Amazon extracts more value from its ecosystem than it contributes. The two techniques by which it achieves this is moving transaction costs to other stakeholders and subsidizing preferred products and stakeholders. In this blog, we show that these are cross-layer subsidies and they are anti-competitive practices.
Two anti-competitive practices of Amazon are using seller data to introduce its own private labels and subsidizing new ventures from AWS revenue.
Amazon collects data from sellers on its Marketplace to launch competing products. This is like a shopping mall that collects all sales transactions from shops in the mall, and then competes with these shops by manufacturing and selling successful products under its own private label — in the same mall. The following diagram shows how this works.
The diagram shows three layers of Amazon’s ecosystem, where any layer provides services to the layers above it. Each layer is a platform that enables transactions of actors in higher layers. Amazon marketplace here collects data from the layer that it supports and uses this to subsidize its own private labels, which sell on its marketplace.
This is different from a retailer selling private labels. The proper analogy is a shopping mall that handles all payments of the shops in the mall, and uses this data to sell private labels in its own shop in the mall.
Subsidizing preferred stakeholders
The other anti-competitive practice is subsidizing new business using the proceeds from Amazon Web Services (AWS). This creates a vicious feedback loop in which indirectly, competitors of an Amazon subsidiary finance price subsidies for the subsidiary. Amazon freight is a platform that matches shippers with freight service companies that operates 30% below market price.
The above diagram shows another three layers in Amazon’s ecosystem. AWS is a hugely successful services that earns revenue from many companies. Amazon uses this revenue to subsidize its own subsidiaries, which not only use AWS but also compete with businesses that also use AWS, and hence indirectly subsidize the Amazon subsidiaries who compete with them. Again, this is unfair competition.
Red Queen dynamics
Platform-enabled subsidies create a Red Queen dynamic for non-favored ecosystem participants: they must run faster to stay at the same place. In this case, they must lower their prices to continue competing, without having the resources that allows the subsidized competitor to compete. At the same time, they do provide the resources —data or money— to subsidize their competitor.
Other mechanisms that create a Red Queen race for non-favored ecosystem participants are the transaction fees that Amazon asks for sales and in-app purchases, the design of matching algorithms and search result pages by Amazon. These designs force third-party sellers to pay for favorable placement in search results and for advertising. As O’Reilly observes, over time this tilts the distribution of value over ecosystem partners in favor of Amazon.
Traditionally, anti-trust regulators tend to look at price for customers only. If a market configuration created by a company does not lead to a higher price for customers, they leave the company alone. However, as argued convincingly by O’Reilly, regulators should consider the allocation of value to all ecosystem participants, not just to customers. And they should track the development of that allocation over time.
Regulators need data to do this. As stated by O’Reilly:
“Data is the currency of these companies. It should also be the currency of regulators. You cannot regulate what you don’t understand. Algorithms may be trade secrets, but their outcomes should not be.”
Regardless of the treatment of sellers, Amazon uses cross-layer subsidies of data and money to prefer its own products and subsidiaries. Current antitrust proposals talk about splitting up Amazon. But splitting the marketplace into two would not end the anticompetitive practice of favouring some ecosystem partners over others, and it may well lead to one of these become a near monopolist.
The relevant split is that between ecosystem layers. Platform companies should not fund nor provide data to favoured participants in the ecosystem that they enable. This surely would disrupt Amazon’s business model, but it would improve the health of its ecosystem.
The current antitrust proposals of the House of representatives do exactly that. The Ending Platform Monopolies Act prohibits a company that operates a marketplace to also trade as shop on this marketplace. The American Choice and Innovation Online Act prohibits platforms to give preferential treatment to their own products. In particular, they cannot use non-public data collected about all transactions on their platform to favour their own products.
Centralized digital ecosystems
Cross-layer data subsidies are made possible by centralizing a digital ecosystem. Amazon has enabled an ecosystem in which it collects so much data that it can create a competitive advantage for any actor in the system. In the next blog, we look closer at the structure of these ecosystems.