A legit point, and it’s not just pollsters who are mindful of it. The people being polled are aware of it too.
Here's Steve Doocy ranting for a full minute about how the polls are probably all wrong because of secret Trump voters. pic.twitter.com/6lWOFfVL4E
— Matthew Gertz (@MattGertz) July 16, 2020
Apparently this has been a recurring theme on Fox News today:
Fox's John Roberts today: "Could there be enough secret Trump voters in the battleground states to take him across the finish line?" pic.twitter.com/xN39BiY2ub
— Lis Power (@LisPower1) July 16, 2020
Monmouth’s poll of Pennsylvania yesterday showing Biden up big was headlined “Biden Leads But Many Anticipate Secret Trump Vote.” The topline number had him ahead of Trump, 53/40, but when people were asked who they thought would actually win the state, the numbers shrank all the way down to a 46/45 advantage — for Trump. Monmouth explored that by following up with the question, “Do you think there are so-called secret voters in your community who support Donald Trump but wont tell anyone about it, or not really?” Nearly 50 percent said yes, and it wasn’t just optimistic Republicans who thought so:
By comparison, just 18 percent thought there were secret Biden voters out there. It’s hard to believe that anyone who lived through the shock of the 2016 election would doubt that there’s a meaningful number of people who support Trump but who might not be connecting with pollsters for whatever reason. Maybe pollsters aren’t finding them. Maybe they have an aversion to answering questions from pollsters on principle. (“Suppression polls!”) Maybe they’re just embarrassed to tell a pollster right now that they’re still pro-Trump given the difficult time he’s had lately but intend to show up on Election Day for him regardless. I wouldn’t bet on Biden to cover half of the 13-point spread Monmouth’s set for him in Pennsylvania.
But it’s one thing to say “yes, there are secret Trump voters out there” and another to say “yes, there are enough secret Trump voters out there to hand him a second term.” Nate Cohn of the Times has a post up today about that, noting that it would take a big polling miss — bigger than 2016 — to turn what looks to be a comfortable Biden victory right now into a narrow Trump win. The key, says Cohn, is that pollsters have learned from some of their mistakes last cycle, especially the core mistake of undersampling white voters without a college degree. That’s Trump’s base; they turned out for him in droves four years ago. Underestimate their support for POTUS and your poll is destined to miss. So pollsters are taking care this time not to underestimate them.
Perhaps most important, many pollsters now weight their sample to properly represent voters without a college degree. The failure of many state pollsters to do so in 2016 is widely considered one of the major reasons the polls underestimated Mr. Trump’s support. Voters without a four-year college degree are far less likely to respond to telephone surveys — and far likelier to support Mr. Trump. By our estimates, weighting by education might move the typical poll by as much as four points in Mr. Trump’s direction.
Though many state pollsters still do not weight by education, far more do than four years ago. The Monmouth poll is one example. The final Monmouth poll of Pennsylvania in 2016, which showed Mrs. Clinton up four percentage points, would have shown her with a two-point lead, 47 percent to 45 percent, if it had been weighted by education, according to Patrick Murray, director of the poll. That alone covers about half of the difference between the actual result and the final Monmouth poll, and it’s a reason to have more confidence in the new Monmouth poll.
There are also fewer undecided voters this year, Cohn points out, which means less uncertainty around the topline numbers. And pollsters are being more cautious about expecting blockbuster African-American turnout this time, having learned from 2016 that the tremendous support from black voters for the Democratic nominee in 2008 and 2012 may have been unique to Obama and won’t be repeated. All of this is a long way of saying that pollsters agree with Doocy — there are secret Trump voters out there, probably a lot of them, possibly even a game-changing amount if antipathy to polls has spread so widely on the right that most Trumpers just won’t answer the phone for them anymore. But precisely because they do agree with Doocy, pollsters are doing what they can to account for those voters. Their polls should be more accurate this time because of that.
They might even be too conservative in estimating the size of Biden’s lead at the moment. New from Gallup:
Double-digit leads for Democrats on party ID aren’t unheard of, says Gallup, but they’re rare and historically have presaged blue waves. Interestingly, it doesn’t appear to be the pandemic that caused the bottom to fall out here for Republicans. They held steady throughout April and May, when New York was being ravaged by COVID-19, and only collapsed in June, a month when we had a (relative) respite from mass infection. What must have driven voters towards Democrats were the protests following the death of George Floyd. Various polls since then have showed broad support for Black Lives Matter and police reform. Since those are issues championed by the left, swing voters may have swung around towards identifying as Democrats in solidarity.
If that’s what happened here then the wide gap between Democrats and Republicans might shrink as those issues recede a bit over the next few months. But maybe not: Although the pandemic apparently didn’t drive this change in partisanship, it may cement the change if it persists deep into fall. That is, someone who moved from independent to Democratic because they identify with the protesters might “lock in” to that new identity if the spread of COVID-19 gets worse on Trump’s watch. Having finally found the motivation to put on one party’s hat, a swing voter will need some persuading to take it off again. Containing the virus might provide that reason. It’s hard to see right now what else might.