State-space risk measurement: an application to renewable energy returns

Abstract

This paper uses state-space methodology for modelling excess returns, risk and dynamics for the WilderHill New Energy Index (NEX). The NEX is a global exchange-traded index for investment in development, production and efficiency of renewable energy. It currently lists 98 companies located in 21 countries; the total capitalization of the index is about 285 billion US$ (www.nexindex.com). The NEX has experienced a substantial growth in the last decade along with the rapid development of the renewable energy sector. According to UNEP (2010) estimations, the total amount of (public and private) new investment in renewable energy increased from 46 billion US$ in 2004 to about 162 billion US$ in 2009. As a result, renewable power generation capacity has increased from about 4% of total power generation to nearly 7% between these two dates. Along with this long-run positive trend, the NEX has been able to offer high returns. Naturally, these returns have been associated with high risk exposure. For instance, the index suffered substantial turbulence between 2007 and the end 2009. This paper is aimed at bringing a deeper understanding of the fundamentals that underpin this behaviour. The models is this paper considers various fundamentals that have been associated with the NEX in various reports. The analysis is carried with weekly data between week 33 in 2001 and week 12 in 2011.This study reports work in progress on two different state-space methodologies for assessing the returns and risk for the NEX. First, I use a multi-factor state-space model with time-varying coefficients to analyze the impact of different fundamental variables on the NEX. This first approached was applied to monthly data in Inchauspe, Ripple & Trueck (2011) and presented at the 34th International Conference by the International Association for Energy Economics. Expanding the research to weekly data suggests problems in the robustness of that model. To remedy this, I propose using a Markov-switching (MS) model that allows for regime inference and dynamic analysis. The MS model is written in a mean-adjusted form. This mean, as well as the covariance matrix, are allowed to shifts over time. This specification allows for assessing the significance of exogenous variables after allowing for shifts in mean NEX returns. As regime shifts in NEX excess returns are associated with a positive trend in the NEX levels, the regimes are labelled as “bull” or “bear” markets that cannot be explained by fundamentals.Earlier literature has proposed using state-space methodology to measure time-varying beta factors in capital asset pricing specification (e.g. Bolleshev, Engle and Wooldridge, 1988; Koopman et al., 2008; van Geloven and Koopman, 2009; Tsay, 2005, p.577). The first model borrows from this approach to specify a multifactor model with time-varying coefficients. In addition, a considerable amount of literature has used Markov-switching models to study univariate dynamics of “bull” and “bear” markets in stock market indices (Gordon and St-Amour, 2000; Maheu and Curdy, 2000; Lunde & Timmerman, 2004, Edwards et al., 2003; Girardin and Liu, 2003; Pagan & Soussonov, 2003). I propose studying the dynamics of possible bull/bear markets after relevant exogenous fundamentals are included in the analysis; this approach has also been popular in the literature (Chen, 2009; Chang, 2009; Guidolin & Timmermann, 2005; Chauvet & Porter, 2001). The Markov-switching model in the second part of this paper allows for distinguishing four distinctive distributions associated with abnormal returns and variances. The information obtained from this estimation is valuable for analysts and investors considering medium-run long positions and provides insights into mean-reverting properties of NEX excess returns.