Oil price shocks on Indian economy: evidence from Toda Yamamoto and Markov regime-switching VAR

Abstract

The study investigates the dynamic impact of linear and non-linear specifications of oil price shocks on macroeconomic fundamentals for an oil-importing emerging economy – India – during the period March 1991 to January 2009. The paper deploys extended vector autoregressive (VAR) model of possibly integrated processes proposed by Toda and Yamamoto, which has its advantage of application irrespective of the variables being stationary or cointegrated. The study further estimates two-state Markov regime-switch VAR model to examine regime shift behaviour of the underlying variables and its relationship. The study finds that inflation and foreign exchange reserve are greatly impacted by oil price shocks. The study also confirms that the movement in oil price is exogenous with respect to the movement of India’s macroeconomic variables and the impact of oil price shocks are asymmetric in nature with negative price shocks having more pronounced effect than positive shocks.

Keywords:

• oil price shocks
• India
• asymmetric impact
• Markov regime-switch VAR
• economy

Notes

^1. This is true by the property of k-State Markov chain process. s(t) equals some particular value, say i, depends on the past only through the most recent past value s(t – 1). (Hamilton 1994) Also, p_i1+ p_i2 + … + p_ik = 1.

^2. Note p₁₁ + p₁₂ = 1.

^3. As mentioned in Section 3, these variables have been considered after logarithmic transformations. Also, Table1 suggests though inf is stationary, oil, expand forex are non-stationary in nature and the first difference of these series made them stationary. Hence doil, dexp and dforex have been taken in MRS-VAR. The reason of choosing four variables are twofold: (i) a two-state MRS-VAR estimation with 7 variables would require 128 parameters to estimate, which is infeasible by estimation property as the sample size is 216; (ii) results of T&Y extended VAR causality suggest that oil price shocks have a significant impact only on inflation, exchange rate and foreign exchange reserve.

^4. MS_Regress package is a set of MATLAB codes which estimate, forecast and simulate univariate and multivariate Markov regime-switching models with time-invariant probability.

^5. Similarly, two states for inflation and changes in foreign exchange reserve can be defined as their high and low phases in the study period. Two states of ‘changes in exchange rate’ represent depreciation and appreciation of Indian rupee to US dollar for the underlying data.