Big problem with polls; pollsters tinker with data

 

Gordon L. Weil

To ward against underestimating support for Donald Trump, many pollsters adjust their survey findings to increase the influence of how voters say they voted in the last election, according to a report by Nate Cohn, the New York Times chief polling analyst.

According to the report, though the pollsters have their own doubts about the validity of this recall-vote adjustment, they use it to avoid the possibility of badly missing the true outcome.  In short, they try to narrow their possible error and may try not to differ from the herd.

One major defect of the adjustment is that there’s a bias in remembering that you voted for the winner, he says.  But there might also be a bias for “changing” your vote, if the candidate’s later actions cause you to regret how you voted.  In any case, voters may not provide a correct answer.

In effect, the implicit conclusion in making this adjustment is that one election is like another.  With Trump running for a third time, the temptation to reach that conclusion may be great.  And it might be correct.

But what about the possibility that this election is different from most elections?   Kamala Harris is not white and not male.  Those are pretty big differences from the past.

And when the Supreme Court toppled Roe v. Wade, it was like a constitutional amendment.  It not only awakened the opposition of many women, but it raised questions about the meaning of personal freedom.  The issue won’t fade away. 

Texas, with one of the strictest laws in the country, challenged the power of the federal government to order it to save lives, claiming it already had such a requirement. But the legal risks for doctors making the life-and-death decision are so great that many won’t perform abortions. Today, the Supreme Court using its phantom docket (decisions with no reasons given) upheld the Texas position and further fueled the issue.

The abortion issue is likely to bring out women to the ballot box.  They may be more numerous than in the past, and they may recruit others.  A majority of voters are women.

The defection of leading Republicans from Trump, despite having endorsed him eight and four years ago, suggests that the recall-vote adjustment does not apply to them. Perhaps other traditional Republicans will follow them.  We might call this the “Liz Cheney effect,” for which no adjustment is made.

There is a wave of new voter registrations this year, as in Maine, and especially among young voters.  How does the recall-vote adjustment work when the margin of victory in a swing state may be less than the number of new voters? We might call this the “Taylor Swift effect,” for which no adjustment was made.

At the end of his analysis, Cohn writes, “A near repeat of the last presidential election is certainly a plausible outcome. In today’s polarized era, who could possibly be surprised by a repeat in Mr. Trump’s third presidential run?”  He concludes, “But if this election is different, in any direction, this year’s polls might not be able to see it coming.”  The pollsters are not reporting data; they are manipulating it.

Another conclusion may be that the pollsters are more interested in protecting their reputations than in making a serious attempt at understanding the electorate.  They may fail at both.