Robots at Work? Pitfalls of Industry Level Data

Research question(s): Does the inclusion of non-robotising sectors lead to a bias in the estimation of robotisation effects on productivity, prices and labour market outcomes when analysing industry level data? • Approach: Empirical estimation using OLS, Panel-Data-, and Instrumental Variable Regression. • Data: Country-industry level data on robot adoption from the International Federation of Robotics (IFR) merged with EU-Klems-data on productivity, prices, and labour market outcomes. • Main result(s): Restricting the estimation sample to sectors which actually experience relevant changes in robotisation (i.e. removing non-robotising sectors) leads to firstly, a reduction in the size of the estimated productivity effects, secondly, no significant effect on prices, and thirdly, drastically changes the estimated employment and wage effects.