The primary->general extrapolation problem

Memo to pundits and campaigns: can we please stop extrapolating from primary election results to the general election? Here are three inferential problems that are already plaguing discussion of the Super Tuesday results:

1. Winning primaries in states dominated by the other party doesn't mean you will be competitive there in November.

Ezra Klein flags a Hillary Clinton press release claiming that her wins in Oklahoma, Tennessee and Arkansas proves she can "compete and win in red states." But as Klein points out, "Democrats always win Democratic primaries in red states." Obama's wins in North Dakota, Kansas, etc. are just as meaningless as a measure of his ability to win those states in a general election. To take a silly example, I'm sure Bob Dole won primaries in "blue states" before getting slaughtered in them during the general election in 1996.

2. Demographic breakdowns of primary voting don't necessarily translate to the general election.

Via Matthew Yglesias, John Judis writes that "Hillary Clinton.... is going to have problems with white male voters" and that "Obama is having trouble with white working-class voters and Latinos" based on primary election results, a pattern that he then extends to the general election, where he suggests both will struggle with those demographic groups.

But as Yglesias points out, this approach has an an epistemological problem:

Judis' method is going to reach the conclusion "Party X is Doomed" any time Party X has two fairly equally matched contenders. Even if the two contenders are both very strong, each is going to look "weak" among whichever groups of voters prefer the other candidate. Conversely, if there are two very weak contenders then they're both going to look "strong" within the demographic groups where their rival is especially weak.

To illustrate the point, Hillary Clinton is performing poorly among African Americans in the primaries, but if she is the nominee, she will surely do well because blacks as a whole are a heavily Democratic constituency (a point Judis concedes). Similarly, I'm not sure that Barack Obama's weakness with Latinos in the Democratic primaries is a significant problem given the way GOP immigration rhetoric is alienating that community.

For this kind of analysis to make sense, we need to make some distinctions among demographic groups (to be fair, Judis makes some of these points). For instance, struggling to win primary votes among groups that aren't particularly favorable to Democrats is a real sign of potential trouble in the general election. Hillary's weaknesses with men and Obama's weakness with white working class voters could be serious problems in the general.

3. Primary turnout levels aren't a measure of general election appeal.

The argument that Yglesias offers in response to Judis is also flawed:

To me, most indications are that the Democrats have two strong contenders. Consider that in Missouri about 552,000 people came out to vote in the GOP primary -- a primary that all three candidates seriously contested. By contrast 800,000 people came to vote on the Democratic side. If you put all five candidates into a single election, Hillary Clinton's second place showing of 395,000 would have trounced John McCain's 194,304 for third place. Both candidates, in short, are good at appealing to large numbers of voters and getting them to show up.

All this means is that they are good at appealing to large numbers of people who vote in Democratic primary elections. Energizing the base is part of a good campaign, but it doesn't prove much about the candidates' ability to attract support from independents and moderates in the other party. Even the independents who cross over and vote in open primaries are not necessarily representative of independents as a whole.

Update 2/7 9:34 AM: Like Yglesias, Media Matters suggests that record Democratic primary turnout shows that the party's candidates have a broad appeal. That just isn't necessarily true.

Comments

[Rob]

Your points are all well taken, though I'd add a caveat with respect to your second comment.

If, as now seems likely, the Democratic primaries will be inconclusive and the nomination will be decided by the superdelegates in what is perceived as a brokered convention, many of the losing candidate's supporters may be disappointed, bitter and disillusioned. That may particularly be a problem if Hillary Clinton is the nominee.

While Hillary would certainly win the African-American vote in November, will it be by the extraordinary margins Democrats have come to expect? And perhaps even more importantly, how many African-Americans would sit out the general election?

And what about the rosy-cheeked youthful voters who have come out for Obama? Won't many of them lose faith in the system and fail to vote in the general election? And among those who do make it to the polls, isn't it probable that there will be substantial support for a John McCain who's perceived (rightly or wrongly) as straight-talking, authentic and likeable versus a Hillary Clinton who's perceived (rightly or wrongly) as the opposite?

[tof]

i think (3) is an open empirical question. but i am not allowed to start any new projects...

[ikl]

Independents voting in the MO primary would have to be awfully unrepresentitive of independents in the state for Obama's 37% win among them not to be a good sign for him . . .