Let's talk a bit about what @VitalikButerin called "anti-network effects". TLDR: There is no empirical basis for the notion that network effects inevitably lead to a winner-takes-all outcome. In most cases, such a conclusion is simply a matter of framing.
One of the iffy folk tales in network economics emerged when Brian Arthur didn't get his papers on increasing returns published in 1985/6 and opined that this must be that this must be bc his ideas were too radical and thus rejected by mainstream economists.
That maverick story was repeated in Mitchell Waldrop's 1992 on Complexity and the early days of the Santa Fe Institute, and like many of those truthy-sounding stories it just refuses to die. It's also mostly nonsense.
First, increasing returns is a concept out of production theory. Simplified, if producing another unit adds more revenue than cost (meaning, it contributes to firm profit) the production is in the range of increasing returns [to scale] and should keep producing.
Once additional costs outrun revenues per unit, we're reaching the realm of decreasing returns, and we should stop adding units to our production plan. Decreasing returns has some nifty properties, esp it is much easier to create models with unique equilibrium points.
So there's a bit of an opportunistic reason to model production under decreasing returns, but there's also the simple observation that typical industrial production works at the upper end of capacity rather than at the lower end and capacity constraints are a major source of DRs.
[Side note: This isn't all up to date in an economy where production also includes such things as software or intellectual property, products with "second-unit" costs of near zero, and no capacity constraints.]
Returns and scale economies are a powerful concept in economics, and can also be applied to the demand side, as Arthur did. In a simple consumer choice model we know diminishing marginal utility, or the concept that people usually like more of a thing, but not that much more...
Arthur and his contemporaries were mostly interested in inter-user scale economies (aka "network effects"), so even if users buy only one unit of a thing, they might derive more utility from others using the same thing. That would be a demand-side increasing returns scenario.
Not to take too much away from Arthur, he was certainly at the forefront of thinking on those things in the early 1980s, but then again it all goes back to Schelling's "dying seminar" critical mass model from 1978.
Schelling's model, a close relative of Keynes' stock market as beauty contest model, is pretty much a returns dynamics flipped onto itself. Academics are more likely to attend a seminar when they perceive surging attendance, and vice versa.
So the seminar is either fully booked, empty, or hits an unstable equilibrium point inbetween (the critical mass point), where the increasing and decreasing forces cancel each other out. Notably, Schelling's model has an upper cap: the capacity of the seminar room.
After this prelude, a couple of things on how structure and context matter in investigating the limits of network effects, and how oft repeated outcomes are more the result of modeling assumptions rather than characteristics of the network itself.
For instance, most academic models and almost all folk tales tackle competition between networks as a endgame between two remaining incompatible competitors in which each consumer has to make a binary choice for either one or the other. VHS vs Beta is the archetype.
But there is almost no case where that captures the facts in the wild. Any technology with popular appeal sees multiple entrants with often significant technological differences, users can install multiple clients, technology and even adoption patterns change over time.
Even though most of the networks we are looking at are technological, the motivating interaction effects are typically behavioral. If we think of social networks, we think of peer-to-peer interaction (one-way or mutual), but the structure of the interaction is quite tricky.
One of Facebook's biggest problems is that its longstanding attempt to replicate every user's real-life network runs against social behavior online. While online we like to sort ourselves into affinity clusters (bubbles) our offline networks are often based on more formal ties.
Continued here:
