AI as Technology Shock: Firm Productivity, Reallocation, and Long-Run Growth

AI, including generative AI, is increasingly viewed as a general-purpose technology with potential implications for productivity growth. Yet its effects may be uneven across firms and may take time to materialise if adoption requires complementary investments in intangible assets (e.g., data, skills, management practices) and organisational change. This project will explore firm-level quantitative evidence on how AI adoption and diffusion reshape productivity dynamics and, through reallocation, affect aggregate productivity and long-run growth.

The project will study a unified mechanism through which AI may affect productivity and long-run growth. It will first examine how AI adoption emerges and diffuses across firms, focusing on the firm characteristics and constraints that shape adoption intensity and timing. It will then estimate the causal effects of AI adoption on firm performance, including productivity, growth, innovation, cost structures, and markups, while accounting for potential implementation lags arising from adjustment costs and complementary investments such as skills, data infrastructure, and organisational change. Finally, the project will assess the aggregate implications of these heterogeneous firm-level effects by analysing how AI adoption reshapes resource allocation across firms, including market-share shifts toward more productive firms, entry and exit dynamics, and changes in productivity dispersion.

The project will take a fully quantitative empirical approach. It will employ firm-level data and credible causal inference strategies, including panel-data methods, event studies, and other quasi-experimental designs. Where appropriate, machine learning and/or NLP may be used to construct transparent measures of AI exposure or adoption intensity and to analyse heterogeneous effects across firms and industries. The emphasis will be on economic mechanisms, credible identification, and replicable empirical analysis rather than on developing frontier AI models.

The candidate will receive training in applied econometrics (causal inference with firm-level data), productivity and firm-dynamics measurement, and practical data skills (Python/R/Stata).

Supervisors

Dr Shang Jiang

Dr Chunping Liu