Some Iowa Statistics, or Why Huckabee Isn't Winning

There have been numerous comments on the incongruity of having Iowa as the first primary state, given how non-representative it is of the nation as a whole. The primary factor is the disproportionate number of evangelical voters in Iowa (though to the largely religion-blind Democratic party this is negligible); the SF Chronicle reported that around 60% of Republican caucus-goers were self-described evangelical Christians according to exit polls, which is an obvious aid to Mike Huckabee. Michael Medved has a post up at Townhall.com analyzing the strength of Huckabee's evangelical boost in Iowa. Unfortunately, in the process of concluding the bias isn't strong at all, he completely screws up the statistical analysis. I'll critique his post first and then offer my own conclusions about the data he presents.

Michael Medved's conclusions are unorthodox if nothing else. He compares Mike Huckabee's support among Evangelicals (46%) to his support among women, the poor, and the young (40, 41, and 40% respectively) as well as the total of 34% of caucus attendees. This leads him to the conclusion that Evangelicals did not differ significantly from Iowa Republicans as a whole. Well, I have news for Mr. Medved: this is because Iowa as a whole is evangelical. The vote counts from women, the poor, the young, and everyone are skewed by the huge proportion of evangelicals present. Thus no real conclusion can be drawn from the data Mr. Medved offers. He does basically the same analysis with Mitt Romney's numbers, making the argument that the difference between 19% of evangelical voters and 24% of overall voters isn't large, and again missing the skew in the overall numbers from the evangelical numbers. He uses this to argue that Romney's "phoniness" alienated voters in Iowa, without any evidence to support the phoniness (I know it's there, but it's sloppy to not show any since he's giving it as THE reason why Huckabee beat Romney); this conclusion might be valid if the statistics pointed that way, but they don't. There goes half his post right there.

Medved also points out here that Huckabee didn't really win the evangelical vote because he didn't get a majority; he got only 46%. But in the same article he points out that the next most popular candidate among evangelicals was Romney with 19% of the vote; the other 4 major vote-getters split 35% of the evangelical vote. In head-to-head contests among evangelicals, Huckabee beat each other candidate by at least a 70-30 split, and it's highly unlikely that Huckabee is the last choice of the 54% of evangelicals who voted for someone else. Medved fails to recognize how dominating 46% of the vote is in a 6-way race, though it doesn't take a rocket scientist to realize the difference between 46% and the average of 16.66...% if votes were divided evenly.

But Medved really gets in over his head with the latter part of his post, where he attempts to blame Huckabee's low numbers among non-evangelicals on anti-religious bias. His justification? The large difference between the non-evangelical number (13%) and the overall vote. That's right, Huckabee's unpopularity among non-Christian conservatives is due to anti-religious bias, and we know this because Huckabee's just so darn unpopular among the non-Christian conservatives compared to everyone else. Of course, I know what he's actually trying to say, which is that Huckabee's unpopularity among non-Christian conservatives is so far out of line with Iowa Republicans generally that there must be some underlying reason, and anti-religious bias is the most likely (though again he provides no evidence for this claim). But the real reason, again, is the huge number of evangelicals. Because evangelicals make up such a huge percentage of Iowa caucus-goers, and because there's such a huge difference between the evangelical and non-evangelical votes (46% to 13%, for reference), the result will obviously be a large difference between the total vote and the evangelical vote. All one can conclude using Mr. Medved's methods is that evangelicals like Huckabee more than non-evangelicals; the rest of his post is a patchwork of unsupported conclusions and poor analysis of statistics.

This is not to say, however, that no conclusions can be drawn from the data Michael Medved provides. First, a trivial calculation: Huckabee's support among evangelicals was 10% greater than total support in Iowa, while the difference between the non-evangelical and total votes was 21%. Since this data would imply that 70% of the caucus-goers were evangelicals, which differs from the 60% estimate by a significant amount, one can conclude that there was a "decline to state" option; furthermore, voters that took that option liked Huckabee about as much as evangelicals. One can guess that the majority of the "decline-to-state" voters were actually evangelicals.

Second, a more significant calculation: What happens if we make Iowa more representative of the nation as a whole? Iowa's evangelical vote represented 60% of the total, but the percentage of evangelicals nationwide is only about 15%. Since the percentage of Iowa evangelicals is about 40%, and evangelicals turned out in greater numbers to vote for the Huckster (a ratio of 3 to 2), let's be generous and guess that the evangelical vote in a more representative state would make up 25% of the total (a 5 to 3 ratio - I AM being generous here). Then we can extrapolate the voting percentages Mr. Medved provided, with 75% of the hypothetical representative state giving Huckabee 13% of their support (non-evangelicals) and 25% of the state giving Huckabee 46% of their support (evangelicals) for a total of 19.8% of the primary vote in a representative state. By contrast, Mitt Romney would get 75% of this state to give him 33% of the vote and 25% of the state to give him 19% of the vote, for a total of 29%. Though Mr. Medved does not link to the source for the numbers he cites, he does say that the other candidates all had far less support from evangelicals (less than 10% each), so their numbers would be even further boosted by this conversion. So the conclusion one can draw from this set of numbers is that the disproportionate number of evangelicals in Iowa actually played a HUGE role in propelling Mike Huckabee to the top. Of course, we knew that already - even if Michael Medved still has no clue.

Anyway, this is why I still think Huckabee has no realistic chance at the Republican nomination. Without the evangelical base, he has nothing that would appeal to conservative voters. My justification of this position another time.