I have beaten this particular horse to death, but if you went looking for a state that is white and working class, Oregon wouldn't be a bad place to end up. Just 1.6 percent of Oregonians are African-American (though it does have its share of Hispanic and Asian voters). It ranks below the national average in income levels while having one of the nation's higher unemployment rates.
And yet, there may be something about Oregon's political DNA that's a little different. As I mentioned previously, Oregon has the most left-leaning Democratic electorate in the country. The problem is that the relationship between ideology and voting patterns has been murky thus far in the primaries. Clinton did quite well for herself in states like Massachusetts and New Hampshire, which also have quite liberal electorates. But more recently, she has tended to excel in conservative areas.
I checked to see whether there was a time trend in the data and in fact there was. When the election cycle began, Obama was actually outperforming Clinton in Republican-leaning districts. In a district with a rating of R+10, for example (one where George Bush performed ten points better than his national averages), Obama would be expected to win by roughly 8 points. But now, the same district would be expected to go to Clinton by 8 points. Conversely, however, Obama would be expected to win a Democratic-leaning district that he might have lost before.
This helps to explain how polling in New Jersey and California, two states with fairly liberal-leaning electorates, now suggests that Obama might have won those states. But it also proved to be helpful in Oregon. My modeling was consistently showing Oregon to be a toss-up state -- leaning only slightly to Obama. And if Oregon had voted back in February, maybe it would have been a toss-up. It might be noted that in Washington's beauty contest primary (which the model does not use directly in its estimates), Obama beat Clinton by just 3 points. And Washington should be a couple of points stronger for Obama than Oregon, as it is wealthier and has a somewhat larger black population.
But now, that isn't how Oregon is likely to vote. Clinton smartly recognized that the states that were scheduled to vote late in the primary process were moderate or conservative-leaning states. As such, she has moved somewhat to Obama's right. That's going to work to her benefit in places like Kentucky, but she's liable to pay a price for it in the one Kerry state that remained on the calender, which is Oregon. Tack on a couple of points for the fact that Obama has engaged the state more actively, and he could be looking at a double digit margin.
* - *
Apart from recognizing this time trend, I made a few additional tweaks to my model. Fundraising is now tracked at the Congressional District level, although as it turns out that you do best with a combination of local and statewide numbers. There is also a relationship between the fundraising numbers and a state's ideology; the more liberal a state's electorate, the more important the fundraising numbers tend to be. We have accounted for this effect as well.
I have also tracked candidate visits at the CD level; a visit may be split between two or more districts if the city a candidate visit straddles more than one Congressional District. As it turns out, a visit to a CD does not appear to have any strong effect on the way that the vote is divided up in that district; it only affects things in the state as a whole. However, candidate visits do appear to have a localized effect on turnout. If you think about a typical campaign event, it is probably designed more to generate enthusiasm among a candidate's supporters than to persuade undecideds, who aren't as likely to attend the event in the first place. This finding would be consistent with that characterization. You draw your supporters to the event itself, hoping to make sure that they turn out. But thereafter, you're mostly benefiting from the earned media, which is not invaluable, whose effects tend to be more dispersed across the state.
It is interesting, by the way, to see the different ways that the candidates tend to plan out the geography of their events. Obama is more inclined to double- or triple up in a particularly important Congressional District, whereas Clinton tends to blanket her states more evenly.
Finally, as Oregon has a decent number of Hispanic voters, we are accounting for that variable as well. Holding everything else like income levels constant, Hispanics perform worse for Obama than what we call our white Anglo-Saxons, but marginally better for him than certain other types of "ethnic" whites. When you see the large margins that Clinton rung up in South Texas, keep in mind that they were very similar to her numbers in West Virginia. Ethnically, the two areas are very different, but economically they are quite similar.
Let me run through Oregon's Congressional Districts:
OR-1 (Northwest / Beaverton). We name the first district, tucked away in Oregon's Northwest corner, after the city of Beaverton, but it also includes portions of Portland. Either way, it's solid Obama country. OR-1 is Oregon's wealthiest and most educated district, its population is relatively young, and it's bohemian and gay-friendly -- all of which are favorable markers for Obama. Still, it isn't especially extreme in most of these departments as compared to the national averages, whereas its 1.3 percent African-American population is exceptionally low. It's not out of the question that Obama could get a 5-2 delegate split here, but our model thinks he'll fall a few votes short. Projection: Obama 62.1, Clinton 37.9; Obama 4-3 Delegate Win.
OR-2 (East / Bend). The second district, making up the eastern three-quarters of Oregon's land mass but only a fifth of its population, is among the most difficult in the country to characterize. It is the third-whitest district in the entire nation, it is very much blue collar, and it also has a high senior population. But Clinton has not fundraised at all well in OR-2, and it has a high veteran population, which is one of Obama's hidden bases of support. The model thinks that, had the Oregon primary been held on Super Tuesday, this district would have gone to Obama by several points. But with Clinton now overperforming in conservative-leaning districts, it sees it as a toss-up, very slightly leaning Clinton. With that said, I wouldn't be surprised to see a result as lopsided as 58/42 for either candidate in this district. Projection: Clinton 50.2, Obama 49.8; Clinton 3-2 Delegate Win.
OR-3 (North / Portland). This is the district most commonly associated with Portland, and also with Oregon's reputation for progressive politics -- it has, for instance, the highest proportion of gay households of any district not associated with a top-15 metropolitan area. In other respects, it should be slightly less favorable to Obama than OR-1; its income levels are only average, for instance. But as this is the most liberal district in the state, there are about 75,000 reasons to think he might overperform here and get to the 61.1 percent he'd need to grab a sixth delegate. Projection: Obama 61.5, Clinton 38.5; Obama 6-3 Delegate Win.
OR-4 (Southwest / Eugene). Obama has spent a lot of time in this district, and that makes a lot of sense because he'll need to turn out the student vote at University of Oregon in Eugene in order to win it and take a majority of delegates. If not for Eugene, OR-4 would be essentially indistinguishable from OR-2, although it's more urban and does not have OR-2's Republican lean. We think Obama will just squeak by. Projection: Obama 51.7, Clinton 48.3; Obama 4-3 Delegate Win.
OR-5 (WNW / Salem). Congressional Districts containing state capitals are not infrequently swing districts, and Salem's OR-5 is no exception. It's a combination of rural Oregon and urban Oregon and its demographics are similar to the state averages, producing results that should closely track Oregon as a whole. To the extent that OR-5 distinguishes itself, it tips slightly younger and wealthier than the rest of the state, but Obama remains a slight underdog to break the 3-3 delegate deadlock. Projection: Obama 56.9, Clinton 43.1; 3-3 Delegate Split.
The model is projecting an Obama margin of approximately 13 points on turnout of about 570,000 voters, which would represent around 70 percent of registered Democrats in the state. Because Oregon's mail balloting has produced relatively strong turnouts in previous elections, this turnout figure may be slightly low, but it is difficult to conceive of turnout of more than about 650,000 in this closed primary. I have, by the way, switched back to estimating turnout as a function of the age-eligible population in the district rather than the number of Kerry voters, which turns out to be the more reliable way to do it.
Unlike some previous contests, Oregon is a delegate-watcher's dream. A delegate could easily change hands in all five of its Congressional Districts, as well as a couple at the statewide level. Obama will need to win Oregon by 8.3 points to pick up an extra statewide delegate, and 16.7 points to pick up a second. Overall, however, we have him coming up just short on a couple of counts and finishing with 29 delegates to Clinton's 23.
Keep in mind that about 4-5 points of our projected margin stems from the fact that Obama has spent more time on the ground in Oregon. Without that campaign activity, the state looked to be just close enough that Clinton must have faced a tough decision about whether or not to campaign seriously there. It's likely, however, Clinton effectively decided her fate when she decided to move to the political center; an issue like the gas tax moratorium plays quite badly in Oregon. Her goal was to have a strong showing in North Carolina and Indiana, perhaps even running the table in every state but Oregon. But Oregon was probably going to have to be sacrificed to enable that strategy.
by Nate Silver @ 12:59 PM