Episode Transcript
Transcripts are displayed as originally observed. Some content, including advertisements may have changed.
Use Ctrl + F to search
0:15
Pushkin. Hi,
0:21
Tim Harford here with two pieces
0:23
of news and a brand new
0:26
mini cautionary tale for you. News
0:28
headline number one. We're back in
0:30
the studio recording a new Monster
0:33
season of fourteen episodes
0:36
of Cautionary Tales. There will
0:38
be wonderful actors, hilarious calamities
0:40
and tragic catastrophes Cautionary
0:43
Tales Season two. Stay
0:46
tuned. News headline
0:48
number two. I have a brand
0:51
new book coming. If you're listening in
0:53
the US or Canada, it's out early
0:55
next year and it's called The Data
0:57
Detective. It's never too early
1:00
to pre order, so please consider doing
1:02
that. But if you're listening anywhere
1:04
else in the world, the book is out
1:07
right now and it's called How
1:09
to Make the World Add Up. On
1:11
the surface, it's a book about statistics,
1:14
but really it's a book about how to think
1:16
clearly about the world. And
1:18
if you want to know what the book is really about, gather
1:22
close. I'm Tim
1:24
Harford and this is
1:26
a special mini episode of
1:29
Cautionary Tales. You
1:38
can make an estimate of how many children
1:40
have been born into a Dutch or Danish
1:42
family by counting the storks nests
1:44
on the roof of their house. In statistical
1:47
terminology, it would be said that a positive
1:49
correlation has been found to exist between
1:51
these two things. These
1:56
are the words of Daryl Huff, the
1:58
author of the most popular book about
2:01
statistics ever written, How
2:03
to Lie with Statistics. But
2:06
of course, storks do not actually
2:08
deliver baby is, no matter how
2:10
strong the correlation may be. Half
2:13
explains, big houses attract
2:15
big and potentially big families, and
2:17
big houses have more chimney pots on
2:19
which storks may nest.
2:22
Of course, sometimes
2:24
it seems you can prove anything
2:27
with statistics, especially
2:29
things that aren't true. Darryl
2:31
Huff's little book is a master class in
2:33
skepticism. He exposes
2:35
mistakes, half truths, and outright
2:38
lies. Why, he explains,
2:41
the crooks already know these tricks, honest
2:44
men must learn them in self defense. Daryl
2:47
Hulf made statistics seem like
2:49
a game, or perhaps a stage
2:51
magician's trick, all could
2:53
fun, but never to be trusted. And
2:56
I worry that we're starting to trust
2:59
nobody. We're starting to believe
3:01
that lying with statistics is all
3:03
anyone ever does. But skepticism
3:06
can easily curdle into cynicism,
3:09
and cynicism can be corrosive
3:13
after all. In nineteen fifty four, the
3:15
very same year that Half published
3:18
How To Lie With Statistics, two
3:20
researchers, Richard Doll and
3:22
Austin Bradford Hill, produced
3:25
one of the first convincing demonstrations
3:28
that smoking cigarettes caused
3:30
lung cancer. They couldn't have figured
3:32
that out without statistics. Even
3:35
to notice the dramatic increase in lung
3:37
cancer rates in the mid twentieth century
3:40
required a statistical view of the
3:42
world. So what
3:44
Darrell Half saw as a trick, Richard
3:47
Doll and Austin Bradford Hill saw
3:49
as a tool, and a tool
3:52
that has saved many millions
3:54
of lives. Two
3:56
visions of statistics had emerged
3:58
at the same time, and they
4:00
very quickly came into conflict, because
4:03
as the evidence on smoking and
4:06
cancer emerged, the tobacco
4:08
companies fought back. When
4:11
their product was so dangerous, so
4:13
demonstrably dangerous, how
4:16
on earth could they persuade people to
4:18
keep smoking? In
4:21
a stroke of twisted genius,
4:23
they realized the answer scientific
4:26
evidence was the enemy, so
4:28
they had to give people reason to doubt
4:30
that evidence without attacking
4:33
it head on. They chose to
4:35
emphasize uncertainty, the
4:37
fact that experts disagree and
4:40
to call for more research. Before anyone
4:42
did anything hasty, they
4:45
manufactured doubt,
4:48
and it turns out that doubt is
4:50
a really easy product to make. A
4:54
couple of decades ago, two psychologists
4:56
carry Edwards and Edward Smith, conducted
4:59
an experiment to test how people
5:01
felt about all sorts of politically
5:04
fraught questions, including
5:06
abortion rights, smacking children,
5:09
and the death penalty for under sixteens.
5:12
They invited people to produce as many
5:14
arguments as they could in favor
5:16
of or against the conclusion, given
5:19
three minutes per topic. Of
5:22
course, people found it easier to produce
5:24
arguments that backed up what they already believed.
5:27
Lots of researchers have found that, but
5:30
more strikingly, Edwards and Smith found
5:32
that disbelief seemed
5:34
to flow more fluidly than belief.
5:37
It's much easier to argue against
5:39
positions you dislike than
5:41
in favor of positions you support.
5:44
Doubt has a special kind of
5:46
power, and it is the easiest
5:49
thing in the world to suggest that you
5:51
should doubt what the statistics are telling
5:53
you. This is why the simplest
5:55
move in any argument today is
5:58
to encourage people to believe
6:00
nothing doubt
6:04
is a powerful weapon. Now,
6:07
statistics are a natural rural target,
6:10
but honest statistics need to be defended
6:13
because while it's easy to lie with
6:15
statistics, it's even easier
6:17
to lie without them. And there
6:19
are plenty of things in this world, from
6:21
the link between smoking and cancer to
6:24
how to deal with the coronavirus epidemic,
6:27
that we simply can't begin to understand
6:29
without statistics. In a
6:31
complex world, they are an
6:33
essential tool. In
6:36
the spring of nineteen sixty five, a
6:39
US Senate committee was pondering
6:41
the life or death matter of
6:43
whether to put a health warning on packets
6:45
of cigarettes. An
6:47
expert witness appeared, and he
6:50
wanted to throw doubt on the scientific
6:52
evidence, so he
6:54
turned to the topic of storks
6:57
and babies. It turns out
6:59
he said that households with more children
7:02
also have more storks on their roofs.
7:05
That old story about babies being
7:07
delivered by storks was true,
7:09
the expert went on, of course
7:11
it wasn't. Correlation is
7:14
not causation, and
7:16
similarly, just because smoking
7:19
was correlated with lung cancer didn't
7:21
mean, not for a moment, that
7:24
smoking caused cancer. Do
7:27
you honestly think there is as casual
7:30
a relationship between statistics linking
7:32
smoking with disease as there is about
7:34
storks, asked the committee
7:37
chair. The expert witness
7:39
replied that the two seemed
7:41
to me the same. The
7:44
witness's name was
7:46
Darryl Huff. He had
7:48
been paid by the tobacco lobby to do what he
7:50
did best, weave together witty
7:52
examples, some statistical savvy,
7:55
and a certain amount of cynicism to
7:57
cast doubt on the idea that cigarettes
8:00
were dangerous. He was even working
8:02
on a sequel to his masterpiece, although
8:04
it was never published. The sequel's
8:07
name was How to Lie
8:09
with Smoking Statistics.
8:13
I understand that statistics can seem
8:15
intimidating, and they are often
8:17
used deceitfully, but we mustn't
8:20
give way to cynicism or
8:22
feel helpless under the deluge of numbers.
8:25
Understanding statistical claims doesn't
8:27
require a deep technical background.
8:30
If you're able to think, then you're able
8:32
to think sensibly about
8:34
statistical claims, which
8:38
is why I've written a book about how to think
8:40
more clearly about numbers and about
8:42
the world. If you're listening in the US
8:44
or Canada, that book is called The Data
8:47
Detective, and it's out early
8:49
next year, and if you're listening anywhere
8:51
else. The book is called How to Make
8:53
the World Add Up, and it's out
8:55
now as a book, an e book, and as
8:58
an audiobook read by me. If
9:01
you like what I do here on Cautionary Tales,
9:03
I really think you're going to love the book. The
9:07
world is an amazing place, and
9:09
these days, whatever you're interested in,
9:11
the chances are you'll understand it much
9:13
better through the lens of statistics.
9:17
Don't be cynical, don't assume
9:19
it's all a lie or a trick. Don't
9:22
be afraid to pick up this statistical
9:24
telescope and gaze around.
9:32
This mini episode of Cautionary Tales
9:35
was written and read by me Tim
9:37
Harford. The producer was Ryan
9:39
Dilley, and the music and sound design
9:42
by Pascal Wise. Cautionary
9:45
Tales is a Pushkin Industry's
9:47
production. I
10:10
don't don't be
Podchaser is the ultimate destination for podcast data, search, and discovery. Learn More