On Bond Returns in a Time of Climate Change.

• Author: Ravina, Alessandro

1. INTRODUCTION

   Climate change is attributed to two different causes: natural climate variability--natural internal processes or external forcings--and human activity that alters the composition of the atmosphere (Intergovernmental Panel on Climate Change, 2014; United Nations, 1992). The consensus of actively publishing climate scientists on anthropogenic global warming is in the 90%-100% range (Cook et al., 2016). The breaking point of human contribution to climate change is usually identified with the industrial revolution since economic development is strictly correlated to energy consumption (Energy Information Administration, 2017; Stern, 2007): the burning of fossil fuels has increased the concentration of atmospheric carbon dioxide (C[O.sub.2]), the most prominent forcing factor, from 280 parts per million (ppm) in preindustrial times to approximately 400 ppm (Wagner and Weitzman, 2016).

   The literature has established a dichotomy of climate change risks. The first category has been labeled "climate risk" (Carney, 2015): changes in extreme climate phenomena--e.g. temperature extremes, high sea levels extremes, precipitation extremes (Intergovernmental Panel on Climate Change, 2014)--are likely to cause serious damages to agriculture, coastal zones, human health, and affect growth (Dell et al., 2014; Pycroft et al., 2016), productivity (Graff Zivin and Nei-dell, 2014; Hallegatte et al., 2015), the value of financial assets and insurance claims. The second category, labeled "transition risk" or "carbon risk", makes reference to the cost of the adjustment towards a low-carbon economy (Caldecott and McDaniels, 2014). Low-carbon transition risk is a multi-faceted concept that includes all drivers of risk linked to the decarbonisation of the economy: a) pollution reducing market-based instruments (a carbon price: a carbon tax, an auction price, or a secondary market price), b) command and control induced technological shifts aimed at a reduction of C[O.sub.2] emissions, e.g. stranded assets or assets that have suffered from unanticipated or premature write-downs, devaluations, or conversion to liabilities (Caldecott et al., 2016), and c) market risk, i.e. market demands for low carbon products (Zhou et al., 2016).

   The impact of a particular market-based instrument, the European Union Emission Trading System (EU-ETS), upon financial values has already been addressed by the literature; nevertheless, efforts pertain primarily to stocks, leaving the bonds field out of the picture. The objective of this paper is to assess the impact of the 2003/87/CE directive, upon which the EU-ETS is based, on European bond returns.

   Literature on the interconnection between carbon pricing and bond values is scant. Man-sanet-Bataller and Pardo (2008) look at the effect of including European Union Allowances (EUAs) in diversified portfolios made up of stocks, bonds, and commodities (Brent and Natural Gas) finding that including phase I and phase II EUAs actually improves the investment opportunity set for market practitioners that have initially invested in traditional assets like stocks and bonds. Koch (2014) studies price linkages between EUAs and market fundamentals and how they vary over time. The correlation between EUAs and a set of assets like oil, gas, coal, electricity, but also stocks and bonds, is analysed in order to explain the variations of price linkages; results show that carbon and financial markets are not segmented: high expected market volatility shifts carbon-stock correlation significantly upwards and carbon-bond correlation significantly downwards. Chevallier (2009) examines the relationship between carbon future returns and changes in macroeconomic conditions, finding that macroeconomic variables such as equity dividend yields, the junk bond premium, the U.S. Treasury bill yields, and the excess return on a globally diversified portfolio of commodities are only loosely related to carbon futures returns. While this scarce body of work aims at finding the determinants of a carbon price, the study of the inverse causal relationship has, to my best knowledge, never been undertaken.

   In order to detect the impact of low-carbon policy--the 2003/87/CE directive which initiated EU-ETS--upon the bond returns of European firms, a Fama and French (1993) framework, for the first time, is employed. Along with the two bond market factors proposed by Fama and French (1993), TERM and DEF, an EU-ETS participation factor is added: GMC. Supplementing classical factors with an environmental factor has already been done in research carried out on the stock market (Gorgen et al., 2017; Oestreich and Tsiakas, 2015; Ravina and Kaffel, 2019). However, some differences in the construction of the environmental factor remain. In this sense, the factor construction closer to the one presented here is found in Ravina and Kaffel (2019). The rationale behind the GMC factor is the following: if we want to measure the impact of the 2003/87/CE directive with a factor, one possibility is to take all firms regulated by the policy, perform carbon accounting for each firm, construct two portfolios, i.e. a high-carbon portfolio and a low-carbon portfolio, and then take the differences of the value-weight returns. Unfortunately, this operation wouldn't permit us to uncover the real green (or carbon) premium because the firms that take part in the EU-ETS are all high-carbon firms. This means that, when we build the two portfolios, the low-carbon portfolio would contain a set of firms which are only slightly less polluting than firms in the other portfolio. The resulting environmental factor would be biased, i.e. negligible in terms of magnitude. In order to cope with the fact that the EU-ETS covers only high-carbon sectors, an alternative is to construct the environmental factor by means of two portfolios, a portfolio composed of EU-ETS liable firms (which I call "carbon" portfolio) and a portfolio composed of EU-ETS exempt firms (which I call "green" portfolio). In this context, while TERM proxies for the common risk in bond returns related to unexpected changes in interest rates and DEF mimics the risk factor in returns related to shifts in economic conditions that change the likelihood of default, GMC (Green minus Carbon) is meant to mimic the risk factor in returns related to low-carbon policy, the 2003/87/CE directive in this case. The new component, the GMC factor, is obtained by subtracting the weekly value-weight carbon bond portfolio returns from the weekly value-weight green bond portfolio returns from the beginning of Phase II (2008) of EU-ETS. The carbon bond portfolio is composed of 25 firms regulated by the 2003/87/CE directive and the green bond portfolio is composed of 25 firms exempted by the 2003/87/CE directive upon which the EU-ETS is based.

   This paper makes the following contributions. Firstly, it is the first time that a factor model is employed to assess the sensitivity of bond returns to low-carbon policy. The sensitivity of bond portfolio returns to the GMC factor has been found to be positive in the case of green portfolios and negative in the case of carbon portfolios. Most importantly, slopes on GMC are highly statistically significant. Secondly, the average value of GMC itself is positive: finding a positive GMC means that in Europe, in the 2008-2018 time-span, there is no carbon premium as some of the literature asserts, but rather a green premium. Such a green premium confirms that the EU-ETS has a positive effect in the financing of the low-carbon transition: the beginning of phase II of EU-ETS--the start date of the study--coincides with both capital outflows from EU-ETS liable firms and capital inflows to EU-ETS exempt firms. Thirdly, evidence is found that the addition of an environmental factor improves the performance of the Fama and French two factor model for bonds, at least in Europe from 2008 onwards. Fourthly, since the literature has recently proposed stress testing, a technique developed for testing the stability of an entity, as an evaluation framework for climate change risks (Bank of England Prudential Regulation Authority, 2015; Fay et al., 2015; Schoenmaker and van Tilburg, 2016; Zenghelis and Stern, 2016), I follow the recent carbon stress test trend and put forward a stress test that is able to indicate the impact of a hypothetical EU-ETS average price upon bond returns. The results show the effects of a plausible, but more severe, average EU-ETS price on both carbon firms and green firms.

   The paper is structured in the following way: Section 2 presents the EU-ETS; Section 3 introduces the model; Section 4 puts forward the data; Section 5 provides the empirical results; Section 6 presents the diagnostics; Section 7 exhibits the carbon stress test; Section 8 concludes.

2. THE 2003/87/CE DIRECTIVE

   The 2003/87/CE directive is at the origin of the European Union Emission Trading System (EU-ETS). The EU-ETS is a market based instrument, launched as a pilot project in 2005, whose objective is to reduce greenhouse gases (GHG) emissions in all European Union (EU) countries as well as Iceland, Lichtenstein and Norway. The three-year (2005-2007) pilot project, phase I, has been followed by a four-year (2008-2012) phase II and a seven-year (2013-2020) phase III. In 2020, at the end of phase III, emissions covered by the EU-ETS, around 45% of the EU's GHG, are expected to be 21% lower than at the start of the pilot project (2005). From the beginning of phase III, the EU-ETS covers more than 11,000 installations consisting of power and heat generation, oil refineries, commercial aviation, and production of steel, iron, aluminium, metals, cement, lime, glass, ceramics, pulp, paper, cardboard, acids and bulk organic chemicals (European Commission, 2015).

   The EU-ETS is a cap and trade system: the European Commission has put a cap on EU-wide GHG emissions which has been progressively reduced. When a...