Carnegie Mellon University

The Causal Structure of the Vector Autoregression in Economics:
A Case Study: Kevin Hoover

Abstract: The vector autoregression is a system of dynamic equations in which all variables are endogenous. Since the 1980s it has become the workhorse of empirical macroeconomics. The original vector autoregression was a purely associational or reduced-form model. Early critics persuaded macroeconomists that counterfactual analysis required identification (at a minimum) of the causal structure of the contemporaneous variables of the dynamic system. The presentation will illustrate the application of graph-theoretic search methods in identifying the contemporaneous causal order in a case study of the United States macroeconomic/monetary system. It will also illustrate how graph-theoretic search methods can be integrated with other econometric search methods and will point out some key problems in applying casual search that are still in need of resolution.