Taxing the Digital Economy Necessitates Moving Beyond Analogy

As the Multistate Tax Commission pushes for a new definition of “automated digital products,” which are provided for sale with minimal human intervention, states have a chance to rethink an outdated tax model that was built for physical goods.

The MTC’s idea for bringing state sales tax codes into the digital realm looks good on paper. But instead of defining digital products by what they are—PDFs, streams, apps—states should define them by what they do.

A Netflix subscription is entertainment, QuickBooks is a productivity tool, and therapy over Zoom is health care. Consumers already view their digital lives in these terms, and state tax codes should catch up. A functional model is one that makes intuitive sense and can scale up as new technologies emerge.

The proposed automated digital product definitional line has a simple policy logic: If you can sell it a thousand times over without adding on a single employee, you’re probably selling something that looks more like a retail good than a professional service.

The problem lies in the minimal human intervention line, which is less a bright-line rule and more of a palm smudge. Is a chatbot that escalates some questions to a live agent still automated? How about artificial intelligence tools that are trained and then retrained by human engineers?

For its part, the MTC admits the weaknesses and leans on international tax bodies such as the Organization for Economic Cooperation and Development and United Nations for guidance. But that merely imports a standard designed for cross-border tax issues into state sales tax codes.

At best, it kicks the clarity question down to auditors, regulators, and courts to provide. At worst, it promises a future with as many as 50 slightly different versions of the same core confusion.

These definitional issues aren’t new—most tax policy involving new technology begins by analogizing the new tech to something older and more clear cut. States spent decades arguing whether code on disk was tangible property, when software as a service was in fact a service, and whether cloud-delivered offerings were some new kind of digital chimera.

Courts and regulators ended up shouldering most of the load, with inconsistent results. The MTC’s minimal human intervention standard risks repeating that cycle.

This repetition of definitional battles reflects a deeper issue: We’re stuck in the goods-versus-services dichotomy—equal parts a product of history, politics, and policy inertia. Sales taxes are consumption taxes, so it makes some sense that goods are taxable, as they are consumed, whereas services aren’t.

But that binary makes less sense in a digital economy where the act of downloading a bunch of bits seems more like the provision of a service, as against a good, unless some consideration is made for how that bunch of bits is used.

States have long struggled with definitions that distinguish taxable digital goods from exempt human services. A movie download looks a lot like a product, but a livestreamed seminar with audience interaction sounds more like a service. Sales tax systems were built to efficiently capture the former and, in most cases, ignore the latter.

That’s why it may be time to stop asking whether a digital offering looks more like a good or a service entirely, and start asking what purpose it serves. A sensible tax code shouldn’t have to care whether a book is bound in paper or delivered in pixels—only whether it’s being consumed as entertainment, research, or something else.

That shift is both cleaner for policymakers and closer to how consumers experience their digital lives. States already treat movie tickets differently from medical services, for example, and this kind of taxonomy mirrors what tax administrations do when they distinguish deductible business expenses according to function rather than form.

It scales more naturally with emerging technologies—an AI research tool needn’t be analogized to a piece of software on a disk. By focusing on what a product does for the end user, states can build a more durable tax policy framework for taxing digital activity.

Of course, even such a function-first tax framework has edge cases. For instance, is a ChatGPT subscription research, productivity, entertainment, or something else? The answer depends on the user, which complicates categorization.

But the point isn’t that every edge case can be neatly solved. It’s that a functional model is at least asking the right question—what purpose this product serves, rather than how intense the human intervention was prior to it making its way out the door to the end user.

To the MTC’s credit, it’s trying to bring order to a tax base that has failed to keep up with the digital economy. That’s no small task. But importing vague standards from international tax regimes won’t give states the clarity they need.

If states are serious about taxing digital products, the clear route is through a focus on function rather than form. Tax rules that mirror real-world uses will be clearer for consumers, fairer, easier to administer, and more adaptable to technologies that haven’t yet come to market.

Andrew Leahey is an assistant professor of law at Drexel Kline School of Law, where he teaches classes on tax, technology, and regulation. Follow him on Mastodon at @andrew@esq.social

Read More Technically Speaking