Co-movements between carbon, energy and financial markets. A multivariate GARCH approach

Mr Nicolas Koch, University of Hamburg

This paper jointly investigates time-variations in carbon-energy market and carbon-financial market co-movements, taking the past six years (2005-2011) of the EU ETS operation as research framework. Our first empirical question focuses on correlation asymmetries under different market uncertainty conditions and asks whether correlations between carbon-energy and carbon-financial markets are exacerbated during episodes of financial turmoil or whether we observe a decoupling of these markets. Our second empirical question centers on structural breaks and asks whether the overall level of correlation between carbon, energy and financial markets changed over time. We argue that structural changes with respect to (i) the market design and (ii) the market maturity of the relatively young carbon market should coincide with changing correlation regimes in the Phase I-to-Phase II period.

We follow Silvennoinen and Teräsvirta (2005, 2009) and adopt the Double Smooth Transition Conditional Correlation (DSTCC) GARCH model that allows the conditional correlation to change smoothly across up to four regimes directly as a function of observable transition variables. We use smooth transition models because they can capture both gradual and sudden changes in correlation patterns, impounding both slowly developing trends (i.e. due to institutional change) and rapid changes in investor expectations (i.e. due to shifts in risk perception). An appealing feature of this model class is that it provides a framework in which constancy of correlations and the existence of links to observable and interpretable economic variables or general proxies for latent factors can be tested in a straightforward fashion.

Our most important findings are as follows: First, we provide convincing evidence for structural breaks in correlations between carbon and energy markets in the run-up to the second phase of the EU ETS starting 2008. In particular, we document that correlations between carbon, on the one hand, and gas, coal and electricity, on the other hand, are four, three and two times as high in Phase II as in Phase I, respectively. Second, we find that carbon and financial markets are not segmented. More importantly, correlation heavily depends on market conditions and the VSTOXX index seems to be a useful state variable that is informative about the uncertainty or risk of common shocks often associated with extreme events that shift correlations. Specially, high expected stock market volatility shifts carbon-stock correlation significantly upwards. In rebuttal, carbon-bond correlation switches to stronger negative correlation. The correlation states corresponding to turbulent periods peak around the Lehman Brother failure and then persist until around mid 2009.