New to LessWrong?

New Comment
13 comments, sorted by Click to highlight new comments since: Today at 11:23 PM

Social desirability is a big one, thanks, I didn't think of that.

A related issue is that every representation we see of reality -- every newspaper, blog, TV show, movie -- is created by a writer or artist, definitionally. So when we look at representations of the world, we're looking at the biased viewpoint of the highly verbal and well-educated. And, of course, these are not all the people. If you live in a highly symbolic/representational world, you may be very mistaken about how things work, not only because of a particular bias in your bubble of representations, but because of a systematic bias in all representations.

(I think Tolstoy is an unusually valuable resource because he writes about the inner lives of the kind of people who would never, ever write or talk about their inner lives.)

Social desirability bias is a part of this -- it biases what we believe towards what people are willing to talk about -- but I think the entire phenomenon is worth thinking about.

Another possibility is that

that less-educated whites, who historically have had a low propensity to vote, turned out in greater numbers than pollsters predicted


In states where white voters tend to be well-educated, such as Colorado and Virginia, the polls pegged the final results perfectly. Conversely, in northern states that have lots of whites without a college degree, Mr Trump blew his polls away—including ones he is still expected to lose, but by a far smaller margin than expected, such as Minnesota.

and in this case I would expect social desirability bias to be stronger among well-educated (is there any research about that in general case?), because the percentage share of their peers that do not like Trump is greater.

It doesn't change the narrative about a lot of things, but it does provide new evidence about polling.

Technically true, but ultimately, it doesn't provide all that much new evidence about polling. The polling error was not abnormally large.

And the direction of the error was known and stated in advance by informed interpreters (538). A fair number of Trump voters would not have been considered "likely" voters based on past non-voting, and that was a systematic bias in the polling estimate rather than something that would affect a few states independently. Pollsters tended to stick with their "likely" filter rather than change it on the assumption that these voters would turn out and vote. They turned out and voted.

I seem to recall seeing Trump doing better in polls of registered voters versus likely voters, but I cannot say I have strong evidence for that and it might have just been comparing a few surveys. Most polls seem to have been of likely voters.

I think social desirability bias almost certainly did mask Trump's support.

The study I linked to is a pretty strong case for it existing. The study randomly assigned voters to complete a poll online or via phone. College educated voters were substantially more likely to support Trump in the online poll. Whether social desirability bias alone accounts for Donald Trump outperforming the polls is another question.

It's a common hypothesis, but I think there are some questions to answer.

1) If social desirability effects were a factor, why didn't they show up in the Republican primaries? Trump slightly underperformed his polling there.

2) If they were a factor, why were they such a small one? And how can we claim to see the effect in such a small polling error? Polls were more accurate this time around than 4 years ago, and I don't think anyone thought there was a social desirability effect in 2012.

If you look at total numbers of votes cast for each candidate in each state, it looks like it was

1) in states like FL, big loss in number of blue voters.

2) in states like PA, uptick in red voter participation

Of these, the first seems more significant - Trump got fewer votes than Bush, Romney or McCain, so his vote total can't have got up that much, while Clinton got 9 million votes fewer than Obama did. Clinton barely got more votes than Kerry did, and he lost.

Since the overall turnout was so low, I would suspect it was a turnout failure throwing off likely voter models, rather than people lying to pollsters. Though of course it could have been.

I think this is a better model here,

"To win voters to his cause, Trump tapped a vein of nationalism that social science calls honour culture"

"Honor Bound, a new book by social psychologist Ryan Brown, lists US states by how concerned residents are about their reputation and their desire to retaliate against perceived slights. The top 10 are South Carolina, North Carolina, Alabama, Georgia, Arkansas, Mississippi, West Virginia, Virginia, Tennessee and Texas. Trump was victorious in all of these except for Virginia.

Romney won 8 of those states. Maybe (no) that tells us something about Republicans in general, but that tells us nothing about how Trump is different from Romney. Did Trump even do better in those states than Romney did? He did much worse in Texas.

OP's link is more about the reason behind voter shyness, where as your link is more about the reason behind voter choice. But yea, I agree mostly with what your article interpretation too.


That was remarkably close to the default assumption...