Nonlinear effects of oil shocks on stock returns: a Markov-switching approach

Abstract

Using Markov-switching models, we investigate whether oil price shocks have nonlinear effects on stock returns. Empirical evidence from a set of international stock indexes suggests that an increase in oil prices has a negative and significant impact on stock prices in one state of the economy, whereas this effect is significantly dampened in another state of the economy. Furthermore, it is shown that changes in oil prices or in oil price volatility do not lead to a higher probability of switching between regimes.

Acknowledgements

I would like to thank the Editor and the anonymous referee for helpful comments and suggestions. Financial support from Dirección Xeral de I+D (Xunta de Galicia) under PGIDIT06PXIB201002PR is gratefully appreciated.

Notes

¹ For example, Jones and Kaul (1996) reported international evidence of a significant negative relationship between oil shocks and stock returns using quarterly data. Sadorsky (1999) showed that US real stock returns were negatively affected by positive shocks to oil price changes or volatility using monthly data. On the other hand, Huang et al. (1996) found no evidence of a relationship between oil future prices and aggregate stock returns, but a significant causality from oil futures returns to individual oil company stock returns. A significant positive relationship between oil prices and oil and gas companies’ returns was reported in Sadorsky (2001) and El-Sharif et al. (2005).

² This empirical framework differs from the multivariate threshold autoregressive model employed in Huang et al. (2005). Instead of assuming the amount of the oil price increase as a threshold variable, the effect of the oil price on stock prices is influenced by an unobserved random variable, namely, the state or regime, which evolves according to a first-order Markov transition process.

³ For example, McQueen and Roley (1993) found that the effect of macroeconomic news on stock prices critically depended on the state of the economy. Schwert (1989) and Hamilton and Lin (1996) showed that the volatility of stock returns was higher during recessions than during expansions. Other authors, such as Chauvet and Potter (2000) and Perez-Quiros and Timmermann (2001), studied model dynamics and nonlinearities in stock market returns employing the Markov regime-switching model. Cyclical variations in stock returns were also identified in Maheu and McCurdy (2000), Edwards et al. (2003), Pagan and Sossounov (2003) and Lunde and Timmermann (2004).

⁴ Mork (1989), Mork et al. (1994), Lee et al. (1995), Hamilton (1996, 2000) and Balke et al. (2002), found that the correlation between oil prices and economic output was statistically significant and negative when oil prices increased, but statistically nonsignificant when oil prices decreased. Raymond and Rich (1996) analysed the nonlinear relationship between oil price shocks and US business cycle fluctuations with a Markov-switching model and reported evidence of nonlinearities.

⁵ Co-movements among series of crude oil prices are analysed in Bentzen (2007) using high-frequency data.

⁶ These tests are extensively explained in Patterson and Ashley (2000), where an easy-to-use toolkit to identify nonlinearities is also developed.

⁷ We applied all the tests directly to the residuals of the linear equation (1) since, as shown by the DW-statistic, there is no serial correlation (an AR(0) model for residuals was selected over any other alternative AR(p) model).

⁸ A more detailed description of this test can be found in Patterson and Ashley (2000).