May 5, 2008
Previewing Indiana
In previous posts here I made county-by-county predictions for Barack Obama and Hillary Clinton’s performance in North Carolina. I found that using race and college education as my two independent variables, I was able to explain 80% of the variance in the vote between Hillary and Obama in states across the South. I applied this to North Carolina, and estimated a 55-45 Obama win.
But what about the other May 6 state, Indiana?
Indiana is more difficult. First, I don’t think the Southern model will work. Not only are there important historical and cultural differences between the North and South that make this difficult to predict, there is also a much smaller county-to-county variance in the black percentage in the North. This may be problematic in many ways, not the least of which is that the white vote may behave differently in a state without a racial history similar to that found in the South. We have seen before that in homogenously white states, the white vote is more favorable to Obama. Wisconsin is a case in point. So we probably need to model things differently.
The problem, though, is there aren’t many good states to use for data here. Almost all the plains states used caucuses. New England is politically and culturally different than the industrial midwest. That leaves PA, OH, MI, WI, and IL. IL is out for obvious reasons, and the MI count is problematic. WI has just been a consistent outlier this entire election season; we haven’t seen it replicated anywhere.
So my dataset is PA and OH counties. This is actually not so bad, as these are two of the most recent caucuses, so they may give a good snapshot of how Northern voters are thinking now. Of course, IN is different than these states; it projects further South (and Southern IN is more culturally “Southern” than Southwest PA or Southeast OH), and is more Republican. More on that later.
I also ran the regression on a congressional district level, just to doublecheck myself.
Using just the AA and college educated vote, our results are less robust than in the South. Remember, in South, these two variables alone explained 80% of the county-by-county difference. Here, using the % college educated and % AA only explains 55% of the county-to-county variance.
Looking at it on the congressional district level, the model performs much, much better. It explains 80% of the variance. The variables are significant, and perform as we would expect: As AA and black percentage increase, so does Obama’s vote share. As college educated voters increase, so does Obama’s vote share.
So this begs the question: How can we strengthen the county results? Let’s start by thinking about the difference between North and South. Interestingly, AA population is worth less in the North than in the South. But we know from exit polls that AAs perform roughly the same way. How is this possible? One explanation is that we also know that there are fewer blacks in these states than in the South, and almost no counties that reach 80-90% AA, as there were in the South. So we explain less of the total variance with that variable.
We also know from looking at our initial regression that college education isn’t as good at explaining the difference among whites as it is in the South. There is still a strong relationship, but it isn’t the stark difference we saw in the South.
My theory for the remaining variance is that in counties where Bush performed the best, Obama’s vote share should increase. This is in part because of his strong performance among Republicans to date. But it is also because, in those counties, the Republican party will be moderated, in part because of its size. The theory is that in heavily Republican counties, the white Democrats will tend to be more liberal. This is something of an analogue of what we see in California, where moderates have largely aligned with the Democrats, leaving a rump Republican party of hardcore conservatives. We have some support for this; Jay Cost has noted that Obama overperformed in central Pennsylvania; this is also the most Republican section of the state. You can also note than in Ohio, Obama performed better in Northwest Ohio than in Southeast Ohio, even though the two areas are not demographically dissimilar.
Why would this be different in the North? Because Presidential voting is a better predictor of local voting there than in the South. Heavily Bush counties don’t correlate with heavily Republican counties in the South. This is why, even though Republicans are likely going to get clubbed today in a special election in Louisiana [yes I wrote this Saturday], not too much should be read into it. Part of it is the trouble with the Republican brand, but part of it is that there are still many voters in the South who will vote Democrat except for President. GOP losses in R+6 MS-04 in 1998, or in R+9 LA-05 in 2000 (or in LA-03 in 2004) did not spell the beginning of the end of the Republican brand; it just meant that re-alignment was incomplete in the South.
Running the regression with Bush’s % of the vote in 2004 improves the strength of the model. Running this on the Congressional level, we explain 87% of the variance (up from 80% with just race and education). Our coefficients all behave as we would expect: As a county becomes blacker, more education, and/or more pro-Bush, Obama improves. By far the best indicator is race. Running this on the county level, we improve from 55% to 70% of the variance explained.
Results
Plugging these numbers from OH and PA into IN, Obama’s best district will be the 7th, which he should win with about 58% of the vote. He should split the Fifth District, which is heavily Republican, and where he should get a boost from the Indianapolis suburbs. He should run about 50-50 in the First, but narrowly lose. This will be interesting to watch. Many have predicted that this will be a strong district for him, because of the high AA population. I have disagreed, because there is also a high white ethnic population. If he overperforms, we may be able to conclude that there is in fact some spillover effect from Chicago media.
After that, the 3rd, 4th, and 2d, all located in the North of the state, should provide a strong performance for him. He should come in below 40% in the 6th, 8th, and 9th, which are the more Southern portions of the state.
The county-by-county results are similar. He does his best in the middle of the state, where Hamilton county is by far his best county, followed closely by Marion (Indianapolis). Other areas of strength are Lake County (Gary), Monroe County (Bloomington) and Allen county (Ft. Wayne). The South is a mass of blue.
I’ve attached maps on both the Congressional and County-by-County basis. Again, if the results on Tuesday look like this, I will trumpet it on the front page. If not, you will never hear about this again. The prediction, by the way, from the model is a 57-43 drubbing by Clinton.
http://race42008.com/2008/05/05/previewing-indiana/trackback/
May 4th, 2008 at 5:34 pm
Can’t say what I think about all that, but I think the 57-43 prediction will not come to fruition. I was born and raised a Hoosier, though no longer live there, but have most of the family still around. I just think there is too much anti-Clintonism in the state for it to be that wide of a margin. My guess is a Clinton win, but not by more than 5 to 6 percentage points. Central IN, north of Indy I think is going to be more Obama than you give credit for, particularly the Kokomo-Marion-Anderson area. Just my gut feeling.
May 4th, 2008 at 6:30 pm
I was also born a Hoosier and I think that Clinton will do very well in Northeast Indiana. Fort Wayne is a very conservative town with lots of blue-collar white Democrats. The only thing people from Allen County like about Chicago is the Cubs.
May 4th, 2008 at 6:41 pm
As usual, an impressive analysis, and I’d give this a stronger shot than your NC predictions of being correct, because Ohio and PA voted recently while all the Southern states were a while ago. There is therefore no good data about how the Wright storm will affect Southern voters in the end, which is a fairly large variable you aren’t accounting for.
May 4th, 2008 at 6:41 pm
As usual, an impressive analysis, and I’d give this a stronger shot than your NC predictions of being correct, because Ohio and PA voted recently while all the Southern states were a while ago. There is therefore no good data about how the Wright storm will affect Southern voters in the end, which is a fairly large variable you aren’t accounting for.
May 4th, 2008 at 6:43 pm
Jonathan,
I worked for Cong. Souder and I don’t think Ft. Wayne is Clinton territory. Most of the people there are Republicans, and probably won’t be voting in the primary. In my experience, the Dems there are pretty darned liberal. The outlying counties are a different story, and she should do very well there.
Steve,
My gut is that it is a bit high on the spread, but then again, Obama is at about 41% in the RCP average, and he rarely does much better.
The counties you describe only give Clinton a slight edge, and remember, many of the Clinton-haterz will be Republicans.
May 4th, 2008 at 7:29 pm
Great work again Sean. Assuming you are right and Obama loses by something close to 14 (or even only by about 10 points) - what does this do to the narrative? For some reason these Democrat superdelegates won’t budge and you keep seeing the trickle every day of another couple proclaiming allegience to Obama. It boggles the mind. I really don’t understand these Democrats. Unless I am completely out of touch (and I don’t think I’m completely out of touch) it seems that Obama is much more likely to fail in the Fall. The blacks and the college kids are going to get over it if the Democrats go with Obama. I can’t for the life of me understand why they are going with someone so demonstrably weak. The only thing I can think of is payback of these Democrats to the Clintons. Is the anger with Bill and Hil over the loss of congress so great that they would go with this wimpy affirmative-action pick of questionable patriotism?
As I said - it just boggles the mind.
May 4th, 2008 at 7:31 pm
I should say the blacks and the college kids are going to get over it if the Democrats **Don’t** go with Obama.
May 4th, 2008 at 7:59 pm
The Democrats are DUMB - they will go with the loser Obama, who is clearly the inferior candidate. They would rather give up the GE, so as not to offend the black and youth vote (which has been leaving him lately). I guess they would rather offend women, blue collar voters, seniors, Catholics and Latinos. He will be easier to defeat in the fall….go figure.
May 4th, 2008 at 8:01 pm
Adam,
Depends on what happens in NC. If he loses this big, but wins by a similar margin in NC, I don’t think the narrative changes. But if he loses IN big and only wins NC by single digits, I think he is actually in some trouble, esp. with a 40-point loss in WV coming around the bend.
Obama isn’t just an affirmative-action pick; he just peaked at the right time. It’s kind of like if Huckabee had peaked in December instead of November. By the time the media got around to examining Obama, it was too late. Plus, the playing field is so slanted against the GOP this year that a relatively weak candidate like Obama could still best the strongest possible candidate we could put up (eg McCain).
May 4th, 2008 at 8:12 pm
Maybe I am too dismissive with the affirmative action bit. I say that only because I suspect we would have never heard about Obama had he not spoken at the convention in 2004. And I have my suspicions of whether or not he would have been given that role had he been named Barry O’Connor. But that’s moot now anyway.
And you’re right - I’m cautiously optimistic abot the Fall. But I have no illusions that it’s a done deal.
What do you think about Puerto Rico? If Hil wins bif there does that change the dynamics? Or will the media just dismiss it because PR really doesn’t have any electoral sway?
May 4th, 2008 at 8:30 pm
PR depends on turnout. It’ll highlight Obama’s problems with Hispanics (who increasingly are an extension of the white working class). It might give Hill an absolute popular vote lead. I dunno. I think IN and NC are more importnat, really.
May 4th, 2008 at 9:27 pm
NYTimes poll shows that Wright was really a gift to Obama - he has a whopping 12 point margin nationwide.
It was over a month ago.
Now it is even more over.
Time for Republicans to focus on Obama.
May 4th, 2008 at 9:33 pm
Sean:
If I remember right, Tom Henry is the new mayor of Fort Wayne, and he ran as a pretty conservative democrat (his GOP rival had ethic’s problems but still). If he endorsed someone, it might tip the city and all of Allen County. I’m just surprised to here that neither Obama or Clinton have gone to Fort Wayne. It is the 2nd biggest city in Indiana after all.
May 4th, 2008 at 9:39 pm
Barack’s share of the white vote has fallen as whites have learned more about him.
Blacks don’t care about anything but his skin color apparently.
May 4th, 2008 at 11:16 pm
Morrissey, that poll is completely useless. It shows Obama +11 on McCain and Hillar +12 on McCain. All it is just a survey of adults, not even registered voters.
May 5th, 2008 at 6:15 am
Talked to my brother in northern IN and there are some interesting Republican sub-ticket primaries. Seems to be a big push, I suppose out of Fort Wayne and maybe South Bend to get some Christians on Senate and House tickets. Thus, don’t expect a large Republican crossover on the Pres primary in the north and northeast in the outlying counties. If “Operation Chaos” is alive there, it will be minimal and hurt Hillary. On the other hand in the central areas, I expect a larger crossover but not due to “Operation Chaos”. Those folk want to see Hillary gone so I suspect will cross over for Obama as I predicted earlier. Thus, I think Hillary’s margin is going to be shrunk in both cases. 5-6% win for Hillary is my prediction. Interesting how some of the sub-ticket races might influence this.
May 5th, 2008 at 6:45 am
Mark my words Obama will lose by no more than 6 and will win NC by AT LEAST 10.
We might be watching the last week of this exciting campaign.
May 5th, 2008 at 8:05 am
Adam,
The behavior of the superdelegates is perfectly understandable. The best thing that could possibly happen to the Democratic party right now would be a major Obama scandal. He took a bribe. Or colluded with Rezko in shadiness. Etc. But, consider the worst scenario. What if a large number of Democrats decide that Hillary’s the best choice? Well how large would this number have to be to theoretically tip the nomination to her? Probably around 70% of superdelegates, best case scenario. But, what if she only gets 60%? What happens to Obama? Well, he starts looking like a bigger and bigger loser, but they can’t quite shake him. That’s a disaster for the Democratic Party. Either they need to shake Obama, welly and truly, or they need to do anything possible to protect him. Since it’s extremely difficult to imagine him losing the nomination, they’ve decided to protect him. And people the most protecting when they’re flailing.
May 5th, 2008 at 5:26 pm
[...] of regressions and other statistics related things that I only have a vague understanding of, Sean predicted that the results of the Indiana primary would be a 14 point win for Hillary Clinton, 57-43. Well, [...]
May 6th, 2008 at 2:37 am
[...] race42008.com » Blog Archive » Previewing Indiana (tags: politicalmaps indiana) [...]
May 6th, 2008 at 9:10 am
[...] really detailed analysis at Race42008 says Hillary takes Indiana by 14, losing North Carolina by [...]
May 6th, 2008 at 6:57 pm
[...] (SW / Bloomington): Prediction: Clinton 57.1, Obama 43.9; 3-3 Delegate Split. Race42008.com: Previewing Indiana Plugging these numbers from OH and PA into IN, Obama’s best district will be the 7th, which he [...]