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

Supplier Churn and Growth

View more insights in Social Sciences - Economics - Microeconomics

View more figures from University of California Los Angeles

View the page of David Baqaee

Read the Study

It is widely believed that access to new suppliers generates a surplus for consumers and that this surplus drives long-run growth and trade. Despite its importance, little direct empirical evidence exists on the magnitude of this surplus. David R. Baqaee discusses this effect’s impact on both micro- and macroeconomic outcomes.

Image courtesy of interviewee. February 1, 2024

Generating Workspace

Export as PDF

This video video discusses the importance of understanding the dynamics of trade and growth in economic theory, challenging the conventional wisdom based on idealized models of perfect competition. It highlights that, in a perfectly competitive market, individual firms’ existence has no substantial impact on market outcomes, which implies that introducing new firms does not contribute to economic growth in terms of GDP per capita. The conversation shifts to how in reality, new firms and products create value for consumers and have significant indirect effects on other firms, influencing their productivity and marginal costs.

The speakers suggest that actual firm behavior deviates from the perfect competition model, as firms do not operate in a vacuum; they influence each other's costs and outputs. For instance, the example of large firms like Ford lobbying for the survival of competitors illustrates the interconnectedness within supply chains, contrary to predictions based solely on perfect competition.

Furthermore, the authors present their research methodology, using detailed datasets from Belgium to measure how changes in supplier relationships (such as the entry or exit of suppliers) directly impact production costs and productivity for firms. Their findings indicate that every 1% change in supplier availability correlates with a 0.3% change in marginal costs, underscoring the significance of these supplier relationships for aggregate economic productivity.

The discussion concludes by linking these micro-level dynamics to macroeconomic implications and suggesting that a notable portion of productivity growth in economies can be attributed to changes in supply chain dynamics. Thus, the research suggests that firm entry and exit decisions may not always internalize the broader economic impacts, indicating potential inefficiencies that could warrant policy intervention. However, any policy recommendations should be approached with caution due to the complexities involved in these economic interactions and the need for further research.

Language:

Export as PDF

Export as PDF
Here are some key quotes from the video:

1. **Motivation for Understanding Economics**:
- "Well, the thing that motivated us really is understanding the deeper mechanics of what drives growth and trade."

2. **On Idealized Worlds in Economics**:
- "Oftentimes economists have a tendency to start by thinking about idealized worlds when they want to think about the world."

3. **Perfect Competition Concept**:
- "If there was an individual firm that disappeared or an individual firm that appeared... it doesn't really affect anybody."

4. **Implications for Growth**:
- "If you're thinking about a situation in which the standards of living are rising over time, you're not gonna be able to get it in this world where new firms are just introduced into the marketplace, but you have perfect competition."

5. **Nature of Trade**:
- "You would expect trade to happen between agents or countries that are very different to one another."

6. **Unique Contributions of Firms**:
- "The fact that it was different and it was better made everybody better off, and it made consumers' standard of living increase."

7. **Measuring Consumer Value**:
- "However, it's very challenging to measure this thing to measure exactly how much value an additional product has for consumers."

8. **Link Between Firms and Growth**:
- "If you are running a business and you can have access to more varieties of suppliers, then that is also gonna be beneficial to you, just like it would be to a consumer."

9. **Impact of Supplier Dynamics**:
- "If that happens, then that raises the marginal cost of production for you by about 0.3%."

10. **Aggregate Productivity Growth**:
- "Our framework says... one component of that missing residual can be explained by churn in the supply chain."

11. **Implications for Policy**:
- "This idea that it could actually be inefficient for General Motors to go out because it causes the suppliers to go out...potentially opens the door for all sorts of intervention."

12. **Complexity of Economic Interventions**:
- "The world is pretty complicated and we need to do more research to try to understand some of these."

These quotes encapsulate the central themes of the discussion about competition, trade, firm dynamics, and economic growth.

Export as PDF

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

00:00.000 --> 00:03.685
Well, the thing that motivated us really is understanding

00:03.745 --> 00:07.525
the deeper mechanics of what drives, growth and trade.

00:08.025 --> 00:10.485
Um, you think about,

00:10.485 --> 00:12.965
oftentimes economists have a tendency to start

00:13.105 --> 00:14.485
by thinking about idealized

00:14.485 --> 00:15.805
worlds when they want to think about the world.

00:16.505 --> 00:21.005
And really, the most, famous benchmark idealized world

00:21.005 --> 00:23.285
that we think about as a starting point is a world

00:23.285 --> 00:25.085
with perfect comp anonymous markets.

00:25.785 --> 00:28.725
Um, and, no individual firm

00:28.825 --> 00:31.525
or no individual customer has any kind of market power.

00:32.385 --> 00:34.165
So if you think about a world like that, a world

00:34.165 --> 00:37.365
with perfect competition, then one thing

00:37.365 --> 00:40.845
that you suddenly realize is if there was an individual firm

00:40.955 --> 00:44.125
that disappeared or, or an individual firm that appeared.

00:44.185 --> 00:46.205
So, like you think about going to the farmer's market,

00:46.205 --> 00:48.405
which is sometimes used as an example of something

00:48.405 --> 00:50.645
that's close to a compe, perfectly competitive market.

00:50.945 --> 00:52.005
If there's a new stall

00:52.005 --> 00:55.165
that opens up in a big farmer's market selling more

00:55.445 --> 00:57.525
tomatoes, it doesn't really affect anybody.

00:57.665 --> 01:00.485
It doesn't affect the market price of tomatoes,

01:00.825 --> 01:04.405
and it doesn't affect anybody's ability to get get tomatoes.

01:05.825 --> 01:10.285
So in that world, you have perfect competition.

01:10.425 --> 01:13.445
As I say, like a necessary consequence of that is

01:13.445 --> 01:16.365
that each individual firm doesn't matter.

01:16.765 --> 01:18.645
Collectively, the firms are gonna matter,

01:18.705 --> 01:20.205
but one by one, they don't matter.

01:20.665 --> 01:23.765
Now, this simple observation has important implications

01:23.825 --> 01:25.765
for growth, long run economic growth,

01:25.825 --> 01:27.245
and for economic trade.

01:27.425 --> 01:29.245
So let's start by thinking about the implications

01:29.305 --> 01:30.805
for, for growth.

01:31.505 --> 01:33.125
If you think of this farmer's market

01:33.225 --> 01:36.325
and you think about adding more fruit stalls,

01:36.325 --> 01:40.285
like one additional guy adding a new fruit stall, it's gonna

01:40.845 --> 01:42.645
increase total output.

01:42.755 --> 01:45.125
It's gonna increase the GDP of the farmer's market,

01:45.425 --> 01:48.365
but it's not gonna increase GDP per person, which is

01:48.365 --> 01:50.485
how we normally think about economic growth.

01:50.905 --> 01:52.645
So if you're thinking about a

01:53.165 --> 01:54.245
situation in which the standards

01:54.245 --> 01:57.045
of living are rising over time, you're not gonna be able

01:57.045 --> 01:58.165
to get it in this world

01:58.175 --> 02:01.765
where new firms just introduce are introduced into the

02:01.765 --> 02:03.525
marketplace, but you have perfect competition.

02:03.915 --> 02:06.685
Okay? What are the implications for trade?

02:06.755 --> 02:08.685
Well, if you think about trade in such a world,

02:08.785 --> 02:12.085
you would expect trade to happen between agents

02:12.185 --> 02:14.845
or countries that are very different to one another.

02:15.305 --> 02:18.005
So the guys who are making tomatoes are not gonna be trading

02:18.005 --> 02:19.245
with other guys who make tomatoes

02:19.245 --> 02:20.485
because they make the same thing.

02:20.985 --> 02:23.765
You would expect the guy who makes bananas to trade

02:23.765 --> 02:25.605
with the guys who make tomatoes and vice versa.

02:26.985 --> 02:29.085
But both of these implications don't seem

02:29.085 --> 02:30.485
to be borne out in the data.

02:30.905 --> 02:32.885
So like, if you think about the growth implications,

02:32.905 --> 02:34.445
you think about, for example, apple

02:35.085 --> 02:38.125
entering the phone market that had an effect

02:38.235 --> 02:42.205
because Apple was not just another company making the exact

02:42.205 --> 02:43.725
same type of phone as everybody else.

02:43.995 --> 02:45.525
They were making something that was different.

02:46.025 --> 02:47.285
The fact that it was different

02:47.385 --> 02:49.565
and it was better made everybody better off,

02:49.665 --> 02:52.405
and it made consumers standard of living increase.

02:53.355 --> 02:55.125
Similarly, when you look at trade,

02:55.545 --> 02:58.485
you don't actually have a situation where mostly countries

02:58.485 --> 02:59.565
that are very different

02:59.565 --> 03:01.325
to each other are trading with one another.

03:01.785 --> 03:02.845
So the picture

03:02.865 --> 03:06.285
of global trade isn't just Saudi Arabia making oil

03:06.345 --> 03:09.205
and selling it to Germany who makes cars and vice versa.

03:09.765 --> 03:12.205
A lot of trade is between countries that produce

03:12.235 --> 03:14.565
that look pretty similar and produce similar goods.

03:14.705 --> 03:17.405
So South Korea is making Samsung phones,

03:17.945 --> 03:20.525
the US is making iPhones, and they're trading them back

03:20.525 --> 03:21.525
and forth with one another.

03:22.545 --> 03:25.485
So the reason economists think these sorts

03:25.485 --> 03:26.725
of things can happen is

03:26.725 --> 03:29.885
because in practice products are not actually the same.

03:30.105 --> 03:33.285
And so new products actually generate some value

03:33.305 --> 03:34.725
for the customer,

03:35.225 --> 03:37.845
unlike the perfectly competitive benchmark.

03:39.405 --> 03:43.905
So, eh, so fundamentally this observation that firms,

03:44.435 --> 03:46.225
individual firms matter,

03:46.565 --> 03:48.945
is gonna have important implications for trade and growth.

03:49.095 --> 03:51.025
Because when you write models of trade

03:51.025 --> 03:52.905
and growth, you need a way to quantify

03:53.125 --> 03:56.345
how different are the new, new products

03:56.345 --> 03:58.105
that are being introduced and the new firms

03:58.105 --> 04:00.745
that enter the market to the preexisting ones.

04:00.745 --> 04:01.985
Because as you can imagine,

04:02.005 --> 04:04.705
that's gonna have an important role to play for whether

04:04.725 --> 04:07.185
or not equilibrium and such models are efficient,

04:07.205 --> 04:09.585
and how much growth can you get from an increase in market

04:09.695 --> 04:10.905
size and so on and so forth.

04:11.085 --> 04:12.905
So that's kind of what we're interested in.

04:12.995 --> 04:15.625
We're interested in, in a, in a concrete way.

04:16.025 --> 04:17.585
I think in the prior literature,

04:17.735 --> 04:19.665
both the theoretical literature on trade

04:19.665 --> 04:21.465
and growth emphasizes this notion

04:21.465 --> 04:24.385
of consumer surplus from new products or new.

04:24.455 --> 04:25.945
This is something that's very important.

04:26.885 --> 04:30.065
So however, it's very challenging to measure this thing

04:30.165 --> 04:33.345
to measure exactly how much value an additional product

04:33.565 --> 04:34.705
has for consumers.

04:36.125 --> 04:40.305
The reason is because when I introduce a new iPhone,

04:41.715 --> 04:43.975
for consumers, there are not, like,

04:43.975 --> 04:46.295
maybe it makes you feel better, it makes you happier

04:46.475 --> 04:48.455
as a consumer to have access to an iPhone,

04:48.635 --> 04:51.215
but there's not some piece of economic data

04:51.325 --> 04:52.935
that I can just go and look at.

04:53.005 --> 04:55.255
Perhaps I could conduct some experiments

04:55.255 --> 04:56.975
and say, well, suppose I raised the,

04:56.975 --> 04:58.775
would you still be willing to pay for it?

04:59.035 --> 05:02.135
And then under some kind of idealized, experimental setting,

05:02.295 --> 05:03.575
I could derive these values.

05:04.115 --> 05:06.695
But in general, the way economists approach the problem

05:06.755 --> 05:10.535
of figuring out the value of a new product has been,

05:11.125 --> 05:13.335
basically to use functional form assumptions.

05:13.555 --> 05:16.015
So you estimate demand systems locally,

05:16.635 --> 05:19.095
and then those demand systems, you just kind of

05:19.725 --> 05:20.775
extrapolate them out.

05:20.795 --> 05:23.255
And then the way you work out consumer surplus is you just

05:23.285 --> 05:24.975
fill in the area under the demand curve.

05:25.445 --> 05:27.615
What we do in this paper is a little bit different.

05:27.805 --> 05:30.215
What we do is we, we make the observation

05:30.215 --> 05:32.815
that these forces are gonna also matter for firms,

05:33.115 --> 05:34.255
not just for consumers.

05:35.115 --> 05:36.735
So if you are running a business

05:37.115 --> 05:41.255
and you can have access to more varieties of suppliers, then

05:41.255 --> 05:43.095
that is also gonna be beneficial to you,

05:43.095 --> 05:44.455
just like it would be to a consumer.

05:44.885 --> 05:47.575
However, for firms, there's a difference.

05:48.115 --> 05:51.055
Um, and the difference that's important for us is

05:51.055 --> 05:53.295
that firms have things called outputs.

05:53.915 --> 05:56.695
So I can measure how productive

05:56.815 --> 05:59.005
A firm is by comparing the quantity

05:59.105 --> 06:01.925
of output it can produce relative to the quantity

06:01.985 --> 06:03.365
of inputs that it's taking in.

06:03.605 --> 06:04.765
I can look at its productivity.

06:04.865 --> 06:06.165
That's not something I can do

06:06.165 --> 06:07.765
for your happiness as a consumer.

06:08.185 --> 06:10.845
So that's kind of like our key into our, our,

06:10.845 --> 06:12.805
entry point into this literature is to say, well,

06:12.905 --> 06:15.325
one way you could estimate the value of new firms,

06:15.455 --> 06:17.565
firms is not to look at the value for consumers,

06:17.585 --> 06:19.365
but to look at the value for other firms.

06:19.785 --> 06:21.885
And there we have good measures,

06:21.905 --> 06:25.325
or at least better measures of value, which is productivity

06:25.545 --> 06:28.365
or the cost of production for the downstream firms

06:28.395 --> 06:31.885
that are being, that are getting access to new products

06:31.885 --> 06:34.925
of, new products and new firms, new firms.

06:34.925 --> 06:37.325
Um, and these things are important, you know, like, so

06:37.325 --> 06:40.445
for example, there's a specific,

06:40.515 --> 06:42.885
example in this literature that people like to talk about,

06:43.055 --> 06:46.285
which is that, in, in the fall

06:46.285 --> 06:49.245
of 2008 when the great financial crisis was unfolding,

06:49.865 --> 06:53.425
the president of Ford Motor Company went to Washington, DC

06:53.525 --> 06:55.425
to, to lobby for government bailouts.

06:56.535 --> 06:58.315
Now, the thing that's interesting is

06:58.315 --> 07:01.315
that Ford was not lobbying for itself.

07:01.535 --> 07:04.155
It was lobbying on behalf of General Motors and Chrysler.

07:04.545 --> 07:05.795
They didn't need a bailout,

07:05.935 --> 07:08.395
but they wanted their biggest competitors to be bailed out.

07:08.735 --> 07:10.635
Now, why was it that Ford wanted these

07:10.635 --> 07:12.195
competitors to be bailed out?

07:12.345 --> 07:14.875
Well, the reason was because they were worried about the

07:15.075 --> 07:16.195
survival of their supply chain.

07:16.615 --> 07:18.715
So they understood that if General Motors

07:18.715 --> 07:21.635
and Chrysler exit the market, then many

07:21.655 --> 07:23.515
of their joint suppliers

07:23.515 --> 07:25.715
that they share together are gonna be knocked out.

07:26.015 --> 07:27.995
And now once those suppliers are knocked out,

07:27.995 --> 07:29.515
then Ford is gonna get into trouble.

07:30.125 --> 07:32.155
Again, this speaks to the nature

07:32.885 --> 07:35.235
departures from the perfectly competitive benchmark,

07:35.235 --> 07:37.035
because in the perfectly competitive world,

07:37.385 --> 07:39.075
Ford wouldn't care if some

07:39.075 --> 07:40.755
individual suppliers are knocked out.

07:40.855 --> 07:42.355
It would just go to the marketplace

07:42.495 --> 07:44.275
and there's other people selling the same thing,

07:44.335 --> 07:47.275
and specific individuals wouldn't have mattered.

07:48.015 --> 07:51.135
So, so this is kind of like where we're gonna look.

07:51.135 --> 07:52.815
We're gonna look at the effects of these things.

07:52.835 --> 07:53.895
For firms, micro

07:53.995 --> 07:57.565
and macro are kind of like, there's kind

07:57.565 --> 08:00.005
of a logical sequence to it, which is that you start

08:00.005 --> 08:01.565
with a micro one and you try

08:01.565 --> 08:04.285
to understand at the micro level for an individual firms,

08:04.395 --> 08:08.085
what are the consequences of getting access to new suppliers

08:08.385 --> 08:10.085
or losing access to suppliers.

08:10.345 --> 08:12.885
So you estimate that and you, you come up with a number,

08:13.305 --> 08:16.165
and then the macro part is to say, okay, well in the,

08:16.585 --> 08:18.645
in the data we see firms forming

08:18.785 --> 08:20.365
and breaking links with one another.

08:21.065 --> 08:23.885
Can we use our estimates from the micro part of the paper

08:24.545 --> 08:28.045
and observations on how many links are being formed

08:28.065 --> 08:30.685
and broken every year to try to quantify

08:30.685 --> 08:32.605
how much this matters for aggregate productivity

08:32.705 --> 08:33.725
or aggregate growth?

08:34.065 --> 08:37.165
The dataset, we, for this paper we worked with,

08:37.225 --> 08:38.565
the National Bank of Belgium.

08:39.075 --> 08:41.765
Belgium is a very, interesting test case

08:41.765 --> 08:43.565
because they have fantastic data.

08:44.385 --> 08:47.165
So, for the Belgian economy,

08:47.745 --> 08:49.685
we have manufacturing surveys

08:49.695 --> 08:53.005
where they track at the individual product

08:53.135 --> 08:54.845
level how many

08:55.935 --> 08:58.565
Units of the different products firms are shipping.

08:59.305 --> 09:02.965
And then from the tax data, we can see all the firms,

09:03.545 --> 09:07.645
the domestic firms that these Belgium firms are interacting

09:07.645 --> 09:09.805
with, who they buy from, who they sell to.

09:10.345 --> 09:15.275
Um, and then we can also see things like, their,

09:15.455 --> 09:18.795
balance sheets and their wage bill and so on.

09:18.975 --> 09:21.115
And so we kind of have all the data

09:21.115 --> 09:24.115
that we would need in order to try to figure out

09:24.345 --> 09:28.075
what is the value of, suppliers

09:28.175 --> 09:30.075
to manufacturing firms in Belgium.

09:30.075 --> 09:32.115
And then that's the thing that we extrapolate

09:32.355 --> 09:33.515
to the rest of the economy.

09:33.975 --> 09:38.315
So the thing that we start with is, the two things

09:38.315 --> 09:40.595
that we need in order to answer our question is, first

09:40.595 --> 09:41.715
of all, we need to see

09:42.895 --> 09:46.155
how good is each Belgian manufacturing firm at

09:46.155 --> 09:47.195
producing outputs.

09:47.495 --> 09:49.955
And so we can do that because we can see what they buy

09:50.015 --> 09:53.275
and we can see what they produce in terms of physical units.

09:53.415 --> 09:55.955
We can see what they produce, not just the the Euro value,

09:56.055 --> 09:58.035
but the actual units that they share.

09:58.385 --> 10:01.235
Okay? The other thing that we need is we need to know

10:01.615 --> 10:03.035
who they're interacting with.

10:03.255 --> 10:05.715
And that comes from the value added tax data

10:05.715 --> 10:07.355
that Belgium has that allows us

10:07.355 --> 10:10.675
to track at annual frequency, all firm

10:10.675 --> 10:12.275
to firm transactions in Belgium.

10:12.275 --> 10:14.195
So we see exactly who your suppliers are.

10:14.195 --> 10:16.875
If you're a manufacturing firm, the final thing

10:16.875 --> 10:19.835
that we need is what's known as a natural experiment.

10:20.255 --> 10:25.195
We need something that exogenously changes access

10:25.195 --> 10:29.835
to suppliers for these firms that is not related

10:30.415 --> 10:32.795
to some other thing that's happening to that firm.

10:33.135 --> 10:34.955
So you couldn't, for example, just go

10:34.975 --> 10:38.755
and look at correlations between firms adding suppliers

10:39.055 --> 10:40.275
and their productivity.

10:40.375 --> 10:42.595
Or in our case, the notion we use is marginal cost.

10:43.215 --> 10:47.115
The reason is because maybe something else happened

10:47.115 --> 10:49.875
to this firm that changed its marginal cost

10:50.215 --> 10:52.555
and caused it to add more suppliers.

10:52.815 --> 10:56.195
So like, let's say you innovate in your own production

10:56.195 --> 10:59.075
process and that has a direct effect on your ability

10:59.095 --> 11:01.275
to produce stuff, but it also means maybe you start

11:01.385 --> 11:04.555
outsourcing some activity or insourcing some activity.

11:04.615 --> 11:05.995
And we don't wanna conflate those.

11:06.015 --> 11:08.995
We don't wanna say that that's the pure effect

11:08.995 --> 11:10.395
of getting access to a supplier.

11:10.575 --> 11:12.755
So what we do is we need what are known as instruments,

11:12.755 --> 11:15.315
basically, something that induces

11:15.935 --> 11:19.635
random variation in access to suppliers, unrelated

11:19.635 --> 11:21.195
to other things that could be affecting the

11:21.195 --> 11:22.635
downstream firms marginal cost.

11:23.335 --> 11:25.875
We, we come up with instruments, additions,

11:25.935 --> 11:28.635
and subtractions of suppliers for the downstream firm

11:29.015 --> 11:31.995
by looking at firm birth and firm deaths.

11:33.035 --> 11:35.535
So, what we do is,

11:35.955 --> 11:38.095
for every firm in our manufacturing survey,

11:38.395 --> 11:42.095
we look at its suppliers who in a given year,

11:42.525 --> 11:43.535
exit the marketplace.

11:44.075 --> 11:46.815
So suppliers that die, then

11:46.815 --> 11:51.175
what we do is we focus on the subset of suppliers that die,

11:51.475 --> 11:52.615
who sales to the

11:52.615 --> 11:54.645
Downstream firm are small.

11:55.825 --> 11:59.005
So what that means is say, I'm your supplier

11:59.105 --> 12:00.965
and I suddenly exit the marketplace.

12:01.745 --> 12:04.325
Now, I don't wanna just look at the effect that has on you

12:04.555 --> 12:05.645
because if you

12:05.665 --> 12:07.925
and I have a very deep relationship with one another,

12:08.295 --> 12:12.285
maybe the reason I exit also is affecting you directly.

12:12.545 --> 12:14.365
In fact, maybe something bad happened to you

12:14.425 --> 12:16.045
and that's what's causing me to exit.

12:16.865 --> 12:20.205
So for that reason, we focus on only suppliers whose sales

12:20.385 --> 12:23.925
to the downstream customer is less than 5%

12:23.985 --> 12:25.045
of their total sales.

12:25.425 --> 12:26.925
So that means that when I exit,

12:27.345 --> 12:29.405
you are a very small customer for me.

12:29.585 --> 12:32.005
So can't be that some idiosyncratic thing

12:32.005 --> 12:34.405
that affected you is what's causing me

12:34.405 --> 12:35.525
to exit the marketplace.

12:36.435 --> 12:39.685
Similarly, when we wanna look at supplier additions,

12:39.905 --> 12:43.405
we look at the entry of new firms into the marketplace

12:43.465 --> 12:44.805
who sell to the downstream firm,

12:45.105 --> 12:48.125
but where the downstream firm is a very small fraction of

12:48.325 --> 12:50.125
that upstream firm's total sales.

12:50.585 --> 12:52.285
So I didn't start my firm

12:52.285 --> 12:54.365
because I wanted to service your demand

12:55.085 --> 12:56.805
'cause you're a very small customer for me.

12:56.965 --> 12:57.965
I just started the firm

12:58.025 --> 12:59.605
and then you have the benefit from that.

13:00.225 --> 13:02.805
So those are the, that's the variation we focus on.

13:02.995 --> 13:05.845
When we looked at this stuff, we found that basically

13:05.945 --> 13:10.845
for every one percentage point of suppliers that you add

13:11.105 --> 13:12.685
or lose exogenously.

13:12.985 --> 13:17.085
So that means take your entire cost basket

13:17.315 --> 13:18.965
that you spend on an annual basis

13:19.505 --> 13:22.845
and imagine 1% of those disappear.

13:23.035 --> 13:25.325
Like there's a supplier who sells you 1%

13:25.325 --> 13:26.685
of the stuff you used to buy

13:26.825 --> 13:28.965
and that guy just disappears, exogenously.

13:29.585 --> 13:32.565
If that happens, then that raises the marginal cost

13:32.565 --> 13:35.845
of production for you by about 0.3%.

13:37.445 --> 13:40.145
So for every 1% of suppliers you lose in terms

13:40.145 --> 13:43.065
of cost shares, your marginal cost of production rises

13:43.065 --> 13:44.105
by 0.3%.

13:45.245 --> 13:48.795
Um, on the other side,

13:48.895 --> 13:53.545
if you gain access to a new supplier, your marginal cost

13:53.545 --> 13:56.985
of production go down by 0.3%

13:57.085 --> 13:58.625
for every 1% that you gain.

13:59.005 --> 14:01.745
Now, there's no reason theoretically

14:01.765 --> 14:03.945
or empirically that it had to be symmetrical,

14:04.445 --> 14:07.305
but we just found that it was, it goes up by 0.3

14:07.305 --> 14:10.225
for every 1% and it goes down by 0.3 for every 1%,

14:10.435 --> 14:11.545
which is substantial.

14:11.545 --> 14:15.305
Just to remind you, the perfectly competitive benchmark

14:15.555 --> 14:19.425
would say the answer should be 0% in either direction.

14:20.045 --> 14:21.345
The macro aspect of this,

14:21.485 --> 14:24.305
as I mentioned when I was talking about this Ford,

14:24.305 --> 14:27.905
general Motors example, if it's the case

14:28.055 --> 14:32.385
that individual firms have effects on other individual

14:32.395 --> 14:33.865
firms at the micro level.

14:35.235 --> 14:38.905
So if my firm, operating

14:38.905 --> 14:43.145
or not operating has a direct effect on your marginal cost

14:43.145 --> 14:45.865
of production, then what that means is

14:47.615 --> 14:50.255
supplier churn where firms add

14:50.275 --> 14:52.885
and drop suppliers is gonna have consequences

14:53.105 --> 14:54.125
for the aggregate economy.

14:54.645 --> 14:56.205
'cause every time a new firm appears,

14:56.825 --> 14:59.565
not only does it have potentially direct effects

14:59.585 --> 15:01.485
for the consumer, which we're not measuring,

15:01.745 --> 15:05.325
but it's gonna have indirect effects through existing firms.

15:05.665 --> 15:08.125
By making the existing firms more productive

15:08.185 --> 15:10.405
and lowering their marginal cost of production,

15:10.675 --> 15:13.845
it's gonna lower prices or it's gonna raise profits.

15:14.775 --> 15:17.595
And either way, that's gonna affect GDP

15:17.705 --> 15:21.115
because if it's higher profits, then income goes up,

15:21.305 --> 15:23.355
aggregate income goes up relative to prices,

15:23.535 --> 15:26.595
and if it's lower prices, then real GDP still goes up

15:26.595 --> 15:29.115
because prices are going down for the same level of income.

15:29.575 --> 15:32.355
And so what we wanna do is we wanna take that into account.

15:32.355 --> 15:35.395
We wanna say, okay, every time a firm appears it's gonna

15:35.395 --> 15:37.115
have some effects on the downstream firms,

15:37.205 --> 15:40.435
every time a firm forms a link for with existing firms,

15:40.505 --> 15:42.195
it's gonna have an effect on downstream firms.

15:42.445 --> 15:44.955
We're gonna try to develop an accounting framework

15:44.955 --> 15:47.995
where we take all of those surpluses into account.

15:48.175 --> 15:50.875
And the way we do this is using the VAT data in Belgium.

15:50.875 --> 15:53.795
So as I mentioned, the value added tax in Belgium allows us

15:53.795 --> 15:55.915
to see for basically the universe

15:55.915 --> 15:57.915
of firms in domestic firms in Belgium,

15:58.815 --> 16:01.355
all the bilateral links that are being formed and broken

16:01.495 --> 16:04.395
and how much is being transacted along any one

16:04.395 --> 16:06.315
of those links at annual frequency.

16:06.775 --> 16:10.895
And so what we can do is we can in a sense just sort of

16:11.405 --> 16:14.855
aggregate, add up all the supply chain links

16:14.855 --> 16:17.255
that are being formed and destroyed year by year

16:17.755 --> 16:19.335
and assign a value to each

16:19.335 --> 16:21.815
of those links based on our micro estimates,

16:22.555 --> 16:25.255
and then add them all up and see what we come up with.

16:25.835 --> 16:28.535
And what we find is that given our micro estimates,

16:28.535 --> 16:31.335
it's very plausible that much

16:31.335 --> 16:32.895
of aggregate productivity growth

16:32.995 --> 16:35.255
or at more than half of it can be attributed

16:35.355 --> 16:36.695
to churn in the supply chain.

16:36.915 --> 16:39.895
So if you think about where aggregate growth comes from in,

16:39.955 --> 16:41.175
in economic models,

16:41.515 --> 16:44.415
it comes either from the accumulation of labor or capital

16:44.715 --> 16:46.655
or it comes from improvements in technology.

16:47.555 --> 16:50.375
So labor and capital is something we can measure.

16:50.595 --> 16:54.215
Its accumulation and it explains, something like

16:54.565 --> 16:56.335
between a half, two thirds

16:56.355 --> 16:59.615
of growth over our sample in, in Belgium.

17:00.035 --> 17:02.575
So that leaves about one half to a third.

17:02.795 --> 17:04.455
That's just technology growth.

17:04.685 --> 17:06.495
It's basically like the bit that we don't know

17:06.495 --> 17:09.295
where it comes from because we see capital on labor rise,

17:09.295 --> 17:10.855
but we see output rise by more.

17:11.795 --> 17:15.015
So our framework says, well, one component of

17:15.015 --> 17:17.735
that missing residual can be explained

17:17.915 --> 17:19.775
by churn in the supply chain.

17:20.545 --> 17:22.365
And, and when we do this aggregation,

17:22.385 --> 17:25.925
we find you could plausibly explain more than half of

17:25.925 --> 17:29.125
that missing residual, which is to say,

17:29.355 --> 17:31.165
something on the order of,

17:31.245 --> 17:33.565
I wanna say maybe about half a percentage point

17:33.565 --> 17:37.605
of growth per year, using, using these estimates.

17:38.345 --> 17:41.285
Our work definitely has implications for policy,

17:41.745 --> 17:44.165
but I wouldn't wanna make any kind

17:44.165 --> 17:45.605
of policy pronouncements.

17:46.145 --> 17:48.885
So what, what our work suggest is because these,

17:49.335 --> 17:53.735
These, churn effects are important.

17:53.915 --> 17:57.535
So like individual firms matter to the customers,

17:57.835 --> 18:01.415
to the downstream firms, then what this means is

18:02.645 --> 18:04.775
when a new firm operates

18:04.775 --> 18:06.975
or introduces a new product, it's not

18:07.975 --> 18:10.375
appropriating all the surplus that it's generating.

18:10.915 --> 18:12.815
So when I make a new product

18:13.075 --> 18:16.175
or when I enter the marketplace, obviously I benefit, well,

18:16.175 --> 18:20.695
I get some profits, but, but I can't capture everything.

18:21.245 --> 18:23.415
Some of it is going to these downstream firms.

18:23.515 --> 18:25.575
Now think about the perfectly competitive benchmark.

18:25.575 --> 18:27.575
This is actually a low counterintuitive in the

18:27.575 --> 18:28.775
perfectly competitive benchmark.

18:29.145 --> 18:31.655
Every firm procreates all the surplus

18:31.655 --> 18:33.215
that it generates for the other firms.

18:33.845 --> 18:35.055
Like the tomato guy,

18:35.405 --> 18:38.975
when he opens up his new tomato stall in the farmer's

18:38.975 --> 18:41.495
market, he's not generating any additional surplus

18:41.495 --> 18:43.495
that he is not capturing the value of

18:43.495 --> 18:45.655
what he produces is exactly what he's selling.

18:46.475 --> 18:48.015
In this world, that's not how it works.

18:48.395 --> 18:52.385
You don't actually internalize all the benefits you generate

18:52.385 --> 18:55.505
because you don't appropriate all the surplus.

18:56.085 --> 18:59.665
So what that means though is annoyingly, is that

19:00.325 --> 19:02.145
the firm's decisions

19:02.405 --> 19:06.825
to enter exit the marketplace are potentially inefficient

19:07.495 --> 19:10.145
because they don't appropriate it auto surplus.

19:10.295 --> 19:12.065
They don't internalize everything

19:12.135 --> 19:13.305
when they make that decision.

19:13.885 --> 19:16.505
And so the fact that there's an inefficiency there,

19:16.505 --> 19:19.105
like this idea that it could actually be inefficient

19:19.105 --> 19:20.585
for General Motors to go out

19:20.585 --> 19:22.625
because it causes the suppliers to go out

19:22.625 --> 19:24.265
and then that causes for it to go out

19:24.365 --> 19:25.985
and they don't internalize any

19:25.985 --> 19:27.665
of those things when they make that decision

19:28.175 --> 19:31.505
that potentially opens the door for all sorts of interven.

19:31.535 --> 19:33.545
This stuff is super complicated

19:33.765 --> 19:36.225
and you need to know a lot of data in order to be able

19:36.225 --> 19:37.305
to intervene correctly.

19:37.565 --> 19:42.145
So that's why I think, I wouldn't say that I wanna advocate

19:42.165 --> 19:44.625
or provide any kind of prescriptions for policy.

19:45.245 --> 19:49.145
It just more so says that the world is pretty complicated

19:50.135 --> 19:51.555
and we need to do more research

19:51.615 --> 19:52.955
to try to understand some of these.

19:52.955 --> 19:55.275
But it does mean that it's possible in theory,

19:56.095 --> 19:58.115
for these things to matter big time

19:58.175 --> 20:00.595
and for there to be a need for, for some kind

20:00.595 --> 20:01.515
of policy intervention.

00:00 - Well, the thing that motivated us really is understanding the deeper mechanics of what drives, growth and trade. you think about, oftentimes economists have a tendency to start by thinking about idealized worlds when they want to think about the world.

00:16 - And really, the most, famous benchmark idealized world that we think about as a starting point is a world with perfect comp anonymous markets. and, no individual firm or no individual customer has any kind of market power. So if you think about a world like that, a world with perfect competition, then one thing that you suddenly realize is if there was an individual firm that disappeared or, or an individual firm that appeared. So, like you think about going to the farmer's market,

00:46 - which is sometimes used as an example of something that's close to a compe, perfectly competitive market. If there's a new stall that opens up in a big farmer's market selling more tomatoes, it doesn't really affect anybody. It doesn't affect the market price of tomatoes, and it doesn't affect anybody's ability to get get tomatoes. So in that world, you have perfect competition. As I say, like a necessary consequence of that is that each individual firm doesn't matter.

01:16 - Collectively, the firms are gonna matter, but one by one, they don't matter. Now, this simple observation has important implications for growth, long run economic growth, and for economic trade. So let's start by thinking about the implications for, for growth. If you think of this farmer's market and you think about adding more fruit stalls, like one additional guy adding a new fruit stall, it's gonna

01:40 - increase total output. It's gonna increase the GDP of the farmer's market, but it's not gonna increase GDP per person, which is how we normally think about economic growth. So if you're thinking about a situation in which the standards of living are rising over time, you're not gonna be able to get it in this world where new firms just introduce are introduced into the marketplace, but you have perfect competition.

02:03 - Okay? What are the implications for trade? Well, if you think about trade in such a world, you would expect trade to happen between agents or countries that are very different to one another. So the guys who are making tomatoes are not gonna be trading with other guys who make tomatoes because they make the same thing. You would expect the guy who makes bananas to trade with the guys who make tomatoes and vice versa. But both of these implications don't seem

02:29 - to be borne out in the data. So like, if you think about the growth implications, you think about, for example, apple entering the phone market that had an effect because Apple was not just another company making the exact same type of phone as everybody else. They were making something that was different. The fact that it was different and it was better made everybody better off, and it made consumers standard of living increase.

02:53 - Similarly, when you look at trade, you don't actually have a situation where mostly countries that are very different to each other are trading with one another. So the picture of global trade isn't just Saudi Arabia making oil and selling it to Germany who makes cars and vice versa. A lot of trade is between countries that produce that look pretty similar and produce similar goods. So South Korea is making Samsung phones,

03:17 - the US is making iPhones, and they're trading them back and forth with one another. So the reason economists think these sorts of things can happen is because in practice products are not actually the same. And so new products actually generate some value for the customer, unlike the perfectly competitive benchmark. So, eh, so fundamentally this observation that firms, individual firms matter,

03:46 - is gonna have important implications for trade and growth. Because when you write models of trade and growth, you need a way to quantify how different are the new, new products that are being introduced and the new firms that enter the market to the preexisting ones. Because as you can imagine, that's gonna have an important role to play for whether or not equilibrium and such models are efficient, and how much growth can you get from an increase in market

04:09 - size and so on and so forth. So that's kind of what we're interested in. We're interested in, in a, in a concrete way. I think in the prior literature, both the theoretical literature on trade and growth emphasizes this notion of consumer surplus from new products or new. This is something that's very important. So however, it's very challenging to measure this thing to measure exactly how much value an additional product

04:33 - has for consumers. The reason is because when I introduce a new iPhone, for consumers, there are not, like, maybe it makes you feel better, it makes you happier as a consumer to have access to an iPhone, but there's not some piece of economic data that I can just go and look at. Perhaps I could conduct some experiments and say, well, suppose I raised the, would you still be willing to pay for it?

04:59 - And then under some kind of idealized, experimental setting, I could derive these values. But in general, the way economists approach the problem of figuring out the value of a new product has been, basically to use functional form assumptions. So you estimate demand systems locally, and then those demand systems, you just kind of extrapolate them out. And then the way you work out consumer surplus is you just fill in the area under the demand curve.

05:25 - What we do in this paper is a little bit different. What we do is we, we make the observation that these forces are gonna also matter for firms, not just for consumers. So if you are running a business and you can have access to more varieties of suppliers, then that is also gonna be beneficial to you, just like it would be to a consumer. However, for firms, there's a difference. and the difference that's important for us is

05:51 - that firms have things called outputs. So I can measure how productive A firm is by comparing the quantity of output it can produce relative to the quantity of inputs that it's taking in. I can look at its productivity. That's not something I can do for your happiness as a consumer. So that's kind of like our key into our, our, entry point into this literature is to say, well,

06:12 - one way you could estimate the value of new firms, firms is not to look at the value for consumers, but to look at the value for other firms. And there we have good measures, or at least better measures of value, which is productivity or the cost of production for the downstream firms that are being, that are getting access to new products of, new products and new firms, new firms. and these things are important, you know, like, so for example, there's a specific,

06:40 - example in this literature that people like to talk about, which is that, in, in the fall of 2008 when the great financial crisis was unfolding, the president of Ford Motor Company went to Washington, DC to, to lobby for government bailouts. Now, the thing that's interesting is that Ford was not lobbying for itself. It was lobbying on behalf of General Motors and Chrysler. They didn't need a bailout, but they wanted their biggest competitors to be bailed out.

07:08 - Now, why was it that Ford wanted these competitors to be bailed out? Well, the reason was because they were worried about the survival of their supply chain. So they understood that if General Motors and Chrysler exit the market, then many of their joint suppliers that they share together are gonna be knocked out. And now once those suppliers are knocked out, then Ford is gonna get into trouble.

07:30 - Again, this speaks to the nature departures from the perfectly competitive benchmark, because in the perfectly competitive world, Ford wouldn't care if some individual suppliers are knocked out. It would just go to the marketplace and there's other people selling the same thing, and specific individuals wouldn't have mattered. So, so this is kind of like where we're gonna look. We're gonna look at the effects of these things.

07:52 - For firms, micro and macro are kind of like, there's kind of a logical sequence to it, which is that you start with a micro one and you try to understand at the micro level for an individual firms, what are the consequences of getting access to new suppliers or losing access to suppliers. So you estimate that and you, you come up with a number, and then the macro part is to say, okay, well in the, in the data we see firms forming

08:18 - and breaking links with one another. Can we use our estimates from the micro part of the paper and observations on how many links are being formed and broken every year to try to quantify how much this matters for aggregate productivity or aggregate growth? The dataset, we, for this paper we worked with, the National Bank of Belgium. Belgium is a very, interesting test case because they have fantastic data.

08:44 - So, for the Belgian economy, we have manufacturing surveys where they track at the individual product level how many Units of the different products firms are shipping. And then from the tax data, we can see all the firms, the domestic firms that these Belgium firms are interacting with, who they buy from, who they sell to. and then we can also see things like, their, balance sheets and their wage bill and so on.

09:18 - And so we kind of have all the data that we would need in order to try to figure out what is the value of, suppliers to manufacturing firms in Belgium. And then that's the thing that we extrapolate to the rest of the economy. So the thing that we start with is, the two things that we need in order to answer our question is, first of all, we need to see how good is each Belgian manufacturing firm at

09:46 - producing outputs. And so we can do that because we can see what they buy and we can see what they produce in terms of physical units. We can see what they produce, not just the the Euro value, but the actual units that they share. Okay? The other thing that we need is we need to know who they're interacting with. And that comes from the value added tax data that Belgium has that allows us to track at annual frequency, all firm

10:10 - to firm transactions in Belgium. So we see exactly who your suppliers are. If you're a manufacturing firm, the final thing that we need is what's known as a natural experiment. We need something that exogenously changes access to suppliers for these firms that is not related to some other thing that's happening to that firm. So you couldn't, for example, just go and look at correlations between firms adding suppliers and their productivity.

10:40 - Or in our case, the notion we use is marginal cost. The reason is because maybe something else happened to this firm that changed its marginal cost and caused it to add more suppliers. So like, let's say you innovate in your own production process and that has a direct effect on your ability to produce stuff, but it also means maybe you start outsourcing some activity or insourcing some activity. And we don't wanna conflate those. We don't wanna say that that's the pure effect

11:08 - of getting access to a supplier. So what we do is we need what are known as instruments, basically, something that induces random variation in access to suppliers, unrelated to other things that could be affecting the downstream firms marginal cost. We, we come up with instruments, additions, and subtractions of suppliers for the downstream firm by looking at firm birth and firm deaths. So, what we do is,

11:35 - for every firm in our manufacturing survey, we look at its suppliers who in a given year, exit the marketplace. So suppliers that die, then what we do is we focus on the subset of suppliers that die, who sales to the Downstream firm are small. So what that means is say, I'm your supplier and I suddenly exit the marketplace. Now, I don't wanna just look at the effect that has on you

12:04 - because if you and I have a very deep relationship with one another, maybe the reason I exit also is affecting you directly. In fact, maybe something bad happened to you and that's what's causing me to exit. So for that reason, we focus on only suppliers whose sales to the downstream customer is less than 5% of their total sales. So that means that when I exit, you are a very small customer for me.

12:29 - So can't be that some idiosyncratic thing that affected you is what's causing me to exit the marketplace. Similarly, when we wanna look at supplier additions, we look at the entry of new firms into the marketplace who sell to the downstream firm, but where the downstream firm is a very small fraction of that upstream firm's total sales. So I didn't start my firm because I wanted to service your demand

12:55 - 'cause you're a very small customer for me. I just started the firm and then you have the benefit from that. So those are the, that's the variation we focus on. When we looked at this stuff, we found that basically for every one percentage point of suppliers that you add or lose exogenously. So that means take your entire cost basket that you spend on an annual basis and imagine 1% of those disappear.

13:23 - Like there's a supplier who sells you 1% of the stuff you used to buy and that guy just disappears, exogenously. If that happens, then that raises the marginal cost of production for you by about 0.3%. So for every 1% of suppliers you lose in terms of cost shares, your marginal cost of production rises by 0.3%. on the other side, if you gain access to a new supplier, your marginal cost

13:53 - of production go down by 0.3% for every 1% that you gain. Now, there's no reason theoretically or empirically that it had to be symmetrical, but we just found that it was, it goes up by 0.3 for every 1% and it goes down by 0.3 for every 1%, which is substantial. Just to remind you, the perfectly competitive benchmark would say the answer should be 0% in either direction. The macro aspect of this,

14:21 - as I mentioned when I was talking about this Ford, general Motors example, if it's the case that individual firms have effects on other individual firms at the micro level. So if my firm, operating or not operating has a direct effect on your marginal cost of production, then what that means is supplier churn where firms add and drop suppliers is gonna have consequences for the aggregate economy.

14:54 - 'cause every time a new firm appears, not only does it have potentially direct effects for the consumer, which we're not measuring, but it's gonna have indirect effects through existing firms. By making the existing firms more productive and lowering their marginal cost of production, it's gonna lower prices or it's gonna raise profits. And either way, that's gonna affect GDP because if it's higher profits, then income goes up, aggregate income goes up relative to prices,

15:23 - and if it's lower prices, then real GDP still goes up because prices are going down for the same level of income. And so what we wanna do is we wanna take that into account. We wanna say, okay, every time a firm appears it's gonna have some effects on the downstream firms, every time a firm forms a link for with existing firms, it's gonna have an effect on downstream firms. We're gonna try to develop an accounting framework where we take all of those surpluses into account. And the way we do this is using the VAT data in Belgium.

15:50 - So as I mentioned, the value added tax in Belgium allows us to see for basically the universe of firms in domestic firms in Belgium, all the bilateral links that are being formed and broken and how much is being transacted along any one of those links at annual frequency. And so what we can do is we can in a sense just sort of aggregate, add up all the supply chain links that are being formed and destroyed year by year and assign a value to each

16:19 - of those links based on our micro estimates, and then add them all up and see what we come up with. And what we find is that given our micro estimates, it's very plausible that much of aggregate productivity growth or at more than half of it can be attributed to churn in the supply chain. So if you think about where aggregate growth comes from in, in economic models, it comes either from the accumulation of labor or capital

16:44 - or it comes from improvements in technology. So labor and capital is something we can measure. Its accumulation and it explains, something like between a half, two thirds of growth over our sample in, in Belgium. So that leaves about one half to a third. That's just technology growth. It's basically like the bit that we don't know where it comes from because we see capital on labor rise, but we see output rise by more.

17:11 - So our framework says, well, one component of that missing residual can be explained by churn in the supply chain. And, and when we do this aggregation, we find you could plausibly explain more than half of that missing residual, which is to say, something on the order of, I wanna say maybe about half a percentage point of growth per year, using, using these estimates. Our work definitely has implications for policy,

17:41 - but I wouldn't wanna make any kind of policy pronouncements. So what, what our work suggest is because these, These, churn effects are important. So like individual firms matter to the customers, to the downstream firms, then what this means is when a new firm operates or introduces a new product, it's not appropriating all the surplus that it's generating. So when I make a new product

18:13 - or when I enter the marketplace, obviously I benefit, well, I get some profits, but, but I can't capture everything. Some of it is going to these downstream firms. Now think about the perfectly competitive benchmark. This is actually a low counterintuitive in the perfectly competitive benchmark. Every firm procreates all the surplus that it generates for the other firms. Like the tomato guy, when he opens up his new tomato stall in the farmer's

18:38 - market, he's not generating any additional surplus that he is not capturing the value of what he produces is exactly what he's selling. In this world, that's not how it works. You don't actually internalize all the benefits you generate because you don't appropriate all the surplus. So what that means though is annoyingly, is that the firm's decisions to enter exit the marketplace are potentially inefficient because they don't appropriate it auto surplus.

19:10 - They don't internalize everything when they make that decision. And so the fact that there's an inefficiency there, like this idea that it could actually be inefficient for General Motors to go out because it causes the suppliers to go out and then that causes for it to go out and they don't internalize any of those things when they make that decision that potentially opens the door for all sorts of interven.

19:31 - This stuff is super complicated and you need to know a lot of data in order to be able to intervene correctly. So that's why I think, I wouldn't say that I wanna advocate or provide any kind of prescriptions for policy. It just more so says that the world is pretty complicated and we need to do more research to try to understand some of these. But it does mean that it's possible in theory, for these things to matter big time

19:58 - and for there to be a need for, for some kind of policy intervention.

Faculti AI can make mistakes