Popular in Social Sciences

Books

City Time: On Being Sentenced to Rikers Island

Independent Scholar David A. Campbell

Culture and Society

The City as a Growth Machine: Toward a Political Economy of Place

New York University Harvey Moltoch

Economics

Gender and Economics

University of California Santa Barbara Shelly Lundberg

American Politics

Keeping the Republic

Boston College Marc Landy

Banks and Policy Organisations

Greening the financial sector

European Investment Bank Matteo Ferrazzi

British Politics

Entrepreneurial thinking: the politics and practice of policy impact

University of Sheffield Matthew Flinders

Culture and Society

Carceral Entrapments: Views from the Prison/Street Interface in India

Indian Institute of Technology Delhi Mahuya Bandyopadhya

Does the Cream Always Rise to the Top? The Misallocation of Talent in Innovation

View more insights in Social Sciences - Economics - Macroeconomics

View more figures from University of Toronto

View the page of Murat Celik

Read the Study

The misallocation of talent in innovation – “missing Einsteins” – has a first-order impact on growth and welfare. Murat Celik combines surname-level empirical analysis and inventor and census micro-data to suggest people from richer backgrounds are more likely to become inventors, but those from high-education backgrounds become more prolific inventors. Therefore rich can become inventors even if mediocre through excessive credentialing spending. Celik discusses the implications of these findings.

Image courtesy of interviewee. April 17, 2024

Generating Workspace

Export as PDF

This video video is a transcript of a talk discussing the question of whether the best and brightest minds in society, particularly in relation to innovation and invention, are being effectively identified and allowed to flourish. The speaker highlights the significance of this question across various professions, focusing on the role of inventors.

Using Albert Einstein as a case study, the speaker illustrates the potential loss to society if talented individuals do not take on roles in innovation due to various barriers. The speaker argues that socioeconomic factors significantly influence who becomes an inventor, emphasizing that individuals from wealthier backgrounds are more likely to enter the field. This video video trend raises concerns about a misallocation of talent, where potentially brilliant inventors from less affluent backgrounds may be overlooked.

The speaker references past economic literature on innovation and growth, noting that while much attention has been devoted to firm-level motivators for research and development (R&D), there has been less focus on the inventors themselves. The study aims to understand factors determining who becomes an inventor and how productive they are once they do.

To explore these issues, the speaker employs a unique methodology that uses surname-based data to infer the socioeconomic backgrounds of inventors and analyze their success rates. The findings suggest that income associated with an inventor's surname significantly predicts their likelihood of becoming an inventor, while educational attainment is a better predictor of productivity among those who already are inventors.

The speaker concludes with suggestions for policy interventions aimed at addressing the misallocation of talent in innovation, such as implementing progressive estate taxation to reduce the advantages of wealth in the invention process. The overall aim is to promote a more equitable system that enables all potential inventors, regardless of their socioeconomic status, to contribute to innovation and economic growth.

Language:

Export as PDF

X-TIMESTAMP-MAP:LOCAL=00:00:00.000,MPEGTS=900000

00:00.215 --> 00:02.645
Thank you very much for the question and inviting me.

00:03.105 --> 00:06.805
So, this question, does the cream rise to the top,

00:07.405 --> 00:12.285
I think is very important for a great number of professions,

00:12.425 --> 00:13.765
not just inventors,

00:14.225 --> 00:18.805
but I focused on this particular question of whether

00:19.585 --> 00:22.525
we always are able to get the best

00:22.545 --> 00:26.485
and brightest minds in a society to,

00:26.945 --> 00:27.965
become inventors

00:28.025 --> 00:32.765
and, benefit from their ideas, their ability to come up

00:32.765 --> 00:34.605
with new technologies, new processes.

00:34.995 --> 00:37.125
That was the question I was interested in.

00:37.585 --> 00:40.925
And, just to give an example, just to give an example,

00:41.625 --> 00:43.725
think of Albert Einstein.

00:43.905 --> 00:45.965
So we all know about him.

00:46.145 --> 00:48.405
We all know he's a prolific physicist.

00:48.545 --> 00:50.645
He has the paradigm shift.

00:50.865 --> 00:55.045
We of a lot of our, technologies,

00:55.825 --> 00:57.445
our ways of thinking to him.

00:57.985 --> 01:01.445
And, what's not as many people know is that

01:01.445 --> 01:03.125
beyond being a prolific scientist,

01:03.505 --> 01:07.285
beyond being a prolific physicist, he has also come up with,

01:07.305 --> 01:09.885
more than 50 patented inventions.

01:10.345 --> 01:13.205
So, he was not just a scientist

01:13.205 --> 01:14.925
that thought about these big ideas,

01:15.145 --> 01:17.885
but he also came up with these applicable ideas

01:17.995 --> 01:20.725
that could kind of, we could see in products

01:21.035 --> 01:23.165
that we can see, in our daily lives.

01:23.905 --> 01:27.805
And, if you kind of look at, Einstein's,

01:27.825 --> 01:31.565
family background, you see that he was no ordinary person.

01:31.865 --> 01:35.885
So actually, he has received a great education in a number

01:35.905 --> 01:38.165
of countries in Germany, Switzerland, Italy,

01:38.545 --> 01:41.245
and his, alma mater was, ETH Zuri.

01:41.705 --> 01:45.485
So he definitely, this received this great education.

01:45.865 --> 01:48.125
And after receiving this great education, he was able

01:48.125 --> 01:52.325
to contribute upon, the body of knowledge that we have.

01:52.865 --> 01:56.935
And the question one, has in mind when,

01:56.935 --> 02:01.545
they see this example is that one, how would the work,

02:01.695 --> 02:05.385
look like if Albert Einstein could not become,

02:05.445 --> 02:08.665
an inventor, but perhaps he became an accountant

02:08.685 --> 02:11.505
or a factory worker, what would we have lost?

02:11.885 --> 02:13.345
And in the case of Albert Einstein,

02:13.375 --> 02:16.105
perhaps we can say we would lose, a lot.

02:16.765 --> 02:18.705
But, when we think further,

02:19.645 --> 02:21.705
how do we know we are not missing out on

02:21.895 --> 02:23.345
potential Einsteins?

02:23.685 --> 02:24.825
So that's the bigger question.

02:24.845 --> 02:28.145
That's what, has motivated me to conduct this study.

02:28.885 --> 02:33.675
And, to give some, brief background on, kind of

02:33.675 --> 02:34.915
what has been done,

02:35.335 --> 02:37.195
in the economics literature on this topic.

02:37.775 --> 02:39.475
So, in the nineties there was,

02:39.475 --> 02:42.875
there were these seminal papers by Philippa Guion,

02:42.885 --> 02:46.515
Peter Hoit and, Paul Romer on endogenous growth.

02:46.895 --> 02:48.635
So we have concluded that

02:49.355 --> 02:51.595
although there can be many sources of economic growth

02:52.735 --> 02:55.475
in the long run, in the very long run, much

02:55.495 --> 02:58.915
of our prosperity also to these new technologies, new ideas,

02:59.185 --> 03:00.185
New products.

03:00.425 --> 03:04.765
And, there has been a surge in interest in,

03:05.285 --> 03:08.645
identifying the bottlenecks in the process

03:08.745 --> 03:10.725
of creating new ideas, new technologies.

03:11.385 --> 03:12.765
Uh, in the past and

03:12.765 --> 03:15.325
before I conducted my study,

03:15.385 --> 03:17.605
the focus was mostly on the firm side.

03:17.905 --> 03:20.325
So what motivates firms to do r and d?

03:20.865 --> 03:25.595
How can we, improve the setting such

03:25.595 --> 03:28.115
that there's more competition, there's more innovation.

03:28.515 --> 03:31.115
A lot of the work was on, firmly on the firm side,

03:31.995 --> 03:36.695
but, all this work was ignoring the other side of the,

03:36.755 --> 03:38.855
equation, which is inventors.

03:39.315 --> 03:43.135
So who are the people who come up with, these new ideas

03:43.355 --> 03:47.655
who work for these firms, who generate these patents,

03:47.685 --> 03:48.695
that the firms use?

03:49.235 --> 03:53.935
And, I decided to look at what determines,

03:54.235 --> 03:55.655
who becomes an inventor

03:56.735 --> 04:00.185
and who becomes a prolific inventor,

04:00.365 --> 04:01.705
not just becoming an inventor,

04:01.705 --> 04:03.305
but a very productive inventor.

04:04.005 --> 04:07.385
And, to, look at this question, you kind

04:07.385 --> 04:09.065
of need two types of data.

04:10.165 --> 04:14.225
Uh, one is, data on who becomes inventors,

04:14.285 --> 04:16.865
and there's, record of innovations,

04:17.075 --> 04:18.345
which we have thankfully.

04:18.565 --> 04:22.385
So in the United States, we have this, micro data

04:22.745 --> 04:24.625
provided by, United States Patent

04:24.685 --> 04:26.585
and Trademark Office,

04:26.585 --> 04:30.185
which contains all information on all the patents since

04:30.665 --> 04:34.945
1976, and all of their citations, all of their inventors,

04:35.205 --> 04:37.145
we have everything we can link them to firms.

04:37.615 --> 04:41.985
It's perfect. What was the missing link, in this is

04:41.985 --> 04:46.905
that we really do not know directly what the backgrounds,

04:47.005 --> 04:48.945
the family backgrounds of these inventors are.

04:49.205 --> 04:52.265
We just see their names, we know their inventions,

04:52.565 --> 04:55.545
we know nothing about their family backgrounds, which

04:56.075 --> 04:59.105
might have some influence on whether they were able

04:59.105 --> 05:02.265
to make it to the inventor pool in the first place.

05:03.085 --> 05:05.745
So the, kind of the novelty I came up

05:05.745 --> 05:07.665
with in the study was to do the following.

05:07.665 --> 05:09.825
So we have no direct information,

05:09.925 --> 05:12.545
but could we somehow find some proxies

05:13.335 --> 05:15.595
to get at their family backgrounds?

05:16.495 --> 05:19.315
And, interestingly, we can,

05:19.535 --> 05:23.115
and, this links to another literature, essentially.

05:23.495 --> 05:27.315
So, in, back in the,

05:28.035 --> 05:31.995
past research has established that your names and surnames,

05:32.375 --> 05:33.595
and your first name

05:33.595 --> 05:36.515
and last name contains a wealth of information

05:37.185 --> 05:40.475
that can tell us about, where you came from.

05:41.255 --> 05:43.835
Uh, and, there's this,

05:44.105 --> 05:45.795
seminar work by Gregory Clark.

05:45.815 --> 05:49.595
He, he, he looks at these rare surnames people

05:49.595 --> 05:50.915
with these rare surnames.

05:51.055 --> 05:54.195
Uh, do they have a higher socioeconomic status today?

05:54.415 --> 05:55.755
And he says, the answer is yes.

05:56.095 --> 05:58.005
So I kind of use the same methodology.

05:58.555 --> 05:59.885
What I did was the following.

05:59.885 --> 06:04.535
So I took all of the inventors, from,

06:04.975 --> 06:06.135
1976 onwards.

06:06.735 --> 06:08.095
I looked at their surnames

06:08.795 --> 06:12.495
and I matched them to the income levels

06:12.675 --> 06:15.775
and education levels associated with their surnames

06:16.845 --> 06:19.335
back in 1930, which is the last year,

06:19.335 --> 06:20.975
we have public data available for,

06:21.355 --> 06:23.455
so this is basically saying the following,

06:24.345 --> 06:27.725
you're an inventor, you have a particular surname.

06:28.205 --> 06:33.085
I go back to 1930, and then I look at what was the income

06:33.745 --> 06:36.085
and education associated with your surname?

06:36.785 --> 06:41.085
And based on that, try to predict, one, your chances

06:41.085 --> 06:44.845
to be an inventor, and two, how productive you are

06:44.865 --> 06:46.885
as an inventor condition becoming one.

06:47.705 --> 06:51.805
So, and you might think of many things, you might say,

06:51.805 --> 06:54.045
this is a very noisy measure, right?

06:54.105 --> 06:57.005
So this is across two to three generations,

06:57.545 --> 06:59.965
and we are collapsing everyone by their surnames.

07:00.385 --> 07:03.485
So you'd expect this to have zero explanatory power.

07:04.975 --> 07:07.985
Very interestingly, it has such a large,

07:08.355 --> 07:11.705
predictive power that I, I was amazed when I first saw this.

07:12.325 --> 07:16.755
So, actually, people with a,

07:16.775 --> 07:21.625
higher income certainly are 24% more

07:21.625 --> 07:22.905
likely to be an inventor,

07:23.935 --> 07:28.155
and 18% more likely to be a very productive inventor.

07:28.535 --> 07:31.035
So even across these two to three regenerations,

07:31.425 --> 07:32.715
despite all the time lag,

07:32.785 --> 07:36.315
despite every po potential mechanism you can think of,

07:36.945 --> 07:39.755
just knowing your surname can tell us a lot about whether

07:39.755 --> 07:40.795
you'll be an inventor

07:41.295 --> 07:43.035
or you'll be a very productive, I invent,

07:43.895 --> 07:46.995
and this kind of tells you the following,

07:46.995 --> 07:50.835
before we dig deeper into it, it says that persistence

07:50.835 --> 07:55.545
of socioeconomic status as captured by becoming an inventor

07:56.405 --> 07:57.705
is immensely high.

07:58.615 --> 08:01.395
So when I go to the full quantitative model

08:01.415 --> 08:05.195
to investigate this, this corresponds to roughly,

08:05.265 --> 08:09.555
persistence between four to 70% across two generations.

08:10.215 --> 08:11.235
So this is very high,

08:11.815 --> 08:16.045
and this kind of suggests maybe who becomes an inventor

08:16.625 --> 08:20.045
relies a lot on your family background rather than your

08:20.145 --> 08:21.325
actual ability.

08:21.785 --> 08:26.085
In this study, I used the, patent data from u

08:26.265 --> 08:29.325
US PTO, the utility patents, grant data.

08:29.945 --> 08:33.165
And, for the inventors, again,

08:33.165 --> 08:36.205
it's coming from the same source, for the,

08:36.225 --> 08:40.765
census data from 1930, I use, the US census

08:41.285 --> 08:42.645
a again, from 1930.

08:42.955 --> 08:47.645
This is 1% of the whole, census data.

08:48.065 --> 08:50.565
So it's really a huge data set.

08:50.565 --> 08:54.725
And the way they are combined is through combining inventors

08:54.825 --> 08:57.815
to potential family backgrounds in 1930

08:57.815 --> 08:58.855
through their surnames.

08:59.315 --> 09:02.295
So this is a, just a surname matching algorithm.

09:02.915 --> 09:05.575
And, kind of the methodology I use

09:05.575 --> 09:08.415
after having constructed this dataset is the following.

09:08.715 --> 09:13.695
So I look at the predictive power of income associated

09:13.735 --> 09:15.695
with your surname back in 1930,

09:16.315 --> 09:19.775
and education associated with your surname back in 1930

09:20.395 --> 09:23.855
on one, your probability to become an inventor.

09:24.325 --> 09:27.375
This is basically the number of inventors

09:27.375 --> 09:30.615
with your surname divided by the total number of people

09:30.685 --> 09:35.625
with the same surname in the United States between 1976

09:35.685 --> 09:36.825
and 2006.

09:37.835 --> 09:41.935
And, then if you became an inventor, I look at,

09:43.255 --> 09:45.535
I measure your, productivity

09:45.915 --> 09:47.735
as an inventor in various different phase.

09:48.035 --> 09:50.935
One could be just look at your total number of patents.

09:51.405 --> 09:53.815
Another is to look at,

09:53.955 --> 09:55.895
how many citations your patents get.

09:56.275 --> 09:59.255
And I also use several different measures,

09:59.255 --> 10:01.335
people in the innovation that I should have come up with,

10:01.365 --> 10:02.895
such as the originality

10:02.995 --> 10:04.615
and generality of your inventions,

10:04.825 --> 10:09.055
which would mean combining knowledge from a vast array

10:09.075 --> 10:10.495
of different technology classes,

10:11.075 --> 10:12.895
or having found a patent

10:13.165 --> 10:16.855
that cited across various different technological fields.

10:17.355 --> 10:22.175
So, and across all of these measures, it turns out,

10:22.325 --> 10:24.975
that, we get a very distinct pattern.

10:26.675 --> 10:29.925
What predicts your probability of being an inventor

10:30.105 --> 10:31.845
and what predicts your success

10:31.845 --> 10:33.965
as an inventor turns out to be different.

10:34.945 --> 10:39.365
So it turns out that your, the income associated

10:39.365 --> 10:42.365
with your surname has a much higher predictive power

10:42.905 --> 10:47.385
in explaining whether you'll become an inventor, whereas

10:47.895 --> 10:51.585
condition becoming an inventor, your, the,

10:51.685 --> 10:53.265
the education associated

10:53.265 --> 10:55.985
with your surname has a much higher predictive power.

10:56.895 --> 10:58.985
This tells us something very remarkable.

10:59.205 --> 11:03.735
It tells us that people who we would like to be inventors

11:04.855 --> 11:09.045
are not the same people as who,

11:09.625 --> 11:12.525
who we select as a society to be inventors

11:12.595 --> 11:13.965
through various mechanism.

11:14.105 --> 11:17.605
It turns out there is this kind of misallocation of talent

11:18.225 --> 11:21.125
in the sense that if you are from a richer background,

11:21.125 --> 11:22.565
you are more likely to be an inventor.

11:23.075 --> 11:26.365
That is not what predicts your success as an I inventor,

11:27.465 --> 11:29.295
which suggests to me

11:29.325 --> 11:31.455
that there is some misallocation going on.

11:31.555 --> 11:36.175
And, kind of the gist of my study focuses on,

11:36.475 --> 11:38.735
one, should we care about this missile location?

11:38.735 --> 11:40.775
Like how large is this misallocation?

11:40.775 --> 11:43.295
And two, can we do something about it?

11:43.675 --> 11:46.655
So can we come up with policies to alleviate,

11:46.835 --> 11:50.415
if not eliminate this, missile location in innovation?

11:50.865 --> 11:52.215
Definitely. So,

11:52.595 --> 11:56.085
and becoming an inventor is a very long process, right?

11:56.185 --> 12:00.765
So if you look at, data on, recent inventors,

12:00.765 --> 12:03.165
what we see is that, two thirds

12:03.165 --> 12:06.165
of inventors in the United States have a master's degree

12:06.165 --> 12:08.885
or higher, and one third of them have a PhD.

12:09.825 --> 12:13.525
So basically the way to get into the inventor profession is,

12:14.085 --> 12:15.325
although there might be exceptions,

12:15.945 --> 12:19.845
you need a high quality education, the graduate degree,

12:19.865 --> 12:20.925
at a good university.

12:20.945 --> 12:23.685
So that, that's the usual pathway, the common pathway

12:24.195 --> 12:25.285
into becoming an inventor.

12:25.285 --> 12:27.405
The garage inventors, we allow them,

12:27.465 --> 12:32.165
but if you look at the data, they account for single digits

12:34.015 --> 12:36.045
below 5% of all inventors.

12:36.625 --> 12:39.685
The most common inventors are working for companies,

12:39.685 --> 12:41.325
to a lesser degree for universities.

12:41.465 --> 12:44.765
So that's the pathway into being an inventor today.

12:45.425 --> 12:49.885
So, what I have in mind is that, if you are,

12:50.105 --> 12:53.565
if you come from a richer background,

12:54.385 --> 12:58.885
you have this, you have this leg up in becoming an inventor,

12:59.375 --> 13:03.245
maybe your family background can,

13:04.505 --> 13:07.165
make you have an easier time getting

13:07.315 --> 13:08.525
into a good university.

13:08.865 --> 13:11.085
And from that, getting into a good graduate program,

13:11.665 --> 13:12.685
so on and so forth.

13:12.745 --> 13:15.685
And it also, I mean, trying to become an I inventor,

13:15.685 --> 13:17.245
this is a superstar profession.

13:17.245 --> 13:20.525
We know that a few inventors are very prolific.

13:20.525 --> 13:21.845
Most of them have one

13:21.845 --> 13:23.125
or two patents to their name,

13:23.345 --> 13:25.325
so it's a very risky profession as well.

13:25.545 --> 13:27.965
So coming from a wealthy background helps,

13:27.965 --> 13:31.205
because if you're risk averse, you'd like to,

13:31.595 --> 13:35.445
pick a more insured profession, maybe become an accountant,

13:35.505 --> 13:39.965
or, I know, another profession

13:39.965 --> 13:42.485
that would give you more secure income without this risk

13:42.505 --> 13:45.525
of trying to become an inventor and now failing, right?

13:46.185 --> 13:49.265
So, given this context,

13:49.405 --> 13:52.145
so the mechanism I had in mind was that,

13:52.445 --> 13:55.545
people from richer families could afford to kind

13:55.545 --> 13:57.305
of take this risk and become inventors,

13:57.365 --> 13:59.905
but once they enter into the pool of inventors,

14:00.215 --> 14:01.545
they're not as prolific.

14:02.045 --> 14:05.245
But the problem here is, as a society,

14:05.305 --> 14:07.645
we don't have infinite inventor jobs.

14:07.885 --> 14:09.965
I mean, even this graduate training, the,

14:09.985 --> 14:12.765
the spots in the best universities,

14:12.835 --> 14:14.365
best graduate programs, they're limited.

14:14.785 --> 14:18.585
So whenever one of these mediocre,

14:18.585 --> 14:22.905
let's say a little bit more mediocre,

14:25.535 --> 14:28.865
inventors, enter into the, inventor profession,

14:29.455 --> 14:32.785
they're kind of blocking the path for other,

14:32.815 --> 14:34.025
perhaps more talented,

14:34.285 --> 14:38.505
but kind of, people with, less means

14:39.005 --> 14:43.145
to compete from getting into the, inventor talent pool.

14:43.165 --> 14:47.155
So basically, this is a question of who we want

14:47.155 --> 14:50.955
as inventors, given our constraints of on

14:51.105 --> 14:53.325
how many inventors we can have to kind

14:53.325 --> 14:54.965
of put numbers on this question.

14:55.025 --> 14:57.005
So we see this, these correlation patterns,

14:57.265 --> 15:00.925
but nothing tells us how, bad it is for our society.

15:00.925 --> 15:04.125
So I wanted to put some numbers on this inefficiency,

15:04.185 --> 15:05.565
should we really care about this?

15:05.625 --> 15:08.045
And, so what can we do about it?

15:08.185 --> 15:11.045
So I, I created a structural model,

15:11.655 --> 15:13.205
which can speak to these facts.

15:13.425 --> 15:17.965
And, the novelty of this model, is mostly on,

15:18.145 --> 15:19.605
the household side of the economy.

15:19.865 --> 15:22.245
So past work had focused on the firm side.

15:22.325 --> 15:23.805
I focused more on the household side.

15:24.265 --> 15:26.485
So, I borrowed this framework from

15:26.995 --> 15:29.445
what we call the heterogeneous agents literature,

15:29.655 --> 15:32.605
which treats individual households separately.

15:32.905 --> 15:35.925
So I keep track of the wealth of these households,

15:35.925 --> 15:36.925
the education level,

15:37.505 --> 15:40.285
and, their unobserved ability,

15:40.615 --> 15:42.725
their unobserved productivity as an inventor,

15:43.345 --> 15:46.325
and then kind of build the,

15:47.495 --> 15:50.505
model in which financial frictions

15:51.245 --> 15:54.305
can prevent individuals from becoming inventors,

15:54.775 --> 15:57.345
even if they will be great inventors otherwise.

15:58.085 --> 15:59.665
And the main mechanism

15:59.695 --> 16:01.905
that gets this is a tournament mechanism.

16:02.965 --> 16:05.225
So basically, you can think of this as,

16:05.225 --> 16:07.865
people trying their best to become inventors.

16:08.445 --> 16:12.305
So maybe you will try, very hard to get good grades,

16:12.395 --> 16:14.305
maybe succeed in the SAT

16:14.725 --> 16:18.865
or do whatever else you might need to increase your chance

16:18.865 --> 16:21.145
of admission to these high quality graduate

16:21.145 --> 16:22.505
programs or colleges.

16:23.165 --> 16:25.895
And, there are

16:26.545 --> 16:29.215
three ways in my model,

16:29.215 --> 16:31.295
through which you can become an i inventor.

16:31.915 --> 16:35.375
One, you could be very innately talented.

16:35.635 --> 16:38.615
So this is like the cream part does the cream rice part.

16:38.945 --> 16:43.415
Maybe you were just lucky to be able to produce great ideas,

16:43.475 --> 16:44.695
you just have a talent for it.

16:45.515 --> 16:47.375
Two, you might have,

16:47.695 --> 16:49.815
obtained a very great education pre-college,

16:50.075 --> 16:53.255
so the end your parents spending can help with that.

16:53.795 --> 16:57.935
And third, there's this additional investment, which I

16:58.475 --> 17:02.495
called credentialing spending, which is basically anything

17:02.725 --> 17:04.215
that you can spend money on

17:04.515 --> 17:06.695
to give you a leg up in these admissions.

17:07.195 --> 17:09.055
And you could think of many things.

17:09.195 --> 17:12.295
Uh, this could simply be your parents being able

17:12.295 --> 17:13.375
to pay for your tuition.

17:13.375 --> 17:14.575
That's the most,

17:14.575 --> 17:17.175
simple case rather than looking for a scholarship.

17:17.175 --> 17:18.655
You, your parents can pay for it.

17:18.915 --> 17:22.895
But also the, we know of these very interesting examples

17:22.895 --> 17:26.775
where parents of some kids open up a million dollar NGO

17:26.835 --> 17:28.375
for them, give them a leg up,

17:28.395 --> 17:32.015
and then they get into, these great universities.

17:32.795 --> 17:35.335
So anything you could manipulate through

17:36.135 --> 17:39.455
spending money on this margin I call expand.

17:40.275 --> 17:42.655
And then, I kind of estimate this model

17:42.655 --> 17:44.495
with the data I have already mentioned.

17:44.715 --> 17:48.175
Uh, I make the model,

17:48.995 --> 17:51.285
exactly replicate the correlation patterns

17:51.285 --> 17:52.485
that I have seen in the data.

17:52.825 --> 17:55.245
So that's going to inform me about the strength

17:55.245 --> 17:57.565
of this credentialing margin as opposed

17:57.625 --> 18:00.005
to more productive components of

18:00.005 --> 18:03.085
what predicts your probability of becoming an inventor.

18:03.585 --> 18:08.415
And, after estimating this model,

18:09.615 --> 18:11.895
I run two experiments.

18:12.595 --> 18:14.575
One is a very hypothetical experiment.

18:14.635 --> 18:18.125
One is, if we could get rid of this credential expanding,

18:18.125 --> 18:21.045
if we could get rid of this, the effect of

18:21.675 --> 18:24.805
your parental wealth on you becoming an inventor,

18:25.845 --> 18:28.975
what would've happened could, like what, what is the cost

18:29.155 --> 18:31.335
to the society of this credential expanding?

18:31.835 --> 18:34.775
It turns out that if we could kind

18:34.775 --> 18:37.215
of completely eliminate this credential expanding,

18:38.305 --> 18:40.285
we would have a much higher growth rate,

18:41.375 --> 18:44.635
but this higher growth rate would come coupled with,

18:44.665 --> 18:47.595
also a little bit increase in equality at the top end

18:47.615 --> 18:48.995
of the income distribution.

18:49.455 --> 18:51.955
So this is basically when we let the best

18:51.955 --> 18:55.075
and the brightest become inventors, actually, you see

18:55.665 --> 18:57.955
much more productive inventors in the society,

18:57.975 --> 18:59.035
and they are all, of course,

18:59.185 --> 19:01.595
paid a little bit more, consequently.

19:01.935 --> 19:04.435
So you have both an increase in efficiency

19:04.855 --> 19:06.235
and an increase in inequality.

19:06.625 --> 19:08.475
However, the net effect is

19:08.475 --> 19:10.875
that basically everybody is better off

19:10.875 --> 19:13.555
because the extra growth,

19:13.695 --> 19:17.035
raises all the bots at the same time so that,

19:17.035 --> 19:20.315
everybody benefits from this despite increase in inequality.

19:21.495 --> 19:24.435
But this is kind of a very abstract thought experiment.

19:24.455 --> 19:27.315
We, we probably could never completely eliminate this

19:27.315 --> 19:28.355
credential, expand it.

19:28.975 --> 19:32.925
So I kind of tried to come up with a policy tool

19:32.925 --> 19:36.085
that we can use in the real world to kind

19:36.085 --> 19:38.285
of strike at the main mechanism

19:38.285 --> 19:40.165
through which this missal location is happening.

19:41.225 --> 19:45.645
And, as I mentioned, the

19:46.385 --> 19:49.725
wealth that you, inherit from your,

19:49.725 --> 19:53.285
parents is basically what's driving this,

19:53.515 --> 19:54.605
missile location here.

19:54.985 --> 19:58.605
So, people with, people who are very wealthy

19:58.705 --> 20:02.965
but not as good, that wouldn't make great inventors,

20:03.625 --> 20:07.415
can kind of, kick out,

20:07.565 --> 20:10.575
potential Einsteins who are coming from poorer backgrounds.

20:11.355 --> 20:15.135
So what if we kind of implemented, tax

20:15.315 --> 20:17.055
or some estate taxation?

20:17.595 --> 20:20.015
So what if we text,

20:20.015 --> 20:24.335
what you can inherit from your parents in a progressive way?

20:24.915 --> 20:28.255
So I calculated this optimal progressive,

20:29.115 --> 20:32.815
estate taxation, implied by my model.

20:33.195 --> 20:35.695
And it turns out that, it is very progressive.

20:35.875 --> 20:40.295
It is so progressive that, 95%

20:40.315 --> 20:43.655
of the households would be better off, with this scheme.

20:43.835 --> 20:47.965
So 95% are net recipients from this transfer scheme,

20:48.185 --> 20:50.925
and just the top 5%, in terms

20:50.925 --> 20:53.805
of b requests are paying into the system.

20:54.505 --> 20:57.965
And despite this, it can still achieve,

20:59.625 --> 21:01.365
one quarter of the increasing growth

21:01.595 --> 21:05.645
that we would have under the more, less realistic,

21:06.145 --> 21:07.885
the less realistic scenario I mentioned.

21:08.425 --> 21:11.995
And, you can achieve this higher growth

21:12.835 --> 21:16.335
simultaneously with having the same equality,

21:16.395 --> 21:19.495
so you don't increase inequality and you get efficiency.

21:19.515 --> 21:22.535
So basically this tells us that the current

21:24.245 --> 21:28.045
situation we are in is not one in which we have an equity

21:28.045 --> 21:29.165
efficiency trade off.

21:29.545 --> 21:33.765
We could get better efficiency without giving up on equity.

21:34.545 --> 21:36.685
So, this is something we could do.

21:36.745 --> 21:41.165
And, something worth, thinking about,

21:41.265 --> 21:44.245
in the context of the United States at least, what,

21:44.305 --> 21:46.085
the main insights, I think

21:46.235 --> 21:49.085
that can be taken from my research is that,

21:49.095 --> 21:50.605
there is a significant amount

21:50.605 --> 21:53.445
of misallocation in innovation in the United States.

21:53.585 --> 21:55.965
So what can we do to combat this?

21:56.345 --> 21:59.325
As I mentioned, we could think of transfer schemes like,

21:59.955 --> 22:00.965
bequest taxation.

22:00.965 --> 22:02.285
That's one way to go forward.

22:03.145 --> 22:07.165
Uh, another way would be to kind of reduce the influence of

22:07.695 --> 22:10.965
these credential expanding, like what you can do

22:11.625 --> 22:14.765
in college admissions to re the process in your favor.

22:15.105 --> 22:19.445
So to the extent we could get rid of the these,

22:19.665 --> 22:23.045
and assure a better allocation of talent into our colleges,

22:23.045 --> 22:27.885
into our graduate programs, we could combat, this,

22:29.265 --> 22:31.405
misallocation of talent, in innovation

22:31.505 --> 22:32.605
and to link it

22:32.705 --> 22:35.245
to the broader literature on economic growth nowadays.

22:35.745 --> 22:39.845
So, there has been, a surge,

22:40.225 --> 22:44.205
in research on what is happening on, our,

22:44.265 --> 22:46.085
what's happening to our productivity growth.

22:46.145 --> 22:48.765
So, what we have observed in the past couple

22:48.765 --> 22:51.845
of decades is that, our productivity growth

22:52.915 --> 22:56.215
has been going down at the aggregate level despite spending

22:56.365 --> 22:58.335
more on research and development.

22:58.635 --> 23:01.095
So we are spending more on research and development.

23:01.455 --> 23:04.855
Actually, we have more inventors than ever more research

23:05.005 --> 23:06.095
workers than ever.

23:06.675 --> 23:10.095
And despite this, our productive growth is lackluster.

23:10.095 --> 23:14.185
It's going down. So there's this paper by,

23:15.045 --> 23:19.865
Chad Jones, Nick Blueman co-authors, it's titled,

23:20.085 --> 23:21.905
are Ideas Getting Harder to Find?

23:22.085 --> 23:25.225
And, their answer is Yes, ideas are getting harder

23:25.285 --> 23:26.745
to find our r

23:26.745 --> 23:30.575
and d efficiencies holding down both at the economy level.

23:31.195 --> 23:34.135
If you look at the firm level for us public firms,

23:34.675 --> 23:39.095
and if you look at individual, sectors like,

23:39.115 --> 23:43.455
pharmaceuticals, semiconductors, we see this fall in, r

23:43.455 --> 23:44.575
and d efficiency everywhere.

23:45.385 --> 23:47.445
One channel that could explain this whole R

23:47.445 --> 23:49.845
and ED efficiency is what I have found.

23:50.345 --> 23:54.805
If we are kind of selecting, not the best people

23:54.905 --> 23:57.405
as inventors, if we are missing out on this potential

23:57.405 --> 23:59.685
license, that would reduce our r and d efficiency

24:00.145 --> 24:01.525
and kind of explain

24:02.805 --> 24:05.385
to an extent why we have lower growth.

24:06.165 --> 24:08.145
And I think, more,

24:09.185 --> 24:10.945
targeted policies could be thought of.

24:11.165 --> 24:12.345
Uh, in this context.

24:13.145 --> 24:14.865
I have investigated a few more,

24:14.965 --> 24:17.105
and it seems to be a very complicated problem.

24:17.245 --> 24:21.065
So, for instance, I have also tried looking at subsidizing,

24:21.525 --> 24:24.505
spending on education, whether that would help.

24:25.045 --> 24:27.465
And, as a first pass, you might think it might help,

24:27.525 --> 24:28.905
and that was my intuition actually.

24:29.285 --> 24:31.505
But when you go through the whole structural model,

24:31.805 --> 24:34.625
it turns out that when you kind

24:34.625 --> 24:38.225
of offer these un-targeted subsidies, the rich again,

24:38.285 --> 24:40.345
are the ones who benefit from this the most.

24:40.805 --> 24:44.545
And since it's a race between everyone in the economy

24:44.605 --> 24:47.545
to become an inventor, you still do not achieve a better,

24:47.765 --> 24:48.985
allocation of talent.

24:49.365 --> 24:50.865
So it's a very delicate question.

24:51.805 --> 24:54.185
Uh, I have some proposals,

24:54.185 --> 24:56.385
but I think more research needs to be conducted

24:56.645 --> 24:59.985
to find very targeted policies that could kind

24:59.985 --> 25:01.865
of discover these missing Einsteins

25:02.125 --> 25:03.865
and get them the education they need,

25:04.205 --> 25:08.065
and so that we can benefit as a society, from their,

25:08.305 --> 25:08.905
brilliant minds.

00:00 - Thank you very much for the question and inviting me. So, this question, does the cream rise to the top, I think is very important for a great number of professions, not just inventors, but I focused on this particular question of whether we always are able to get the best

00:22 - and brightest minds in a society to, become inventors and, benefit from their ideas, their ability to come up with new technologies, new processes. That was the question I was interested in. And, just to give an example, just to give an example, think of Albert Einstein. So we all know about him. We all know he's a prolific physicist. He has the paradigm shift.

00:50 - We of a lot of our, technologies, our ways of thinking to him. And, what's not as many people know is that beyond being a prolific scientist, beyond being a prolific physicist, he has also come up with, more than 50 patented inventions. So, he was not just a scientist that thought about these big ideas, but he also came up with these applicable ideas that could kind of, we could see in products

01:21 - that we can see, in our daily lives. And, if you kind of look at, Einstein's, family background, you see that he was no ordinary person. So actually, he has received a great education in a number of countries in Germany, Switzerland, Italy, and his, alma mater was, ETH Zuri. So he definitely, this received this great education. And after receiving this great education, he was able to contribute upon, the body of knowledge that we have. And the question one, has in mind when,

01:56 - they see this example is that one, how would the work, look like if Albert Einstein could not become, an inventor, but perhaps he became an accountant or a factory worker, what would we have lost? And in the case of Albert Einstein, perhaps we can say we would lose, a lot. But, when we think further, how do we know we are not missing out on potential Einsteins? So that's the bigger question.

02:24 - That's what, has motivated me to conduct this study. And, to give some, brief background on, kind of what has been done, in the economics literature on this topic. So, in the nineties there was, there were these seminal papers by Philippa Guion, Peter Hoit and, Paul Romer on endogenous growth. So we have concluded that although there can be many sources of economic growth in the long run, in the very long run, much

02:55 - of our prosperity also to these new technologies, new ideas, New products. And, there has been a surge in interest in, identifying the bottlenecks in the process of creating new ideas, new technologies. in the past and before I conducted my study, the focus was mostly on the firm side. So what motivates firms to do r and d? How can we, improve the setting such

03:25 - that there's more competition, there's more innovation. A lot of the work was on, firmly on the firm side, but, all this work was ignoring the other side of the, equation, which is inventors. So who are the people who come up with, these new ideas who work for these firms, who generate these patents, that the firms use? And, I decided to look at what determines, who becomes an inventor and who becomes a prolific inventor,

04:00 - not just becoming an inventor, but a very productive inventor. And, to, look at this question, you kind of need two types of data. one is, data on who becomes inventors, and there's, record of innovations, which we have thankfully. So in the United States, we have this, micro data provided by, United States Patent and Trademark Office,

04:26 - which contains all information on all the patents since 1976, and all of their citations, all of their inventors, we have everything we can link them to firms. It's perfect. What was the missing link, in this is that we really do not know directly what the backgrounds, the family backgrounds of these inventors are. We just see their names, we know their inventions, we know nothing about their family backgrounds, which might have some influence on whether they were able to make it to the inventor pool in the first place.

05:03 - So the, kind of the novelty I came up with in the study was to do the following. So we have no direct information, but could we somehow find some proxies to get at their family backgrounds? And, interestingly, we can, and, this links to another literature, essentially. So, in, back in the, past research has established that your names and surnames, and your first name

05:33 - and last name contains a wealth of information that can tell us about, where you came from. and, there's this, seminar work by Gregory Clark. He, he, he looks at these rare surnames people with these rare surnames. do they have a higher socioeconomic status today? And he says, the answer is yes. So I kind of use the same methodology. What I did was the following.

05:59 - So I took all of the inventors, from, 1976 onwards. I looked at their surnames and I matched them to the income levels and education levels associated with their surnames back in 1930, which is the last year, we have public data available for, so this is basically saying the following, you're an inventor, you have a particular surname. I go back to 1930, and then I look at what was the income

06:33 - and education associated with your surname? And based on that, try to predict, one, your chances to be an inventor, and two, how productive you are as an inventor condition becoming one. So, and you might think of many things, you might say, this is a very noisy measure, right? So this is across two to three generations, and we are collapsing everyone by their surnames. So you'd expect this to have zero explanatory power. Very interestingly, it has such a large,

07:08 - predictive power that I, I was amazed when I first saw this. So, actually, people with a, higher income certainly are 24% more likely to be an inventor, and 18% more likely to be a very productive inventor. So even across these two to three regenerations, despite all the time lag, despite every po potential mechanism you can think of, just knowing your surname can tell us a lot about whether you'll be an inventor

07:41 - or you'll be a very productive, I invent, and this kind of tells you the following, before we dig deeper into it, it says that persistence of socioeconomic status as captured by becoming an inventor is immensely high. So when I go to the full quantitative model to investigate this, this corresponds to roughly, persistence between four to 70% across two generations. So this is very high, and this kind of suggests maybe who becomes an inventor

08:16 - relies a lot on your family background rather than your actual ability. In this study, I used the, patent data from u US PTO, the utility patents, grant data. And, for the inventors, again, it's coming from the same source, for the, census data from 1930, I use, the US census a again, from 1930. This is 1% of the whole, census data. So it's really a huge data set.

08:50 - And the way they are combined is through combining inventors to potential family backgrounds in 1930 through their surnames. So this is a, just a surname matching algorithm. And, kind of the methodology I use after having constructed this dataset is the following. So I look at the predictive power of income associated with your surname back in 1930, and education associated with your surname back in 1930 on one, your probability to become an inventor.

09:24 - This is basically the number of inventors with your surname divided by the total number of people with the same surname in the United States between 1976 and 2006. And, then if you became an inventor, I look at, I measure your, productivity as an inventor in various different phase. One could be just look at your total number of patents. Another is to look at, how many citations your patents get.

09:56 - And I also use several different measures, people in the innovation that I should have come up with, such as the originality and generality of your inventions, which would mean combining knowledge from a vast array of different technology classes, or having found a patent that cited across various different technological fields. So, and across all of these measures, it turns out, that, we get a very distinct pattern.

10:26 - What predicts your probability of being an inventor and what predicts your success as an inventor turns out to be different. So it turns out that your, the income associated with your surname has a much higher predictive power in explaining whether you'll become an inventor, whereas condition becoming an inventor, your, the, the education associated with your surname has a much higher predictive power. This tells us something very remarkable.

10:59 - It tells us that people who we would like to be inventors are not the same people as who, who we select as a society to be inventors through various mechanism. It turns out there is this kind of misallocation of talent in the sense that if you are from a richer background, you are more likely to be an inventor. That is not what predicts your success as an I inventor, which suggests to me that there is some misallocation going on.

11:31 - And, kind of the gist of my study focuses on, one, should we care about this missile location? Like how large is this misallocation? And two, can we do something about it? So can we come up with policies to alleviate, if not eliminate this, missile location in innovation? Definitely. So, and becoming an inventor is a very long process, right? So if you look at, data on, recent inventors, what we see is that, two thirds

12:03 - of inventors in the United States have a master's degree or higher, and one third of them have a PhD. So basically the way to get into the inventor profession is, although there might be exceptions, you need a high quality education, the graduate degree, at a good university. So that, that's the usual pathway, the common pathway into becoming an inventor. The garage inventors, we allow them, but if you look at the data, they account for single digits

12:34 - below 5% of all inventors. The most common inventors are working for companies, to a lesser degree for universities. So that's the pathway into being an inventor today. So, what I have in mind is that, if you are, if you come from a richer background, you have this, you have this leg up in becoming an inventor, maybe your family background can, make you have an easier time getting into a good university.

13:08 - And from that, getting into a good graduate program, so on and so forth. And it also, I mean, trying to become an I inventor, this is a superstar profession. We know that a few inventors are very prolific. Most of them have one or two patents to their name, so it's a very risky profession as well. So coming from a wealthy background helps, because if you're risk averse, you'd like to,

13:31 - pick a more insured profession, maybe become an accountant, or, I know, another profession that would give you more secure income without this risk of trying to become an inventor and now failing, right? So, given this context, so the mechanism I had in mind was that, people from richer families could afford to kind of take this risk and become inventors, but once they enter into the pool of inventors, they're not as prolific.

14:02 - But the problem here is, as a society, we don't have infinite inventor jobs. I mean, even this graduate training, the, the spots in the best universities, best graduate programs, they're limited. So whenever one of these mediocre, let's say a little bit more mediocre, inventors, enter into the, inventor profession, they're kind of blocking the path for other, perhaps more talented,

14:34 - but kind of, people with, less means to compete from getting into the, inventor talent pool. So basically, this is a question of who we want as inventors, given our constraints of on how many inventors we can have to kind of put numbers on this question. So we see this, these correlation patterns, but nothing tells us how, bad it is for our society. So I wanted to put some numbers on this inefficiency, should we really care about this?

15:05 - And, so what can we do about it? So I, I created a structural model, which can speak to these facts. And, the novelty of this model, is mostly on, the household side of the economy. So past work had focused on the firm side. I focused more on the household side. So, I borrowed this framework from what we call the heterogeneous agents literature, which treats individual households separately.

15:32 - So I keep track of the wealth of these households, the education level, and, their unobserved ability, their unobserved productivity as an inventor, and then kind of build the, model in which financial frictions can prevent individuals from becoming inventors, even if they will be great inventors otherwise. And the main mechanism that gets this is a tournament mechanism.

16:02 - So basically, you can think of this as, people trying their best to become inventors. So maybe you will try, very hard to get good grades, maybe succeed in the SAT or do whatever else you might need to increase your chance of admission to these high quality graduate programs or colleges. And, there are three ways in my model, through which you can become an i inventor.

16:31 - One, you could be very innately talented. So this is like the cream part does the cream rice part. Maybe you were just lucky to be able to produce great ideas, you just have a talent for it. Two, you might have, obtained a very great education pre-college, so the end your parents spending can help with that. And third, there's this additional investment, which I called credentialing spending, which is basically anything that you can spend money on

17:04 - to give you a leg up in these admissions. And you could think of many things. this could simply be your parents being able to pay for your tuition. That's the most, simple case rather than looking for a scholarship. You, your parents can pay for it. But also the, we know of these very interesting examples where parents of some kids open up a million dollar NGO for them, give them a leg up,

17:28 - and then they get into, these great universities. So anything you could manipulate through spending money on this margin I call expand. And then, I kind of estimate this model with the data I have already mentioned. I make the model, exactly replicate the correlation patterns that I have seen in the data. So that's going to inform me about the strength of this credentialing margin as opposed

17:57 - to more productive components of what predicts your probability of becoming an inventor. And, after estimating this model, I run two experiments. One is a very hypothetical experiment. One is, if we could get rid of this credential expanding, if we could get rid of this, the effect of your parental wealth on you becoming an inventor, what would've happened could, like what, what is the cost to the society of this credential expanding?

18:31 - It turns out that if we could kind of completely eliminate this credential expanding, we would have a much higher growth rate, but this higher growth rate would come coupled with, also a little bit increase in equality at the top end of the income distribution. So this is basically when we let the best and the brightest become inventors, actually, you see much more productive inventors in the society, and they are all, of course,

18:59 - paid a little bit more, consequently. So you have both an increase in efficiency and an increase in inequality. However, the net effect is that basically everybody is better off because the extra growth, raises all the bots at the same time so that, everybody benefits from this despite increase in inequality. But this is kind of a very abstract thought experiment. We, we probably could never completely eliminate this

19:27 - credential, expand it. So I kind of tried to come up with a policy tool that we can use in the real world to kind of strike at the main mechanism through which this missal location is happening. And, as I mentioned, the wealth that you, inherit from your, parents is basically what's driving this, missile location here. So, people with, people who are very wealthy

19:58 - but not as good, that wouldn't make great inventors, can kind of, kick out, potential Einsteins who are coming from poorer backgrounds. So what if we kind of implemented, tax or some estate taxation? So what if we text, what you can inherit from your parents in a progressive way? So I calculated this optimal progressive, estate taxation, implied by my model. And it turns out that, it is very progressive.

20:35 - It is so progressive that, 95% of the households would be better off, with this scheme. So 95% are net recipients from this transfer scheme, and just the top 5%, in terms of b requests are paying into the system. And despite this, it can still achieve, one quarter of the increasing growth that we would have under the more, less realistic, the less realistic scenario I mentioned. And, you can achieve this higher growth

21:12 - simultaneously with having the same equality, so you don't increase inequality and you get efficiency. So basically this tells us that the current situation we are in is not one in which we have an equity efficiency trade off. We could get better efficiency without giving up on equity. So, this is something we could do. And, something worth, thinking about, in the context of the United States at least, what, the main insights, I think

21:46 - that can be taken from my research is that, there is a significant amount of misallocation in innovation in the United States. So what can we do to combat this? As I mentioned, we could think of transfer schemes like, bequest taxation. That's one way to go forward. another way would be to kind of reduce the influence of these credential expanding, like what you can do in college admissions to re the process in your favor.

22:15 - So to the extent we could get rid of the these, and assure a better allocation of talent into our colleges, into our graduate programs, we could combat, this, misallocation of talent, in innovation and to link it to the broader literature on economic growth nowadays. So, there has been, a surge, in research on what is happening on, our, what's happening to our productivity growth. So, what we have observed in the past couple

22:48 - of decades is that, our productivity growth has been going down at the aggregate level despite spending more on research and development. So we are spending more on research and development. Actually, we have more inventors than ever more research workers than ever. And despite this, our productive growth is lackluster. It's going down. So there's this paper by, Chad Jones, Nick Blueman co-authors, it's titled, are Ideas Getting Harder to Find?

23:22 - And, their answer is Yes, ideas are getting harder to find our r and d efficiencies holding down both at the economy level. If you look at the firm level for us public firms, and if you look at individual, sectors like, pharmaceuticals, semiconductors, we see this fall in, r and d efficiency everywhere. One channel that could explain this whole R and ED efficiency is what I have found. If we are kind of selecting, not the best people

23:54 - as inventors, if we are missing out on this potential license, that would reduce our r and d efficiency and kind of explain to an extent why we have lower growth. And I think, more, targeted policies could be thought of. in this context. I have investigated a few more, and it seems to be a very complicated problem. So, for instance, I have also tried looking at subsidizing,

24:21 - spending on education, whether that would help. And, as a first pass, you might think it might help, and that was my intuition actually. But when you go through the whole structural model, it turns out that when you kind of offer these un-targeted subsidies, the rich again, are the ones who benefit from this the most. And since it's a race between everyone in the economy to become an inventor, you still do not achieve a better, allocation of talent.

24:49 - So it's a very delicate question. I have some proposals, but I think more research needs to be conducted to find very targeted policies that could kind of discover these missing Einsteins and get them the education they need, and so that we can benefit as a society, from their, brilliant minds.

Faculti AI can make mistakes