Thursday, October 27, 2016

Do Clinton's Supposedly Negative Poll Numbers Mean Anything?

by Neil H. Buchanan

If there is one narrative that has taken root during the long course of the 2016 U.S. presidential campaign, it is the idea that both Hillary Clinton and Donald Trump are widely loathed.  It has become a staple not just of political commentaries but of late-night comedy to point to polls that seem to show that both candidates are historically unpopular.

But a factually true statement can be still be grossly misleading and damaging.  Lazily pointing to some poll numbers to attack both Clinton and Trump is apparently good sport.  Unfortunately, it also distorts the way that people think about the candidates.  Even worse, it allows voters to say, "Well, I don't want to make a choice, because they're both so bad.  Everybody thinks so."  It has thus become a self-reinforcing distortion.

Much of my discussion here will involve criticizing the use of public opinion polls, so let me first clear up a few possible misunderstandings.  Most importantly, I am not buying into the usual criticisms of polls as a general matter, for example, the claim that polling is inherently inaccurate.

The fact is that, as a matter of forecasting election results, polling has become impressively precise.  Although there are always unscientific (which usually means deliberately biased) polls that prove nothing, the last few elections have shown that careful analysis of independent polling can lead to accurate predictions.  Candidates who are behind in the polls disparage those polls, of course, but that means nothing.

Moreover, it is not true that polling cannot capture "soft" ideas.  For example, a relatively new branch of social science known as "happiness research" has made it possible to try to measure people's well-being in ways that stand up to replication and that can prove helpful to policymakers.

Properly done, therefore, public opinion polling can quite usefully improve our knowledge of the world.  Even so, it is still true that even the best pollsters ask questions that are ambiguous, and too many people (including the pollsters themselves) are willing to over-interpret the results.  A few examples will make the point.

Last month, I wrote a guest piece for a British publication called The World Financial Review (TWFR). (The column is, unfortunately, behind a paywall.)  There, I quoted from a Pew Research Center report from July -- "Campaign 2016: Strong Interest, Widespread Dissatisfaction."  Pew is one of the top polling organizations in the world, but it is not immune to analytical errors.

Summarizing a national survey of voters, Pew wrote that "large numbers of the supporters of both Trump and Clinton view their choice as more of a vote against the opposing candidate than an expression of support for their candidate."  This, according to the report, was "another sign of voter discontent."

But is it, really?  Is it not possible that at least some people are very happy with their candidate but simply dislike the other person more?  I happen to know that to be possible, because I am one of those people.

As the campaign has moved along, I have become ever more impressed with Hillary Clinton, to the point where I now view her as a potentially great president.  On the other hand, although I would not have thought it possible, my opinion of Trump is now lower than it was a week, a month, or a year ago.

Because Trump scares me so much, my honest answer to Pew's question -- "Are you voting more for your candidate or against the other candidate?" -- would have to be that I am voting against Trump.  Yet that response does not at all comport with the conclusion that Pew draws that both candidates are reviled.

Moreover, the pure hyper-partisanship of the current election all but guarantees that poll numbers are going to look more negative than in previous elections.  To find "high negatives," a pollster does not even have to do what Pew did.  All you have to do is contrast the number of people who dislike Clinton and Trump with the number of people who disliked Bush and Dukakis or McCain and Romney.  This is a new era.

Consider another example of a poorly worded question that can be badly misinterpreted.  Over the summer, The Guardian published "The Death of Neoliberalism and the Crisis in Western Politics."  Because I am a critic of neoliberalism, I suspected that this would be a pleasant read.

In some ways, it was indeed an insightful piece.  At its conclusion, however, the article cited an unspecified poll to support the claim that "[r]oughly two-thirds of Americans agree that 'we should not think so much in international terms but concentrate more on our own national problems.'"

But what does that really mean?  The author of The Guardian's piece presented it as proof that the xenophobia that Trump has tapped into is going to continue to express itself in "opposition to the hyper-globalisers."  Possibly, or possibly not.

A respondent to that poll could be saying that "we should not be getting involved in more foreign wars, given the problems with poverty here at home," even though she wants the U.S. to fight poverty abroad as well.  Another person might sincerely believe that we should close our borders and ignore the world entirely.  That poll's results do not prove what the author claims they prove, because two people with widely divergent views could give the same answer.

Similarly, the Washington Post published an article last month citing poll results showing that "[o]nly 21 percent of Latinos say the GOP cares about their community," whereas "70 percent say that Trump has made the Republican Party 'more hostile' to them."

How could thirty percent of Latinos not believe that Trump has made his party more hostile to them?  Perhaps they think that the Republicans could not be more hostile than they were before Trump came along, so that his arrival has done nothing to change their views of the Republican Party.  That would not mean that they have a benign view of Trump, but only that they already saw the Republican Party for what it had become.

Another poll, cited by a columnist in The New York Times last month, asked people to respond to the following statement: "It bothers me when I come in contact with immigrants who speak little or no English."

Could I honestly agree with that statement?  Yes.  It bothers me because it reminds me that I never learned another language, which I regret.  It bothers me because I would like to be able to communicate with everyone.  It bothers me because I worry that they will view my clumsy attempts to communicate with them as condescension or worse.

None of that would mean that I am uncomfortable with immigration, or that I am in favor of English-only laws, or any of the rest of the nativist line.  But if I were being completely honest, I would have to say that I agree with the statement in the poll.

And this is where things become especially interesting.  People can be quite savvy, and they often know what the "right" answer is to many poll questions.  Some Latino voters who hate Trump might know to say "yes" to the question of whether Trump has made Republicans more hostile to Latinos, because that is the way to register anti-Trump sentiment.

Similarly, someone in my position would be a fool to agree with the statement that coming into contact with non-English-speaking immigrants is uncomfortable.  Respondents frequently know how to register their approval or disapproval.

Thus, even if a Democrat is genuinely worried about the racism that has surfaced during Obama's presidency, she would be crazy to give an honest answer to a polling question that asks whether "the country has been moving in the right direction or the wrong direction since Obama took office."

When interpreting such polls, therefore, we really do not know how many people answer "naively honestly" or "strategically honestly" (much less dishonestly).  And the difference matters.

All of which brings us back to the meme that "both Clinton and Trump are historically unpopular."  As I noted above, my complaint is not with polling in general, but with bad questions and poor interpretations of responses.  Fortunately, just this week a helpful analysis was published by the poll-meisters at FiveThirtyEight.

Under the helpful title, "Clinton Voters Aren't Just Voting Against Trump," Harry Enten examines apples-to-apples polling from 1980 onward regarding voters attitudes about their candidate and against the other candidate.  Because this is a comparison over time, it is possible that any bias introduced by my concern above -- that being "more against" Candidate B can obscure the degree of affection for Candidate A -- might be relatively constant over time.

If so, Enten's analysis is especially interesting, because he notes that although "56 percent of Clinton’s voters ... affirmatively supporting her may not seem like a lot, ... it’s about average for a presidential candidate."  (Consider just how amazing it is that Clinton is hitting the historical averages, given how much the anti-Clinton narrative has taken hold even among Democrats who plan to vote for her.)

Moreover, Enten writes, "The most interesting thing about these numbers is how few of Trump’s supporters are his fans. No candidate since 1980 has had a lower percentage of voters say they plan to cast a vote for their candidate."

Again, that latter result could simply be a measure of just how much hatred there is out there for Hillary Clinton.  But viewed in historical context, these results do paint a rather different picture from the threadbare "two terrible choices" story line.  I am not saying that Enten's polls are better than other polls, but I am saying that his analysis is simply more careful and complete.

And remember, this is all in a media environment where voters have been told for months that they are supposed to hate both candidates.  That there is nonetheless so much positive support for Clinton has to be a good sign for her presidency.


Greg said...

If I try and read between the lines of these different results, I think they're saying something that seems obvious but perhaps isn't.

Hillary Clinton is a semi-polarizing candidate. Democrats like her as much as any other normal candidate, but don't love her. Republicans loathe her.

Donald Trump, on the other hand, is a universally reviled candidate. Democrats dislike him about the same as they would dislike any other Republican candidate, but the difference is that Republicans don't like him much either.

This does support the idea that both Clinton and Trump are more unpopular than the typical candidate, but for different reasons. Clinton is unpopular because she has less support from the opposite party than a typical presidential candidate. Trump is unpopular because only a minority of his own party likes him very much. I don't know if this idea is supported by the favorability ratings of the two candidates when favorability is separated by party.

Michael C. Dorf said...

Greg: I disagree with the suggestion that Trump is disliked by Democrats "about the same as they would dislike any other Republican candidate." Based on my admittedly unscientific observations, numerous Democrats (including myself) regard Trump with a kind of existential dread that was not in play with the likes of McCain or Romney and that wouldn't have been in play with Jeb Bush, John Kasich, or Marco Rubio. (Ted Cruz is sui generis.) I think that what Democrats find especially objectionable about Trump is much the same as what many Republicans find objectionable about him: his character and personality traits that make him ill-suited to be president, quite apart from whatever policies he happens to favor.

Joe said...

The fact that reasonable, I say this without sarcasm, people like Greg says things like that is depressing. It's simply untrue but such a basic misunderstanding is all too present.

The other half is somewhat non-conventional wisdom too, since the idea there is that she isn't as popular even among fellow Democrats is fairly popular. I think that is exaggerated but there is some truth to it. It's balanced though because she does have a segment that really like her. I figure some segment of Democrats will find a reason to dislike most candidates but Clinton does turn some off that others might not.

It does seem true to me that Republicans loathe her more -- they have a long history there of building up the dislike as compared to someone like Obama. It would be strongly partisan otherwise but if Obama let's say was running, I think at least a few would find it harder to say "got to vote for Trump, you know Obama!" Or, someone new. After four or more years of battle, some might be riled up.

OTOH, behind the scenes, some actually do say they respect her skills. (Her fellow Republican senators respected her on that level.)

Greg said...

I think people are misinterpreting my post. Anecdotally, I would agree that Trump is a whole other level of bad, and people seem to know it. I erroneously equated amount of dislike with number of people who dislike, which was a mistake on my part.

My point was, I'm not sure the idea that Trump is unusually disliked by Democrats is supported by the statistics, and certainly isn't supported by Prof. Buchanan's reference. I see a few possible reasons for that.

1.) Approval ratings tend to be pro/con types of things, and don't really reflect that Trump is considered to be an existential threat vs. just a bad choice, so long as the same numbers of people disapprove.

2.) If it were really true that more Democrats dislike Trump than the typical Republican candidate, that would imply that Clinton is actually a more-liked than average Democratic candidate *among Democrats.* I don't think that's true, so I would interpret the Harry Enten numbers to imply that from a numbers perspective, Trump is disliked by about the same number of Democratic voters as a typical Republican. I could be wrong and Clinton is in fact an unusually popular candidate among Democrats.

3.) Trump does have some success with white union voters who actually believe that he can bring manufacturing jobs back to the USA. That probably lowers the percentage of Democratic voters who are voting against Trump vs. for Clinton.

Of course, all the polling biases already discussed by Prof. Buchanan also probably play a role here, as some voters who are really voting against Trump will instead say they are voting for Clinton as they know this is the "right" answer.

Joe said...

"erroneously equated amount of dislike with number of people who dislike"

If so, not quite sure what "misinterpreting" is involved.

As a matter of raw theory, I would be willing to accept that numerically let's say McCain will be "disliked" in some raw fashion by comparable number of Democrats as Trump. But, that is not how me and Prof. Dorf took "like her as much."

I "dislike" various things to the degree of not supporting them in some fashion. But, "as much" to me was a matter of degree. On that level, I do think some additional Democrats are not supportive of Trump. If Republicans aren't, the Democrats who are somewhat Republican-curious (those willing to think about them & sometimes vote for them ... let's say a McCain in 2000) probably will not be.

This is probably a sliver so I'm not sure how much it pops up in polling. Democrats are going to typically support Democrats. This seems fairly obvious. As to #2, I don't know how that necessarily follows -- you can dislike two people. See many Republicans not "liking" Clinton suddenly because of Trump.

Anyway, given your opening correction, not sure exactly how much is left.

Asher Steinberg said...

I think what you're evading is that Clinton *does* have historically high unfavorable ratings for the likely victor of a presidential election - something which Trump's only being down a touch more than the margin by which Obama defeated Romney tends to support. One would have to imagine that were Obama allowed to run for a third term, he'd do quite a bit better against Trump than he did against a perfectly unoffensive, amply qualified generic Republican.