Granular Instrumental Variables

Abstract

In many settings, there is a dearth of instruments, which hampers economists’ ability to investigate causal relations. We propose a quite general way to construct instruments: “granular instrumental variables” (GIVs). In the economies we study, a few large firms or countries account for a large share of economic activity. As they are large, their idiosyncratic shocks affect aggregate outcomes. This makes those idiosyncratic shocks valid instruments for aggregate shocks. We provide a methodology to extract idiosyncratic shocks from the data, this way creating GIVs. Those GIVs allow us to then estimate parameters of interest, including causal elasticities.

We first illustrate the idea in a basic supply and demand framework: we achieve a novel identification of supply and demand elasticities, based on idiosyncratic shocks to supply or demand. We then show how the procedure can be adapted to handle many enrichments. We provide initial illustrations of the procedure with two applications. First, we measure how shocks to domestic banks causally affect sovereign yields. We document how negative shocks to Italian banks adversely affect Italian government bond yields, and vice-versa. This gives the first causal measure of the “doom loop” between banks and sovereign yields. Second, we estimate short-term supply and demand elasticities in the oil market. Our estimates match well existing estimates that use much more complex and labor-intensive (e.g., narrative) methods.

We sketch how GIVs could be useful to estimate a host of other causal parameters in economics, particularly in aggregate macro-finance contexts where instruments are usually very rare.