Nate Silver (full) | Conversations with Tyler

Nate Silver (full) | Conversations with Tyler


TYLER COWEN: Nate doesn’t need much of an
introduction. He is a phenomenon in the areas of data, sports, politics, online media—all
the growth sectors, basically. When I think of your work, I think of you as dedicated
to the idea of numbers and data and wanting to apply that to as many different areas as
possible. If you had to say of all the areas of human
life, where can data bring the biggest improvements, what would your answer be?
NATE SILVER: That’s a pretty heavy question. I should have taken half an hour to think
about that. The answers are probably obvious in some sense where health is an area where
I’ve not done a lot of work personally, but I’m sure it’s incredibly valuable.
Doctors are not known for being terribly analytics driven. I don’t know the culture enough
as to know why. In terms of areas that I would like us to
focus on at FiveThirtyEight a little bit more than we do now, criminality and criminal justice
is an interesting area, in part because you have lots of issues with data. If you want
to know how many police officers are killed, how many people are killed by police officers,
you don’t really know that very well. Education is an area where I suspect you have
a lot of data used poorly as well as data used well.
Urban planning is something we’re fascinated by. We did a big analysis of Uber data that
New York City spent like $2 million to conclude, but we concluded on our own in a week or two
of work, which is that Uber, by and large in New York, was not adding cars to the streets,
at least not in Manhattan. COWEN: We’re in a law school right now.
If we applied a lot more data to the law, what kind of improvement could you imagine
we might come up with, just tentatively? SILVER: See, I think that might be the last
field where— [laughter]
SILVER: Where you would have a lot of—and I don’t say that in a pejorative way at
all. But a lot of the advantage of working with data sets and becoming more adept at
it is that you get an answer that’s at least approximately right.
Whereas, the legal sector, I think, relies more on precision. You want a very precise
and possibly wrong answer, which is what you’re trying to avoid sometimes when you’re doing
statistical analysis. COWEN: I sometimes wonder how much data do
people want. As part of my prep for this, I went back to your high school yearbook,
and I took a look at the quotation you left. It’s from Macbeth.
It goes as follows: “Then the liars and swearers are fools, for there are liars and
swearers enough to beat the honest men and hang up them.”
SILVER: It’s a little self-righteous, but you’re entitled to that when you’re in
high school, I think. [laughs] COWEN: I have no objection to the sentiment,
but I’ve read papers which show when you give a lot of people the chance to view the
quality of their hospital or doctor, they’re not interested. As a citizenry, how much data
do you think people want? Do you think it’s a kind of entertainment
where sports, betting, politics, that kind of horse race, it’s fine, but real data,
do people want to see data on how good or how honest they actually are, or is it more
like the Macbeth quotation? SILVER: I’m like that. My partner got really
into 23andMe and wanted all this detail. They don’t actually tell you all that much. But
I don’t want to have to stress about a bunch of things that I can’t necessarily effect.
I don’t know. The notion of empowering people to make better
decisions with their own health is a noble notion. I guess I’m enough of a free marketer
that I say you should give people the information whether they use it well or not. It’s their
right to have it. I’m not sure I have a firm conclusion about whether it leads to
better decisions or not. Again, my impression is that among doctors
and hospital administrators, they’re not terribly data driven either, despite their
obviously rigorous work in other respects. COWEN: I think of you as a superforecaster,
to use Philip Tetlock’s term. Do you think you can beat prediction markets? Not all the
time, but a smidgen above average? If this were a game and we were all investors, at
the end of 30, 40 years, you’d have some excess returns, just like Cliff Asness, one
of our earlier guests. SILVER: [laughs] Maybe by a very small amount,
but not by enough to make up the variance. It depends on what market you’re talking
about. The markets in politics are not all that liquid and not all that sophisticated,
necessarily. I know the various sports algorithms that
we have at FiveThirtyEight have tended to beat Vegas—not by a lot, but we’ll win
52 percent of the time and so forth on average. COWEN: That is a lot.
SILVER: It is kind of a lot. This is what I spend a lot of my time thinking about, the
dynamics between how markets can be—it’s amazingly arrogant, in some sense, for anyone
to think that they can beat markets. [laughs] At the same time, the more worshipful we become
of markets, then the less useful they become, as well.
A lot of time I’ll have people say, “Well, I know, say, Donald Trump’s gonna win, because
he’s up to 52 percent.” He’s at lower than that now, but at one point, he was up
to 55 percent to win the GOP nomination at Betfair. That doesn’t really add any value
to the conversation. I’m more interested, as a person and as
a researcher and journalist, in providing information that then other people can aggregate,
as opposed to the other way around. If you talk about, “How good are political markets?,”
the first question is, “How good are markets?” The answer to that is “pretty good,” but
when the distribution of error is not very linear, when they’re off, they can be off
by a lot. Are you the person who knows when they’re off? That’s harder to do, potentially.
COWEN: What are the differences between forecasting and futurism, and do you have any predictions
for the year 2050? They don’t have to be great. They just have to be better than the
market. We’ll take a 52 percent prediction and go home and celebrate.
SILVER: I’m mildly pessimistic in some ways. COWEN: What’s the biggest source of your
pessimism? SILVER: [laughs] There’s probably some survivorship
bias in the United States, and thinking about how our way will persevere forever and ever
and ever. We were talking backstage about how you go to Asia and I go to Asia—not
as often as you. If you want to feel optimistic about civilization, then go there.
Some of it is thinking about, frankly, this Donald Trump phenomenon.
COWEN: I’ve heard of him. SILVER: Yeah. It just made me consider that
a lot of assumptions a lot of people made about how American politics work are really
based on a relatively narrow slice of history, post–World War II through 2000 or so, maybe
even briefer, 1980 through 2000. It’s not really a lot of history.
In many other contexts, there are all types of places around the world where nationalism
is a much bigger phenomenon than it is in the United States. Race and racism is embedded
in a great deal of political turmoil in the United States.
In some ways, I kind of wondered after the Great Recession, “How come we haven’t
seen more social upheaval?” Maybe we’re seeing that a little bit delayed. It’s more
of a revolution of rising expectations. At the same time, there is such a tendency now
to focus on—in politics, people focus on a very small number of stories that are not
representative of the big picture. There is a lot of wonderful news in the world
in terms of poverty rates going down globally, income inequality going down, diseases being
eradicated, but I wondered, to some extent, how much the media culture tends to focus
a lens on negative aspects of society, lower people’s happiness level, and all this type
of stuff. COWEN: More optimistically, how about love
and sex? Do you think data can improve matching? Should we just follow the algorithms, or do
you think that’s a perpetual dead end and all the algorithms really do is force you
chose someone, give you a phony reason and get you out of your indecision?
SILVER: I think that— [laughter]
SILVER: The market would say that people find online services fairly useful. Maybe it removes
some spontaneity. I met my partner at a bar, which almost feels old-fashioned now.
COWEN: But it could be like these pills that are sold, the online services, like a placebo
effect. SILVER: There’s a lot of over-optimization.
That’s a problem across almost any sector you’d want to talk about where data is being
used. You’re optimizing for a short-term equilibrium, and it’s much harder to measure
the long term. Before, if you couldn’t measure anything
at all, then maybe your heuristics aren’t that bad. But if you can say, “What’s
going to make me really happy tomorrow? In my business, what’s going to get my website
the most traffic 36 hours from now?,” isn’t necessarily the best decision in the long
run. You can measure the short run and not the
long run, and measure some things, not others. That can make you quite myopic and is a bigger
problem than people realize, perhaps. COWEN: You mentioned Donald Trump a moment
ago. I had told quite a few people I didn’t think Trump could get very far. It’s not
obvious that I was right. [laughter]
COWEN: Paul Krugman said pretty early on that Trump had quite a good chance. What is it
that Paul Krugman saw that I didn’t? SILVER: I was one of the Trump skeptics, too.
Let me say I thought you’d ask a version of this question.
COWEN: But I wasn’t allowed to blame you, so I put it on myself. I’m still not sure
you were wrong, but there’s something we didn’t see.
SILVER: That is important, I think. I got a little frustrated, because a lot of people
were saying, “Trump’s instantly going to evaporate in the polls.”
If you go back and look at what we wrote, we said, “That could happen, but there are
also a lot of candidates—Pat Buchanan, and so forth, Ron Paul, Rick Santorum—who will
get 20, 25, 30-something percent of the electorate, and we have a high-floor, low-ceiling type
of candidate.” That could still wind up being true.
With that said, for one thing, we’re dealing with a fairly small sample of relevant elections.
People look at, in the primaries, going back to 1972. One very basic lesson is that when
you have a sample size, let’s say it’s roughly 15, there’s nothing you can do to
make it not a sample size of 15. No matter how compelling you can make your
rationalization to say, “Well but, you know, we have theory as well as empirics here,”
still, 15 cases is 15 cases. I think maybe making people more cautious about saying “Unlikely”
versus “Never.” Now, the record will show we said “Unlikely” and not “Never,”
but still it’s a lot of things to think about.
You talked about what superforecasters are supposed to do—
COWEN: That’s you, yes? SILVER: Yeah. You start with priors. You can
say, “The prior is that candidates like Donald Trump tend not to win the nomination.
So what signs could I find that would violate that assumption?” It’s not necessarily
performing well in early polls. Lots of candidates who are flashes in the pan—I guess it’s
a little tautological. Lots of unusual candidates have done well
in early polls. Lots of unusual candidates have won Iowa or New Hampshire, not usually
both, but one or the other. It’s the ability to consolidate the field after that, by becoming
the consensus choice of the party that’s been more unusual. That assumption still might
prove to be true. I did think though that I, and a lot of people,
overrated the ability of the Republican Party to stop what I think is, in some ways, a radical
insurgency within the GOP. COWEN: The party’s weaker than you thought.
What other judgments about the world do you feel you or I should revise that we once held?
Paul Krugman I think would say, “Republicans are more racist than many people believe.”
It wouldn’t be my take in particular, but it’s a candidate.
SILVER: To be honest, that’s a little bit of what I wanted to resist. Actually, I think
one lazy heuristic that, in my thinking about Trump I use is (there are exceptions—Paul
Krugman, Norman Ornstein—who have been very consistent for a long time), but I thought
the people who were pro-Trump were generally not people whose opinions I would weight as
highly. And I think that’s lazy and possibly quite dangerous.
COWEN: If he chooses to run, does Michael Bloomberg have a chance?
SILVER: Give me a probability. COWEN: If I go to the betting markets, this
morning I think I saw 2.8 percent. Is that too high or too low, relative to—
SILVER: A 2.8 percent chance of becoming president? COWEN: President.
SILVER: That’s probably about right. Obviously, in some ways, the climate could be as fertile
as ever for some type of third candidate running, but Bloomberg, I don’t know. Number one,
I’m not sure he differentiates all that well from Clinton, with whom he has a lot
in common policy-wise, and Trump, with which he’s kind of the same character.
But the most basic problem is that in an election between Sanders and Trump or Clinton and Trump,
everything is quite left of center. Trump, when he was thinking about running as an independent
in 1999–2000, had an eccentric platform. It involved single-payer health care, a wealth
tax. He was anti-immigration, even then, but pro-choice.
He said explicitly, “I’m not bound by any party, really. I’ll probably reconsider
my stances if I become the Republican nominee.” To me, the more viable candidate in that case
would be a Mitt Romney–Condi Rice ticket or something like that.
COWEN: Let’s move past the esoterica and give the people what they really want to hear.
Let’s go back to 1968, in the World Series, the manager of the Detroit Tigers, Mayo Smith,
he took Mickey Stanley, pulled him out of center field and put him in at shortstop so
that Jim Northrup could play center and the recovering Al Kaline, who was the better hitter,
future Hall of Famer, could play in right field.
No one had ever done this before. You’re advising at that time. How do you start thinking
about that problem? You know what I’m talking about, right? You’re from Michigan, of course
you do. SILVER: I do know what you’re talking about.
We know a lot more now about the value of defense than we did in the late ’60s or
even than we did 10 years ago. If anything, defense turns out to be quite a bit more important
than people would have thought. It was an ironic way when some of the conventional scout
wisdom was confirmed, as data got more advanced and more sophisticated.
If you want to get really complicated, you say the Tigers had Denny McLain, Mickey Lolich.
They had a strikeout-heavy pitching staff. Maybe worry about defense a little bit less.
Tiger Stadium is a park conducive to low batting averages to begin with, a little high home
run totals. It’s not obvious to me that it was the best
move, but it is interesting that, all of a sudden, baseball teams and football teams
become, in general, more strategically correct when they have more on the line.
In the World Series, closer usage is a lot better, when you bring your best guy in in
the eighth inning or the seventh inning. You see in the NFL teams will go for two more
often in the playoffs, go for it on fourth down more often in the playoffs. Which is
a hint that when the stakes are low, culture tends to prevail. When the stakes are high
and the outcome of the game is all that matters, then things are different.
COWEN: Given your view on Mickey Stanley and the Detroit Tigers, who is the underrated
candidate in the Republican race this year? [laughter]
COWEN: Just to impose a kind of consistency on you.
[laughter] SILVER: I don’t know.
COWEN: You’ve got to go long on someone. You have a hundred dollars to bet.
SILVER: I think the markets are fairly close to correct right now. But I’ve been a Rubio
optimist for a while, on the theory that he is the only candidate who really has appeal
to all the various sectors and constituencies within the GOP, which may be a fraying party,
but still, he has the highest favorability ratings in the party.
He speaks the language of conservatives without being too extreme, but the big question in
the election right now is, “Where is Trump’s ceiling?” If you start out at 25 percent,
like he did in Iowa, that’s one thing. Thirty-five percent is much closer to the
point where he’d be hard to stop, but even now, you see in New Hampshire, even though
he won with 35 percent of the vote, half of the Republicans there said they would not
want him as their nominee, so the question is, “Can the non-Trump candidates organize
themselves into one candidate?” And then does he stop at 35 percent, or 40, or 45,
or 51? If he stops at 51, then it kind of doesn’t matter.
I suppose I think Rubio at—what is he at? Three to one?
COWEN: Something like that. SILVER: I think he should be more at two to
one or something, so not a dramatic mispricing. COWEN: What do you think of the Ted Cruz theory
that this election is not about swing voters, that you actually need a somewhat less compromising
candidate to bring out the silent conservatives who maybe don’t vote, and that there’s
a lot of them, and that Cruz is more electable than Rubio?
You hear this. There’s data on this, and what’s your opinion of that data?
SILVER: We talked about priors, or Occam’s razor. If you look at lots and lots of Senate
races over time—and the reason why that’s relevant is because, number one, senators
are easier to measure their ideology, because they legislate, and number two, we have a
much larger sample size. You can see that there’s a price for extremism.
Not a price that can’t be overcome, if we go into a big recession or if Clinton or Bernie
has huge problems, but Cruz would probably cost you three or four points relative to
the median generic Republican. COWEN: Last I saw, Bernie’s share was at
something like 17 cents. At that price, do you go long or short?
SILVER: Probably short, but I think it’s also not dramatically mispriced. The thing
people miss is that unlike on the GOP side where Trump’s at least passed the first
test—he has people who are out there, willing to vote for him. There was some doubt about
that, especially after Iowa, where he underperformed his polls.
Sanders, we haven’t really seen. Can he win states that are not very white and very
liberal? Maybe he can. Nevada seems to be pretty close. I’m just saying, we haven’t
really received that much information that would make you update your priors about Sanders
all that much. He also probably has to win by, with a little
bit of room to spare. If it’s a tie, then Clinton will probably win on the basis of
superdelegates. If she loses by a couple of points, but it’s close enough—so if you’re
an underdog in a football game, and you lose unless you win by more than a field goal,
it actually reduces your win probability quite a bit.
If you’re the favorite already, then maybe it wouldn’t, but that’s tricky. When underdogs
win, they tend to win narrowly, and if Clinton wins some of the races she should have won
or should have lost—because superdelegates turn a narrow Bernie win by a field goal into,
“Oh, after further review, we’re going to have an overtime quarter. The superdelegates
will weigh in instead,” then I don’t know. I think 10 or 15 percent, somewhere in that
range is probably about right. COWEN: We may come back to politics, but let’s
turn to a nobler endeavor, sports. [laughter]
COWEN: You’re a fan of baseball, and I’d like to ask you, of all the different baseball
records, which is the one that is most impressive to you, or the most a statistical aberration,
and try to stay a bit modern. We both know in 1889, Hoss Radbourn won 59 games.
Start with Wilson’s—was it 36 triples in 1912? That, and up through the modern age.
What’s the most statistically impressive baseball record, and why?
SILVER: I think the biggest outlier is the number of intentional walks that Barry Bonds
drew. I forget what year it was, 2001, where he had like 161 intentional walks, and the
next closest player is 50? [laughter]
SILVER: There is just no other example I can think of in sports where the record holder
has three times the sum total of the nearest highest player.
COWEN: What do you think of streaks? Do streaks impress you more or less than most people?
DiMaggio’s hitting streak, consecutive games played streak, Cal Ripken. Johnny Vander Meer,
two no-hitters in a row. When will that be done again?
SILVER: I’m sure you’ve read this is another area where the simple sabermetric answer might
not have been totally correct, but there was a lot of talk for a long time about how the
hot-hand theory was false, and how basically things are random to a first approximation.
Now there’s more evidence arguing against that, and in fact, it’s a classic mistake,
but if you have a test that has low power, then you may mistake an ambiguous result for
a negative result instead. It appears now that there is some streakiness,
as you would expect. You would expect that there is some variation in human behavior
from day to day. It’s kind of amazing that being professional athletes, there’s less
streakiness than you might think, but still—and now we’re getting, actually, data, too,
that’s less noisy. We now get data, for example, for Major League hitters—how hard
the ball comes off their bat. We have some unpublished research that a colleague
of mine is doing, but it looks like you can maybe predict batting average up or down,
or on-base average 20 or 30 points from a baseline, which in baseball terms is pretty
relevant. If you have a guy hitting a leadoff spot because
he’s a 0.370 OBP hitter, and he’s really a 0.340 based on his current condition, then
he should maybe be demoted down to the eighth spot in the lineup instead. This is true of
a lot of things where the first cut from data is overly simplified. You refine over time
to something which is a little bit more nuanced. COWEN: What do you think is the strongest
piece of data-based evidence we have that sports analytics work? Let me give you an
example. As you know, the Houston Rockets are run by
Daryl Morey, a numbers-intensive guy. He’s from MIT, seems to be super smart, and right
now, the big debate in Houston is which of their two star players they should trade,
or maybe both. They may not even make the playoffs, so how good is the best regression
showing data analytics even work in sports? SILVER: The Golden State Warriors might be
one of the best examples, where— COWEN: But, examples. Let’s say it increases
your variance, so the good examples will look really good, but as a predictor, how hard
should we be selling it? What’s the average impact, the marginal impact?
SILVER: You talked a little bit before about how hard it is to be marketed, so there’s
a little bit of this, too, in sports. Actually, a colleague of mine, Ben Lindbergh, just wrote
a book that’s not published yet, but that I was reading about. [The Only Rule Is It
Has to Work is available in May.—Ed.] He and Sam Miller, another baseball stathead,
were given, basically, ownership of a minor league team in Sonoma, California, for a year.
They kind of had carte blanche, but they encountered baseball culture in a very head-on way, and
yeah, the team did pretty well. It was a very obscure minor league, but if the baseline
is making half of your decisions right, and you make 55 percent of your decisions right,
or 53 percent right, because you’re using analytics, that’s a pretty big gain at the
margin. Yet, in a sport as noisy as baseball, it’s
going to take a lot of time to show up. Basketball is less noisy, I will grant you, but in general,
I think the Spurs are fairly analytics friendly, and the Rockets and the Warriors. Basketball
is probably the best example. COWEN: Let me ask you about a sport which
I find totally baffling, soccer. [laughter]
COWEN: There’s not much of a natural time unit. That’s why it’s hard to squeeze
in commercials, and not many points. There are assists, not many. They’re not always
well defined. Defense, my goodness, more confusing to this
American than cricket, yet market salaries of soccer players are determined. There’s
some papers on market salaries. The small amount of data we have seem to predict those
salaries very well. The fact that soccer behaves in normal ways,
does this mean A, we can still get it all done with limited data, or does it mean B,
having more data doesn’t really help us that much, and all of sports is a bit like
soccer, and we’re ultimately just throwing up our hands and saying, “My goodness, it
might as well be cricket”? SILVER: I feel like I’m answering all your
questions the same way, which is analysis is far from perfect, it’ll make lots of
mistakes. At the same time, pretty good is hard to beat, [laughs] sometimes.
The heuristics that develop over time about how to value players at different positions
in soccer—they kind of appropriately don’t value goalkeepers all that much, which the
analytics seems to bear out, for example. I do think, though, in soccer, that we’re
at the very early stages, and partly there hasn’t really been very much data collected
at all. It’s not like in the NBA, where you had blocks and steals, very crude defensive
stats. Literally, and we built the system for ESPN a few years ago, where the only data
that’s been kept in the long term in soccer is goals and bookings, red cards and yellow
cards. It’s not like we have assists, until recently.
It’s not like we have tackles. You can maybe get time on the pitch if you parse play-by-play
records. We don’t have crossing passes. I think there’s still a lot of room for
upward improvement in soccer. COWEN: We live in a global economy, with billions
of laborers. Why don’t more of them learn the knuckleball? Wilbur Wood, Hoyt Wilhelm.
They were not athletic. The pay’s pretty high. R. A. Dickey won the Cy Young award
in 2012 with a knuckleball, which he taught himself.
People put labor into a lot of endeavors, but why so few knuckleballers? Why isn’t
that a more regularized statistical process? Why so lumpy?
SILVER: I have two answers. One answer is that there are diminishing returns in the
number of knuckleballers in the league, so when you have a second knuckleballer in the
National League, when you already have R. A. Dickey, that already discernibly affects
the success of both pitchers. But I think there is a second thing, which
is that sports tends to engender conformism. A lot of walks of life do, and that’s the
whole tension, again, that comes up, is on the one hand, knowing that the market is usually
pretty good, but on the other hand, that there are powerful biases to conform.
COWEN: Trump is like the knuckleballer of politics then?
SILVER: Yeah, that’s not a bad— [laughter]
COWEN: He throws the knuckleball at Jeb Bush. Jeb is baffled. Maybe now he’s finally coming
up with a response. [laughter]
SILVER: The other kind of statistical way to think about it is that when you have something
which is unusual, then there’s a bigger error bar around it, so you could say, “Well,
I think Trump is, on average, going to get less of the vote than someone who’s much
higher than him in the polls,” but he has a much longer tail on the other side, and
so therefore that’s one reason to not be dismissive of him, at least in the early going,
when everyone is kind of a long shot, at least. COWEN: In all of these interviews, in the
middle segment, we do a little game. It’s called “Underrated versus Overrated.”
I name a few things, and you tell me if you think they’re underrated or overrated, and
you’re free to pass on any you don’t want to have an opinion on.
New York City, the Upper East Side. Overrated or underrated?
[laughter] COWEN: Are they really that happy in Seinfeld,
and how did they afford that apartment, anyway? SILVER: I think a little overrated. But like,
but look, New York is actually extremely efficiently priced.
We did an analysis for New York Magazine a few years ago where we tried to say what’s
the best neighborhood. The problem is that cost accounts for an R-squared of like 0.93
with our quality index or something. It’s really hard to improve your lot in New York,
at least at the neighborhood level. But I don’t know, it’s hard to find good
hole-in-the-wall places to eat in the—although there’s some ramen shops that are getting
better. I, like, basically, judge everything by food.
COWEN: The shortest avenue in Manhattan, Fourth Avenue, it’s what, six blocks long.
SILVER: Yeah. COWEN: Overrated or underrated?
SILVER: I think underrated. You probably saw my list a few weeks ago where, it’s short
but it’s in a very dynamic section of town. COWEN: Even with all the used bookstores closed
down? SILVER: Length isn’t everything.
COWEN: Length isn’t everything, OK. The idea of legalizing drugs. Overrated or
underrated? SILVER: By this crowd, probably rated properly.
I mean, I don’t know. I’m enough of a lowercase L libertarian where I think that
the government ought to have a stronger reason to intervene in choices that people are making
instead of a lesser reason, necessarily. To me, it clearly makes no sense to treat
marijuana as being a more serious substance than alcohol, for example. I don’t think,
in my heart of hearts, if I were running for office or in the Senate or something, that
I would vote for a bill to legalize heroin or cocaine, but decriminalizing it, perhaps.
But I don’t know. I think the consequentialist case, for a long time, probably underrated
and may have gone a little bit too far in the other direction. But again, I would say
if it’s close, then you give people the choice.
COWEN: The musical group My Bloody Valentine, they put out an album in 2013 after almost
a 20-year hiatus. Few people expected this album. It was called MBV. Overrated or underrated?
SILVER: The group or the album? COWEN: The album. We know the group’s underrated,
right? SILVER: Yeah, the group’s one of my favorite
bands. COWEN: Sure, of course.
SILVER: I think the album was properly rated. COWEN: Singapore. Overrated or underrated?
SILVER: Underrated except by you. [laughter]
SILVER: But I saw exactly why you would like it. Right? It has great food, it’s like
a little laboratory experiment. It’s a fascinating—we were talking before
about how Singapore is a city that, we hypothesized that if you have a few constraints that might
seem slightly strange, then maybe having a few strange constraints is helpful.
When you talk about the strange—my sister lived in Germany for a while. She’s like,
“Well, yeah, if you’re there in the shop and they’re going to close at four, it doesn’t
matter if you have a huge line of groceries, the store will close down and you have to
put your stuff back, and ‘Sorry, you can come back tomorrow,’” which kind of seems
irrational. But Germany has weird little quirks and constraints
and yet seems to be doing, I don’t know, fairly well in certain ways, or Scandinavia
or something. If you give up a little bit of freedom to
have more freedom, it’s like a Björk lyric, which is, “I thought I could organize freedom,
how Scandinavian of me.” Singapore feels a little bit that way, too.
COWEN: I have a big compound question. It’s about a few things. Feel free to touch on
it what you want. But one of them is sports, one is fantasy sports, and one is gambling.
They’re all interrelated. What’s your take on—what do these actually
do for us? How socially productive are they? You’re of a mixed mind on drug legalization.
I could ask a comparable question about gambling, either legalization or liberalization. But
what’s the net social externality on this mix of sports, fantasy sports, gambling?
What’s your view? I know it’s maybe fun for you.
SILVER: I’m biased, I make my living as a—you know.
One of the things about the regular fantasy football league that you’re in with your
friends is that you get a lot out of it. You get to hang out with people you might not
see very often, especially as you get older, into your 30s or 40s or whatnot. You get to
watch a lot of games with more of a rooting interest.
The thing about daily fantasy sports is that a lot of that is really taken away. It’s
very much a brute force approach to watching sports.
I did this for a few weeks and then got bored with it. But basically I had a program that
would randomly generate high-scoring lineups. Then you scrape that data, and you load it
up, 200 lineups at a time, and it kind of took all the joy out of it.
But that’s not quite what you’re asking, really. Right?
COWEN: Right. [laughter] What if people just watched sports? Like,
I watch sports. If it’s a good game, I enjoy it. I don’t feel I need to gamble on it.
I don’t want rotisserie, I don’t want—I want to read analytics and I want to read
you or your people at FiveThirtyEight on the game, which is great, and then watch the game
and I’m done. If they’re good players, I’m happy.
What am I missing, if anything? SILVER: For one thing, gambling and fantasy
sports are a good way to teach people applied analytics. I’m not being jokey, like, I
think this has probably a measurable benefit to society.
But look, it’s the case where, unlike drug legalization—where there are not a lot of
countries where drugs apart from marijuana are even there fairly rarely—worldwide,
people are much more relaxed about gambling, and it’s normalized.
You can go to the betting shop, any Ladbrokes anywhere in the United Kingdom and place a
bet and it doesn’t seem to ruin society. Maybe you have in low-paying leagues or in
tennis, the occasional betting scandal which is not great. But I think it’s a way for
a substantial number of people to enjoy sports and develop critical thinking skills.
Again, I say if it’s close, let people do it, and I feel that way about gambling. But
in that case, you do have examples of many, many Westernized countries where spending
on sports is legal, and there seems not to be, at least, grave societal harm.
COWEN: You Run the website, FiveThirtyEight. It’s your vision, you founded it, you developed
it. You took it to ESPN. Over those years, what would you say is the most important thing
you’ve learned about managing? SILVER: Basically there are three strategies,
three fundamental strategies of management when you have a disagreement with something
that your, one of your employees is doing. One of which is you can give up. Right? You
can say, “Well, I’m not going to pick this battle to fight, because there’s a
consequence to lowering this person’s morale, or I’m tired, I have other issues.” So
you can capitulate. Number two, you can fiat. You can say, “Well,
sorry, but I’m ultimately the one who signs your checks,” or my boss signs their checks,
but, “This is the line of authority, and we are not going to publish that article.
I’ll explain my mind later on.” Number three is you can try and persuade instead.
Which sounds perfect except persuasion is really, really time consuming.
Figuring out which ones of those three tactics to use and in what ratio is important. I actually
found though, overall, that there’s like a little bit more value in micromanagement
than I thought. Not about everything, but strategically saying, “I’m going to spend
a lot of time going into detail on this one.” Or, I guess it’s just mentoring, I guess,
is a way to put it. COWEN: Which sports coach or manager are you
most like? Vince Lombardi, Gregg Popovich? Who do you draw inspiration from? Do you think
about it in these terms? SILVER: I’m not arrogant enough to compare
myself to Popovich. But I’m like, I’m laissez-faire, but when I weigh in on something,
I’ll weigh in pretty directly. I think you do have to pick your battles a little bit.
You have to hire really well. But it’s a culture of creatives and a culture
of journalists, and journalists are strange and wonderful people, and data journalists
are still journalists too, but you have to kind of trust people to make their own decision.
A big thing, too, is kind of figuring out which one of my deputies, the other managers
and editors on the staff, what’s my agreement ratio with them. It’s incredibly valuable
to have someone who, without your intervention, agrees with you 80 percent of the time. Then
the 20 percent of the time that they don’t agree with you, that they’re right as often
as not. If it goes to 95 percent, then they’re a
sycophant, and it’s probably bad. If it goes to 60 percent, well, then, you might
as well do the work yourself. Figure out the people who will listen to you, but also challenge
you at the right times. COWEN: You mentioned food before. Let’s
take a data intensive approach to food. You’re trying to find a good place to eat. What is
the underrated statistic about a restaurant that you will consult or advocate others consult
in this endeavor? SILVER: This is a fairly basic one. But I’d
rather look, if you’re looking at Yelp or TripAdvisor, the number of reviews is a better
signal than the average star rating. Especially the number of reviews relative to how long
a place has been open. We’ve done some work on this, too, where
when you’re drawing from a more diverse segment of people—there’s something I
want to vent about how, like, every book on Amazon, in the long run, gravitates toward
having four stars. A lot of 9/11 conspiracy books are rated pretty well on Amazon, because
only the conspiracists bother to read them. Whereas, Othello or Macbeth or something,
everyone reads, a lot of kids have to read it for homework when they don’t want to,
necessarily, so they’ll leave a bad review there, potentially. But I think that problem
is more acute than people might realize when it comes to restaurants, where a place is
notorious for drawing people who might not like that cuisine as much.
People also, when I go —and I used to do more Yelping and stuff like that—if I go
to a mom-and-pop place in a small town somewhere and it’s not very good, there’s almost
no way that I’m going to leave a negative review for that place.
I don’t want to hurt anyone’s feelings, I don’t want to—there’s actually studies
showing that Yelp reviews can, like, a one-star Yelp review can cost like thousands of dollars
in business for a restaurant that has under 50 reviews or something like that.
COWEN: In New York City, is there better food on the avenues or the streets?
SILVER: I read you on this. I think the—but New York is weird, because there are really,
there are kind of three New Yorks from a culinary perspective. Right? There’s rich Michelin-starred
New York, there’s kind of hip Soho and Williamsburg New York, and there’s ethnic New York, for
lack of a better term. Making sure that you have a mental list of
places from all three types of those. There’s some rules that work well in one of those
lanes that don’t work well in the others, necessarily.
In the very high-end restaurants of New York, it’s like so competitive that I think your
rule about, order the weirdest thing on the menu, I think there are parts of New York
where that probably isn’t true, because it’s so hyper competitive that the menu
couldn’t afford to lead people astray. Sometimes the thing that the menu is very
clearly pointing you toward in New York is the kind of thing that you would want to order
instead. That might not be true if you go out to Queens or something like that, where,
frankly, pound for pound, probably the food is better than Manhattan or Brooklyn.
But yeah, you could write a whole book and maybe I will, especially if we have Trump
win the election or something, maybe I’ll write a whole book about heuristics for eating
food in New York. COWEN: I said before and I’ll tell the whole
crowd, one of my dreams is that someday you write the quantitative history of New York
City. This would be one of my favorite books. Question about the weather reporting. There’s
some evidence that there’s what is called wet bias, so storms are over-forecast. Why
is that? Is this even true? SILVER: It’s true the further downstream
you go, so the local meteorologist here in Virginia or in Washington on TV, they want
to get higher ratings. COWEN: They’re trying to scare us.
SILVER: They’re trying to scare you, yeah. COWEN: It’s like they want the Iraq War,
so to speak, so that people turn on CNN. SILVER: The irony though is the data the government
produces is very well calibrated and doesn’t really have a wet bias. There are a few individual
like models for winter weather that do. But I don’t know, it’s been kind of interesting,
in my shoes, going from someone who was a total outsider, to someone who has more reputational
risk. To a first approximation, I think it might
make someone a worse forecaster, potentially. By the way, another thing about the Trump
thing I’ve been thinking about is—so my early view, that Trump had a very low chance—not
zero, but very low—of winning the nomination was not based on any formal model, per se.
I wonder what if I had even like a fairly bad model instead?
The good thing about building a statistical model is that it commits you to rules, right?
Instead of just kind of saying, “Well, early polls aren’t very predictive and your prior
is it currently probably won’t win, therefore, probably not.”
It pins you down and says, “Well, OK, early polls aren’t predictive, but at what point
do they become more predictive?” When Trump went from being at 25 percent in the polls
to 35 percent after Paris and San Bernardino, how significant is that?
To have an answer that is set up by an algorithm you designed ahead of time is actually maybe
more helpful than people would think. The long way of saying this is that I’m
not sure that I’m any better than the average pundit unless I have a model. The disciplining
effect of a model, doing your thinking in advance, and setting up rules of evidence
is probably quite important. COWEN: I have a question about the economics
and sociology of sports. This has puzzled me for a while. You may have thought about
this. But I’m struck by the relatively small number of professional athletes who have come
out as being gay. In Hollywood it’s a lot of people, even
in Washington, which is a very conservative town, I wouldn’t say it’s a lot of people,
but it happens in a quiet kind of way. In sports, why is there so little? If we applied
some kind of economic or statistical model, in which sports would you expect to see the
new breakthroughs coming when they come? SILVER: I’m sure there are a lot of athletes
in the closet. I don’t assume that it’s four or five percent or whatever the population
average is. I assume it’s a fair amount lower than that, but I don’t know, I think
people forget about how much the economics change when you’re talking about people
who are in the 0.001 percent of something. Where the fact that, until fairly recently,
until maybe a few years ago, and in many parts of the country, obviously, still now. Until
fairly recently, growing up gay is something that was, if not traumatic, at least required
a lot of bandwidth, it requires a lot of energy. Because the fact that, for example, there’s
data from Freakonomics about how hockey players who are born in January, just because they
start a little bit earlier than their peer group, that’s a very powerful effect versus
being born in November or December instead. If something that minor can have that profound
an effect, where I don’t know if it’s twice as many NHL players from January as
December, then something as important to your identity as being gay in a society that until
recently didn’t accept it, that’s a comparative disadvantage.
Maybe there are also correlations on what kinds of skills and traits people have, I
don’t know, but we’ll see. The prediction, if that theory is true, is
that as it’s become more normalized, and now people who are growing up in middle school
and high school where being gay is not as much of a disadvantage. Then you’d expect
from that generation there to be substantially more gay athletes.
COWEN: In which sport will that happen first? What’s the implied prediction? We see a
bit of it in women’s tennis, right? SILVER: Women’s tennis.
COWEN: Individual sports, maybe, over team sports? Yes? No?
SILVER: Yeah, you would think that in tennis and golf you might see it first. The NBA,
where talent is so manifest and one player can make so much difference. LeBron James
could come out as gay tomorrow and I think it would not hurt his ability to get a max,
max plus contract at all. COWEN: But it could hurt endorsements.
SILVER: It could hurt endorsements. He is kind of a high default, but I don’t know.
I think it’s no longer about kind of the marketing side of it so much as the fact that
it just kind of—sports is still a very conformist culture. Some of the reason I might say the
NBA is I think it’s a little bit more individualistic as a culture. Guys are free to express themselves
more. Listen to baseball players talk. They’re
boring as hell. Right? Kevin Durant or something, these guys are smart and they’re interesting
to listen to. COWEN: Kareem Abdul-Jabbar, right?
SILVER: Kareem Abdul-Jabbar. COWEN: Yeah.
SILVER: So I would think basketball might be a sport where you’d see it relatively
soon. COWEN: Let me ask you a general question about
forecasting, and I worry about this in the context of finance. I see a lot of money managers,
so there’s Ray Dalio at Bridgewater. He saw one basic point about real interest rates,
made billions off of that over a great run. Now it’s not obvious he and his team knew
any better than anyone else. Peter Lynch, he had fantastic insights into
consumer products. Use stuff, see how you like it, buy that stock. He believed that
in an age when consumer product stocks were taking off.
Warren Buffett, a certain kind of value investing. Worked great for a while, no big success,
a lot of big failures in recent times. Is it possible the so-called true model is
always shifting, and there’s a kind of selection bias where different forecasters are elevated
and they have their run for three, five, however many years? Then the true model shifts and
what they’re good at isn’t valued. We turn them over and replace them with other
forecasters. As, like, our best forecaster, do you worry
about this? SILVER: Sure. Even if you are skeptical about
the efficiency of markets, if you have a great gig and you’re picking up hundred dollar
bills off the ground, then boy if you can extend that by three or five years without
adapting and evolving, that’s on the extreme high end I think. Three or five years is a
very long and fortunate run. That’s part of why even though now we’re
very immersed in the election cycle, it’s part of why I wanted to make sure that FiveThirtyEight
was not just an election site. We’re going to blow an election sooner or later. We might
blow this one. To be doing a whole diverse array of things both intellectually and commercially
is important. The follow-up to that is, “Are there people
who have the skills to find the next underweighted opportunity?” Maybe, that’s trickier.
I think a lot of people have one or two really good insights, and if you’re very lucky
that can take you a long way. COWEN: Here’s a related worry. It’s clear
in the data, stock market volatility is correlated with itself over time. If you have some volatile
days, you’re likely to get more. That’s pretty clear. That’s another way to say
those returns, for a while, are hard to forecast and stay hard.
This year politically, it’s already a big surprise to me, to a lot of people. Could
it be the case we’re entering a new era where political volatility is higher and basically
all forecasters will just do much worse than they have been doing?
SILVER: It’s possible. Again, I go back in saying what people take to be the equilibrium
baseline condition may actually have been an outlier, instead. You have this relatively
stable, long boom politics and economics from the ‘50s to the ‘90s or the early 2000s,
thereabouts, and that could potentially reverse itself.
Again, looking at examples outside of the United States I think is instructive. Maybe
I’m more of a believer in American exceptionalism than I thought. You see constituencies that
are Trumpian in different parts of Europe and have been extant for a long time, so maybe
America just got really lucky for 50 years. COWEN: Nassim Taleb has a hypothesis that,
in some ways, the world is getting weirder. There’s the example of plane crashes. Planes
used to crash a lot for pretty normal reasons. The engine would fall apart. Obviously, we
invested more resources in making planes safer. I just read in the Wall Street Journal last
year, there were actually zero deaths from jetliner crashes other than terror attacks.
We have strange events like the Germanwings pilot flying into the Alps. Malaysian Air
disappears, no one knows why. The events people talk about, we’re left with only the weird
ones. Do you think we’re headed toward a future
where we’re only going to be talking about weird, very hard to forecast events, precisely
because we get good at avoiding a lot of problems and mistakes?
SILVER: For sure. There’s some stupid metaphor I use in the book, where one of the problems
with comparing how shortstops play for example, is that you always evaluate players who are
on the edge of their range. Can they make the spectacular diving catch?
To the first approximation, everyone is equally good at the edge of their range. The question
is how much territory do they cover in between the nonspectacular plays, that we can miss
potentially? It’s probably more true—one reason why I like when we forecast sports,
is you have a chance to build up your sample size.
A perfectly routine Wizards versus Cavaliers game, where we have the Cavs favorite at home
and they win, that counts. You get hundreds and hundreds of those on the course of the
season, whereas in politics you’re more drawn to the spectacular and the weird events.
A lot of models are good heuristics, when conditions are fairly normal. They don’t
deal all that well with the edge cases, because they’re fully designed, or because you have
nonlinearity, or because they have small sample sizes, or whatever else. How well do models
deal with the weird cases versus other types of heuristics, I’m not sure. Maybe the advantage
is more in the baseline cases instead. COWEN: Other than skilled with data, what
are the personal qualities of good predictors? SILVER: You have to have a certain mistrust
of conventional wisdom, and that’s a tricky thing. On the one hand we know that I’m
not that smart, that this room is way way way way smarter than me, and a market is way
way way way smarter than me. At the same time people are social beings, and they behave
in herds sometimes. This is easier in politics than almost any
other field, because the political press corps literally is kind of a herd. It’s the perfect
example of it. You have a few hundred journalists who travel around together, who are all reading
one another on Twitter, who are all talking to one another.
It’s not 500 really smart people. It’s one or two really smart people, and 489 followers
instead. I don’t know. We get ourselves in a little bit of trouble I think at FiveThirtyEight
at times, because we are fairly combative. For a long time I thought, “Well this is
kind of part of my personality, and the kind of more happy warrior data side is more part
of it too.” They’re actually kind of sides of the same
coin. When you read the New York Times or the Post, not basic factual statements where
they say, “Today, Donald Trump was in Arizona,” but when there’s a piece of analysis that
isn’t necessarily obvious, to say, “Boy, there might be a 40 percent chance that that’s
basically wrong.” That leaves you in a weird place kind of.
But to believe that is, I think, the source of a lot of the healthy skepticism that we
have and also some of our failings sometimes. COWEN: Now let me get to the question that
maybe the crowd most wants to hear. Who will be the next president of the United Arab Emirates?
[laughter] Now this is a trick question because it’s
a hereditary monarchy, but here’s my background question. Intelligence agencies and scholars
did very poorly forecasting the Arab Spring and did very poorly forecasting ISIS.
So you’re put on the case, someone from Washington, McLean, wherever. They call you
in. They say, “What variables should we be looking at to understand the Middle East
that we’re underweighting right now?” I know it’s a tough question, but who will
be the next president of the United Arab Emirates? Will there be a next president? How do you
think about what’s happening there? Black swans or a regularized process?
SILVER: We have all certain compromises. I don’t know that much about international
politics, even enough to have fun sitting like this to speculate all that much. I flew
via Emirates Airlines. It’s the extent of my knowledge about the UAE pretty much. [laughs]
COWEN: Not this election cycle but four more years out, this nation, what’s your best
pick for who will be elected president? SILVER: Who will be president in 2020?
COWEN: Correct. SILVER: I mean the boring pick is probably
Hillary Clinton still. COWEN: Number two, next best pick?
SILVER: I think it’s close between Donald Trump and Marco Rubio. [laughs] Although I
think Trump might be a one-termer. COWEN: If that.
[laughter] COWEN: Who is the most likely next vice president?
SILVER: John Kasich maybe seems tailor-made for the vice presidential role.
COWEN: Even if you think Hillary is more likely to win, he may be the single individual most
likely to be the next vice president. Is that the right way to frame it?
SILVER: I think that might be. Hillary has a very long list to pick from and a lot of
tactical objectives that she would want to fulfill. I think it’s probably a shorter
list for the GOP. COWEN: Can we apply data analysis to figure
out the next Supreme Court pick—again not to know who it will be but to get that 52
percent edge up on what other people are thinking? SILVER: Potentially. There are some fledgling
attempts at Supreme Court analytics although this is also a case where we’re in a sample
size of zero, where you have a nominee who’s very unlikely to be confirmed but there’s
still high political stakes. My uninformed guess would be maybe Srinivasan who was confirmed
97 to nothing. I would tend to think that—my hunch and
this is just a hunch, the theory is that either Obama nominates someone with unimpeachable
credentials and makes Republicans look very unreasonable or he makes a pick that trolls
Republicans and plays to Democratic base. I’m more of a believer in the former as
Obama’s mode of doing things. I think he’d push things in their way and have someone
who might just be at the risk of pissing off the liberal base, but the Republicans have
to look ridiculous opposing, as opposed to the other way around.
COWEN: Now my last question before we get to the crowd. As you said before, we have
a lot of same interests, food, travel, sports, not sure if politics counts as one of mine
but in a broad sense politics. You’ve taken a lot of trips, some for work, some vacation.
If you apply data analysis to those trips, what do you learn about what makes for a good
trip and what can you do or what can we all do to have better trips?
SILVER: I just love travel so much. I had an unintended experiment where I went to Hawaii
for, I guess, two Christmases ago. For some reason I sat on my phone and my phone didn’t
work. We were flying through Portland for some reason or flying New York to Kansas City
to Portland to Honolulu, don’t ask why. But the day I was in Portland, I was panicked.
We drove, we got to this strip mall on the edge of town and they’re like, “You have
to wait in line two hours for us to replace your phone.” So I didn’t have a phone
in Hawaii, and it was the most amazing thing pretty much.
[laughter] COWEN: You’ve repeated that experience each
subsequent trip. SILVER: No.
[laughter] SILVER: But no. I was in Thailand, by contrast,
this Christmas, and I had to build the goddamn primaries election model, and so yeah, your
enjoyment goes down a lot. A little bit of work, working 20 percent of the time, I think
reduces your enjoyment by 70 percent. [laughter]
COWEN: Here’s how we’re going to do questions. We have two mics, one on each side. We’ll
run two queues. I will alternate. These are questions for Nate to speak to. They are not
statements. If you start making a speech, or a statement, I will cut you off, even though
we do not have Kareem Abdul-Jabbar here, this will suffice, so please just ask a question.
It’s fine to introduce yourself, if you wish, and then Nate will respond. I will start
over here. First question, please. AUDIENCE MEMBER: Hi, my name is Caleb. We
talked a little bit earlier about superforecasters, and I was wondering if you’ve ever considered
incorporating the work of superforecasters into FiveThirtyEight?
SILVER: You mean literally the guys who wrote the book, or?
AUDIENCE MEMBER: Or getting a market of superforecasters to help you make your models better.
SILVER: I guess I find crowdsourcing sort of boring. As a journalist, I find it boring.
Even though if you’re in a business setting, that’s exactly what you should do. We actually
are doing that a little bit with the Oscars this year.
We found eight different people who created different models, and we’re seeing how well
they do. Of course six awards is not anywhere near a sufficient sample size to deal with
something like that. I don’t know. I’m very process-driven
as a person. For me, a lot of the joy is in thinking through the process of it, so reading
a book like this is really useful, because I talk about a process I think is pretty great.
But to actually publish the projections, in some sense, is almost beside the point, in
the sense that you’re still probably dealing with sample sizes that are too small to really
tell you all that much. When you start to do that, then I think it
takes folks away from thinking about process and heuristics for forecasting. That’s kind
of an unsatisfying answer, I guess. COWEN: We’ve printed out a lot of Nate’s
columns. Many of them are here, so there’s plenty you can ask about. Next question.
AUDIENCE MEMBER: Hi. I’m Michael Willie, thanks for coming. If it winds up being Clinton
versus Trump, is that the first time where we’ve had two candidates with the highest
unfavorables going against each other? SILVER: Yeah, I would think so. There have
been a few candidates—I think Romney was basically breakeven when he was nominated,
as Obama was, but Clinton is negative 10, and Trump is negative 25, or something. There
probably will be some reversion to the mean in both cases.
Remember, one reason why—and I’d say Trump’s a fairly heavy underdog if he wins the nomination.
But it’s a conditional probability. Conditional on having won the GOP nomination, Trump will
have had to display some staying power, some acumen, because sooner or later he will have
to get beyond 35 percent to win 50 percent or so, and probably will have done something
to improve his image with people who are not in his core constituency.
But yeah, it would be pretty unprecedented, certainly. I wonder if you have to adjust—in
baseball, you have to adjust stats for the era, where if you’re in the home run or
steroid era, then 50 home runs doesn’t matter as much. Maybe now, Obama with a 48 percent
approval rating, is at a 56 park-adjusted approval rating. I’m not sure, maybe.
[laughter] COWEN: Next question.
AUDIENCE MEMBER: Hi. My name’s Tom. Earlier, Tyler brought up the question, “How much
data do people want?” A two-part question. Does the amount of data that people want,
is that influenced by the way data is presented? The second part would be, what advice would
you give as far as presenting data, or visualizing it?
SILVER: Visualizing it might be some of the advice. People seem to learn a lot better
from visualization. One thing I think a lot about as a journalist is preferring simple
models to more complicated models. There are other virtues of simple models. People can
also take it too far. But as a journalist, for example, to have
something I can say this is a benchmark, and I understand what it’s doing, and I can
explain what it’s doing, and I can also understand what the limitations of it might
be, and so I know which direction to lean relative to that baseline. It’s more useful
than a place where you just say, “Well, we fed some data into a random number generator,
or a magic machine, and here’s what it spit out.”
For example, I highly prefer—this is going off on a tangent—but I prefer regression-based
modeling to machine learning, where you can’t really explain anything. To me, the whole
value is in the explanation. But I do think likewise, people, when you explain it and
say, “Hey, we probably have the same interests here in mind,” to say this is actually pretty
simple once you start peeling away the BS. To me, that approach works a lot better in
the long run than the approach of saying, arguments from authority, “Well, this is
rigorous, and empirical, and objective, so therefore believe the numbers.” I think
explaining to people why it’s actually not all that complicated, and why you’re making
very defensible assumptions how that leads you to an answer that might surprise them.
COWEN: Next question. AUDIENCE MEMBER: Frank Manheim, School of
Policy. Could you put some numbers on the criteria that the median voter would use in
the United States to elect major politicians, like president? For example, emotions, personal
acquaintance, rational concepts, information, and so on?
SILVER: The classic political science answer is that people are deeply concerned about
the economy, and that the economy makes up 50 percent or so of what people vote about.
There’s room to dispute that. There’s esoteric critiques that maybe these models
are overfit. But leaving that aside for now. One of the reasons why I was initially skeptical
about Trump is that America has a history of not nominating candidates, and electing
candidates, rightly, who are blatantly unfit for office.
[laughter] COWEN: I have a softball follow-up question
to that. [laughter]
COWEN: We’re in the state of Virginia. To the best of my knowledge, you’re the only
person to have calculated correctly, what is the chance if you are a voter in the state
of Virginia, that your vote will sway a presidential election?
SILVER: It was pretty high. I think it was—oh, as an individual voter?
COWEN: Individual voter. What’s the chance that your vote in the state of Virginia will
matter? If you don’t remember, I do, but it’s from your paper.
SILVER: It’s like 10 percent divided by 4 million or something?
COWEN: It’s 1 out of 10 million, the highest of any state. So if you’re going to vote
anywhere, vote here. [laughter]
COWEN: Next question. AUDIENCE MEMBER: Hi, Tyler O’Neil, a reporter
with PJ Media. My question is, you mentioned how difficult it is, the weakness of empirical
models when predicting presidential elections. Is it possible to look at congressional elections,
House of Representatives races, and draw more information in modeling from those?
SILVER: Yeah, I think we would say that even though it’s less sexy to predict Senate
races or congressional races that having larger quasi-independent samples, that would be the
better test, ultimately. Although even there, the errors—so we saw in the Senate races
last year how the polls were off, on average, by three or four points, which is pretty bad.
It’s happened before. The problem is that all those errors were
in the same direction, so Republicans won when a lot of races around the country that
they were underdogs in—not huge underdogs, except Virginia was almost a major upset.
But yeah, it’s a much purer form of the exercise to do data mining on congressional
elections and weighing polls versus fundamentals and whatever else.
COWEN: Next question. AUDIENCE MEMBER: Hi. My name’s Harold Walbert.
You mentioned some things like limited data, limited observations, nonlinearity, and things
of that nature that make traditional tools like statistics and econometrics difficult.
What are your thoughts on more computationally intensive methods like agent-based modeling
for dealing with these things, like you mentioned herd behavior, that make some of these analyses
more difficult? SILVER: Agent-based modeling is interesting,
and there is some, if you can simulate the underlying mechanisms. This is how weather
forecasting works, by the way, is that weather forecasting is not particularly statistically-driven
as we would think of it. They actually are creating a physical model of the atmosphere
which they are resolving mathematically. If you have reason to know exactly how certain
people would behave and how they behave as a system, then agent-based modeling could
give you insights you couldn’t get from regression analysis.
On the other hand, if you’re somewhat wrong about those assumptions, then the things could
go very haywire in a hurry. When I’m building models myself now, I spend a lot more time
thinking about the edge cases, say, let’s put some really weird inputs in here that
are on the edge of plausible, and see how the model responds to those.
Maybe you have a function that’s approximately linear, but can’t be at the edge case. It
would say that, for example, if you have a model saying that Hillary Clinton will get
106 percent of the vote in Washington, DC, or something, against Trump, I used to think,
“Well, who really cares? She’s going to win DC anyway.”
COWEN: She may. [laughs] SILVER: She may, right? You can vote twice
in some parts of DC. [laughter]
SILVER: But now, that bothers me more, so I’m trying to think more about the correct
functional form of a model that would apply when the going gets weird, because when the
going gets weird is when things are interesting, anyway.
COWEN: We have four minutes left. Next question. AUDIENCE MEMBER: I’m Holly Richard. I’m
an intern at the House of Representatives. Do you believe Facebook and Twitter, where
people create their own newsfeed, has led to possibly confirmation bias, and has led
to people choosing more extreme views of political ideologies such as socialism, nationalism,
Marxism? SILVER: Perhaps. Though I would also say that
the traditional two-dimensional political spectrum is a strange and contrived thing,
too. It’s the result of a very messy process of coalition building between parties.
I mentioned reasons to be pessimistic earlier. A reason to be optimistic as a fan of democracy
is that you are seeing voice given to quirkier ideologies that are no less intellectually
coherent in the kind of Democratic versus Republican axis that we have in the United
States. I believe in the notion of a filter bubble,
where people surround themselves where they’re getting like information and not confronting
themselves with unpleasant facts, necessarily. You saw that a lot during the 2012 election,
where the polling was a lot more straightforward than it is this time around, and people still
were cherry-picking data to tell themselves that Romney might win.
You saw Democrats doing the reverse, by the way, in the 2014 midterms, more or less. But
yeah, as someone who’s a critic of media, I think the way people consume media is important,
and has probably fairly large effects on our politics.
COWEN: Last question. AUDIENCE MEMBER: Hi, Mike Bliley. I’m a
law student here, and I get my coverage of the election exclusively from FiveThirtyEight.
[laughter] AUDIENCE MEMBER: I do that largely because
of the unbiased nature, except for Harry’s unabashed love for Chris Christie.
[laughter] AUDIENCE MEMBER: I noticed that specifically
in your debate coverage, one of the things that you all always mention is that the mainstream
media’s portrayal of the debate matters more than anything else. When they say that
someone wins, that coverage carries, and then at the end of those pieces, you and your staff
put together grades for how the candidates did.
You may see where I’m going with this. You strike me as someone who would rather predict
rather than influence, but do you see yourself playing into this zeitgeist where you could
carry some weight in this election? SILVER: That’s why the primaries, although
they’re fun, are a little tricky. The general election people are fairly sensible and retreat
to their corners, but the primaries are so momentum-driven, that it’s a little bit
weird. I’m sure people do read what we say and so forth. It’s not the type of influence
that I want. At the same time, the fact is that all news
coverage is influential, and I would say at the very least, we promise some self-awareness,
that we’re aware that the way the events are covered by the press can affect voters’
views. Sometimes the press can be surprised that it doesn’t go the way they expect,
but you can have these big feedback loops. I’m surprised how difficult it is. I think
one big edge we have—I’m glad that you read us—but I think one big edge we have
over, say, the New York Times or something is that we can talk about the media as a political
actor. We are the media, too, and so I’m aware
of the circularity of that. Frankly, I think one reason why during the primaries sometimes
the conservative sites are more interesting to me than liberal sites, is that they also
start off being more suspicious of the media, sometimes in ways that I think are wrong,
like about the polls in 2012. But I think having that skepticism and seeing
the media as a political actor instead of a benevolent umpire is to a first approximation
the right way to do things, and that’s reflected in our coverage, I guess sometimes at the
risk of being a little bit hypocritical, potentially. We do try and be very transparent about what
we think is a fact, what’s an opinion, what’s an analysis, what is a provocation. One reason
why I like your blog is that you have a lot of provocations. Sometimes you put the Tyrone
label on it, but it’s clear what they are. It’s clear that they’re provocations meant
to incite discussion and debate, and so we’ll have a few of those, too, at times.
Speaking in the first person, I think, is important, and breaking from the voice of
God where “A storm cloud gathered on New Hampshire today, and the voters decided that—.”
Speaking as a subjective individual trying to understand what the objective world is
like is a lot of what we’re all about. It’s not for everyone, but I think that
should be reflected at least in the tone and approach of our coverage, even where we wind
up getting things wrong in the end. COWEN: Here’s Nate’s book. Read Nate’s
site. Nate, thank you for a great chat. SILVER: Thank you.

7 Comments

  1. In response to Tyler's comments regarding Buffett and Lynch's investment out-performance:
    See 2:30–2:40 in vid below:
    https://www.youtube.com/watch?v=eE-r0eOyvEA

  2. Silver deserves props for navigating Cowen's terrible, clever little questions and still managing to provide interesting and insightful answers.

  3. The best line of the conversation:

    1:10:25 "One of the reasons why I was initially skeptical about Trump is that America has a history of not nominating candidates and electing candidates who are blatantly unfit for office."

Leave a Reply

Your email address will not be published.


*