Climate risks can be broken down into two broad categories, namely physical risks associated with the direct impact of climate change on business models, and transition risks, which include the impact of policy changes, reputational impacts and shifts in market preferences, norms and technology. When it comes to transition risks, given that the transition to a low-carbon economy – whatever the scenario or trajectory envisioned – will inevitably involve an increase in carbon prices, carbon risk exposure has become a particularly critical concern for the investment industry.
Our research programme draws a fundamental distinction between treating carbon as a risk attribute instead of a risk factor.
The standard procedure to approach the problem of carbon risk measurement involves treating carbon as a relevant (non-financial) attribute of the firm. In other words, the idea is to measure carbon emissions at the company level, and then obtain a carbon footprint or carbon intensity estimate at the portfolio level as a weighted average of emissions normalized by the company market cap or revenue, respectively.
This approach raises, however, a number of concerns. First of all, while carbon emission estimates are relatively consistent across data providers when we limit the analysis to scopes 1 and 2, extending the analysis to scope 3, although desirable, leads to a high degree of heterogeneity. Second, a carbon footprint is a measure of the company’s contribution to climate change, which a priori is not the same as its exposure to climate change. Finally, this approach does not lead to a straightforward estimation of the change in portfolio value that would arise under different scenarios for carbon prices.
Our research programme will analyse these concerns at both the conceptual and empirical levels. It will also focus on analysing whether cross-sectional differences in carbon footprint estimates explain cross-sectional differences in risk and performance, after controlling for standard factor exposures. Turning to the time-series dimension, we will also explore whether meaningful carbon momentum strategies can be implemented as a follow-up to significant efforts to curb carbon emission as measured by substantial changes/improvements to measures of scopes 1, 2 or 3.
To alleviate these concerns, one can instead switch to a different perspective, where carbon price is explicitly treated as a risk factor. The idea here is to use a suitable statistical procedure to measure the sensitivity or exposure of a stock return or portfolio return to changes in carbon prices. The resulting quantity, which can be regarded as a carbon beta, can then be used to estimate the expected impact of a given change in carbon prices on the portfolio return.
To check for the consistency – or lack thereof – between the two approaches, we propose to analyse whether companies with a high (respectively low) carbon footprint also tend to be companies with high (respectively low) carbon beta, and we will explore the portfolio implications of these findings, for example by exploiting any information between discrepancies in terms of contribution and exposure to climate change. In an attempt to add robustness to carbon beta estimates, we also propose to introduce non-parametric conditional factor models that treat carbon beta as a function of the carbon footprint. Preliminary tests suggest that a non-linear relationship exists between carbon beta and carbon footprint, and our research will explore the portfolio implications of these findings.
Our research programme draws a fundamental distinction between managing carbon risk exposure without carbon derivatives and attempts to manage this exposure with carbon futures and option contracts.
The idea here is to use a combination of suitably designed security selection and portfolio construction methods to build an equity portfolio with the desired carbon risk exposure. More specifically, our objective is to propose a robust low-carbon portfolio construction framework and analyse the trade-off between reducing the portfolio carbon risk exposure (which can be defined either as carbon footprint or carbon beta) and increasing the tracking error of the portfolio with respect to the benchmark. When the relationship between carbon beta and carbon footprint is not monotonic, these dimensions can be regarded as two distinct degrees of freedom: for a carbon footprint at a given target level, one may seek to minimize carbon beta, and for a given target level of carbon beta one may seek to minimize the carbon footprint. We also propose to test two non-mutually exclusive carbon risk exposure reduction strategies, namely selection/exclusion (positive or negative screening, with or without country/sector/factor neutrality constraints) and optimization. On the latter dimension, we plan to test various portfolio optimization techniques including minimum variance, risk parity, max diversification, etc.
As a substitute for – or rather a complement to – the previous approach, it would also be possible to use carbon derivatives to reach a target level of carbon risk exposure. As far as futures markets are concerned, we first propose to conduct a thorough statistical analysis of basis risk, including analysis of the speed of convergence between spot prices and futures prices on the expiration date of the futures contract, derivation of the optimal hedge ratio, etc. We also plan to consider portfolio experiments where we measure the impact of a given change in carbon prices on (1) the value of a reference portfolio with no carbon risk management, (2) the value of the same portfolio re-engineered using selection and optimization methods so as to reduce/eliminate carbon risk exposure, and (3) the value of an immunized portfolio defined as the reference portfolio plus a suitable exposure to carbon futures contracts.