Serializing Energy and Growth Review Paper

I'm updating my 2004 working paper on energy and economic growth to submit to the annual Ecological Economics Reviews published in the Annals of New York Academy of Sciences. I did publish a short version of the working paper as a chapter in the Encyclopedia of Energy but I didn't submit it to a journal because some of my fundamental ideas about energy and growth began to change just after I finished the paper. Following John Quiggin's example, I'm thinking to post here sections of the paper as I finish them. Unfortunately, my title is way boring by comparison. Please comment if you think I've left out key papers and ideas. I will acknowledge you in the final version if I use the material you suggest. The sections will just appear in the order I finish my first draft. Anyway, the first section to be posted is below. There is a large literature in this area and many papers contain reviews of the previous literature, so I only cite the few papers which I think are most critical.

Empirical Testing of the Causal Relationship between Energy and Growth

In this section we look at the empirical evidence on whether there actually is a causal relation between the two variables using studies that are not grounded in a single theory, potential mechanism, or school of thought. The main approach has used reduced form time series models that do not specify structural linkages between energy and output.

Ordinary linear regression or correlation methods cannot be used to establish a casual relation among variables because when two or more totally unrelated variables are trending over time they will appear to be correlated simply because of the shared directionality. Even after removing any trends by appropriate means, the correlations among variables could be due to causality between them or due to their relations with other variables not included in the analysis. Two methods for testing for causality among time series variables are Granger causality tests (Granger, 1969) and cointegration analysis (Engle and Granger, 1987). See Enders (1995) for an accessible introduction to these methods, and Hendry and Juselius (2000) for their application to energy economics. It is now understood that in the absence of cointegration between the variables a Granger causality test on a VAR in levels is invalid. Ohanian (1988) and Toda and Phillips (1993) showed that the distribution of the test statistic for Granger causality in a VAR with non-stationary variables is not the standard chi-square distribution. This means that the significance levels reported in the early studies of the Granger-causality relationship between energy and GDP may be incorrect, as both variables are generally integrated series. If there is no cointegration between the variables then the causality test should be carried out on a VAR in differenced data, while if there is cointegration, standard chi-square distributions apply when the cointegrating restrictions are imposed. Thus testing for cointegration is a necessary prerequisite to causality testing. A Granger causality test using a VAR in first differences will reflect the short-run marginal effect between the variables but is subject to possible omitted variables bias if the lack of cointegration implies that variables essential to cointegration were omitted from the model.

This literature first emerged in the late 1970s. Ozturk (2010) provides an extensive list of energy-growth causality and cointegration studies. Early studies relied on Granger causality tests on unrestricted vector autoregressions (VAR) in levels of the variables, while more recent studies use cointegration methods. Another key characteristic that distinguishes between studies is whether a bivariate model of energy and output or a multivariate framework is used. A third way to differentiate among models is whether energy is measured in standard heat units or whether some type of indexing method is used to account for differences in quality among fuels.

The results of the early studies that tested for Granger causality using a bivariate model were generally inconclusive (Stern, 1993). Where nominally significant results were obtained they indicated that causality runs from output to energy. However, in many cases results differed depending on the samples used, the countries investigated etc.

Stern (1993) tested for Granger causality in a multivariate setting using a vector autoregression (VAR) model of GDP, capital and labor inputs, and a Divisia index of quality-adjusted energy use in place of gross energy use. The multivariate methodology is important because reductions in energy use are frequently countered by the substitution of other factors of production for energy and vice versa, resulting in an insignificant overall impact on output. When both the multivariate approach and the quality adjusted energy index were employed, energy was found to Granger cause GDP.

Most analysts believe that capital, labor, and technical change play a significant role in determining output, yet early studies used only energy as an independent variable. When variables known to be important are omitted from the model, there will be no cointegration and a spurious regression will result. Results are frequently sample dependent in the face of omitted variables and non-cointegration (e.g. Stern and Common, 2001). This may explain the very divergent nature of the early causality literature. On the other hand, Zarnikau (1997) and Oh and Lee (2004) found using bivariate models that a Divisia index of energy use Granger causes GDP in the US and Korea, repectively.

These results are supported by Hamilton (1983) and Burbridge and Harrison (1984), who found that changes in oil prices Granger-cause changes in GNP and unemployment in VAR models whereas oil prices are exogenous to the system. More recently, Blanchard and Gali (2008) used VAR models of GDP, oil prices, wages, and two other price indices, to argue that the effect of oil price shocks has reduced over time. Hamilton (2009) deconstructs these arguments to show that past recessions would have been mild or have merely been slowdowns if oil prices had not risen. Furthermore, he argues that the large increase in the price of oil that culminated in 2008 was a major factor in causing the 2008-2009 recession. Paradoxically, as in the short-run the elasticity of demand for oil and other forms of energy is low, the main short-run effects of oil prices are expected to be through reducing spending by consumers and firms on other goods, services, and inputs rather than through reducing the input of energy to production (Hamilton, 2009; Edelstein and Killian, 2009). Therefore, models using oil prices in place of energy quantities may not provide much evidence regarding the effects of energy use itself on economic growth.

More recent work has used cointegration methods to investigate the E/GDP relationship. The first such study was conducted by Yu and Jin (1992). Again, the results of this and subsequent studies differ according to the regions, time frames, and measures of inputs and outputs used. It would seem that if a multivariate approach helps in uncovering the Granger causality relations between energy and GDP a multivariate approach should be used to investigate the cointegration relations among the variables. When multivariate cointegration methods are used, a picture emerges of energy playing a central role in determining output in a diverse set of developed and developing nations. Stern (2000) investigated the time series properties of GDP, quality weighted energy, labor, and capital series, estimating a dynamic cointegration model using the Johansen methodology. The cointegration analysis showed that energy is significant in explaining GDP. It also showed that there is cointegration in a relationship including GDP, capital, labor, and energy. The multivariate analysis shows that energy Granger causes GDP either unidirectionally or possibly through a mutually causative relationship depending on which version of the model is used. Oh and Lee (2004) and Ghali and El-Sakka (2004) apply Stern’s (1993, 2000) methodology to Korea and Canada, respectively, coming to exactly the same conclusions, extending the validity of Stern’s results beyond the United States. Lee and Chang (2008) and Lee et al. (2008) use panel data cointegration methods to examine the relationship between energy, GDP, and capital in 16 Asian and 22 OECD countries over a three and four decade period respectively. Lee and Chang (2008) find a long-run causal relationship from energy to GDP in the group of Asian countries while Lee et al. (2008) find a bi-directional relationship in the OECD sample. A number of other analysts also demonstrate the role of omitted variables in the energy -income relation (e.g. Glasure, 2002; Hondroyiannis et al., 2002; Masih and Masih, 1997). Taken together, this body of work suggests that the inconclusive results of the earlier tests of Granger causality are probably due to the omission of necessary variables - either the quantities of other inputs (and quality adjustment of the energy input) or energy prices. The marginal product of energy, and hence energy prices, should reflect the quantities of other inputs used together with energy, which Stern (2000) found were necessary to achieve cointegration.


References
Blanchard, O. J. and J. Galí (2008) The macroeconomic effects of oil price shocks: why are the 2000s so different from the 1970s? In: J. Galí and M. J. Gertler (eds.) International Dimensions of Monetary Policy, University of Chicago Press.
Burbridge, J. and A. Harrison (1984). “Testing for the effects of oil prices rises using vector autoregressions.” International Economic Review 25: 459-484.
Edelstein, P. and L. Kilian (2009) How sensitive are consumer expenditures to retail energy prices? Journal of Monetary Economics 56(6): 766-779.
Enders, W. (1995). Applied Econometric Time Series. New York: John Wiley.
Engle, R. E. and C. W. J. Granger (1987). “Cointegration and error-correction: representation, estimation, and testing.” Econometrica 55: 251-276.
Ghali, K. H. and M. I. T. El-Sakka (2004) Energy use and output growth in Canada: a multivariate cointegration analysis, Energy Economics 26: 225-238.
Glasure, Y. U. (2002). “Energy and national income in Korea: Further evidence on the role of omitted variables.” Energy Economics 24: 355-365.
Granger, C. W. J. (1969). “Investigating causal relations by econometric models and cross-spectral methods.” Econometrica 37: 424-438.
Hamilton, J. D. (1983). “Oil and the macroeconomy since World War II.” Journal of Political Economy 91: 228-248.
Hamilton, J. D. 2009. Causes and Consequences of the Oil Shock of 2007–08. Brookings Papers on Economic Activity. 2009(1): 215-261.
Hendry, D. F., K. Juselius (2000). “Explaining cointegration analysis: Part 1.” Energy Journal 21(1): 1-42
Hondroyiannis, G., S. Lolos, and E. Papapetrou (2002). “Energy consumption and economic growth: assessing the evidence from Greece.” Energy Economics 24: 319-336.
Lee, C.-C. and C.-P. Chang (2008) Energy consumption and economic growth in Asian economies: A more comprehensive analysis using panel data, Resource and Energy Economics 30(1): 50-65.
Lee, C.-C., C.-P. Chang, P.-F. Chen (2008) Energy-income causality in OECD countries revisited: The key role of capital stock, Energy Economics 30: 2359–2373.
Masih, A. M. M. and R. Masih (1997). “On the temporal causal relationship between energy consumption, real income, and prices: some new evidence from Asian-energy dependent NICs based on a multivariate cointegration/vector error-correction approach.” Journal of Policy Modeling 19: 417-440.
Oh, W. and Lee K. (2004). “Causal relationship between energy consumption and GDP revisited: the case of Korea 1970–1999.” Energy Economics 26: 51–59.
Ohanian, L. E. (1988). “The spurious effects of unit roots on vector autoregressions: A Monte Carlo study.” Journal of Econometrics 39: 251-266.
Ozturk, I. (2010) A literature survey on energy-growth nexus, Energy Policy 38: 340-349.
Stern, D. I. (1993). “Energy use and economic growth in the USA, A multivariate approach.” Energy Economics 15: 137-150.
Stern, D. I. (2000). “A multivariate cointegration analysis of the role of energy in the U.S. macroeconomy.” Energy Economics 22: 267-283.
Stern, D. I. and M. S. Common (2001). “Is there an environmental Kuznets curve for sulfur?” Journal of Environmental Economics and Management 41: 162-178.
Toda, H. Y. and P. C. B. Phillips (1993). “The spurious effect of unit roots on vector autoregressions: an analytical study.” Journal of Econometrics 59: 229-255.
Yu, E. S. H. and J. C. Jin (1992). “Cointegration tests of energy consumption, income, and employment.” Resources and Energy 14: 259-266.
Zarnikau, J. (1996) Reexamination of the causal relationship between energy consumption and gross national product, Journal of Energy and Development 21: 229-239.