'Nonlinear causality between crude oil price and exchange rate: A comparative study of China and India' - A failed replication (negative Type 1 and Type 2)

Evidence published in this journal by Bal and Rath (2015) purports a bidirectional nonlinear causality between oil price and India's exchange rate and, for China, unidirectional nonlinear causality running from exchange rate to oil price. Their entire testing protocol and ensuing results rest upon claims that all the variables contain a unit root. We raise several critical issues and revisit the order of integration of the series as well as their cointegration and Granger causality properties through a 'pure replication' and a 'reanalysis'. Contrary to Bal and Rath (2015), when we repeat their estimated model with their specification of the Ng and Perron (2001) unit root test on their data, we find that their oil price series (ROL) is level stationary (negative replication Type 1), a result which makes all their subsequent results biased and misleading. Our reanalysis confirms that ROL is 1(0), linearly as well as nonlinearly. We also find that the basic bivariate model proposed by Bal and Rath (2015) fails to produce statistically robust and stable cointegrating patterns. Nonlinear causality tests confirm the absence of any nonlinear causality for both countries (negative replication Type 2). (C) 2016 Elsevier B.V. All rights reserved.