Popular in Law

Human Rights Law

Human Rights and Corruption: Problems and Potential of Individualising a Systemic Problem

Max Planck Institute Anne Peters

Law

A statistical approach to law school citation rankings

University of Virginia Joshua Fischman

Constitutional Law

What is constitutional ideology?

University of Oxford Nick Barber

Civil Procedure Law

Politics, Identity, and Class Certification on the U.S. Courts of Appeals

University of Pennsylvania Sean Farhang

Law

Diversity and Social Mobility in the UK legal profession

Law Society I. Stephanie Boyce

Internet Law

The Law of AI for Good

University of San Diego Orly Lobel

Human Rights Law

Transforming Restorative Justice for Transitional Settings

University of Nottingham Kerry Clamp

Data Valuation and Law

View more insights in Professional - Law - Corporate Law - Internet Law

View more figures from University of Southern California

View the page of Jordan Barry

Read the Study

Next in Corporate Law

Banking Law and Climate Change: Key Legal Issues

Alessandro Gullo, Ender Emre, Mario Tamez | International Monetary Fund

Back to the future: the identification of market power in the Big Data Sector

Andrea Lista | University of Southampton

Jordan Barry proposes a unified framework for valuing data that will promote consistent valuations across fields of law.

Image courtesy of interviewee. April 18, 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 video discusses a joint project focused on the valuation of data, particularly in the context of digital transformation in business and law. It highlights the increased value of data due to the rise of artificial intelligence (AI) and machine learning, which allow businesses and governments to analyze large amounts of data to identify previously hidden relationships. This video video analysis creates economic value, enabling companies to make targeted decisions and improve profitability.

The paper advocates for using real options analysis as a framework to value data. Real options capture the strategic flexibility that data provides businesses, allowing them to adapt, experiment, and change direction based on insights derived from data. This video video gives a concrete example illustrating how the ability to shut down a project after assessing its performance can significantly enhance its attractiveness and overall value.

However, challenges exist in applying real options analysis due to the complexity and uncertainty involved in data valuation. The author acknowledges that while precise mathematical models can be difficult to establish, the framework offers valuable insights into the inherent flexibility that data provides.

The author also notes the legal implications of data valuation, particularly in contexts like data breaches, mergers and acquisitions, and antitrust law, where a clear understanding of data value is crucial. This video video emphasizes the need for a unified approach to data valuation across different legal frameworks to avoid inconsistencies and inefficiencies.

Ultimately, the paper argues that experimenting with real options as a valuation method can yield useful heuristics and improve legal consistency, making the application of the law more sensible and accurate in dealing with data-driven decisions.

Language:

Export as PDF

Export as PDF
Here are some key quotes from the video regarding the value of data and its implications in business and law:

1. **On the Importance of Data**:
- "Data has become much more valuable than it used to be."
- "Data is a big money item at this point."

2. **On AI and Data Relationships**:
- "Using statistical techniques to analyze data and find relationships that weren't previously visible."
- "The key insight is that data gives you new and valuable choices that you didn't have before."

3. **On Real Options Analysis**:
- "We think that the right answer is real option analysis."
- "Real options capture the value of that flexibility."

4. **On Strategic Flexibility**:
- "The ability to take different actions to change your behavior and adjust course is really valuable."
- "The project looks like a really good idea because... you can keep the project going and make $10 million every year."

5. **On Data Valuation Challenges**:
- "How do you value data? That’s a tricky question that’s engendered quite a bit of confusion."
- "This is tricky because how do you value data as sort of a tricky question that’s engendered quite a bit of confusion already."

6. **On the Legal Implications of Data Value**:
- "There are a number of areas of law for which the question of how do you value data is going to be important."
- "Data breaches... how much have you lost? That’s basically a valuation question."

7. **On the Need for a Unified Framework**:
- "We have opposed a solution: a unified framework that we think can be applied across all these different areas and the framework is based on real options."

8. **On the Complexity of Real Options**:
- "Getting correct answers for real options questions is really hard."
- "The mathematics behind valuing stock options... is significant and complicated."

9. **On the Potential Benefits of This Approach**:
- "We think that our approach... helps isolate what it is about data that makes data valuable."
- "Experimenting could be a good way to find those heuristics."

10. **On Addressing Inconsistencies in Law**:
- "Improving consistency across courts could have long-term value and help make the law more sensible and more accurate."

These quotes encapsulate the central themes of discussing the economic value of data, the role of AI, the importance of flexibility in strategy through real options analysis, the complexities of valuing data for legal purposes, and the call for a unified approach across different legal areas.

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

00:00.405 --> 00:00.565
Absolutely.

00:00.565 --> 00:02.205
So this paper comes from, it's a joint project

00:02.235 --> 00:04.165
with Danny Soel, my colleague is in the office

00:04.165 --> 00:05.285
next door to me, great guy.

00:05.945 --> 00:09.045
And the context is there was a USC symposium

00:09.625 --> 00:11.525
on at the LAR view here at USC

00:12.065 --> 00:14.725
on digital transformation in business end in law.

00:14.725 --> 00:16.845
And there really has been quite a change in this area.

00:17.545 --> 00:21.605
So data has become ex much more valuable than it used to be.

00:21.625 --> 00:23.365
And there's kind of a confluence of a couple

00:23.385 --> 00:26.525
of different things that drive this sort of the short answer

00:27.145 --> 00:29.445
is sort of ai, which includes machine learning

00:29.465 --> 00:31.965
and you know, sort of this broad commentary,

00:32.335 --> 00:37.215
which is basically at a high level using

00:37.215 --> 00:40.495
statistical techniques to analyze data

00:40.555 --> 00:43.495
and find relationships that weren't previously visible.

00:43.915 --> 00:45.575
And then to use that data that those,

00:45.575 --> 00:46.895
those relationships you identify

00:47.315 --> 00:48.895
for purposes that generate money.

00:49.235 --> 00:51.475
So lemme back up a little bit.

00:51.855 --> 00:53.515
So the drivers here are,

00:53.515 --> 00:56.195
first there's just a lot more data than there used to be.

00:56.975 --> 01:01.355
So in 2021, over a trillion megabytes a day

01:01.735 --> 01:03.075
of new data was generated

01:03.075 --> 01:04.995
and that number continues to go up since then.

01:05.255 --> 01:08.515
So we're talking about just a phenomenal quantity of data.

01:09.015 --> 01:11.035
And then secondly, computers are getting

01:11.035 --> 01:12.835
increasingly powerful over time as well.

01:13.215 --> 01:14.675
So the ability to process

01:15.055 --> 01:17.515
and act upon this data has been going up

01:18.125 --> 01:19.525
tremendously as well.

01:19.905 --> 01:24.365
And then third, you know, AI is not new, conceptually,

01:25.685 --> 01:28.265
but the increasing speed of collecting data

01:28.325 --> 01:31.865
and ability to process that data have made these tools of

01:32.385 --> 01:34.945
artificial intelligence increasingly powerful

01:35.045 --> 01:36.825
and useful over time.

01:37.285 --> 01:39.665
And so we've seen like this whole

01:39.855 --> 01:41.345
ecosystem build up around data.

01:41.395 --> 01:44.745
There are data brokers, there are, you know, providers

01:44.845 --> 01:47.185
and users and this whole complicated ecosystem

01:47.185 --> 01:50.105
and it's created just a tremendous amount of economic value.

01:50.105 --> 01:53.385
Data is a big money item at this point.

01:53.385 --> 01:54.745
There's new products and services.

01:55.365 --> 01:58.945
Uh, businesses use AI and data. Governments use AI and data.

01:59.325 --> 02:00.465
So lots of end users.

02:01.085 --> 02:04.665
Um, and so that's kind of the very high level approach.

02:04.685 --> 02:06.145
So where does this value come from?

02:06.605 --> 02:08.665
At a high level, what you basically do is you can,

02:08.665 --> 02:11.465
you collect the data and then you conduct statistical

02:11.785 --> 02:14.785
analysis to try and find relationships hidden in the data.

02:15.165 --> 02:16.985
And then you exploit those relationships.

02:17.815 --> 02:20.515
So for example, maybe you would go out

02:20.515 --> 02:23.555
and collect data on what lots of different people buy

02:24.255 --> 02:26.275
and you identify people who buy your product

02:26.975 --> 02:28.715
and you figure out sort of what do they look like,

02:28.745 --> 02:29.795
what other things do they buy,

02:29.795 --> 02:31.475
what sort of habits do they have?

02:31.855 --> 02:34.035
And then you look through your pool of consumer data

02:34.055 --> 02:35.275
and you find, all right, well

02:35.275 --> 02:36.875
what other consumers are out there

02:37.375 --> 02:39.395
who look a lot like people who buy my product

02:39.855 --> 02:41.035
but aren't buying my product.

02:41.485 --> 02:42.395
Maybe these are people

02:42.395 --> 02:43.555
who would be interested in my product.

02:43.765 --> 02:45.635
Maybe these are people I should be advertising to

02:45.655 --> 02:47.755
and trying to reach and I can target those

02:48.035 --> 02:49.155
consumers or future sales.

02:49.455 --> 02:53.115
So that's like a pretty classic use of ai

02:53.215 --> 02:55.555
and that's been quite lucrative for a number of companies.

02:56.135 --> 02:58.595
The key insight of what I'm saying here is that data

02:59.095 --> 03:00.205
Gives you new

03:00.305 --> 03:02.925
and valuable choices that you didn't have before.

03:03.505 --> 03:05.085
That's the way we think about it is that

03:05.085 --> 03:07.405
because you have this data, you can do things

03:07.405 --> 03:10.965
that weren't available to you without it and so forth.

03:11.305 --> 03:12.685
And so how do we value that?

03:12.745 --> 03:14.125
How do you think about something that gives you choices?

03:14.715 --> 03:18.765
Well we think that the right answer is real option analysis.

03:19.465 --> 03:23.085
So again, high level stuff for a second, just the ability

03:23.585 --> 03:26.045
to take different actions to change your behavior

03:26.145 --> 03:28.925
and to adjust course that's really valuable.

03:29.195 --> 03:30.725
That strategic flexibility

03:31.185 --> 03:35.085
and real options also called strategic options capture,

03:35.115 --> 03:37.325
that they capture the value of that flexibility.

03:37.425 --> 03:39.805
How much is it worth that you can change course,

03:39.805 --> 03:42.765
that you can expand your operations or shrink them.

03:43.025 --> 03:45.645
You can accelerate the project timeline or slow it down.

03:45.985 --> 03:48.045
You can switch strategies or change technologies.

03:48.175 --> 03:50.445
These are kind of common classic real options

03:50.465 --> 03:52.885
and we think that this is actually a very good model

03:53.265 --> 03:54.405
for data valuation

03:54.405 --> 03:56.085
because of the core of data is

03:56.085 --> 03:58.165
that it gives you new choices you didn't have.

03:58.545 --> 04:01.365
And real options are what get at that question.

04:01.745 --> 04:03.805
Let me give you another more concrete example

04:04.105 --> 04:05.125
to help anchor this.

04:06.055 --> 04:07.545
Suppose that you are a business

04:07.765 --> 04:09.465
and you're thinking about launching a new product

04:10.285 --> 04:12.665
and you don't know if the product is going

04:12.665 --> 04:13.985
to make you money or not.

04:14.235 --> 04:15.785
Let's say there's like a 50 50 chance

04:15.895 --> 04:17.825
that this will be a valuable product or not.

04:18.085 --> 04:19.985
So let's say if it works out, you know,

04:19.985 --> 04:23.675
the good case you make $10 million every year, you know,

04:23.675 --> 04:25.475
forever or for 10 years or whatever,

04:25.895 --> 04:27.955
and the other half of the time things don't work out

04:28.055 --> 04:31.035
and you lose $20 million a year every year

04:31.035 --> 04:32.195
you have this product on the market.

04:32.865 --> 04:34.355
Well if you look at these as,

04:34.535 --> 04:36.115
you know, I must do this forever.

04:36.215 --> 04:38.035
If I make the product I have to commit to it forever,

04:38.585 --> 04:39.595
then you shouldn't do it

04:39.745 --> 04:41.795
because half the time you make 10 a year,

04:41.795 --> 04:44.035
other time you lose 20 a year on average

04:44.035 --> 04:45.355
you're losing money every year.

04:45.625 --> 04:49.075
It's a bad idea. Now let's give you a new choice.

04:49.615 --> 04:51.675
Assume that you can shut the project down

04:51.725 --> 04:53.315
after a year if you wanna do it.

04:53.785 --> 04:56.395
Well now in the good case, you can keep the project going

04:56.395 --> 04:57.955
and make $10 million every year.

04:58.375 --> 05:01.355
In the bad case, you run it one year, you lose $20 million,

05:01.375 --> 05:02.555
you cut your losses and you quit.

05:03.435 --> 05:04.575
So in this case, the product,

05:04.795 --> 05:06.135
the project looks like a really good idea

05:06.165 --> 05:08.655
because on the plus side, you make $10 million a year

05:08.755 --> 05:10.615
for 10 years or forever or for whatever.

05:10.995 --> 05:12.695
And on the bad case, you only lose the one

05:12.695 --> 05:13.815
year of $20 million.

05:14.205 --> 05:16.895
That becomes a very positive net present value project.

05:17.165 --> 05:18.335
It's something you're gonna want to do

05:18.715 --> 05:21.255
and all that we've changed from making this project a big

05:21.255 --> 05:22.535
money loser, stay away.

05:22.915 --> 05:25.775
No bad, bad idea to great idea, make lots

05:25.775 --> 05:28.655
of money on average is given you one new choice.

05:29.195 --> 05:30.615
That's a real option we've added

05:30.835 --> 05:33.175
and that really increased the value of this project to you.

05:33.435 --> 05:35.975
And this illustrates more generally how the choice

05:35.995 --> 05:38.095
to the ability to change direction, to tweak

05:38.095 --> 05:39.895
what you're doing to make adjustments

05:39.895 --> 05:42.175
and change course is really valuable.

05:42.245 --> 05:45.055
Totally. So the academic

05:45.115 --> 05:47.455
and practitioner literature on data valuation

05:47.595 --> 05:48.655
is pretty complex.

05:49.235 --> 05:51.415
Um, for example, there's some good literature on on data

05:51.445 --> 05:53.335
brokers, which suggests that the cost

05:53.335 --> 05:54.415
of data have gotten lower.

05:54.695 --> 05:56.855
A good place to start there is,

05:57.045 --> 05:59.525
Goldfarb and Tucker, 2019.

05:59.535 --> 06:01.645
There our paper is called Digital Economics.

06:01.985 --> 06:05.005
It appeared in, volume 57 of the Journal

06:05.005 --> 06:06.165
of Economic Literature.

06:06.785 --> 06:09.965
Um, in terms of how data is actually used to create value

06:09.985 --> 06:11.805
and the ecosystem sort of that are developed around that.

06:12.075 --> 06:14.485
There's a large literature, I'm gonna mention a couple

06:14.485 --> 06:17.285
of papers, but I'm certainly gonna, not intentionally,

06:17.545 --> 06:18.925
but I'm sure I will leave out some,

06:19.075 --> 06:21.125
some significant works 'cause it's a big literature.

06:21.355 --> 06:23.205
Here are a couple of just selections.

06:23.705 --> 06:24.965
Um, Ian CDE

06:24.965 --> 06:28.085
and Lanie have a book, competing in the age

06:28.085 --> 06:29.205
of AI strategy

06:29.225 --> 06:31.485
and leadership when algorithms and networks run the world.

06:31.585 --> 06:32.925
That's a 2020 book.

06:33.505 --> 06:35.205
Um, ner Purina

06:35.205 --> 06:38.405
and j have a paper in the Strategy Science Journal,

06:38.405 --> 06:40.125
volume four from 2019.

06:40.625 --> 06:42.885
Uh, what is different about digital strategy from

06:42.885 --> 06:44.565
quantitative to qualitative change?

06:45.585 --> 06:48.485
And then another good paper on the theory

06:48.485 --> 06:52.525
of ecosystems in this area is, Jack Abit is Geno

06:52.645 --> 06:54.605
and Gower, towards the theory

06:54.605 --> 06:57.285
of ecosystems in Strategic Management Journal,

06:57.285 --> 06:59.125
volume 39 from 2018.

06:59.745 --> 07:03.165
And of course I should also mention, Parker Van Alstein

07:03.165 --> 07:07.185
and Jang, their paper is called Platform Ecosystems,

07:07.205 --> 07:08.585
how Developers Invert the Firm.

07:08.925 --> 07:11.225
And that's from, volume 41

07:11.225 --> 07:14.865
of Management Information Systems quarterly, from 2017.

07:15.485 --> 07:18.145
So that's just a, a smattering of literature that might be

07:18.145 --> 07:19.865
of interest to people who care about this topic.

07:20.205 --> 07:22.985
On the law side, there hasn't really been a lot

07:22.985 --> 07:26.425
that looks at how to value data, for legal purposes.

07:27.105 --> 07:29.425
Absolutely. So there are a number of areas of law

07:29.525 --> 07:33.425
for which the question of how do you value data,

07:33.525 --> 07:36.545
is going to be important and will come up.

07:37.665 --> 07:40.125
So one simple example is in the context

07:40.265 --> 07:42.325
of data breaches, right?

07:42.345 --> 07:43.805
So if your data gets breached

07:43.805 --> 07:45.165
and you, you know, you have data and it's supposed

07:45.165 --> 07:46.765
to be kept private and then it gets out,

07:47.265 --> 07:48.325
how much have you lost?

07:49.075 --> 07:51.965
Well how much, like that's basically a valuation question.

07:51.965 --> 07:53.765
How much is it worth as an exclusive data source

07:53.865 --> 07:55.325
or not and so forth.

07:55.745 --> 07:58.285
So I think that's pretty directly implicated there.

07:58.675 --> 08:00.445
Then there are other times when data sort of serves

08:00.445 --> 08:01.485
as a general asset.

08:01.905 --> 08:04.565
So for example, if you're going to sell your company,

08:05.345 --> 08:08.045
and m and a law, you have obligations if you're auctioning

08:08.165 --> 08:09.645
your company off to kind of get the highest price

08:09.645 --> 08:11.205
and make sure that you're getting an appropriate

08:11.585 --> 08:12.685
payoff for your shareholders.

08:13.075 --> 08:14.765
That includes thinking about what you're thinking about,

08:14.765 --> 08:15.845
what your assets are worth.

08:16.145 --> 08:17.965
So if you have data that's an important asset,

08:18.145 --> 08:19.725
you've gotta think about whether you're getting paid

08:19.725 --> 08:21.605
for it in an appropriate way, what the price

08:21.605 --> 08:24.005
of the sale is gonna be good for the shareholders,

08:24.185 --> 08:25.205
and bankruptcy.

08:25.415 --> 08:27.165
Maybe you have data as an asset there.

08:27.225 --> 08:29.565
And then how is, how much is it worth when it gets sold off

08:29.585 --> 08:31.605
or gets valued or assigned to creditors?

08:31.945 --> 08:33.165
That's an important question.

08:33.585 --> 08:36.445
Uh, many taxes involve figuring out the valuation of things.

08:36.625 --> 08:39.085
So like estate taxes for example, value items

08:39.265 --> 08:40.685
or gift taxes potentially.

08:41.065 --> 08:43.565
Um, there are also various sort of digital taxes

08:43.665 --> 08:44.925
of different types that can

08:45.125 --> 08:46.365
implicate this sort of thing as well.

08:46.865 --> 08:48.285
So it's a live question there.

08:48.815 --> 08:51.485
There are a number of antitrust context which has come up.

08:51.585 --> 08:54.605
So there are a lot of areas of law in which valuing data is

08:54.605 --> 08:57.365
Significant and having a a unified way

08:57.365 --> 08:58.365
to think about this is important.

08:58.865 --> 08:59.885
And this is tricky

08:59.885 --> 09:03.405
because how do you value data as sort of a tricky question

09:03.405 --> 09:06.405
that's engendered quite a bit of confusion already.

09:07.145 --> 09:10.405
So it's worth thinking about sort of what is data, you know,

09:10.525 --> 09:12.925
'cause it's not just the data that gets used in this process

09:13.065 --> 09:14.365
of we talked about earlier.

09:14.745 --> 09:17.285
You know, there's the AI algorithms that apply the data

09:17.705 --> 09:19.125
and they're sort of how you implement

09:19.315 --> 09:20.685
what you've learned about the data.

09:20.685 --> 09:23.045
Like maybe I am really good identifying which customers I

09:23.045 --> 09:25.205
should target, but am I actually good at then making the

09:25.445 --> 09:26.765
advertising for them or things like that.

09:27.225 --> 09:29.365
So that's stuff you have to worry about. So focusing on sort

09:29.365 --> 09:31.725
of the data piece of it is one important point

09:31.725 --> 09:33.085
of clarity, which is significant.

09:33.585 --> 09:35.165
And then of course another piece of it is

09:35.165 --> 09:36.405
what does it mean to own the data?

09:36.435 --> 09:37.965
What legal rights do you have

09:38.065 --> 09:40.845
as the data quote unquote owner or user or what have you

09:41.465 --> 09:43.085
and which rights are at play

09:43.105 --> 09:44.685
and what are each individual right worth?

09:44.705 --> 09:46.005
Or maybe we'll have the value of the bundle.

09:46.505 --> 09:48.165
That's an important question as well.

09:48.225 --> 09:50.045
So there's a lot of confusion on that and

09:50.205 --> 09:52.885
whenever you have confusion, you're gonna have a lot

09:52.885 --> 09:54.725
of opportunity for inconsistency.

09:55.145 --> 09:56.605
Uh, and you could have this, you know, both

09:56.605 --> 09:58.165
between the literature and the case law,

09:58.165 --> 09:59.165
which could be problematic.

09:59.705 --> 10:01.205
Uh, and then of course I mentioned there's a bunch

10:01.205 --> 10:02.285
of different areas of law.

10:02.285 --> 10:03.925
They're gonna have to consider this data question

10:04.225 --> 10:06.565
and they might take up different approaches to valuing data,

10:06.695 --> 10:09.605
which wouldn't be great because presumably data is worth

10:10.115 --> 10:12.885
some amount and you want the law to get it right

10:12.885 --> 10:14.125
and to at least be consistent

10:14.125 --> 10:16.565
and you wouldn't necessarily want, say tax law

10:16.565 --> 10:19.205
to value data differently from bankruptcy law or so forth.

10:19.475 --> 10:20.925
That seems like it could be problematic.

10:21.265 --> 10:24.285
And of course we also have different jurisdictions.

10:24.585 --> 10:26.805
So maybe like one state values it a different way

10:26.805 --> 10:28.325
and a different state values in a different way

10:28.595 --> 10:30.285
that seems like it wouldn't be great either.

10:30.285 --> 10:31.805
That could be problematic as well

10:31.825 --> 10:33.005
for many of the same reasons.

10:33.505 --> 10:36.565
And you put this kind of commun confusion leading

10:36.565 --> 10:39.685
to inconsistency that can result in inefficiency.

10:40.225 --> 10:43.605
So you can get a bunch of bad results, you can get rules

10:43.605 --> 10:44.645
that don't line up correctly

10:44.645 --> 10:46.765
where things are treated differently across the law

10:46.765 --> 10:47.885
in ways that are unfortunate.

10:48.225 --> 10:49.685
And you also create the opportunity

10:49.825 --> 10:52.845
for sophisticated parties to take advantage of the law.

10:53.265 --> 10:55.005
Uh, if the law isn't consistent

10:55.005 --> 10:56.005
and there's holes you can do,

10:56.205 --> 10:57.565
regulatory arbitrage is the phrase

10:57.865 --> 10:59.085
that's used in the literature.

10:59.385 --> 11:01.805
You can basically, they kind of minor changes to

11:01.805 --> 11:04.365
what you're doing that don't change the economics,

11:04.625 --> 11:06.845
but that shift the legal regime that you're dealing with

11:06.845 --> 11:07.885
that creates opportunities

11:07.945 --> 11:09.205
for improving your situation

11:09.475 --> 11:10.965
without changing your economics.

11:11.055 --> 11:13.525
Which seems like something we generally,

11:13.755 --> 11:15.005
well it's not always bad,

11:15.025 --> 11:16.645
but it seems like it can create some

11:16.645 --> 11:17.845
problems, put it that way.

11:18.145 --> 11:19.965
So we have in our paper opposed a solution.

11:20.345 --> 11:22.085
We have sort of a unified framework

11:22.085 --> 11:24.325
that we think can be applied across all these different

11:24.325 --> 11:27.325
areas and the framework is based on real options.

11:28.005 --> 11:29.805
Absolutely. So lemme tell you why real options,

11:29.805 --> 11:31.125
we think it can be a good solution.

11:31.465 --> 11:34.045
So as a starting point, we think it's actually the right way

11:34.045 --> 11:36.525
to think about data and why it makes it valuable.

11:36.845 --> 11:38.365
I think like if you wanna get the right answer

11:38.365 --> 11:40.485
to a question, you have to ask the right question.

11:41.385 --> 11:44.925
And so what are data? Like what, why is data useful?

11:44.925 --> 11:46.525
Well, we think it's useful because it lets

11:46.525 --> 11:47.605
you do things you couldn't do.

11:48.645 --> 11:51.145
And that's what Real Options gets at.

11:51.145 --> 11:52.905
It gives you data, gives you new opportunities,

11:53.345 --> 11:56.925
New choices, new things that you couldn't do previously

11:56.925 --> 11:58.165
that you now can do.

11:58.545 --> 12:00.125
And that's where the value comes from.

12:00.505 --> 12:02.685
And by asking that question, we think we're honing in on

12:02.685 --> 12:05.005
what data is contributing and why it is valuable.

12:05.385 --> 12:07.445
We also think this is a very flexible framework.

12:07.705 --> 12:09.965
You can apply it in a lot of different contexts.

12:10.465 --> 12:12.845
Uh, it's expansive, it's generalizable.

12:13.265 --> 12:15.205
So that's the good side of the ledger,

12:15.205 --> 12:17.005
which we think is is pretty good.

12:17.315 --> 12:20.565
Yeah, so one thing you could do is you could basically say,

12:20.565 --> 12:22.485
all right, we have this data and we have a sense of

12:22.485 --> 12:25.045
how it could be applied to,

12:25.225 --> 12:27.085
affect our new product development

12:27.185 --> 12:28.725
or our customer acquisition.

12:29.065 --> 12:30.485
And so you basically have to do some kind

12:30.485 --> 12:32.605
of discounted cashflow app model where you'd say, all right,

12:32.675 --> 12:34.205
here are the choices available to us

12:34.205 --> 12:35.765
and how it could increase our revenues

12:35.765 --> 12:37.845
or profits over time and go from there.

12:38.225 --> 12:40.125
So that's kind of the, you know, that's sort of a battle

12:40.125 --> 12:41.165
of the expert sort of story,

12:41.165 --> 12:43.085
but you least have the experts speaking the same language

12:43.355 --> 12:47.045
with the same framework, kind of able to talk to each other.

12:47.105 --> 12:49.045
And then the courts get a better sense

12:49.045 --> 12:51.245
of when people are kind of all on the same starting point.

12:52.215 --> 12:53.955
Uh, it gets easier for courts sort of weigh

12:53.955 --> 12:55.035
what different people are saying and reach

12:55.035 --> 12:56.195
more sensible conclusions.

12:56.975 --> 13:00.435
It also constraints what people can say, into a more

13:01.395 --> 13:03.155
a bounded domain, which is helpful.

13:03.575 --> 13:05.235
So it's not all good news.

13:05.935 --> 13:09.715
The chief limitation involved with using real options.

13:09.715 --> 13:12.435
The way we describe is that getting correct answers

13:12.455 --> 13:15.115
for real options questions is really hard.

13:15.855 --> 13:17.915
So let me explain this a little bit more.

13:18.495 --> 13:20.235
Um, when I talk about options, the thing

13:20.235 --> 13:21.275
that most people think of a lot

13:21.275 --> 13:22.715
of the time is financial options.

13:22.815 --> 13:25.155
You know, like stock options buying and selling stock.

13:26.015 --> 13:28.915
So the classic model

13:29.135 --> 13:32.715
for valuing stock options is the Black Shoals model.

13:33.575 --> 13:35.315
And that's really hard.

13:35.415 --> 13:37.995
The mathematics behind that is significant and complicated.

13:38.115 --> 13:39.155
I mean literally deriving

13:39.155 --> 13:41.635
that model was a Nobel Prize winning achievement.

13:41.985 --> 13:44.395
Like not easy, okay?

13:44.855 --> 13:48.275
Uh, and real options are way worse in terms

13:48.275 --> 13:49.435
of figuring them out because there's

13:49.435 --> 13:50.555
so many more things you can do.

13:50.585 --> 13:52.275
It's not just do I buy it on this date?

13:52.425 --> 13:53.915
There's a lot of choices you can

13:53.915 --> 13:54.955
make and when you make the choices.

13:55.415 --> 13:57.595
So it gets really quite difficult

13:57.615 --> 13:59.595
to get precise mathematical answers.

14:00.055 --> 14:02.715
And given how difficult the math can be, you know, one,

14:02.715 --> 14:03.995
you have to know the inputs

14:03.995 --> 14:05.355
and to organize the inputs is tricky.

14:05.775 --> 14:07.675
Uh, that of course is going to be difficult

14:07.735 --> 14:09.075
for a court to administer.

14:09.415 --> 14:11.795
You've gotta be worried about parties and manipulating them

14:11.795 --> 14:13.635
because if it's complicated, you know, the people

14:13.635 --> 14:16.155
who know more and generally parties will know more about

14:16.155 --> 14:19.635
this, then courts will, then you have to worry about

14:19.665 --> 14:22.195
ways they might manipulate numbers and so forth.

14:22.375 --> 14:24.435
So that's really challenging,

14:24.815 --> 14:26.915
but we're still pretty optimistic

14:27.105 --> 14:30.835
because for one thing, you don't have to be perfect

14:31.295 --> 14:32.875
to be better than the alternative.

14:32.875 --> 14:34.435
Like the goal is to make things better.

14:35.175 --> 14:37.135
And you know, I think

14:37.135 --> 14:38.575
that the real options framework does

14:38.575 --> 14:39.655
highlight a couple of things.

14:40.075 --> 14:42.695
It does make clear for one thing that data is valuable,

14:42.915 --> 14:45.255
having real options is in fact useful

14:45.835 --> 14:47.935
and that's going to be something that will be clear.

14:48.355 --> 14:49.375
And I think there are some

14:49.375 --> 14:50.815
factors. Courts are good at dealing

14:50.815 --> 14:52.720
With factor tests, something they deal with all the time,

14:53.145 --> 14:55.125
and there are factors that affect valuation.

14:55.625 --> 14:56.765
So for options,

14:56.965 --> 14:58.565
a longer lived option is more

14:58.845 --> 14:59.885
valuable than a shorter lived option.

14:59.885 --> 15:02.125
Having more time in which you can make choices is better.

15:02.545 --> 15:04.965
So in the data context that translates to

15:04.965 --> 15:06.365
how long is the data useful for?

15:06.365 --> 15:07.885
Does it have a long useful life?

15:08.195 --> 15:10.325
Does knowing this help you make decisions for many years

15:10.325 --> 15:12.725
to come or is it only good for like the next week?

15:13.025 --> 15:14.165
That's an important feature.

15:14.515 --> 15:15.685
Another feature is sort

15:15.685 --> 15:17.565
of the informational uniqueness of the data.

15:17.945 --> 15:19.765
Can I get the same information without this

15:19.765 --> 15:20.885
data through some other means?

15:21.465 --> 15:22.485
So is this telling me something

15:22.485 --> 15:23.605
I wouldn't be able to get elsewhere?

15:23.865 --> 15:26.165
Is it really adding choices that I couldn't

15:26.165 --> 15:28.445
otherwise get that's valuable.

15:28.985 --> 15:30.805
Um, how uncertain is the future?

15:31.075 --> 15:33.005
Usually when you're valuing the future?

15:33.665 --> 15:35.245
Um, uncertainty is bad.

15:35.585 --> 15:37.405
So if you're thinking about like if I'm loaning you money,

15:37.865 --> 15:41.445
the more risky you are, the less I am going

15:41.445 --> 15:42.605
to value your payments in the future.

15:42.875 --> 15:45.045
Similarly with stocks like uncertainty

15:45.045 --> 15:46.325
is considered a bad thing.

15:46.475 --> 15:49.085
Generally speaking options aren't like that

15:49.085 --> 15:50.205
because the ability

15:50.225 --> 15:52.965
to change direction becomes more valuable depending on

15:52.965 --> 15:55.605
how wide a spread outcomes can have.

15:56.065 --> 15:57.965
So if whatever I do, the result is the same,

15:58.445 --> 15:59.645
changing directions is not valuable,

15:59.905 --> 16:01.005
but if, you know,

16:01.165 --> 16:04.965
changing direction can change my fortunes a lot, then

16:05.425 --> 16:06.965
the option becomes more valuable.

16:06.985 --> 16:08.405
So the more uncertainty there is,

16:08.625 --> 16:09.965
the more valuable the option

16:10.345 --> 16:13.485
and interest rates can also play in here as well in sort

16:13.485 --> 16:15.725
of the same kind of way that they play into kind

16:15.725 --> 16:18.165
of standard financial option valuation as well.

16:18.705 --> 16:21.445
So we think that this is an area where experimentation

16:21.445 --> 16:24.005
and practice might actually produce a lot

16:24.005 --> 16:25.285
of useful heuristics.

16:25.745 --> 16:29.485
It might be the case that we can't value these directly,

16:29.705 --> 16:31.365
but we can find sort of proxies

16:31.475 --> 16:33.365
that get at the same concept in a good way

16:33.665 --> 16:35.005
and that could be a good way forward.

16:35.265 --> 16:37.125
And there's some literature on this that suggests

16:37.125 --> 16:39.645
that private firms do sort of similar things.

16:39.935 --> 16:43.125
There are a bunch of metrics firms use when making capital

16:43.565 --> 16:46.405
allocation decisions, which aren't great from like a

16:47.005 --> 16:48.125
straightforward finance theory.

16:48.155 --> 16:50.365
Like how fast do you get your money back from a project

16:50.905 --> 16:53.085
is not something that would normally enter into a discounted

16:53.085 --> 16:54.325
cash flow analysis per se.

16:54.705 --> 16:56.725
But if you're thinking about from an option perspective,

16:56.745 --> 17:00.165
it actually ends up correlating reasonably well,

17:00.955 --> 17:03.365
with kind of, option value.

17:03.545 --> 17:06.005
And so the frequent use, the frequent

17:06.815 --> 17:10.365
firms making frequent use of those kinds of heuristic,

17:10.535 --> 17:11.845
might indicate that that's a way

17:11.845 --> 17:13.725
that firms are basically accounting for option value.

17:14.025 --> 17:15.005
And you could see sort of imagine

17:15.005 --> 17:16.165
courts doing something kind of similar.

17:16.675 --> 17:18.845
I'll also throw in that because it's hard to value,

17:19.275 --> 17:21.405
it's hard for firms to get exactly right also.

17:21.865 --> 17:24.405
So in that sense, maybe in a similar boat to courts,

17:24.405 --> 17:25.885
which could reduce gamesmanship a little bit.

17:26.665 --> 17:28.285
And honestly at the end of the day,

17:28.425 --> 17:30.885
if this approach doesn't pan out, like maybe we try it

17:30.885 --> 17:32.005
and it just turns out to be too

17:32.005 --> 17:33.885
hard, that'd be good to know too.

17:34.145 --> 17:35.925
We can act accordingly. Like maybe we have to sort

17:35.925 --> 17:37.245
of abandon the field in some level

17:37.265 --> 17:38.525
and find another way to do it.

17:38.785 --> 17:41.005
And there's a nice analogy here I think,

17:41.025 --> 17:42.525
to the stock options case.

17:42.945 --> 17:44.765
So, I'm a, I do a lot of tax,

17:44.765 --> 17:45.845
so I'm gonna give a tax example.

17:46.315 --> 17:48.845
Section 83 of the tax code,

17:48.925 --> 17:51.885
covers when employees receive property as compensation.

17:51.945 --> 17:53.765
So example is like you're an executive

17:54.065 --> 17:56.285
and you get stock in the company as part

17:56.285 --> 17:57.365
of your compensation package.

17:58.145 --> 18:01.605
So if you get stock options, the rule is different.

18:01.605 --> 18:02.925
Like normally if you get stock,

18:03.385 --> 18:05.485
we tax you when you receive it, assuming you get

18:05.485 --> 18:07.765
to like once it's clear, you get to keep it, we tax you.

18:07.825 --> 18:09.725
I'm oversimplifying of course, but that's the general idea.

18:09.725 --> 18:12.605
If you're gonna keep it, we tax you with stock options.

18:12.605 --> 18:13.885
That's generally not the rule.

18:14.225 --> 18:15.245
And presumably that's

18:15.245 --> 18:17.285
because it's really hard to value stock options

18:17.285 --> 18:19.045
and the service was getting beaten up over it.

18:19.145 --> 18:20.285
So they just gave up, they said, you know what,

18:20.285 --> 18:21.365
we're not gonna value stock options.

18:21.365 --> 18:24.125
We're gonna wait until you exercise and actually get stock.

18:24.465 --> 18:26.165
And at that point we'll tax you.

18:26.545 --> 18:28.085
So they sort of left the options field.

18:28.085 --> 18:29.845
Unless of course the option is easy to value,

18:29.845 --> 18:31.045
like if it's publicly traded,

18:31.235 --> 18:32.445
then we'll just tax you right away.

18:32.445 --> 18:35.005
But otherwise we'll wait till the ground is more favorable

18:35.005 --> 18:37.005
and tax you then and one could imagine

18:37.785 --> 18:40.085
not literally the same approach in the data contest Exactly,

18:40.625 --> 18:44.055
but analogs, and that would be good to know too.

18:44.275 --> 18:46.695
So to wrap things up, we think that our approach,

18:46.785 --> 18:49.615
which we think about data valuation using real options,

18:49.675 --> 18:51.415
we think it's a good way to think about this problem.

18:52.035 --> 18:54.975
We think it helps isolate what it is about data

18:55.045 --> 18:56.495
that makes data valuable

18:56.675 --> 18:58.055
and lets you focus it on the value

18:58.055 --> 19:00.055
of the data itself versus the value of other things

19:00.565 --> 19:02.495
like the algorithms you're using

19:03.155 --> 19:05.455
to analyze the data and things like that.

19:05.715 --> 19:09.055
You know, you know, to analogize a little bit, how much

19:09.055 --> 19:10.975
of it is the ingredients versus the value

19:10.975 --> 19:13.295
of the recipe versus the value of the chef and so forth.

19:13.295 --> 19:14.815
We're really isolating the ingredients,

19:14.815 --> 19:16.615
the data itself, which we think is good.

19:16.955 --> 19:18.815
Uh, we think this is an area where there might be a lot

19:18.815 --> 19:21.935
of useful heuristics, even if we don't approve,

19:22.305 --> 19:25.695
apply like a formal real options analysis.

19:25.975 --> 19:27.735
Ultimately there might be some really good heuristics

19:27.735 --> 19:29.255
that approaching it from this perspective

19:29.255 --> 19:30.535
would ultimately yield.

19:30.835 --> 19:33.375
And we think that experimenting could be a good way

19:33.375 --> 19:34.615
to find those heuristics.

19:34.915 --> 19:37.935
And improving consistency across courts could have long-term

19:38.025 --> 19:41.975
value and help make the law more sensible and more accurate.

Unlocking Value Through Data Transformation

00:00:00 - Absolutely. So this paper comes from, it's a joint project with Danny Soel, my colleague is in the office next door to me, great guy. And the context is there was a USC symposium on at the LAR view here at USC on digital transformation in business end in law. And there really has been quite a change in this area. So data has become ex much more valuable than it used to be. And there's kind of a confluence of a couple of different things that drive this sort of the short answer is sort of ai, which includes machine learning and you know, sort of this broad commentary,
00:00:37 - statistical techniques to analyze data and find relationships that weren't previously visible. And then to use that data that those, those relationships you identify for purposes that generate money. So lemme back up a little bit. So the drivers here are, first there's just a lot more data than there used to be. So in 2021, over a trillion megabytes a day of new data was generated
00:01:05 - So we're talking about just a phenomenal quantity of data. And then secondly, computers are getting increasingly powerful over time as well. So the ability to process and act upon this data has been going up tremendously as well. And then third, you know, AI is not new, conceptually, but the increasing speed of collecting data and ability to process that data have made these tools of artificial intelligence increasingly powerful
00:01:37 - And so we've seen like this whole ecosystem build up around data. There are data brokers, there are, you know, providers and users and this whole complicated ecosystem and it's created just a tremendous amount of economic value. Data is a big money item at this point. There's new products and services. businesses use AI and data. Governments use AI and data. So lots of end users. and so that's kind of the very high level approach. So where does this value come from?
00:02:08 - you collect the data and then you conduct statistical analysis to try and find relationships hidden in the data. And then you exploit those relationships. So for example, maybe you would go out and collect data on what lots of different people buy and you identify people who buy your product and you figure out sort of what do they look like, what other things do they buy, what sort of habits do they have? And then you look through your pool of consumer data and you find, all right, well what other consumers are out there
00:02:39 - but aren't buying my product. Maybe these are people who would be interested in my product. Maybe these are people I should be advertising to and trying to reach and I can target those consumers or future sales. So that's like a pretty classic use of ai and that's been quite lucrative for a number of companies. The key insight of what I'm saying here is that data Gives you new and valuable choices that you didn't have before. That's the way we think about it is that because you have this data, you can do things

Valuing Data Through Real Options Analysis

00:03:11 - And so how do we value that? How do you think about something that gives you choices? Well we think that the right answer is real option analysis. So again, high level stuff for a second, just the ability to take different actions to change your behavior and to adjust course that's really valuable. That strategic flexibility and real options also called strategic options capture, that they capture the value of that flexibility.
00:03:39 - that you can expand your operations or shrink them. You can accelerate the project timeline or slow it down. You can switch strategies or change technologies. These are kind of common classic real options and we think that this is actually a very good model for data valuation because of the core of data is that it gives you new choices you didn't have. And real options are what get at that question. Let me give you another more concrete example to help anchor this. Suppose that you are a business
00:04:10 - and you don't know if the product is going to make you money or not. Let's say there's like a 50 50 chance that this will be a valuable product or not. So let's say if it works out, you know, the good case you make $10 million every year, you know, forever or for 10 years or whatever, and the other half of the time things don't work out and you lose $20 million a year every year you have this product on the market. Well if you look at these as, you know, I must do this forever. If I make the product I have to commit to it forever,
00:04:39 - because half the time you make 10 a year, other time you lose 20 a year on average you're losing money every year. It's a bad idea. Now let's give you a new choice. Assume that you can shut the project down after a year if you wanna do it. Well now in the good case, you can keep the project going and make $10 million every year. In the bad case, you run it one year, you lose $20 million, you cut your losses and you quit. So in this case, the product, the project looks like a really good idea because on the plus side, you make $10 million a year
00:05:10 - And on the bad case, you only lose the one year of $20 million. That becomes a very positive net present value project. It's something you're gonna want to do and all that we've changed from making this project a big money loser, stay away. No bad, bad idea to great idea, make lots of money on average is given you one new choice. That's a real option we've added and that really increased the value of this project to you. And this illustrates more generally how the choice to the ability to change direction, to tweak what you're doing to make adjustments
00:05:42 - Totally. So the academic and practitioner literature on data valuation is pretty complex. for example, there's some good literature on on data brokers, which suggests that the cost of data have gotten lower. A good place to start there is, Goldfarb and Tucker, 2019. There our paper is called Digital Economics. It appeared in, volume 57 of the Journal of Economic Literature. in terms of how data is actually used to create value

Valuing Data: Legal Perspectives and Literature

00:06:12 - There's a large literature, I'm gonna mention a couple of papers, but I'm certainly gonna, not intentionally, but I'm sure I will leave out some, some significant works 'cause it's a big literature. Here are a couple of just selections. Ian CDE and Lanie have a book, competing in the age of AI strategy and leadership when algorithms and networks run the world. That's a 2020 book. ner Purina and j have a paper in the Strategy Science Journal, volume four from 2019.
00:06:42 - quantitative to qualitative change? And then another good paper on the theory of ecosystems in this area is, Jack Abit is Geno and Gower, towards the theory of ecosystems in Strategic Management Journal, volume 39 from 2018. And of course I should also mention, Parker Van Alstein and Jang, their paper is called Platform Ecosystems, how Developers Invert the Firm. And that's from, volume 41
00:07:15 - So that's just a, a smattering of literature that might be of interest to people who care about this topic. On the law side, there hasn't really been a lot that looks at how to value data, for legal purposes. Absolutely. So there are a number of areas of law for which the question of how do you value data, is going to be important and will come up. So one simple example is in the context of data breaches, right?
00:07:43 - and you, you know, you have data and it's supposed to be kept private and then it gets out, how much have you lost? Well how much, like that's basically a valuation question. How much is it worth as an exclusive data source or not and so forth. So I think that's pretty directly implicated there. Then there are other times when data sort of serves as a general asset. So for example, if you're going to sell your company, and m and a law, you have obligations if you're auctioning your company off to kind of get the highest price and make sure that you're getting an appropriate payoff for your shareholders.
00:08:14 - what your assets are worth. So if you have data that's an important asset, you've gotta think about whether you're getting paid for it in an appropriate way, what the price of the sale is gonna be good for the shareholders, and bankruptcy. Maybe you have data as an asset there. And then how is, how much is it worth when it gets sold off or gets valued or assigned to creditors? That's an important question. many taxes involve figuring out the valuation of things. So like estate taxes for example, value items or gift taxes potentially. there are also various sort of digital taxes
00:08:45 - implicate this sort of thing as well. So it's a live question there. There are a number of antitrust context which has come up. So there are a lot of areas of law in which valuing data is Significant and having a a unified way to think about this is important. And this is tricky because how do you value data as sort of a tricky question that's engendered quite a bit of confusion already. So it's worth thinking about sort of what is data, you know, 'cause it's not just the data that gets used in this process of we talked about earlier.

Legal Complexity in Data Ownership and Valuation

00:09:17 - and they're sort of how you implement what you've learned about the data. Like maybe I am really good identifying which customers I should target, but am I actually good at then making the advertising for them or things like that. So that's stuff you have to worry about. So focusing on sort of the data piece of it is one important point of clarity, which is significant. And then of course another piece of it is what does it mean to own the data? What legal rights do you have as the data quote unquote owner or user or what have you and which rights are at play and what are each individual right worth? Or maybe we'll have the value of the bundle.
00:09:48 - So there's a lot of confusion on that and whenever you have confusion, you're gonna have a lot of opportunity for inconsistency. and you could have this, you know, both between the literature and the case law, which could be problematic. and then of course I mentioned there's a bunch of different areas of law. They're gonna have to consider this data question and they might take up different approaches to valuing data, which wouldn't be great because presumably data is worth some amount and you want the law to get it right and to at least be consistent and you wouldn't necessarily want, say tax law
00:10:19 - That seems like it could be problematic. And of course we also have different jurisdictions. So maybe like one state values it a different way and a different state values in a different way that seems like it wouldn't be great either. That could be problematic as well for many of the same reasons. And you put this kind of commun confusion leading to inconsistency that can result in inefficiency. So you can get a bunch of bad results, you can get rules that don't line up correctly where things are treated differently across the law
00:10:48 - And you also create the opportunity for sophisticated parties to take advantage of the law. if the law isn't consistent and there's holes you can do, regulatory arbitrage is the phrase that's used in the literature. You can basically, they kind of minor changes to what you're doing that don't change the economics, but that shift the legal regime that you're dealing with that creates opportunities for improving your situation without changing your economics. Which seems like something we generally, well it's not always bad, but it seems like it can create some problems, put it that way.
00:11:20 - We have sort of a unified framework that we think can be applied across all these different areas and the framework is based on real options. Absolutely. So lemme tell you why real options, we think it can be a good solution. So as a starting point, we think it's actually the right way to think about data and why it makes it valuable. I think like if you wanna get the right answer to a question, you have to ask the right question. And so what are data? Like what, why is data useful? Well, we think it's useful because it lets you do things you couldn't do.
00:11:51 - It gives you data, gives you new opportunities, New choices, new things that you couldn't do previously that you now can do. And that's where the value comes from. And by asking that question, we think we're honing in on what data is contributing and why it is valuable. We also think this is a very flexible framework. You can apply it in a lot of different contexts. it's expansive, it's generalizable. So that's the good side of the ledger, which we think is is pretty good. Yeah, so one thing you could do is you could basically say,

Valuing Data Through Real Options Framework

00:12:22 - how it could be applied to, affect our new product development or our customer acquisition. And so you basically have to do some kind of discounted cashflow app model where you'd say, all right, here are the choices available to us and how it could increase our revenues or profits over time and go from there. So that's kind of the, you know, that's sort of a battle of the expert sort of story, but you least have the experts speaking the same language with the same framework, kind of able to talk to each other. And then the courts get a better sense of when people are kind of all on the same starting point.
00:12:53 - what different people are saying and reach more sensible conclusions. It also constraints what people can say, into a more a bounded domain, which is helpful. So it's not all good news. The chief limitation involved with using real options. The way we describe is that getting correct answers for real options questions is really hard. So let me explain this a little bit more. when I talk about options, the thing that most people think of a lot of the time is financial options.
00:13:26 - So the classic model for valuing stock options is the Black Shoals model. And that's really hard. The mathematics behind that is significant and complicated. I mean literally deriving that model was a Nobel Prize winning achievement. Like not easy, okay? and real options are way worse in terms of figuring them out because there's so many more things you can do. It's not just do I buy it on this date? There's a lot of choices you can
00:13:55 - So it gets really quite difficult to get precise mathematical answers. And given how difficult the math can be, you know, one, you have to know the inputs and to organize the inputs is tricky. that of course is going to be difficult for a court to administer. You've gotta be worried about parties and manipulating them because if it's complicated, you know, the people who know more and generally parties will know more about this, then courts will, then you have to worry about ways they might manipulate numbers and so forth. So that's really challenging,
00:14:27 - because for one thing, you don't have to be perfect to be better than the alternative. Like the goal is to make things better. And you know, I think that the real options framework does highlight a couple of things. It does make clear for one thing that data is valuable, having real options is in fact useful and that's going to be something that will be clear. And I think there are some factors. Courts are good at dealing With factor tests, something they deal with all the time, and there are factors that affect valuation.
00:14:56 - a longer lived option is more valuable than a shorter lived option. Having more time in which you can make choices is better. So in the data context that translates to how long is the data useful for? Does it have a long useful life? Does knowing this help you make decisions for many years to come or is it only good for like the next week? That's an important feature. Another feature is sort of the informational uniqueness of the data. Can I get the same information without this data through some other means? So is this telling me something I wouldn't be able to get elsewhere? Is it really adding choices that I couldn't

Valuing Future Uncertainty in Financial Options

00:15:28 - how uncertain is the future? Usually when you're valuing the future? uncertainty is bad. So if you're thinking about like if I'm loaning you money, the more risky you are, the less I am going to value your payments in the future. Similarly with stocks like uncertainty is considered a bad thing. Generally speaking options aren't like that because the ability to change direction becomes more valuable depending on how wide a spread outcomes can have. So if whatever I do, the result is the same,
00:15:59 - but if, you know, changing direction can change my fortunes a lot, then the option becomes more valuable. So the more uncertainty there is, the more valuable the option and interest rates can also play in here as well in sort of the same kind of way that they play into kind of standard financial option valuation as well. So we think that this is an area where experimentation and practice might actually produce a lot of useful heuristics. It might be the case that we can't value these directly,
00:16:31 - that get at the same concept in a good way and that could be a good way forward. And there's some literature on this that suggests that private firms do sort of similar things. There are a bunch of metrics firms use when making capital allocation decisions, which aren't great from like a straightforward finance theory. Like how fast do you get your money back from a project is not something that would normally enter into a discounted cash flow analysis per se. But if you're thinking about from an option perspective, it actually ends up correlating reasonably well,
00:17:03 - And so the frequent use, the frequent firms making frequent use of those kinds of heuristic, might indicate that that's a way that firms are basically accounting for option value. And you could see sort of imagine courts doing something kind of similar. I'll also throw in that because it's hard to value, it's hard for firms to get exactly right also. So in that sense, maybe in a similar boat to courts, which could reduce gamesmanship a little bit. And honestly at the end of the day, if this approach doesn't pan out, like maybe we try it and it just turns out to be too
00:17:34 - We can act accordingly. Like maybe we have to sort of abandon the field in some level and find another way to do it. And there's a nice analogy here I think, to the stock options case. So, I'm a, I do a lot of tax, so I'm gonna give a tax example. Section 83 of the tax code, covers when employees receive property as compensation. So example is like you're an executive and you get stock in the company as part of your compensation package. So if you get stock options, the rule is different. Like normally if you get stock,
00:18:05 - to like once it's clear, you get to keep it, we tax you. I'm oversimplifying of course, but that's the general idea. If you're gonna keep it, we tax you with stock options. That's generally not the rule. And presumably that's because it's really hard to value stock options and the service was getting beaten up over it. So they just gave up, they said, you know what, we're not gonna value stock options. We're gonna wait until you exercise and actually get stock. And at that point we'll tax you. So they sort of left the options field. Unless of course the option is easy to value, like if it's publicly traded, then we'll just tax you right away. But otherwise we'll wait till the ground is more favorable

Valuing Data Through Real Options Analysis

00:18:37 - not literally the same approach in the data contest Exactly, but analogs, and that would be good to know too. So to wrap things up, we think that our approach, which we think about data valuation using real options, we think it's a good way to think about this problem. We think it helps isolate what it is about data that makes data valuable and lets you focus it on the value of the data itself versus the value of other things like the algorithms you're using to analyze the data and things like that.
00:19:09 - of it is the ingredients versus the value of the recipe versus the value of the chef and so forth. We're really isolating the ingredients, the data itself, which we think is good. we think this is an area where there might be a lot of useful heuristics, even if we don't approve, apply like a formal real options analysis. Ultimately there might be some really good heuristics that approaching it from this perspective would ultimately yield. And we think that experimenting could be a good way to find those heuristics. And improving consistency across courts could have long-term
00:19:38 -

Faculti AI can make mistakes

Research

Policy

Analysis