Causality Testing and Data Quality: Effects of Error-Induced Misspecification

The paper examines the effects of measurement error on a well-known causality-testing procedure, that due to Pierce and Haugh. This is shown to be extremely sensitive to small random variations, which may induce the appearance of, as well as obscure, causality patterns, through causing misspecification of ARIMA filters. Spectral-analytic techniques are used to explain why errors may induce "causality", and why the quality of economic data is likely to lead to misleading results.