Get access

Causal Inference by Independent Component Analysis: Theory and Applications


  • We thank conference/seminar participants at Universite des Antilles et la Guyane, University of Pisa, University of Sussex, Universitat Rovira i Virgili, Scuola Superiore Sant'Anna, London School of Economics, as well as Jonathan Temple (the editor) and two anonymous referees for useful comments. All remaining errors are our own. Alex Coad gratefully acknowledges financial support from the ESRC, TSB, BIS and NESTA on IRC grants ES/H008705/1 and ES/J008427/1, and from the AHRC (FUSE project). Doris Entner and Patrik Hoyer were supported by the Academy of Finland project #1125272.


Structural vector-autoregressive models are potentially very useful tools for guiding both macro- and microeconomic policy. In this study, we present a recently developed method for estimating such models, which uses non-normality to recover the causal structure underlying the observations. We show how the method can be applied to both microeconomic data (to study the processes of firm growth and firm performance) and macroeconomic data (to analyse the effects of monetary policy).