Artificial Intelligence Technology, Task Attribute and Occupational Substitutable Risk: Empirical Evidence from the Micro-level

Abstract

 Artificial intelligence (AI) technology has emerged as a new general-purpose technology (GPT) in recent years. However, the impact of AI technology on firm productivity, employment, and workforce composition is not well understood. This study uses a micro-level panel dataset of Taiwanese electronics firms, which are listed on the Taiwan Stock Exchange (TSE) or the Over-the-Counter (OTC) market for the period 2002-2018. We employ the keyword-matching method to identify AI-related patent classifications, used patents capturing AI innovations, and match-listed electronics firms to AI patents to construct a panel dataset. Empirical estimations illustrated that AI technology is significantly and positively associated with firm productivity. We also adopt various techniques of the generalized method of moments (GMM) for the dynamic panel data model to deal with endogeneity and obtain similar results. Our analyses may yield useful implications for R&D and labor policies. We also describe how our measures can be useful to researchers and policy‐makers interested in identifying the effect of AI on markets.