Q: Campaigns would have the —
SHAW: Campaigns would have, almost universally, Eisenhower, Kennedy, Johnson, Nixon, Reagan, Carter. You know, all the way through the Bush’s, Clinton, and even up to the Dubya elections, George W. Bush, campaigns would have these five categories, with the two base categories on the ends. And then these three categories in the middle. [00:16:00] And they would have states that kind of leaned towards them — you know, they wanted to shore up — states that probably leaned towards the opposition, that might be worth taking a crack at, and then states that really seemed to be determinative.
And what’s interesting, I’ve had these conversations with some of my friends and how Electoral College strategy has changed over time, whether it’s changed over time. And one thing that struck me is that even if you’re on a campaign that has very little chance of winning, you know, the McGovern campaign in 1972, or the Mondale campaign in ’84, or the Dole campaign in ’96, that you’ve got to figure out what your minimum winning coalition would look like, even if you may have a hard time — if that’s unlikely, you still need to figure out the combination of states that get you to 270. So even if you’re George McGovern or Bob Dole, you’ve got a strategy. And you’re going to concentrate resources in those states that get you to 270.
Now, if you’re on the opposite side, if you’re Reagan in ’84, or Clinton in ’96, you can expand [00:17:00] your strategy, but if — at its core, that’s an Electoral College strategy. And there’s some consistency over time to this. Within — so when you identify those small set of states that you think are going to be critical to getting to 270, then the question is, how do you allocate resources within that small band of states?
Q: Let me ask you this: how do you decide what — and the term that’s usually used for what you’re describing is “battleground states,” right?
SHAW: “Battleground states,” or “swing states.”
Q: How do you determine which those states are?
SHAW: Yeah, this has actually evolved over time. I’ve gone back and done some research in the Eisenhower elections, in the Kennedy and Johnson elections, Nixon elections, and a lot of it is based on previous electoral history. So you take a look at the last electoral map. And states that were close, you assume are going to be close moving forward.
Now, what we did in 2000 actually, and this actually a product of some of the work I had done with Fred Steeper, [00:18:00] is you can actually take states and look at their votes over time, compare election to election. And if you do something we do in political science fairly regularly and that’s run a factor analysis, you actually find that the loading of states breaks down into two factors. Sometimes elections — the structure of states, that is, the rank ordering, which states are most Republican to which states are least Republican — has a structure that kind of loads on a similar factor, and on other instances, the structure loads on another factor. And if you look at the details, what you find is that in some elections, the southern states end up being really pro-Republican. They’re at the top of the list, the Mississippis and the Alabamas. In another set of elections, Utah, Wyoming, Nebraska end up being at the top of that list. You know, this is all based on historical data.
But it occurred to — I ran this with Fred Steeper in 1992. It occurred to Steeper that, and me [00:19:00], that some elections seem to structure more on ideological terms, that is the rank-ordering of states was a reflection of how conservative ideologically the states are. And those elections that loaded there were elections like 1980, and 1964. On the other factor, you saw elections in which the underlying party distribution in the state was the dominant factor in terms of rank order. And those are elections like 1972, or 1976. What seems to — and then there are elections that fall in that 90 degree space between those two factors.
But one of the things that Fred and I were interested in in ’92, and one of the things I mentioned, and ran for Karl repeatedly in 2000 was, what kind of election were we looking at, in ’92 or 2000? Is it an ideological election in which states — you know, the Border South states for instance, become much [00:20:00] more Republican, or is it a party election, in which case the Border South states become battleground states like Kentucky and Tennessee? And we found in 2000, initially, that if you ran some of the various scant state polling data that we had in late 1999, or early, very early 2000, that the race looked like it was going to fall almost on a 45 degree angle, that it was not an ideological race, not a party race, but something almost right in between, a hybrid. And if you do that, it produced a hypothetical rank order of states, according to the Republican potential. And we used that list and updated it continuously. When we would get new polling information in, we would feed the polling information into the factor analysis to see if there was an adjustment in the vector that 2000 projected into.
Now eventually, you know, I mentioned and been obsessing with historical results. But eventually, you get enough polling information [00:21:00] that you can discard the historical data, except as a check, and use the rank order based on contemporaneous polling information. And what we ended up getting was, when enough state polling data had come in, we found that, essentially the rank order was very close to the vector we had projected.
So, I think what’s happened over time, this was becoming the case in 2000 and 2004, was that polling information almost supersedes historical results. The question is, so for instance, if you got a state that’s gone Democrat five times in a row, like West Virginia, for instance, heading into 2000, but you get polling information, or modeling information that suggests maybe it’ll go a different way this time, at what point do you believe it? At what point do you believe those data? And we did have a couple of cases with that. We had West Virginia, which consistently popped up as a pretty promising Republican state [00:22:00] with the polling data. And then we had a state like Tennessee, which we had initially written off because it was Al Gore’s home state. But we kept having empirical evidence suggesting that it would be ridiculous to write it off.
So those are two examples, and I think as we move forward past 2000, 2004, you’ll see states that pop up consistently in this manner that Republican keep going after even though they look tough. Wisconsin is one that comes to mind. Wisconsin hasn’t gone Republican a lot of years, but both polling data and the historical data, you know, and the structural data suggest that it probably ought to be on the target list.
By the way, I would suggest that, one state that, where there was a discrepancy was Pennsylvania. The Electoral College, historical factor analysis that I had run suggested Pennsylvania was not a good state. The polling data suggested it was. I kind of wish we’d gone with the historical data. (laughter) Because we wasted a lot of time and money [00:23:00] in Pennsylvania, and I’m not sure Pennsylvania was ever winnable.