May 9, 2008
Wha’ Happen, Part Deux
So the question du jour is, “what went wrong” on Tuesday for the Clinton campaign? Why did she underperform?
North Carolina
You will recall that a couple of weeks ago, I sought to perform a regression analysis that would explain the North Carolina results. I came up with a good one. I found that by looking at all the counties in the South, an increase in African American (”AA”) population in a given county meant a .83 point increase in Obama’s performance. An increase of 1% in the college educated percentage meant that Obama’s performance would increase by .94% (plus a constant of 17.5).
Across the South, that explained 79% of the county-to-county variance in Obama’s performance.
It predicted that Obama’s map should look like this:
The bluest counties are the counties where Clinton was expected to get more than 65% of the vote. The greenest are the counties where Obama was expected to get more than 65% of the vote.
Here’s what it looked like in reality:
As you can see, the match is pretty good. Clinton overperformed in counties in Eastern North Carolina, especially toward the Outer Banks. Obama overperformed in Buncombe county, where the college education/AA screen wouldn’t necessarily pick up the high concentration of hippies in Asheville. He also overperformed in the counties in between Charlotte and Chapel Hill. I’m not sure about the demographics of these counties, but I thought it was interesting.
At any rate, North Carolina is more like the South than the rest of the South. The college-educated coefficient remained at .94, meaning that in North Carolina, a 1% increase in college-educated dwellers in a county meant that Obama would get an extra .94% of the vote. The AA coefficient was down, but only slightly — .74. This may represent an increased racial polarization in highly AA counties. The coefficient was about the same – 16.3%. But the r-square increased to .87, meaning that almost all of the county-to-county variance can be explained solely by race and college education. Which is stunning, when you think about it.
Indiana
For Indiana, things were more difficult. In Ohio and Pennsylvania, I had to add a variable for Bush performance to improve the r-square. It still barely explained 70% of the variance, and predicted a 14% win for Hillary.
But the constant wasn’t significant, meaning that while the model itself could be correct for how different variables affected Obama’s performance, we were unsure where the base Obama vote was.
Ultimately, that’s what happened here.
The internal regression for Indiana against Obama’s results ended up giving identical coefficients to the Ohio/PA coefficients. It was the base vote that was off. You can see this in the map above. The areas where Obama was expected to be mediocre ended up being strong for Obama. Some of the areas where he was expected to be weak, but not awful, ended up being mediocre Obama counties. He especially overperformed in the small and medium cities, such as Ft. Wayne, South Bend, and Evansville.
I tried a variety of variables to improve the performance, and just couldn’t. I tried rural%, unemployment, working class, some college education, tried putting in a dummy variable for appalachia into the PA/OH numbers, and various other theories, and came up with nothing.
Nor does proximity to Illinois explain it. Obama only did a few points better in Lake County than the model expected. He did substantially better in places like Allen County and Vandenburgh County, which are on the other side of the state from Chicago.
In the end, I think the difference is the difference between Southeast Ohio and Northwest Ohio. Both are demographically similar, with lots of working class whites, and few minorities. But Northwest Ohio was substantially more receptive to Obama. That seems to have occurred in Indiana as well.
Going Forward
So I think the answer to what went wrong lies in a rare miscalculation by the campaign. They began to believe their own press releases, and thought they had a shot in NC and would do well in Indiana. As such, they mismanaged expectations. The press began to turn the expectations into a narrow Clinton win in NC and a blowout in IN.
The Clinton campaign should have gone off of what I predicted a few months ago, here. Even with a 12-point loss in NC, and a narrow win in Indiana, she could still win the popular vote. That should have been the pitch. That a rough patch was coming, but that she could still prevail.
In reality, the results in IN and NC represent something of a comeback by Clinton. You will recall that it was not that long ago that she was trailing by 20 in NC polls and by 5 in Indiana. She should have stuck to that story, rather than allowing expectations to be played up. THAT is what doomed her, not the loss in NC and near-loss in IN.
So we’re left with this:
Her two best states are coming up in KY and WV. Had she played the expectations game better, she could be set up for a true comeback next week. But as it stands, it will be very, very difficult to dig out of this hole (but probably not impossible).
And with that, I think maybe I’ll go roll myself a dwarven fighter/mage. If I roll double-zeroes twice, maybe I’ll get psionic abilities!!!!
http://race42008.com/2008/05/09/wha-happen-part-deux/trackback/
May 8th, 2008 at 8:46 pm
You meant KY not IN at the end.
May 8th, 2008 at 8:55 pm
“What a difference a day makes . . . Twenty-four little hours . . ”
Obama was really on the ropes before Tuesday. He was physically and mentally exhausted and the media was ready to write him off (that is indeed how fickle the media is). But after a big win in NC and a pretty good performance in Indiana given predictions, he was “Barack-Star” today. I live in DC and there was a buzz about him being in town and that buzz turned into near pandemonium when he showed up on Capitol Hill.
He also gave two good interviews, on CNN and NBC. In both he had regained his sure footing. Superdelegates watching him today are probably confident that he has recovered fully. Frankly, both Clinton and McCain seem almost secondary as Obama commands the stage.
May 8th, 2008 at 8:56 pm
A poll out of WV today had Hillary up 40!
I think after PR, she’ll ehave the lead in the popular vote. On Chris Matthews today, Howard Wolfson, the main pr guy, emphatically stated that they will count PR in the popular vote. Matthews was aghast. But the Clintons know she stands to net at least 400K from PR. I think she nets at least 300K from WV/KY. That’ll be enough to give her the lead in the popular vote.
That + FL/MI will be enough to take her through to August. The Supers are a bunch of wimps. If they haven’t come out for Obama by now, you think they will after the 20th after he’s just lost WV and KY by 30? After he falls behind in the popular vote after Puerto Rico.
Hillary will take this all the way to Denver, wher I predict she’ll lose on the 1st ballot, like Reagan in ‘76.
May 8th, 2008 at 9:26 pm
Curious, Sean.
Do you roll with actual dice? Or a computer program?
May 8th, 2008 at 9:34 pm
Mad props to the one who can link up to the McCain-O’Reilly video that ran tonight. My money is on LJ to deliver it first.
May 8th, 2008 at 9:45 pm
A dwarven mage? Is that allowed now?
May 8th, 2008 at 10:01 pm
#6.
I picked up on that, too. I’d go for it.
May 8th, 2008 at 10:07 pm
Look you just stuffed up.
No big deal.
Popular vote matters NAUGHT btw!
May 8th, 2008 at 10:09 pm
Hillary has weakened her own legitimate argument by overreaching on Florida and Michigan; thereby giving an already hostile media ample ammunition to dismiss her otherwise valid claim as frivolous. Instead of citing the RNC’s punishment of seating half the states’ delegates as precedent and calling for a new and fair primary in Michigan — one in which Obama is on the ballot — she’s making the absurd assertion that she has “received more votes by the people who have voted than anyone else.” Equally a stretch, Clinton now contends that the real finish line is 2209 delegates; not 2025. That new figure includes not half, but all of both Florida’s and Michigan’s delegates. Yes, Virginia, the woman has balls…
May 8th, 2008 at 10:29 pm
I agree Aron, but there are two points to add to your argument. One is that even including MI and FL Obama still leads in the popular vote, just by a slimmer margin. Two, its worth repeating that Clinton like Obama said MI and FL would not count because they went out of turn. She tries to claim its about the voters, but what about her credibility?
Regarding the Obama plan to declare victory on May 20, its actually an astute move because it comes before the DNC meets at the end of the month and before the final contests. It is predicted that on May 20 he will clinch a majority of delegates. Thus, Clinton could win 100% of the remaining delegates and still be behind. That means that those superdelegates who have stated their intention to support the delegate leader will come out in support of Obama. Pelosi has also said the delegate leader should be the nominee so its possible she will also announce her support. Thus, any idea of Clinton going to the convention will end and at the end of the month the party can feel free to fully seat the MI and FL delegations.
May 8th, 2008 at 10:53 pm
He won’t lead in the popular vote after WV, KY and PR vote.
You think Bill and Hillary give 2 fs about what Nancy Pelosi does?
As for Obama declaring victory on the 20th, I’d love to see it. I can imagine the headlines:
Obama gets shellacked in KY, loses by 35, declares victory
May 8th, 2008 at 11:02 pm
I’m pretty sure dwarfts can’t be dual-classed, either.
May 8th, 2008 at 11:38 pm
the mccain-oreilly interview:
http://www.youtube.com/watch?v=JyZumcv78O8
http://www.youtube.com/watch?v=EsyY0SU0WIQ
May 8th, 2008 at 11:43 pm
Sean,
By the way…good work! Your map has caught the eye of Jay Cost.
Not Quite Yet
http://www.realclearpolitics.com/printpage/?url=http://www.realclearpolitics.com/horseraceblog/2008/05/not_quite_yet_1.html
May 8th, 2008 at 11:59 pm
You all need to convert to 3rd edition. It’s where the cool geeks are now.
May 9th, 2008 at 12:09 am
I can’t wait to see all the media types on the beach in San Juan as Bill goes to every Barrio on the Island.
She nets at least 400K votes.
Chris Matthews and Tim Russert in Hawaiian shirts and shorts as Hillary stuns the world by taking Obama’s popular vote lead and sending us to Denver in a stalemate.
May 9th, 2008 at 12:11 am
You folks do realize that Dem primary voters may decide the race needs to be over for the sake of the party, and start leaning toward Obama rather than Clinton. I think we have seen some of that on Tuesday already.
May 9th, 2008 at 12:21 am
the mccain-oreilly interview:
http://www.youtube.com/watch?v=JyZumcv78O8
http://www.youtube.com/watch?v=EsyY0SU0WIQ
May 9th, 2008 at 12:39 am
In reviewing the internals from the past two SurveyUSA polls for Kentucky, I just noticed that Hillary’s lead in Louisville (30% of KY’s electorate) has more than doubled in the past week. What was a 49-42 percent, seven point Clinton lead on April 29 is now a 15-point, 54-39% advantage over Obama in the Louisville metropolitan area.
If, and when Hillary wins Kentucky by 35, rest assured Russert and Olbermann will say Clinton still couldn’t cover the six touchdown spread.
May 9th, 2008 at 4:45 am
A dwarf with psionic abilities. May it never be!! Gnomes perhaps.
May 9th, 2008 at 5:21 am
Very true Metro in 17.
And if the pv didn’t matter for us in 2000 why would it matter now for the Dems when it’s about DELEGATES?
May 9th, 2008 at 6:23 am
Dwarves used to be allowed to be fighter/thieves. And way back in the 1st edition I think they could also be fighter/assassins.
May 9th, 2008 at 7:59 am
You folks do realize that Dem primary voters may decide the race needs to be over for the sake of the party, and start leaning toward Obama rather than Clinton. I think we have seen some of that on Tuesday already.
Yes.
I wonder how much of Clinton’s WV lead will be cut into.
Will they try to save her? Or will they abandon her for the sake of unity?
If it’s even a 50/50 split, that seriously hurts her.
Although every single Clinton supporter I know (and I know a lot) wants to save her and are furious at Obama. They’re in utter denial about Hillary’s chances.
We’ll see…
May 9th, 2008 at 8:35 am
12. I’m pretty sure dwarfts can’t be dual-classed, either.
In 2nd ed, dual-classed, no. That was reserved for humans (& 1/2 elves, I think). Multi-classed, yes, but not the fighter/mage combo. I won’t go into the distinction. I don’t go as far back as Indy, however, so I can’t speak to 1st ed. Your psionics, however, would be pretty good with that roll!
15. You all need to convert to 3rd edition. It’s where the cool geeks are now.
MiddleSnu, I think it’s 3.5 now, with 4th scheduled to come out soon/already has come out. Can’t speak to either one.
Ok, now that the geekfest is out of my system…
Looking at that map and making assumptions, that’s an awful area for Sen Obama, and is exactly why Sen Clinton isn’t getting out (no news there). I don’t think any of these will matter in the end, however. The convention will come, Sen Clinton will try to seat FL and MI, and probably will get some of their delegates seated (again, no news). The supers are going to break for Sen Obama based on two factors: 1) They believe Sen Obama can win over Sen Clinton’s supporters easier than she could win over his (as she’d be seen as stealing the nomination from him if she got it), and 2) they don’t understand why his baggage might be viewed as worse than hers, and he’s got more charisma.
Either way, though, Dems are in deep trouble early in this one. Either choice, at this point, has lost such a significant part of the base that they need to take a considerable portion of the independents, and my humble opinion is that Sen McCain just won’t lose the middle to either one. He’ll lose some off the right for things like immigration and his disdain for parts of the base that disagree with him, but not as much as either Dem candidate will lose.
May 9th, 2008 at 8:59 am
17. Possibly. I don’t think we saw much evidence of it on Tuesday (except maybe in small-to-medium cities in IN), but we may see it in WV on Tuesday.
21. Because its only about delegates up to the point of the pledged delegates. The question is how the Superdelegates should vote. In order to lay a legit claim to the nomination, she has to give the supers some type of lead in some type of vote count. That’s why the popular vote is important to her. Plus, arguing Bush’s p.o.v. is not really what Obama wants to do in a Dem primary.
May 10th, 2008 at 6:26 pm
Wow. I totally forgot the distinction between dual-classed and multi-classed characters.