Roebroeck Seth Valdes-SosaV. Walsh and A. Cowey. Transcranial magnetic stimulation and cognitive neuroscience.Nat Rev Neurosci, 1(1):73–9, 2000.W. W. S. Wei. <strong>Time</strong> <strong>Series</strong> Analysis: Univariate and Multivariate Methods. Addison-Wesley, Redwood City, 1990.Halbert White and Xun Lu. Granger causality and dynamic structural systems. Journalof F<strong>in</strong>ancial Econometrics, 8(2):193–243, 2010.N. Wiener. The theory of prediction. In E.F. Berkenbach, editor, Modern Mathematicsfor Eng<strong>in</strong>eers. McGraw-Hill, New York, 1956.F. Wies<strong>in</strong>ger, P. F. Van de Moortele, G. Adriany, N. De Zanche, K. Ugurbil, and K. P.Pruessmann. Potential and feasibility of parallel mri at high field. NMR Biomed, 19(3):368–78, 2006.E. Yacoub, N. Harel, and K. Ugurbil. High-field fmri unveils orientation columns <strong>in</strong>humans. Proc Natl Acad Sci U S A, 105(30):10607–12, 2008.104
JMLR: Workshop and Conference Proceed<strong>in</strong>gs 12:95–114, 2011<strong>Causality</strong> <strong>in</strong> <strong>Time</strong> <strong>Series</strong>Causal Search <strong>in</strong> Structural Vector Autoregressive ModelsAlessio MonetaMax Planck Institute of EconomicsJena, GermanyNad<strong>in</strong>e ChlaßFriedrich Schiller University of Jena, GermanyDoris EntnerHels<strong>in</strong>ki Institute for Information Technology, F<strong>in</strong>landPatrik HoyerHels<strong>in</strong>ki Institute for Information Technology, F<strong>in</strong>landmoneta@econ.mpg.denad<strong>in</strong>e.chlass@uni-jena.dedoris.entner@cs.hels<strong>in</strong>ki.fipatrk.hoyer@hels<strong>in</strong>ki.fiEditors: Flor<strong>in</strong> Popescu and Isabelle GuyonAbstractThis paper reviews a class of methods to perform causal <strong>in</strong>ference <strong>in</strong> the framework ofa structural vector autoregressive model. We consider three different sett<strong>in</strong>gs. In thefirst sett<strong>in</strong>g the underly<strong>in</strong>g system is l<strong>in</strong>ear with normal disturbances and the structuralmodel is identified by exploit<strong>in</strong>g the <strong>in</strong>formation <strong>in</strong>corporated <strong>in</strong> the partial correlationsof the estimated residuals. Zero partial correlations are used as <strong>in</strong>put of a searchalgorithm formalized via graphical causal models. In the second, semi-parametric,sett<strong>in</strong>g the underly<strong>in</strong>g system is l<strong>in</strong>ear with non-Gaussian disturbances. In this casethe structural vector autoregressive model is identified through a search procedurebased on <strong>in</strong>dependent component analysis. F<strong>in</strong>ally, we explore the possibility ofcausal search <strong>in</strong> a nonparametric sett<strong>in</strong>g by study<strong>in</strong>g the performance of conditional<strong>in</strong>dependence tests based on kernel density estimations.Keywords: Causal <strong>in</strong>ference, econometric time series, SVAR, graphical causal models,<strong>in</strong>dependent component analysis, conditional <strong>in</strong>dependence tests1. Introduction1.1. Causal <strong>in</strong>ference <strong>in</strong> econometricsApplied economic research is pervaded by questions about causes and effects. Forexample, what is the effect of a monetary policy <strong>in</strong>tervention? Is energy consumptioncaus<strong>in</strong>g growth or the other way around? Or does causality run <strong>in</strong> both directions? Areeconomic fluctuations ma<strong>in</strong>ly caused by monetary, productivity, or demand shocks?Does foreign aid improve liv<strong>in</strong>g standards <strong>in</strong> poor countries? Does firms’ expenditure<strong>in</strong> R&D causally <strong>in</strong>fluence their profits? Are recent rises <strong>in</strong> oil prices <strong>in</strong> part caused byspeculation? These are seem<strong>in</strong>gly heterogeneous questions, but they all require someknowledge of the causal process by which variables came to take the values we observe.c○ 2011 A. Moneta, N. Chlaß, D. Entner & P. Hoyer.