The Economic Consequences of Internet Censorship (33c3)

33c3 pre-roll music

Herald: Err ...

H: ... a talk would be good, right?

applause

Do you want to give a talk?

Toni: Aah, it’s a little early
but I’ll try.

Herald: Okay, guys, well, I found someone
who’s willing to give a talk!

laughter and applause

That is most excellent.
So, if you ever asked yourself,

I’ve got this big regime and
I’m rolling out internet censorship,

what does my economy do?

There are people in here
asking that question, right?

There’s always someone at Congress
who’s asking some question.

Well, you came to the right place,
and as part of her PhD thesis work

Toni is going answer that question,
hopefully, to a satisfactory point.

Please give a warm round of applause!
applause

Toni!
ongoing applause

Toni: Okay, thanks everyone for being
here, I hope you can all hear me

correctly. And I’m glad to be here
and to be presenting

some part of my thesis to day.
Now, this is ongoing work

so I’m really grateful for any kind of feedback
that you guys would have

and I’m really only presenting this
as kind of a first try,

because when I looked at the topic
of internet censorship

and what that could mean for an economy,
I really didn’t find anything academic

and I was quite surprised: it seemed
like a very obvious question to me,

because I was looking mostly
at China at the beginning.

And I read a lot of newspaper articles
and I talked to a lot of businessmen

who told me: “Well, doing business
in China is very difficult”

and I think China is really
holding itself back by having

this big censorship thing going.

But no one really looked into
how it is holding itself back

or if it is even holding itself back.

So there is really
very, very little research.

And we don’t even have an agreement among
economists or business studies people

about what impact the internet has
on the economy. So if you want to ask:

“So what does internet censorship do
to an economy?” it seems pretty obvious

to first ask: “What does the internet do to
an economy?” and we don’t even know that.

That was quite surprising to me and I’m
going to be talking about the reasons

for that a little bit later on. But in
general, I was thinking of a research

question to ask which for me is: “Does
internet censorship reduce economic welfare?”

Now, not all of you are economists,
so some of you might think of welfare

more as the transfer payments
that a state gives to its poorer people.

But for economists, economic welfare
is defined as the consumer

and producer surplus. So basically, the
difference between what something costs

and what you can sell it for
is the producer surplus.

The difference between
what you would be willing to pay

and what you’re actually paying
is your consumer surplus.

Now let’s assume I have a laptop
and I bought this.

And I would have been willing to pay
€ 1500 for this laptop because

I think it’s a very good product,
it’s by Lenovo that makes good laptops.

But actually I got it for like €800
or €900. That would mean

my personal consumer surplus
is something like €600 or €700.

And if we add up everyone’s
individual consumer surplus

we get the economic welfare surplus.

So first, I was trying to figure out
what does the internet mean

for the economy. And I’ve said that there
is really no good agreement on that.

Now, a very crude measure that I found is
how much does "the Internet economy"

contribute to GDP?
Now, what is "the internet economy"?

It wasn’t very clear in the research
that I’ve read. It seems to be sort of

online retail, and possibly some other
internet-enabled services?

Possibly but not necessarily
internet advertisement revenue

is reflected in this. But because it was
BCG, which is a big consulting agency

that basically published this research
they weren’t very diligent about

their methods, basically.
So we can see, well it seems that the UK

has a pretty big part of internet economy
as part of GDP.

That’s probably mostly because of
online retail which is bigger in the UK

than in most other countries we look at.
And we see that there is

a small difference between
developed and developing market averages

when looking only at the G20 countries.
But this seems like a very

dissatisfactory answer because first
of all, I don’t know the methods,

so I can’t really say
whether this is actually good.

And secondly, GDP is actually
not a good measure

for what we are trying to measure because
a lot of the stuff that the internet creates,

a lot of the value the internet creates
isn’t captured by GDP at all.

One example is free online courses.
Most of the online courses you can take

on the web are actually free.
And most of them are not ad-enabled.

So most of them don’t really have
advertisements in the general sense.

So classical economics basically says:
“Well, they don’t really create any value.”

But if you’ve ever taken
one of these online courses,

and maybe you’ve been lucky
and took a good one

you would actually… I would say that
some of the courses I took,

they created some value for me.
So one of the ways to look at this

is actually to think about time as
something that has opportunity cost.

So if I’m spending my time doing this
online course I’m not spending it

e.g. earning money. I’m also not
spending it doing something leisurely

that is fun for me.
And these guys, Brynjolfsson

– I’m sorry I don’t know
how to pronounce it exactly,

he sounds Swedish, possibly –
and ohh, in 2012

they tried to get an idea of
how much consumer surplus

these online courses actually create.
Which isn’t at all

reflected in the GDP.
And you see that in some models

it would be 5% of GDP
for these online courses alone.

Even if we take their more... most conservative
model which is $4.18 billion

on average for the years 2008-2011,

that’s still a pretty significant chunk
of economic welfare

that’s somehow being created
that is not reflected in GDP

because GDP is only stuff
that you actually pay money for.

Another example that we
might think of is Wikipedia.

Now Wikipedia has a certain cost of
operating: obviously the servers and stuff.

But because most people contributing
to Wikipedia are actually volunteers

the cost of operating
does not really reflect

the true value Wikipedia creates.
And one of the…

even if you don’t want to say…
even if you don’t agree

that time has opportunity cost, what
about the money that you don’t spend

on encyclopedias? How many of you guys
have encyclopedias at home?

OK, that’s more than I expected!

How many of you guys have
recent encyclopedias at home?

That’s a little less, this is kind of more
what I was expecting.

And now, my family also… we also have
an encyclopedia at home.

I think it’s from 1985 or something.
And before this encyclopedia

we would regularly update an encyclopedia,
we would regularly go out and buy

a new encyclopedia because
knowledge changed, obviously.

But ever since probably 1990,
we just didn’t bother.

So, assuming an encyclopedia might,
like a physical book, might cost €100.

And assuming sort of 2/3
of all households in Germany

have had an encyclopedia at one point.

We’re looking at 13 million households
at this point.

Now you don’t buy an encyclopedia
every year but you might buy it

every ten years. So in order to simplify
this we can say, every year

1.3 million households buy
an encyclopedia on average.

1.3 million times €100,
so we’re at €130 million

of economic welfare, of something that
people were willing to spend money for

that they’re not spending money for anymore
because of Wikipedia, because now that

we have Wikipedia most of the encyclopedias
aren’t actually useful for us anymore

because the knowledge that we have,
the knowledge that they would have

would be outdated very, very soon and
Wikipedia tends to be more up to date.

Well, that was from the consumer’s side.
But what about the business side?

There’s a lot of research on whether the
internet actually increases productivity

for businesses or not. Well, I don’t really
want to go into that debate because

it’s a really long tedious debate that is
kind of focused on “Well, you did this

method wrong”, or “You did this wrong”,
and “Well, I don’t think your argument

makes sense”. So it’s very… I don’t like
this kind of debate. I really like to go

deeper in things. But one of the things
that I found was that a lot of businesses

do rely on the internet by now. Now
we can see on this graph that most firms,

overall about 70% of firms actually
use the email to communicate.

Now email obviously only works
if you have internet, so they need

some sort of access to internet in order
for their current business model to work.

Now this was just some short ideas on
sort of what can the internet mean for

the economy. And now I want to talk about
Internet censorship, just a little bit.

Now, I’m not a censorship expert. I’m just
someone who read a lot of papers about it,

and who was very interested in what kind
of effects this has beyond sort of

the obvious “people don’t have access
to political information”.

So first a definition. ‘Internet censorship’
is the controller suppression

of what can be accessed, published
or viewed on the Internet

enacted by regulators or on their own
initiative. Now, in trying to conceptualize

internet censorship, for me, personally,
there’s two dimensions that are

very important. One is how targeted
is this internet censorship?

Now, you could, in theory, basically
have internet censorship

that is very, very targeted,
which you see in some cases.

Or you can have censorship
that isn’t targeted at all, like in Egypt.

They just decided to close the internet
down, basically, for a day.

That isn’t very targeted censorship,
obviously. The other thing to look at

is how widespread is it? So if you are
a business or if you’re a normal consumer

how probable is it that you would come (?)
something that’s censored?

Now, obviously, if you’re in China it’s
a lot more probable that you would

try to access something that’s censored
than if you’re in Germany. Even though

Germany also does some censorship.
And the way I like to conceptualize it is

to be kind of on a continuum. So I don’t
look… I don’t say “Well, either

there’s censorship or there isn’t
censorship”. What I’m trying to say is

“Censorship has a big spectrum
of things that can happen”.

These are some types of Internet censorship
that have different sort of implications.

I don’t want to go through them in detail
because I think we’ve heard some really

interesting talks on Internet censorship
already. But this is kind of

interesting or important for the model
that I’m trying to build.

But before trying to build my model,
first some more motivation.

I was trying to look at “is there any
evidence that it would have

an economic impact?”. And there actually
is a study that’s conducted by sort of

lobbying organizations, so obviously
should be taken with a grain of salt.

But it is quite interesting, and it shows
that there seems to be a correlation

between freedom and how good
the economic impact of internet is.

This is just a simple correlation. You can
see that there’s a really good line

going through it. They did do some
controlling for GDP per capita, so

for development level. But it still seems
quite rudimentary, to be honest.

The data that they use is quite bad
because it is very, very…

it’s just not finally granular enough, and
a lot of it is kind of… someone rating…

so “How do you think the economic…”,
“How do you think Internet

impacts the economy in this country?”
And then this is the data that they use,

to some degree. So it seemed very…
it didn’t really seem like a good, final answer.

So I’m trying to set up my own model.
And in my model I have a government

that chooses the type of censorship. And
for this type of censorship that it chooses

it pays a cost. Because we all know
censorship can be very expensive.

And in my model for now the only type of
expenses that I calculate are actual

manpower and technology expenses. I don’t
calculate reputation expenses at this point.

There is… there are firms in n industries.
Now this n is kind of not a fixed number

but instead is a number that can fluctuate
depending on the kind of country

I’m trying to model. And these industries
distinguish themselves by their

information intensity, or what I like
to call ‘information intensity’. Basically

I look at information as a commodity.
And what I’m trying to decide, or

the way I distinguish different kinds of
industry is how important is information

as a commodity, as opposed to other kinds
of commodities that are important

for this industry. So let’s look at
information intensity equals Zero.

Like if we don’t really… if information
as a commodity really isn’t important,

especially sort of conveyed information,
transmitted information. We can

think of traditional agriculture. Now
I know today’s agriculture tends to be

large-scale, and there’s a lot of
technology involved. But if you look at

very traditional agriculture that we
still might see happening in some parts

of Africa there usually is very, very
little information transmission involved.

And most of the information transmission
that is involved is actually mostly through

word of mouth. So that would be a case of
information intensity of very close to Zero.

And then if we look at information intensity
of 1 where basically the internet is

the most… or information is the most
important commodity. Internet businesses

themselves would… obviously qualify here,
– sorry – like, let’s look at Facebook

and other kinds of businesses like this.
And in between we have sort of industrial

companies in the modern world.
Now if we’re closer to the Zero end

of the spectrum we might be
at 0.2 .. 0.3, something like this,

we might be in traditional garment
factories. They do have information needs,

they get their cuts and stuff from the
Internet by now, or by email.

But once they have them they basically stay
the same for a couple of weeks or months.

So there’s a very low information
requirement. On the other side,

closer to 0.8 or something
like that we have high-tech,

especially software manufacturing,
so to speak. Information and being able

to transmit this information is very
important. Now, in between we might look

at traditional industrial companies
like automobile manufacturing

that might be somewhere in between.
And before the game, or before…

or at the first run of the model
‘service level’ and ‘globalization level’

are randomly distributed. The information
intensity of industries is also kind of

randomly distributed, but not in a true
random fashion. Because when looking

in the wild, sort of what kind of
economies exist, most of them…

the information intensity of one
industry is kind of correlated with

information intensities of other industries
in this country. Like in Germany

we’re very known for a certain type
of industry that we have quite a lot of,

which is manufacturing, very high-technology
manufacturing. So we have more industries

in this area but we have less traditional
agriculture, for example.

So having a true random distribution
wouldn’t work. In addition the service level

and the globalization level are randomly
distributed as kind of external variables.

Obviously, this is a simplification because
I can’t really start at the beginning like

I can’t say: “Oh well, I’ll start,
I don’t know, 2000 BC

with a very blank economy, and then
something happens and something happens

and something happens”. That’s just not
realistic. So in order to get a better idea

of what happens with different types of
economies, what I’m doing is I’m running

this game or this model again and again.
And having these random parameters

basically changed everytime.
So on average there should be…

there should be usable results.

Now what this is actually missing
is the consumer as a labourer.

So I don’t really have ‘labour’ reflected
in here. A more complete model would have

that reflected. But it’s not the most
interesting aspect of my model, so

I’m not presenting this here, basically.

Now, let’s look at what this would
mean for firms. In my model

what kind of things would I expect
thinking through it logically which is

always the first step when trying to model
something. First of all if we have

an information intensity of something
greater than Zero but smaller than One.

Because the information intensity being
close to One is kind of a special case

that I’ll be talking about later on.
Internet censorship increases the cost

and uncertainty of information.
And of course that is more important

the more important information is
for this certain industry.

So for a traditional garment factory
internet censorship might be a lot

less important than for a semiconductor
factory that has to receive

new blueprints every day or every month
or something. The second thing is

the more globalized the economy as a whole
is the more costly internet censorship

will be. Similar reasoning.

And another thing for firms is the
less focused the censorship

the higher the cost. Now this assumes that
the censorship or the goal of censorship

usually isn’t to turn down firms or to
make sure that firms don’t succeed.

So if censorship is very focused
firms tend to be affected less

which makes their associated cost less.
Now of course we can argue, well,

firms can circumvent censorship, and they
can do that for sure. But it is expensive

to do that. If you’ve ever tried a VPN
in China e.g., first, buying the VPN

is expensive. Then, having someone sort of
make sure that the VPN works is expensive,

every couple of months you need to change
it because the Chinese Government decides,

well, this VPN shouldn’t work anymore. So
it’s a very expensive and uncertain thing,

really. For firms in
‘information intensity = 1’

it obviously also increases the cost
of operating. Some of these firms actually

carry out some censorship for governments.
We have seen that happening more recently.

But there might actually be some firms
that have a relative advantage, especially

domestic firms often have a relative
advantage due to the censorship because

they know the regulators better, they know
how to deal with it, they might have

less need to circumvent, actually.
And even if they do need to circumvent

it’s easier for them because
they speak the language etc.

This is actually a special case that I’ll
be talking about a little bit later as well.

For the government – I’ve said
that censorship is costly. But moreover,

the more targeted and accurate censorship
is the more manpower and technology intensive

it actually is. This is a finding by
Leberknight et al. in a research paper.

I think they’re electrical engineers, and
they calculated through different types

of censorships and how expensive it would
be to scale them up. So that is actually

a really interesting finding because
it shows that for governments

having sort of less targeted censorship
is less costly. But this is the kind of

censorship that is actually most affecting
in a negative way to firms,

in an economy. So that’s kind of not
a result that we would really want

because the incentives don’t line up in
that way. And economists love to talk

about incentives, obviously. Now for
consumers, they would obviously get

less benefits through the internet, the
benefits that I’ve talked about before.

And also businesses often pass on the cost
to consumers.

Now however, some countries
still benefit from internet censorship.

I’ve talked mostly
about why it’s costly to do it,

and I think it is costly in most cases.

But developing countries that start out at
low service and low globalization levels

usually have… in these kind of situations
internet censorship has less of an impact,

less of a negative impact.
And censorship can actually act

as protectionism. In information intensive
industries governments can use this kind

of censorship to push domestic industries
and enable catch-up growth. Now there

are a couple of further prerequisites.
First of all, the country needs to be

large enough so that these
information intensive industries

have a domestic market as well.

Obviously. And then also only
targeted censorship can serve as

protectionism. The only other way would be
if you decided on a domestic intranet and

basically closed your entire intranet off
to the world. Which is kind of difficult.

But what about the long-term effects
of that? Would they still be positive

for the government? Now, I’m using
‘positive’ in a very… sort of something

that should be taken with a grain of salt,
obviously. And what I did is I looked

at China. Obviously, I’m a China watcher.
So I’m really interested in China. And

this is kind of where my interest started.
And I’m really trying to find a framework

where China isn’t the exception but
instead China kind of fits into the model.

What we see is the Chinese government has
outsourced much if its censorship to these

internet companies. Baidu, Sina weibo,
Tencent probably would not exist by now,

actually, if the censorship didn’t exist.
And what we actually see now is that

WeChat e.g. is going global. It has
more functionality than Whatsapp

and they’re trying to get out. But as I’ll
be talking about later on a little bit

the censorship is starting to be a problem
for these companies that used to benefit.

There’s some things about Chinese… about
the character of Chinese Internet censorship

that is relevant here. But what about
the future? Now first it’s difficult to

innovate with this kind of censorship. And
this kind of insular education that we see

also makes innovation, real innovation,
very difficult. In China e.g. Github

is blocked most of the time. That makes
kind of collaborating, especially in

coding environments, very, very hard.

Second, we see more global internet enabled

supply chains in the world. So if we have
these global Internet-enabled supply chains

having internet censorship turns out to be
more of a disadvantage the more globalized

these supply chains actually become. And
information becomes the most important

commodity all throughout China. Now this
of course also makes Internet censorship

more costly for the economy. What about
possible positives? So what could work

in the Chinese government’s favour? First,
the Chinese intranet is actually pretty

attractive to most people. Most people
don’t try to go outside, even like

they don’t even know that they can’t. They
just don’t want to do it. Second, the IoT,

where machines communicate with each other
doesn’t need to be affected because

most of the censorship that we see
happening could be reworked in a way

that doesn’t affect machine-to-machine
communication. And that wouldn’t be

a problem for what the censorship intends
to do which is sort of suppress political

opposition. And a third, the government
wants an economy more focused on domestic

consumption. So if they want to do this
then censorship might actually be good

for that. Now, for me, what I found out
when doing this research is first,

standard economic models really aren’t
suited for this kind of question. Because

they tend to use GDP, and I’ve told you
why GDP really is not a good measure

for that. Second, the next step that
I’ll be doing is agent-based modeling.

But I would really like to feed my models
with some reliable data. And I can’t

really find any of that. I can find some
data going back a couple of years

on, like, is there censorship, is there
no censorship. But I can’t really find any

good data that distinguishes between
different types of censorship, which would

be really important for the kind of
research that I really want to carry out

in the future. Thank you, guys. If you
have questions you can ask now or

you can come to me later, you can
of course also send me an e-mail.

I’m always happy to talk about this topic.

applause

Herald: Thank you very much for this talk.
We have six microphones at the floor level

here, so if you have questions we have
a very brief amount of time.

Please line up at the microphones.
We have microphone no. 2 over here.

Question: I want to mention one thing.
Always when talking about China censorship

this censorship applies to China main
land. So it’s not Hong Kong and not Taiwan.

Toni: Yes.

Question: And my question I want
to ask is:

What do you think about productivity
of work?

So e.g. if you shut down Facebook do you
think this would increase working

productivity?
Toni laughs

applause
Toni: That’s a really interesting question,

and something that I haven’t seen anywhere
in literature. There is a big literature

discussion about what the internet as such
means for productivity, and that’s

kind of both ways. Now, one of the things
to look at is that just because you

shut down Facebook doesn’t mean you
shut down any sort of social network.

And I do think that if people use Facebook
and suddenly aren’t able to use it anymore

they would probably spend their resources
trying to find new ways to access Facebook

which would probably not exactly
improve their productivity.

Herald: Next question
from microphone no. 2.

Question: Would it make sense to have
a model where firms use information

as an input to a production function and
then model censorship as a kind of tax

on that. That will seem like standard new
classical micro-econ one-on-one stuff?

Toni: That would make sense. I’ve actually
looked at this. One of the problems with

doing that is that information
as a commodity

is very difficult to be used in this new
classical way because you usually assume

that everything is kind of friction-less.
And if things are friction-less then

information can’t really be a commodity
because you assume that information

basically gets transferred immediately,
and without any sort of censorship. So

we can talk about this a little bit later.
Maybe you have some ideas that

I haven’t found yet.
It would be interesting.

Herald: And the next question,
as well, from microphone no. 2.

Question: So, going the same direction:
for GDP is rather defined what is

the optimization problem for a government.
For your further approaches what would be

the optimization that a government like
China does then. If you say e.g. Wikipedia

which leaks out to all over the world but
what is the government optimizing then?

Toni: What I’m looking at is economic welfare
as defined as producer and consumer surplus.

And I assume that the government’s goal
is to optimize economic welfare for both

producers, consumers and also for itself
as a producer and as a consumer.

Question: So your criticism is more like
you don’t have a good proxy,

using GDP for economic welfare?

Toni: Yes, yes.
Okay. Thank you.

Herald: I’m afraid we’re all out of time.

Please give a warm round
of applause to Toni!

applause

post-roll music

Subtitles created by c3subtitles.de
in the year 2017. Join, and help us!