Man Without Qualities

Tuesday, October 22, 2002

A Sea-Change Signifying Nothing in November?

Peggy Noonan recently opined that "Even though everyone says Sept. 11 changed everything in America, I'm not sure we've fully noticed how much it's changed everything. And here's a paradox: All that change may well yield a kind of stasis, at least immediately, at least in the midterm elections." Ms Noonan may very well be right to see all changed, changed utterly, but passing in November as a terrible stillness.

Here I want to raise the possibility of another kind of change born of September 2001 - a sea-change in the American voter, and how that sea-change might induce a lot more change than Ms. Noonan and other pundits foresee:

Those are pearls that were his eyes:
Nothing of him that doth fade
But doth suffer a sea-change
Into something rich and strange.

Politicians, pundits such as Pete DuPont and pollsters are addressing the upcoming election as one held in ordinary times - with all efforts made, not to ignore the consequences of September 11, but to treat them as ordinary events to be analyzed and dealt with by ordinary means. For the most part such people can do nothing else - their skills and tools are derived from and dependent on past experience - mostly past, ordinary experience. But no matter how much we would prefer to act and believe otherwise, these are not ordinary times, nor this an ordinary election, nor are the consequences of September 11 ordinary in any sense.

Simply put: A necessary reliance on past voter trends and behavior make the predictions of politicians, pollsters and pundits - even those as astute as Ms. Noonan and Mr. DuPont - far more likely to be way off the mark in this election than is normally the case.

There are signs that the widely but vaguely perceived change in voter attitudes may have consequences in these elections that are not being clearly identified by ordinary methods. For example, politicians who have treated the current mix of national issues as a matter of principle, such as Senator Paul Wellstone may have benefited even while going against the public by voting against the Iraq war resolution. But politicians who have tried to treat that mix in ordinary Beltway fashion - most notably, Senator Daschle - have clearly suffered severely. In Senator Daschle's case, this seems to have shown itself in polling, but only indirectly - through a serious and probably fatal drop in South Dakota voter support for junior South Dakota Senator Johnson. Pollster John Zogby attempts to "explain" this effect by breathlessly suggesting we are seeing "A proxy race in South Dakota - Johnson (D) vs. Thune (R) really equals Daschle vs. Bush!" Mr. Zogby advances this view even as he also advances the view that "While some view the Florida Gubernatorial contest as a reprise of the 2000 Bush-Gore race (with Jeb Bush as a stand-in for his brother and Democrat Bill McBride as a replacement for Al Gore), in reality the race is a referendum on the incumbent." Left unexplained is the question of why the South Dakota race should become a "proxy" vote on President Bush, where the Florida race - in which the President is obviously much more closely identified with his own brother, the Governor, is "in reality ... a referendum on the incumbent." The obviously ad hoc nature of these views makes them almost a joke. Mr. Zogby detects some significant changes in voter attitudes, but he has no plausible model to explain those changes. And he far from alone among pollsters.

And even some polling seems to suggest major electoral effects. For example, if it is true that voter attitudes have changed profoundly, the advantage of incumbent office holders, who were selected in prior elections on the basis of prior voter attitudes, should weaken. And, indeed, some pollsters believe that "incumbents in trouble virtually everywhere." And it is at least possible that the huge jump in early voting in Democrat-heavy Texas counties represents an increase in Democrat participation rather than merely a shift in how people vote (as noted by Mickey Kaus), in which case the substantive change could be far from marginal.

An explanation as to why pollsters are likely to be having unusually large problems with this election probably starts with the related questions of poll sample selection ("built-in bias") and what John Zogby calls the polling industry’s "dirty little secret," ... the non-response rate. In the 1980s, about 35 percent of those called refused to answer. By 1999, the non-response rate was as high as 65 percent. That these factors may more profoundly skew polling results in this election than in the past has been noted by commentators on both the left and the right, although they draw different conclusions. For example, the conservative commentator linked above invokes both factors in noting that some polling practices may skew results against conservatives: Polling over the weekend tends to favor Democrats...[C]itizens who are married with children are 2-to-1 more likely to be Republican. These individuals are too busy on weekends to talk to a pollster for 20 minutes. However, the linked liberal commentator writes:

An analysis of the 2000 results shows that a couple of polling outfits that present their findings on the internet and mainstream as non-partisan, are obviously biased, and that their polls for 2002 and 2004 should be questioned. Other mainstream media polls in 2000 were not too wrong, especially if you look at only their final numbers, but taking a look at their cumulative findings, a bias emerges. Overall, the bias in the polls is built into the data, in the sense that they enable the myth that the US voter is turning more conservative, when the facts clearly show the opposite. That the Bush team is chasing a shrinking, over-weighted in the polls, conservative minority is not the news you'll be likely to hear from the media going into 2004, but it's true. .... [T]he problem of almost all of the national polling outfits [is], "Simply put, we had too many Republicans in our sample." ... Yes, contrary to the facts, the media does project that Republicans and Democrats are equal --another faulty trend assumption-- even though the 2000 results showed a 3% difference still exists. The root of the problem is that the polls continue to posit an extra 3% conservative Republican bias.

The other related problem, or what pollster John Zogby calls the polling industry’s "dirty little secret" is the non-response rate. In the 1980s, about 35 percent of those called refused to answer. By 1999, the non-response rate was as high as 65 percent (Consumers’ Research, September 2000). Who might not be responding? Probably latinos and african-americans, progressives, libertarians.... Who is more apt to respond? Social conservatives that want their views known, and want their views to influence society.

Pollsters "correct" for these and other skewing factors, generally using methods not revealed to the public. Ordinarily such "corrections" are pretty good. But such "corrections" depend on using devices derived from past voter behavior to first modify the methodology used to obtain raw polling data, and then sometimes modifying the raw data itself to obtain the final result. But if there has been a profound change in voter attitudes, it seems likely that using "correction" devices based on past voter behavior is simply more likely to create of enlarge errors than in prior elections. For example, the liberal commentator linked above criticizes the Portrait of America poll's supposed built-in bias towards conservatives and notes how other national polls follow a different (but still flawed, in the view of that commentator) methodology:

POA did not weight by party, but by political leanings (conservative, moderate, liberal), and accepted up to 40% conservative as a valid representative sample of data. "For a variety of reasons, our firm has never weighted by party. If they had weighted the data before the election to include an equal number of Republicans and Democrats," their results would have shown Bush up by 2%, which just about what the average of the nationally accepted polls wrongly projected. So, strip the poll of its conservative leaning, and it still sits with a Republican leaning result, just like the national polls. Yes, contrary to the facts, the media does project that Republicans and Democrats are equal --another faulty trend assumption-- even though the 2000 results showed a 3% difference still exists.

But looking at what seems to be happening in South Dakota, for example, how confident does one feel about castigating that 3%?

Another pollster (cited by the conservative commentator linked above) believes that averaging between registration and turnout in the last presidential election, rather than relying on the respondent’s self-description—is more accurate than polling "registered voters" or "adults." But does the World Trade Center now being a large hole in lower Manhattan and the nation preparing to go to war in the Middle East cause one to have more or less confidence in the significance of "averaging between registration and turnout in the last presidential election" in determining likely turnout?

I am not suggesting that current poll results must be wrong, or, if there are bigger errors here than usual, that those errors must favor one party or the other (consider Senator Wellstone, for example). But I do think that the pollsters' ability to correct reliably for many polling problems has been weakened by a profound change in voter attitudes - a change whose profundity, if not its likely effects, seems to be widely observed by many pundits.

But, while pundits such as Ms. Noonan and Pete DuPont do agree on the existence of the change in voter attitude, they can't seem to identify practical consequences, and I am concerned that their tentative conclusion that these elections will yield little change is more a reflection that their approaches - which are based on past, ordinary experience - are simply coming up short. Similarly, Walter Shapiro writes that "after traveling around the country, a political reporter comes home with a fuzzy impression of the national mood," and notes that with "so little movement visible on the political battlefield, this campaign reporter worries that he's missing something big. It's almost like an old war movie where the veteran sergeant says, ''It's quiet out there . . . too quiet.'' He goes on to observes:"With so much at stake, it is hard to find a candidate talking in standard English instead of using buzzwords tested in focus groups." Other thoughtful pundits sense that America today is "more serious" and that the events of September 11 have helped to check some serious national "illusions." If any of that is true, it should have a serious effect on the elections.

But those may not be effects that the ordinary machinery is designed to detect in advance.

Comments: Post a Comment