Central planning


I used to work for an e-commerce startup that sold cheap clothes manufactured in China at large markups to American customers. Most of my job involved “planning”: providing recommendations on how many clothes of each size/color/style we’d need to buy to keep our inventory healthy. Buying too few clothes, either in aggregate or for a specific “stock keeping unit” (SKU), would create pockets of out-of-stock inventory, reducing our potential sales. Conversely, buying too many clothes would waste money by loading up inventory behind SKUs unlikely to deplete in the near future. (Our current “supply chain crisis” is one example of what transpires when this compromise cannot be successfully struck.)

One thing struck me as strange, in hindsight: my introductory economics textbooks had virtually nothing to say about a person like me, a “planner”. These textbooks spend inordinate time talking about “markets” and “marketplaces”: places where actors — buyers and sellers — making rational decisions cause the market to clear. In this picture, my company would be one of these actors: a seller of goods in the marketplace for (cheaply-made but highly marked-up) clothes. But, of course, there are actors within actors: how successfully my company sold its wares depended on the ability of its planners. The standard economic picture ignores individuals within firms —the intra-firm dynamics, in favor of examining firms within markets—the inter-firm dynamics.

This simplification is not unreasonable. It would certainly be reassuring if the success of a market did not depend on aptitude (or luck) of a solitary coder. And, as the logic goes, even if the planning decisions of a single company are poor, this does not entail a market failure. Quite the opposite. In the short-term, my company’s inventory being out-of-stock represents an opportunity for its competitors to make up the shortfall. In the long-term, if I do a systematically poor job at planning our inventory, my company would experience either systematically higher costs, lower sales, or both, and eventually be driven out of business. This constant churning — the exit of uncompetitive, sclerotic firms and the entry of new, vibrant ones — is what is said to ensure a healthy marketplace where customer needs are satisfied. Even if there might be failures of “local” planning (or, in other words, even if I did my job poorly), the market in aggregate functions well: because it is not centrally planned, and therefore these local failures do not spread to the entire system.

We can ignore what happens within a firm if the firm is small relative to the market. But what if that is not the case? There was a slightly troll-y article in Jacobin a couple of years ago called, “Yes, a Planned Economy Can Actually Work”. The argument was, essentially, that if giant firms like Alibaba, Amazon, and Walmart are able to plan demand successfully, then the government should be able to as well. Perhaps “previous state planners in the Soviet Union and the early People’s Republic of China failed due to insufficient information”, but that limitation no longer applies in this glorious new age of data and AI.

Setting aside the question of whether state planning is either technically feasible or a good idea (I’m more sanguine on the former than the latter), the interesting point, to my mind, is that the traditional picture of capitalism and markets is breaking down. Once, perhaps, firms operated within markets. Now, increasingly, firms are the markets, or, even more surprisingly, markets operate within firms. Employees at large technology companies are making decisions about markets that have profound effects on the finances of the thousands or even millions of actors that participate in those markets. These employees are central planners. And, unlike with me, their effects on markets should not necessarily be ignored or assumed away.

Typical accounts of anti-competitive behavior by large technology companies have largely centered on a conventional framing of these companies as participants within a market. In Lina Khan’s famous article, Amazon’s Antitrust Paradox, she devotes considerable attention to problems of predatory pricing, horizontal integration, vertical integration, and the like. In one example (that I’ve written about myself), she recounts Amazon’s efforts to drive a competitor, Quidsi (owners of diapers.com) out of business by systematically underpricing their diapers, causing Quidsi to accrue large losses and eventually capitulate to Amazon’s takeover bid. This is certainly awful behavior, and I don’t mean to underplay it. But, to reiterate, the idea Amazon competes against other e-commerce companies is somewhat antiquated. It is not an actor in a market for buying and selling diapers online: it is the market. As a consequence, many of the problems with Amazon are not manifestations of market concentration, but instead of market manipulation.

Let’s focus on Amazon Prime, the paid subscription of $139/year that gets you access to free 2-day shipping on Amazon, its entire video and music catalog, discounts/cash back at Whole Foods, and so on. Amazon’s strategy has been to make Prime such a compelling value proposition that American households could not imagine a different way to shop. Matt Stoller writes,

[Bezos said] the point of Prime was to use free shipping “to draw a moat around our best customers.” The goal was to get people used to buying from Amazon, knowing they wouldn’t have to worry about shipping charges. Once Amazon had control of a large chunk of online retail customers, it could then begin dictating terms of sellers who needed to reach them.

This is a marketplace, by any definition: on one side are the Prime customers, and on the other are the thousands of sellers who wish to reach them. But it is a marketplace within Amazon, not outside of Amazon. Amazon sells goods within this marketplace, to be sure. But it also extracts fees from the market actors, and sets the rules of the marketplace, and these latter sources of revenue are increasingly driving Amazon’s bottom line. (Khan writes, “third-party sellers’ share of total items sold on Amazon rose from 36% in 2011 to over 50% in 2015. That figure has only in recent years, with the 2018 number being almost 60%.)

What are some examples of the “market manipulation” I alluded to above? The obvious one is that if Amazon is both selling its own products within a marketplace and running that same marketplace, there is an obvious incentive to privilege its own goods above its competitors. ProPublica found that Amazon price comparisons are misleading for non-Prime customers, where Amazon’s “rankings omit shipping costs only for its own products and those sold by companies that pay Amazon for its services”. Amazon boosts the position of its own products in search even if they do not represent the best value for the customer. Amazon also uses its marketplace data to design knockoffs of popular products and then undercut the originals on price, one example being Amazon’s shoddy Allbirds lookalikes.

Some, like Elizabeth Warren, have argued that actors running a platform should not be able to sell goods on that platform; instead, Amazon should be turned into a “platform utility”. Fair enough. But I think even this understates the problem. Even if Amazon sold nothing on its own making on its platform, opportunities for manipulation would exist, and, for Amazon Marketplace’s central planners, would likely be irresistible.

Amazon provides incentives for sellers to join the “Fulfilled by Amazon” (FBA) program, wherein Amazon handles the logistics of shipping goods and processing returns. The odds of a seller appearing in Amazon’s search “buy box” without being in FBA are minuscule. Amazon has essentially complete discretion over what fees to charge sellers, either as a subscription (recurring revenue) or on every purchase (as a percentage of the sale price). Combining the FBA and other fees, Matt Stoller finds that “nearly half the revenue of a seller go[es] to Amazon”. And Amazon charges additional fees, not covered by the aforementioned categories, to sellers to promote themselves in Amazon’s search results. Cory Doctorow writes,

Remember when Amazon’s screen real estate was given over to “Customers who bought this also bought this” and “Customers who viewed also viewed”? Today those slots are filled with “Sponsored products related to” and “Brands related to this category.”

In other words, Amazon has converted its “customer-centric” personalization system, which emphasized the products it predicted you would like best, into an auction house, where the products that have paid the most come first.

It is easy to get lost in the details. The important thing to remember is that Amazon has tremendous power over the participants in the marketplace, and can use that power to extract all sorts of fees, whether or not they are called as such. (And this is true whether or not Amazon is selling its own goods in the marketplace or not.) In Amazon’s case, the impact is borne primarily by the merchants (sellers), but it is also worth noting that customers (Prime members) have seen rapid price increases, from $59/year in 2005 to $139/year now, far outpacing inflation. It is also likely that many of the fees that Amazon charges to sellers are indirectly passed on to consumers in the form of higher prices. (Sadly, there is no such thing as free shipping.)

It is not just Amazon. Every major technology company runs a marketplace. And every such company has product managers thinking about how to make it worse by extracting more money from its participants. Twitter, Google, and Facebook run ads marketplaces. (Google, according to a lawsuit filed by the Texas attorney general, lied about its ads auctions, claiming they were “second price” but actually charging third price and pocketing the difference.). Amazon, Ebay, and Walmart run e-commerce marketplaces. Airbnb runs a marketplace for vacation and long-term stays. Uber and Lyft run marketplaces that pair riders and drivers. Even companies, like Spotify, that had historically refrained from such behavior are now growing interested in the easy rents that come with being a marketplace owner, or, rather, a market manipulator. In 2019, Spotify announced its own marketplace:

*We personalize these new album recommendations based on your listening taste, combined with human curation. With an upcoming test we’re running in the US, we’re giving artists and their teams the ability to directly tap into this process and connect with the fans that care most about their music.

In this test, we will let artist teams pay to sponsor these recommendations, giving them the power to tell their listeners on Spotify — across both our Free and Premium tiers — about their latest release.

Did you catch that? Spotify is now going to make money off artists, not just listeners, by charging artists the privilege of being recommended. It is no different from Amazon charging Energizer for its buy box for a customer looking for batteries, or Google charging Expedia for its top search position for a customer looking for a cheap getaway. It’s easy money, and, if there’s no real alternative, the market participants will grumble but ultimately comply.

The point I hope I’ve made is that it is not enough to crack down on market concentration or vertical integration or conflicts of interest posed by marketplaces shilling their own stuff. These are indeed problems, but, in my view, somewhat ancillary. The real problem is the marketplace. The magical self-organized marketplace in which firms and customers settle on a market-clearing price is simply not relevant to big tech. Markets are increasingly now within firms, not the other way around. And, instead of self-organization, we have central planning. The question is whether we want those central planners to be product managers at Amazon or regulators in the government.


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s