The Kids Markets Aren't Alright

Dec 7, 2023

TL;DR: The most profitable ventures too often resemble landlords.

There is an inherent economic tension between efficiency and equality. Utilitarian models with linear value on consumption often favor extremely high levels of inequality, in which massive capital is invested by a small number of sophisticated entities to increase overall productivity. But the world is more accurately modeled by sublinear utility. (The alternative is not usually considered.) In the Rawlsian extreme -- the min function -- equality is paramount over economic efficiency.

The Rawlsian extreme is also a bad model, as the world doesn't wholly revolve around its single unhappiest inhabitant. So, the liberal ideal balances somewhere in the middle, where we justify modest levels of inequality with the overall gains to society they produce, and the exact balance is adjusted through perpetual debate.

But implicit in this liberal contract is that the concentration of capital is used to increase the overall surplus. Otherwise, the Marxist comparison of capitalists to dynastic landlords gains pertinence -- and why should society pay rent in eternity to landlords?

Part 1: Natural Selection of Moats

A moat is anything that protects a business from competition. I prefer to frame moats as having the broadest possible definition: that for every person or company in the world, there is a set of moats that explains why they are not competing with you. Under this maximalist definition, the margin a company can derive is determined entirely by the quality of its moats.

Commonly discussed moats include unusual domain expertise or interest, access to capital or talent, customer demand, or perhaps simply accumulated physical, digital, or legal assets, like factories, codebases, or patents. But there are many others, too. Physical stores are protected from competition by geography, allowing them to carve out a local market that is less competitive -- and thus prices for the same goods vary wildly between New York and New Mexico. State-owned companies are usually protected from competition by men with guns. And all companies are protected from competition by the innate cost, effort, and risk of starting a new enterprise.

If one thinks of companies as living, changing organisms, one can imagine a kind of natural selection process at play on the moats themselves. Moats usually erode over time from competition, and companies either adapt or die. Intel depended on their DRAM manufacturing moat until Japanese companies learned to compete in the 1980s. Intel was forced to reinvent itself around microprocessors, which then turned out to have even stronger moats. Companies are more Lamarckian than Darwinian. Even so, those that lose too many moats suffer a fate little better than bankruptcy: commoditization.

An oft-overlooked economic impact of computers was that they led to the realization of new kinds of corporate moats. These moats are so deep that the companies which have mastered them are, to use Marc Andreessen's term, “quasi-monopolies.” These companies face relatively weak competitive pressures, because their businesses inherently perpetuate themselves over competitors. The deepest seem to be scalable distribution moats: for example, those which either maximize aggregation or minimize interoperability.

Aggregation is a generalization of network effects which describes businesses which derive value by mediating between different entities and capturing a fraction of the value that passes through. Some aggregative businesses, like Grindr, are driven by relatively homogenous network effects. Other businesses may have more complex aggregative structures, with two-sided or even three-sided markets. But in all cases, the moat imposed by the aggregation emerges from (A) the surplus being maximized by having all users on a single platform, and (B) the massive scale of the coordination problem in transitioning onto another platform.

Interoperability is, as of late, probably less well-considered than aggregation but probably more important. The key point is that for products to be truly in substitutive competition with each other, they need to really be almost identical. For example, any two bread toasters on Amazon under $25 are likely in competition, because they'll have very similar feature sets: a dial to adjust toast darkness, and a lever to lower the toast and turn it on. But this level of similarity almost never happens in software, because even the simplest software is far more complicated than any physical good sold to consumers. Software with identical features but different user interfaces are not necessarily interchangeable. When one adds in data or vendor lock-in, or even just subtle forms of friction (disabling certain features when users don't purchase complementary products), the seller can create essentially arbitrarily fine-grained pressure on the purchaser. Oracle and IBM, which continue to make vast margins selling protection payments -- whoops, I mean mainframes -- represent this class well.

The main issue with these moats is that they are uncorrelated (or in the latter case, truly anti-correlated) with the quality of the product itself. Apple was happy to degrade the iMessage experience for hundreds of millions of its own users in order to throw around its market power and put pressure on the consumers outside its ecosystem. And as profits become divorced from the consumer experience through these structural moats, consumers suffer.

These moats have proven so foundational for the technology industry that they can be found in every corner of it. Database providers make data ingress free but egress extraordinarily expensive to discourage competition. Google and Facebook depend mostly on aggregation, whereas Apple relies on both; discouraging interoperation through their control of the hardware and operating system, and aggregating in their service sector like the App Store. NVIDIA's moat comes as much from owning CUDA and preventing software interoperation with other GPU providers as it does from hardware performance.

Although the phenomenal complexity and scalability of technology makes these moats easier to realize, they're not constrained to technology, either: John Deere's lock-in on repair is a clear attempt to bring Apple's business model to the world of farm equipment. I worry companies outside of technology will continue to be inspired by these principles.

Part 2: AggreCo, Inc

Let's take a moment to imagine an idealized aggregation business: AggreCo, Inc. AggreCo has several million American users who use it every day to drive most of their income, and AggreCo taxes 30% of the value exchanged on their platform. Additionally, AggreCo is completely dominant in their platform's space, with 95% saturation (though AggreCo strongly insists that they are but one small piece of a much larger market). The financials of the company speak for themselves: they spend 20% of their revenue on personnel, 30% on marketing their core business, and 50% turns to net income. Yet despite their thousands of employees and countless micro-optimizations, the core product hasn't really changed in the last five years.

The question is: is it right, from the perspective of society, for AggreCo to take this tax on the value driven through their platform?

On the one hand, the aggregation provided by AggreCo has created tens of billions of dollars of annual value to society by lowering friction. Furthermore AggreCo bet their entire business on building out this new market, and the company worked nights and weekends to get to this point -- and if society wants to encourage such a work ethic, it ought to make the rewards commensurate to the risk.

On the other hand, the founders, investors, and early employees are all wealthy beyond their imagination, and the company hasn't really done anything in the last five years other than keep the wheel turning. It spends more on marketing than on its own product. At what point does the motivation provided by a marginal yacht become less important to society than its marginal cost?

The best answer that I can come up with centers on a notion of value over replacement. (A more principled approach would use a Shapley value but this is well known to be intractable in practice; leave-one-out is a good approximation.) Let's consider two different potential backgrounds to the AggreCo business.

  1. AggreCo was the lone company that believed the market and its tremendous surplus value could be unlocked, spent vast amounts of capital to perfect the product and convert a critical mass of users, and had to be rescued with last-minute capital from its dedicated investors a few times before the company finally reached profitability. In fact, the first two years after the launch of the product were net losses because the user base grew so quickly, and it took another two years for AggreCo to make up those losses.
  2. The AggreCo market was extremely obvious to everyone, and AggreCo and its nine competitors all started within about six months of each other. AggreCo simply won through aggressive expansion and pricing to drive their competitors out of business.

In the former case, AggreCo drove tremendous value above replacement: the market and all its surplus simply would not have existed without AggreCo for at least another decade; thus, I can justify AggreCo capturing any fraction of the surplus value for the first decade. In the latter case, AggreCo added essentially zero value beyond the world without the company, and it would have been broadly better for society if they and most of their competitors had worked on a different problem instead. AggreCo's tremendous profitability is not an indication of the quality of the company; it's a consequence of the structure of the market. I see little reason for society to so richly compensate the venture in the second case.

Part 3: What to Do?

I don't mean to imply that aggregative businesses should not exist, nor that companies should be required to collaborate with their competitors to commoditize themselves. I am genuinely sympathetic that Apple does not particularly want to devote extra effort, for example, to make AirPods fully-featured with competing devices. And, I certainly believe that the hard work and risk that goes into creating differentiated value should be richly rewarded by society. My fear is that, as in the hypothetical AggreCo example, these distribution moats distort markets far beyond this differentiated value.

This distortion is not necessarily a problem. When these profits drive large-scale product investment and improve consumer experience, it may be a relatively efficient way for society to invest in and deploy new advances! But if distribution is the goal, and the minimum product investment is conducted to ensure continued distributional dominance, then these structural moats pose a real threat to the social contract of capitalism.

In any case, I certainly wouldn't advocate the wholesale slaughter of the companies that depend on these moats. I think we should aim to have our cake and eat it, too.

First, even if the quasi-monopolies prove, on the whole, are a net drain, many individually may be positive, and many more may be able to be reformed. I don't think that regulating Apple as a utility would be likely to lead to much good.

That isn't to say that such reforms would be easy. Trillions of dollars of annual revenue and hundreds of billions of dollars of profit depend on these moats; these companies will certainly have to be dragged kicking and screaming. And some collateral damage is surely guaranteed.

I think there could be value in developing a basket of metrics to track the degree by which companies depend on these relatively harmful moats. One wouldn't want to be over-reliant on specific metrics that could be gamed, but with a sufficiently broad basket of high-level metrics, I think one could get a good picture.

One example metric would be the ratio of money spent on marketing to money spent on product development. A company spending lots of money on marketing instead of developing its products is by no means damning, but it is circumstantially concerning.

Another such metric would be to measure the correlation of product purchases to help identify lock-in. For example, if consumers strongly bifurcate into either buying many products from a single large company or very few, then this is an indication that adoption is substantially driven by existing distribution. Again, this isn't damning -- perhaps all the company's products really appeal to a very specific demographic! But one can answer quite directly if the company is very large and serves many demographics.

Of course, the engineering of these metrics must be done with extreme care, to prevent them from being gamed or causing collateral damage. Fortunately, they only need to be applied or audited at very large scales, where this is reasonably affordable to do. The EU is presently taking an extreme version of this approach: to simply identify violators (“Gatekeepers”) directly. I think this is better than nothing, even if I would prefer a more subtle adjustment of incentives and market structures. But I look forward to the result of the experiment.


First: society's best chance at influencing companies towards the broad well-being of society runs through aligning incentives, and this active process is necessary to maintain the social contract which underpins the moral grounding of capitalism. Companies will build the best moats they can, and there's no reason to think that all are innately good for society. Each has its advantages and disadvantages, and adjusting the strength of different kinds of moats is an important and underused lever of public policy.

Distributional moats represent a particularly extreme case of those which are both deep and poorly aligned with society's interests. Key distributional moats center around aggregation (generalized network effects) or minimizing interoperability (adding friction to competitors' products), and can lead to highly distorted market dynamics which are bad for consumers and society at large. To this extent, we ought to take a good look at the processes that leads to these moats and modify the structure of these markets and the incentives of the companies which dominate them.