We combine a large administrative panel dataset for the German labor market with patent data from PatStat to study the causal effects of innovation on labor market outcomes at the local-sectoral level from 2004 to 2019. We do so by building an instrument for local-sectoral trends in innovation that exploits the network of patent citations. According to our estimates, in innovative sectors, the average wage increases while the dispersion of the distribution decreases. This is reflected by a consolidation of routine workers' wages in more innovative sectors. We hence investigate how labor market institutions, namely Collective Bargaining Agreements (CBA), affect innovation returns on wages at the micro-level. In particular, we exploit the decentralized structure of the German collective bargaining system. By exploiting the regional variation of innovation within the same industry, we find that routine workers affected by bargained wage rises also benefit from innovation returns, while workers in other occupations are not responsive to innovation through the channel of collective bargaining.

Since its conception, relatedness has emerged as a ductile framework to study countries’ and regions’ productive and technological capabilities. More recently, this framework has faced a growing number of critiques, both for its policy implications, and its conceptual outlines. Scholars have been largely scrutinizing relatedness for its theoretical and methodological limitations (Whittle and Kogler, 2020; Bathelt and Storper, 2023). This work explores the feasibility of a Shift-Share Instrumental Variable (SSIV) design in the relatedness density framework. On the one hand, it establishes a solid empirical strategy for relatedness analysis. On the other hand, it redefines relatedness as a measure of exposure, reframing relatedness theory in the light of such intuition. The relatedness density of technological sector in a region, indeed, represents how exposed it is to aggregate shocks in technological coevolution. This work assesses whether and under which assumptions we can identify the effects of relatedness density using an exogenous-shifts SSIV design. We use EPO patent data on 635 technological classes (IPC 4-digit codes). First, we underline the necessary identification assumptions for such a SSIV and a set of balancing tests on the technologies’ characteristics. We, hence, identify quasi-experimental variations from exogenous shifts, by leveraging cooccurrences of technologies across patents in a different technological space. We conclude that our proposed SSIV is relevant and exogenous and can be used to rigorously analyze a relatedness density framework.