Disentangling measures of carbon risk

Decarbonizing Everything

• Alexander Cheema-Fox, Bridget Realmuto LaPerla, George Serafeim, David Turkington, & Hui (Stacie) Wang
• Financial Analysts Journal
• A version of this paper can be found here
• Want to read our summaries of academic finance papers? Check out our Academic Research Insight category

What are the research questions?

This was an exploratory study of various measures typically used in the investment industry to gauge climate risk and build investment strategies. Carbon intensity is a measure of climate risk and reflects how efficiently, in terms of carbon usage, a company generates $1 of sales. It is a scalable measure, standardized across all companies, and relies on company-reported data, which is a major advantage of this measure.  Lower carbon intensity is considered desirable and a low score indicates that a company has mitigated its carbon profile.  As a result, those firms tend to exhibit lower exposures to climate change risks (regulatory, reputational, and so on).

The study was conducted from the perspective of investors who are increasingly using a variety of data to manage climate risk by means of portfolio construction.  The authors examined 3 measures of climate risk including (1) the operational carbon (OC) measure –uses company reported carbon emissions from operations scaled by company revenue, obtained from S&P/ Trucost; (2) the total value chain (TVC) measure — uses the OC plus emissions from upstream and downstream activities, also from S&P/Trucost; and (3) analyst ratings (AR) decarbonization factor–obtained from MSCI.

The ratings may or may not incorporate the two carbon intensity measures, but certainly include judgments as to a variety of factors related to climate issues that are characteristic of a specific company.  The factors are then combined in a subjective weighting scheme to arrive at a final rating that reflects the analysts’ evaluation of how the company manages the risks and opportunities with respect to climate change.

The universe consisted of 2,087 firms, across 43 industries, with data on all 3 measures, and observed between 2013 and 2020. Primary among the questions addressed was the issue of consistency among the 3 measures and how they performed on a risk-adjusted basis when included in the portfolio construction process. Long (short) strategies were referred to as decarbonization factors (DF), although explicitly not considered necessarily as priced capital market risk factors. They are simply an indication of each company’s ability to perform and compete in a low carbon environment (better/worse) as indicated by each measure (low/high score).

What are the Academic Insights?

Here is a summary of the main findings:

1. There is quite a bit of variability between the 3 scores within and across industries.  Industries scoring high in the OC factor included electric utilities, independent power & renewable electricity producers and multiutilities.  High scores on the TVC factor included machinery, food products, and insurance. Industries with high scores on the AR factor included biotechnology, pharmaceutical, and media.  Low scores on AR included oil, gas & consumable fuels, metals & mining, and food products.


2. Use of the 3 factors can produce various types of biases in the resulting portfolio. For example, small cap and more profitable firms are overweighted when value-weighted OC and TVC factors are used, while equally-weighted AR factors tilt towards larger cap and more “conservative” companies.  The spread of carbon intensity between the long and short legs of portfolios was preferable (greater carbon efficiency) with the OC and TVC measures when compared to AR. However, all 3 measures did indeed reduce carbon exposure.  OC and TVC exhibited a 78% overlap in holdings, with 50% overlap with the AR factor.


3. When climate change risks were relatively more financially material, factor returns were more positive.  The reasoning: high-carbon intensive businesses will face growing costs, slowing demand, carbon taxes, regulations on emissions, and other disruptions.  Although the effect was not present when the AR measure was used.  It seems that when carbon intensity is high, there is more herding among analysts and their opinions exhibit much less variability in financially material industries.  Analyst opinions were less effective at differentiating across companies when carbon risk was very high.


4. The measures should be used in combination. For example, the authors suggest using the OC factor for industries with high dispersion in TVC and low correlation between OC and TVC. For industries with low dispersion in the TVC factor, use the AR measure. In backtests, such a strategy produced excess returns of 177 to 332 bps when compared to returns from using any of the 3 measures in isolation. The cumulative abnormal returns to a combined, rules-based metric and the 3 individual measures are shown in Figure 3 below.

Why does it matter?

Investors certainly care about the potential risks tied to dramatic climate shifts. In terms of contribution, this article provides a much-needed perspective for investors on the merits of various measures of climate risk.  The bad news takeaway is significant: the measures do NOT overlap and each one leads to an identifiable bias within portfolios constructed by those individual measures. On the other hand, analysts seem to add value above and beyond the 2 carbon intensity measures with the exception of companies with high carbon intensity dispersion. Kudos to the analysts on this one. Good to know: fundamental analysis has merits when evaluating E(SG)-style portfolios.

A caveat or two: The average properties of the trading strategies were tested only over the full sample of data. There were no out-of-sample results presented. The reader should refrain from inferring any type of guarantee that the out-of-sample performance will reflect the numbers reported here. Excess returns were calculated using the Fama-French 5-factor (market, size, value, investment, quality, and momentum) risk model.

The most important chart from the paper

Abstract

We analyze how the use of different climate risk measures leads to different portfolio carbon outcomes and risk-adjusted returns. Our findings are synthesized in a rules-based investment framework, which selects a different
type of climate metric across industries and weighs industries in the portfolio based on the variability of carbon outcomes among firms within each industry. We conclude that analyzing the merits and applicability of various climate data can help investors manage climate risk while increasing risk-adjusted returns.