Romney vs. Perry – Statistical Analysis

I’m working on a statistical model to analyze a Romney vs. Perry primary, should it come down to just those two.  My starting point is the 2008 Republican primary… I’m assuming that a Perry voter looks a whole lot like a Huckabee voter… to me, this seems like a reasonable assumption given Mike and Rick are both very evangelical southern governors with a populist streak.

So far, I’ve found that I can get a very powerful model of predicting Huckabee’s performance in a state in 2008 based on just 3 variables:  % of white evangelicals in that state, % of Mormons in that state, and % of the vote Obama received in the general election in that state.

Then, if I go back and apply those 3 variables on each state (using no other information whatsoever), I come up with the following breakdown of states for 2012.  To me, the results seem very plausible (especially taking into consideration what the model is blind to… like the fact that Perry is from Texas, so the Lone Star State should be considered a lock for him rather than a lean) and suggest that we could be in for a very long primary.  I’ll have more on this in the weeks ahead.

Perry Locks (263 delegates)
W. Virginia
Perry Leans (480 delegates)
S. Carolina
Romney Locks (493 delegates)
New Jersey
Rhode Island
New York
Romney Leans (386 delegates)
New Mexico
New Hampshire
Battleground States (674 delegates)
N. Carolina
N. Dakota
S. Dakota


  1. Firecracker says:

    Wyoming, Michigan and Nevada should be considered a LOCK for Mitt Romney. There is no way he is going to lose either of those states and he will win big.

    Montana should be considered leans Romney. He won the caucuses last time around there.

  2. Not sure about a lock, but I would say all 3 of those states at least lean Romney in real life. In terms of the model, it has no way of knowing that Mitt’s dad was governor of Michigan and but for that connection I think it would be a tossup state. This is just my first stab at a model… I’ll try to account for such things later. One other one I will definitely be looking into in predicting primary results is whether it is a closed primary (which favors Perry) or an open primary (which favors Romney).

    To account for cases like Wyoming, I might need to do some tweaking to make the Mormon % a more powerful predictor… right now, my model just looks at Wyoming and sees a state with more white evangelicals than Texas and more conservative than Texas, both indicators that favor Perry. And there are some Mormons there (which is why my model classifies it as a battleground state rather than a Perry state), but it is not quite Utah or Idaho.

    Help me out here… what other variables can you think of that would distinguish why one can predict that Wyoming is a Romney state rather than a tossup state?

  3. Also, to be fair, I think there are states I am classifying as battlegrounds (North Carolina jumps off the page) that should probably be in the “leans Perry” column.

  4. The RNC adopted rules this year to make the March primaries award delegates based on proportional representation. As part of the effort to extend the primary season and give more states the chance to be counted, the party decided that states that vote in March could not be winner-take-all states. If you really want to model the primary contest accurately, you’ll need to take this rule change into effect.

    Another factor this year may be the “Christine O’Donnell effect.” I’m sure that moderate and establishment Republicans in Delaware never believed that a flake like Christine O’Donnell would win the GOP primary. Granted, I probably would have voted for flaky Christine over liberal RINO Mike Castle in the primary if I were a Delaware voter, but they didn’t have any good choices. The 2012 GOP primaries may see a higher turnout among Republicans who typically haven’t put as much effort into the primaries. Responsible conservatives who might have been less motivated in the past will want to ensure that we don’t nominate more kooks. This factor will work for Mr. Romney and against Rick Perry.

    I live in Louisiana, and I’m certain that Louisiana will go strongly for Perry.

  5. Yep, good point Bill… you can’t just add up delgates in the various states because in many cases, the loser will still receive some. It is hard to dig into that at this point… many states don’t even know when their primary is going to be, let alone what sort of distribution of awarding delegates they are going to use. But I hope to eventually get to the level of sophistication to be able to consider all of that.

  6. And yeah, my definitition of “lock” is pretty strict. States like Maryland and Nevada (for Romney) and Mississippi and Louisiana (for Perry) might be in the “lean” category for now, but I wouldn’t read too much into that.

  7. I will give more thought to your question regarding variables later this evening. Modeling is big work. I’m just a housewife. 😉 Good thing there’s someone like you to take on the task.

  8. Yeah, think about it. So far, other variables I’ve tried include other religious groups (black protestants, catholics), income, and education level and none of those proved better than the ones that I already have. But if you have any others you think I should try, let me know.

    I have a couple other ideas of tweaks that I think have potential… there’s nothing like the sheer glee of an actuary getting destracted working with spreadsheets at the office fantasizing about the spreadsheets he’ll get to play with once he gets home. 😉

    (FYI, that WAS a joke).

  9. One I thought of was unemployment in that state. It would probably weigh more into a model for R v D, but it might be an issue here.

    Another thing I considered is whether or not that state was R or D from 2008.

  10. % of the vote Obama received in 2008 is one of my variables, so I think that covers your 2nd one.

    Your 1st one is interesting. Which way would you expect that to work (are states with lower unemployement Romney states or Perry states?) Seems like there are examples of both on both ends of the spectrum (Romney states NV and CA are high but UT and NH are low. For Perry, same thing… SC and MS are high, OK is very low).

  11. Everybody can skip the Romney v Perry modeling – Romney will be, nay, IS the 2012 nominee. That was decided on Super-Duper Tuesday 2008, the same point when the Next-In-Line™ maxim was proven able to withstand the biggest challenge to it in the persona of John McCain.

    As for the 2016 nominee (I have no confidence in Romney’s ability to beat Obama), that too will be decided before the 2012 circus rolls into Wisconsin. The only question on that front is whether the minority half of the Bipartisan Party-In-Government (which gave us both McCain and Romney) will be able to stop Perry.

  12. “Next in line” has seemed to work out in many cases, but I don’t think it is a hard and fast rule. Look at how lucky McCain got in 2008.

    1) If Rudy tries at all in NH (which he should have) he would have probably either won or, at the least, pulled away enough voters from McCain to allow Romney to win.

    2) If Fred Thompson would have abandoned his hopeless campaign a little sooner, Huckabee would have gotten more votes in SC and beat McCain there.

    If either of those things would have happened, McCain never would have gathered enough momentum to win in Florida (as it is, he still won by less than 5% even getting the endorsement from Governor Crist… remember, that was back when Charlie still had credibility in conservative circles).

    Barring another candidate entering, we know that Perry will win a lot of Southern states and we know that Romney will win a lot of Eastern and Western states. For Perry to win, he needs either to score a big win on Romney turf on Super Tuesday or before (like CA or NY) or he needs to win almost all the midwestern states and eek out a very close win. Both are possible, but neither is likely IMO. (That’s the next thing I hope to demonstrate in a future post…. the built in advantages of Romney based on the order and delegate distribution of the various states.)

    So in the end I agree, history probably will play itself out as it always has this time… but I don’t think it is a given. For example, if Perry can’t win on the coasts this cycle and is basically a southern regional candidate in 2012, I don’t see how that would change in 2016 or 2020.

  13. As MSNBC put it last night this is a choice between dumb and dumber. I am truly disappointed that Dick Cheney is not running. It would do his heart so much good.