An Autocrat's Dilemma: Boosting Domestic Business Profits at the Expense of Research Quality

Scott Wallsten (00:01):

Hi, and welcome back to Two Think Minimum, the podcast of the Technology Policy Institute. Today is Tuesday, November 21st, 2023 and I'm Scott Wallsten, President of the Technology Policy Institute. The dictator's dilemma refers to an authoritarian government challenge of trying to use information and communications technology for economic development while also controlling the information available to the population. The biggest country that fits that description is of course China, to talk about that we're excited to have with us Professor Meicen Sun, who has new detailed empirical research on this issue, quantifying the trade-off. Professor Sun is an assistant professor in the School of Information Sciences at the University of Illinois Urbana Champaign. Her research examines the political economy of information, the geopolitics of data and information policy. She served as a fellow on the World Economic Forum's, global Future Council in China, and is an affiliated faculty with the MIT FutureTech. She recently completed her PhD at MIT, followed by a postdoctoral fellowship at Stanford University. She's bilingual in English and Chinese and has also written stories, plays and music and staged many of her works in China, Singapore and the US. Meicen, thank you for joining us today.

Meicen Sun (01:03):

Thank you, Scott. It's a great pleasure.

Scott Wallsten (01:05):

So let's start off just talking about your paper. Give us a quick overview of what you found and why it matters.

Meicen Sun (01:13):

Absolutely. In short, my research is about the impact of internet control in China, but I'm not looking at the political impact of the internet control. That is a topic that has been studied many a time by political scientists, by political communication scholars. Instead, I am actually looking at the impact of internet control on productivity and on innovation. More specifically, what I looked at in this paper is a major wave of internet control shock in China that happened in 2014. What I found was two things. The first thing that I found was that this 2014 internet control shock is associated with a 25% increase in revenue for Chinese domestic data-intensive firms. So firms that employ data as a primary factor production over the Chinese domestic data non-intensive firms. There you can sort of see this at face value, this seems to lend support for claims such as the USTR’s – the allegation that China has been using, for example, internet control/ internet censorship as a form of trade barrier, but just in a digital context.

Of course I got curious and I extended this analysis to the US-China context. So I brought in basically the US firms during the same period of time and I extended this analysis. And surprisingly or not, the result actually did not translate to the bilateral context. If anything, the Chinese data-intensive firms do not seem to have done better than their US counterparts during this time period. If anything, they actually seem to have done a little worse. What this suggested to me in turn was that there are powerful countervailing forces at play apart from this beneficial effect for the Chinese data-intensive firms. And that gets to the second thing that I found with my research where I looked at the impact of the internet control shock on not on productivity, not on firm performance, but on innovation, more specifically on research output.

There I discovered that the same 2014 internet control shock in China has incurred 10-13% reduction in research quality for Chinese researchers, conditional on how dependent their discipline is on internet access, basically in producing knowledge. And again, I extended, this time, this analysis to the US-China context and this time it did translate. So again, the 2014 internet control shock seems to be associated with up to 23-24% decline in research quality for Chinese researchers vis-a-vis/relative to their US counterparts regardless of what discipline they are in. So holding constant the discipline, there is a 23-24% marginal reduction in research quality, which I think is a huge one.

Scott Wallsten (04:23):

So before we go on to talk a little more about the substance of it, tell us a little bit about this internet control shock. What do you mean by that? What was it? I think most people know about the great firewall of China, but they may know exactly what you're referring to.

Meicen Sun (04:40):

Yeah, so note that the Great Firewall, so to speak, has varied tremendously in its strength over time. Sometimes you would see a lot of websites being blocked very strictly; at other times you seem to have a period of relaxation, for example. So to get at this variability over time, the first thing that I did was that I mined the data from this website called greatfire.org. What they do is that they have been using physical servers located within mainland China to routinely test thousands and thousands of websites, domains, their accessibility from within China on a pretty regular basis. I mined their data,which literally tested the accessibility of all these multitude of domains over time. What I found was that there was, around end of May, early June, 2014, lots of websites across the board were all of a sudden subject to a very major wave of throttling/disruption that lasted for almost two years.

Scott Wallsten (05:54):

What was the cause for that then? Why did they do that then?

Meicen Sun (05:57):

Presumably it was – the timing of it – was right before the 25th anniversary of the Tiananmen Square incident. So that sort of attests to the primary motivation of this being a political event, or exogeneity to market and to innovation if we're interested in the econometrics.

Scott Wallsten (06:19):

Some of us are very interested in the econometrics and some people not so much, but explain just a little bit about why this change and what we call a shock, why that matters to your research.

Meicen Sun (06:33):

So to understand the impact of this, why we should care about this internet control, we really, I think, have to first go back to understand what internet control is about. And ultimately I think we need to understand internet control as a form of information flow restriction. That gets to the question of “what is there in the flow of information?” We can think of information essentially as containing many different things. There are many different things that are bound up in the same flow of information. And for our purpose here, I think there are at least two things that we can distinguish. There is the flow of data as a primary factor of production for the so-called data-driven industries that employ data, for example, for training their core algorithms that underlie their core products and services from which they derive a major stream of their revenue.

So that's one thing. But then in the same information flow, you also have ideas, right? The flow of ideas, particularly the politically charged ideas that remain to be the primary motivation for sovereign governments to impose this kind of information/internet control for maintaining, for example, domestic stability. But then there's also the flow of knowledge that is the main driver for innovation for domestic researchers such as, for example, a piece of technical know-how, a piece of cutting edge research. These remain to be a principal driver for innovation. So to the extent that both data and knowledge are bound up in this information flow, when a government imposes an information/internet control shock, such as the one we witnessed in China in 2014, at least two things should happen.

Meicen Sun (08:37):

The first thing is that this constitutes a form of knowledge flow restriction. What happens there is that the domestic knowledge-intensive workers such as ourselves, we'd like to think – basically people who rely on the access to the existing stock of knowledge in producing their own, now they cannot access the same stock of knowledge from the outside world to the same extent as before. I'm not saying that they cannot access anything, but what I'm saying is that the rate, it's the rate of such access that is going to make all the difference. To put this more concretely, imagine you go to a country under internet control, you are used to browsing, let's say five, 10 articles an hour, easy. And now you can only access, because Google Scholar is blocked for example, or if not blocked, throttled let's say. And so are the journal, your favorite journal websites, and then your rate of this access drops to, let's say, one to two articles an hour if you're lucky.And that slowed rate of knowledge access in turn leads to your slowed rate of knowledge production as a result. That's the knowledge side of the story. So this is a form of knowledge barrier, this internet control. 

Now we can see it as a form of knowledge barrier, but then on the data side, what happens there is quite different. What happens there is that now, the domestic consumers, they cannot access the foreign digital products and services at the same rate as before. Again, to use an example, you go to one of these countries, you find that your favorite apps are blocked/throttled. So what am I going to do? To the extent that an indigenous equivalent of these apps and websites, to the extent that they exist, you are going to switch to these domestic substitutes of these foreign digital products and services. And this should then lead to both an increase in the sales revenue as well as a supply of data as a raw material for these domestic data-intensive sectors. And there we have, I think, a pretty familiar trade barrier story just in a digital context.

Scott Wallsten (10:48):

Do you view this as a short run benefit, long-term cost kind of situation?

Meicen Sun (10:53):

I would say so just based on what I found. I think it is a pretty straightforward – if we look at the timelines – I think it's a pretty straightforward story, where on the knowledge barrier side, because what I found was that everybody loses, everybody loses from this internet control shock as a form of knowledge barrier.

Meicen Sun (11:22):

Everybody loses. It's just a matter of extent. Whereas the benefits accrued to these domestic data-intensive firms, they tend to be more immediate. So yes, to your question, I think the answer is yes.

Scott Wallsten (11:41):

At what point now, so now we're getting into sort of more speculative areas. What do you think the relevant timeline is for sort of seeing the effects on innovation? I mean you're talking about the effects of something that happened in 2014, which is almost 10 years ago. Well I guess the effects happened less than that, but do you think we would expect to all see some kind of slowdown in innovation as a result of that by now? Would they see the effects or is it later? Is it going to be so on the margin we can't really detect it? What do you think?

Meicen Sun (12:18):

Well, if you ask me, I think the recent signs that we have been seeing about the Chinese economy, I think they actually have had no small part to do with this very thing that has been happening in the infosphere, in the Chinese information system where you have this big blockade, big sort of blockage of information flowing in and out of the country. I think that actually  has a lot to do with the sort of downturn trajectories that we've seen about the Chinese economy. The thing is, I think the irony there is we've all heard about the double circulation strategy on the part of China that was very much triggered by the US sanctions against China a few years ago. I think the irony is that, sure, there's now this internal circulation. That strategy is ultimately about self-reliance. It's self-reliance on the internal market over the external market. It's about the internal circulation of capital. But I think long before that, since long before that, at least 2014 if not even earlier than that, we've had this internal circulation of information and of knowledge that has been going on the whole time. And that's got to have some kind of impact given just the magnitude of this impact across the board. We're not talking about, recall what I just said, we are not talking about just some restricted, some small number of fields. We're really talking about a negative impact across the board that's got to have some kind of impact in some form or fashion down the road. And I think what we are seeing now is really part and parcel of that effect playing out.

Scott Wallsten (14:06):

Do you think that there's some risk of it turning into a self-perpetuating cycle where the leaders see the economy slowing and so then they think, well we have to boost our domestic industry, although not that they've been so friendly to their own domestic companies and by restricting information flow even more, which would give you the short-term benefit, but again exacerbating the long-term innovation cost?

Meicen Sun (14:34):

Yeah, I definitely see a possibility in that. I definitely think that is one of the likely scenarios. Number one, the first thing is there could be a lot of ground for complacency on the part of the government over there. Why? Because overall, actually China's innovation output has been on the rise, very steady rise, particularly if we look at certain fields in AI that are very much data-fueled. There's a huge data endowment domestically within China. That I think has, you can see, offset some of this decline in the capacity for innovation as the result of this decline/decreased access to external knowledge. So the planners could look to this positive trajectory and think, okay, that's pretty good. Any additional advantage we would gain from relaxing internet control is probably not going to be worth the lessened degree of political control that we are going to have as a result of this. I think that's a very plausible argument that they're going to make. I think, on the whole, the state is aware of this detriment on innovation in a very, very generic kind of sense because otherwise we would not have seen episodes where, for example, websites like GitHub, they actually got unblocked following protests of the coders and programmers. The government has been made aware of this drag on innovation from a measure like internet control many a time. But the question of whether they will take some kind of corrective measure here and now, whether they would be sufficiently incentivized to take a corrective measure, that's a completely different question. And there are at least three pretty important reasons as to why they would not be incentivized to do something about it, despite what somebody looking from the outside would think that okay, they would absolutely want to do something to correct this.

So first is a matter of timeline. If we look at the timeline for this kind of innovation loss from a decreased access to external knowledge to materialize basically into monetary terms, we're almost necessarily talking about a longer timeline than, for example, the benefits accrued to these domestic digital technology firms on the one hand, and more importantly I think, the timeline for the domestic threats to domestic instability that remain to be the primary motivation for the government to impose this kind of control in the first place. So when weighed – it's not that the innovation premise is not important – but you're just going to weigh it against these much more kind of pressing concerns and threats. And unfortunately in this context that often has to take a backseat. 

The second factor here is we can think in terms of who the winners and losers are from this internet control. So we know who the winners are, right? They are the technology firms, the digital technology data-intensive firms, many of which are quite powerful financially and politically, and also the political rulers themselves who stand to benefit from this added degree of control for the purpose of maintaining their regime security and domestic stability. Well the losers are – who are we talking about? We're talking about researchers, scholars, professors, students, all in all a pretty powerless group who also tend to be more diffused both in terms of them being scattered around the country at research institutions and universities and also in terms of their interests. So there you have a classic kind of collective action conundrum where you have this asymmetry in the ability to exert some kind of influence over the political decision-making process of the small group of leaders. 

And the last factor here is, okay, now let's say even if the government is fully aware of this detrimental effect and it's super willing to do something about it, and they're also fully on board with my findings on this differential impacts on different academic fields. The thing is, what are they going to do? How are they going to do it? Because essentially in order to take a corrective action to compensate for this kind of loss in innovation capacity as a result of this knowledge access barrier, you would be attempting to arrive at a dollar amount per unit decline in knowledge production due to the decline in knowledge access for each of the knowledge domains. So that's already a mouthful for me. Imagine how complex and inefficient a process that is going to be in practice when you're actually trying to implement something, a corrective measure of sorts. Because as you can imagine, the default policy instrument there would be to increase R&D spending, but differentially – differentially across different academic fields. Even for a simple tariff for a specific good in a well-defined sector, if you are just trying to transfer from the domestic producers to compensate the domestic consumers, that already tends to get very inefficient very quickly, let alone something as complex, nuanced and dynamic as something as trying to compensate the researchers for their loss in knowledge access, which also is, it varies from sector, from field to field. And also it's rapidly changing too. It's a highly dynamic situation.

Scott Wallsten (20:54):

So the way you're describing it, the firms are generally on the side of or generally agree with the government, maybe not for the same reasons, but they appreciate that protectionist aspect of it, and the losers of course are the researchers and people nobody cares about. But are firms that clear that firms are on the side of less information coming into the country? So many of them rely on flows of information here in the West. Broadly speaking, firms are very much against data localization practices. So why would that be different in China or is it just that, the short-term profit implications outweigh any of that? And in particular, since I think we're required by law now to talk about AI in every single conversation, that's an extremely data intensive field. And so firms working on AI, do they feel differently from other areas?

Meicen Sun (21:53):

Yeah, so I think you hit the nail on the head, I think the firms’ attitude toward this kind of protectionist barrier of internet control, I think they are not at all clear-cut. Because I think you are absolutely right that firms also lose as innovation players, right? Firms, sure, they are producers of digital products and services, but at the same time, in order to do that, they also have to innovate most of all. And I think they are certainly also bearing the brunt from this kind of information flow restriction, generally less so than academics because they tend to be, I mean in general, better positioned to take advantage of resources to circumvent this internet control compared to a professor at a regular university, for example. So the source of their benefit remains primarily to be from the data dividend from the domestic consumers who are made to switch to their products because the foreign products and services are blocked.That's the trade protection story that we just told earlier. 

The second caveat I would throw in is the state and the so-called tech giants, they actually don't have the best kind of relationship over the years as if the 2021 crackdown has not reminded us enough, even though I considered them to be the winners of the story. I think it's a tenuous one. Do not think at the end of the day they have that much sway over the government, over the decision to, whether or not to impose internet censorship. I still think internet censorship/internet control in that country remains to be primarily about maintaining domestic stability. I think in the past few years, the firms from the perspective of this data dividend that they were able to acquire made them kind of welcome this as a nice byproduct. And the same thing for the government. It's very much an afterthought as opposed to, oh, let's impose these nationwide throttling campaigns just so that our own Chinese big tech can forge ahead. I do not think that is what was going on.

Scott Wallsten (24:25):

Well, so even if in China the reasons for doing it are different, it seems like it would play out the same way here and you would see the same, I mean in principle, it could be profitable in the short run to restrict information in any country, right? Do you think we're seeing any of that today in other countries outside of China? 

Meicen Sun (24:47):

India. India is the country that certainly comes to mind because India has been blocking, in their context, in their case, it takes a different form because how they do it is primarily through the blocking of apps, mobile apps on mobile phones. Also taking into account how the vast majority of Indian internet users actually access the internet through their mobile devices as opposed to desktops. So we are talking about hundreds and hundreds of apps, mostly from China, but there are also other apps, in the last two or three years. And that's a huge number. You have reports about how this blocking pattern seems to have corresponded with, for example, how the Indian firms, the Indian technology firms have been doing. You could definitely potentially make a trade protectionist story there. The bottom line really is we tend to think of this as the dictator's dilemma between letting in too much information, risking political discontent versus letting in too little information, risking economic stagnation. But I think in the digital age, the dynamics that we just laid out between, this trade-off between this more immediate benefit, the more immediate benefit from data dividend versus the longer term, but subtler kind of detriment from this decreased access to knowledge, it really could apply to everybody. And I think that is a sort of admonition to all states in the world, regardless, essentially nobody is really immune because this dynamic could play out as readily in a democracy as in an autocracy.

Scott Wallsten (26:40):

Right? It's interesting that India is the first country, is the country that you think of since because of their long history of import substitution, this sort of seems like another reincarnation of that same approach, which has not exactly been successful in the past.

Meicen Sun (26:55):

And I think in the current case we're talking about, I think one reinforcing factor in India's case that would make this strategy particularly appealing is to think in terms of the data endowment of India as a country. India, first of all, it has a huge population, but I think equally critically, even more importantly, I would say in the past decade, India has seen a vast increase in internet connectivity for its population. And that I think makes a huge difference because this whole import substitution story in the digital context that we just told, it depends critically on the number of users, the concentration of domestic users, and how much of a data dividend that you could plausibly acquire by blocking out foreign digital products and services. I think that's a key piece to the puzzle.

Scott Wallsten (27:53):

Do you think it's related to how globalized a firm's network of consumers and suppliers is, I mean because the incentives as they're laid out, like you said could apply anywhere. And so you could imagine an administration here saying this could be really good to juice the economy and the firm saying, yeah, that sounds good to us too, but we aren't, well at least on the firm side, we aren't seeing that. And does that have to do with the way that their businesses are spread around the world, do you think?

Meicen Sun (28:27):

I absolutely think that is a factor too, because in the Chinese case, right, China really, I mean if you are speaking from the perspective of the state planners over there, they really sort of lucked out in terms of timing. I think just as important as this internet barrier, we also have to take into account things like language barrier for foreign digital products and services to have a reasonable market share in China. I think the language barrier really has played a role that oftentimes gets underestimated versus if you want to break into an equally large market that is predominantly English speaking, I think you would have a very different story there potentially. And in terms of the US case, the US now remains to be the biggest exporter of these digital goods and services. And again, that also makes things quite different. If we are talking about whether a country is, where they are on the supply chain and whether they are predominantly importing or exporting these digital goods, I think that the story is going to look quite different.

Scott Wallsten (30:04):

So I mean this is all related to digital trade and recently the US trade representative's office announced that it would no longer oppose data localization rules and we don't know what's going to happen with that because they haven't talked about it anymore since. And people in Congress across parties are upset about it. But how would that relate to your results and what would you say to policymakers who are thinking about the potential effects of such rules?

Meicen Sun (30:35):

Yeah, I think we probably are aware that domestic concerns over, for example, regulation of the so-called big tech and things like AI regulation and the potential implications, things like algorithmic accountability. These concerns have been the driver for this recent reversal or retreat on the part of the USTR. Obviously, there's been a lot of talk about this being an emboldening move for China and emboldening for countries looking to adopt China's approach to internet policy, to this kind of trade barrier in the digital space. I definitely see where they're coming from and I think there is, that's a legitimate line of argument to be made. On the other hand, I think this could also be a little overblown –  this threat, this threat that the China model is going to take over， China is going to be all over the place with its Digital Silk Road strategy that is going to see even less resistance among countries in the world, now that the pressure from the US is gone. 

The thing is we have seen, I think that doesn't have all to do with the pressure from the US. We have seen in the recent years, especially in the most recent year or two, countries have been, a lot of countries, also different countries on the political spectrum. You have democracies, but also autocracies that are relaxing their data localization restrictions. We're talking about, for example, India has been relaxing its data local localization restrictions and so has Saudi Arabia very recently in the very recent months and of all countries, China has taken pretty serious steps in backtracking on its data localization/cross-border data flow restrictions. And the thing is, I think one simple explanation is countries have now, several years in, they have now realized and seen the very real material detriment, very real drag on their own economy from imposing something as cumbersome as these measures. And naturally I think they're just doing something out of self-interest. And that's probably good news for the US because sure, I think the US pressure in some ways may have catalyzed some of these moves, but I think even on their own, I think countries, they are recognizing how these measures may be already shooting themselves in the foot.

Scott Wallsten (33:29):

So the US past advocacy for free trade and data will offset the US' attempt to shoot itself in the foot. Is that what you're saying?

Meicen Sun (33:40):

We can write a research paper on that. That would be a quantitative study.

Scott Wallsten (33:47):

So we're almost out of time, but before we go, how did you decide to work on this particular topic? I mean, you've been studying now for a long time and brought you to it.

Meicen Sun (33:57):

Sure. Funny enough, I was, the impetus for this current study that finally materialized into my dissertation, it actually came from my experience around the 2010s when I was in China doing fieldwork on a completely unrelated topic, on a completely different topic. And there, of course, I immediately noticed something different about the internet. Something was off with the internet connection there. So on the very micro level, on a very individual level, two things that struck me immediately was, basically literally what I eventually hypothesized as the two key mechanisms. One is the substitution effect where when something was blocked, when a foreign website was blocked, people who would otherwise not use a domestic app, they now would have to download, for example, WeChat or whatever app there was. So that's the first thing. 

And the second thing is this, like I said, this decrease in the rate of knowledge access.The example that I gave about browsing articles online, that very much came from my personal experience. And my thought then was if this is how much of a drag that it is having on my productivity as one small researcher’s experience, how that is going to aggregate on a population of billions, we've got to see some kind of effect on the aggregate. Even if we do not see it now, eventually at some point, maybe sooner than later, we are going to see something materializing. And of course, when I first attempted to propose this as a dissertation topic, there was quite a bit of resistance, I'll just put it mildly. And that was 2016, 2017. It wasn't that long ago.

Scott Wallsten (35:58):

That's really interesting. Why did they resist? That seems like even at the time, it must have seemed like a great topic.

Meicen Sun (36:07):

It's a very positional kind of argument: This is our identity and we just don't do something like this. I think the years since – first, misinformation campaigns and the role of digital platforms that became squarely a political science topic. And by now, also, this digital trade  barrier has finally, I think, become a legitimate line of inquiry – I think after too long if you ask me. But I think that just goes to show the kind of inertia that often unfortunately takes place in academia and how entrenched, I think, academics can be about prevailing theories, about sticking to the theories. We can only make incremental progress and we have to maybe more so in my fields than in yours, but we have to write a 10-page literature review first of all. Short of that, we can't even vet you as a legitimate scholar. On that point, I have to make a shout-out to my committee members who really took a risk by standing behind me, one of who is Yasheng Huang. My committee members, Professors Kenneth Oye, In Song Kim, and Yasheng Huang from MIT. I really owe so much of what I have accomplished to them, to their courage.

Scott Wallsten (38:37):

Well, and I'm glad you persisted because it's great work that needs to be continued. And one more question. So you mentioned in your bio that you're also an artist, music and plays. And tell me a little bit about that. Did you ever consider that as a career or was it always more of a hobby?

Meicen Sun (39:00):

Yeah, I mean, I am a classically trained musician. I play the piano. But pretty quickly after I got my performance diploma, I just went into composing and that's where when I actually started to have a lot of fun. So I started doing that in China and then in Singapore and then in the US, started writing plays and directing plays and acting in my plays. It just gives me a lot of joy. But theatrical arts, unfortunately, it's very hard to do it on your own as I've come to realize. It's kind of almost hard to make it into a career because when you're in school, you have the venue and the people you can work with, and then the school sort of supports you for everything. But once you're out of school, it becomes really, really hard, which is a bit of a shame because I still feel very connected to that part of me.

It's very much still in me. For example, when I watch movies, I never want to watch movies with other people because the playwright in me would keep rewinding just to critique on the directing just to, if the directing was particularly good, I would just replay that for 20 times and nobody wants to watch movies with me. So I think I'm still very much in touch with that side of me. But I think you are right. I would say especially I think in the initial years of me being a political scientist researcher, I definitely, I was in hiding very much. I did not want to be seen almost as a creative person, as if that would detract from my legitimacy as a researcher – as a positivist kind of empirical researcher – which is no longer the case now. I write that in my bios these days because I think I've come to acknowledge that if it's something that's very salient to my identity, then it should be out there. People should know that about me too.

Scott Wallsten (41:07):

I think creativity is a key part of good research, otherwise, it's nothing interesting.

Meicen Sun (41:13):

I completely agree with you. And it's what catalyzes new combinations – if we think of innovation essentially as new combinations of pieces of information. And I do think me having a sort of creative mindset has been, I think, a huge plus – seeing some connections all of a sudden in the middle of the night just because, whereas if I were to follow a more conventional kind of path or a conventional way of thinking, I don't think almost nearly as many ideas would have come to me – in the middle of the night typically.

Scott Wallsten (41:49):

Yeah. Well this was really great. I really appreciate you taking the time to talk to me and I hope we will talk again soon.

Meicen Sun (41:58):

Thank you, Scott. The pleasure was all mine. Thank you for having me. Thanks.