Causal relationship between energy consumption (EC) and GDP: A Markov-switching (MS) causality

This paper re-examines the causal relationship between energy use and Gross Domestic Product (GDP) in the United States for the period 1960-2005. To that end, we use Markov-switching vector autoregressive (MS-VAR) models, rather than vector autoregressive (VAR) models, which allows for regime shifts. These models are capable of detecting changes in the relationship between variables; in addition, the coefficients of the model are time dependent and they depend on the states of the variables. Therefore, in contrast to VAR and vector error correction models (VECM), which assume a stable relationship, the relationship between the variables could be different in the separate regimes. Results from the estimation of MS models show changes in the pattern of causality relationship between GDP and energy use. That is, we found evidence of bidirectional Granger causality (GC) between the variables in the first regime, while there is no GC between the variables in the second regime. The first regime consists of 1971-1975,1977-1982, 1989-1995, and from 2001 to the end of the sample. This regime includes the energy crises in 1970s, the recessions in the early 1980s, 1990s, and the recession in 2001. (C) 2011 Elsevier Ltd. All rights reserved.