Popular in Social Sciences

Political Science

V-Dem’s Conceptions of Democracy and Their Consequences

University of Notre Dame Michael Coppedge

Central Monetary Authorities

Macroeconomic modelling of CBDC: a critical review

European Central Bank Ulrich Bindseil

Social Identities and Divisions

A new explanation of antisemitism: Jew hatred as a civilisational phenomenon

Boston University Liah Greenfeld

Central Monetary Authorities

Financing Private Credit

Federal Reserve Bank of New York Nina Boyarchenko

Economics

Failing Banks

Massachusetts Institute of Technology Emil Verner

Culture and Society

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

New York University Harvey Moltoch

American Politics

Innovation and Strategic Network Formation

View more insights in Social Sciences - Economics - Microeconomics

View more figures from Boston University

View the page of Krishna Dasaratha

Read the Study

Next in Economics

Do renewable energy investments create local jobs?

Natalia Fabra | University Carlos III Madrid

An early indicator for anomalous stock market performance

Thomas Gries | Universität Paderborn

Firms face a choice between secrecy, which protects existing intellectual property, and openness, which facilitates learning from others. Their decisions determine interaction rates between firms. Krishna Dasaratha discusses the interaction rates in a learning network.

Image courtesy of interviewee. June 28, 2024

Log-in or Sign-up to Faculti
Currently viewing this subject insight as a guest. You have insight(s) remaining for this month. Login to view 8000+ figures on the platform.

Generating Workspace...

Export as PDF

This video discusses the importance of social networks in the spread of innovative ideas and information, particularly in technological settings like Silicon Valley compared to Route 128 in Boston. It highlights research by Anna Lee Sian, which shows that Silicon Valley's success is due to its highly connected network facilitating idea sharing, whereas Route 128's companies tended to be siloed and less collaborative. The author’s research seeks to understand the factors that determine whether social networks are dense and innovative or sparse and secretive.

Utilizing tools from game and network theory, the author models how firms decide on interacting and investing in research and development, with the incentive to balance learning from others while protecting their own ideas. The findings suggest that certain social network structures experience "phase transitions," meaning small changes in behavior can shift a network from being sparse to highly connected, leading to significantly different innovation outcomes.

The research argues for the need for more interaction among firms to promote innovation, suggesting that government interventions, such as public institutions or labs that do not prioritize secrecy, can enhance connectivity and idea flow. The author also reflects on the role of intellectual property rights in enabling or hindering collaboration, noting that while patents can protect ideas, they may also create barriers to idea sharing.

In summary, the author emphasizes that increasing firm interaction may yield substantial benefits for innovation, advocating for policy measures that encourage collaboration rather than merely subsidizing research and development. Future research directions proposed include exploring how diverse groups contribute to innovation through networking.

Language:

Export as PDF

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

00:00.285 --> 00:04.065
An important part of innovation in many settings can be

00:05.075 --> 00:07.025
ideas and information spreading

00:07.025 --> 00:09.585
through social networks of various forms.

00:10.005 --> 00:11.865
And then these ideas

00:12.285 --> 00:14.905
and this information getting incorporated into new

00:14.905 --> 00:16.465
technologies or inventions.

00:17.565 --> 00:20.025
And so this is an idea that there's a lot

00:20.025 --> 00:22.065
of empirical evidence from in different

00:22.065 --> 00:23.225
fields in social science.

00:24.285 --> 00:25.465
And a real starting point

00:25.485 --> 00:29.025
for my work is qualitative research by Anna Lee Sian,

00:29.025 --> 00:31.425
where she compares Silicon Valley

00:31.765 --> 00:35.095
and the Boston area Route 1 28,

00:35.095 --> 00:37.215
technology industry in the 1980s.

00:37.315 --> 00:41.095
And what she finds is Silicon Valley looks a lot more like

00:41.095 --> 00:43.575
what you alluded to, a highly connected cluster with a lot

00:43.575 --> 00:47.165
of sharing of ideas that leads to,

00:47.225 --> 00:48.645
all sorts of new breakthroughs.

00:48.955 --> 00:50.645
Silicon Valley, sorry,

00:50.645 --> 00:53.085
route 1 28 had successful companies initially,

00:53.385 --> 00:56.195
but they were much more siloed in

00:57.565 --> 01:00.075
their individual firms, had much less interaction

01:00.075 --> 01:01.715
between them in the long run.

01:02.015 --> 01:05.395
That's an explanation for why the computer industry

01:05.395 --> 01:06.915
and the Boston area didn't last

01:07.015 --> 01:08.995
and Silicon Valley became what it is today.

01:09.695 --> 01:11.755
And so what my research is doing is it's trying

01:11.755 --> 01:14.955
to think about from a theoretical perspective,

01:16.165 --> 01:18.775
what might determine when we get these highly connected

01:18.775 --> 01:21.975
clusters, when do we get sort of dense social networks

01:21.975 --> 01:23.855
with a lot of interaction and spread of ideas

01:24.435 --> 01:25.855
and when do we get sparse

01:25.855 --> 01:27.895
or social networks that don't drive that as much?

01:28.635 --> 01:31.575
And what I do is I try and use tools from game theory

01:31.575 --> 01:34.055
and from network theory to understand what the types

01:34.055 --> 01:37.585
of social networks we might get are going to look like

01:37.725 --> 01:40.945
and which of these worlds we end up in.

01:41.085 --> 01:43.745
And so the basic point there is that firms

01:43.885 --> 01:45.985
and inventors have reasons to be secretive.

01:45.985 --> 01:50.555
They can also benefit from interacting with each other

01:50.575 --> 01:51.595
and learning from each other.

01:51.875 --> 01:56.345
I try and understand how do those incentives drive the types

01:56.345 --> 01:59.545
of networks we end in my work, as part of the literature

01:59.545 --> 02:02.385
that's developed in the last 30

02:02.485 --> 02:07.345
or so years in microeconomic theory, trying to look at

02:07.605 --> 02:12.045
how do networks matter for all sorts of economic,

02:12.675 --> 02:14.685
political, social science situations.

02:15.905 --> 02:20.365
And essentially the first question this literature asked is

02:20.505 --> 02:22.445
how can we model how networks form

02:22.545 --> 02:25.565
and especially how can we model how networks form trying

02:25.585 --> 02:26.765
to think about the incentives

02:27.145 --> 02:29.805
of the different agents in the network to actually interact

02:29.805 --> 02:31.045
with form lengths with each other?

02:32.285 --> 02:35.665
And so the initial approach, which was really the start

02:35.665 --> 02:38.955
of this network theory literature, was

02:38.955 --> 02:40.875
to think about settings where people have a lot

02:40.875 --> 02:42.915
of control over their links with each other.

02:43.295 --> 02:47.075
And so the basic model there is each person decide

02:47.095 --> 02:48.275
who do they wanna be connected to?

02:48.295 --> 02:51.075
So who do I wanna be friends with or working with

02:51.135 --> 02:52.435
or signing a treaty with?

02:53.775 --> 02:56.435
And two people or countries

02:56.435 --> 02:58.795
or firms end up connected to each other. They both say

02:58.795 --> 02:59.950
They wanna connect with each other.

03:01.115 --> 03:02.815
And this has been a great model

03:02.915 --> 03:04.495
for describing a lot of situations.

03:04.495 --> 03:07.375
Maybe you wanna describe treaties between different nations.

03:07.375 --> 03:09.975
Maybe you wanna describe formal research collaborations.

03:11.885 --> 03:14.855
What I'm doing is saying this is assuming a lot about the

03:14.855 --> 03:17.655
control people have over their connections

03:17.715 --> 03:19.975
and their knowledge of who's connecting with who.

03:20.855 --> 03:23.795
And often in a lot of social situations, if we're sort

03:23.795 --> 03:26.475
of talking to people informally forming friendships,

03:27.185 --> 03:28.675
there's a lot more uncertainty.

03:28.705 --> 03:30.315
It's a lot harder to control exactly

03:30.375 --> 03:31.755
who you end up talking with.

03:32.455 --> 03:35.155
And so what I'm using is random networks tools

03:35.175 --> 03:38.275
for probability theory to say what types

03:38.275 --> 03:39.795
of networks do we end up with?

03:39.895 --> 03:42.275
If people decide essentially how much to interact

03:42.275 --> 03:44.235
with each other in formal terms,

03:44.235 --> 03:46.795
they're deciding the probabilities of connections

03:46.795 --> 03:47.955
with different types of people,

03:48.615 --> 03:51.355
but they don't have exact control over

03:51.975 --> 03:53.435
who exactly they connect with.

03:54.375 --> 03:57.115
Um, and it turns out in a lot of ways that's a model

03:57.115 --> 04:01.075
that looks more like other ones we u use in other parts

04:01.075 --> 04:04.435
of economics where we can use sort of calculus type tools

04:04.575 --> 04:07.935
to solve for what networks we end up actually getting.

04:08.265 --> 04:12.495
Basic idea of the model is you have a large number

04:12.495 --> 04:15.095
of different firms, you can think of these as small startups

04:15.095 --> 04:19.535
or inventors, and they're all deciding how much

04:19.535 --> 04:21.455
to invest in research and development

04:21.475 --> 04:23.655
and almost also how much to interact with each other.

04:23.715 --> 04:25.775
And that's really the choice that I care most about.

04:26.915 --> 04:30.975
Um, and so each of these firms traces a trade off

04:31.955 --> 04:34.025
where they wanna make new technologies

04:36.155 --> 04:37.295
to make those technologies.

04:37.365 --> 04:39.015
It's helpful to learn ideas from others,

04:39.195 --> 04:41.735
but they don't want others to learn from them

04:41.735 --> 04:43.375
because they can potentially compete.

04:44.515 --> 04:47.415
And the way the new technologies work in my model is

04:47.415 --> 04:50.735
essentially there's building blocks which are basic ideas

04:50.825 --> 04:54.065
which can be combined into technologies.

04:54.165 --> 04:56.705
And so this is borrowing from a great paper

04:56.925 --> 05:01.065
by Marty Weitzman, which, called Recombinant Growth,

05:01.075 --> 05:02.345
which is saying one

05:02.345 --> 05:05.265
of the ways we should be thinking about innovation in the

05:05.265 --> 05:08.825
economy is by combining different ideas to form new

05:08.845 --> 05:13.105
and more complicated, new more complicated technologies.

05:14.725 --> 05:16.945
And so what I'd then do in this, that's the basic structure

05:16.945 --> 05:18.105
of the model, what I try

05:18.105 --> 05:20.745
and ask is what are the networks that we end up with

05:21.965 --> 05:23.785
in this model end up looking like

05:24.875 --> 05:26.495
if we're at what's known as equilibrium?

05:26.495 --> 05:31.455
So if each firm or each inventor is choosing how much

05:31.455 --> 05:34.455
to interact and how much to invest optimally given

05:34.455 --> 05:35.455
what other people are.

05:37.145 --> 05:41.445
And an important part of my results is that it turns out

05:41.445 --> 05:42.765
that these types

05:43.065 --> 05:45.965
of network structures exhibit what's sometimes known

05:45.965 --> 05:46.965
as phase transitions.

05:47.785 --> 05:48.965
And what this means is

05:48.965 --> 05:52.165
that there's a point at which you transition from having

05:52.165 --> 05:53.245
these sparse networks

05:53.245 --> 05:54.605
to having these highly connected

05:54.605 --> 05:56.045
clusters with a lot of innovation.

05:57.145 --> 05:58.565
And it's called a phase transition

05:58.565 --> 06:01.005
because actually this transition can happen quite sharply.

06:01.025 --> 06:04.245
So a pretty small change in behavior can have quite big

06:04.915 --> 06:07.165
impacts for the amount of innovation.

06:07.185 --> 06:08.925
And so this is by analogy with physics.

06:10.185 --> 06:12.085
Um, for example, when we transition,

06:12.085 --> 06:14.925
when water transition from a liquid to a gas

06:15.185 --> 06:18.525
or a liquid to a solid, this happens sharply.

06:18.525 --> 06:20.005
You change the temperature a little bit.

06:20.065 --> 06:22.925
And we go from being in one state to another

06:23.465 --> 06:25.685
in the same way these networks can have very different

06:25.885 --> 06:28.085
structures when people change their behavior a little.

06:28.865 --> 06:30.485
And so the main result ends up being that

06:30.485 --> 06:34.615
what you can often get is the equilibrium is

06:34.755 --> 06:35.895
at this phase transition.

06:35.915 --> 06:39.015
And so the intuition for this is that these different types

06:39.015 --> 06:41.415
of networks that have different structures above

06:41.415 --> 06:45.985
and below this critical threshold,

06:46.375 --> 06:48.225
turn out to lead to quite different incentives.

06:48.285 --> 06:50.265
So essentially if you have a highly connected cluster,

06:50.265 --> 06:52.465
this is gonna make firms wanna be more secretive.

06:53.485 --> 06:54.545
If you don't have one

06:54.545 --> 06:56.785
that firms on the other hand gonna wanna try

06:56.785 --> 06:59.425
and interact more to sort of find ideas from different parts

06:59.425 --> 07:00.665
of the network to combine them

07:01.205 --> 07:02.865
and make productive new technologies.

07:04.005 --> 07:08.305
And so I think the, to to wrap up what the,

07:09.335 --> 07:11.225
what this space transition result means.

07:11.625 --> 07:13.065
I think the way to think about it's

07:13.065 --> 07:16.505
that you're almost getting these highly connected clusters

07:16.505 --> 07:17.905
that can lead to a lot of innovation.

07:18.725 --> 07:22.705
But in the absence of some sort of sort of intervention

07:23.125 --> 07:25.825
by government or by other types of institutions,

07:26.445 --> 07:29.225
the market itself may have too many incentives privately

07:29.225 --> 07:31.225
to be secretive to quite deliverance.

07:31.925 --> 07:35.105
The high level takeaway from the result I just told you is

07:35.105 --> 07:38.225
that it could be that getting a little bit more interaction,

07:38.225 --> 07:40.265
getting firms to be a bit less secretive

07:40.265 --> 07:42.865
and more collaborative could have very larger rewards.

07:42.885 --> 07:45.865
And so I have several results sort of quantifying that,

07:45.865 --> 07:47.665
saying the benefits can be quite big.

07:48.485 --> 07:51.905
And then a question that comes up is what types of,

07:52.495 --> 07:55.065
what types of policies might actually deliver this?

07:55.845 --> 07:57.905
And so the first thing I ask about is a

07:57.905 --> 07:58.945
policy that we see a lot.

07:58.945 --> 08:01.145
What if you just subsidize research and development?

08:02.135 --> 08:06.035
So pay pay firms to do more research in various ways.

08:06.695 --> 08:11.035
And what I find is this has some benefits, it does sort

08:11.035 --> 08:14.755
of improve welfare, but in my model it's not fixing

08:14.815 --> 08:15.835
the first order problem.

08:15.935 --> 08:19.075
And so what I find is you really wanna fix these sort

08:19.075 --> 08:20.235
of incentives to interact

08:20.235 --> 08:21.955
and get these highly connected clusters,

08:22.585 --> 08:26.155
getting the right amount of individual firm RD

08:26.315 --> 08:28.595
so subsidizing that can have very big benefits.

08:29.335 --> 08:31.315
So then I ask, how can you actually change

08:32.005 --> 08:33.795
these network formation patterns?

08:34.655 --> 08:38.915
And one thing I argue can help with this is,

08:39.095 --> 08:40.395
and this is in line with sex

08:40.395 --> 08:43.945
and Ann's explanations for the differences between Boston

08:44.085 --> 08:47.945
and Silicon Valley, is public institutions that aren't

08:47.945 --> 08:49.065
as concerned as secrecy.

08:49.085 --> 08:52.025
And so this could be government labs or universities.

08:52.845 --> 08:56.165
If you have say Stanford University interacting

08:56.165 --> 08:59.685
with many researchers, then they can potentially lead

08:59.685 --> 09:01.805
to the emergence of these highly connected clusters.

09:02.425 --> 09:04.245
And it's not so much because of the research

09:04.245 --> 09:05.405
that they're doing, but it's

09:05.405 --> 09:07.365
because they can serve as what I call informational

09:07.365 --> 09:10.805
intermediaries, learning ideas potentially from one private

09:10.805 --> 09:15.045
firm and fasting them along to another one that can be sort

09:15.045 --> 09:16.645
of the central component of one

09:16.645 --> 09:18.725
of these highly connected clusters.

09:19.435 --> 09:21.325
There's a huge amount of work

09:22.215 --> 09:26.395
in economics thinking about incentives towards innovation

09:26.815 --> 09:28.755
and secrecy

09:28.755 --> 09:30.355
and competition within sort

09:30.355 --> 09:32.835
of more traditional property rights frameworks.

09:33.375 --> 09:36.435
And what I told you so far is trying

09:36.435 --> 09:38.435
to think less about formal property rights,

09:38.435 --> 09:40.315
formal collaborations and more about

09:41.305 --> 09:43.275
what I think is something

09:43.275 --> 09:45.835
that I think has been less emphasized in a lot of,

09:45.835 --> 09:47.355
at least the economic theory literature,

09:47.355 --> 09:50.195
which is more informal types of spread of information

09:50.215 --> 09:51.715
and collaboration and so on.

09:52.215 --> 09:54.755
And so, so far what I've told you about, I'm assuming

09:54.755 --> 09:57.635
that firms don't have any property rights,

09:58.295 --> 10:01.275
and that ideas are only spreading informally.

10:01.375 --> 10:03.635
Of course that's not how the real world works.

10:03.635 --> 10:06.515
There's a mix of more informal spread of ideas

10:06.615 --> 10:08.835
and formal property rights and collaborations.

10:09.535 --> 10:10.995
And so I have a section on the paper

10:11.025 --> 10:13.355
that I would say is just trying to explore some

10:13.355 --> 10:14.595
of the forces that show up

10:15.645 --> 10:17.175
when you include both of these in.

10:17.195 --> 10:18.495
So what my patents do,

10:19.555 --> 10:21.535
and I think the point there is just

10:21.535 --> 10:23.535
that patents can be helpful, they can solve some

10:23.535 --> 10:26.495
of these problems, but they also have outside.

10:26.515 --> 10:29.295
So one advantage to patents in my model is if firms can

10:29.295 --> 10:31.215
patent their ideas, then they may not need to worry

10:31.235 --> 10:33.725
as much about secrecy.

10:34.105 --> 10:36.085
And so they may be more willing to interact.

10:37.345 --> 10:40.365
On the other hand, if firms patent their ideas,

10:40.425 --> 10:42.725
it can make it more difficult for other firms to use them.

10:42.725 --> 10:44.845
Of course you can have licensing agreements and so on,

10:44.865 --> 10:48.645
but it'll baseline, it may be that firms interact more,

10:50.145 --> 10:52.085
but the ideas they're learning are ideas

10:52.155 --> 10:53.565
that they can't incorporate

10:53.585 --> 10:55.125
or are concerned about incorporating

10:55.125 --> 10:57.525
because they may get sued for infringement

10:57.545 --> 10:59.685
or may have to pay a licensing fee and so on.

10:59.685 --> 11:03.445
And so I think the basic point here is baseline world is

11:04.425 --> 11:06.125
one without formal property rights,

11:06.615 --> 11:08.925
there are ways in which formal property rights may be able

11:08.925 --> 11:12.165
to help in some cases, but they're likely not going to fix

11:13.315 --> 11:15.265
these problems completely in many cases.

11:16.005 --> 11:18.145
Um, and for example, often it could be

11:18.145 --> 11:20.185
that having some patents can be helpful,

11:20.405 --> 11:22.745
but you wanna sort of restraint patents

11:22.745 --> 11:24.265
to particular subset of items.

11:24.905 --> 11:29.265
I think high level takeaway is motivated by a lot

11:29.265 --> 11:33.355
of great empirical work on how the spread of ideas

11:33.895 --> 11:35.435
matter for innovation.

11:35.735 --> 11:37.995
I'm trying to plug that into a more

11:37.995 --> 11:39.155
formal theoretical model.

11:39.295 --> 11:41.035
And what that finds is

11:41.035 --> 11:44.035
that there can be potentially quite big benefits

11:44.055 --> 11:45.435
to getting more interaction.

11:46.375 --> 11:48.835
Um, and that can often be a first order policy

11:49.385 --> 11:50.915
concern in thinking about how

11:50.915 --> 11:52.995
to encourage innovation, sometimes

11:53.150 --> 11:55.600
Potentially more important than things like targeting

11:56.305 --> 11:58.525
how much research is an individual company doing

11:59.305 --> 12:01.725
and then tried to point out that

12:03.025 --> 12:06.615
firms individual incentives to be secretive can mean

12:06.615 --> 12:09.015
that markets could be very inefficient

12:09.015 --> 12:10.055
at providing innovation.

12:10.885 --> 12:13.375
There's a number of ways one could try to fix that.

12:13.555 --> 12:17.975
One that I've tried to particularly advocate for is,

12:18.735 --> 12:21.295
building up more public institutions

12:21.295 --> 12:23.175
that can play an important role in innovation,

12:23.435 --> 12:25.695
but other types of approaches like property right,

12:25.695 --> 12:29.655
could work in some settings, future directions.

12:29.775 --> 12:32.895
I think something I've been particularly interested in is

12:34.005 --> 12:37.335
essentially the model I told you about is one where

12:38.945 --> 12:42.815
firms are treated as apia asym, symmetric.

12:43.875 --> 12:46.195
I have a little bit of some extensions in the paper,

12:46.675 --> 12:48.915
relaxing that a bit, but I think there's a lot of richness

12:50.855 --> 12:54.995
in how social networks connect people from different groups.

12:55.095 --> 12:57.875
So for example, there's a lot of recent research on how

12:58.915 --> 13:02.065
immigrants can play an important role in innovation

13:02.085 --> 13:03.185
and I think a lot of that's

13:03.185 --> 13:05.265
because of sort of spread of ideas, combination of ideas.

13:05.605 --> 13:08.705
So something I've been excited about thinking about is how

13:09.905 --> 13:14.295
about how do different types of, people, inventors,

13:14.655 --> 13:15.935
companies interact with each other?

13:16.195 --> 13:19.655
How might those differences matter for this process

13:19.795 --> 13:22.175
of the spread of in combination of ideas.

"The Dynamics of Innovation: Exploring the Impact of Social Networks on Technological Advancement"

00:00:00 - An important part of innovation in many settings can be ideas and information spreading through social networks of various forms. And then these ideas and this information getting incorporated into new technologies or inventions. And so this is an idea that there's a lot of empirical evidence from in different fields in social science. And a real starting point for my work is qualitative research by Anna Lee Sian, where she compares Silicon Valley and the Boston area Route 1 28, technology industry in the 1980s. And what she finds is Silicon Valley looks a lot more like
00:00:43 - of sharing of ideas that leads to, all sorts of new breakthroughs. Silicon Valley, sorry, route 1 28 had successful companies initially, but they were much more siloed in their individual firms, had much less interaction between them in the long run. That's an explanation for why the computer industry and the Boston area didn't last and Silicon Valley became what it is today. And so what my research is doing is it's trying to think about from a theoretical perspective, what might determine when we get these highly connected clusters, when do we get sort of dense social networks
00:01:24 - and when do we get sparse or social networks that don't drive that as much? And what I do is I try and use tools from game theory and from network theory to understand what the types of social networks we might get are going to look like and which of these worlds we end up in. And so the basic point there is that firms and inventors have reasons to be secretive. They can also benefit from interacting with each other and learning from each other. I try and understand how do those incentives drive the types of networks we end in my work, as part of the literature that's developed in the last 30
00:02:07 - how do networks matter for all sorts of economic, political, social science situations. And essentially the first question this literature asked is how can we model how networks form and especially how can we model how networks form trying to think about the incentives of the different agents in the network to actually interact with form lengths with each other? And so the initial approach, which was really the start of this network theory literature, was to think about settings where people have a lot of control over their links with each other.
00:02:47 - who do they wanna be connected to? So who do I wanna be friends with or working with or signing a treaty with? And two people or countries or firms end up connected to each other. They both say They wanna connect with each other. And this has been a great model for describing a lot of situations. Maybe you wanna describe treaties between different nations. Maybe you wanna describe formal research collaborations. What I'm doing is saying this is assuming a lot about the control people have over their connections and their knowledge of who's connecting with who. And often in a lot of social situations, if we're sort
00:03:27 - there's a lot more uncertainty. It's a lot harder to control exactly who you end up talking with. And so what I'm using is random networks tools for probability theory to say what types of networks do we end up with? If people decide essentially how much to interact with each other in formal terms, they're deciding the probabilities of connections with different types of people, but they don't have exact control over who exactly they connect with. and it turns out in a lot of ways that's a model that looks more like other ones we u use in other parts of economics where we can use sort of calculus type tools

"Phase Transitions in Innovation Networks: Balancing Secrecy and Collaboration Among Firms"

00:04:08 - Basic idea of the model is you have a large number of different firms, you can think of these as small startups or inventors, and they're all deciding how much to invest in research and development and almost also how much to interact with each other. And that's really the choice that I care most about. and so each of these firms traces a trade off where they wanna make new technologies to make those technologies. It's helpful to learn ideas from others, but they don't want others to learn from them because they can potentially compete. And the way the new technologies work in my model is
00:04:50 - which can be combined into technologies. And so this is borrowing from a great paper by Marty Weitzman, which, called Recombinant Growth, which is saying one of the ways we should be thinking about innovation in the economy is by combining different ideas to form new and more complicated, new more complicated technologies. And so what I'd then do in this, that's the basic structure of the model, what I try and ask is what are the networks that we end up with in this model end up looking like if we're at what's known as equilibrium? So if each firm or each inventor is choosing how much
00:05:34 - what other people are. And an important part of my results is that it turns out that these types of network structures exhibit what's sometimes known as phase transitions. And what this means is that there's a point at which you transition from having these sparse networks to having these highly connected clusters with a lot of innovation. And it's called a phase transition because actually this transition can happen quite sharply. So a pretty small change in behavior can have quite big impacts for the amount of innovation. And so this is by analogy with physics. for example, when we transition,
00:06:15 - or a liquid to a solid, this happens sharply. You change the temperature a little bit. And we go from being in one state to another in the same way these networks can have very different structures when people change their behavior a little. And so the main result ends up being that what you can often get is the equilibrium is at this phase transition. And so the intuition for this is that these different types of networks that have different structures above and below this critical threshold, turn out to lead to quite different incentives. So essentially if you have a highly connected cluster, this is gonna make firms wanna be more secretive.
00:06:54 - that firms on the other hand gonna wanna try and interact more to sort of find ideas from different parts of the network to combine them and make productive new technologies. And so I think the, to to wrap up what the, what this space transition result means. I think the way to think about it's that you're almost getting these highly connected clusters that can lead to a lot of innovation. But in the absence of some sort of sort of intervention by government or by other types of institutions, the market itself may have too many incentives privately to be secretive to quite deliverance. The high level takeaway from the result I just told you is
00:07:38 - getting firms to be a bit less secretive and more collaborative could have very larger rewards. And so I have several results sort of quantifying that, saying the benefits can be quite big. And then a question that comes up is what types of, what types of policies might actually deliver this? And so the first thing I ask about is a policy that we see a lot. What if you just subsidize research and development? So pay pay firms to do more research in various ways. And what I find is this has some benefits, it does sort of improve welfare, but in my model it's not fixing the first order problem.

"Facilitating Innovation Through Informal Collaboration and Public Institutions"

00:08:19 - of incentives to interact and get these highly connected clusters, getting the right amount of individual firm RD so subsidizing that can have very big benefits. So then I ask, how can you actually change these network formation patterns? And one thing I argue can help with this is, and this is in line with sex and Ann's explanations for the differences between Boston and Silicon Valley, is public institutions that aren't as concerned as secrecy. And so this could be government labs or universities. If you have say Stanford University interacting
00:08:59 - to the emergence of these highly connected clusters. And it's not so much because of the research that they're doing, but it's because they can serve as what I call informational intermediaries, learning ideas potentially from one private firm and fasting them along to another one that can be sort of the central component of one of these highly connected clusters. There's a huge amount of work in economics thinking about incentives towards innovation and secrecy and competition within sort of more traditional property rights frameworks. And what I told you so far is trying
00:09:38 - formal collaborations and more about what I think is something that I think has been less emphasized in a lot of, at least the economic theory literature, which is more informal types of spread of information and collaboration and so on. And so, so far what I've told you about, I'm assuming that firms don't have any property rights, and that ideas are only spreading informally. Of course that's not how the real world works. There's a mix of more informal spread of ideas and formal property rights and collaborations. And so I have a section on the paper that I would say is just trying to explore some of the forces that show up when you include both of these in.
00:10:19 - and I think the point there is just that patents can be helpful, they can solve some of these problems, but they also have outside. So one advantage to patents in my model is if firms can patent their ideas, then they may not need to worry as much about secrecy. And so they may be more willing to interact. On the other hand, if firms patent their ideas, it can make it more difficult for other firms to use them. Of course you can have licensing agreements and so on, but it'll baseline, it may be that firms interact more, but the ideas they're learning are ideas that they can't incorporate or are concerned about incorporating because they may get sued for infringement
00:10:59 - And so I think the basic point here is baseline world is one without formal property rights, there are ways in which formal property rights may be able to help in some cases, but they're likely not going to fix these problems completely in many cases. and for example, often it could be that having some patents can be helpful, but you wanna sort of restraint patents to particular subset of items. I think high level takeaway is motivated by a lot of great empirical work on how the spread of ideas matter for innovation. I'm trying to plug that into a more
00:11:39 - And what that finds is that there can be potentially quite big benefits to getting more interaction. and that can often be a first order policy concern in thinking about how to encourage innovation, sometimes Potentially more important than things like targeting how much research is an individual company doing and then tried to point out that firms individual incentives to be secretive can mean that markets could be very inefficient at providing innovation. There's a number of ways one could try to fix that. One that I've tried to particularly advocate for is,

"Exploring the Role of Social Networks in Innovation and Idea Exchange"

00:12:21 - that can play an important role in innovation, but other types of approaches like property right, could work in some settings, future directions. I think something I've been particularly interested in is essentially the model I told you about is one where firms are treated as apia asym, symmetric. I have a little bit of some extensions in the paper, relaxing that a bit, but I think there's a lot of richness in how social networks connect people from different groups. So for example, there's a lot of recent research on how
00:13:19 - and I think a lot of that's because of sort of spread of ideas, combination of ideas. So something I've been excited about thinking about is how about how do different types of, people, inventors, companies interact with each other? How might those differences matter for this process of the spread of in combination of ideas.

Faculti AI can make mistakes

Research

Policy

Analysis