Macroeconomic Structure and Oil Price Shocks at the Industrial Level

Abstract

This paper analyses the role of the macroeconomic structure in the response of industrial output to an oil price shock in six OECD countries. The modelling of the macroeconomic structure is important in examining the effect of an oil price shock on the industry-level output, since the analysis of the transmission mechanisms helps us to better understand the response of industrial output to such a shock. Thus, cross-country differences found in the responses of industrial output to oil price shocks within the European Monetary Union can be partially explained by differences in the transmission mechanisms of such shocks.

Keywords:

• Oil price shock
• identified VAR
• manufacturing industries

JEL CLASSIFICATIONS :

• E32
• Q43

Acknowledgements

I thank the editor and two anonymous referees for their helpful comments and suggestions. This work is part of a research project financed by the Bank of Spain and initiated during my stay as a postdoctoral research fellow there. The views expressed in this paper are those of the author and do not necessarily reflect the position of the Bank of Spain.

Notes

¹See Hamilton (1983, 1996, 2003), Gisser and Goodwin (1986), Mork (1989), and Hooker (1996), among others, for the US; and Burbidge and Harrison (1984), Mork et al. (1994), Jiménez-Rodríguez and Sánchez (2005, 2009), and Kilian (2008) for countries different from the US.

²We find other authors who analysed the US industry-level effects of oil price shocks on other research topics (creation and destruction of manufacturing job, Davis and Haltiwager, 2001; inventory-sales, Herrera, 2007; and stock returns, Kilian and Park, 2009).

³Bohi (1989) compared the industrial impact of oil prices in Japan, Germany, the UK, and the US, although his comparisons were informal.

⁴Total industrial production is used as a proxy of economic activity. It is worth stressing that this variable and GDP are highly correlated and that the latter is only available at quarterly frequency.

⁵Real effective exchange rate is defined such that an increase means a real appreciation of the currency considered. An appreciation of the real exchange rate is expected to hurt the country's external competitiveness.

⁶The use of non-linear transformations of oil price to analyse the impact of oil price shocks on aggregate output has been theoretically justified by the inter-sectoral adjustments with which they are involved (see Jiménez-Rodríguez & Sánchez, 2005, and references therein). Thus, we do not apply such transformations in our study, given that our main goal is to analyse the impact at the industrial level, and their use may give rise to biased results.

⁷The macroeconomic block includes, apart from total industrial production and real oil prices, other variables to capture some of the most important transmission channels through which oil prices may affect real economic activity indirectly, in part by inducing changes in economic policies. These channels include the effects of oil prices on inflation and exchange rates, which then induce changes in real economic activity. Our macroeconomic block also incorporates a monetary sector that can react to inflationary pressures.

⁸In this paper we do not perform an explicit analysis of the long-run behaviour of the economy. By doing the analysis in levels we allow for implicit cointegrating relationships in the data, and still have consistent estimates of the parameters. For further discussion in this issue, see, for example, Sims et al. (1990), Hamilton (1994), and Ramaswamy and Sløk (1998).

⁹Davis and Haltiwanger (2001) also did the same assumption in their study about the effects of oil price changes and other shocks on the creation and destruction of US manufacturing jobs.

¹⁰The likelihood ratio tests (based on the so-called block-exogeneity test) indicate that the block-recursive restrictions (i.e. R ₁₂(L)=0) cannot be rejected for most industries and most countries. Even in the cases of rejection, the estimate of the variance-covariance matrix of the macroblock, Ω₁₁, is not significantly altered by the presence of the specific industrial output. Thus, we estimate under the restriction that macroeconomic variables are not affected by the specific industrial output.

¹¹It is worth noting that the non-recursive structure (contrary to the recursive one) allows us here to consider the contemporaneous interactions between the interest rate and the exchange rate, and the non-reaction of the interest rate contemporaneously to changes in output and inflation (see Sims & Zha, 1998), as well as the contemporaneous interactions between the interest rate and money stock (see Kim & Roubini, 2000).

¹²Our baseline results are robust to other identifying schemes. An Appendix with these results is available from the author upon request.

¹³Notice that A ₂₁ and Ω₂₁ contain the same numbers of elements, and so we do not impose restrictions. The same situation happens with A ₂₂ and Ω₂₂.

¹⁴The OECD database ‘Indicators for Industry and Services’ is available from 1975:1 to 1998:12 based on International Standardized Industrial Classification (ISIC) Revision 2 and from 1990:1 to 2001:9 based on ISIC Revision 3. We use data from ISIC Revision 2 and do not extend up to 2001:9 given that OECD advises that the two classification systems (ISIC Revision 2 and ISIC Revision 3) are not fully compatible and, thus, comparisons have to be made with caution.

¹⁵As in Jiménez-Rodríguez (2008), we start the Spanish sample in 1980:1, given that the Spanish policy of subsidizing oil prices (implemented up to the end of the 1970s) may have distorted the effects of oil price shocks on the Spanish industries.

¹⁶Notice that the growth tendency of real oil prices from 2002 onwards differs from previous oil price shocks, which were mainly caused by sizeable disruptions to the supply of oil. The factors driving the oil price increases over the last few years were basically caused by demand factors. See Kilian (2009) for further discussion.

¹⁷The finding that higher oil prices adversely affect the UK manufacturing activity is in concordance with the result reported by Jiménez-Rodríguez and Sánchez (2005) for overall economic activity.

¹⁸Jiménez-Rodríguez (2008) considers a recursively identified bivariate VAR model with real oil prices in domestic currency and specific industrial output as variables (entering into the bivariate model in that order), and this model is estimated by the maximum likelihood estimation approach.

¹⁹We have verified that the same conclusion can be obtained by analysing whether the responses resulting from the bivariate model are within the 90% confidence intervals calculated around the responses of our multivariate model. These results are available from the authors upon request.

²⁰It is worth stressing that the macroeconomic block does not lead to more homogeneous output responses across industries in those countries for which cross-industry heterogeneity is found using the bivariate model (France, Germany, and Spain).

²¹The only exception to this is Italy, where the monetary aggregate only lessens after one year.

²²Looking at the forecast error variance decomposition, we observe that the main two sources of volatility (apart from oil prices) for most industries are interest rates and real effective exchange rates in all countries.

²³Authors such as Bohi (1989) and Bernanke et al. (1997) argued that overall economic downturns observed in the US after oil price shocks are caused by a combination of direct impacts of the shocks themselves and the monetary responses to them.

²⁴An Appendix with all these results is available from the author upon request.

²⁵See Jiménez-Rodríguez (2008) for further details regarding the influence of the industrial structure.