Multivariate Volatility In Environmental Finance aHoti, S., aM. McAleer and bL.L. Pauwels aSchool of Economics and Commerce, University of Western Australia bEconomics Section, Graduate Institute of International Studies, Geneva, Email: [email protected] Keywords: Environmental risk; financial risk; multivariate conditional volatility; persistence; shocks; symmetry; spillovers. EXTENDED ABSTRACT In this paper, we analyse empirically the time- varying conditional variance (or risk) associated There exist several important benchmark indices in with investing in leading sustainability-driven firms environmental finance, some computed by well- using multivariate models of conditional volatility. known financial index providers such as the Dow As the concept of environmental risk has had Jones group or the FTSE group, and others by several different interpretations in the economics independent agencies specializing in environmental literature, we use the definition given in Hoti, and ethical issues in finance. The main feature of McAleer and Pauwels (2005a): these sustainability indices is that they are constructed from a selection of financial stocks “Environmental risk is the volatility according to sustainable economic, environmental, associated with the returns to a variety of social and ethical criteria. The resulting environmental sustainability indexes.” sustainability indices are meant to be representative of the diversity of industries and size of firms in the Models of the conditional variance, or risk, of a market, at the national, regional and international time series have long been popular in the financial level. This paper builds on earlier empirical work econometrics literature. Three of the most popular investigating conditional volatility or risk inherent in models to capture the time-varying volatility in two major financial time-series indices featuring financial time series are the Generalised ethical and environmental screening. Moreover, the Autoregressive Conditional Heteroscedasticity trends and volatility of two prominent financial (GARCH) model of Engle (1982) and Bollerslev indexes, namely DJIA and S&P500, are analysed in (1986), the GJR model of Glosten, Jagannathan and the same manner to provide a comparison of the time Runkle (1992), and the Exponential GARCH series performance of the two types of indexes. We (EGARCH) model of Nelson (1991). Multivariate examine symmetric and asymmetric effects of extensions of GARCH models are also available in shocks at the multivariate level, and we investigate the literature, such as the Constant Conditional the presence and the importance of multivariate Correlation (CCC) GARCH model Bollerslev effects in conditional volatility in each of these (1990), Vector Autoregressive Moving Average indices as a way to analyse their relative inherent GARCH (VARMA-GARCH) model of Ling and risk. We further investigate empirically the existence McAleer (2003), and VARMA Asymmetric of risk spillovers across these four indexes. GARCH (VARMA-AGARCH) model of Hoti, Chan and McAleer (2002). Environmental issues have become increasingly important in economic research and policy for To date there seem to have been only a few sustainable development. Such issues are tracked by empirical studies of such sustainability indexes. It is the Dow Jones Sustainable Indexes (DJSI) and only recently that time-varying models of Ethibel Sustainability Index (ESI) through financial heteroscedasticity have been applied to market indexes that are derived from the Dow Jones sustainability indexes (see Hoti, McAleer and Global Indexes and Standard & Poor’s (S&P). The Pauwels (2005a)). The plan of the paper is as environmental sustainability activities of firms are follows. Section 1 presents the environmental assessed using criteria in three areas, namely sustainability indexes, namely Dow Jones economic, environmental and social. Risk (or Sustainability and the Ethibel Sustainability Index uncertainty) is analysed empirically through the use and discusses their key features. Multivariate of conditional volatility models of investment in conditional volatility models for daily indexes are sustainability-driven firms that are selected through presented in Section 2. The data are described in the DJSI and ESI (for further details see Hoti, Section 3, and the empirical results are analysed in McAleer and Pauwels (2005a,b)). Section 4. Some concluding remarks are given in Section 5. 2225 1. SUSTAINABILITY INDEXES approximate the sector weights of the S&P Global 1200, such that each regional sub-component 1.1 Dow Jones Sustainability Indexes (DJSI) accounts for a share of ESI Global. Specifically, Asia-Pacific accounts for 11% in ESI Global, the Dow Jones Sustainability Indexes (DJSI) Americas 57%, and Europe 32%. commenced in 1998, and report on the financial performance of leading sustainability-driven firms The ESI Global tracks 162 companies in 19 worldwide (for a discussion of the DJSI indexes, see different countries. In the same manner as the Dow Hoti, McAleer and Pauwels (2005a,b)). These Jones Sustainability Indexes (DJSI), the Ethibel sustainability indexes were created by the Dow Jones Sustainability Indices are calculated as both price Indexes, STOXX Limited and the SAM group. and returns indices in USD and EUR, yielding a total of 16 indices (further information about the The main purpose of the DJSI is to provide asset ESI and the regional ESI can be found at managers with a benchmark to manage sustainability www.ethibel.org). portfolios, and develop financial products and services that are linked to sustainable economic, The selection of companies is based on a set of environmental and social criteria. DJSI indexes positive criteria which examine the best-in-sector quantify the development and promotion of and best-in-region companies. The screening and sustainable values on the environment and society by research methodology concentrates on two the business community. They also enable the elements of corporate social responsibility, namely: promotion of sustainability within the private sector (1) sustainable development, guiding the research by informing investors about firms that behave in an on a specific company over environmental, internal environmentally sustainable manner. and external social, economic and ethical aspects and policy; and (2) stakeholder involvement, which As for the Dow Jones Global Indexes, the DJSI translates to dialogue between the ESI and the features the same methods for calculating, reviewing stakeholders during the research. and publishing data. The DJSI is used in 14 countries, with 50 licenses having been sold to asset 2. MULTIVARIATE MODELS OF managers. There are two sets of DJSI indexes, CONDITIONAL VOLATILITY FOR namely the DJSI World and the DJSI STOXX SUSTAINABILITY INDEXES (which is a pan-European index). The latter index is also subdivided into another regional index, namely The primary empirical purpose of the paper is to DJSI EURO STOXX, which accounts solely for model the DJSI and ESI indexes and their Euro-zone countries. Dow Jones Sustainability associated volatility for the period 31 December World Index (DJSI World) is constructed by 1997 to 1 September 2005. This approach is based selecting the leading 10% of sustainability firms on Engle’s (1982) development of time-varying (which number more than 300) in the Dow Jones volatility (or uncertainty) using the autoregressive Global Index, which covers 59 industries over 34 conditional heteroskedasticity (ARCH) model, and countries. The composite DJSI World is available in subsequent developments associated with the four specialised subset indexes, which exclude ARCH family of models (see, for example, the companies that generate revenue from (1) tobacco, recent survey by Li, Ling and McAleer (2002)). Of (2) gambling, (3) armaments or firearms, and (4) the wide range of univariate conditional volatility alcohol, in addition to the three previously models, the two most popular have been the mentioned items. symmetric generalised ARCH (GARCH) model of Bollerslev (1986) and the asymmetric GARCH (or The DJSI World is reviewed annually and quarterly GJR) model of Glosten, Jagannathan and Runkle to ensure consistency. It also accommodates (1992), especially for the analysis of financial data. potential changes in the behaviour and status of Several other theoretical developments have companies which could affect their sustainability recently been suggested by Wong and Li (1997), performance (such as bankruptcies, mergers and Hoti, Chan and McAleer (2002), Ling and McAleer takeovers). The index comprises companies from 60 (2002a,b) and Ling and McAleer (2003). A industry groups and 18 market sectors. comparison of the structural and statistical properties of alternative univariate and multivariate 1.2 Ethibel Sustainability Indices (ESI) conditional and stochastic volatility models is given in McAleer (2005). The Ethibel Sustainability Index (ESI) is composed of four regional indices, namely ESI Global, ESI Two constant conditional correlation models, Americas, ESI Europe and ESI Asia-Pacific. The namely the symmetric VARMA-GARCH model of indices are calculated and maintained by Standard & Ling and McAleer (2003), and the asymmetric Poor's (S&P), following S&P methodology. The ESI VARMA-GARCH (or VARMA-AGARCH) model regional indices are designed in such a way as to of Hoti, Chan and McAleer (2002), are estimated 2226 using daily data on two sustainability indexes and in which εη= h for all