Episode Transcript
Transcripts are displayed as originally observed. Some content, including advertisements may have changed.
Use Ctrl + F to search
0:00
It's a new ghost burger
0:02
from Carl's Jr. It's a
0:04
juicy, char-boiled Angus Beat Burger.
0:06
Eeee-fi- Melty Ghost Pepper Cheese.
0:08
Eeee-fi- Crispy Bacon Trippy Spicy
0:10
Soul-Scorching Sauce Burger. Peekity Spicy
0:12
Sauce! And that's the ghost
0:15
that haunts the recording booth. I also
0:17
like Carl's Jr. Burger. You've said that
0:19
before, Jeb. Save me one? Here. I
0:22
don't have any tea. Grab a ghost burger
0:24
for a limited time. Only at Carl's Jr.
0:26
Beat Burger, Get Burger. Available for a limited
0:28
time at participating restaurants. So,
0:33
Emily Guskin. Because you're the
0:35
deputy polling director here at The Post,
0:39
I want to know what it's like
0:41
to be asked about a poll over
0:43
the phone. Like, let's say
0:45
that I have been chosen to answer
0:47
a poll that's being conducted by The
0:49
Washington Post. What does that look
0:51
like? Well, first your
0:53
phone's going to ring for this poll. Okay, let's
0:55
just do it for real. Okay, so, doo doo
0:58
doo, ring ring ring ring. So I pick up
1:00
the phone. Hello? Hello, I'm Emily. We're not selling
1:02
anything. Just doing an opinion poll on interesting subjects
1:04
in the news. I'm sorry, I'm a little busy
1:07
right now. How long is this going to take?
1:09
Very short. We'd love to have your opinion. May
1:11
I please speak from our team? This
1:14
is she. Wonderful. Again, my
1:16
name is Emily. Before we continue, are you
1:18
driving or doing anything that requires your full
1:20
attention right now? Uh,
1:23
I mean, I'm like washing the dishes,
1:25
so I guess not. So, super, we
1:27
will go ahead then. That's
1:32
my co-hosts, Martine Powers, speaking with
1:35
Emily Guskin, our deputy polling director.
1:37
If you're like me or Martine, you
1:39
may be wondering about all these political
1:42
polls you've been hearing about right now.
1:44
So, we brought Emily into our
1:47
studio to help us make sense
1:49
of election polling this year, and
1:52
to also give us an idea of what it
1:54
actually feels like to take a poll over the
1:56
phone. Are you registered to vote in
1:58
the District of Columbia or not? this
8:00
really fundamental aspect of polling
8:02
called sampling. And sampling is
8:04
figuring out that sample of
8:06
people who will reach out
8:08
to and those people respond to the
8:10
poll questions. And the
8:12
sample that we reach out to
8:15
should be representative of the total
8:17
universe that we're surveying. So if
8:19
it's, say, Virginia voters, the sample
8:21
that we're reaching out to, which
8:23
is not every single voter in
8:26
Virginia, it's just people that we
8:28
randomly chose to respond to the
8:30
survey needs to reflect everyone there.
8:32
So we want to ensure that
8:34
the population is reflective. An
8:39
analogy I like to use is making
8:41
soup. If you're making soup
8:43
for your friends who are coming over, you're
8:45
not going to taste the entire bowl before they
8:48
get there because then you've got nothing to serve
8:50
them. And you would be very, very full. And
8:52
you'd be a crappy hostess. If
8:55
you take a ladle and stir it up so
8:57
that you can mix it up and make sure
8:59
that the salt is all over and
9:01
that the beans are all over and not just
9:03
in one part totally, and then take a tiny
9:05
taste and be like, yeah, that's the ticket. And
9:07
you make sure in the spoon that you've got
9:09
a little bean and you got a little piece
9:12
of carrot and you make sure that it's got
9:14
the, you know, that it's a nice mix.
9:18
It's a little, little hot bath of
9:20
public opinion. And if it needs more salt,
9:22
you go back and you get more
9:24
salt in there. Okay.
9:26
So that helps me understand what a sample
9:29
is. But then getting back
9:31
to this idea of this poll that's
9:33
out in the field, once you
9:36
get the responses back from people, then
9:38
what happens? Then we get a
9:40
data set that comes back
9:42
to us and it just
9:44
looks like a list of numbers
9:47
and nonsense to most people. And
9:50
we are able to analyze that data
9:52
by looking at the different responses, but
9:54
also breaking them down by crosstab. So
9:56
at the end of all those issue
9:58
questions, we also ask people, you
12:00
look and it's just a two percentage
12:02
point difference, that's not huge. You
12:04
can be an educated consumer and look at what the
12:07
margin of error is in a poll and know
12:09
that it's a close race. Right
12:12
now, the presidential race is a pretty
12:14
close contest nationally and in several key
12:16
states, but in some states, it's not
12:18
as close. We did a poll in
12:21
Virginia and we found Harris leading Trump
12:23
by eight percentage points. And we can
12:25
comfortably say that she's leading there because
12:27
that eight percentage point difference was larger
12:29
than twice our error margin in that
12:31
poll. So
12:34
you're talking about how we should know
12:36
to trust a poll. And
12:38
I think one other kind of marker
12:40
of trust is the name
12:42
that's attached to that poll. We hear
12:44
these different organizations like Ipsoceana,
12:47
Pew, and I don't know really
12:49
what any of these places are
12:51
or what they do other than
12:53
I guess release a lot of
12:56
polls, but how should we be
12:58
thinking about the names that are attached to these
13:00
polls? There's a lot of polling
13:02
outlets out there and a lot of them are
13:04
good. I don't expect people to
13:06
really remember who to trust
13:08
and who not to, but I think there's a few
13:11
things that people can look for when they're reading polls.
13:13
If it's done by a big media organization,
13:15
even one that's not the post, you can
13:18
be pretty confident in them trying to do a
13:20
really good job, right? New
13:22
York Times poll, Wall Street Journal poll, an Associated Press
13:25
poll, like these are all and
13:27
then TV networks too. And
13:30
you can also see how much information they've
13:32
shared about that poll. So if there is
13:34
a methodology statement that talks about margin of
13:37
error and sample size and things of that
13:39
nature, that can give you some more confidence.
13:41
Also look at the question wording. A poll
13:43
should release the wordings of their questions. You
13:46
don't have to have a master's degree in
13:48
polling methods to understand if a question is
13:51
fairly written and if something is
13:53
leading or trying to get someone to respond in a
13:55
poll. In a certain way. But
13:57
also it's really important to just look and
13:59
compare. There are outlets that share a bunch
14:02
of different polls in one space, and you
14:04
can look and see. Sometimes
14:06
the outlier actually ... Sometimes an
14:09
outlier is not an outlier, right? If a race
14:11
changes a lot, maybe the
14:13
race changed. But if there's been
14:15
like five polls in one state
14:17
and four of them show a
14:20
two-point difference or a one-point difference or
14:22
an even, and one shows a 10-point
14:24
lead, that 10-point lead might
14:26
just be an outlier, but it could also
14:28
show that the race is changing. So it's
14:31
good to keep watching, to see poll averages.
14:33
The Washington Post has a poll average in
14:35
a bunch of swing states
14:37
and also nationally, and you can compare just
14:39
to that average to see what you think
14:41
and to see what others are measuring. After
14:49
the break, Martine and Emily talk
14:51
about the public's trust issues with
14:53
polls and what pollsters
14:55
have done to improve since 2016. We'll
14:59
be right back. It's
15:24
a new ghost burger from Carl's Jr. It's a juicy
15:26
charbroiled Angus beef burger. Melty
15:49
ghost pepper cheese. Crispy
15:52
bacon, trippy, spicy, soul-scorching sauce
15:54
burger. And
15:57
that's the ghost that haunts the recording booth. Also
16:00
like us, Junior Burgers. You've said
16:02
that before, Jeb. Save me one?
16:04
Here. I don't have any
16:06
tea. Grab a ghost burger for a limited
16:08
time. Only at Carl's Junior. Need burger? Get
16:11
burger. Available for a limited time at participating
16:13
restaurants. I
16:16
want to acknowledge the elephant in the room
16:18
here, because I think when
16:20
you ask a lot of people about polling,
16:22
whether polling can be trusted, they
16:24
will often bring up instances
16:27
or situations or elections where they felt like
16:29
the polls weren't accurate. And one that comes
16:31
to mind, obviously, for many people is 2016,
16:33
where there is this
16:35
widespread belief or feeling that
16:38
people didn't see Trump's
16:41
win coming, or that the polling didn't
16:43
reflect that, and that that's a reason
16:45
to not trust the polls
16:47
going forward. What would you say to someone who thought
16:49
that? In
16:52
2016 and 2020, polling was kind of far
16:55
off. We're a self-critical group of folks' pollsters,
16:57
and we meet every year at a conference
16:59
and basically analyze what has happened in previous
17:01
years and try to improve on what we've
17:03
done before. And we really want to get
17:06
it right. Political horse-shares polling is
17:08
particularly hard, because we need to figure
17:10
out exactly who is turning out to
17:12
vote in an election. And we don't
17:14
know until after election day who voted.
17:17
In 2016, a lot of polls showed Hillary
17:20
Clinton ahead of Trump, but there
17:22
were some that we're a lot closer to, and
17:24
it was a very close election. One
17:27
way we've tried to improve since then
17:29
is waiting by education. So
17:31
we talk about ensuring that
17:33
the population is reflective. That
17:35
election in particular, and subsequent
17:38
elections, it's seen that
17:40
people with higher educations tend to vote
17:42
Democratic, and that wasn't necessarily the case
17:44
with white people in previous elections. So
17:47
making sure we have more balance and
17:49
are able to do that. We also
17:51
look at non-response issues more, trying
17:54
to ensure that there's no
17:57
group of people that's systematically
17:59
not responding. to surveys, so it's important
18:01
to look at that because this
18:03
country chooses its president not
18:05
based on popular vote nationally
18:08
but on the electoral map,
18:10
which is very confusing and
18:12
means that we have to do a lot
18:14
more work to see what's happening. That
18:17
I think was part of the reason that 2016 was
18:19
surprising to so many people. Hillary Clinton was
18:21
ahead in the popular vote in so many
18:23
polls leading up to the election, that
18:26
we don't choose our president in the
18:28
US based on popular vote, but there
18:30
was a lot more polling
18:32
being done at the national level. It
18:35
was a really close election and we've
18:37
tried to improve on those results. In
18:39
2018, polls are better, in 2020 they
18:42
were worse, and our goal
18:44
is to improve and to get that right. And
18:47
it seems to me at least that by
18:49
and large, like in recent elections, polls
18:51
really have gotten it right. In
18:54
2018, 2022, the midterms, you know
18:56
that there was a real matchup between polling
18:58
and what actually transpired in the election. We
19:00
like to get it right. What
19:03
are some of the common mistakes that you see
19:05
in people interpreting polls incorrectly? I
19:08
think it's good to have a skeptical hat on
19:10
when you're seeing something from an advocacy
19:13
group or someone who has skin in the game. They
19:15
can ask questions in certain ways to get
19:17
the answers that they want. News
19:20
pollsters don't want to do that.
19:23
Candidates and political partisans often will
19:26
release a poll if it looks
19:28
good for their candidate, but
19:30
not if it looks bad for
19:32
their candidate and they'll be choosy about that. You
19:35
might say, oh, this congressional candidate
19:37
is up 10 points according to them, but they
19:39
might have had 15 polls before
19:41
then where it was neck and neck and they
19:43
just kept that to themselves because they're not a
19:45
public pollster and they don't really have to release
19:48
that. What you're saying is trying to be
19:50
aware of the folks who are
19:52
putting out this poll. Whether they have skin in the
19:54
game, whether they're trying to advocate for something or
19:56
are they're on one side or the other and
19:58
that they have a good candidate. an
20:01
incentive to create a skewed picture. Absolutely, and
20:03
I get a lot of that in my
20:05
inbox. Those people are sending
20:07
journalists a lot of stuff all the time.
20:09
I can imagine. One more
20:11
thing that has come up relatively recently
20:13
that I've been wondering about is
20:15
this term prediction models, which
20:18
to me sounds like a different way of saying poll,
20:20
right? That like, this is
20:22
the thing that's going to tell you who's going to win and who's going
20:24
to lose. What is a prediction
20:26
model? And how
20:28
should we be thinking about those? Prediction
20:31
models take polls and put them in
20:33
a computer. This sounds like
20:35
explaining something to my dad, but just put
20:38
into a computer, the machine will take care
20:40
of it. Pips and bops and do the
20:43
math. No, my dad
20:45
is very good at understanding these things. He
20:47
does a good job of explaining stuff to
20:49
the other people in their 55-plus community about
20:52
polling. It's like he's doing God's work. Prediction
20:55
models and polls are different. Prediction
20:58
models use polls along with
21:00
other indicators that they figure
21:03
out and put in to
21:05
look at what might happen in the future.
21:08
Polls are snapshots of what is happening
21:10
right now. We're focused on today with
21:12
a poll. Prediction models,
21:15
you have to believe in the indicators
21:17
that they're expecting to have an impact
21:19
on the future. So at
21:21
the post, we don't have a prediction
21:23
model, but we average polls, which is
21:25
a bit different. So we're not looking
21:27
into the future. We're trying to use what
21:30
we have to give us the best picture of
21:32
what is happening today. So we
21:34
include both state and national polls, and
21:36
we weigh them using information about who
21:38
is putting that poll out to give
21:40
us the best estimation of the true
21:43
state of that race right as it
21:45
is right now. So
21:48
you've talked about how polls
21:50
work, why
21:52
polls can be trusted when they come
21:54
from reputable sources and when they have
21:56
large enough sample sizes, and how polls
21:59
can be trusted. polls can be really
22:01
insightful. But I'm
22:03
curious more from a self-care
22:06
standpoint, how much
22:08
do you think people should be spending
22:10
their day-to-day following the ups and downs
22:12
of this poll and that poll and
22:15
whether or not a
22:17
half-points percentage change in one direction or
22:19
the other, how much do you
22:21
think that is informing the
22:25
wider public in a useful way versus
22:27
maybe it's too much, maybe you shouldn't,
22:30
be living and dying by whether or not
22:32
your preferred candidate happens to have
22:34
gotten a slightly more positive poll in the last 24
22:37
hours? So I've been
22:39
working in polling since 2008, that
22:42
was my first presidential election, working
22:44
in polling, and
22:47
let me tell you that self-care
22:49
is really important. And
22:52
living and dying by your candidate's
22:54
polls can be exhausting and probably
22:57
not healthy, but please keep reading
22:59
our stuff. I
23:01
think it's good to be aware of
23:03
it, but be aware of it in
23:06
context. And still polling is the best
23:08
way to measure what a population thinks.
23:11
And I think it's important to have
23:13
compassion for voters. We have
23:15
undecided voters, that's a fact. That
23:18
can be- And I think some people find that
23:20
very frustrating. I think people find that really hard
23:23
to believe. Like at this point, how can you
23:25
not be sure whether or not you want to
23:27
vote for Donald Trump? Either you do or you
23:29
don't, and what more information do you need? People
23:32
are complicated and some of them are low information,
23:34
but a lot of them aren't. And they have,
23:36
if you voted Republican for your entire life, it
23:39
might be hard for you to consider voting for
23:42
a Democrat now, and vice versa. The
23:44
people who respond to our surveys and the
23:47
Americans overall, I'm so thankful to them. Because
23:49
a lot of us are in our bubbles,
23:51
and a lot of us only talk to
23:53
people who we see eye to eye on
23:56
or have similar life experiences to us. And
23:59
a poll can open. up our eyes
24:01
to how other people think and other
24:03
people feel and how they live their
24:05
lives. And there's
24:07
a lot of people out there who
24:09
aren't like us and they deserve to
24:11
have their voices heard just as much
24:14
and maybe even more so if they're more interesting
24:16
and they're a bit unsure about who
24:19
they're going to vote for and why
24:21
they're unsure about who they're going to
24:23
vote for. I think it's a great
24:25
responsibility for us to measure people's opinions
24:28
because no one else is really doing
24:30
that than pollsters and we see
24:32
a lot of news articles that quote experts
24:34
but I think the American people really are
24:36
the experts and they're the people that we
24:39
do need to hear from and understand how
24:41
they feel and think and why they're doing
24:43
what they're doing. Emily,
24:47
I never thought that you would be able to
24:50
make me feel emotional about polling. Thank
24:52
you so much for sharing that. Thank you so
24:54
much. Emily
24:57
Guskin is the deputy polling director for
24:59
The Post. She spoke to
25:01
my co-host, Martin Powers. Before
25:15
I leave you, I wanted to share some news
25:17
that I've been following today. We're
25:20
still learning about how devastating Hurricane
25:22
Helene has been. The
25:24
death toll is now over 150 and
25:27
that number is expected to rise. My
25:30
colleague Brianna Sacks is on the ground right
25:32
now in southern Georgia in a
25:35
town called Valdosta. This
25:37
area was still recovering from a
25:39
hurricane last year when Helene
25:42
hit. Brianna was inside
25:44
of a fast food restaurant
25:46
when she met Jesse Johnson-Chiroux.
25:49
Like her neighbors, Johnson-Chiroux has
25:51
been dealing with trying to feed
25:53
her family and she hasn't
25:55
had power for a week. The
28:00
kind of immunity is reserved
28:02
for official presidential acts. This
28:04
filing is intended to show Trump can
28:06
still face trial by arguing
28:09
he was acting as a political
28:11
candidate, not as the president. And
28:14
there could be a political fallout
28:16
from this filing. My colleague, politics
28:18
reporter Aaron Blake will unpack that
28:20
and more tomorrow in his podcast
28:22
The Campaign Moment. Usually
28:24
you can find The Campaign Moment on
28:26
Fridays in our post-reports feed, but we'll
28:28
have a different episode here tomorrow. So
28:31
head to The Campaign Moment podcast feed
28:34
and subscribe so you'll be sure to catch it. That's
28:39
it for post-reports. Thanks for listening.
28:42
Today's show was produced by Ariel Plotnick
28:44
with help from Bishop Sand. It
28:47
was mixed by Sam Baer and edited
28:49
by Rina Flores. Thanks
28:51
also to Scott Clement and Brianna Sacks.
28:54
I'm Elahe Izzadi. We'll be
28:57
back tomorrow with more stories from The Washington Post.
29:26
Bye bye.
Podchaser is the ultimate destination for podcast data, search, and discovery. Learn More