RESEARCH & PROJECTS
Evidence over assertion.
Selected work connecting consequential questions, quantitative methods and economic judgment. Additional projects will be shared when the evidence is ready.
APPLIED ECONOMIC RESEARCH
Modelling U.S. Labor Productivity in the Age of AI
When does investment in technology begin to show up in productivity?
The question
Technology can change what businesses are capable of doing. But can we see that change in the productivity of the broader U.S. economy, and how long might it take?
What the evidence suggests
Immediate productivity gains were difficult to detect. The evidence was more informative when the analysis allowed economic relationships to change over time and technology-related effects to emerge with a delay.
28% lower prediction error for the best-performing adaptive model compared with a conventional baseline. This is an improvement in forecasting accuracy, not a measured AI productivity gain.
Why it matters
Measuring returns on AI and digital investment requires patience and a view of the surrounding system: complementary technology, workflows, organizational readiness and adoption. The study supports investigating these connections; it does not establish that AI caused the observed changes.
Methods & limitations98 quarters of U.S. data · Comparative forecasting · Careful interpretation
Method. The study compared time-series forecasting approaches, including ARIMA, ARIMAX, VAR, rolling VAR, state-space and structural-break models, using technology investment, R&D, data-processing costs and electricity prices as indicators.
Limits. These are proxies for the technology environment, not direct measures of AI adoption. The late-2019 shift overlaps with pandemic-era disruptions. This Harvard graduate-course project was not peer-reviewed and does not establish that AI caused changes in productivity.