Closing the Loop: Distribution, Consumption, and Macroeconomic Closure under Transformative AI
Year: 2026
Type: article
Abstract: Task-based AI scenarios can map automation into productive capacity, factor shares, wages, and labor reallocation without uniquely determining current expenditure or realized activity when income shifts sharply across households and factors. This paper develops an empirically disciplined closure diagnostic for the 2030 scenarios in Korinek and others (2026). A corrected accounting decomposition shows that, in their Extreme scenario, aggregate labor income is almost unchanged relative to no AI even though cognitive labor income falls sharply, because gains to other workers nearly offset the loss. Using observed ranges for the transmission of persistent income shocks, cash-spending responses, recent corporate payout ratios, Federal Reserve ownership data, and household leverage benchmarks, I report deterministic rather than probabilistic sensitivity sets. The first-round current-consumption response relative to no AI is positive in the Modest scenario but not sign-identified in the Extreme scenario. Holding AI output fixed, however, the shift from labor to capital robustly lowers current consumption within the calibrated envelope. A minimal goods-market closure shows why investment, monetary policy, and external demand are necessary to translate either result into output. The central conclusion is therefore about identification: a task model can identify a production frontier without identifying the macroeconomic path used to reach it. The replication package (code, input tables, outputs, figures) reproduces all reported quantitative results: run python code/01_empirical_closure.py and python code/checks.py.
Author: Wyszomirski, Paweł
Institution: WSB University
Language: English
Topic: Economic Theory and Policy
Domain: Social Sciences
Open Access status: green
OpenAlex ID: W7234262901
