Local Policy Effects in the Presence of Regional and Policy Spill-Overs
Final Report Abstract
Local policies are a very important and widely used instrument in the toolboxes of policymakers worldwide. For instance, most countries set tax rates on income, property or profits at subnational levels. Moreover, subsidies targeted at specific regions, so-called place based policies, are common means to stimulate economic activity in under-developed areas. This project was set out to study the effects of those local policies, while paying special attention to the regional and political spill-overs triggered by them. In order to do so, I exploited various features of German institutional setting, where more than 11,000 small municipalities, can independently make important fiscal decisions such as setting tax rates or increasing spending. In a first paper, we study the spillovers of a regional capital subsidy to East German counties after the reunification. The so-called Gemeinschaftsaufgabe Verbesserung der regionalen Wirtschaftsstruktur (GRW) is Germany’s most important regional policy, mainly targeted at East German manufacturing firms in poorly performing counties. Exploiting quasi-experimental variation in the maximum fraction of covered investment costs, we find that subsidy cuts reduce investment, employment and wages of low-skilled workers. Assuming symmetry, these findings imply that place-based policies can stimulate local economic activity of treated firms. Importantly, we demonstrate substantial spill-overs: 60% of the employment effects are absorbed at the higher regional level of the commuting zone. Moreover, there is suggestive evidence of relocation of employment within county across sectors. These spill-overs can explain why we do not find significant policy effects on county-level GDP or unemployment. Hence, the subsidy is ineffective in boosting under-developed regions. Instead it redistributes incomes within greater regions across counties, industries and worker groups. In the empirical model, we adjust the classic and widely-used event study design to account for multiple treatments of various intensities. In a methodological spin-off paper, we discuss more generally how to adjust the canonical event study design to non-standard institutional settings. In the course of our argument, we also point to common caveats and pitfalls when setting up event study designs and distributed-lag models. We derive the following main results: First, event study designs and distributed-lag models are numerical identical leading to the same parameter estimates after correct reparametrization. Second, binning of the endpoints of the effect window allows identification of both secular time fixed effects and dynamic treatment effects in the presence of unit and time fixed effects even when no never-treated units are present. Third, classic event study designs using dummy variables as event indicators can be naturally generalized to models that account for multiple events of different sign and intensity of the treatment, which are particularly interesting for research in labor economics and public finance. In an ongoing project, I study the regional mobility of workers. Evidence on worker mobility is scarce and typically confined to very specific high-income individuals. However, knowing about the mobility responses of workers across the skill-distribution, including the average worker, is crucial for policy makers, for instance to design optimal policies to alleviate regional inequality, such as place-based policies. The goal of the paper is to provide such empirical evidence on individual mobility across the skill distribution. Exploiting the German institutional set-up with municipalities setting local business tax rates and using a change in local business taxes as a demand shock, I estimate the local labor supply elasticities of various worker groups. First evidence suggests that high-skilled and young workers are relatively more mobile than other worker groups. The two projects studying political spill-overs are still at early stages due to unanticipated data problems and a lack of time.
