Climate Risk and Equity Valuation: Should Investors Worry?

      Sizing Up the Climate Threat to Global Equity Values – A Novel Approach

EDHEC-Risk Climate Impact Institute's Scientific Director, Riccardo Rebonato, presents key findings from new research paper, "How Does Climate Risk Affect Global Equity Valuations? A Novel Approach," supported by Scientific Beta.

In this paper we adapt established valuation techniques in an innovative way in order to assess how the value of global equities can be affected by physical climate damage for different degrees of aggressiveness of the abatement policy. The topic is of obvious importance for investors. However, it is also of great relevance to prudential regulators, who want to understand how the climate-sensitive assets held by Systemically Important Financial Institutions may deteriorate in value and, by so doing, endanger the liquidity and solvency of the institutions, and threaten financial stability.

Our original contribution to assessing the impact of climate change and the value of global equities is based on combining three distinct features in a coherent valuation framework:

• a fully probabilistic approach, which puts around economic and climate uncertainty center stage;
• a focus on the importance of state-dependent discounting; and
• an emphasis on the combined analysis of physical and transition climate risk.

This paper builds on these features.

Firstly, we address the state dependence of discounting. Investors (and often regulators) have tended to use expected discounted cashflow models to arrive at the impact of climate change on equity valuation. While the approach is intuitive and theoretically sound, the current implementations make assumptions and simplifications that are far from innocuous. All expected discounted cashflow models arrive at their price estimates by combining two components: an estimate of the future cashflows; and a way to discount these future cashflows to today. For this second task, it is common practice to use sometimes time-dependent, but seldom state-dependent discount factors. We show in this paper that climate damages impair cashflows in a very state-dependent way (for instance, higher damages tend to be incurred in states of high economic activity), and that therefore state-dependent discounting plays an important, and much-neglected, role in arriving at valuation.

The second point of departure between our treatment and standard practice lies in how we handle transition and physical climate risk. Most of the analysis carried out by academics and practitioners has been focussed on transition risk (broadly speaking, the costs to business arising from complying with the regulatory measures to curb greenhouse-gas emissions) rather than physical risk (this being, for the most part, the direct damages arising from unabated climate change). This is consistent with the view that transition risk is `front loaded' and that the impact of physical risk, which is expected to materialise in the more distant future, is dampened by discounting. Whether climate damages are truly so distant depends, of course, on the appropriate discount factor and this naturally links with the state-dependent discounting perspective alluded to above. So, in our study we take into account both transition and physical risk, without judging a priori whether one or the other component should be more important for equity valuation (as it turns out, we will show that the impact of physical climate risk on equity prices is far from negligible). We note that in our approach transition costs and physical damages are estimated jointly. The exact split between the two components of equity valuation depends on the details of the preferences. However, we present in Section 3 a high-level discussion of the relative importance of the two factors. To summarise, when the abatement policy is robust, the effect on equity valuation is modest, physical and transition costs are similar, and they are both a small fraction of GDP; if we abate little, physical damages are, naturally, much larger, and dominate the transition costs. So, ‘when it really matters’, physical damages are more important.

The third way in which our approach differs from common practice is that our results are obtained within a proper probabilistic setting. The impact of climate change on asset prices has often been carried out using as reference the scenarios prepared by institutions such as the Intergovernmental Panel on Climate Change (IPCC), the International Energy Agency (IAE), or the Network for the Greening of the Financial Sector. Broadly speaking, these approaches create a single realisation of macrofinancial variables for each narrative, and do not associate any probabilities to the various narratives. In the words of the latest NGFS (2022) report ``the NGFS scenarios are not forecasts. They are intended to explore the range of plausible futures (neither the most probable nor the most desirable) for the assessment of financial risk''.[1]

Since no probabilities are attached to the scenarios, and since these do not span the possible `sample space' (what may happen),[2] carrying out the expectation part of the discounted cashflow models becomes impossible: when an asset is priced in a single-path approach, its cashflows become deterministic, and the uncertainty is accounted for by choosing an `appropriate' discount factor. This single number (for instance, the weighted average cost of capital) reproduces by construction observed market prices, but is not `transportable' to cases, such as climate damages, for which it has not been calibrated.

The absence of a probability dimension to scenarios is not limited to the IPPC/NGFS/IEA approaches: Bingker and Colesanti (2022) in their review of climate transition risk tools, remark that ``[a] major caveat is that (...) none of the tools provide output values by default as a probability distribution or as an estimate with associated confidence intervals.'' Our approach differs radically from these approaches because we draw on the best available economic and climate-damage literature to obtain a full probability distribution of future outcomes, from which an expectation can be obtained.

The probability distributions that are central to our approach could in principle be of two types: conditional and unconditional. The former are associated with a given abatement policy, the latter average over the probabilities of these policies. In this work, we present conditional estimates, but make no judgement on the likelihood of the different policies. In the conditional mode, our choice of abatement policy can be related to the Representative Carbon Pathways of the IPCC scenarios, but the key difference is that we take into account the full uncertainty in economic and climate outcomes associated with each abatement path. We are carrying out independent research to estimate these probabilities (eg, from fiscal, monetary and technological constraints), and, ultimately, the unconditional distribution of policy options.[3] The key message is that, by employing a model that (conditionally or unconditionally) `knows about' the dispersion of outcomes, we arrive at more negative losses than approaches that take average pathways as inputs: expectations of averages are not equal to averages of expectations.

Taking the probabilistic dimension into account is important. Our results indicate clearly that adjustments to equity prices can be most profitably analysed along two distinct perspectives: their magnitude and their uncertainty. The latter has been downplayed in many estimates of the impact of climate on investment portfolio (see, in this respect, the discussion in Rebonato (2023)). This is both unhelpful and dangerous, because it implicitly suggests a much greater degree of model precision than is actually achievable. Our results show that unavoidable uncertainty about economic growth cascades into uncertainty in physical damages, in equity dividends and in the appropriate discount rates. The additional uncertainty about the damage function and the physics of the problem further increases the width of the probability distributions we obtain. Tipping points add, but are not the only contributors, to the dispersion of cashflows that we obtain.[4] This matters for valuation, because, ceteris paribus, the greater the dispersion of cashflows, the greater the associated risk premium.

In addition to our focus on state-dependent discounting, on uncertainty and on a joint treatment of physical and transition costs, another distinguishing feature of our approach is that we work `from the top down'. By this we mean that we arrive at a price impact by modelling the effect of climate change on global GDP, and by treating global equities as (leveraged) claims on consumption.  This contrasts with a bottom-up approach, which would see us focus on one equity security or sector at a time and combine scenario information with issuer characteristics to produce granular predictions of climate change impacts and opportunities. There are, of course, pros and cons to both approaches, but, at the very least, the aggregate estimates that we arrive at (and that do not rest on difficult-to-verify assumptions on how individual securities or sectors will be affected by climate change) should provide a consistency check for the sum of the granular estimates.

These points of departure from common valuation practice explain why our estimates of the impact of climate change on equity valuation are different (and often higher) than most assessments: because we treat transition risk physical risk jointly and consistently;[5] because our approach allows for state-dependent discounting; and because it incorporates the probabilities of these states.

As for the magnitude of our results, we find that the difference in equity valuation with respect to a world without climate damages mainly depends:

• on the aggressiveness of the emission abatement policy (the slower the abatement, the greater the downward repricing);
• on the presence or otherwise of tipping points with relatively low threshold temperatures; and
• item on the extent to which Central Banks are able and willing to lower rates in states of economic distress (low consumption).

The difference in equity valuations between a no-climate-damage world and a world with climate damages can be significant, ranging from less than 10% if prompt and robust abatement action is taken, rising to more than 40% in a close-to-no-action case. In the presence of climate tipping points, this range widens from less than 10% for robust abatement to more than 50% in the case of very low emission abatement. We find that a severe impact on equity valuation can be obtained with very plausible combinations of policies and physical outcomes, and that there is considerably more downside than upside risk. We also find that, for all parameter choices, robust abatement policies strongly limit the impact of climate change on equity valuation.

The first two determinants of equity valuations (the aggressiveness or otherwise of the emission abatement policy, and the severity of climate damages) are not surprising. What is less widely appreciated is how strongly the state-dependence of discounting affects valuation by `state dependence' we mean dependence of the rate at which future cashflows are discounted on then-prevailing state of the economy. So, expected future cashflows obviously matter for valuation (and these directly depend on the severity of damages and on the aggressiveness of the chosen abatement policy); however, we show that, when it comes to valuation, how these impaired cashflows are discounted is also extremely important.

We have remarked that our estimates are different, and often higher, than what has been often reported in the literature. We stress, however, than, in arriving at our valuation, we have consistently made conservative choices: for instance, we have used a relatively tame damage function; we have chosen values for the equity leverage at the lower end of the academic consensus; and we have modelled investors preferences in such a way as to combine empirical realism with a relatively high discount rate (a lower discount rate would make the effects we describe more pronounced). Also, in our model interest rates always fall in periods of low economic activity; in reality, there may be circumstances (such as high inflation or high public debt) when these accommodative policies may not be forthcoming. This would also make our estimate of equity losses more severe.

Footnotes

[1] Emphasis added.

[2] This is particularly true for the NGFS scenarios, which almost exclusively rely on a particular socioeconomic narrative, the so-called `Middle of the Road' (SSP2) possible world.

[3] See in this respect a non-technical description of our approach in Rebonato (2024).

[4] Roughly speaking, a climate tipping point is a critical temperature threshold which, when exceeded, can lead to a significant, fast, and often irreversible shift in the climate system. Given the limited scope for adaptation, they can result in dramatic and widespread impacts.

[5] In our approach transition risk is reflected in the cost of following a given abatement policy. More aggressive reductions of emissions entail greater transition costs, but reduce physical damages. At the aggregate level we can therefore equate transition risks with transition costs. We do not consider the additional costs associated with a disorderly transition, but these would make, if anything, the impact on equity valuation even stronger.

References

Bingker, J. A., and C. Colesanti (2022): “Taming the Green Swan: a Criteria-based Analysis to Improve the Understanding of Climate-related Financial Risk Assessment Tools,” Climate Policy, 22 (3), 356–370.

NGFS (2022): “NGFS Scenarios for Central Banks and Supervisors,” Network for Greening the Financial System, pp. 1–48.

Rebonato, R. (2024): “What Climate Scenarios Currently Lack, and How to Fix Them,” The Conversation, 1–4.

Rebonato, R. (2023): “Portfolio Losses from Climate Damages: A Guide for Long-Term Investors,” EDHEC-Risk Climate Impact Institute Position Paper, pp. 1–39.