By Rich Lowry
Sunday, October 18, 2020
The polling aggregator on the website RealClearPolitics
shows the margin in polls led by Joe Biden in a blue font and the ones led by
Donald Trump in red. For a while, the battleground states have tended to be
uniformly blue, except for polls conducted by the Trafalgar Group.
If you are a firm believer only in polling averages, this
isn’t particularly meaningful, but if you are familiar with Trafalgar’s successes
in 2016, when (unlike other pollsters) it had Trump leading in Michigan and
Pennsylvania and, in 2018, Ron DeSantis winning his gubernatorial race, it is
notable. Regardless, it’s worth knowing why one pollster is departing from
nearly everyone else.
To this end, I checked in with Robert Cahaly, who is
predicting a Trump victory, on the latest edition of The Editors
podcast. This piece is based on our conversation.
Cahaly was born in Georgia and got involved in politics
going door-to-door as a kid. He started a political-consulting firm with some
others in the late 1990s. Around 2008, he says, they realized that the polling
they were getting wasn’t very good, so they started doing their own. He says
they got good, accurate results in the races they were working.
In the 2016 primaries, they started putting out some of
their own polls. “Our polls ended up being the best ones in South Carolina and
Georgia,” Cahaly says. “So we started studying what it was that made those so
different.”
Then there was the breakthrough in the 2016 general
election. “We ended up having an incredible year,” he says. “I mean, we got
Pennsylvania right. We got Michigan right. We had the best poll in five of the
battleground states in 2016. And I actually predicted 306 to 232 on the
electoral college. And we went from doing a little bit of polling on the side
to that [being] our primary business in about 24 hours. And since then, that’s
what we’ve been doing.”
Much of Trafalgar’s approach focuses on accounting for
the so-called social-desirability bias. As Cahaly puts it, that’s when a
respondent gives you “an answer that is designed to make the person asking the
question be less judgmental of the person who answers it.” Cahaly notes that
this phenomenon showed up as long ago as the 1980s, in the so-called Bradley
effect, when the African-American mayor of Los Angeles, Tom Bradley,
underperformed his polling in a gubernatorial race. It has been a hallmark of
the Trump era and is one reason other pollsters missed the impending victory of
Ron DeSantis over Andrew Gillum in the 2018 Florida gubernatorial race.
“I’ve got to get past what you want to say in public and
get to what you really feel,” Cahaly says. “Because what’s in your heart is
going to be what’s on that ballot.”
There are a number of methodological differences in how
Trafalgar goes about its work.
One is the number of questions on its surveys. “I don’t
believe in long questionnaires,” Cahaly says. “I think when you’re calling up
Mom or Dad on a school night, and they’re trying to get the kids dinner and get
them to bed, and that phone rings at seven o’clock — and they’re supposed to
stop what they’re doing and take a 25- to 30-question poll? No way.”
Why does that matter? “You end up disproportionately
representing the people who will like to talk about politics, which is going to
skew toward the very, very conservative and the very, very liberal and the
very, very bored, “Cahaly explains. “And the kind of people that win elections
are the people in the middle. So I think they miss people in the middle when
they do things that way.”
According to Cahaly, most polls are more than 25
questions. He keeps it between seven and nine, so respondents can answer in a
matter of minutes.
Then there is how the questions are asked. “We do not
like to do all live calls,” Cahaly says.
This goes back to the social-desirability bias. People
with opinions that are unpopular “don’t want to be judged by somebody on the
phone that they don’t know.” If this was always true, it’s particularly so now:
“They’ve seen all this stuff of people being shamed for their opinion, people
losing their jobs.”
So Trafalgar mixes up how it contacts people, and
especially wants respondents to feel safe in responding. “We use collection
methods of live calls, auto calls, texts, emails, and a couple that we call our
proprietary digital technology that we don’t explain, but it’s also digital,”
Cahaly says. The point, he continues, is to “really push the anonymous part —
this is your anonymous say-so.”
Another factor, is that “conservatives are less likely to
participate in polls in general,” he says. “We see a five-to-one refusal rate
among conservatives.” That means “you’ve got to work very hard to get a fair
representation of conservatives, when you do any kind of a survey.”
Trafalgar also goes about building its list differently.
One thing the firm noticed in its polling in the Georgia and South Carolina
primary in 2016, Cahaly says, is “people voting who didn’t know how to use the
touch machines, people showing up who hadn’t voted in 15 years.”
It went out of its way to build a list including these
kinds of low-propensity voters, “knowing,” per Cahaly, “that the other
pollsters probably weren’t even reaching out to these people.” The firm has “a
fingerprint” of characteristics meant to find these hard-to-identify voters.
Cahaly excoriates pollsters who use exit polls from the
previous election to determine the demographics of the current electorate.
“Exit polls can give you a sense of how people are voting,” he notes. “But how
many people of a certain age, ethnicity, geography turned out? You ain’t got to
guess at that. It’s a knowable number. And every single state maintains those
statistics.”
He also has no use for relatively small sample sizes. “I
think this is important,” he says. “We don’t do a state with less than a
thousand. You see these polls, 400, 500, 600 people for a state. I don’t buy
that. Your margin of error is far too high.”
Trafalgar tries to avoid so-called weighting to get the
partisan mix of respondents right. A traditional pollster might want to get,
say, 35 percent Republicans to have a balanced survey, but he comes up short
because Republicans are less likely to respond. If only, say, 22 percent of
Republicans answer, they are given additional weight to make up for the
shortfall.
“The better you do at getting an even sample,” according
to Cahaly, “the less weighting you have to do.”
One problem with weighting is that Republicans “who don’t
like Trump can’t wait to answer a poll,” he says. “So immediately, within the
22 percent, they’ve probably overrepresented it, the anti-Trump Republicans,
the Never Trumper types. Well, when you weight that up from 22 to 35, now you
have skewed an already bad representation sample. So that’s kind of inherently
how they can be so off.”
As a general matter, he discounts national polls. First,
because the race for the presidency is won state by state, not on the basis of
the national vote. Second, because all the methodological difficulties involved
in getting a balanced, representative sample in a state poll of 1,000 people
are magnified in a national survey. “It’s easily skewable at that point,” he
says. “You start making assumptions.”
So how does he see the 2020 race? Fundamentally, as a
motivation race, rather than a persuasion race, with perhaps 1.5 percent, at
most, of the electorate undecided in battleground states.
The likeliest Trump electoral path to victory involves
winning the battlegrounds of North Carolina, Georgia, Florida, Arizona, and
either Michigan or Pennsylvania among the former Blue Wall states (assuming he
doesn’t lose states such as Iowa or Ohio).
This is Cahaly’s breakdown: He believes Trump will win
North Carolina and Florida and discounts Biden’s chances in Georgia because the
Republican-base vote is too big there (the same is true in Texas).
As for Arizona, “I think Trump has the lead,” Cahaly
says. “I think [Republican senator Martha] McSally has some ground to make up.
I see her about five points behind Trump, but I think Trump will probably win
the state. And win it by a couple of points or more. And if he wins it big
enough, McSally has a shot.”
Trump isn’t there yet in Pennsylvania, according to
Cahaly. “Right now, we’ve got him down in Pennsylvania,” he says, “I think if
it were held today, the undecided would break toward Trump and there’d be some
hidden vote. He’d probably win Pennsylvania. But I’m going to give a caveat on
only Pennsylvania. I believe Pennsylvania to be the No. 1 state that Trump
could win and have stolen due to voter fraud.”
In Michigan, Trafalgar has Trump ahead. “I think he will
win Michigan,” Cahaly says, citing fear of the Democratic economic agenda.
Overall, Cahaly sees another Trump win. “If it all
happened right now,” he maintains, “my best guess would be an Electoral College
victory in the high to 270s, low 280s.”
There it is. Among pollsters, you heard it from Cahaly
first, and perhaps exclusively — a position he’s been in before.
No comments:
Post a Comment