Robots at work? Pitfalls of industry-level data

In their seminal paper, Graetz and Michaels (2018) find that robots increase productivity, lower output prices, and adversely affect the share of low-skilled labor. We demonstrate that these effects are partly driven by the sample composition and argue that focusing on manufacturing industries yields more credible results regarding the overall economic effects of robotization. The results show that focusing on robotizing industries leads to a sizable drop of the productivity effects, halving the effect size for labor productivity. Pronounced consequences from the sample choice occur for wage effects that are reversed from significantly positive into significantly negative. Controlling for demographic workforce characteristics proves to be essential for the significant labor productivity effects and leads to the reversal for wages.