A conversation with Erik Brynjolfsson & Jon Bruner
Platform Economies: A Conversation with Erik Brynjolfsson and Jon Bruner
Jon Bruner: Networked, on-demand businesses are transforming and reorganizing firms and labor across the entire economy. How would you characterize the broad movement altogether?
Erik Brynjolfsson: As you digitize and network things, you’re able to take processes, codify them, and then replicate them. That means you often get winner-take-most markets, with huge increases in efficiency and less need for labor input—or capital input, for that matter.
One of the things we see is that sales and profits often look like power laws. There’s a distribution with a small number of really big winners and a long tail of companies or individuals that are filling in niche markets. That’s changing the overall distribution of income a fair amount, especially at the very top.
You also get more Schumpeterian competition in the sense that someone who has a better idea is more easily able to overtake incumbents very quickly, so there’s less stability in market shares.
All of this is something we’ve seen on the Internet and in software and media for some time, but as more and more industries digitize, the same economics are coming to the entire economy.
Let’s walk back through some of these factors. When you mention Schumpeterian competition, are you referring to acceleration of the rate at which newcomers replace incumbents?
Exactly. A big part of it is that there’s more turbulence. Another part of it is that when the winners win, they tend to win big: instead of getting 15 or 20 percent market share, they get 70, 80, or 90 percent market share.
But that’s not necessarily something that the winners can sit back and count on for years to come. If there’s another product or market disruption that comes along and serves consumer needs better, they can lose market share just as rapidly as the incumbents they replaced. So it’s a combination of rapid scaling up and winning big when you win, but also that winners are much more precarious.
Is that a consequence of network effects—that, as new firms gain market share, they gain an advantage in continuing to increase their market share further?
It’s a combination of things. Network effects are a big part of it, and networks tend to lead to power laws. That’s a well-established phenomenon. But digitization is also a big part of it, and digitization brings much lower search costs and enormous economies of scale. Those things together also lead to this rapid scaling up and down.
For instance, lower search costs mean that if a company has a slightly better cost structure and charges a slightly lower price, or if it has a slightly better quality product, then consumers can discover it much more rapidly.
Ignorance and geography, which were two big barriers that protected incumbents in the past, are much less effective now, and the only real protection is having a better value proposition for consumers.
So the Internet is encouraging the rise of large, consolidated national brands that consumers can discover quickly, that aren’t bound to any particular area, that can compete directly with everyone else in the market, and that offer low cost of switching for consumers.
Yes. In particular, geography becomes less and less relevant, in many markets. Instead of having a lot of local retailers with their own little market shares, you end up getting a national or international brand and company that can cut across the world. Uber is operating internationally, for instance. Or in a more mundane way, you see roll-ups of laundromats and restaurant chains. They compete not just on branding, but importantly on supply chains and business process improvements. They get some efficiencies by coordinating the manufacturing and distribution, and that’s increasingly making geography less relevant or even irrelevant.
And this in turn makes it possible to service niche markets as well, by consolidating what would have been impractically small markets into serviceably large ones?
Exactly. If only one customer in a million is interested in your product, it’s hard to sustain your business in a city of just a million people. But if you’re addressing a market of a billion people, now you’ve got a thousand customers, and that’s a niche market that may be worth serving.
This brings to mind plenty of business-to-business services that have found the scale they need to operate efficiently in specialized markets. Digi-Key, for instance, can stock millions of arcane electrical components and ship them overnight to engineers anywhere. Services like that have, in turn, transformed the way that engineers develop electronics.
When you can order what you need wherever you are, you’re enjoying the consolidation of a global marketplace. Adam Smith famously said that specialization is limited by the extent of the market, and this is a perfect example. As the market gets bigger there’s more room for specialization and you get more fine-grained subdivisions in niche markets, both for consumers and for B-to-B. That is yet another way these new markets drive productivity and efficiency gains: when people specialize, they can get better at some component as opposed to trying to be a jack-of-all-trades.
As this effect—niche markets becoming serviceable—comes up against the Schumpeterian competition that you mentioned, are we just going to see the big national brands knock the small companies out whenever they see a new niche grow? What is keeping Amazon.com from entering Digi-Key’s line of business?
Some people like to focus on one end of the power law and other people like to focus on the other end. I’ve written papers about both ends: I’m happy to say I wrote some of the first papers documenting the value of product variety on the Internet. But I’ve also written about the concentration at the other end.
Imagine a power-law curve: there’s value all along that curve. One of the reasons Amazon has been successful is that they have so many more niche products available than the average store in just about any category, but they also make a lot of money up at the head of the curve with big blockbuster winners. It’s a mistake to think of only one part of that phenomenon. There’s money to be made up and down the power law curve. Different people will have different expertise, and people will try things out. Some of the things that you’d think would be a niche product, lo and behold they get discovered and they get really popular.
It’s kind of like “Gangnam Style,” with billions of downloads. In advance you wouldn’t have known that would be a total blockbuster, but the Internet makes it easy to discover and test all sorts of possible new products and services.
So the rise of new businesses depends on the willingness of some people to take different risks and explore different models that the incumbents aren’t using?
That’s exactly right. This is an economy where people take risks, and it’s hard for any of them to know in advance which one of them is going to hit the jackpot. I think entrepreneurs and investors increasingly look at strategies almost like buying lottery tickets. They try out different things and let the market speak and then adjust rapidly.
Another big phenomenon that’s part of this is that the cost of experimentation has gotten so much lower. Rather than spend months or years carefully planning out the exact perfect product or offering, it’s often a good strategy to test the market with a number of different variants relatively early, and if you have a flexible enough business model and production process, you can respond to what the market’s telling you and scale up the parts that are in demand.
That’s the big difference between bits and atoms: bits are so much more flexible and so much easier to modify. The traditional business approach with lots and lots of planning made sense in the 20th century because there were huge costs to setting things up, but now you can modify things much more quickly.
Going back to the long tail versus the head of the tail, are the firms that operate in the long tail just structured differently from the ones at the head of the curve, so that the firms serving the long tail can do so effectively in ways that the big ones can’t?
I don’t know that they’re structured differently; there are so many opportunities out there that there’s no one entrepreneur—no matter how brilliant they are—who’s going to think of all of them, so this is a job for crowdsourcing. Creativity is one of the things that machines are not very good at. This is a job for a few billion brains on planet earth to work on, with each of us thinking of some product or service that may create value and then going out and testing them. Some of us are luckier and others of us are less lucky in being able to come up with the ones that scale up and are valuable, but it’s just a matter of having a marketplace now where it’s much easier for people to try things.
Every startup’s founding mythology seems to include an episode where the founders are summoned to Steve Jobs’s office and told that they should sell or else be crushed when he builds the same product. How do these companies survive? Obviously they do, and they end up providing services that beat the incumbents.
Some do! You know, I can’t give blanket advice to always turn down that offer from the big company. I’ve known a few companies that did turn it down and things didn’t work out, so there’s no guarantee that you’re going to make it on your own.
It’s hard to know in advance, and that’s one of the reasons that venture capitalists invest in a portfolio of companies knowing that some of them are going to make it and some aren’t, and they do their best to filter in the ones that they think are more likely to succeed. We talked about “Gangnam Style,” but an even more extreme one was “Flappy Bird.” That’s not necessarily the highest quality game but something about it just caught on.
There’s art as well as science, or luck as well as skill.
And none of those ever would have made it through some sort of formal project proposal process inside a large company. So, an individual was willing to take a risk to create these things—a different kind of risk than a large firm was willing to take, and the marketplace embraced it.
Exactly. So we now have a global platform where it’s relatively easy for people to throw stuff out there and reach potential consumers without going through the filters of bureaucracy or senior management. The rules are different in that kind of a world, and companies that embrace that approach will have more opportunities to build the big hits—or, for that matter, the niche products.
Speaking of individual creators, let’s go to your observations on labor input and capital input and how this is changing that equation. There are some obvious technological advances that reduce the need for startup capital. In other places, like AirBnB or Uber, you still need startup capital—you’re buying a car or outfitting a venue that you’re renting. Do these platforms just shift the capital requirements from large companies onto individuals who have to raise and apply this capital?
No. It’s not just a zero-sum shift-around. There may be a little of that going on, but the main phenomenon is a real, genuine reduction. There are two ways that’s happening. One is the move from atoms to bits, and businesses based more on bits are immensely cheaper to scale up. There are lots of situations where the bits have become relatively more important.
But the other thing is that as transaction costs and search costs get lower, you’re able to take your existing capital and deploy it much more efficiently. In the case of Uber, a lot of those cars were sitting in someone’s driveway 90% of the time, not creating value for anybody. So Uber is tapping into a capital asset that was underutilized. Ditto with AirBnB; because of lower search and transaction costs, it’s easier to find value-creating matches where the benefit to the user is greater than the cost to the provider.
Those matches have always existed, but it was just too expensive to find them. Somebody would have needed to spend thousands of hours interviewing people on what their needs were and what rooms were available. But by having a digital platform, the costs get much lower so you get more of those matches. Even with the same amount of capital, you get more capital services.
So, I don’t think this is just a shift-around. It’s genuine; it’s creating more capital services out of the same stock of physical capital.
One of the reasons Uber and AirBnB have been successful is that they have very effective algorithms that replace human intuition and heuristics for pricing and matching supply and demand. But incumbents could also improve their algorithms. What is the fundamental difference between AirBnB and, say, a Starwood that has an excellent pricing algorithm and prices hotel rooms dynamically?
There are lots of differences, but I would focus on the way AirBnB taps into unused capital assets: there are apparently large numbers of rooms, apartments, houses sitting idle that have a higher-value use. Without building new hotels or new hotel rooms, you can use the existing capital stock, but in order to do that you have to be able to match knowledge.
Friedrich Hayek wrote a great paper called “The Use of Knowledge in Society” back in 1945, arguing that the most important problem an economy has to solve is reconciling an idiosyncratic knowledge of values: what my value for a room is on Saturday, and what your capacity to rent a room is on that Saturday. That information tends to be widely dispersed; you can’t just summarize it with one statistic because the statistic is meaningless. What you need is the actual detailed knowledge of which person and which room, which location, which time.
Technology, specifically information technology, has dropped by orders of magnitude the costs of matching those two kinds of knowledge, those two kinds of information, and AirBnB is one of the many companies that have taken advantage of that.
So AirBnB’s value is in matching capacity to demand that could never have been matched efficiently before.
Exactly. It’s substituting knowledge and lower search costs for physical capacity. In the old model, there were empty rooms sitting around out there, but you’d build an extra hotel room because you didn’t know where those empty rooms were. Building a physical room out of bricks and mortar is pretty costly. Identifying an unused room that already exists used to be even more costly than building a new room, but now it’s cheaper.
The reason we know it used to be even more costly than building a room is that Starwood is successful, and has been successful for decades. It was just too expensive to find empty rooms that already existed, so it was cheaper to build a whole new room.
Plus, Starwood’s brands solve another important aspect of the information problem.
When I say “find a room,” I’m using that as shorthand for a much more complicated process, which is finding a room that’s clean and reliable, where an axe murderer doesn’t live, where your money won’t be stolen. You really have to come up with a reputation system and a large bundle of knowledge. It’s not enough to just have the room. And Starwood’s branding was one way of offering assurance on all those dimensions.
Which AirBnB also offers through a new kind of reputation system.
Exactly. Another important breakthrough was being able to keep track of people’s ratings and reputations and eliminate some of the information asymmetry that would cause markets to break down.
Another classic paper by George Akerlof talked about how you can have a complete market breakdown when there’s a lot of information asymmetry. The rating system and other aspects of technology have reduced that information asymmetry and allowed markets to thrive where they previously broke down.
These platforms have also entailed changes in the way that people sell their labor. Perhaps you were an employee of a car service and now you’re a self-employed entrepreneur who markets your driving services through Uber. So far, does the evidence suggest that labor is doing better or worse under the new system?
Ultimately, what employees will make is less a function of the system per se and more a matter of the market’s overall demand for that set of skills. Take Uber drivers; if Uber tried to substantially raise the fee that they pay to drivers, that would draw in a lot more people who want to be drivers, and they would on average spend a lot more time driving around waiting for customers. So their equilibrium wage wouldn’t go up by much, if at all.
An example of that is at San Francisco’s airport: it’s a great place to pick up customers because they tend to have a fairly long ride up to San Francisco. Uber drivers are smart enough to know that, so they go to the airport and wait a substantial amount of time to get those rides. Their hourly wage ends up being not much different than it would be if they were driving around elsewhere. When there’s a higher amount paid, then more people enter until there’s a new equilibrium and it equilibrates with what they could get elsewhere.
So, I think that the wages earned by Uber drivers are more a function of what’s happening in the broader economy, and the demand for people with that set of skills and what their alternatives are, and less a matter of Uber-specific policies.
To the extent to which we have an economy that creates more work for people with that skill set in lots of different ways—not just through Uber, but through all the other businesses as well—then we’ll drive up the wages of those kinds of people. To the extent to which those jobs get automated and eliminated in the economy more broadly, then every company will tend to pay less for those kinds of workers.
Does it appear that these workers are being compensated for the added risk of having to buy their own car and go into on-demand business for themselves?
By and large, they’re making rational decisions based on their alternatives. Some of them are probably making some mistakes, but on average they’re freely choosing whether to do that job versus some other job, so I would defer to their judgment: they think it’s a better alternative than whatever their next better alternative is.