Source: Bäck et al., “Return of the Solow-paradox in AI? AI-adoption and firm productivity,” Center for Innovation Research, January 2022.
Commentary: Previous research quantifying the relationship between AI adoption and firm productivity has been sparse. Due in part to the novelty of the technology and a lack of historical data, opponents of AI and robotization interpret the available literature to suggest that AI adoption shares little relationship to productivity growth. However, a recent study from the Center of Innovation Research at Lund University overcomes previous issues with data availability on AI implementation by analyzing job advertisements seeking candidates with AI-related skills as a way to approximate firms implementing AI technologies. With proxy data quantifying the AI content of firms, their model matches ORBIS firm-level data on productivity to assess the impact of AI adoption on productivity.
Their analysis over panel data from 2016 to 2019 identified a strongly positive, statistically significant relationship between AI and productivity, while simultaneously identifying nuances that help explain previous literature challenges to quantify this relationship. The positive impact on firm productivity associated with AI adoption became more pronounced in later years, indicating that previous sparseness in the literature is explained by the immaturity of AI technology that had not yet fully developed its productivity-enhancing benefits. Time-lag implementation in the regression also showed a delay of three years between the point of AI adoption and return on productivity growth, indicating an investment delay for AI that could also explain previous literature gaps. This report illuminates that investments in AI do in fact carry a positive impact on productivity and provides useful context to dispel misconceptions of AI’s utility in today’s labor market.
