Allocative Efficiency and the Returns to Innovation

TEXT START: When producers in a network compete in quantities, an innovator's profit contains none of the change in allocative efficiency that its innovation causes, whether a gain or a loss, because its own choices are optimal and other effects on its profit are transfers.

The Dissection

This is an accounting correction, not a systemic diagnosis. The paper strips allocative gains and losses out of the innovator’s private return, rejects the lazy practice of counting rivals’ lost sales as social damage, and argues that spillovers make the social return to R&D vastly exceed the private return. Its strongest contribution is exposing how market prices misassign credit for innovation.

But it remains trapped inside a functioning market model: firms, output, markups, R&D competition, and welfare aggregation remain the central objects. It measures who receives the surplus while ignoring whether humans retain any necessary role in producing it.

The Core Fallacy

The paper confuses a more accurate surplus calculation with preservation of the economic order.

Under the Discontinuity Thesis, the decisive question is not whether innovation generates allocative efficiency or whether its social return is 5.39 times its private return. The decisive question is whether AI severs the mass employment → wage → consumption circuit. A larger social return can coexist with mass displacement, extreme ownership concentration, and the collapse of productive participation. Correcting the accounting does not restore human bargaining power, employment, or sovereignty.

The paper identifies a leak in the price system while leaving the load-bearing structure unexamined.

Hidden Assumptions

• That output and allocative efficiency are adequate proxies for social welfare.
• That firms and R&D games remain the relevant units after cognitive work becomes automatable.
• That innovation’s labor-displacing effects are either negligible, reversible, or outside the welfare calculation.
• That higher measured social returns can be converted into broad social benefit through ordinary institutions.
• That ownership of productive AI capital does not determine who receives the gains.
• That markets can continue coordinating human participation after AI achieves durable cost and performance superiority.
• That estimated spillovers remain meaningful when innovation can reorganize entire sectors rather than marginally improve them.
• That aggregate welfare can rise without addressing the collapse of the majority’s claim on production.

These assumptions bypass P1, P2, and P3: cognitive automation dominance, coordination failure, and productive participation collapse.

Social Function

Classification: partial truth, prestige signaling, and ideological anesthetic.

The paper is not nonsense. Its correction to business-stealing and spillover accounting is potentially valid within its model. Its anesthetic function is narrower: it keeps attention on whether innovation is undercounted in national welfare while avoiding the more dangerous question of who owns the innovation and what happens when labor is no longer economically necessary.

The reported 5.39-to-1 social/private return gap is therefore not reassuring. Under DT logic, it is evidence that private markets can underprice innovations whose benefits—and displacements—are system-wide.

The Verdict

A rigorous welfare-accounting refinement attached to an obsolete battlefield map. It may improve estimates of innovation’s social value, but it does not challenge the Discontinuity Thesis; it reinforces the central danger by showing that private returns can diverge sharply from system-wide effects. When the innovation is AI, correcting the ledger will not save post-WWII capitalism. It will merely document the surplus generated as human productive necessity disappears.