12/19/2008 10:04 am ET Updated Dec 06, 2017

Penn's WSJ Column

Former Clinton pollster Mark Penn has a new Wall Street Journal column named (what else?) "Microtrends" and written "with" E. Kinney Zalesne, co-author of the book of the same name.

Gawker responds with its usual snarky flair but makes a reasonable point, if you can get past the gratuitous nastiness: This week's column spends more time hawking Penn's book than supporting his arguments with data. They point out, for example, that Penn/Zalesne attach this footnote their first use of the term "Impressionable Elites:"

* "Impressionable Elites" is the term we used for educated, affluent people who focus more on personality than issues when it comes to evaluating political decisions. For more, please see pages 131 to 135 in "Microtrends."

There's a stranger footnote. The central contention of the column is that Bernard Madoff's bilking of a lot of very rich people "proves the point" that "elites have become more impressionable -- more removed from everyday problems, more trusting of what they hear, and more likely to adopt unthinking viewpoints based on brand or emotion."

The data that supports this claim?

[O]ur research** shows that the top 1% is heavily swayed by gut and impression, not numbers and facts. They vote more on the basis of personality in campaigns; buy products more on the basis of brands; and invest more on the basis of the tip than on sound logic.

And what research is that? Another footnote tells us:

** A PSB [Penn, Schoen & Berland] poll of 806 telephone interviews among likely 2008 presidential voters, including an oversample of 400 very likely Democratic presidential primary voters.

Two issues here. First, sample size? I will leave it to readers to decide exactly how many interviews such a sample will yield for the "top 1%." Presumably, non-voters fall disproportionately in the bottom 99%, so it may be that the wealthiest or best educated 1% of the population contributed slightly more than 1% of this sample. Still, I still have a hard time seeing how that top 1% subgroup could have yielded much more than n=15, unweighted.

Second, consider the reference to the "oversample" of likely "Democratic presidential primary voters." It certainly sounds like a survey done during or before the Democratic primaries on behalf of Mr. Penn's most well known recent client. It is not hard to imagine such a survey including tests of negative messages about Obama ("numbers and facts") that failed to persuade his very wealthy supporters (for an example of such a message test -- click here, scroll to pp. 11-12). Please tell me that the referenced evidence of elite impressionability depends on more than that. Please.

Update: Reader DS points out that the Microtrends chapter referenced in that first footnote includes results that give a good clue as to the evidence referenced in the second footnote. Specifically, on pp. 132-133, Penn and Zalesne write (the table is my reproduction of the one in the book):

This isn't just my gut. Let's look at the data.

A standard poll question I ask in campaigns is what people consider most important in voting for a candidate: (1) issues, (2) character, or (3) experience. I ask it because I know that all three are important in a leader, and that it can be tough to rank them.

According to a recent poll we did, a large plurality of voters -- 48 percent -- believe that a candidate's stand on the issues is most important, with character a distant second at 32 percent. That preference for issues holds steady whether or not voters have been to college, whether or not they are religious, and across race. Where it does not stay constant, however, is across income. Once voters reach the magic line of $100,000 per year, their priority shifts to character, by a significant margin. As the table below shows, people earning under $100,000 prioritize issues over character by a serious 51 to 30 percent. But once they reach $100,000, the switch, to character over issues, 45 to 37 percent.


That is a 29-point swing. A shift barely ever gets clearer in polling.

So if this is "research" referenced in the footnote, there's a good chance that the data used to make inferences about the "top 1%" was based on the 25% or so of voters that report an income over $100,000. Also, as DS puts it in his email, there isn't any more survey evidence on the book on this point: "no regression analysis, no experiment, just a simple, dumb question."

As it happens, the soon-to-be-extinct LA Times/Bloomberg poll asked a similar question in their September 2008 campaign survey. I have reproduced the results below. Notice two things: First, the Times/Bloomberg question showed 31% choosing "all" or "some" of the categories. The PSB had no such category and reported only 1% as "don't know." If allowing interviewers to take "all of the above" as an answer draws in nearly a third of the respondents, we have a pretty good clue that the underlying attitudes are soft.


Second and more important, the Times/Bloomberg poll showed a clear pattern by party and ideology: Democrats and Liberals were more likely to choose "positions on issues," while Republicans and Conservatives were more likely to choose "experience" and, to a lesser extent, "character." Here's a theory: Voters perceived McCain as strong and Obama as weak on "experience," so McCain supporters were more likely to say they consider experience important and Obama supporters less likely to choose that answer. That result is consistent with my experience that voters rarely answer such questions in a completely abstract way. They tend to answer with high profile candidates in mind, especially when the question comes in the context of a survey that has already asked about those candidates.

So before making too much of the pattern in Penn's cross-tab, I would want to see whether any other variables show similar correlations and whether controlling for other variables (through a regression analysis) would explain the pattern by income.

And of course, as DS points out in his email, there are many sources of political survey data that would allow a more complete test of this thesis. He says, for example, that he examined ANES data "to see if highly educated / high income respondents were more likely to be impressed by character qualities rather than issues," but found no connection.