Sectoral spillovers in the macroeconomy
Macroeconomists are typically concerned with changes in aggregate economic activity. Traditionally, the drivers of these aggregate fluctuations are considered to also be of aggregate nature, such as productivity, aggregate demand or monetary policy shocks. Most of the literature on economic fluctuations outright disregards the possibility of microeconomic shocks to specific sectors or firms to generate relevant aggregate fluctuations. A common line of reasoning suggests that with a large number of sectors, idiosyncratic shocks average out so that in the aggregate no significant fluctuations occur.
The recent COVID-19 pandemic however shows that this argument does not always apply: sectoral shocks causing supply chain disruptions—such as in the semi-conductor industry—became prominent macroeconomic narratives to explain the business cycle. These types of disruptions or shocks originate in some parts of the economy and then spill over to other sectors and generate effects that become of macroeconomic relevance.
Already prior to the COVID-19 pandemic, a growing literature has shown that such idiosyncratic shocks could explain a more significant share of observed aggregate fluctuations than traditionally thought. One of the main challenges this literature faces is to empirically distinguish sectoral shocks with aggregate consequences from truly aggregate shocks, and to use these insights for causal inference. This is especially relevant for policy analysis, as a clear determination of shocks' origins is crucial for policy-makers to design appropriate policies.
However, most empirical quantifications that measure the contributions of sectoral shocks to real or nominal fluctuations typically rely on theoretical modeling choices and calibration. The common approach is to embed multiple sectors into standard macroeconomic general equilibrium models and then empirically analyze the macroeconomic importance of sectoral shocks within this narrowly defined theoretical framework. This is problematic because theoretical models necessarily involve lots of simplifying assumptions. Ideally, the evidence explaining the sources of aggregate fluctuations should not hinge upon those. Hence, a large part of this dissertation deals with the challenge to disentangle sectoral shocks from other drivers of aggregate fluctuations without relying on too much economic theory or calibration. The aim is to provide empirical answers that guard against theoretical misspecification concerns, yet still enable causal inference.
The main idea behind the identification strategy used throughout this dissertation is as follows: network data (e.g. an input-output table) contains strong information on how idiosyncratic, sector-specific shocks propagate cross-sectionally. The propagation pattern of an idiosyncratic shock differs to those of other shocks depending on the sector where it originates, since every sector connects differently to the network. This heterogeneity manifests in sector rankings that rank sectors from most to least affected. Different sectoral shocks are associated with different sector rankings. This allows to distinguish sector-specific shocks from one another, but also from aggregate shocks. Aggregate shocks can be seen as a combination of sector-specific shocks. This combination leads to a mixture of cross-sectional patterns that arise upon individual idiosyncratic shocks. As a consequence, aggregate shocks exhibit cross-sectionally distinct responses to sector-specific ones. This distinction breaks the observational equivalence between sectoral shocks with aggregate consequences and aggregate shocks.
Moreover, the cross-sectional rankings associated with individual sector-specific shocks are often robust to variations in model specifications and calibration. In other words, different theoretical models can imply very different quantitative effects that sectoral shocks have cross-sectionally. I however show that sectoral shocks have similar effects in how sectors are relatively affected. The ranking of sectors that are most affected by a given sectoral shock is often the same across different theoretical models. This key insight is why sector rankings are powerful restrictions for empirically identifying sector-specific shocks without relying on the specifics of a theoretical model. These sector rankings are hence crucial concepts used within this dissertation.