Automation and unemployment: Does collective bargaining moderate their association?

The stock of robots used in industrial production in the OECD more than doubled over the last two decades. Empirically, the direction of the association between automation and (un-)employment varies across countries. Which factors explain this cross-country variation? We argue that differences in collective bargaining systems play a role. We structure the collective bargaining systems of 37 OECD and EU countries by the degree of coordination of their collective bargaining on the one hand, and by the strength of labor unions on the other hand. These results in four types of collective bargaining systems: highly coordinated with strong unions; highly coordinated with weak unions; weakly coordinated with strong unions and weakly coordinated with weak unions. We use a dynamic panel data approach to investigate whether the association between increased automation and the unemployment rates of different societal groups differs across collective bargaining systems. Our findings are consistent with the view that increased automation is positively associated with unemployment in countries where collective bargaining is weak. In coordinated systems the association is muted, notably for workers with medium skill levels, that is, for the group of workers which is frequently seen to be especially prone to be “automated away”. We cannot unveil indications of insider-outsider behavior of labor unions.