MGI Research

Forward Thinking on measuring the value of the digital age with Avinash Collis

In this episode of the McKinsey Global Institute’s Forward Thinking podcast, co-host Michael Chui talks with Avinash Collis, professor of digital economy at The University of Texas at Austin, and Digital Fellow at Stanford and MIT. He covers topics including:

  • One of the greatest inventions of the 20th century
  • How to run economics experiments to figure out how much digital services are worth
  • How much value digital services provide that GDP doesn’t capture

An edited transcript of this episode follows. Subscribe to the series on Apple Podcasts, Amazon Music, Google Podcasts, Spotify, Stitcher, or wherever you get your podcasts.

Podcast transcript

Michael Chui (co-host): Janet, have you ever been paid not to do something?

Janet Bush (co-host): Wow, I wish I had. But, Michael, you really do ask some interesting questions.

Michael Chui: It’s what we do! We’re podcast hosts.

Janet Bush: Well, no, actually I can’t think of a time when I was paid not to do something.

Michael Chui: Well, believe it or not, today’s guest has won awards for doing just that—for paying people not to use social media, web search and other digital services. We had a previous guest, Diane Coyle, who talked about how difficult it is to measure the value of digital products and services when they are provided for free. Avi Collis, with whom I spoke, talks about paying people not to use digital services to measure how much they’re actually worth to people, and what that means for the economy.

Janet Bush: Fascinating. Applied economics, I guess. I really want to hear more about this.

Michael Chui: Avinash Collis is a professor at the McCombs School of Business at The University of Texas at Austin. Avi, welcome to the podcast.

Avinash (Avi) Collis: Hi, Michael. Thanks for having me on the podcast.

Michael Chui: I’d love to start with your story. Where did you grow up? And what did you study in school? How did you end up becoming a professor at Austin?

Avinash (Avi) Collis: I grew up in Hyderabad, India. And growing up in the ’90s in Hyderabad, I saw firsthand how important information technology was in transforming society and the economy. The city went from a pretty average city in India to one of the most important cities not only in India, but across the world.

That really made me inspired to study computer science in college in India, so that’s what I did. And then afterwards, I got a master’s degree in France and Germany when I switched to studying economies and more specifically using economics to study technology.

During my master’s, I realized that there was also this whole world within business schools where you do research. Business schools of course teach MBA students, but on the side, the professors at business schools also do research, where they get to answer cool questions.

That got me interested in research in business academia. And because I was interested in the intersection of economics and technology, I applied and got into MIT to go and study with Professor Erik Brynjolfsson, who is now at Stanford, and who was at MIT back then.

I got my PhD at MIT, and my dissertation research was on measuring all the value which we get from free digital goods, like Facebook and Google and Wikipedia, etcetera. So that’s what I did during my PhD, and then I graduated in 2020. Now I’m a professor at UT Austin.

Michael Chui: Super exciting story, and I definitely want to get into some of the research that you’ve been doing. But you glided over a little piece of your history there. If I recall correctly you had a computer science degree, I think from IIT Delhi, right, a very prestigious institution? And then France, Germany, and econ. A lot of things happened there. What happened?

Avinash (Avi) Collis: Yeah, so in my college there was an opportunity to do a study abroad semester, and I spent some time in France. I really liked the country, the language, and the food, and basically everything. And there was an opportunity at the school where I was studying to do a double degree with a university in Germany. I did not have much of an econ background before my master’s, because I was studying engineering, so this particular school in Germany, in Mannheim, had a pretty strong economics department. So I basically did like a double degree between my school in Paris and in Mannheim.

Michael Chui: Why the subject switch to econ? How did that happen?

Avinash (Avi) Collis: What I realized is, as an engineer, you focus all your energy on developing a specific part of a product. Let’s say the “buy now” button on Amazon.

It’s a very important piece of the platform, and there’s a lot of engineering which goes into building it. But in that process, with some of my internship experience, what I realized is I’m more interested in the bigger picture rather than the specific engineering challenges.

And then around that time, I also read The Second Machine Age, by [Andrew McAfee and] Erik Brynjolfsson, who later ended up being my PhD adviser. And that book, I must say, was one of the books which inspired me to learn and study economics and use that to study technology. That’s how it ended up happening.

Michael Chui: It’s very dangerous, what one reads.

Avinash (Avi) Collis: Yeah.

Michael Chui: Speaking of professional career, one of the pieces of research for which you’re best known is innovation on GDP as a way to measure the economy. A lot of our listeners have heard about GDP forever, and it’s such an important metric. When it’s growing, things seem to be good. When it’s shrinking, there’s a recession. And it could feel like this GDP metric has been around forever, since Roman times or something like that. But it’s actually surprisingly recent.

Number one, what is GDP? And who invented it? What is this thing?

Avinash (Avi) Collis: First of all, before I go into my own research, I definitely want to clarify that I’m a big fan of GDP. It is called one of the greatest innovations of the 20th century, because it’s been instrumental in how we think about the economy.

GDP, or at least the way we think of GDP now in modern times, was created by the economist Simon Kuznets in the 1930s. It’s actually around 100 years old now, I guess. This was basically created after the war, when they wanted to have a systematic way of measuring the overall production within a country or within an economy.

The simple definition of GDP is basically that it measures the market value of all the final goods produced in the economy, all the things which we buy and sell on the market, which we consume, and not intermediary goods, which are used to produce other types of goods, but the final goods.

This is an age-old problem. GDP measures production really well. It doesn’t measure well-being: how well we are from various things which we are consuming. For most of the 20th century, the things which we were consuming were physical goods, so the production and well-being were a little bit at least correlated with each other.

Let’s say you double your consumption of a particular type of food, or you go and buy a new car, so you contribute towards GDP and also your well-being also increases, because now you have access to a new good.

But in the 21st century, if you use a free app on your phone, and if we increase the usage of this app, then that doesn’t really show up in GDP—but it increases our well-being at the same time. So this disconnect between GDP and well-being gets a little bit challenging now, in the digital age, and that’s what set up the foundation of some of my dissertation research.

The fundamental issue, I guess, is digital products have zero marginal cost. It doesn’t take any money to create an extra copy of, let’s say, a digital file. Take a piece of music, for example. You save it as an MP3 file. You make hundreds of copies of it. You can do that for free. So the marginal cost is zero, which means that in a competitive market, prices of these digital goods oftentimes end up being zero as well, which is why we see online platforms like Facebook and Google Search and Wikipedia and Twitter—all of these are free to consumers.

You don’t pay a price to use them. And that’s when it becomes challenging.

Let’s say you show an ad and then you advertise a certain product. You click on that ad, go and buy that product, and that would show up in GDP. But how much time you spend on that app—how much time you spend on Facebook or how many searches you do on Google, and how that impacts your overall well-being—that part of the equation is not captured in GDP.

And that’s a big chunk of what we do every day. We spend so much time online. This podcast we’re recording right now, I don’t know how much you paid for it, but I didn’t pay anything to download and use it.

And it’s creating a lot of value, because I’m enjoying chatting with you, but our conversation right now, it’s not showing up anywhere in the economy. It doesn’t show up in GDP, or it doesn’t show up anywhere, but it’s clearly increasing my well-being, and I hope it’s increasing your well-being—you are also enjoying it.

That’s the disconnect between GDP and well-being, especially when we look at digital goods.

Michael Chui: Are consumers getting more value out of these things that they’re not paying for?

Avinash (Avi) Collis: Definitely. A good thought experiment for that would be: imagine a life without Google Search. Imagine you had to live without being able to search online. I cannot do my job, and I’m guessing most knowledge workers cannot do their jobs.

In addition to that, it’s not like we remember a long URL or something and go and type it in the browser directly. Google Search is the first step to going anywhere online.

We don’t get a lot of value from using these apps, just by looking at how much time people spend. There is an interesting study done by Hal Varian and others at Google in collaboration with, I think, some economists at Michigan. It was a randomized field experiment, where what they did is, they had a treatment group and a control group, and they gave both of these groups a bunch of questions. And they had to go and find answers to these questions.

The control group had to go into a physical library, search for whatever book they want to, and then find answers to some questions. The treatment group could go and search them on Google. So you can directly quantify how much time Google is saving, in terms of finding answers to the most popular things people search for.

What they found is Google Search saves a ton of time. And if you convert that time into dollars, it ends up being a very big amount. These products are definitely very important to our everyday life.

What we do is, we run online experiments where we pay people to stop using some of these products, pay people to stop using Facebook or Google or Wikipedia or WhatsApp or some of these apps, and see how much do we have to pay someone so that they stop using a particular product. And that is basically how much they value that product or how much consumer surplus they get from it.

Michael Chui: Wait, wait, wait. How does this work? How do you pay somebody not to use one of these?

Avinash (Avi) Collis: There’s no market price for it. You don’t go and pay something in the market, so if you want to put a dollar number on valuation of these goods, we need to think of other ways. In the past, as I mentioned, this Google study, people are trying to look at time spent on various products and services.

But time spent is not really a good proxy for things. It’s hard to convert that into dollars. Just to give you an example, if a search engine is not that great, you’ll spend more time finding the answer to your question. That doesn’t mean you value that more. Less time is better for search. Time spent wasn’t a good proxy, so we wanted to come up with a new way of measuring this. This is what I come up with, which is running experiments online.

Just to give an example, the first experiment we did was with Facebook, paying people to stop using Facebook. We get a random sample of the US online population by partnering with survey companies like YouGov, or running ads on Google, for example, and recruiting people into a study.

That’s how we get samples. And then with Facebook, for example, what I found in doing my PhD was, you can monitor when someone is online on Facebook. If you add someone as a friend on Facebook, you can see when they’re last online. We pay someone to stop using Facebook and check that they haven’t actually used Facebook for, let’s say, a week or a month, and at the end of the period, once they have complied with whatever they said they would, they get the cash. If and only if they comply.

If they have broken the contract and logged in, we see that, because we’ve been tracking them, so we see that, and then they don’t get the cash. This is how we run our online experiments.

Michael Chui: How much did people have to be paid?

Avinash (Avi) Collis: That’s a great question. When we first started running these experiments in 2016, ’17, the median American required around $40 to $50 to give up Facebook for a month. That’s for the median person. There was around 10 to 20 percent of the sample who were not willing to give up, no matter what we gave them.

We offered up to $1,000, and still around 10 to 20 percent were not willing to give up. And we were really very surprised by this. We wanted to make sure they understand the question, so we asked them.

It seems like they get the question, but some people just cannot live without social networks, because that’s the only way they stay in touch with some contacts. And interestingly, with Facebook, when we dig deeper into the data, what we find is—it’s probably not surprising now, but back then, it was surprising to us, which is, it’s actually the older people who value Facebook much more than the younger people.

The reason was older people did not have other alternatives, like other substitutes, to stay in touch with, let’s say, their grandkids or the rest of the family members. Younger people would readily give up Facebook because already back then, they had Instagram and other apps, and now they have TikTok and even more apps. So there were definitely some interesting demographic differences there.

Michael Chui: I’m curious—you mentioned before that a lot of these platforms have business models predicated on advertising. Advertising does show up in GDP. Isn’t advertising good enough? Does that not correspond with the value that these consumers are getting from using these systems?

Avinash (Avi) Collis: Actually, there is research done by Bill Nordhaus, who won the Nobel Prize, and what he finds is, if you look at how much money platforms actually capture from all the value they create, it’s only around 2 to 3 percent of the total value they create.

If you look at ad revenues, platforms like Wikipedia, for example, zero ads, so whatever value it’s creating doesn’t get captured anywhere. It’s all purely consumer surplus. If you take something like Facebook, in the US, I think Facebook now makes like $120 to $130 on average, per year, but globally, it’s more like—I might be wrong, but at least as of a few years back, the Facebook average revenue per user globally was more like $20 to $30, I think, or lower.

But if you look at how much value people get from Facebook, from our calculation, it’s more like $500 to $600. So people get much more value from using these digital goods, compared to how much money these firms can capture through ads.

The underlying mechanism—there’s no law which says that how clever a platform is at showing ads and making you click, that’s not really correlated with how much value you get from that platform itself.

You can have a pretty average platform, but with very interesting ads, so you might click more on the ads. Or you might have a great platform but might not show enough ads or not good ads, so there’s no correlation between how much ad revenue they can make and how much consumer surplus a platform generates.

Michael Chui: What other interesting things did you find? You looked across different platforms. You started this work a half decade ago?

Avinash (Avi) Collis: Yeah. Some interesting results include looking at instant messaging. In the US—again, I’m from India, so for me, it’s shocking, but maybe it’s not for you, but most Americans still use text, SMS, to message people. That was interesting to me, because I thought WhatsApp would have a very high valuation, but in the US, not that many people use WhatsApp. The people who do use WhatsApp really value it a lot. In our sample, it’s almost double that of Facebook, those who do use WhatsApp.

But globally, what we found is WhatsApp is super valuable, especially in European countries. One study we did in the Netherlands, the median person required €400 to €500 to give up WhatsApp for a month. It’s a really big number.

We did several follow-up interviews, and what we found is WhatsApp is essential in Europe and in many countries, including India, not only for personal reasons, like staying in touch with friends and family, but also just professional reasons. You actually get work done over WhatsApp in many countries. You text your coworkers on WhatsApp, and you stay in touch. Let’s say you text your babysitter on WhatsApp. It’s super essential to survive. So that was another interesting result we found.

One hypothesis, which I need to test more carefully, what I find is many of these digital goods are creating a lot of value not just in rich Western countries, but in many countries across the world.

Take India, for example. Lots of people use WhatsApp. Lots of people use Facebook. They also spend a lot of time on these apps. They get the same experience as someone in the US or in Europe in terms of the app features, etcetera.

So the consumer surplus lower-income people are getting, compared to the higher-income people, it’s pretty comparable. I think if you look at the welfare side of the equation, not just the production side, looking at GDP numbers, but also the welfare side, looking at consumer surplus, my hypothesis is actually these digital technologies are creating a lot of welfare, across the board, both richer countries and poorer countries, higher-income people and lower-income people. And some of the difference and inequality, if you look at welfare, might actually be lower, given how much value these goods are creating to people across the world.

That is something I need to test it more carefully, but we do observe this, comparing, let’s say, the US and European countries. If you look at raw GDP numbers, some of the countries we study in Europe, their GDP, on average, it’s around half that of the US, but the value they’re getting from WhatsApp and Facebook, it’s comparable to the US or even more for the average person.

Michael Chui: This is a bit of a tricky technical question, but I am curious. I know sometimes people describe consumer surplus as the “willingness to pay.” And what you measure is the willingness to be paid to not use something. Does that actually mean the same thing as “I would pay 400 or 500 bucks per month in order to have access to a service”?

Avinash (Avi) Collis: That’s a good question. When we ask the reverse question, like, “How much would you pay for access to Facebook,” I’m pretty sure Facebook must’ve done experiments internally as well on that question.

And the answer for the median person is zero, right? Because the average person, there are lots of substitutes out there to being—most people use Facebook to stay in touch with their social network. But they’re on multiple platforms at the same time, to stay in touch with their network. So if you start charging for a product, and as I mentioned earlier, these are all digital goods with zero marginal cost, so in a competitive market, prices would go down to zero.

It’s very hard to charge a subscription fee, which is why we don’t see that happening in the market, as well. So in our case, we look at willingness to accept instead of willingness to pay. Like willingness to accept, to give up a particular good—and you’re right, the raw numbers are not comparable to willingness-to-pay figures, but what we are more interested in is changes in these numbers over time.

Instead of looking at the levels, we look at changes over time. So with Facebook, that’s what we find. Consumer surplus from Facebook, it’s falling over time. Consumer surplus from Instagram, it’s increasing over time. And from consumer surplus from search engines, we also see it’s increasing over time. We can look at deltas, changes over time, rather than the absolute magnitudes.

Michael Chui: What does that mean for that number to be increasing over time or decreasing over time? How do I interpret that?

Avinash (Avi) Collis: If the consumer surplus increases over time, that means people are better off. So that’s a better proxy for measuring well-being. If consumer surplus from you being able to use Facebook, it’s increasing over time, that means you’re overall better off. And you can use that as a measure for well-being.

Having said that, this is all still within the standard economic framework. As you know from basic economics, there are also some negative externalities, which are not captured in such measures. And one aspect is, of course, digital addiction.

Some of these digital apps might be addictive. In terms of dollars, consumer surplus might be a big positive number, but some of these apps might not be good for you when you take the addiction into account.

Some of these digital apps might be addictive. In terms of dollars, consumer surplus might be a big positive number, but some of these apps might not be good for you when you take the addiction into account.

In our approach, we’re still staying within the standard neoclassical economics framework, measuring consumer surplus, measuring changes over time. But in some parallel research, I’m also trying to measure other aspects of well-being, including your overall happiness or subjective well-being.

It’s a little bit challenging, because the best way to do that is still basically asking the survey question, some version of a question asking how happy you are on a scale of one to ten.

It’s by definition subjective, because it’s hard to quantify that. But it’s still better than nothing. The main takeaway from all this research, the objective is not to come up with one number, which is substituting GDP. That’s not the objective.

What I and what we are trying to say is, GDP does a really good job at measuring production, and then our measure measures consumer surplus while still staying within the standard neoclassical framework. It looks at well-being, it looks at economic well-being, and then, of course, we also need to do these happiness surveys to try to measure subjective well-being.

The idea is instead of looking at one number, if you’re a policy maker, for example, instead of just looking at one number, you need to be looking at this dashboard of metrics and what my adviser Erik calls—if you look at, when you’re driving in your car, you don’t have one number.

You have a dashboard with several metrics, and you observe all of them. You observe how much gas you have. You observe what the speed limit is, and all that, to come up with your decision. In a similar way, policy makers need to be looking at a range of metrics.

Michael Chui: But that said, you have created a dial, which you’re calling “GDP-B,” is that right?

Avinash (Avi) Collis: Yes, exactly. GDP-B, where the B stands for “benefits,” that is built based on these surveys we have been doing, trying to measure the consumer surplus from various digital goods.

And in ongoing work, we are expanding that work, not only to digital goods but also other types of goods. By definition, like any type of good you consume, GDP only looks at how much price you pay for it, but the consumer surplus measure would look at how much extra value you get, beyond what you pay for a good. That part is not captured in GDP, but that part is what we’re trying to capture in the GDP-B metric.

Michael Chui: So how much bigger is GDP-B, when you include these consumer surplus measures, versus GDP?

Avinash (Avi) Collis: This is an ongoing research agenda, and we are trying to scale up our measurement, but just looking at Facebook, for example, a single digital good, Facebook, what we find is accounting for consumer surplus from Facebook alone would increase GDP-B, not GDP but the GDP-B metric, by around 0.02 to 0.05 percentage points every year.

That’s a pretty significant number, given that it’s just one app. If we account for all of the digital goods, including search engines and email and all that, it would be a pretty big number. There are some estimates out there, including some estimates we have put out, where if you add up all the consumer surplus from various digital goods, it ends up being around $20,000 to $30,000 for the median person, every year.

The median American’s income is around $50,000, so around half of that would be on top of the GDP metric, which is pure welfare coming from the digital goods.

Michael Chui: But that suggests, if I’m not mistaken, that just looking at the consumer surplus of digital goods, your GDP-B metric potentially could be 1.4 times as large as GDP without the B?

Avinash (Avi) Collis: Yes. Exactly. That’s correct. Having said that, we are measuring these metrics, starting now and into the future, but if you extrapolate that into the past, some of the surplus from these digital goods is probably transferred from other types of goods in the past.

For example, we don’t call people as much anymore, because we just message them on Facebook or on WhatsApp or whatever. So I wouldn’t say that this is all newly created welfare from digital goods.

Some of that has existed in the past in other forms, which has shifted from physical to digital. Actually, a really good example of that is the encyclopedia industry. You used to spend thousands of dollars to buy various volumes of Britannica. And now you get Wikipedia for free.

There, actually, it’s a good example, because GDP would actually go down, because you don’t spend money anymore, but your overall well-being increases. Whatever welfare or consumer surplus Wikipedia is creating, that’s pure welfare, on top of what GDP was even capturing, because GDP would have actually gone down.

Michael Chui: So in the perfect case of the new set of national accounts, you would actually have to account the net of the substitutes—the things that some of the digital services substituted for—and make sure that that all, see how that all netted out?

Avinash (Avi) Collis: And you can do that in an easier way by doing this—let’s say, doing our surveys annually, and looking at changes over time, and that would account for some of these substitution effects.

Michael Chui: Is anybody doing that?

Avinash (Avi) Collis: We are trying to do it, but we are just one research team, and so we’re trying to raise some money, scale it up. I have presented this work at the US BEA, but I also presented it at the IMF, Bank of England, and some other national banks across the world.

We’re trying to popularize this approach. And trying to get some adoption, but I think there is still a lot of work before national statistical agencies can adopt it. So we are trying to do all the research to help make it easier for others to adopt it.

Having said that, most national statistical agencies, including the US BEA, they are severely underfunded, understaffed. To create or to keep track of another metric on top of what they already do, that would require resources. That would depend on if they can get enough funding or not.

Michael Chui: And I’m curious, you touched on this a moment ago, but you focused a little bit on digital products and services, but there are a lot of nondigital products and services. Have you used these types of techniques to estimate the consumer surplus associated with those? Breakfast cereals, for instance, is one that a lot of economists talk about.

Avinash (Avi) Collis: It’s great you bring up breakfast cereal, because as a benchmark, we actually did a study; we actually asked people, “How much would we have to pay you to stop eating breakfast cereal for a year?” What we found is the median person I think asked for around $40 to $50 to stop eating breakfast cereal for a year.

But if you look at how much money they spend on breakfast cereal, I think the median person also spends around $20 to $25 or something. Once you account for these numbers, like the consumer surplus estimate and how much you pay for it, it ends up being very proportional to each other. But this is for the case of breakfast cereal.

It’s a physical good. It’s a relatively competitive market, so prices and welfare and consumer surplus and willingness to pay in all these metrics, they are proportional to each other. Over time, if you eat more breakfast cereal, you spend more on it, but also your consumer surplus increases proportionally.

What we found was for things like breakfast cereal, GDP does a pretty good job at measuring production, and you can use that as a good proxy also for consumer surplus.

But for other types of nondigital goods like healthcare, those are big open questions where we are trying to do studies now. Healthcare is one market. Another thing is just environmental goods. Like how much better off you are or worse off you are, compared to the past year, in terms of all the changes which have happened in the environment. Right now, we are trying to do studies to measure those other types of nonmarket goods as well.

Michael Chui: This is really interesting to me. Some of the research that we’ve been doing, in the sense that the amount of surplus, and not just consumer surplus, but call it “customer surplus,” too, in a B2B context, we’ve often found to be multiples of what companies or individuals are paying. Particularly for digital goods, but not only; for services as well. And yet your example on cereals is a much lower percentage.

Is that just because cereal companies are far better at capturing the value they’re creating for their customers? Or are there patterns here about what percentage of the overall value that’s being created by a product and service can a company manage to charge for?

Avinash (Avi) Collis: That’s a great question. I don’t have a good answer to it, but yeah, at least it looks like in the case of cereals, the companies are pretty good at capturing around half of the total surplus they are generating, and the other half is going to consumers.

And you’re right; for some other types of goods, that ratio seems to be very different. As I told you earlier, for digital goods and for all types of technological innovations, previous research shows that only around 3 percent goes to the firm, and the rest, 97 percent, goes to consumers. I think these are all interesting open questions. I don’t have an answer to it right now.

Michael Chui: We’re recording this in April of 2022. There’s a lot of discussion about inflation. How do you think about inflation in GDP-B? Do you have a separate type of inflation on the benefit side?

Avinash (Avi) Collis: Again, I don’t have a good answer to that question, but I think one thing we need to remember is of course, like GDP-B and welfare and all this, it’s a good metric. But at the end of the day, you also need food to survive.

You also need basic goods and services, for your survival. And that’s what the CPI basket aims to track—most commonly consumed goods—and try to see changes in prices.

Inflation metrics, I think, are really—you know, we should really worry about them. But yeah, on the other end, in terms of benefits, we saw during the COVID pandemic, the pandemic was a massive shock to the economy, but if you look at the digital side of it, actually, all of us increased our consumption of digital goods. Like most of us overnight switched from in-person meetings to Zoom.

So welfare from digital goods clearly must have increased a lot during the pandemic. But in terms of converting that into some kind of inflation measure, that would be an ambitious question for the future.

Michael Chui: I love the fact that you grounded it in the fact that these economic measures are just measures of things, and we’re people, and we have to eat and we have to connect with others and all of those things. So if you don’t mind, I’d love to do a lightning round of quick questions, quick answers.

Avinash (Avi) Collis: Yeah.

Michael Chui: Other than your own, what is your favorite metric in economics?

Avinash (Avi) Collis: I would pick productivity as my favorite metric. As I said in our talk before, with all the caveats, it doesn’t measure digital well, but I think at what it is measuring, it does a good job. Just staying within the production framework, it’s doing a great job. So I would pick productivity.

Michael Chui: MGI loves productivity as well. What’s your least favorite metric in economics?

Avinash (Avi) Collis: I can see positives and negatives to various metrics. It’s hard to pick the least favorite, because I do see value in all the metrics I can think of right now.

Michael Chui: What would you add to GDP-B if you were to create a GDP-C? Joking, but what else would you like to measure?

Avinash (Avi) Collis: I think environment is something I would really try to measure well. It’s one of the biggest challenges we’re facing. Questions related to climate change, and we still don’t have good ways of measuring these things.

We have like ad hoc studies focusing on a specific part of the environment and looking at some specific region, but I think environment, and measuring environment well and systematically, across countries, like that is something I would love to do.

Michael Chui: What economics paper do you wish you had written?

Avinash (Avi) Collis: The one paper which comes to my mind, because it’s one of my favorite papers of all time, is this paper by Nick Bloom in India. They wanted to measure if management matters to firms or not.

They went to India, recruited a few hundred companies, and then they did an experiment on companies. They created a randomized control trial. Half the companies got free management consulting from consultants they hired. They just paid for those consultants. The companies got training. The other half did not get training. And then they measure revenues and productivity in various outcomes. And they find a big positive impact of management practices on productivity. I really loved that paper, not only because it’s a cool paper and it’s ambitious, but it also shows that what we teach in business schools actually has some value on firms’ bottom line.

Michael Chui: Which economics paper are you glad you did not write?

Avinash (Avi) Collis: Oh, that’s a tough one. There was a series of papers in the ’80s and early ’90s, basically saying that technology doesn’t matter. Basically, they weren’t seeing any impact on productivity.

Computers started to enter, lots of people started getting computers, but we weren’t still seeing any impact in productivity, and there were a series of papers written showing that IT doesn’t matter. And I’m glad I didn’t write those papers.

And now same with AI, right? There are some papers which say that AI probably doesn’t matter, and I wouldn’t bet on those papers. I think even if you don’t see some impact now, it’s going to come in the next decade or so.

Michael Chui: Yeah, some of that work, famously called the Solow paradox, Bob Solow described it. He actually ended up working with MGI later on, with complementary work to Erik’s, as well, showing that it did actually matter.

What did you learn that surprised you most about the pandemic? Or during the pandemic?

Avinash (Avi) Collis: What surprised me most is most of the decisions made by policy makers like, let’s say, lockdown policies or mandates, or which places to shut down, which places to reopen, and all those decisions. It doesn’t look like they took a data-driven approach to come at these decisions.

And that was really surprising to me, given that this is a pandemic happening in the 21st century. We have pretty rich pieces of data available. I did some research of my own, early on in the pandemic. Like providing a data-driven framework to decide which places should be shut down, which places should be reopened.

Michael Chui: You’re a college professor; what would you recommend that a college student study today?

Avinash (Avi) Collis: I would highly recommend a combination of econ and computer science. And the best place to actually do that is in a business school. I’m biased, but in a business school, we have lots of programming courses. Like we have lots of data analytics courses.

I did a CS undergrad myself, and CS undergrad is great if you want to think about the theory behind CS. But if you actually want to apply it in practice, I think a business school is a great place to study a mix of econ and CS, which is, I think, where most of the jobs moving forward will be.

Michael Chui: I think in VC circles, we call that “talking your book.”

Avinash (Avi) Collis: Yeah.

Michael Chui: What would you be doing if you weren’t a professor?

Avinash (Avi) Collis: I think I would be in some kind of a policy role, either for an organization like the IMF or World Bank or OECD or one of those organizations, or even working for the government. I really enjoy policy-related questions, which I try to answer in my research. So if I wasn’t a professor, I think I would be in some kind of a policy-related role.

Michael Chui: And what’s one piece of advice you’d give to listeners of this podcast?

Avinash (Avi) Collis: One piece of advice I would give is, there is a lot of data out there now, and it’s really fun to go and play with it instead of just consuming. Consuming news and consuming research is nice and fun, but there’s also lots of data out there, and you can go and play with it yourself.

Michael Chui: Avi Collis, thanks so much.

Avinash (Avi) Collis: Thanks, Michael. It was really great to have a chat with you.

Explore a career with us