4.12.2008
Clinging to Guns
@ 4:54 PM
Top ten states the country, by percentage of gun ownership.
1. Wyoming 59.7%
2. Alaska 57.8%
3. Montana 57.7%
4. South Dakota 56.6%
5. West Virginia 55.4%
6. Mississippi 55.3%
7. Arkansas 55.3%
8. Idaho 55.3%
9. Alabama 51.7%
10. North Dakota 50.7%Among the seven states on this list that have held primaries or caucuses so far, Obama has won 6; Clinton has won only her home state of Arkansas.
Today's Polls, 4/12
@ 3:01 PM
We have only one poll today, but it's an interesting one, as Rasmussen has Barack Obama tied with John McCain in North Carolina -- Hillary Clinton is 11 points behind McCain.
As I alluded to yesterday, Obama appears to get a bounce in states where he runs an active primary campaign. What we don't know yet -- and what we might not know for a long time -- is whether this bounce is permanent or temporary. Regardless, North Carolina has moved up to 6th on Obama's Swing State List, right behind Virginia.
EDIT: The charts have finally been updated.
As I alluded to yesterday, Obama appears to get a bounce in states where he runs an active primary campaign. What we don't know yet -- and what we might not know for a long time -- is whether this bounce is permanent or temporary. Regardless, North Carolina has moved up to 6th on Obama's Swing State List, right behind Virginia.
EDIT: The charts have finally been updated.
4.11.2008
Regions
@ 9:53 PM
Anybody like this?
The idea is to create regions that are politically coherent -- sharing similar demographics and generally tending to vote together. And if you're playing at home, each region must have at least four states and be geographically contiguous, although you're allowed to cheat slightly on Alaska and Hawaii.
EDIT: Or this?
The idea is to create regions that are politically coherent -- sharing similar demographics and generally tending to vote together. And if you're playing at home, each region must have at least four states and be geographically contiguous, although you're allowed to cheat slightly on Alaska and Hawaii.
EDIT: Or this?
Clinton's Unity Bounce
@ 7:38 PM
Pursuant to last night's post about a potential Unity Bounce for Barack Obama, it's only fair to run Clinton's numbers too. Presently, 19% of Obama supporters say they'll vote for McCain in November. This is smaller than the percentage of Hillary supporters who say they'll vote for McCain (28%), but nevertheless fairly substantial. What if half of those supporters (9.5%) actually wound up voting for Clinton?
Clinton gains about 50 electoral votes, and 3.2 percent in the popular vote tally. Her win percentage improves from 38.5% to 60.0%. Across the board, these numbers are about 70% as large as Obama's Unity Bounce -- which is a predictable result, since there are about 70% as many Obama defectors as there are Clinton defectors.
This would produce an awfully interesting election on the state-by-state level:
Clinton would be in a strong, if vulnerable, position. Formally, we have Clinton projected to win Kerry states, plus Florida, Ohio, New Mexico, Arkansas and West Virginia, but minus New Hampshire and Wisconsin. That would give her exactly 300 electoral votes. But there also a lot of states colored white, or a very pale shade of color, indicating toss-ups. Her win percentage is 62% in Michigan, 53% in Florida, and 55% in Oregon; New Mexico and Washington are also below 70%. If she lost Florida and Oregon, for instance, she'd fall 3 electoral votes short of the nomination. On the other hand, she'd also have some further pickup opportunities: both Nevada and Missouri are at 48%, and Wisconsin at 49%. Pennsylvania, meanwhile, gets all the way up to 83%, at which point she might be able to win it without a serious investment of resources. It would be a losable election, but one that would be hers to lose.
Clinton gains about 50 electoral votes, and 3.2 percent in the popular vote tally. Her win percentage improves from 38.5% to 60.0%. Across the board, these numbers are about 70% as large as Obama's Unity Bounce -- which is a predictable result, since there are about 70% as many Obama defectors as there are Clinton defectors.
This would produce an awfully interesting election on the state-by-state level:
Clinton would be in a strong, if vulnerable, position. Formally, we have Clinton projected to win Kerry states, plus Florida, Ohio, New Mexico, Arkansas and West Virginia, but minus New Hampshire and Wisconsin. That would give her exactly 300 electoral votes. But there also a lot of states colored white, or a very pale shade of color, indicating toss-ups. Her win percentage is 62% in Michigan, 53% in Florida, and 55% in Oregon; New Mexico and Washington are also below 70%. If she lost Florida and Oregon, for instance, she'd fall 3 electoral votes short of the nomination. On the other hand, she'd also have some further pickup opportunities: both Nevada and Missouri are at 48%, and Wisconsin at 49%. Pennsylvania, meanwhile, gets all the way up to 83%, at which point she might be able to win it without a serious investment of resources. It would be a losable election, but one that would be hers to lose.
Today's Polls, 4/11
@ 5:46 PM
The headliner today is a Rasmussen Reports survey of Pennsylvania, which contains good news for both Democrats in the Keystone, with Hillary Clinton leading John McCain by 9 points and Barack Obama leading him by 8. Clinton's numbers have consistently been pretty good in Pennsylvania -- she's been ahead in the last four polls by margins ranging from 3 points to 9 -- whereas Obama's have been all over the board. But this could potentially be a sign that Obama's general election numbers tend to improve in states where he runs an active primary campaign. It will be interesting to see whether there is any similar effect for the Democrats in Indiana and North Carolina.
Rasmussen also surveyed Louisiana: Obama trails McCain by 11 there, and Hillary by 22. This is something of a reversal from the numbers we usually see in the Deep South, where Clinton has generally outpolled Obama. One thing to watch for is that polls in non-competitive states can affect our regression equations as much as polls taken anywhere else, so Clinton for instance has taken a hit to her West Virginia numbers as a result of this survey, which has some commonalities with Louisiana in terms of education levels and religiosity.
Finally (and speaking of non-competitive states): in Arizona, McCain has a 22-point lead over both Democrats according to a Northern Arizona University poll.
ALSO, eagle-eyed observers will note that Minnesota has been moved from the Prairie region to the Great Lakes reason, which affects nothing but aesthetics, but it felt weird not having the Land of 10,000 Lakes grouped with its Great Lakes brethren.
Rasmussen also surveyed Louisiana: Obama trails McCain by 11 there, and Hillary by 22. This is something of a reversal from the numbers we usually see in the Deep South, where Clinton has generally outpolled Obama. One thing to watch for is that polls in non-competitive states can affect our regression equations as much as polls taken anywhere else, so Clinton for instance has taken a hit to her West Virginia numbers as a result of this survey, which has some commonalities with Louisiana in terms of education levels and religiosity.
Finally (and speaking of non-competitive states): in Arizona, McCain has a 22-point lead over both Democrats according to a Northern Arizona University poll.
ALSO, eagle-eyed observers will note that Minnesota has been moved from the Prairie region to the Great Lakes reason, which affects nothing but aesthetics, but it felt weird not having the Land of 10,000 Lakes grouped with its Great Lakes brethren.
If it worked for Joe Klein...
@ 5:42 PM
Whoa, Ben Smith can really move traffic. To clarify an issue that came up in the Politico.com thread:
Yes, this is a part-time job. The thing is I really like my full-time job, which is why I blog pseudonymously.
But yes, I sometimes wish that I that I hadn't named myself after a chili pepper. It's Poblan-o, by the way, not Poblan-a.
Yes, this is a part-time job. The thing is I really like my full-time job, which is why I blog pseudonymously.
But yes, I sometimes wish that I that I hadn't named myself after a chili pepper. It's Poblan-o, by the way, not Poblan-a.
Contemplating the Unity Bounce
@ 2:30 AM
Chuck Todd opines that the general election numbers for the Democrats are presently at their floor:
Todd is not necessarily predicting how long the bounce will last. (Indeed, the term 'bounce' can have two somewhat opposite connotations, sometimes signifying a temporary shift in the polls and sometimes a permanent one; we might be better off using a term like 'bubble' in the first case and a term like 'shock' in the latter). But he's saying the election will wind up somewhere between where it is now, and where it is now plus 10 points for the Democrat. Those are his floor and his ceiling.
This assessment seems somewhat optimistic from where I sit; I agree with Todd's ceiling for the Democrats, but I think the floor is lower in the event of a badly-managed campaign. With that said, if I had to guess whether the polls for the Democrat are higher or lower on June 15th than they are now, I would guess higher -- if the nomination has been determined by then.
But let's try and be a little more specific about this. If it's Obama who locks down the nomination, most of his bounce would presumably come from among the 28 percent of Clinton-supporting Democrats who are presently telling pollsters that they plan to vote for John McCain. What if, say, half of these Clinton defectors -- 14 percent of her voters overall -- actually wind up casting their ballot for Barack Obama in November? How would his electoral math change?
To get at this, we first need to estimate the number of Clinton-supporting Democrats. I will do this by multiplying the percentage of voters who identify as Democrats in each state by the percentage of Democrats who supported Clinton in that state's primaries or caucuses. (For states that have not yet had a primary, I use current polling averages from pollster.com. For the two states that have neither had a primary nor have any polls -- these are Montana and South Dakota -- I assume Clinton's support to be 40 percent). Then, I multiply the result by 14 percent. For example, the calculation would go as follows in Ohio:
35% of the electorate identifies as Democrat x
54% of Democrats supported Clinton in the primary x
14% of these voters are showing up as McCain voters in the polls, but will eventually vote for Obama.
That works out to 35% x 54% x 14% = 2.6% in Ohio. So -- Obama would gain 2.6 points in the Ohio polls? Actually, he'd gain twice that amount -- 5.2 points -- because when a defector switches back to Obama, not only is he gaining a vote, but McCain is losing one. So what I've done is to add 5.2 points to Obama's polling numbers in each of his Ohio polls, and rerun my simulations. I also performed the same calculation in each of the other 49 states (the precise amount of the bounce varies some from state to state; it's as high as 8.0 points in Arkansas, but as low as 1.1 points in Idaho).
So what happens when we do this? The effects are relatively dramatic.
Obama gains a net of 4.6 points in the popular vote -- from 0.4 points down against McCain, to 4.2 points ahead. Perhaps more importantly, his projection improves by 65 electoral votes, and he goes from winning the election 44% of the time to 73% of the the time.
We can also look at what these extra votes would do for Obama on a state-by-state basis:
That is perhaps not all that dramatic a difference visually, but look at where the colors are changing. Obama goes from having a 46% chance of winning Pennsylvania to a 74% chance, from a 39% chance of winning Ohio to a 65% chance, and from a 50% chance of winning Michigan to a 77% chance. Under this scenario, the most closely-contested states on election day would be Virginia, North Dakota, Florida, and Montana, all of which have been GOP strongholds in recent elections, and any of which would be the cherry on top of Obama's electoral sundae.
You might assume that, being a professed supporter of Barack Obama, I'm running these numbers as a means of suggesting that Hillary Clinton should concede the nomination. But that's not really the case. What matters, after all, is not what Clinton does, but what her supporters do. My feeling is that it's very important to Clinton supporters (and likewise, to Obama supporters) to feel like their candidate has gotten a fair fight. I don't know whether it's disrespectful to Clinton to suggest that she should concede, but I do think it's disrespectful to her supporters.
The key concept here is legitimacy. Obama supporters are used to thinking about legitimacy solely in terms of a potential Hillary Clinton nomination (e.g. "Hillary would not be legitimate if the superdelegates overturned the pledged delegates"). However, from a Clinton supporter's point of view, there is sort of a sliding scale of legitimacy with respect to a potential Barack Obama nomination. If Obama is to wind up with anything near this 65-EV "unity bounce", he'll have to be toward the top of that scale.
The singular easiest way for him to do would be to win Pennsylvania. If that happened, Obama would have won a major victory on Clinton turf; her supporters might be disappointed by the result of the nomination process, but they'd see it as manifestly fair and legitimate. Failing that, the next best way would be for Obama to win both Indiana and North Carolina. Failing that, gain a net positive number of pledged delegates and popular votes between Pennsylvania, North Carolina, and Indiana, at which point the nomination would probably be conceded. Failing that, win the overall popular vote plus the pledged delegate count. And failing that, lose the popular vote, but win based on pledged delegates. At some point along this spectrum, of course, we'd have gotten to the point where many Clinton supporters would not see an Obama nomination as manifestly legitimate.
What needs to be understood about all the money that Obama is pouring into Pennsylvania, and all the time he's spending in Indiana, is that it's not so much a primary strategy as a general election strategy. Few things that Obama could do would increase his chances of winning the general election more than winning the Pennsylvania or Indiana primaries. The Unity Bounce is likely to be much more substantial if Obama closes out the nomination process with a victory, rather than winning it sort of inertially.
In the meantime, Obama needs to be very, very careful about sending any sort of message that Clinton should end her campaign. If Clinton were to concede while she was perceived as having a legitimate path to the nomination, Obama's nomination would feel force-fed to many of her supporters. Clinton, for her part, needs to be more careful than she has been about undermining the legitimacy of what Obama has accomplished so far. Arguments that she might think are dogwhistles to the superdelegates might in fact be picked up by many of her supporters -- and they might act on those arguments in November.
This cycle is inherently relatively favorable to the Democrats, and we can see what sort of upside they have if Democrats simply vote along party lines. But both campaigns have had some blind spots in terms of thinking about how the arguments they're making now could play out in November. I don't understand why the messaging from both Democratic campaigns isn't a simple "we want to win every vote we can, and we're taking it one vote at a time."
Currently polls show McCain either narrowly ahead or even with both Hillary Clinton and Barack Obama. It is impressive considering how poorly the GOP, and specifically the president, are viewed by the public.Todd's expectation is that if the Democratic candidate wraps up the nomination within the next 4-6 weeks, he can expect to see a bounce of about 10 points in the polls. And yes, I'm using the male pronoun "he", because "next 4-6 weeks" means "not long after North Carolina and Indiana". The only candidate who can wrap up the nomination within that time frame, barring some unforeseen scandal or contingency, is Barack Obama.But it is a faux lead. If the de facto Democratic nominee is clear within the next 4-6 weeks, that person will see a poll bounce. And according to GOP pollster Steve Lombardo, it could be one heck of a bounce, like post-convention. He anticipates the Democratic candidate will move up 10 points once the primary race is over. [...]
The initial bounce will set the polling numbers – the floor and ceiling – for the Democrats, who clearly have the generic advantage this cycle. Those parameters will dictate the morale within the GOP base.
If McCain’s is hanging in, behind by 10 or so points, then it is clear he will have a shot. If the bounce pushes the Democratic nominee to as much as a 15 point lead, it may be very demoralizing to the GOP.
Todd is not necessarily predicting how long the bounce will last. (Indeed, the term 'bounce' can have two somewhat opposite connotations, sometimes signifying a temporary shift in the polls and sometimes a permanent one; we might be better off using a term like 'bubble' in the first case and a term like 'shock' in the latter). But he's saying the election will wind up somewhere between where it is now, and where it is now plus 10 points for the Democrat. Those are his floor and his ceiling.
This assessment seems somewhat optimistic from where I sit; I agree with Todd's ceiling for the Democrats, but I think the floor is lower in the event of a badly-managed campaign. With that said, if I had to guess whether the polls for the Democrat are higher or lower on June 15th than they are now, I would guess higher -- if the nomination has been determined by then.
But let's try and be a little more specific about this. If it's Obama who locks down the nomination, most of his bounce would presumably come from among the 28 percent of Clinton-supporting Democrats who are presently telling pollsters that they plan to vote for John McCain. What if, say, half of these Clinton defectors -- 14 percent of her voters overall -- actually wind up casting their ballot for Barack Obama in November? How would his electoral math change?
To get at this, we first need to estimate the number of Clinton-supporting Democrats. I will do this by multiplying the percentage of voters who identify as Democrats in each state by the percentage of Democrats who supported Clinton in that state's primaries or caucuses. (For states that have not yet had a primary, I use current polling averages from pollster.com. For the two states that have neither had a primary nor have any polls -- these are Montana and South Dakota -- I assume Clinton's support to be 40 percent). Then, I multiply the result by 14 percent. For example, the calculation would go as follows in Ohio:
35% of the electorate identifies as Democrat x
54% of Democrats supported Clinton in the primary x
14% of these voters are showing up as McCain voters in the polls, but will eventually vote for Obama.
That works out to 35% x 54% x 14% = 2.6% in Ohio. So -- Obama would gain 2.6 points in the Ohio polls? Actually, he'd gain twice that amount -- 5.2 points -- because when a defector switches back to Obama, not only is he gaining a vote, but McCain is losing one. So what I've done is to add 5.2 points to Obama's polling numbers in each of his Ohio polls, and rerun my simulations. I also performed the same calculation in each of the other 49 states (the precise amount of the bounce varies some from state to state; it's as high as 8.0 points in Arkansas, but as low as 1.1 points in Idaho).
So what happens when we do this? The effects are relatively dramatic.
Obama gains a net of 4.6 points in the popular vote -- from 0.4 points down against McCain, to 4.2 points ahead. Perhaps more importantly, his projection improves by 65 electoral votes, and he goes from winning the election 44% of the time to 73% of the the time.
We can also look at what these extra votes would do for Obama on a state-by-state basis:
That is perhaps not all that dramatic a difference visually, but look at where the colors are changing. Obama goes from having a 46% chance of winning Pennsylvania to a 74% chance, from a 39% chance of winning Ohio to a 65% chance, and from a 50% chance of winning Michigan to a 77% chance. Under this scenario, the most closely-contested states on election day would be Virginia, North Dakota, Florida, and Montana, all of which have been GOP strongholds in recent elections, and any of which would be the cherry on top of Obama's electoral sundae.
You might assume that, being a professed supporter of Barack Obama, I'm running these numbers as a means of suggesting that Hillary Clinton should concede the nomination. But that's not really the case. What matters, after all, is not what Clinton does, but what her supporters do. My feeling is that it's very important to Clinton supporters (and likewise, to Obama supporters) to feel like their candidate has gotten a fair fight. I don't know whether it's disrespectful to Clinton to suggest that she should concede, but I do think it's disrespectful to her supporters.
The key concept here is legitimacy. Obama supporters are used to thinking about legitimacy solely in terms of a potential Hillary Clinton nomination (e.g. "Hillary would not be legitimate if the superdelegates overturned the pledged delegates"). However, from a Clinton supporter's point of view, there is sort of a sliding scale of legitimacy with respect to a potential Barack Obama nomination. If Obama is to wind up with anything near this 65-EV "unity bounce", he'll have to be toward the top of that scale.
The singular easiest way for him to do would be to win Pennsylvania. If that happened, Obama would have won a major victory on Clinton turf; her supporters might be disappointed by the result of the nomination process, but they'd see it as manifestly fair and legitimate. Failing that, the next best way would be for Obama to win both Indiana and North Carolina. Failing that, gain a net positive number of pledged delegates and popular votes between Pennsylvania, North Carolina, and Indiana, at which point the nomination would probably be conceded. Failing that, win the overall popular vote plus the pledged delegate count. And failing that, lose the popular vote, but win based on pledged delegates. At some point along this spectrum, of course, we'd have gotten to the point where many Clinton supporters would not see an Obama nomination as manifestly legitimate.
What needs to be understood about all the money that Obama is pouring into Pennsylvania, and all the time he's spending in Indiana, is that it's not so much a primary strategy as a general election strategy. Few things that Obama could do would increase his chances of winning the general election more than winning the Pennsylvania or Indiana primaries. The Unity Bounce is likely to be much more substantial if Obama closes out the nomination process with a victory, rather than winning it sort of inertially.
In the meantime, Obama needs to be very, very careful about sending any sort of message that Clinton should end her campaign. If Clinton were to concede while she was perceived as having a legitimate path to the nomination, Obama's nomination would feel force-fed to many of her supporters. Clinton, for her part, needs to be more careful than she has been about undermining the legitimacy of what Obama has accomplished so far. Arguments that she might think are dogwhistles to the superdelegates might in fact be picked up by many of her supporters -- and they might act on those arguments in November.
This cycle is inherently relatively favorable to the Democrats, and we can see what sort of upside they have if Democrats simply vote along party lines. But both campaigns have had some blind spots in terms of thinking about how the arguments they're making now could play out in November. I don't understand why the messaging from both Democratic campaigns isn't a simple "we want to win every vote we can, and we're taking it one vote at a time."
4.10.2008
Small States, Big Advantage
@ 7:54 PM
One thing today's Alaska poll reiterated, in which Obama ran fully 22 points better against John McCain than Hillary Clinton, is that we're generally going to see much larger differences in the performances of the two candidates in smaller states, which may have quirkier and less heterogeneous demographics.
Keeping that in mind, here is one way to look at the relative electoral strengths of Clinton and Obama. What I've done is to take the difference in Win Percentage between the two candidates in a given state, and multiply them by the state's number of electoral votes. For example, our model projects Obama to win Nevada 35% of the time that Hillary does not (69% versus 34%), so we multiply this 35% by Nevada's 5 EV, and come up with a 1.75 electoral vote advantage for Obama. (I'm tired, so I hope this explanation suffices). If we run that calculation for all states, here are the top ten net electoral vote grabs for each candidate:
I think you can make arguments either way about whether one or another of these positions is stronger from a resource allocation and risk-hedging perspective.
Keeping that in mind, here is one way to look at the relative electoral strengths of Clinton and Obama. What I've done is to take the difference in Win Percentage between the two candidates in a given state, and multiply them by the state's number of electoral votes. For example, our model projects Obama to win Nevada 35% of the time that Hillary does not (69% versus 34%), so we multiply this 35% by Nevada's 5 EV, and come up with a 1.75 electoral vote advantage for Obama. (I'm tired, so I hope this explanation suffices). If we run that calculation for all states, here are the top ten net electoral vote grabs for each candidate:
Obama Clinton
1. Colorado 5.74 1. Florida 6.54
2. Iowa 4.48 2. Arkansas 4.08
3. Washington 4.44 3. Pennsylvania 4.06
4. Wisconsin 3.50 4. Ohio 3.64
5. Oregon 2.92 5. West Virginia 2.91
6. Illinois 2.73 6. Missouri 1.98
7. Virginia 2.72 7. Massachusetts 1.97
8. California 1.85 8. Tennessee 1.57
9. Connecticut 1.85 9. New York 1.47
10. New Hampshire 1.83 10. Louisiana 1.31The general pattern here is that Obama does better in small states by large margins, whereas Clinton does better in large states by small margins -- a pattern which is eerily reminiscent of what's happened in the Democratic primary campaign. So, for example, even though Hillary has an advantage in Pennsylvania, this is worth no more in terms of electoral votes than Obama's advantage in Iowa, because Hillary's advantage over Obama in Pennsylvania is small, whereas Obama's advantage over Hillary in Iowa is large.I think you can make arguments either way about whether one or another of these positions is stronger from a resource allocation and risk-hedging perspective.
Today's Polls: Alaska!
@ 4:15 PM
An interesting set of polls today, and the first time in a while that there's enough volume that we need to embed a chart to keep up with everything:
Yes, Rasmussen surveyed Alaska. And a good thing they did, because the race is potentially competitive there. The modest 5-point deficit they show for Obama is consistent with what SurveyUSA showed back at the end of February, as well as the inklings of our regression model. Still, it is arguably more meaningful coming from Rasmussen, a pollster whose results tend to be a little more conservative and a little bit more oriented around traditional red-blue lines.
It seems to me that there are two things that could move Alaska into the category of a true toss-up state. One would be the entry of a third-party candidate like Bob Barr. Alaska has always had strong affinity for third-party candidates, and if such a candidate were to siphon more votes from McCain than Obama, the state becomes very competitive.
The second contingency would be a well-run Senate campaign by Mark Begich, who is challenging longtime GOP incumbent Ted Stevens, and who (also according to Rasmussen) trails by just one point in that endeavor. It almost seems as though the Senatorial coattails could carry the Presidential race, rather than the other way around. It is hard to imagine Obama winning Alaska if Mark Begich loses it. On the other hand, if Begich were to focus his campaign around Stevens' ethical lapses, that dovetails fairly well with Obama's message. Also of note: Alaska is the youngest state in the country, an odd match for the octogenarian Stevens and the septuagenarian McCain.
Moving onward, in ascending number of electoral votes: Rasmussen also shows a relatively competitive race (Obama -5) in Montana, which Bill Clinton carried in 1992. Demographically, Montana is fairly similar to Alaska. While it lacks a competitive Senate race, it does have more of a Democratic party infrastructure, and for that reason is more likely to be treated as a competitive state by the Obama campaign.
Obama has consistently outpolled Clinton in New Mexico, and this poll is no exception. That comes as a little bit of a surprise to the regression model, which thinks that heavily Hispanic and relatively impoverished New Mexico should be a better state for Hillary Clinton. But that's not what the polls have said so far, and Hillary only won the primary there by the slimmest of margins.
This Wisconsin poll should not be taken all that seriously: it's slightly out of date, and it consists of a small sample of just 400 adults (not even registered voters). Although St. Norbert College has been polling Wisconsin for a while, it gets an appropriately low weighting.
Finally, our reality check in all of this is Ohio, where the numbers haven't moved much from Rasmussen's previous poll that had both Democrats down 6. Hillary, who has fared better than this in other Ohio surveys, remains a slight favorite in the state according to our averages, and Obama a slight underdog.
Yes, Rasmussen surveyed Alaska. And a good thing they did, because the race is potentially competitive there. The modest 5-point deficit they show for Obama is consistent with what SurveyUSA showed back at the end of February, as well as the inklings of our regression model. Still, it is arguably more meaningful coming from Rasmussen, a pollster whose results tend to be a little more conservative and a little bit more oriented around traditional red-blue lines.
It seems to me that there are two things that could move Alaska into the category of a true toss-up state. One would be the entry of a third-party candidate like Bob Barr. Alaska has always had strong affinity for third-party candidates, and if such a candidate were to siphon more votes from McCain than Obama, the state becomes very competitive.
The second contingency would be a well-run Senate campaign by Mark Begich, who is challenging longtime GOP incumbent Ted Stevens, and who (also according to Rasmussen) trails by just one point in that endeavor. It almost seems as though the Senatorial coattails could carry the Presidential race, rather than the other way around. It is hard to imagine Obama winning Alaska if Mark Begich loses it. On the other hand, if Begich were to focus his campaign around Stevens' ethical lapses, that dovetails fairly well with Obama's message. Also of note: Alaska is the youngest state in the country, an odd match for the octogenarian Stevens and the septuagenarian McCain.
Moving onward, in ascending number of electoral votes: Rasmussen also shows a relatively competitive race (Obama -5) in Montana, which Bill Clinton carried in 1992. Demographically, Montana is fairly similar to Alaska. While it lacks a competitive Senate race, it does have more of a Democratic party infrastructure, and for that reason is more likely to be treated as a competitive state by the Obama campaign.
Obama has consistently outpolled Clinton in New Mexico, and this poll is no exception. That comes as a little bit of a surprise to the regression model, which thinks that heavily Hispanic and relatively impoverished New Mexico should be a better state for Hillary Clinton. But that's not what the polls have said so far, and Hillary only won the primary there by the slimmest of margins.
This Wisconsin poll should not be taken all that seriously: it's slightly out of date, and it consists of a small sample of just 400 adults (not even registered voters). Although St. Norbert College has been polling Wisconsin for a while, it gets an appropriately low weighting.
Finally, our reality check in all of this is Ohio, where the numbers haven't moved much from Rasmussen's previous poll that had both Democrats down 6. Hillary, who has fared better than this in other Ohio surveys, remains a slight favorite in the state according to our averages, and Obama a slight underdog.
Obama: the Zima-track candidate?
@ 8:12 AM
Gallup has more fascinating data on the "education gap" between Hillary Clinton and Barack Obama. As I inferred yesterday, this is something that carries on quite profoundly to the general election. Hillary performs fully 11 points better against McCain than Obama does among voters with a high school education or less. But Obama performs 6 points better than Hillary among adults with some college, 10 points better among college graduates, and 13 points better among those with postgraduate educations.
Rather than analyzing these numbers by themselves, let's compare them to previous election cycles, and see how other Democratic nominees performed in these categories. I was able to track down exit poll data for 1992-2004, as well as 1976 and 1980.
So much great stuff to think about here. The profile of the typical Democratic voter almost completely flipped from 1976 to 2004. But let me try and analyze this in very broad strokes.
In 1976, Carter won by carrying the high school group (which constituted a larger proportion of the electorate back then) by an 11 point margin; he also narrowly won voters with some college. But he lost to Gerald Ford among college graduates (the exit polls did not yet distinguish between 'regular' college grads and those with postgraduate educations).
Carter lost every category in 1980, which is what happens when you run into a wave like Ronald Reagan's. But he performed particularly badly among the "some college" voters, which he lost by 20 points in 1980 after having won in 1976. These folks may be your Reagan Democrats.
Bill Clinton was able to win twice by bringing back Carter's high-school only voters into the fold, but also adding a new group that hadn't previously been friendly to Democrats: highly-educated professionals and elites. In between these two groups, however, Bill did not do quite as well: he lost college graduates without advanced degrees even in 1996. Nevertheless, this provides one basic template for Democratic victory. If you win the elite vote plus significant segments of the working-class vote (e.g. minorities and unionized workers), you can ignore all those soccer moms in the middle.
Gore tried to replicate the Clinton model -- only, he lost a few points across the board, and wound up cutting it a few hundred votes too close in Florida. Kerry also tried to replicate the Clinton model, but didn't come all that close, losing the high school only voters for the first time since 1980. So where does that leave us today?
Clinton's numbers look the most like Al Gore's -- except, she lacks his support (for now she does anyway) among highly-educated voters. So one scenario this leads to is a case of electoral death by a thousand cuts. Clinton is presently an underdog in Wisconsin, Iowa, Washington, and Oregon, four states that Gore carried in 2000, and that have well-educated electorates. She is also an underdog in New Hampshire, which Kerry won in 2004.
Let's say that Clinton starts with Kerry's 252 electoral votes and then adds Ohio and Arkansas -- that gets her up to 278 EV:
Kerry = 252
+ Ohio = 272
+ Arkansas = 278
Clinton could afford to lose 9 electoral votes from this total, assuming that ties break for the Democrats. If she loses 10 electoral votes, she's out of luck. Wisconsin alone would do the trick, as would Minnesota, as would Washington; so would Oregon + New Hampshire. Hillary has a little more cushion if she wins Florida, which has 7 more electoral votes than Ohio, but still, losing any two states from the {WI, MN, WA, OR} group would be enough to cost her the election; she'd need to pick up a buffer somewhere like New Mexico or West Virginia. Hillary could even win both Ohio and Florida (and Arkansas) and still lose the election if she was swept in all four of WI, MN, WA and OR. Obviously, this latter scenario is not the one that John McCain would want to be banking his chips on, but these are some of the risks that underperforming among highly-educated voters could pose for Clinton.
Obama's coalition, on the other hand, at first appears to resemble the worst aspects of John Kerry's. He certainly does no better than Kerry among high school only voters. However, notice that Obama does perform significantly better than Kerry within the "donut hole" that Reagan opened up in 1980 -- voters with some college (this classification usually includes students), and voters with college degrees, but not postgraduate educations.
These are neither exactly your wine-track voters nor your beer-track voters; they're sort of your Zima-track voters. And there are significant numbers of them in states like Colorado and Iowa, where Obama is a favorite. Winning the Kerry states plus those two alone would leave him exactly 1 EV short, but he could pick that up with any of New Mexico, Nevada, or Virginia, or more exotically, one of Nebraska's electors. True, Obama would still need to hold onto Pennsylvania and Michigan for this math to work out for him. On the other hand, he is arguably less dependent on those states than is Hillary, and he should not have to play defense in states like Washington, Minnesota and New Hampshire, where Hillary will.
The unanswered question is whether you'd rather have to make up a deficit with higher-education voters, as Hillary does, or lower-education voters, as Obama does. Pew has found that low-education voters are indeed low-information voters, so there's a line of thinking that as these voters learn more about Obama, he'll pick up enough of them to win the election. On the other hand, a black candidate might face more challenges in this department than another one might. But I'd still rather be playing this hand than Clinton's, and have to win over high-education voters, who already know plenty about me and have decided I don't have their vote.
Rather than analyzing these numbers by themselves, let's compare them to previous election cycles, and see how other Democratic nominees performed in these categories. I was able to track down exit poll data for 1992-2004, as well as 1976 and 1980.
So much great stuff to think about here. The profile of the typical Democratic voter almost completely flipped from 1976 to 2004. But let me try and analyze this in very broad strokes.
In 1976, Carter won by carrying the high school group (which constituted a larger proportion of the electorate back then) by an 11 point margin; he also narrowly won voters with some college. But he lost to Gerald Ford among college graduates (the exit polls did not yet distinguish between 'regular' college grads and those with postgraduate educations).
Carter lost every category in 1980, which is what happens when you run into a wave like Ronald Reagan's. But he performed particularly badly among the "some college" voters, which he lost by 20 points in 1980 after having won in 1976. These folks may be your Reagan Democrats.
Bill Clinton was able to win twice by bringing back Carter's high-school only voters into the fold, but also adding a new group that hadn't previously been friendly to Democrats: highly-educated professionals and elites. In between these two groups, however, Bill did not do quite as well: he lost college graduates without advanced degrees even in 1996. Nevertheless, this provides one basic template for Democratic victory. If you win the elite vote plus significant segments of the working-class vote (e.g. minorities and unionized workers), you can ignore all those soccer moms in the middle.
Gore tried to replicate the Clinton model -- only, he lost a few points across the board, and wound up cutting it a few hundred votes too close in Florida. Kerry also tried to replicate the Clinton model, but didn't come all that close, losing the high school only voters for the first time since 1980. So where does that leave us today?
Clinton's numbers look the most like Al Gore's -- except, she lacks his support (for now she does anyway) among highly-educated voters. So one scenario this leads to is a case of electoral death by a thousand cuts. Clinton is presently an underdog in Wisconsin, Iowa, Washington, and Oregon, four states that Gore carried in 2000, and that have well-educated electorates. She is also an underdog in New Hampshire, which Kerry won in 2004.
Let's say that Clinton starts with Kerry's 252 electoral votes and then adds Ohio and Arkansas -- that gets her up to 278 EV:
Kerry = 252
+ Ohio = 272
+ Arkansas = 278
Clinton could afford to lose 9 electoral votes from this total, assuming that ties break for the Democrats. If she loses 10 electoral votes, she's out of luck. Wisconsin alone would do the trick, as would Minnesota, as would Washington; so would Oregon + New Hampshire. Hillary has a little more cushion if she wins Florida, which has 7 more electoral votes than Ohio, but still, losing any two states from the {WI, MN, WA, OR} group would be enough to cost her the election; she'd need to pick up a buffer somewhere like New Mexico or West Virginia. Hillary could even win both Ohio and Florida (and Arkansas) and still lose the election if she was swept in all four of WI, MN, WA and OR. Obviously, this latter scenario is not the one that John McCain would want to be banking his chips on, but these are some of the risks that underperforming among highly-educated voters could pose for Clinton.
Obama's coalition, on the other hand, at first appears to resemble the worst aspects of John Kerry's. He certainly does no better than Kerry among high school only voters. However, notice that Obama does perform significantly better than Kerry within the "donut hole" that Reagan opened up in 1980 -- voters with some college (this classification usually includes students), and voters with college degrees, but not postgraduate educations.
These are neither exactly your wine-track voters nor your beer-track voters; they're sort of your Zima-track voters. And there are significant numbers of them in states like Colorado and Iowa, where Obama is a favorite. Winning the Kerry states plus those two alone would leave him exactly 1 EV short, but he could pick that up with any of New Mexico, Nevada, or Virginia, or more exotically, one of Nebraska's electors. True, Obama would still need to hold onto Pennsylvania and Michigan for this math to work out for him. On the other hand, he is arguably less dependent on those states than is Hillary, and he should not have to play defense in states like Washington, Minnesota and New Hampshire, where Hillary will.
The unanswered question is whether you'd rather have to make up a deficit with higher-education voters, as Hillary does, or lower-education voters, as Obama does. Pew has found that low-education voters are indeed low-information voters, so there's a line of thinking that as these voters learn more about Obama, he'll pick up enough of them to win the election. On the other hand, a black candidate might face more challenges in this department than another one might. But I'd still rather be playing this hand than Clinton's, and have to win over high-education voters, who already know plenty about me and have decided I don't have their vote.
Makeover
@ 3:55 AM
Hope you like it. There are a few things that need to be tweaked but it's about 93% there. Thanks to Robert G. for his work on the new style sheets.
4.09.2008
WNBC/Marist: New York a Swing State?
@ 12:59 PM
The Swamp cites a WNBC/Marist poll that shows McCain highly competitive with both Democrats in the Empire State. In fact, he is leading Barack Obama by 2 points, while Hillary Clinton leads McCain by the same small margin.
We try and keep an open mind here, but ... I'm going to go ahead and call this one an outlier. For one thing, Marist has a below-average pollster rating. For another, this survey asked another question about a McCain-Condi Rice versus Clinton-Obama and Obama-Clinton tickets. If the Condi question was asked before the regular Presidential trial heat questions, that could influence perceptions of the latter -- a classic case of question order biasing the results.
We try and keep an open mind here, but ... I'm going to go ahead and call this one an outlier. For one thing, Marist has a below-average pollster rating. For another, this survey asked another question about a McCain-Condi Rice versus Clinton-Obama and Obama-Clinton tickets. If the Condi question was asked before the regular Presidential trial heat questions, that could influence perceptions of the latter -- a classic case of question order biasing the results.
We know even less than we think (a methodological note)
@ 12:45 PM
Way back when I started this, I identified three sources of error that might make election returns different from the polls: sampling error, state-specific movement, and national movement. The latter two types of error can be grouped together, and might not really be thought of as error at all; they're an inevitable consequence of surveying an election days, weeks or months before it occurs.
Then there is sampling error, which is intrinsic to the business of doing surveying. The problem is that actual errors are never as good as what the official margin of error advertises. There is a fourth (or if you prefer, second) source of error, which is methodological error introduced by the pollster. Conceptually:
We are of course aware of methodological error; that is why some pollsters get higher ratings than others.. However, we had not previously been accounting for it in our simulations; we were accounting for it in our averages, but not in the probability distributions around those averages. Beginning now, however, I am now modeling the error term we use in the simulations based on real, historical data. This means the simulations will now account for methodological error in addition to sampling error.
In plain English: we have introduced more uncertainty into our simulations. The practical upshot of this is that the state-by-state win percentages are regressed more toward the mean, particularly in states with limited polling data (although, this issue doesn't go away even if we have a large amount of polling, because the error terms tend to be correlated with one another across different survey). The overall effect of this adjustment is negligible, but it does boost the win percentage by a point or two for the losing candidate in each state. (So, Obama's win percentage goes up a little bit in Alaska, for instance, but goes down in Delaware).
Then there is sampling error, which is intrinsic to the business of doing surveying. The problem is that actual errors are never as good as what the official margin of error advertises. There is a fourth (or if you prefer, second) source of error, which is methodological error introduced by the pollster. Conceptually:
We are of course aware of methodological error; that is why some pollsters get higher ratings than others.. However, we had not previously been accounting for it in our simulations; we were accounting for it in our averages, but not in the probability distributions around those averages. Beginning now, however, I am now modeling the error term we use in the simulations based on real, historical data. This means the simulations will now account for methodological error in addition to sampling error.
In plain English: we have introduced more uncertainty into our simulations. The practical upshot of this is that the state-by-state win percentages are regressed more toward the mean, particularly in states with limited polling data (although, this issue doesn't go away even if we have a large amount of polling, because the error terms tend to be correlated with one another across different survey). The overall effect of this adjustment is negligible, but it does boost the win percentage by a point or two for the losing candidate in each state. (So, Obama's win percentage goes up a little bit in Alaska, for instance, but goes down in Delaware).
If 538 were trying to predict the PA primary... [UPDATED]
@ 9:49 AM
...which, rest assured, it is not. But I thought you might like to see what the race would look like with our weighting formulas.
One BIG difference: I have changed the half-life on the polls from the 30-day number that we use for the general election to 10 days for the primaries. So, recentness is at more of a premium (the specific number, however, is a WFG).
UPDATE: This now includes the new PPP poll that shows Clinton +3. Although that's a nominally better result for Clinton than PPP's previous poll, which had Obama ahead by two, this points to one of our differences in philosophy which is that we see this as more good data from a (relatively) good pollster over a large sample -- and more data showing a tight race. This method is now calling Clinton's edge at 6.9 points (she was at 7.5 previously).
One BIG difference: I have changed the half-life on the polls from the 30-day number that we use for the general election to 10 days for the primaries. So, recentness is at more of a premium (the specific number, however, is a WFG).
UPDATE: This now includes the new PPP poll that shows Clinton +3. Although that's a nominally better result for Clinton than PPP's previous poll, which had Obama ahead by two, this points to one of our differences in philosophy which is that we see this as more good data from a (relatively) good pollster over a large sample -- and more data showing a tight race. This method is now calling Clinton's edge at 6.9 points (she was at 7.5 previously).
The Education Gap, Part II
@ 6:39 AM
Gallup:
This is what we were talking about the other day. There is fully a 60-point swing between non-Hispanic white Democrats with no college education (who prefer Hillary by 30 points) to those with postgraduate educations (who prefer Obama by 29 points). The education gap is more substantial than the gender gap, the income gap (which may be entirely an artifact of the education gap), the white-Latino gap (ditto), or the age gap (although, the latter is close), and anything but the black-white gap. And it may be the most important of any of these "gaps" insofar as carrying over to the general election.
I still want to emphasize that this is neither inherently a good thing or a bad thing for either candidate. Hillary both does better than your usual Democrat among low-education voters and does worse than your usual Democrat among high-education voters. If the former were true but not the latter, she'd have wrapped up the nomination by now. If the latter were true but not the former, Obama would have wrapped up the nomination in February.
This is what we were talking about the other day. There is fully a 60-point swing between non-Hispanic white Democrats with no college education (who prefer Hillary by 30 points) to those with postgraduate educations (who prefer Obama by 29 points). The education gap is more substantial than the gender gap, the income gap (which may be entirely an artifact of the education gap), the white-Latino gap (ditto), or the age gap (although, the latter is close), and anything but the black-white gap. And it may be the most important of any of these "gaps" insofar as carrying over to the general election.
I still want to emphasize that this is neither inherently a good thing or a bad thing for either candidate. Hillary both does better than your usual Democrat among low-education voters and does worse than your usual Democrat among high-education voters. If the former were true but not the latter, she'd have wrapped up the nomination by now. If the latter were true but not the former, Obama would have wrapped up the nomination in February.
African-American turnout in the primaries
@ 1:34 AM
In my previous post, I suggested the African-American proportion of the electorate in the Democratic primaries has been equal to about 175% of the percentage of African-Americans in that state. This off-the-cuff estimate was slightly too high; the actual ratio so far has been 153%.
That's a chart of all 31 Democratic primaries and caucuses so far in which exit polls have been taken. As you can see, African-American turnout is relatively easy to predict. On the other hand, it should be easy to predict, so I'm interested in whether there's anything systematic about those states that tend to fall above or below the regression line.
Below are the turnout statistics for all states so far in which African-Americans make up at least 5 percent of the overall population. I've also included a couple of additional statistics: Bush's share of the white vote in 2004 (per CNN exit polls), and the number of days that Obama spent campaigning in that state, out of the last 30 days before the election (per the Washington Post; I have excluded debates from the analysis but included all other types of events).
We see that the turnout ratio is generally highest in the South, where Bush won a larger share of the white vote (did you know that Bush won 85% of whites in Mississippi in 2004? Wow). This, also, is entirely predictable. If most whites in the state are Republican, that means a relatively greater share of the Democratic primary electorate will be black.
The interesting question is whether Obama can improve black turnout by campaigning actively in a state. There appears to be some evidence of this: black turnout was conspicuously low in some states like New York, Oklahoma, Arkansas and Florida, in which Obama mounted a limited or dispirited campaign. If we regress the turnout ratio on "whitebush" and "obamadays", the Obama variable does show up as statistically significant at the 85% level (though, not at the 90% or 95% level). Specifically, the regression model specifies the following equation
ratio = whitebush * .013 + obamadays * .032 + .547
If we like, we can try and use this equation to predict the African-American turnout in some of the upcoming primaries. Of course, we don't know exactly how much time Obama is going to spend in each state, but let's say that he winds up spending portions of 15 of the final 30 days before the election in Pennsylvania, 7 in North Carolina, and 5 in Indiana. We can then extrapolate African-American turnout in these states:
So, the regression would estimate 17-18% African-American turnout in Pennsylvania (Survey USA's guess was 14%), 35% in North Carolina (Survey USA's guess was 32%), and 14% in Indiana (Survey USA's guess was 11%). So Survey USA's estimates in Pennsylvania and elsewhere might be low -- if Obama continues to campaign actively in these states and energizes black voters. Or they might be not, as this is a rather speculative exercise. The more conservative assumption is that black turnout in Pennsylvania will resemble the ratio established in Ohio -- about 150% of the state's African-American population share -- which would point toward a number of 15% or 16%.
That's a chart of all 31 Democratic primaries and caucuses so far in which exit polls have been taken. As you can see, African-American turnout is relatively easy to predict. On the other hand, it should be easy to predict, so I'm interested in whether there's anything systematic about those states that tend to fall above or below the regression line.
Below are the turnout statistics for all states so far in which African-Americans make up at least 5 percent of the overall population. I've also included a couple of additional statistics: Bush's share of the white vote in 2004 (per CNN exit polls), and the number of days that Obama spent campaigning in that state, out of the last 30 days before the election (per the Washington Post; I have excluded debates from the analysis but included all other types of events).
We see that the turnout ratio is generally highest in the South, where Bush won a larger share of the white vote (did you know that Bush won 85% of whites in Mississippi in 2004? Wow). This, also, is entirely predictable. If most whites in the state are Republican, that means a relatively greater share of the Democratic primary electorate will be black.
The interesting question is whether Obama can improve black turnout by campaigning actively in a state. There appears to be some evidence of this: black turnout was conspicuously low in some states like New York, Oklahoma, Arkansas and Florida, in which Obama mounted a limited or dispirited campaign. If we regress the turnout ratio on "whitebush" and "obamadays", the Obama variable does show up as statistically significant at the 85% level (though, not at the 90% or 95% level). Specifically, the regression model specifies the following equation
ratio = whitebush * .013 + obamadays * .032 + .547
If we like, we can try and use this equation to predict the African-American turnout in some of the upcoming primaries. Of course, we don't know exactly how much time Obama is going to spend in each state, but let's say that he winds up spending portions of 15 of the final 30 days before the election in Pennsylvania, 7 in North Carolina, and 5 in Indiana. We can then extrapolate African-American turnout in these states:
So, the regression would estimate 17-18% African-American turnout in Pennsylvania (Survey USA's guess was 14%), 35% in North Carolina (Survey USA's guess was 32%), and 14% in Indiana (Survey USA's guess was 11%). So Survey USA's estimates in Pennsylvania and elsewhere might be low -- if Obama continues to campaign actively in these states and energizes black voters. Or they might be not, as this is a rather speculative exercise. The more conservative assumption is that black turnout in Pennsylvania will resemble the ratio established in Ohio -- about 150% of the state's African-American population share -- which would point toward a number of 15% or 16%.
4.08.2008
Tonight's Polls, 4/8
@ 11:56 PM
I thought this was a poll-less day in so far as general election matchups go, but we do have a couple.
Firstly, in Washington, Survey USA has Obama leading by 7, but Clinton trailing by 1, against John McCain. These results are quite consistent with other polling we've seen in Washington, and not otherwise noteworthy.
Secondly, in Pennsylvania, a commenter tipped us to a new Strategic Vision poll, which, while it contains relatively good news for Obama in the primaries, shows Clinton faring much better in the general election (+3 versus -7). The general election numbers are a near-match for Strategic Vision's 3/29 poll of Pennsylvania, which showed Clinton leading McCain by 6 points but Obama trailing him by 5. A couple of additional notes/comments after the break.
#1. A number of people have asked me what I think about the primary polls in Pennsylvania; I think you should all read Mark Blumenthal. I don't particularly believe that there's actually been movement toward Clinton; Survey USA seems to show that, but three or four other pollsters show continued movement toward Obama. I do believe it's possible that Survey USA's turnout model, which seems more pessimistic for Obama, may be closer to the mark. In particular, their 58/42 female/male split seems highly consistent with previous primaries, and perhaps more reasonable than the 54/46 split they had before. However, their estimate of 14% African-American turnout may be low. The general rule of thumb is that African-American turnout in a Democratic primary is about 175% of the state's African-American population as a whole, which would imply turnout closer to 18 or 19%.
EDIT:Actually, the ratio is closer to 150% than 175%. So, Survey USA's estimate is not unreasonable. More on this in a moment.
#2. (Technical) I've made a slight adjustment to the 'recentness factor' based on some testing of previous election results that I've done. The half-life for discounting polls is now set at 30 days, regardless of the number of days until the election. I owe you guys a more thorough explanation on this and a couple of other things, but basically -- this turns out to be just as accurate as the method I'd been using before (actually, a little bit more so), and it has the virtue of being less complicated. The practical upshot of this for the time being is that older polls are discounted somewhat more heavily.
Firstly, in Washington, Survey USA has Obama leading by 7, but Clinton trailing by 1, against John McCain. These results are quite consistent with other polling we've seen in Washington, and not otherwise noteworthy.
Secondly, in Pennsylvania, a commenter tipped us to a new Strategic Vision poll, which, while it contains relatively good news for Obama in the primaries, shows Clinton faring much better in the general election (+3 versus -7). The general election numbers are a near-match for Strategic Vision's 3/29 poll of Pennsylvania, which showed Clinton leading McCain by 6 points but Obama trailing him by 5. A couple of additional notes/comments after the break.
#1. A number of people have asked me what I think about the primary polls in Pennsylvania; I think you should all read Mark Blumenthal. I don't particularly believe that there's actually been movement toward Clinton; Survey USA seems to show that, but three or four other pollsters show continued movement toward Obama. I do believe it's possible that Survey USA's turnout model, which seems more pessimistic for Obama, may be closer to the mark. In particular, their 58/42 female/male split seems highly consistent with previous primaries, and perhaps more reasonable than the 54/46 split they had before. However, their estimate of 14% African-American turnout may be low. The general rule of thumb is that African-American turnout in a Democratic primary is about 175% of the state's African-American population as a whole, which would imply turnout closer to 18 or 19%.
EDIT:Actually, the ratio is closer to 150% than 175%. So, Survey USA's estimate is not unreasonable. More on this in a moment.
#2. (Technical) I've made a slight adjustment to the 'recentness factor' based on some testing of previous election results that I've done. The half-life for discounting polls is now set at 30 days, regardless of the number of days until the election. I owe you guys a more thorough explanation on this and a couple of other things, but basically -- this turns out to be just as accurate as the method I'd been using before (actually, a little bit more so), and it has the virtue of being less complicated. The practical upshot of this for the time being is that older polls are discounted somewhat more heavily.
4.07.2008
Barr, the Libertarians and Georgia
@ 7:58 PM
The American Spectator has an interesting article today on Bob Barr, the former Georgia congressman who is gunning for the nomination of the Libertarian party. Barr has an interesting (and somewhat fluid/amorphous) set of political characteristics: anti-war, fiscally conservative, increasingly sympathetic toward civil libertarian causes (he's joined the ACLU), but also with some cultural conservative bona fides -- he was a ringleader in passing the Defense of Marriage Act in 1996 (a position he could now presumably try and justify from a federalism standpoint). Basically, he's from the Republican wing of the Libertarian party.
It's not certain that Barr is going to win the Libertarian party's nomination, for which he'll be competing with Mike Gravel (who won't win it) and a number of Libertarian-party insiders. It's also not completely out of the question, I guess, that Ron Paul could enter the running. But if he does, where might Barr have the biggest impact?
Below, I've listed the leading states by percentage of the general election vote that went for Michael Badnarik, the Libertarian candidate in 2004.
2004 Libertarian Party Vote Share
2004 Third Party Vote Share (Libertarian vote in parenthesis)
UPDATE: More good background on this from the Atlanta-Jorunal Constitution. Barr doesn't seem to care if he plays the role of spoiler.
It's not certain that Barr is going to win the Libertarian party's nomination, for which he'll be competing with Mike Gravel (who won't win it) and a number of Libertarian-party insiders. It's also not completely out of the question, I guess, that Ron Paul could enter the running. But if he does, where might Barr have the biggest impact?
Below, I've listed the leading states by percentage of the general election vote that went for Michael Badnarik, the Libertarian candidate in 2004.
2004 Libertarian Party Vote Share
I'll take "states that begin with the letter 'I'" for $200, Alex! Actually, that list is a little weird. I don't think that Indiana has ever really been thought of as a third-party stronghold, for instance. But one thing we need to contend with here is that some people just want to vote for a third-party ... and they're indifferent about which third party they vote for. In Indiana, there were only a couple of third parties on the ballot, so Badnarnik might have picked up a lot of the generic anti-establishment/protest vote. It might be more instructive to look at the total third-party vote in 2004.
Indiana 0.73%
Idaho 0.64%
Illinois 0.62%
Arizona 0.59%
Alaska 0.54%
Texas 0.52%
Massachusetts 0.52%
Wyoming 0.48%
Washington 0.42%
2004 Third Party Vote Share (Libertarian vote in parenthesis)
Alaska 3.42% (0.59%)That's a little closer to what I was expecting to see, with a concentration of states in the Interior West. If the American Spectator is right that Barr would drain more votes from McCain than Obama, Alaska and Montana bear watching. And although it hasn't recently had a heavy third-party vote, so might Georgia. If Barr could peel the votes of some culturally conservative whites from McCain in Georgia, who might mistake him for one of their own, Obama might have a path to victory on the strength of the black vote and upscale whites in suburban Atlanta.
Utah 2.47% (not on ballot)
Montana 2.37% (0.38%)
Vermont 2.26% (0.35%)
Wyoming 2.06% (0.42%)
Rhode Island 1.91% (not on ballot)
Maine 1.85% (0.27%)
Connecticut 1.74% (not on ballot)
Nevada 1.65% (not on ballot)
South Dakota 1.65% (0.25%)
UPDATE: More good background on this from the Atlanta-Jorunal Constitution. Barr doesn't seem to care if he plays the role of spoiler.
Clinton and Education Levels
@ 6:08 PM
Pursuant to the tease I put out this morning, I've been doing some further investigation into the regression model, and identified a variable that has some pretty interesting effects: a state's educational level, as measured by the average number of years of completed schooling per adult, according to the US Census Bureau.
As you can see from the graphs, there is essentially no relationship between a state's educational level and Clinton's performance in the polls; there is a rather strong relationship for Barack Obama. But it's actually Clinton's graph, rather than Obama's, that's unusual relative to how things have gone in the recent past. If you drew John Kerry's graph in 2004, it would look much more like Obama's than Clinton's.
Put differently, relative to John Kerry, Clinton performs worse in highly-educated states, and better in poorly-educated states. This turns out to be one of the more significant variables in her regression model; in fact, if you look JUST at the Kerry vote and education levels for Clinton, you do very nearly as well as if you consider all the other variables that our model evaluates.
Electorally, this is a bit of a wash for Clinton, but it might require a somewhat different allocation of resources than a Democrat like Obama would use. Following are the most and least educated states:
By the way, once we include the education variable, the Southern Baptist variable (which is changing to 'Evangelicals' -- I'll explain in a moment) drops out of Clinton's equation. It appears that Clinton does not have a particular advantage (relative to your usual Democrat) either in the South, or among evangelicals. Instead, she does better among low-education voters than most Democrats usually have, and there tend to be more of these voters in the South. But in a state like Georgia, which is significantly better educated than most of its neighbors, Clinton has performed quite badly in the polls.
I also tested whether its income levels, rather than educational levels, that appear to be the driving force behind this. It isn't. When both variables are included, the education variable remains highly statistically significant for Clinton, while the income variable drops out.
---
So long as I was including the education variable, I did a little bit of additional maintenance on the regression model:
1. Firstly, the variable for Southern Baptists was replaced with a variable for evangelicals. The Southern Baptist variable was always a little bit of a mess, as it was a hybrid of two different estimates of religious population. But I came across some very good, reliable-looking data on the number of evangelicals from the Association of Religion Data Archives (it's worth a few minutes of your time to explore the site). In addition to being a bit 'cleaner', this variable also turns out to have a slightly stronger relationship with Obama's polls than the old Southern Baptist variable (as I stated above, neither variable is significant in Clinton's model once we account for educational levels). Of note: the ARDA does not consider predominantly black churches in its definition of Evangelicals; these are white evangelical Protestants.
2. The two variables 'Democrat' and 'Independent', which represent party identification in 2004 CNN Exit Polls, were replaced with one variable, 'Partisan', which represents the percentage of Democrats less the percentage of Republicans (so in Arkansas, where 41% of the electorate identified themselves as Democrat and 31% as Republican, the partisan index is 41-31 = +10%). This is a cleaner way to do things, as the two variables we had before ('Democrat' and 'Independent') were somewhat intercorrelated; there doesn't appear to be any such thing as an 'independent spirit' that causes states with a high proportion of independents to behave differently from their overall party leanings. Hillary's support is oriented slightly more strongly around this partisan axis than John Kerry's; Obama's is oriented significantly less so.
3. (Very, very technical). The program is now performing a proper stepwise regression instead of just wiping out a whole bunch of variables at once; there was no reason we weren't doing this before, but it took me a bit of time to figure out the programming. Also, we've slightly lowered the threshold wherein we include a variable from 85% statistical significance to 80% statistical significance.
The site will be updated momentarily with the revised regression model included. Clinton appears to do very slightly better with the new model; there is no real change for Obama.
UPDATE:
For Obama, the parameters currently included in the regression are as follows:
Dropped: Education (highly correlated with Obama fundraising), AfAmerican.
For Clinton, the parameters are:
Dropped: AfAmerican, Evangelical, $_Obama.
As you can see from the graphs, there is essentially no relationship between a state's educational level and Clinton's performance in the polls; there is a rather strong relationship for Barack Obama. But it's actually Clinton's graph, rather than Obama's, that's unusual relative to how things have gone in the recent past. If you drew John Kerry's graph in 2004, it would look much more like Obama's than Clinton's.
Put differently, relative to John Kerry, Clinton performs worse in highly-educated states, and better in poorly-educated states. This turns out to be one of the more significant variables in her regression model; in fact, if you look JUST at the Kerry vote and education levels for Clinton, you do very nearly as well as if you consider all the other variables that our model evaluates.
Electorally, this is a bit of a wash for Clinton, but it might require a somewhat different allocation of resources than a Democrat like Obama would use. Following are the most and least educated states:
Average Years of Educational Attainment
Adults >=25
1. Colorado 13.46
2. Massachusetts 13.39
3. Maryland 13.37
4. New Hampshire 13.36
5. Vermont 13.33
6. Connecticut 13.31
7. Washington 13.28
8. Minnesota 13.25
9. Alaska 13.18
10. Montana 13.14
50. West Virginia 12.16
49. Kentucky 12.21
48. Mississippi 12.24
47. Arkansas 12.25
46. Louisiana 12.32
45. Texas 12.38
44. Alabama 12.42
43. Tennessee 12.43
42. South Carolina 12.52
41. Nevada 12.60So this helps to explain why, for instance, Clinton has struggled in the polls in Colorado and New Hampshire, and why McCain has been close to her in a couple of polls of Connecticut. The regression model now has more confidence in the polls in those states. On the other hand, we can also see why Clinton has polled relatively well in states like West Virginia and Tennessee, and there may be opportunities for her there.By the way, once we include the education variable, the Southern Baptist variable (which is changing to 'Evangelicals' -- I'll explain in a moment) drops out of Clinton's equation. It appears that Clinton does not have a particular advantage (relative to your usual Democrat) either in the South, or among evangelicals. Instead, she does better among low-education voters than most Democrats usually have, and there tend to be more of these voters in the South. But in a state like Georgia, which is significantly better educated than most of its neighbors, Clinton has performed quite badly in the polls.
I also tested whether its income levels, rather than educational levels, that appear to be the driving force behind this. It isn't. When both variables are included, the education variable remains highly statistically significant for Clinton, while the income variable drops out.
---
So long as I was including the education variable, I did a little bit of additional maintenance on the regression model:
1. Firstly, the variable for Southern Baptists was replaced with a variable for evangelicals. The Southern Baptist variable was always a little bit of a mess, as it was a hybrid of two different estimates of religious population. But I came across some very good, reliable-looking data on the number of evangelicals from the Association of Religion Data Archives (it's worth a few minutes of your time to explore the site). In addition to being a bit 'cleaner', this variable also turns out to have a slightly stronger relationship with Obama's polls than the old Southern Baptist variable (as I stated above, neither variable is significant in Clinton's model once we account for educational levels). Of note: the ARDA does not consider predominantly black churches in its definition of Evangelicals; these are white evangelical Protestants.
2. The two variables 'Democrat' and 'Independent', which represent party identification in 2004 CNN Exit Polls, were replaced with one variable, 'Partisan', which represents the percentage of Democrats less the percentage of Republicans (so in Arkansas, where 41% of the electorate identified themselves as Democrat and 31% as Republican, the partisan index is 41-31 = +10%). This is a cleaner way to do things, as the two variables we had before ('Democrat' and 'Independent') were somewhat intercorrelated; there doesn't appear to be any such thing as an 'independent spirit' that causes states with a high proportion of independents to behave differently from their overall party leanings. Hillary's support is oriented slightly more strongly around this partisan axis than John Kerry's; Obama's is oriented significantly less so.
3. (Very, very technical). The program is now performing a proper stepwise regression instead of just wiping out a whole bunch of variables at once; there was no reason we weren't doing this before, but it took me a bit of time to figure out the programming. Also, we've slightly lowered the threshold wherein we include a variable from 85% statistical significance to 80% statistical significance.
The site will be updated momentarily with the revised regression model included. Clinton appears to do very slightly better with the new model; there is no real change for Obama.
UPDATE:
For Obama, the parameters currently included in the regression are as follows:
Variable Coeff t-score
$_Obama 8.13 4.35
Kerry 0.59 4.22
Evangelical -0.34 3.47
Partisan -0.36 2.41
$_McCain -7.11 2.17
$_Clinton -2.52 1.65
Constant 3.17 1.63Dropped: Education (highly correlated with Obama fundraising), AfAmerican.
For Clinton, the parameters are:
Variable Coeff t-score
Kerry 0.63 7.47
Education -7.37 3.61
$_McCain -3.60 1.54
$_Clinton 1.33 1.36
Partisan 0.14 1.33
Constant 94.63 3.59Dropped: AfAmerican, Evangelical, $_Obama.
Graph(s) of the Day
@ 8:35 AM
This is a comparison of Obama's current polling average in each of the 50 states to the average level of educational attainment among adults in that state.
Now, here's Clinton's.
Make of that what you will.
Now, here's Clinton's.
Make of that what you will.
The Momentumless Primary
@ 12:30 AM
A week ago, Barack Obama led Hillary Clinton by 10 points in the Gallup Daily Tracking Poll. I asked the readers whether they expected today's poll -- it's April 6th -- to show a lead of greater than 10 points for Obama, or fewer.
I was sort of trying to goad people into taking the over. After all, we're all used to hearing about the importance of "momentum" and trendlines. Obama had gone, over the span of about 10 days, from being 7 points behind, to roughly tied, to 4 points ahead, to 10 points ahead. Wasn't it logical that he'd be some number larger than 10 points ahead after another week's worth of "momentum"? Wasn't the headline that Obama was pulling away with the race?
But most people did not take the bait, and said they expected Obama to be fewer than 10 points ahead. Which, as it turns out, was the right answer. In today's Gallup Tracker, Obama is 3 points ahead of Hillary Clinton:
Most people who took the under cited the principle of "regression to the mean". And this is an underrated dynamic in analyzing polls. Ninety-five percent of polls, theoretically, fall within the margin of error established by the pollster. The flip side to that is that 5 percent don't.
Five percent -- one out of 20 polls -- doesn't sound like a lot. But consider how many polls come across the wire in a given week. Rasmussen and Gallup each release three tracking polls per day, seven days per week: an Obama-Clinton primary tracker, and general election trackers for both Obama-McCain and Clinton-McCain. That's (7 x 3 x 2 =) 42 polling results right there. Beyond that, we're averaging about 15 new state-by-state general election polls each week. Those polls are surveying two matchups each (Clinton-McCain and Obama-McCain), so that's another 30 contests. On top of that, we might have another 10 or 15 primary election polls in states like Pennsylvania or Indiana, and two or three of the "name brand" national pollsters like CBS/NYT or Pew will weigh in each week with their periodic updates.
So in a given week, we're probably looking at somewhere between 75 and 100 polling results. (By my count, there have been 86 different Presidential election-related polls released publicly since last Monday). Not only might some of these polls be "wrong" (that is, fall outside their margin of error). Some of them will be wrong. It's a mathematical certainty. We should expect to see four or five "bad" polls each week. And that's assuming that the pollsters are doing everything perfectly, which they aren't (not by a long shot).
Of course, it can be a dangerous game to try and guess which polls are the outliers and which ones aren't. A lot of people believed that the Selzer/DMR New Years' Eve Iowa poll was an outlier -- but it turned out to be just about the only one that predicted the Iowa caucus results correctly.
Before Iowa, however, the behavior of the polls was quite volatile. The dynamics of the campaign were changing on a day-to-day basis. Moreover, the pollsters were forced to poll during the Christmas/New Years Holiday, a period they've generally avoided, and nobody had any idea how to model turnout, with no voting yet having been conducted.
The polls remained quite volatile up through and including Super Tuesday. At some point in the middle of February, however, the polls fell into something of a steady state. Obama took the lead from Clinton in the Real Clear Politics average on roughly February 10th. He has held on to that lead ever since. But he has never held a particularly large lead. It's essentially always been somewhere between 1 point, and 6 or 7 points -- sometimes toward the top of that range, and sometimes toward the bottom, but always between those goalposts. This becomes even more clear when looking at the Pollster.com averages, which tends to smooth out some of the day-to-day fluctuations in the polls.
Pollster.com has Obama's numbers moving up steadily up through about the 15th of February. And then they plateau. They don't peak -- they don't go down any -- they just plateau. The same is true for Hillary Clinton's numbers. In fact, Clinton has been stuck at about the same level of support, somewhere in the 42-46% range, since Labor Day.
Now, it's a truism that primary polls are generally more volatile than general election polls. In fact, they are often much more volatile. If we tried to apply the fivethirtyeight.com model, which is designed to predict general election outcomes, to predict primary results instead, it might well lead us in completely the wrong direction (among other things, we'd need to place much more of a premium on recentness if we were trying to evaluate the primaries).
But there are specific reasons why primary polls tend to volatile -- and I would argue that those reasons no longer really apply in the Democratic nomination race:
1. In the primaries, voters generally have weak preferences between the two candidates. After all, you are asking voters to choose from among two or more candidates in their own party. Many voters will have a favorable impression of all of these candidates. That means, however, that it doesn't take much to move a voter from one candidate to the other.
2. In the primaries, voters generally have limited information about the candidates. Presidential candidates usually come from the ranks of the nation's 150 U.S. senators and governors. Even relatively well-informed voters aren't keeping tabs on most of these elected officials; in fact, they'd probably have trouble naming more than 20 or 30 of them. So a significant element of the primary process is simply voters becoming better informed about the candidates and their positions, and the polls can be quite volatile until that process is complete.
Neither of these axioms really hold any longer for Senators Clinton and Obama. Each candidate has nearly 100% name recognition (this has been true for a long time for Hillary, of course). They have been campaigning for more than a year. They have poured a collective $250 million dollars into the campaign, and probably gotten well upwards of a billion dollars apiece in terms of free media exposure. Voters know Clinton and Obama inside and out. They aren't really learning anything new about them -- and for that matter, Clinton and Obama aren't even really bothering to talk about policy these days (instead, all of the discussion is about the metaphysics of the nomination process).
Moreover, preferences between the two candidates have hardened. Both candidates have experienced some recent deterioration in their favorability/unfavorability ratings (see charts below). Although we don't necessarily know how much of this is Democrats turning against their own candidates, versus independents and Republicans being turned off, the general anecdotal sentiment I pick up among Democrats these days is "Don't you get it? We've made up our minds. We're just waiting for this to end". People are bored, and people don't change their votes when they're bored.
This is not to say that there is never any day-to-day movement in the polls. Clearly, for instance, the Jeremiah Wright controversy depressed Obama's numbers for a period of 7-10 days. But the polls fell back into equilibrium pretty quickly. And other events, like Hillary's big victory in Ohio on March 4 -- or, for that matter, Barack Obama's win in Wisconsin -- didn't seem to have moved the polls at all.
But what about Obama's recent upward movement in North Carolina and Pennsylvania? Doesn't he have momentum there?
Well, his numbers have certainly improved in those states. (Although, it partly depends on where you start measuring from; Obama had closed to within 4 points of Clinton in a Rasmussen poll of Pennsylvania on 2/26.) But I don't know that I'd really call it momentum, which I tend to think of as a sequential process in which one result favorably affects the next. It's not like Obama's moving up in Pennsylvania because he won Wyoming three weeks ago.
Instead, what I think we're seeing in Pennsylvania and North Carolina is something different: organization and ground game. Obama's numbers have moved upward in the last few weeks before the primary in essentially every single state that he's campaigned in. But this movement has been local rather than national (otherwise he really would have wrapped the nomination up by now). It seems to me that Obama has some systematic advantages on the ground, including some combination of the following:
1. Higher density of advertising, and/or more effective advertising.
2. Higher concentration of field offices and/or superior outreach efforts and/or greater enthusiasm from paid and unpaid volunteers.
3. Better attended and/or better staged and/or more persuasive campaign events.
4. Superior handling and/or more favorable treatment from local media (consider his substantial advantage in newspaper endorsements).
5. Possibly, superior data mining and microtargeting (this last one is pure speculation on my part, as it's not something a campaign is ever going to want to talk about).
This is not to say that Sen. Clinton is a poor retail campaigner. On the contrary, I think she's an above-average one. She seems frankly to have more endurance than Obama, and her campaign has shown an ability to dictate the course of key media cycles, even if they do not necessarily win them. These assets may be particuarly valuable during the last 72 hours of the campaign. But, based on the almost predictable way in which Obama seems to improve his polling numbers in the T-minus 7-to-30 days period, as we're seeing in Pennsylvania now, we have to give some deference to Obama's ground game.
...
The interesting thing is that, even though the national polls haven't changed, the futures markets have.
Clinton's chances of winning the nomination are now trading at about 13%, which, incidentally, is just about where she was prior to the Texas and Ohio primaries. After shooting up to around 30% following her victories in those states, her stock has had a fairly linear progression back to that 13% level (interrupted by a couple of 'shocks': the market reacted strongly to the Jeremiah Wright story and the Bill Richardson endorsement).
There are, I think, three things that account for this. Firstly, the traders in these futures markets, many of whom are based offshore, are not all that insightful amount American politics. They tend to react too strongly to the media narrative of the day, and not strongly enough to the more glacial forces that really win and lose campaigns. (For example, the market moved a lot after a Clinton adviser told Politico she had no more than a 10 percent chance to win, a non-event that did not affect the fundamentals of the race at all). Secondly, there was a lot of "hidden" bad news for Clinton during this period, such as the continued decline in her superdelegate advantage, and the failure to secure revotes in Florida and Michigan.
However, there is something else going on too. In order for her to have a chance at winning the nomination, Clinton needs there to be volatility in public sentiment. Or, if you prefer, she needs there to be momentum. To get more specific about it, she needs the Pennsylvania result to impact the North Carolina and Indiana results. Otherwise, what will happen is something like this: Clinton will win Pennsylvania by about 6 points, then she will lose North Carolina by about 16 points, but win Indiana by about 7 points (these are the present polling averages in those states). By my math, that would result in a net swing of 3-4 delegates, and 25K-50K popular votes, to Obama between now and May 6th. Although the rest of the schedule would be nominally favorable to Clinton -- I show her picking up a net of 15-30 delegates after May 6th on the strength of big wins in West Virginia, Kentucky, and possibly Puerto Rico -- at that point the campaign would probably be over.
So the fact that the national polls have been stubbornly unmoving is almost as bad news for Clinton as if the polls were breaking away from her. At least if the polls were moving away from her -- that would indicate a volatile electorate. But instead, it's as though we're playing out the last couple of tricks in a euchre hand after the point has already been won. A lot of the value in the Clinton stock, back when it was trading at 30%, was in the possibility that something dramatic would happen. Every day that something doesn't happen is bad news for Clinton. By the way, if the Clinton campaign is reading these numbers the same way I am -- I would look for them to make news this week.
I was sort of trying to goad people into taking the over. After all, we're all used to hearing about the importance of "momentum" and trendlines. Obama had gone, over the span of about 10 days, from being 7 points behind, to roughly tied, to 4 points ahead, to 10 points ahead. Wasn't it logical that he'd be some number larger than 10 points ahead after another week's worth of "momentum"? Wasn't the headline that Obama was pulling away with the race?
But most people did not take the bait, and said they expected Obama to be fewer than 10 points ahead. Which, as it turns out, was the right answer. In today's Gallup Tracker, Obama is 3 points ahead of Hillary Clinton:
Most people who took the under cited the principle of "regression to the mean". And this is an underrated dynamic in analyzing polls. Ninety-five percent of polls, theoretically, fall within the margin of error established by the pollster. The flip side to that is that 5 percent don't.
Five percent -- one out of 20 polls -- doesn't sound like a lot. But consider how many polls come across the wire in a given week. Rasmussen and Gallup each release three tracking polls per day, seven days per week: an Obama-Clinton primary tracker, and general election trackers for both Obama-McCain and Clinton-McCain. That's (7 x 3 x 2 =) 42 polling results right there. Beyond that, we're averaging about 15 new state-by-state general election polls each week. Those polls are surveying two matchups each (Clinton-McCain and Obama-McCain), so that's another 30 contests. On top of that, we might have another 10 or 15 primary election polls in states like Pennsylvania or Indiana, and two or three of the "name brand" national pollsters like CBS/NYT or Pew will weigh in each week with their periodic updates.
So in a given week, we're probably looking at somewhere between 75 and 100 polling results. (By my count, there have been 86 different Presidential election-related polls released publicly since last Monday). Not only might some of these polls be "wrong" (that is, fall outside their margin of error). Some of them will be wrong. It's a mathematical certainty. We should expect to see four or five "bad" polls each week. And that's assuming that the pollsters are doing everything perfectly, which they aren't (not by a long shot).
Of course, it can be a dangerous game to try and guess which polls are the outliers and which ones aren't. A lot of people believed that the Selzer/DMR New Years' Eve Iowa poll was an outlier -- but it turned out to be just about the only one that predicted the Iowa caucus results correctly.
Before Iowa, however, the behavior of the polls was quite volatile. The dynamics of the campaign were changing on a day-to-day basis. Moreover, the pollsters were forced to poll during the Christmas/New Years Holiday, a period they've generally avoided, and nobody had any idea how to model turnout, with no voting yet having been conducted.
The polls remained quite volatile up through and including Super Tuesday. At some point in the middle of February, however, the polls fell into something of a steady state. Obama took the lead from Clinton in the Real Clear Politics average on roughly February 10th. He has held on to that lead ever since. But he has never held a particularly large lead. It's essentially always been somewhere between 1 point, and 6 or 7 points -- sometimes toward the top of that range, and sometimes toward the bottom, but always between those goalposts. This becomes even more clear when looking at the Pollster.com averages, which tends to smooth out some of the day-to-day fluctuations in the polls.
Pollster.com has Obama's numbers moving up steadily up through about the 15th of February. And then they plateau. They don't peak -- they don't go down any -- they just plateau. The same is true for Hillary Clinton's numbers. In fact, Clinton has been stuck at about the same level of support, somewhere in the 42-46% range, since Labor Day.
Now, it's a truism that primary polls are generally more volatile than general election polls. In fact, they are often much more volatile. If we tried to apply the fivethirtyeight.com model, which is designed to predict general election outcomes, to predict primary results instead, it might well lead us in completely the wrong direction (among other things, we'd need to place much more of a premium on recentness if we were trying to evaluate the primaries).
But there are specific reasons why primary polls tend to volatile -- and I would argue that those reasons no longer really apply in the Democratic nomination race:
1. In the primaries, voters generally have weak preferences between the two candidates. After all, you are asking voters to choose from among two or more candidates in their own party. Many voters will have a favorable impression of all of these candidates. That means, however, that it doesn't take much to move a voter from one candidate to the other.
2. In the primaries, voters generally have limited information about the candidates. Presidential candidates usually come from the ranks of the nation's 150 U.S. senators and governors. Even relatively well-informed voters aren't keeping tabs on most of these elected officials; in fact, they'd probably have trouble naming more than 20 or 30 of them. So a significant element of the primary process is simply voters becoming better informed about the candidates and their positions, and the polls can be quite volatile until that process is complete.
Neither of these axioms really hold any longer for Senators Clinton and Obama. Each candidate has nearly 100% name recognition (this has been true for a long time for Hillary, of course). They have been campaigning for more than a year. They have poured a collective $250 million dollars into the campaign, and probably gotten well upwards of a billion dollars apiece in terms of free media exposure. Voters know Clinton and Obama inside and out. They aren't really learning anything new about them -- and for that matter, Clinton and Obama aren't even really bothering to talk about policy these days (instead, all of the discussion is about the metaphysics of the nomination process).
Moreover, preferences between the two candidates have hardened. Both candidates have experienced some recent deterioration in their favorability/unfavorability ratings (see charts below). Although we don't necessarily know how much of this is Democrats turning against their own candidates, versus independents and Republicans being turned off, the general anecdotal sentiment I pick up among Democrats these days is "Don't you get it? We've made up our minds. We're just waiting for this to end". People are bored, and people don't change their votes when they're bored.
This is not to say that there is never any day-to-day movement in the polls. Clearly, for instance, the Jeremiah Wright controversy depressed Obama's numbers for a period of 7-10 days. But the polls fell back into equilibrium pretty quickly. And other events, like Hillary's big victory in Ohio on March 4 -- or, for that matter, Barack Obama's win in Wisconsin -- didn't seem to have moved the polls at all.
But what about Obama's recent upward movement in North Carolina and Pennsylvania? Doesn't he have momentum there?
Well, his numbers have certainly improved in those states. (Although, it partly depends on where you start measuring from; Obama had closed to within 4 points of Clinton in a Rasmussen poll of Pennsylvania on 2/26.) But I don't know that I'd really call it momentum, which I tend to think of as a sequential process in which one result favorably affects the next. It's not like Obama's moving up in Pennsylvania because he won Wyoming three weeks ago.
Instead, what I think we're seeing in Pennsylvania and North Carolina is something different: organization and ground game. Obama's numbers have moved upward in the last few weeks before the primary in essentially every single state that he's campaigned in. But this movement has been local rather than national (otherwise he really would have wrapped the nomination up by now). It seems to me that Obama has some systematic advantages on the ground, including some combination of the following:
1. Higher density of advertising, and/or more effective advertising.
2. Higher concentration of field offices and/or superior outreach efforts and/or greater enthusiasm from paid and unpaid volunteers.
3. Better attended and/or better staged and/or more persuasive campaign events.
4. Superior handling and/or more favorable treatment from local media (consider his substantial advantage in newspaper endorsements).
5. Possibly, superior data mining and microtargeting (this last one is pure speculation on my part, as it's not something a campaign is ever going to want to talk about).
This is not to say that Sen. Clinton is a poor retail campaigner. On the contrary, I think she's an above-average one. She seems frankly to have more endurance than Obama, and her campaign has shown an ability to dictate the course of key media cycles, even if they do not necessarily win them. These assets may be particuarly valuable during the last 72 hours of the campaign. But, based on the almost predictable way in which Obama seems to improve his polling numbers in the T-minus 7-to-30 days period, as we're seeing in Pennsylvania now, we have to give some deference to Obama's ground game.
...
The interesting thing is that, even though the national polls haven't changed, the futures markets have.
Clinton's chances of winning the nomination are now trading at about 13%, which, incidentally, is just about where she was prior to the Texas and Ohio primaries. After shooting up to around 30% following her victories in those states, her stock has had a fairly linear progression back to that 13% level (interrupted by a couple of 'shocks': the market reacted strongly to the Jeremiah Wright story and the Bill Richardson endorsement).
There are, I think, three things that account for this. Firstly, the traders in these futures markets, many of whom are based offshore, are not all that insightful amount American politics. They tend to react too strongly to the media narrative of the day, and not strongly enough to the more glacial forces that really win and lose campaigns. (For example, the market moved a lot after a Clinton adviser told Politico she had no more than a 10 percent chance to win, a non-event that did not affect the fundamentals of the race at all). Secondly, there was a lot of "hidden" bad news for Clinton during this period, such as the continued decline in her superdelegate advantage, and the failure to secure revotes in Florida and Michigan.
However, there is something else going on too. In order for her to have a chance at winning the nomination, Clinton needs there to be volatility in public sentiment. Or, if you prefer, she needs there to be momentum. To get more specific about it, she needs the Pennsylvania result to impact the North Carolina and Indiana results. Otherwise, what will happen is something like this: Clinton will win Pennsylvania by about 6 points, then she will lose North Carolina by about 16 points, but win Indiana by about 7 points (these are the present polling averages in those states). By my math, that would result in a net swing of 3-4 delegates, and 25K-50K popular votes, to Obama between now and May 6th. Although the rest of the schedule would be nominally favorable to Clinton -- I show her picking up a net of 15-30 delegates after May 6th on the strength of big wins in West Virginia, Kentucky, and possibly Puerto Rico -- at that point the campaign would probably be over.
So the fact that the national polls have been stubbornly unmoving is almost as bad news for Clinton as if the polls were breaking away from her. At least if the polls were moving away from her -- that would indicate a volatile electorate. But instead, it's as though we're playing out the last couple of tricks in a euchre hand after the point has already been won. A lot of the value in the Clinton stock, back when it was trading at 30%, was in the possibility that something dramatic would happen. Every day that something doesn't happen is bad news for Clinton. By the way, if the Clinton campaign is reading these numbers the same way I am -- I would look for them to make news this week.
4.06.2008
Today's Polls, 4/6
@ 12:47 PM
Rasmussen has a couple of new polls out this morning that remind each Democrat of a couple of their biggest weaknesses in the general election.
In Iowa, Barack Obama leads John McCain by 4 points, but McCain leads Hillary Clinton by 15 points. Iowa has consistently featured just about the largest Obama-Clinton performance gap of any state not subject to home state effects. In fact, it has the largest gap of any state if we use win percentage as the metric; Obama is presently at 86% to win Iowa, and Hillary just 15%:
Meanwhile, in Tennessee, Rasmussen has McCain leading Clinton by 14, and Obama by 27. Obama continues to perform brutally poorly in the Appalachian region of the country, both in the primaries and in general election polls. However, for Clinton the news is not good either; Tennessee was a state that we'd regarded as potentially competitive for her.
In Iowa, Barack Obama leads John McCain by 4 points, but McCain leads Hillary Clinton by 15 points. Iowa has consistently featured just about the largest Obama-Clinton performance gap of any state not subject to home state effects. In fact, it has the largest gap of any state if we use win percentage as the metric; Obama is presently at 86% to win Iowa, and Hillary just 15%:
Iowa 71% (Obama outperforms Clinton)It seems probable to me that some of the difficulties that Clinton has had in Iowa might stem from the perception her campaign sort of threw the state's caucus process under the bus. If so, I'll bet that some of the damage done is reparable, although not without some significant investment of resources on the ground.
Arkansas 69% (Clinton outperforms Obama)
Colorado 65% (Obama outperforms Clinton)
Oregon 46% (Obama outperforms Clinton)
New Hampshire 43% (Obama outperforms Clinton)
West Virginia 42% (Clinton outperforms Obama)
Washington 41% (Obama outperforms Clinton)
Meanwhile, in Tennessee, Rasmussen has McCain leading Clinton by 14, and Obama by 27. Obama continues to perform brutally poorly in the Appalachian region of the country, both in the primaries and in general election polls. However, for Clinton the news is not good either; Tennessee was a state that we'd regarded as potentially competitive for her.
Saturday Night Nerdiness
@ 2:01 AM
Those of you with better things to do -- there's still time to avert your eyes. The following point is very technical.
By popular (?) demand, I have somewhat reduced the weight of the regression-based "poll" in determining our state-by-state polling averages. The weight is now set so as to be equal to the weighted average of the reliability ratings of polls that make up the regression. Confused yet?
Say we have three polls, one of which has a reliability rating of 1.0, the second of which of which has a rating of 0.8, and the last of which has a rating of 0.2. The regression analysis itself is calculated with these weights in mind: for example, the 1.0-rated poll will have five times more influence on the outcome of the regression than the 0.2-rated poll.
To determine the average rating of a poll in the regression, we essentially have to multiply the poll ratings by themselves, and then divide them by the sum of the ratings. For example, in the example given above, the weighted average poll rating would be calculated as follows:
The principle here is that the regression is only as good as the sum of its parts. If the polls that make up the regression are old or unreliable, the regression itself will be less reliable.
It seems to me that this is a "fair" way to determine the weight given to the regression. It is not necessarily the optimal way from the standpoint of predicting the outcome of the election. Quite frankly, I suspect that if anything, we'd wind up with better predictions if we gave the regression more, rather than less, weight. But, since I haven't yet done the empirical work to back up this claim, the fairness concern prevails.
By popular (?) demand, I have somewhat reduced the weight of the regression-based "poll" in determining our state-by-state polling averages. The weight is now set so as to be equal to the weighted average of the reliability ratings of polls that make up the regression. Confused yet?
Say we have three polls, one of which has a reliability rating of 1.0, the second of which of which has a rating of 0.8, and the last of which has a rating of 0.2. The regression analysis itself is calculated with these weights in mind: for example, the 1.0-rated poll will have five times more influence on the outcome of the regression than the 0.2-rated poll.
To determine the average rating of a poll in the regression, we essentially have to multiply the poll ratings by themselves, and then divide them by the sum of the ratings. For example, in the example given above, the weighted average poll rating would be calculated as follows:
1.0 x 1.0 = 1.0 +In this example, the weighted average poll rating would be .84. As it turns out, the weighted average rating of all the polls that make up the regression is presently .77, and so this is the weight you'll see applied to the regression in determining the state-by-state averages (this number may bob upward or downward slightly over time).
0.8 x 0.8 = 0.64 +
0.2 x 0.2 = 0.04
============
= 1.68 / (1.0 + 0.8 + 0.2) =
= 1.68 / 2.0 = .84
The principle here is that the regression is only as good as the sum of its parts. If the polls that make up the regression are old or unreliable, the regression itself will be less reliable.
It seems to me that this is a "fair" way to determine the weight given to the regression. It is not necessarily the optimal way from the standpoint of predicting the outcome of the election. Quite frankly, I suspect that if anything, we'd wind up with better predictions if we gave the regression more, rather than less, weight. But, since I haven't yet done the empirical work to back up this claim, the fairness concern prevails.
Subscribe to:
Posts (Atom)
