Are Polls Underestimating Trump’s True Support?
Some voters shy about liking Trump
The press in 1980 did their best to scupper Ronald Reagan. They said he was an “extremist” and that he would “would divide America along racial, religious, and regional lines”. They said Reagan was a “dangerous cowboy” with his finger on the nuclear button. They seethed, raged, insinuated.
The Republican establishment joined in the calumnies. The strain on party brotherhood was so bad that after the primaries, one-time Republican candidate John Anderson split from the party and ran as a NeverReagan.
It was thus unfashionable to admit liking Reagan, and so not a few kept their mouths shut.
Presidential polls might have reflected this Reagan shyness. In the month before the election, polls had Carter up an average 44% to Reagan’s 40%. Anderson hovered around 9%, which left about 7% of voters unaccounted for. Were some of these 7% shy Reagan voters?
The final averages right before the election gave Reagan the edge, 47% to Carter’s 44% and Anderson’s 8%. That left only 1% unaccounted for.
The final tallies gave Reagan 51% of the popular vote, Carter 41%, and Anderson 7%, with the remaining 1% spread over novelty candidates.
There is a huge discrepancy here. Polls showed Reagan with 4% less support than he actually had, and Carter with 3% more and Anderson 1% more. These errors could have been caused by Reagan supporters unwilling to tell pollsters their true preference, but they also could have been because of built-in biases of the polls themselves. These biases should not surprise given that many polls are conducted or commissioned by mainstream media outlets, whose sins and biases do not need recounting.
Shy Trump Voters?
At this writing most polls show Hillary nearly tied with Trump, yet there is a suspicion that, like in 1980, some voters are shy about admitting that they like Trump. If this is so, the polls exaggerate Hillary’s true support.
Is anybody who is for Trump coy? If so, how many secret supporters are there? Or are the polls biased?
It’s easy to imagine scenarios where a pollster queries a citizen who is reluctant to say he’s voting for Trump. The college professor or student called on campus, an employee polled at any company in San Francisco or Los Angeles, a canvasser knocking at the door at a certain address in Chappaqua, New York when the lady of the house is in residence, and so on.
Hillary said half of Trump’s supporters are a “basket of deplorable.” Many on the left agree with these harsh words; deviation from leftist ideology is not countenanced.
So rather than trigger a social justice warrior by announcing their Trump preference, some surely keep their mouths shut.
On the other hand, as the election nears and, for instance, the NeverTrump camp realize how horrible the alternative is, and adding in the common knowledge that Americans like a winner, liking Trump grows easier. The polls, as in 1980, are tightening. Even so, there is still a sense polls under-count Trump’s true base.
Shy Brexit Voters
Disentangling voter shyness from poll biases is not easy. Modern polls over emotionally contentious questions suggest shyness is not negligible. Journalist Michael Tracey reminds us that six weeks before the Brexit vote, which the media and majority of the establishment hysterically disfavored, “Remain” led by 4% with “Undecideds” at 14%. But the final tally was 52% for “Leave,” a huge discrepancy and 8-point swing.
In the Brexit case, poll bias is not a likely explanation because Brexit polls were not sampled in the complex way presidential polls are. The results of the Brexit polls were also simple, in the sense that the numbers released were close to the actual numbers received in the polling process. By these comments I mean that the numbers released to the public in presidential polls are not in their raw form; they have been manipulated by various statistical models (this article explains how).
Scientific Polls? Nah
Now, despite what you might have heard, there is no such thing as a “scientific poll.” Or, rather, all polls are equally scientific. But not all polls are equally good. That public poll numbers are actually the result of statistical models means there is plenty of opportunity for bias and error to creep in.
This is well illustrated by no less than the New York Times, which recently gave four polling groups the same raw data. If polling were a rigid science, the answers should have been the same. But they weren’t. Results ran from Hillary +3 to Trump +1, a 4-point swing, a discrepancy more than large enough to change the outcome of the election (especially considering details about the Electoral College which needn’t detain us here).
There is also the possibility that some biases are intentional, as in so-called push polls, or because the samples are finagled in a preferred direction. Unscrupulous pollsters know that simply showing a candidate is ahead causes some people to favor that candidate (Americans like a winner). And if that candidate is shown to be well ahead of her true support, others will be discouraged from voting (why bother?). But surely the mainstream media would never lie to us, right?
Whatever the polls show, there is no certain way to say anything about their performance until after the election. See you on the other side.