N.Y.: Will Bob Barr and Ralph Nader have any appreciable effect anywhere? Say, Georgia?
Charles Franklin: Probably not, given that Obama's effort to make Georgia competitive in the summer doesn't seem to have paid off. It would have to be awfully close for Barr to matter. Of course, in a Florida 2000-like case, any third-party votes at all matter.
Nate Silver: Barr's candidacy pretty much has washed out. The only thing the third-party candidates might do is serve as sort of a buffer for racist Democrats who don't want to vote for Obama.
Washington: How difficult is it for pollsters to quantify the support for Nader and Barr when their numbers are usually within the margin of error?
Charles Franklin: When their names are included they get a few percent, but by Election Day that usually declines a good bit. If you just let people tell you they will vote for "someone else" that number turns out to be closer to the actual third-party vote in most polls. It reflects real intention to vote rather than a passing dislike for the Democrat and Republican.
Albuquerque, N.M.: The Aug. 4, 2000, comparison that Charles linked to earlier in the chat shows about six-point pro-GOP swings in both 2000 and 2004 right about now (if I'm reading the chart right). Is there anything in the data suggesting such a swing will (or won't) happen again this year?
Charles Franklin: That's a big unknown -- my point would be that large swings have happened in the past two, though earlier in the campaigns. I think we underestimate the potential for large shifts, though the closer to Election Day we get, the less time for them to run their course, obviously.
Denver: There appears to be a significant time delay in poll movements between national and state polls, with state polls lagging. I've noticed this in comparing national and state polls coming out on the same day, so I don't think it is just that we poll the states less frequently.
Charles Franklin: I'm not sure there is that much of a lag. We've been showing both in some recent daily updates and the moves match pretty well.
The bigger issue is that noncompetitive states seldom are sampled so it is the competitive states that dominate state polls. Nonetheless we see that even the solid blue and red states move in parallel with the national. I take that as good evidence of national forces at work across states.
Charlottesville, Va.: I'm curious as to both of your thoughts on Appalachia. What is the significance of Appalachia in a general election? Is it worth campaign resources to invest in Appalachia as a geographic region with tentacles in Virginia, North Carolina, Tennessee, Ohio, Pennsylvania and West Virginia, rather than state-by-state? Thanks! Best wishes to you both.
Nate Silver: The Obama campaign has been playing with a pretty big map, but the one area that they clearly seem to have excluded is Appalachia. There are arguments that they should have staffed up West Virginia, where some polling has had it reasonably close, but I think their feeling is that they had such a frustrating time there during the primaries that they don't want to fight an uphill battle again. The dire state of the economy may help them from getting completely blown out in those areas, however, as might any last-minute campaigning by the Clintons.
Essex Junction, Vt.: McCain seems to have argued vigorously against some polls as "outliers," which would suggest that he's afraid of excessive discouragement. Any evidence that happens?
Charles Franklin: I think all campaigns are sensitive to a "bad poll," and the easy response is to trash them. One reason we show all the polls in our charts is so you can see for yourself what's out of line with other recent results.
Nate Silver: The one particular thing that McCain may be afraid of is for Republican activists to start focusing more on congressional races, which is sort of what happened with Bob Dole in 1996.
Santa Rosa, Calif.: Can polls account for absentee or early voting? For example, some in Ohio already have voted.
Charles Franklin: Many ask in states with big early vote percentages.
Nate Silver: It shouldn't be that difficult to account for these voters. As Charles says, you should begin to see an "already voted" category broken out in a lot of these polls.
Highland Park, N.J.: According the RealClearPolitics the most recent Texas poll was in late August -- and McCain was up by 10 points. Any ideas of more recent polls in Texas?
Charles Franklin: I expect that no one thinks Texas is likely to change.
Poplar Bluff, Mo.: Will Missouri be a true bellwether state, as in the past several elections?
Nate Silver: I tend do doubt it. The region of the country that I call the "highlands" -- ranging from Missouri through Tennessee and Kentucky and West Virginia -- is the one place where the Republicans have held their ground while the rest of the country is becoming bluer. As a result, Missouri is moving away from the median of the electorate.
Charles Franklin: Trending for McCain for a while now. It hasn't yet shown the upturn Obama has enjoyed in the past two weeks of national data.
Charles Franklin: Thanks to you all. It has been fun. Sorry if we didn't get to everything.
Nate Silver: Happy that you all could join us today. It's going to be a busy five weeks!
washingtonpost.com: Live Discussion: Analyzing Sarah Palin's Rhetoric (washingtonpost.com, Live NOW)
washingtonpost.com: Upcoming Discussion: Tucker Carlson vs. Ana Marie Cox (washingtonpost.com, 2 p.m. ET)
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.