How Is Econometrics Changing? (Josh Angrist, Guido Imbens, Isaiah Andrews)
-
0:00 - 0:03♪ [music] ♪
-
0:04 - 0:06- [Narrator] Welcome to
Nobel Conversations. -
0:07 - 0:08In this episode,
-
0:08 - 0:12Josh Angrist and Guido Imbens
sit down with Isaiah Andrews -
0:12 - 0:15to discuss how the field
of econometrics is evolving. -
0:16 - 0:19- [Isaiah] So Guido and Josh,
you're both pioneers -
0:19 - 0:22in developing tools for
empirical research in economics. -
0:22 - 0:23And so I'd like to explore
-
0:23 - 0:25sort of where you feel like
the field is heading, -
0:25 - 0:28sort of economics, econometrics,
the whole thing. -
0:29 - 0:31To start, I'd be interested to hear
-
0:32 - 0:35about whether you feel like
sort of the way in which -
0:35 - 0:39the local average treatment
effects framework sort of took hold -
0:39 - 0:42has any lessons for how
new empirical methods in economics -
0:42 - 0:44develop and spread
or how they should. -
0:45 - 0:46- [Josh] That's a good question.
-
0:47 - 0:48You go first.
-
0:48 - 0:49(laughter)
-
0:50 - 0:53Yeah, so I think
the important thing -
0:53 - 0:59is to come up
with good convincing cases -
0:59 - 1:02where the questions are clear
-
1:02 - 1:06and where kind of the methods
apply in general. -
1:06 - 1:08So one thing I--
-
1:08 - 1:12Kind of looking back
at the subsequent literature, -
1:12 - 1:17so I really like the regression
discontinuity literature -
1:17 - 1:20[where there were] clearly a bunch
of really convincing examples -
1:20 - 1:21and that allowed people to kind of
-
1:22 - 1:27think more clearly, look harder at
the methodological questions. -
1:27 - 1:29Kind of do clear applications
-
1:29 - 1:31that then allow you
to kind of think about, -
1:31 - 1:34"Wow, do this type of assumption
seem reasonable here? -
1:34 - 1:38What kind of things do we not like
in the early papers? -
1:38 - 1:40How can we improve things?"
-
1:40 - 1:42So having clear applications
motivating, -
1:43 - 1:46these literatures,
I think it's very helpful. -
1:47 - 1:49I'm glad you mentioned
the regression discontinuity, Guido. -
1:49 - 1:53I think there's a lot of
complementarity between IV and RD, -
1:55 - 1:57Instrumental Variables and
Regression Discontinuity. -
2:00 - 2:03And a lot of
the econometric applications -
2:03 - 2:05of regression discontinuity
-
2:05 - 2:07are what used to be called
"fuzzy" RD, -
2:07 - 2:12where, you know, it's not discrete
or deterministic at the cutoff, -
2:12 - 2:15but just the change
in rates or intensity. -
2:15 - 2:19And and the late framework helps us
understand those applications -
2:19 - 2:21and gives us a clear interpretation
-
2:21 - 2:25for say, something like,
in my paper with Victor Lavy, -
2:25 - 2:28where we use Maimonides'
rule, the class size cutoffs. -
2:29 - 2:30What are you getting there?
-
2:30 - 2:32Of course, you can
answer that question -
2:32 - 2:34with a linear
constant effects model, -
2:34 - 2:36but it turns out
we're not limited to that, -
2:36 - 2:40and an RD is still very powerful
and illuminating, -
2:41 - 2:42even when, you know,
-
2:42 - 2:46the correlation between the cutoff
and the variable of interest, -
2:46 - 2:49in this case class size,
is partial, -
2:49 - 2:51maybe even not that strong.
-
2:52 - 2:55So there was definitely a kind of,
a parallel development. -
2:55 - 2:56It's also interesting,
-
2:57 - 3:00you know, nobody talked about
regression discontinuity designs -
3:00 - 3:01when we were in graduate school,
-
3:01 - 3:05it was something that other social
scientists were interested in, -
3:06 - 3:10and that kind of grew up
alongside the late framework -
3:10 - 3:15and we've both done work on
both applications and methods there -
3:15 - 3:18and it's been very exciting
to see that kind of develop -
3:18 - 3:20and become so important.
-
3:20 - 3:24It's part of a general evolution,
I think, towards, you know, -
3:24 - 3:28credible identification strategies
causal effects... -
3:29 - 3:30less, you know, making econometrics
-
3:30 - 3:33more about causal questions
than about models. -
3:34 - 3:39In terms of the future, I think one thing
that LATE has helped facilitate -
3:39 - 3:43is a move towards more creative
randomized, trials, where, -
3:43 - 3:44you know, there's something of interest,
-
3:46 - 3:51it's not possible or straightforward
to Simply turn it off or on -
3:51 - 3:53but you can encourage it
-
3:53 - 3:58or discourage it. So, you subsidize
schooling with financial aid, for example, -
3:59 - 4:00so now we have a whole
-
4:00 - 4:02Framework for interpreting that.
-
4:03 - 4:04And,
-
4:04 - 4:07and it kind of opens
the doors to randomized, -
4:07 - 4:09Trials of things that that maybe would,
-
4:10 - 4:10you know,
-
4:10 - 4:14not have seen possible before we've,
-
4:14 - 4:20we've used that a lot in the work. We do
on schools in our in the blueprint Lab -
4:20 - 4:27at MIT were exploiting random assignment
and in very creative ways, I think. -
4:28 - 4:32Related to that. Do you see sort
of particular factors that make for -
4:32 - 4:34useful research and econometrics.
-
4:34 - 4:37You've alluded to it?
-
4:37 - 4:40Having a clear connection to
problems that are actually coming up. -
4:40 - 4:45And empirical practice is often a good
idea. I'll send it. Always a good idea. -
4:46 - 4:50I often find myself sitting in an
economy metrics Theory, seminar. -
4:51 - 4:52Say the Harvard MIT seminar
-
4:53 - 4:57and I'm thinking what problem is
this guy solving who has this? -
4:57 - 5:00This problem and you know,
-
5:02 - 5:05sometimes there's an
embarrassing silence if I ask -
5:05 - 5:08or there might be a
fairly contrived scenario. -
5:09 - 5:12I want to see where the tool is useful.
-
5:12 - 5:15There are some purely foundational tools.
-
5:15 - 5:18I do take the point, you
know, there are people who are -
5:18 - 5:22working on conceptual
foundations of you know, -
5:23 - 5:25it's more becomes more like
mathematical statistics. -
5:26 - 5:27I mean, I remember an early example
-
5:27 - 5:28I believe that that I,
-
5:28 - 5:32you know, I struggled to understand was
the idea of stochastic Equity continuity, -
5:32 - 5:37which my one of my thesis advisors Whitney
knew he was using to great effect and -
5:38 - 5:40I was trying to understand
that and there isn't really. -
5:41 - 5:45It's really foundational. It's not
but an application that's driving that -
5:46 - 5:47at least not immediately
-
5:49 - 5:53but but most things are not like that
and so there should be a problem. -
5:54 - 5:57And the I think it's on the it's on
-
5:57 - 6:00On the, the seller of that sort of thing,
-
6:00 - 6:04you know, because there's opportunity
cost the time and attention and effort -
6:04 - 6:07to understand things to, you
know, it's on the seller to say. -
6:07 - 6:09Hey, I'm solving this problem
-
6:09 - 6:13and and here's a set of results
that show that it's useful. -
6:13 - 6:15And here's some insight that I get.
-
6:16 - 6:17As you said, Josh, great,
-
6:17 - 6:21sort of there's been a move in the
direction of thinking more about causality -
6:21 - 6:23in economics and empirical
work in economics, -
6:23 - 6:25any consequences of sort of the Wilds,
-
6:25 - 6:27the spread of that view. That
surprised you or anything. -
6:27 - 6:31UB was downsides of sort of the way that
he could empirical, economics has gone -
6:32 - 6:32sometimes.
-
6:32 - 6:38I see somebody does Ivy and they get a
result which seems implausibly large. -
6:39 - 6:40That's the usual case.
-
6:42 - 6:44So it might be, you know,
-
6:44 - 6:49an extraordinarily large causal effect
of some relatively minor Intervention, -
6:49 - 6:53which was randomized or for
which you could make a case that -
6:53 - 6:57that there's a good design.
And then when I see that, -
6:57 - 6:58That and,
-
6:58 - 6:58you know, I think,
-
6:58 - 6:59you know,
-
6:59 - 7:03it's very hard for me to believe that this
relatively minor intervention has such -
7:03 - 7:04a large defect,
-
7:04 - 7:07the author. Well, sometimes
resort to the local average, -
7:07 - 7:09treatment effects theorem and say,
-
7:09 - 7:13wow, these compliers, you know,
they're special in some way. -
7:13 - 7:18And, you know, they just benefit
extraordinarily from this intervention -
7:18 - 7:22and I'm reluctant to take that
at face value. I think, you know, -
7:22 - 7:24often when effects are too big,
-
7:24 - 7:27it's because the exclusion
restriction is failing. So -
7:27 - 7:32Don't really have the right endogenous
variable to scale that result. -
7:32 - 7:36And so I'm not too happy to see
-
7:36 - 7:39you know, just sort of
a generic heterogeneity -
7:39 - 7:44argument being used to excuse something
that I think might be a deeper problem. -
7:45 - 7:47I think it played somewhat
of an unfortunate roll pin. -
7:47 - 7:52The discussions kind of between reduced
form and structural approaches where -
7:53 - 7:54I feel that wasn't quite
-
7:55 - 7:59right. The instrumental
variables assumptions are -
8:00 - 8:05at the core structural assumptions about
Behavior. They were coming from economic -
8:07 - 8:10thinking about the economic
behavior of agents, -
8:10 - 8:15and it's somehow it got
pushed in a Direction. -
8:15 - 8:18That I think wasn't
really very helpful. If -
8:19 - 8:22the way I think, initially the
-
8:23 - 8:27we wrote things up. It was it was describing
what was happening, there was set of -
8:28 - 8:32methods. People were using be
clarified what those methods were doing -
8:33 - 8:38and in a way that I think
contain a fair amount of insight, -
8:39 - 8:45but it somehow it got pushed into a corner
that I think was not necessarily very -
8:45 - 8:49or even just the language of
reduced form versus structural. -
8:49 - 8:51I find kind of funny in
the sense that the right -
8:51 - 8:53the local average treatment
effect model, right? -
8:53 - 8:55The potential outcomes
model is a nonparametric. -
8:55 - 8:56Structural model,
-
8:56 - 8:59if you want to think about it, as
you sort of suggested, he does. -
8:59 - 8:59So, there's something,
-
9:00 - 9:04there's something a little funny about
putting these two things in a position when -
9:04 - 9:07yes, well, that language, of
course, comes from the area, the -
9:07 - 9:1070s equations framework that we inherited.
-
9:10 - 9:12It has the advantage that people seem
-
9:12 - 9:15to know what you mean
when you use it, but might -
9:15 - 9:18That people are hearing different. Different
people are hearing different things. -
9:18 - 9:21Yeah. I think I think veggies
Farmers had become use -
9:21 - 9:23a little bit of the
pejoratives. Okay? Yeah. -
9:23 - 9:28The word, which is not really quite
what it was originally intended for. -
9:30 - 9:34I guess something else that strikes
me in thinking about the effects of -
9:34 - 9:38the local average treatment effect
framework is that often folks will appeal to -
9:38 - 9:42a local average, treatment effects
intuition for settings. Well, beyond -
9:42 - 9:45ones, where any sort of formal
results has actually been -
9:45 - 9:50Shhhhht. And I'm curious given
all the work that you guys did to, -
9:50 - 9:53you know, establish late results in
different in different settings. I'm curious -
9:53 - 9:58any thoughts on that. I think there's
going to be a lot of cases where -
9:58 - 10:02the intuition does get.
You get you some distance, -
10:03 - 10:08but it's going to be somewhat limited
and establishing formal results. There -
10:08 - 10:13may be a little tricky and there may
be only work in special circumstances, -
10:13 - 10:14you need.
-
10:15 - 10:20And you end up with a lot of formality
that may not quite capture the intuition -
10:20 - 10:23sometimes I'm somewhat uneasy with them
and they are not necessarily the papers. -
10:23 - 10:25I would want to ride that the
-
10:25 - 10:30but I do think something do intuition
orphaned US capture part of the -
10:30 - 10:31of the problem.
-
10:33 - 10:36I think, in some sense we were
kind of very fortunate there -
10:37 - 10:40in the way. The late paper go handle. It.
Don't know if that, actually the editor, -
10:41 - 10:42made it much shorter
-
10:42 - 10:46and that then allowed us to kind of
focus on very clear, crisp results -
10:47 - 10:50where if, you know, this,
-
10:50 - 10:54this is somewhat unfortunate tendency in
the commercialization of having the papers. -
10:55 - 10:59Well, you should be able to fix that, man.
I'm trying to take some time to fix that. -
10:59 - 11:03I think this is an example where it's sort
of very clear that having it. Be sure. -
11:03 - 11:08It's actually impose that no paper can
be longer than the late paper that wow. -
11:09 - 11:14Great. At least no Theory. No Theory Pig.
Yeah, and I think, I think they're well, -
11:14 - 11:17I'm trying very hard to get
the papers to be shorter. -
11:17 - 11:19And I think there's a lot of value
-
11:19 - 11:23today because it's often the second
part of the paper that doesn't actually -
11:24 - 11:26Get you much further
and understanding things -
11:27 - 11:32but and it does make things much
harder to read and, you know, -
11:32 - 11:34it sort of goes back to
-
11:34 - 11:38how I think he kind of a trick should
be done to you should focus on the see. -
11:39 - 11:41It should be reasonably
close to empirical problems. -
11:42 - 11:44They should be very clear problems.
-
11:45 - 11:49But then often the the theory
doesn't need to be quite so long. -
11:49 - 11:49Yeah,
-
11:51 - 11:53I think they had things have
-
11:54 - 11:55On a little off track.
-
11:56 - 11:58The relatively recent change has been a
-
11:58 - 12:02seeming big increase in demand for
people with sort of econometrics. -
12:02 - 12:05Causal effect, estimation
skills in the tech sector. -
12:05 - 12:09I'm interested either of you have
thoughts on sort of how that's gonna -
12:09 - 12:12how that's going to interact with
the development of empirical methods, -
12:12 - 12:14or Empirical research, and
economics. Going forward, sort of -
12:15 - 12:21whether sort of a meta point, which
is there's this new kind of employer -
12:22 - 12:26the Amazons and the Uber and, you know,
-
12:26 - 12:28Riser world
-
12:28 - 12:29and I think that's great.
-
12:29 - 12:33And I'd like to tell my students about
that, you know, especially at MIT. -
12:33 - 12:37We have a lot of computer science
Majors. That's our biggest major -
12:37 - 12:43and I try to seduce some of those folks
into economics by saying, you know, -
12:43 - 12:46you can go work for these,
-
12:46 - 12:48you know companies that
people are very keen to -
12:49 - 12:51work for because the work seems exciting,
-
12:52 - 12:56you know that the skills that you get in
econometrics are are as good or better. -
12:56 - 13:01Better than than any competing discipline
has to offer. So you should at least -
13:01 - 13:04take some econ, take some
econometrics. And some econ. -
13:05 - 13:07I did a fun project with a uber
-
13:08 - 13:13on labor supply of Uber drivers and was
very, very exciting to be part of that. -
13:13 - 13:15Plus. I got to drive for Uber for a while
-
13:16 - 13:21and I thought that was fun tonight. I did
not make enough that I was attempted to -
13:21 - 13:25give up by a mighty job, but
I enjoyed the experience. -
13:25 - 13:26I see a
-
13:26 - 13:31Cho challenge to our model
of graduate education here, -
13:32 - 13:37which is if we're trading people
to go work at Amazon, you know, -
13:38 - 13:43it's not clear. Why? You know, we should
be paying graduate stipends for that. -
13:43 - 13:45Why should the taxpayer effectively
-
13:46 - 13:51be subsidizing? That our graduate education
in the u.s. Is generously subsidized? -
13:51 - 13:56Even in private universities. It's
ultimately there's a lot of public money. -
13:56 - 13:59Me there. And I think the
traditional rationale for that is, -
14:00 - 14:04you know, we were training, Educators and
Scholars, and there's a great externality -
14:04 - 14:06from the work that we do.
-
14:06 - 14:10It's either the research externality,
or a teaching externality. -
14:10 - 14:15But, you know, if many of our students
are going to work in the private sector, -
14:16 - 14:22that's fine, but that maybe their
employers should pay for that. -
14:22 - 14:25He says, so different from
people working for a Consulting. -
14:26 - 14:27Trust me.
-
14:27 - 14:33It's not clear to me that the number
of jobs in academics has changed. -
14:33 - 14:38It's just, I feel like this is a
growing sector whereas Consulting, -
14:38 - 14:42your right to raise that, it might
be the same for for Consulting. -
14:43 - 14:44But this,
-
14:44 - 14:48you know, I'm placing more and
more students in these businesses. -
14:48 - 14:50So, it's on my mind in
a way that I've sort of, -
14:51 - 14:56you know, not been attentive to consulting
jobs, you know, Consulting was always, -
14:56 - 15:00It's important and I think they'll so
there's some movement from Consulting back -
15:00 - 15:03into research. It's a little more fluid.
-
15:03 - 15:04The,
-
15:04 - 15:05a lot of the work in the
-
15:06 - 15:10in both domains. I have to say,
it's not really different but -
15:10 - 15:14you know, people who are working
in the tech sector are doing things -
15:14 - 15:17that are potentially of scientific
interest, but mostly it's hidden. -
15:17 - 15:21Then you really I have to say, you know,
why, why is the government paying for this? -
15:22 - 15:26Yeah, although yeah, I mean taquitos point,
I guess it. There's a, there's a data. -
15:26 - 15:30Question here of it has the sort
of total nanak. It sort of say -
15:30 - 15:31private
-
15:31 - 15:34for-profit sector employment of econ Ph.D.
-
15:34 - 15:38Program graduates increased or has
it just been a substitution from -
15:38 - 15:40finance and Consulting towards tack.
-
15:40 - 15:44I may be a reaction to something
that's not really happening -
15:44 - 15:48so bad. I've actually done some work
with some of these tech companies. -
15:49 - 15:52So I don't disagree with Justice
point that we need to think -
15:52 - 15:55a little bit about the funding model
whose it was in the end paying for the -
15:56 - 15:59It education. But from a
scientific perspective. -
16:00 - 16:04The only do these places have
have great data and nowadays. -
16:04 - 16:07They tend to be very careful
with that for privacy reasons, -
16:08 - 16:09but also have great questions.
-
16:10 - 16:11I find it very
-
16:12 - 16:13inspiring kind of to listen to
-
16:13 - 16:16the people there and kind of see
what kind of questions they have -
16:16 - 16:17and often their questions.
-
16:18 - 16:21That also come up outside of these.
-
16:21 - 16:25These companies have a couple of
papers with the rights in the chat. -
16:26 - 16:26And then
-
16:26 - 16:32as soon as an atheist kind of where we
look at ways of combining experimental data -
16:32 - 16:34and observational data, and can it there.
-
16:36 - 16:39Rights Chetty was interested
in what is the effect -
16:39 - 16:45of Early Childhood programs on outcomes
later in life? Not just kind of test scores, -
16:45 - 16:48but on earnings and stuff, and
we cannot be developed methods -
16:49 - 16:52that would help you shed
light on that, on the some, -
16:53 - 16:55in some settings and the same problems.
-
16:56 - 17:00Came up kind of in this
tech company settings. -
17:01 - 17:04And so for my perspective, it's
-
17:04 - 17:08the same kind of a stocking two
people doing a protocol work. -
17:08 - 17:12I tried to kind of look at these
specific problems and then try to come up -
17:12 - 17:18with more General problems that we
formulating the problems at a higher level. -
17:18 - 17:23So that I can think about solutions
that work in a range of settings. -
17:23 - 17:26And so from that perspective, the
-
17:26 - 17:30His with the the tech companies I
just very valuable and very useful. -
17:31 - 17:31It's know.
-
17:32 - 17:34We do have students. Now spent
-
17:34 - 17:37doing internships there and
then coming back and writing -
17:37 - 17:43more interesting thesis, as a
result of their experiences there. -
17:45 - 17:48If you'd like to watch more
Nobel conversations, click here, -
17:48 - 17:50or if you'd like to learn
more about econometrics, -
17:51 - 17:53check out Josh's mastering
econometrics series. -
17:54 - 17:56If you'd like to learn more about he do.
-
17:56 - 17:58Josh and Isaiah check out
the links in the description.
- Title:
- How Is Econometrics Changing? (Josh Angrist, Guido Imbens, Isaiah Andrews)
- ASR Confidence:
- 0.80
- Description:
-
- Video Language:
- English
- Team:
Marginal Revolution University
- Duration:
- 18:03
Show all