I have argued in my last post, Value Investing As Software Eats The World, that value investors these days can no longer avoid coming to grips with subtle differences between information-based and physical goods that might affect their moat and valuations. As Hal Varian (now chief economist at Google) and Carl Shapiro explain in their excellent (albeit slightly dated) book, Information Rules: A Strategic Guide to the Network Economy, while the classic principles of economics remain valid in concept, they do have some interesting implications due to the unusual cost structure of information-based goods. In this post I would like to highlight some of these implications that seem to me to be particularly relevant from an investor’s point of view.
Perhaps the most important of these differences is that information goods usually have high fixed costs of initial production but practically zero costs of reproduction. Think of the creation and production of a new song, a book, a movie, or a new software program. All the costs have to incurred upfront, before the first item is sold. But after that initial item is shipped, reproducing it is just a matter of copying the digital bits and transmitting them over the internet, for hardy a few cents per copy these days; and even that low cost is exponentially approaching zero! Since the marginal cost of an information good is nearly zero, age-old economics principles confidently predict that its price will also head towards zero, in the presence of any significant competition. This implies that information goods cannot really be priced on the usual “cost of production plus profit” basis of physical goods — their cost of production is practically zero.
Another unusual aspect is that information goods often have to be experienced in order to be valued. Consider, for instance, how much would you pay for a news article? If it turns out that you already knew the news in it, it no value to you; but you cannot know this until you have actually read it! Similarly, a piece of music has to be heard before one can properly put a value to it. Producers of information goods have to solve this conundrum in pricing; this often requires giving away some version of it for free and investing heavily in building a reputation for quality. The high quality of past editions of, say, The New York Times is the reason we agree to pay for today’s newspaper, even before knowing whether it contains anything of value.
A more consequential implication of near-zero variable costs is that an information-based products can be reproduced in billions of units with very little additional investment: a successful product can scale to massive volumes with surprising speed and efficiency. Value investors of the old school, used to the slow and steady expansion needed for scaling up physical goods, are often blindsided and tend to “miss” moated companies such as Google and Facebook, as they go from huge losses (due to high fixed costs of producing the first item) to billions in profits within a matter of just a few years (due to minimal costs of scaling up).
Zero incremental costs also explains the differential pricing often found in information-based products, where the same core product is sold at vastly different prices in multiple versions with superficial variations, including free versions. This abundance of digital copies must, however, deal with the harsh reality of that ultimate scarcity — human attention. Metrics related to the consequent “battle of eyeballs” are often used as a proxy for competitive advantage at the early stages of a new digital product. Such metrics are often scorned by traditional value investors, who prefer to measure actual cash earnings, but I think they are missing the point; at their early stages of production, metrics of attention capture are perfectly rational measures of value for information-based goods, due to the unusual economics of networks that applies to them.
The economics of networks
Notice that information-based goods have to interconnect with many complementary products in order to function: a browser has to work with an operating system; a smartphone with a mobile carrier; an app with an app store; a TV show with a cable network, etc. Such interconnections eventually lead to entire networks; and once a product is entrenched into the fabric of a customer’s network, it is very difficult to persuade that customer to switch to a newer product, even if it is quite a bit better or cheaper than the older one. These high costs of switching due to networks creates a strong moat against competition, once an information product has achieved a critical mass of acceptance.
The implication is clear: to really understand the information economy, it is necessary to master the economics of networks.
Think of buying the very first fax machine – it was useless by itself because it could talk to no other fax machine, but as soon as other businesses also purchased fax machines, it became more and more valuable. Similarly, the first user of Facebook had nobody to communicate with, but the value of the Facebook network goes up with every additional Facebook user that joins — the benefit to the nth user rises in proportion to (n-1), the number of previous users. The inventor of the Ethernet, Robert Metcalf was one of the first people to point out a mathematical implication of this: the total value of a network is proportional to the square of the number of connected users in the network [n*(n-1)]. This suggests that, if all other things are equal, as Facebook grew its user base 10 times, its value grew not 10 but 100 times! (Of course all things are not really equal – a user in a poor economy is not as valuable as a user from a rich one; nevertheless, the value of a network increases to very large numbers surprisingly quickly.)
The term “network” does not mean that wires or wireless connections have to exist between the nodes. The term is used metaphorically whenever there are interdependencies between nodes, which is almost always for information-based goods. For example, as more and more office workers adopted Microsoft Word in the 1990s as their editor of choice, the more valuable it became for a new user to adopt it, and the value of the “Word network” kept rising faster than the size of the installed base. Network effects from office workers dependent on Word, Excel, and PowerPoint explains why even a fading Microsoft still commands over 300 billion in market value. Any challenger product for processing a document or spreadsheet immediately runs into the problem that no one else in the company can work with it; it is extremely difficult to persuade everyone to move simultaneously to a new kind of document or spreadsheet editor, even if it is somewhat better than Word or Excel – a classic example of the powerful moat created by network effects.
Traditional value investors repeatedly underestimate this powerful arithmetic of networks. The market value of such companies can grow exponentially even as their P/E ratio shrinks arithmetically; they can change from apparently insanely expensive multiples to quite reasonable valuations all too quickly.
Tipping points and positive feedback loops – the dynamics of networks
Producers of physical goods usually face diminishing returns from the physical and bureaucratic costs of being too big. The demand for such a product usually is not too influenced by the number of previously sold products. Information-based goods, on the other hand, usually have decreasing supply costs as their fixed costs are amortized over larger volumes; in addition they benefit from increasing demand for every incremental user, due to the network effects noted earlier. These double-barreled positive feedback loops can create runaway winners due to the extreme economies of scale involved. As noted by Geoffrey Moore in his prescient book, Inside The Tornado, such companies have an unusual growth pattern: they grow slowly at first before reaching a “tipping point” of acceptance; products that cross this point can show explosive growth afterwards. The demand for such products is also subject to self-fulfilling loops: a product that is expected to win often obtains more and more followers, following a herd-like dynamic of adoption.
This economics of such increasing returns to scale, have long been discussed but only recently have elegant models developed by economists such as Paul Krugman, Paul Romer, and Brian Arthur, enabled them to be tackled mathematically. Taken to extreme, increasing returns can lead to a winner-takes-all monopoly for some products, famously Microsoft’s Windows and Intel’s x86 chips; Facebook’s social network is showing similar adoption dynamics these days. A generative model for how such networks grow has been described by the physicist Albert-Laszlo Barabasi in his insightful book, Linked: The New Science of Networks. In Barabasi’s “preferential attachment” model, even a slight preference of a new node to join one of the bigger hubs of an existing network causes the big hubs to get even bigger. Such a growth process can explain many of the “power laws” often found in a variety of network-like phenomenon, from the size of cities to the popularity of blogs. Barabasi also found that a second factor, intrinsic quality, can explain the occasional exception to this “big get bigger” pattern, provided that the challenging product’s quality is an order of magnitude better than the existing product. A similar notion was first posited by Intel’s legendary CEO, Andy Grove, who observed that a new technology that is, say, only 10% better often fails to make a dent to a well established incumbent, but one that is 10 times better stands a good chance.
In light of such non-linear growth dynamics, it is easy to see why information-based companies spend so heavily in their early stages, sometimes even giving away their initial product in order to establish the impression that they have already won. Growing to critical mass quickly is often crucial to success, monetization can wait; an established network of users is much easier to monetize later on. Clear examples of this strategy are Google and Facebook: both patiently waited for many years before monetizing their growing base of users, and were quickly and hugely profitable once they turned on their monetization engines. Similarly, new products such as Twitter, Instagram, and Whatsapp are still growing their user networks, deliberately delaying monetization for now. While many value investors roll their eyes at the absence of current profits, the delay strategies are rational due to the strange economics of networks.
Invading the network moat
If there are such strong positive feedback loops as I have described above, how could Facebook topple Myspace, the earlier social network? In general, how can a challenger win in the face of the tremendous moat created by the high switching costs of networks?
The co-founder of PayPal and an early investor in Facebook, Peter Thiel, has shared some fascinating insights on this topic in his recent book, Zero to One. He thinks it is crucial for the new network to start by targeting a small universe of users that it can hope to dominate. Facebook, for example, initially targeted only Harvard university students, and quickly came to be perceived as the social network on the campus. Growth then involves expanding this concentric circle of users, all the while remaining dominant within it. Facebook expanded by growing into the adjacent markets of other universities; only much later did it expand to the general set of users outside its initial university focus. By then its tremendous momentum had a sense of inevitability that helped it topple Myspace, and seize the lead to become the dominant network of today.
Similarly, Amazon initially targeted just selling books online, and quickly came to dominate this niche. It then expanded into other media; only much later did it take on the much bigger eCommerce market. Consequently, it has enjoyed the benefits of self-fulfilling expectations — being perceived as the inevitable winner throughout.
It seems to me that another way a challenger can win the network battle against a well-established incumbent is by linking to an even larger network. Microsoft successfully toppled Netscape’s web browser by tightly linking their Internet Explorer to Windows, thus leveraging the massive network of the Windows installed base against Netscape’s then dominant browser network.
This appears to be happening these days to eBay’s dominant network of third party sellers. Amazon is trying to leverage its massive 250 million first party customers into its smaller but rapidly growing third party marketplace. It has opened up its existing (first party) product catalog and its massive shipping infrastructure (“Fulfillment by Amazon”) to third party sellers. I predict that it is only a matter of time before this battle tips decisively towards Amazon and against eBay.
A series of monopolies
Perhaps due to diminishing returns on scaling production of physical goods, the older industrial economy has been characterized by a set of stable oligopolies (a few large companies dominate the production of cars, airlines, chemicals, miners, oil, paper, etc.). In contrast, the presence of positive feedback loops in demand as well as supply often leads to the winner establishing a winner-take-all monopoly-like position in the economy of information-based goods. However, such monopolies do not necessarily last forever. The demise of MySpace and Netscape shows that a determined challenger with a 10x better product, or a clever linkage to an even bigger installed base, can overcome an existing network and establish a new monopoly.
I expect the information age will likely be governed by a series of monopolies, each successive disruptor toppling the older incumbent. This is an important pattern to note, since Buffett has shown during the past fifty years how monopolies can lead to amazing compounded investment returns. I think, however, that digital monopolies will be a lot more volatile — positive feedback can lead to vicious cycles as well, if a challenger succeeds in unbundling a monopoly incumbent. Value investors need to be very cognizant of the powerful forces of disruption that can destroy an existing monopoly and create a new one in its stead.
Disclosure: As of the time of writing this post, I am long many of the information economy stocks mentioned in this post, including Amazon, Google, Facebook, Twitter, Intel, and Microsoft. This post is not meant to be and should not be construed as investment advice of any sort. Investing is extremely difficult, the risk of permanent loss is high, and past results are meaningless in the future.