Edito by Riccardo Rebonato, Professor of Finance, EDHEC-Risk Institute
Over the past decade, scores of studies have tried to provide an answer to this question by estimating empirically the magnitude of the climate risk premium. Unfortunately, these empirical studies are plagued by a problem that, for the moment, remains insurmountable: we just do not have enough data to say anything empirically meaningful. As I tell my students, measuring a risk premium is like detecting the direction of a gentle steady breeze in the middle of a hurricane: it can be done, but it requires many decades worth of data, collected over several business cycles, in a variety of different market conditions. Alas, climate change has only caught the attention of investors (and can therefore be expected to be impounded in prices) in the last 15 years or so. As luck would have it, in this period market conditions (with the Great Recession, the exceptional monetary conditions that ensued, and, to cap it all, the COVID years) have been anything but typical. So, as if to mirror the climate juncture, the financial ‘hurricanes’ in the midst of which we are supposed to discern the direction of the climate risk-premium breeze have been unhelpfully violent.
Add to this the fact that, as climate awareness has taken hold in the investing community, a secular tilting of investments away from fossil-intensive and into green sectors and industries has recently gathered pace. Were the exceptionally low and high returns earned by brown and green assets, respectively, due to the climate risk premium, or a mechanical effect of the investment stampede? It is simply too early to tell.
Fortunately, this does not mean that we can’t say anything about something as fundamental as the effect of climate change on the investment environment. It simply means that, for once, taking a fundamental, first-principles perspective can be more useful than getting our hands dirty with data. This theory-based approach will not tell us whether to buy stock X or short security Y, but it may provide us with a compass to navigate the choppy investment waters in the years of climate change. So, while we are still at a stage when hard, crisp predictions about investment outcomes are premature, we can nonetheless try to understand what we have to watch out for on the way to making these hard and crisp predictions.
So what do asset pricing theory and climate science say about asset returns? The first observation is that holders of securities have pieces of (electronic) paper that entitle them to the fraction of what the economy produces that goes to the providers of capital. If we assume this fraction to be constant, then the first determinant of future returns is how much the economy will produce in the future. If economic output is affected by climate change in ways that are not embedded in today’s prices, then the impact on realized returns will be substantial. This is what I call the ‘size-of-the-pie’ effect, and it has all to do with (P-measure) expectations. The key point here is that climate risk is not at all like idiosyncratic risk, in the sense that it cannot be diversified away in the limit of an infinitely large portfolio. This being the case, the first-order question is: how bad do we expect the climate damage to be?
Now, there are two major conceptual frameworks to model the economic damage created by climate risk. The first points directly to bad climate outcomes as a main source of risk for the economy: in this view of the world, if the impact of climate change is going to be large, it will be because sea levels will rise so much, or hurricanes will be so violent, or working environments will become so unbearably hot as to cause direct damage to capital or to productive capabilities.
The second framework takes a very different perspective and looks at severe climate damage as a consequence of exuberant economic activity (because in a not-yet decarbonized world, the more we produce the more we emit). In doing so it reverses the cause-effect link. In this view, we suffer extensive damages exactly because we are producing (and hence emitting) ‘too much’. Therefore, if we look at climate damage from this perspective, uncertainty about climate damage is ultimately due to our uncertainty about how strong the economy will be in the decades to come.
Simplifying a bit, the first view of where these damages come from borrows from the literature of economic ‘disasters’ made famous by Barro (2006, 2009) and used to explain the equity risk premium. They key ingredient in this account is that these disasters affect not just consumption, but also consumption growth (à la Bansal and Yaron, 2004 and Bansal, Kiku and Ochoa, 2016). The second strand is the route taken by the original Integrated Assessment Models, of which DICE is the best known.
Now, irrespective of the directionality of the causal link between economic activity and climate damage, negative climate outcomes will negatively (positively) affect those sectors of the economy that are adversely (favourably) exposed to climate risk. So, as a truism, assets with a positive (negative) ‘beta’ to climate shocks will decrease (increase) in value in response to one, whatever its origin. However, this is where the similarities between the two channels of causation end, and it is risk premia that drive the wedge between the two views of climate damage. Risk premia are not a small correction. If we look at equities, we know very well that we cannot make any sense of their valuations unless we take risk premia into account (the equity risk premium, after all, is huge, of the order of 6%). So understanding risk premia can be absolutely essential to understanding valuations. But, when it comes to the climate risk premium, this poses a bit of a problem, because the catastrophist and the growth-mediated views of climate damage make diametrically opposite predictions. The reasoning goes as follows.
The first thing to keep in mind is that risk premia, when positive, are the compensation we require from securities that pay well in good states of the world (when we ‘do not need the money’); when risk premia are negative, they are the insurance premia that we are ready to pay to enjoy good payoffs from ‘hedging securities’ in states of low consumption. So, if we believe in the climate disaster models, climate damages will tend to materialize when consumption is low – after all, in these models, consumption is low precisely because of the climate disasters. Assets with a positive climate beta – ie, assets that pay poorly when the climate shock occurs – will therefore require a positive risk premium. And since equity owners of wind turbines and storage batteries will do well in these low-consumption states of the world, the wind turbines and storage batteries will act as insurance, which means that they will command a negative risk premium.
Switch now to the world where climate damages are caused by exuberant economic expansion (via the attending emissions). These are, almost by definition, states of high consumption, and the same wind turbines and storage batteries are now offering good pay-outs when we ‘don’t really need the money’. And, if the economy is in a deep recession, the climate damages will be small, and the positive-beta green assets will fare badly. As a result, it is now negative-beta assets that require a positive risk premium (a low price) to be held in equilibrium.
Which view of the world is (more) right? In which direction does the arrow of causality point? It is still too early to be able to tell with certainty, but this is an area of vibrant research, both at the EDHEC-Risk Institute and in the wider academic world. When my research colleagues and I couple the latest physical models of the climate system (as in Joos et al., 2013 and Dietz et al., 2021) with a state-of-the-art version of a new-generation Integrated Assessment Model, we find that, in the absence of uncertainty in the growth of the economy, climate damage only seriously affects economic output if threshold effects come to the fore – where ‘threshold effects’ is a blanket term covering both tipping points and near-irreversible losses beyond some critical level. In other words, if climate damages are just an unwelcome by-product of economic output (high when we are rich and low when the economy goes through a rough patch), the size-of-the-pie effect is small. From a reverse-stress-testing perspective, we ‘need’ threshold effects for climate change to have a strong effect on output – threshold effects that may be triggered by, but become independent of, economic growth.
So the next question is: how bad can these non-linear effects be? We do not know for sure, but we know enough to worry. We know, for instance, that in the relatively recent past ‘’roughly half the north Atlantic warming since the last ice age was achieved in only a decade” and that, according to Alley & Clark (1999), ‘’large, abrupt climate changes have repeatedly affected much or all of the earth, locally reaching as much as 10° C change in 10 years”. In a similar vein, Lang et al. (1999) report that “[c]hanges of up to 16° C and a factor of 2 in precipitation have occurred in some places in periods as short as decades to years”. And Andersen et al. (2004) carry out a high-resolution study of the climate for the northern hemisphere during the last glacial period, and measure via O18 isotopic proxies changes of more than 10° C within 40 years.
How likely are these major climate changes? Again, we cannot know for sure, but while early studies by the IPCC suggested that ‘large-scale discontinuities’ would only be likely for temperature increases greater than 5° C, the latest estimates from the same source suggest that tipping points could be breached even for temperature increases between 1° and 2° C.
So our calculations confirm that, yes, states of high economic growth are associated with high emissions and hence high levels of damages (in ‘good states of the world’). However, if significant tipping points are ‘lurking out there’, then the catastrophist model begins to play an important role as well in generating substantial economic losses in low-consumption states of the world, thereby reversing the sign of the risk premium.
What about the term structure of risk premia? Will long-dated or short-maturity securities be affected more strongly by climate surprises? Can we glean some information about the sign of the risk premium by making the problem a bit more complex? To some extent, we can. The key factor now is how the economy responds to a climate surprise. If the serial autocorrelation of economic growth is positive, then there is a long-term risk to consumption, because today’s shock, of whatever sign, will affect consumption growth in the same direction. This, after all, was the key intuition beyond Bansal and Yaron’s seminal work to explain the equity risk premium puzzle and the interest-rate puzzle. If this is the case, then long-maturity assets – whose value today depends significantly on distant future cashflows – will be affected more severely by climate shocks than short-dated assets. And, if the main uncertainty for the economy comes from the climate damages (the ‘tipping points out there’), then the negative risk premia for wind turbine-like assets will become more negative the longer the life of the asset. Conversely, if the main uncertainty is about the strength of the economy, wind turbine-like assets will command a positive risk premium that increases with the maturity of the asset.
Everything, of course, works in reverse if we have instead a negative coefficient of serial autocorrelation – if there is, that is, mean reversion (faster growth) after a negative economic shock. Now, the existence of mean reversion can be interpreted as the effectiveness or otherwise of human adaptation to climate change – a topic about which there has been much theoretical (and ideological) debate, but little hard evidence. Fortunately, there have recently been significant advances in the estimation of the term structure of risk premia for asset classes that are exposed to climate risk, such as equities (van Binsbergen et al., 2012; Van Binsbergen and Koijen, 2017) and real estate (Giglio et al., 2015; Giglio, Kelly and Stroebel, 2021). In both cases, researchers have found positive risk premia that decline with the horizon (see Giglio et al., 2021). So the climate catastrophic view of the world coupled with persistent (non-mean-reverting) losses seems at least to be embedded in current market prices.
This is not the end of the story. Yes, if van Binsbergen, Giglio and their colleagues are correct, we can conclude that the prices of climate-sensitive assets seem to be arrived at by assuming large, directly climate-induced and positively correlated economic losses, and that our ability to adapt will be limited. But investors’ understanding of the climate system, and of the nature of the damages climate change will inflict on the economy, can be no better than that of the scientists who devote their lives to studying these matters – and, currently, this understanding is very limited, especially when it comes to the magnitude of the damages. It is very possible that in the relatively short term this price-embedded information will be revised (climate models, for instance, are undergoing changes so quickly that the reduced-form models which were state-of-the-art in 2016 are by now unusable). Which means that our expectations about climate damages, about their causal link to economic growth, and about our ability to adapt are likely to change substantially over a relatively short span of time.
The bottom line is that at least one prediction can be made with confidence. Since the variability in the price of a security comes both from the revision of cashflow expectations (the size-of-the-pie bit) and the repricing of risk (this is the risk premium component), what we can expect for sure is significant volatility for the sectors more exposed (positively and negatively) to climate change.
There are, as usual, many qualifications and caveats. First of all, I have only discussed the effect of climate change on asset prices. When Integrated Assessment Models suggest that, in the absence of threshold effects, the impact of climate change on economic output would be limited, the conclusions only refer to the size-of-the-economic-pie effect. There is, of course, far more to climate change than a loss of GDP, as global warming can have devastating effects on very poor populations (eg, the 150 million people who live in the Sahel). Unfortunately, the global output these populations currently produce is so small that it hardly makes a difference to security valuations. However, indirect effects on wealthier parts of the world could materialize via, say, forced migrations. The fact that these spill-over effects are just too difficult to model does not mean that they may not turn out to be very important.
And, lastly, I have focussed in this note on physical risk and said almost nothing about transition risk. One way to model transition risk with Integrated Assessment Models is by using non-optimal abatement paths (say, too sluggish an abatement pace first, followed by a catch-up period). This is, indeed, one of the areas of active research for the ERI Climate programme.
 These considerations draw on the excellent review article by Giglio et al. (2020).
 In our version of the DICE model, we allow for simultaneous stochasticity for up to four variables (including the all-important Total Factor of Production and damage exponent in a recursive-utility (Epstein-Zin) framework to disentangle risk aversion from elasticity of intertemporal substitution, with a process for consumption consistent with the pricing of the equity and debt markets (as in Bansal and Yaron, 2004)). We also allow for Bayesian learning of the physical parameters, as in Rudik (2020).
Alley, R. B. and P. U. Clark. 1999. The Deglaciation of the Northern Hemisphere: A Global Perspective. Annual Reviews of Earth and Planetary Sciences, 27, 149–182
Andersen, K. K. 2004. High-Resolution Record of Northern Hemisphere Climate Extending into the Last Interglacial Period. Nature, 147–151
Bansal, R., D. Kiku and J. Ochoa. 2016. Price of Long-Run Temperature Shifts in Capital Markets. NBER working paper series, working paper 22529, 1–54
Bansal, R. and A. Yaron. 2004. Risks for the Long Run: A Potential Resolution of Asset Pricing Puzzles. Journal of Finance, 59(4), 1481–1509
Barro, R. J. 2006. Rare Disasters and Asset Markets in the Twentieth Century. Quarter Journal of Economics 121(3), 823–866
Barro, R. J. 2009. Rare Disasters, Asset Markets and Welfare Costs. NBER working paper series, working paper 13690, 1–44
Dietz, S., F. van der Plog, A. Rezai and F. Venmans. 2021. Are Economists Getting Climate Dynamics Right and Does It Matter? Journal of the Association of Environmental and Resource Economists, 8(5), 895–921
Giglio, S., M. Maggiori, J. Stroebel and A. Weber. 2015. Climate Change and the Long-Run Discount Rate: Evidence from Real Estate. CEPR Discussion paper, working paper 10958, 1–83
Giglio, S., B. Kelly and J. Stroebel. 2021. Climate Finance. Annual Review of Financial Economics, 13, 15–36
Joos, F., J. S. Fuglestvedt et al. 2014. Carbon Dioxide and Climate Impulse Response Functions for the Computation of Greenhouse Gas Metrics: A Multi-Model Analysis. Review of Financial Studies, 27(10), 3060–3097
Lang, C., M. Leuenberger, J. Schwander and S. Johnsen. 1999. 16 degrees C Rapid Temperature Variation in Central Greenland 70,000 years ago. Science, 286, 934–937
Van Binsbergen, J. H., W. Hueskes, R. Koijen and E. Vrugt. 2013. Equity Yields. Journal of Financial Economics, 110, 503–519
Van Binsbergen, J. H. and R. S. Koijen. 2017. The Term Structure of Returns: Facts and Theory. Journal of Financial Economics, 124, 1–21