The Productivity J-Curve

In the AI Story

Brynjolfsson articulated the J-curve framework most fully in collaboration with Daniel Rock and Chad Syverson, drawing on historical analogies (particularly the electrification of American industry) and contemporary data on information-technology investments. The framework explains why the productivity statistics showed the 1987 paradox during early computerization and then the late-1990s acceleration: the early period was the dip phase, the acceleration was the rise phase.

The complementary investments the framework identifies include organizational restructuring, training and human capital development, process redesign, and the development of new business models that capitalize on the technology's capabilities. These investments are costly and produce their returns only when the technology is mature enough, and the organization adjusted enough, to exploit them. During the investment phase, measured productivity actually declines — firms are spending resources on adjustment rather than production — even though the underlying capability is being built.

For AI, the J-curve framework predicts a multi-year lag between deployment and aggregate productivity effect. Organizations are currently in the dip phase: investing in AI tools, restructuring workflows, training workers, and absorbing the organizational disruption that AI deployment requires. The aggregate productivity surge the framework predicts will arrive in several years — by most estimates, between 2027 and 2032 — and its magnitude will depend on how successfully organizations complete the complementary investments.

Schor's framework engages the J-curve in a specific way: the dividend's arrival is delayed behind capability deployment, meaning that the institutional window for capturing it is narrower than raw productivity projections suggest. By the time aggregate productivity rises enough to make the dividend unambiguously visible in national accounts, the work-spend cycle will have substantially absorbed it through expanded output and intensified work. The window for institutional redesign is therefore not the period after the surge arrives, but the current dip phase — the period during which the complementary investments are being made and organizational patterns that will persist into the surge are being established.

Origin

Developed most fully in Erik Brynjolfsson, Daniel Rock, and Chad Syverson, "The Productivity J-Curve: How Intangibles Complement General Purpose Technologies," NBER Working Paper (2018, revised 2020).

Drawing on the historical productivity research of Paul David, Alexander Field, and Robert Gordon on electrification and previous general-purpose technologies.

Key Ideas

Characteristic temporal pattern. Dip during complementary investment phase, sharp rise as investments mature and exploitation begins.

Complementary investments. Organizational restructuring, training, process redesign, business model development — costly in the short term, critical for long-term gains.

Historical precedent. Documented for electricity and information technology; structurally predicted for AI based on technology's general-purpose character.

AI delayed surge. Aggregate productivity effect expected in 2027–2032, with magnitude depending on organizational adjustment quality.

Narrowed institutional window. Dividend capture must occur during the current dip phase, not after the surge arrives, because the work-spend cycle absorbs the surge as it emerges.

Debates & Critiques

Critics argue that AI may not follow the J-curve pattern — that its productivity effects may be more concentrated (in specific sectors or specific labor categories) and less broadly diffused than electricity or IT. Others argue that AI's complementary investments are in some ways easier (lower organizational restructuring cost) and in some ways harder (greater uncertainty about appropriate workflows) than previous general-purpose technologies.