On The "GOP is Demoralized" Consensus
Posted by JAY COST | E-Mail This | Permalink | Email Author
In yesterday's entry, I discussed the much-vaunted Republican GOTV machine. I asserted that there is little evidence to conclude whether or not it has been efficacious. Oh sure - we have media/journalistic accounts and anecdotes that outline Republican activities in 2004. And these indeed sound very impressive. But it is one thing to outline what the Republicans do, and it is quite another thing to measure the effectiveness of those doings. And we simply lack the data for the latter task.
But this is not to say that we cannot draw some reasonable conclusions about it. The fact that political actors - on both sides of the aisles - believe that it is efficacious says something important.
I left yesterday's post with this question: is the GOP base less "spirited," and therefore less susceptible to mobilization? Unfortunately, I do not have an answer to this question. What I intend to outline here could be classified as a critique of the media consensus on the matter. All in all, I suspect that the base is - to some unknown extent - dispirited. My intention here is not to rally my own laundry list of polls to "prove" GOP spiritedness. Rather, I intend to criticize the method media analysts have used to "prove" GOP dispiritedness.
What follows are two rejoinders to the arguments typically offered by the media to justify the dispirited storyline.
1: Underdetermination: When bloggers are critical of media polls, they almost always criticize the one element of a poll that is least susceptible to criticism: the external validity of the sample. This is precisely what pollsters - at least those who follow industry standards - are qualified to do. It is their technical specialty, and so it always seems odd to me to hear non-specialists discuss the external validity of a sample (e.g. the partisan makeup of a set of respondents) - especially when there are other, more fertile, grounds for critique.
The most salient critique stems from the fact that pollsters - at least as far as I know - are not political psychologists. The pundits who interpret their polls most certainly are not. And, while they are quite qualified to put together externally valid samples - they are not necessarily qualified to interpret their own results. And I see a lot of lousy interpretations of polling data in the press - which is to say that I see many inferences from polling data that are (a) deemed to be necessary when they are not or (b) deemed to be probable when they are not.
The inferences drawn from media polling about Republican spiritedness speak precisely to this point. For instance - people look at the crosstabs for a question like: "How much attention have you been paying to the 2006 campaign?" and see that Democrats are registering a much higher positive response rate than Republicans. From this, they infer that Democrats are more likely to vote than Republicans. That is all this is - it is an inference: the response to the polling question is not the voting act itself, nor does a certain answer to the question obligate the respondent to vote or not vote. To move from answers to the question to estimates of the voting act is to draw a causal inference. You are inferring that that which caused the answer to the question will also cause the choice to vote or not vote: Republican voters are not planning to vote and, accordingly, are not paying attention; Democratic voters are planning to vote and, accordingly, are paying attention.
But is this necessary? Is it even probable? Here is another hypothesis that is entirely consistent with the data: Republican voters assess that the climate is a negative one for their party, and they are paying less attention to politics because they do not like interacting too much with a negative environment -- but when their Republican candidates start to campaign, they will liven up, as those candidates are - by dint of their advertisements - offering something much more positive. That seems to me to be an equally reasonable inference - perhaps even more reasonable, as it recognizes the importance of candidate campaigns in stimulating the electorate.
Do not take this the wrong way. I am not pushing this particular hypothesis over the consensus one. What I am doing is asserting that the consensus hypothesis is underdetermined. There are other, equally reasonable, inferences to be drawn from these media polls - and these equally reasonable inferences are at odds with the media consensus. With the data we have, there is no way to arbitrate between the different interpretations -- because media polls are explicitly designed for maximum news value, not maximum Truth value. They are not like the American National Elections Survey - where we could cross-reference a whole host of questions to test competing hypotheses more thoroughly. We cannot do that with media polls - hence the problem of "underdetermination."
Let me add a general warning. It is one thing to draw an inference about voter preferences. That seems to me to be fairly straightforward. If you call up a voter in Findlay, Ohio on his way out to vote - and he tells you he is going to vote for Sherrod Brown, you can draw a very reasonable inference from that response: the guy is voting for Brown. However, we are not discussing something so simple. Voter spiritedness seems to me to be a very thorny theoretical issue, one that gets to the heart of voter psychology, which is a complicated and difficult subject. It requires much training in statistics, method, and existing scholarly research. When I came to graduate school, I initially focused on political philosophy. I "jumped" to American politics several years ago, and found that -- all in all -- the reading load was much easier to bear. It is just easier for me to understand Robert Dahl's Who Governs? than it is to understand David Hume's Treatise of Human Nature. It is less abstract and more obviously related to the day-to-day of my life. Except for the public opinion/vote choice literature. I almost met my end in my public opinion graduate course reading Sniderman, Brody and Tetlock's Reasoning and Choice!
Public opinion is a very hard subject -- wickedly hard, given that at first glance it seems downright easy. The reason it is so complicated, I think, is that those who study it are ultimately interested in how the average voter thinks. However, if you have advanced to a level of political knowledge where you begin to wonder how the average voter thinks, you have probably lost much of your "sympathy" (in the Humean sense) for the cognition methods of the average voter! The difference between the two of you now is information -- as you acquire more and more political information, you begin to make political decisions/conclusions in a radically different manner than the average voter. Thus, experts -- broadly defined -- must study the average voter as one studies the "other." Political scientists are aware of this, and so develop very rigorous theories that treat the average voter with this kind of respect. Political pundits seem to me to have no awareness of the difference between themselves and the average voter, and so proceed to blithely make all kind of factual/methodological mistakes in "interpreting" him/her. I think they ultimately go wrong because generalize their own cognition processes to the rest of the electorate, which I think is very egotistical - but that is a subject for another day!
Simply stated, I think that most of those in the media who try to "do" voter psychology without lots of years of training in it are probably talking out of their you-know-what's.
2: The Ecological Fallacy: It is looking increasingly like the difference between a Democrat-run House and a Republican-run House is coming down to who wins the following districts in the Midwest: IN 02, 08, 09, KY 04, OH 01, 15, 18. In terms of mobilization, this is a very specific slice of the GOP electorate. To what extent does it differ in politically relevant ways from the rest of the GOP-inclined electorate? Indiana and Kentucky, according to SurveyUSA, have a relatively high estimation of Bush. Ohio has a relatively low estimation of Bush. Are its GOP voters different in terms of spirit in the same way? If they are, then the ecological fallacy has reared its ugly head once again: when the extent of spiritedness varies within the whole nation, we cannot draw inferences about the parts of the nation from surveys of the whole nation.
Those national polls that tell us about GOP spiritedness are almost always polls of the national GOP base. Common sense indicates that the GOP base is probably somewhat dispirited everywhere -- so the difference between California base voters and Kentucky base voters is probably one of degree only. But the size of the degree is impossible to judge.
This is a species of a general problem that has dogged the media in their endeavors to explain the House (and, to a lesser extent, the Senate) contests that are quickly approaching. This battle is being fought in, at most, 40 House districts, or little less than 10% of the nation. How can we so blithely use national survey data to draw inferences about the situation in this small sub-sample? This is invalid reasoning!
One might call this the Washington Redskins fallacy - the Skins so often like to cherry pick good players/coaches from other teams, ostensibly under the conclusion that, since Steve Spurrier (for example) was the head coach of the Gators, and the Gators were such a good team, Spurrier will make the Skins a good team. This does not follow - as we all know. The mistake is called the ecological fallacy: you cannot draw inferences about parts of a whole from data of the whole.
Again - do not misunderstand me. I am not arguing for one interpretation of GOP spiritedness over another. My broad point in this entry is that we lack the data to make a valid inference either way because (a) the data leaves our preferred hypothesis underdetermined and (b) the data does not speak to electoral outcomes, as the latter will be determined by a tiny portion of the whole.