Small businesses are a pillar of capitalism and of a healthy economy. They provide a path into the middle and upper-middle class for people who don’t have elite connections or a fancy college degree. And they distribute capital income broadly, so that the gains from business ownership aren’t concentrate among a few wealthy shareholders.

Retail is a big component of small business. In recent decades, big chains have pushed out mom-and-pop stores. But there has been hope that e-commerce would provide a lifeline for small retailers.

There’s just one problem. E-commerce in the U.S. is increasingly dominated by a single platform — Amazon.com Inc. As of 2018, the e-commerce giant had grabbed about half of the online retail market:

That wouldn’t necessarily be a problem for small retailers if Amazon simply provided a venue that allowed small businesses to connect with customers.

But increasingly, Amazon sells its own products, including private-label goods, that compete with the offerings of independent merchants on its platform. A recent paper by economists Feng Zhu and Qihong Liu observed Amazon’s behavior over time, and found that it tends to introduce products in niches that smaller merchants did the work of discovering by finding out what consumers like. Amazon then piggybacks on their efforts. This kind of tactic could be increasingly important as Amazon makes its own private-label products. This is similar to a classic move used by supermarkets — observe which products sell well, then introduce private-label brands to try to grab some of those markets. But technology has given online platforms superior tools to outcompete their suppliers.

One of these tools is search. Customers look for products using Amazon’s internal search function. Independent sellers can try to take sales from rivals by buying placements in the search results for a rival’s brand-name product — for example, when a customer searches for Purina dog food, she might see a promoted result for Kibbles ‘n Bits. But Amazon doesn’t allow other sellers to compete with its products this way — if you search for an Amazon product, you’re invariably going to see an Amazon product first. But if you search for another company’s product, you might see an Amazon product promoted at the top of the list. Amazon now is experimenting with a feature that could, if adopted, automatically include Amazon products in searches. And search isn’t the only advantage a platform has in the digital age; Amazon also collects potentially crucial sales and marketing data that it can choose not to share with third-party merchants.

Of course, any online retailer could do the same. But the world of e-commerce is subject to stronger network effects. When you buy something from a brick-and-mortar retailer, you tend to go to one that’s conveniently located, whether it’s a Wal-Mart, a Target, a Best Buy or a local convenience store. But online, unless you’re searching for a specialized product, there’s often no reason to go anywhere but Amazon. This naturally tends to push the platform market toward winner-take-all. And it’s exacerbated by Amazon’s practice of requiring merchants not to offer their products more cheaply on any other platform — a type of agreement known as a most-favored nation provision or MFN. This means that merchants who want to sell their products online have no choice but to be on Amazon, the biggest platform, and play by its rules.

Some, such as business professor Andre Hagiu, argue that Amazon would be foolish to out-compete its merchants, because this would deter them from offering their products on Amazon in the future. But researchers have long known that this logic doesn’t necessarily apply in the presence of dominant market power; back in 2000, economists Joseph Farrell and Michael Katz showed that a monopoly can have incentives to confiscate the profits from the innovation of companies who produce complementary products, thus stifling innovation. Tech publisher Tim O’Reilly argues that even if eating the third-party ecosystem isn’t a good long-term decision, a platform may be tempted to do it anyway just for the short-term profits.

So if Amazon is chewing up the small-business world — and the e-commerce world in general — what’s to be done? One approach is to identify and ban Amazon’s specific anticompetitive practices, as European authorities are trying to do in the case of data sharing. Another approach is to try to introduce competition into the e-commerce platform space by banning MFNs and other anticompetitive practices. But these efforts may be inadequate, since the former involves a continuous cat-and-mouse game between regulators and Amazon, and the latter probably won’t be enough to overcome the strong network effect driving the concentration of online retail.

Another alternative, of course, is simply to break up Amazon. But before such a drastic step is taken, economist Hal Singer argues, antitrust authorities should consider a gentler alternative — a nondiscrimination regime. This would basically allow any third-party merchant to lodge a complaint with the Federal Trade Commission or another independent tribunal. Although only larger merchants would have the resources to lodge such complaints, any victories they won would benefit smaller businesses as well, by curbing Amazon’s anticompetitive stratagems.

As e-commerce becomes a more important part of the U.S. retail landscape, questions like this will inevitably become more pressing. If small businesses are to survive in the digital age, the online economy can’t be allowed to be winner-take-all.