Preferential Attachment and Consumer Surplus in Networked Economies

Why the battle between WhatsApp and Telegram is more important than you think.

First of all, what is preferential attachment? Preferential attachment is a process in which some quantity, typically some form of wealth or number of users on a social platform, is distributed according to how much a firm, product or individual already has, so that those who for example have already captured a large user-base will see it grow disproportionately compared to those who have not.

Processes of preferential attachment generally feed into increasing returns, defined colloquially as “that which has, gets; that which does not, gets less.” This happens because the vast majority of a population facing a shared problem prefer to assume that choices made by the more qualified among them are more reliable.

When was the last time you judged somebody’s success by the car they drove as opposed to demanding their income statements? Why does the cost of a diamond engagement ring converge to the price of the average monthly salary of anybody, anywhere in the world? Signals matter.

It turns out that individuals rely less on processing and analyzing information personally and more on the signals generated by the influencers in their networks that they trust because life is complex and relying on authority figures is often cheaper than putting in the work yourself (think about how the market might move when Soros or Greenspan initiate a put option on an otherwise bullish market).

So what are increasing returns in the economic sense? Increasing returns are defined from the perspective of the agent or firm that receives them, whereas preferential attachment is the process that generates those returns. The degree to which a firm’s economic returns are increasing in scale and for how long defines to what extent they are a monopoly. Insomuch as preferential attachment contributes to these returns, it is a process that generates monopolies. Keep this in mind.

Monopolies, as we all know, are bad for consumer surplus. By definition, a monopoly precludes the existence of firms which could have otherwise taken form under more competitive circumstances. Fewer firms mean fewer choices for consumers. Fewer choices means companies can get away with keeping prices up and quality down. While the Cournot Duopoly model shows that oligopolies do end up converging to prices that lie between monopoly and competitive prices, they do so over time-scales that do not preclude the possibility of additional negative (and often unintended) consequences.

Proponents of monopolies argue that long-period protection of key intellectual property is a boon to innovation, but the fact is that monopolies kill innovation faster than even governments do (Cass, ICER, 2012). Reagonomics is dead. The innovation argument as trickle-down is baseless. Innovation bubbles up from the bottom. Move along.

An agent-based simulation demonstrating preferential attachment on a network.

Try it out yourself. Press Setup, then Go. Press Resize Nodes to visualize the importance of a given node. Importance is defined as its in-degree, or the number of connections coming into the node. Now monitor the Degree Distribution. Do you see a Bell curve, a standard normal distribution? Not likely. What you see is a Power Law, where a small proportion of all nodes receive disproportionately more of the available input than most others.

What this basically tells us that in any process in which at least some of the decision to consume, share or utilize a given resource is impacted not just by one’s own rational determination but by the decisions and opinions of others — whether that be the emergence of a new social network application, consumer product, or cultural trend —we are likely to see a distribution of the rewards that looks something like the 80/20 rule, the 90/10 rule or worse, the 99/1 dichotomy we’ve all come to know through the Occupy Wall Street movement. This is inequality at its natural finest, but nature ought not always imply goodness.

Let’s take an example of firm formation in an industry that wrests heavily on preferential attachment. In a world in which your choice of messaging application is restricted to either WhatsApp or Telegram, which would you choose? Most people will think first about what proportion of their current friends who use messaging applications use which particular product, because that’s the most important value-add for a chat app: “If my potential friends list when using WhatsApp is already larger or of a higher quality than it is for Telegram, I’ll choose WhatsApp.”

No harm done. If everybody ends up making their decision this way, however, it’s natural to see that initially insignificant differences in potential friends lists will ultimately spiral wildly out of control and result in, at least for our example, WhatsApp’s total hegemony over the market (as of today, WhatsApp, with 1 billion monthly active users, has 10 times as much traction as Telegram, the third largest competitor; Kik has 200 million; Messenger (owned also by Facebook) has about 200 million as well. Basically, the allocation of long-run market power in industries dominated by preferential attachment come down to a function that combines at least some early decisions based on market quality considerations, but then is wholly consumed by the mostly random decisions of the industry’s earliest marginal users.

But why should most firms be forced to compete within such an exceptionally small initial time frame or risk losing out on participating essentially altogether? This isn’t just unfair to firms. It’s unfair to future users of the early winning platform. What if all of a sudden a new market consideration comes into play? What if all of a sudden, data privacy matters because the amount of data generated by the application becomes massive and advertisers and insurance companies learn that they can prey on uneducated or otherwise leverageless consumers?

Well, with only one or at most two viable applications on the market, you don’t have the choice to use another in a way that is even remotely comparable to your accustomed quality standards. So what happens? People, having already decided which app they like to use the most, discount the value of the new market consideration, and some may even actually deny it matters. In an alternate Universe, however, where a larger amount of smaller firms offering a wider diversity of choices had actually existed, this is something these same people may have actually cared about, but unless you have access to a Time Machine you’ll never know.

In effect, this dichotomy between present and hypothetical alternative preferences is a restriction on your own future consumer surplus that depends on logic that is not even good enough to be wrong. Furthermore, when a new company actually does come out with a solution to address the new market consideration, i.e. data privacy viz. end-to-end encryption, the cost of adopting a new platform means convincing all of your friends to join, but they won’t join unless all of their friends join. Who makes the first move?

I’ve tried it myself! In over three years, I’ve managed to convince all of 4 friends to adopt Telegram. When I don’t like the service at a Greek restaurant I walk down the street and choose the nearest Italian. Sure, it’s not the same, but I can easily make a choice that allows me to consume a similar amount of real value. And if a few of my friends agree, the switching cost of convincing the entire entourage is not so high. When I want to take advantage of a new feature that I really care about on a social network, I truly can’t do so unless the majority of my friends join me. So a small desired increase in quality is offset by an dramatic decrease in the thing that matters to me most, i.e. my friends list. In a free market, this is not the kind of dilemma that should ever exist.

Any solutions? Well, one way to bring power back to consumers could be to stop storing users in central databases altogether. This would also include their relationships with others. What if I could access any application using the same user account, anonymously should I so choose, and what if that user account was stored in such a way that left me totally in charge of the way its data was processed, perhaps on a server owned by me or by a collective of like-minded other users? Virtual private servers such as Urbit accomplish this technically but there is currently no indication thats uch platforms will ever reach widespread market adoption.

What if I could personally set the rate to charge credit ratings agencies, insurance companies, and advertisers for access to my data? Markets would form. This could be the basis of a Digital New Deal, and perhaps the first widely adopted Universal Basic Income. The implications are many, and would totally eviscerate the unearned differential market power of recent Silicon Valley megaliths such as Facebook and AirBnb, as well as future others. Unfortunately, the only way to implement this would involve government intervention or at least third-party regulation in the form of a Universal Single Sign On (SSO) API and the politics could be complicated (Addendum: I wrote this before Blockchain was in vogue). It would be a monumental task, but I think it would worth it in order to make markets freer.

Today’s tech leaders are canny. They claim to “make the world a better place” despite engaging in every effort to “move first”, to chase scale, to build virality into their products, to form a stranglehold on network effects. The entire effort is a bit of an oxymoron.

The case can be made for the opposite, of a possible extra-normal deceleration of increasing returns via positive feedback loops of a more destructive nature, as in Uber’s recent PR spat over surge pricing in the face of the “Muslim ban”, or more traditionally as Schumpeter’s Gales of Creative Destruction. In practice, however, the threat is not credible. It was easy to depose the largely non-technologically savy taxi industry, but will it be easy for the next generation of entrepreneurs to depose the Ubers of the world, and their venture-capital funded offshoots?

As for the recent virality of #DeleteUber, the story will peter out and things will go back to normal in a few weeks. Firms that have secured their position at the top are not going away anytime soon because for contemporary digital social networks, user-switching costs become totally insurmountable once a certain threshold level of adoption has been sufficiently capitalized. This invariably feeds into the next wave of monopolistic product offerings.

Frankly, this is a new kind of game. Even if these firms one day fail, their shareholders, board members, and original investors will always be in a better position than potentially more technically qualified competitors to build up the next big monopoly.

Who cares if monopolies crumble if new ones merely form in their place? If the underlying power dynamics never change?

Look on my works, ye Mighty, and despair!” Of that colossal creation, boundless that lair, The full and shining Empire stretch far away. Ozymandias lives.

Martin is Lead Project Manager at Selotec. You should follow him on Twitter.


Cass, R. A. (2012). Antitrust for High-Tech and Low: Regulation, Innovation, and Risk. International Center for Economics Research.

Wilensky, U. (2005). NetLogo Preferential Attachment model. Center for Connected Learning and Computer-Based Modeling, Northwestern University, Evanston, IL.

Wilensky, U. (1999). NetLogo. Center for Connected Learning and Computer-Based Modeling, Northwestern University, Evanston, IL.