“Climate change poses significant challenges for the global economy and the financial system. The public rightly expects us to work to ensure the financial system is resilient to climate-related financial risks.” Federal Reserve Board Chair Jerome H. Powell, October 21, 2021.
Supervisors around the world wish to ensure that the financial system is resilient to climate change (see, for example, Chair Powell’s comment above). To that end, current best practice is to formulate detailed long-run climate scenarios and then ask whether financial institutions, especially banks, can withstand the losses associated with them. These scenarios typically map the path of surface temperature, sea level, and the resulting economic damage over the next 30 or 40 or 50 years.
However, financial-system stress arises from sudden, widespread changes in the value and perceived quality of leveraged intermediaries’ assets, while climate change is likely to remain gradual over decades. As a result, skeptics reasonably doubt that climate change poses systemic financial risk sufficient to merit the use of scarce supervisory resources and a costly testing apparatus. To quote John Cochrane: “[B]anks did not fail in 2008 because they bet on radios not TV in the 1920s. Banks failed over mortgage investments they made in 2006.”
Fortunately, we now have low-cost, high frequency, forward-looking tools for monitoring climate-related sources of financial instability. In this post, we use one such tool to identify episodes in which the potential influence of climate change on systemic resilience may be worthy of attention. We also look at how an aggregate measure of financial system vulnerability evolves over time.
Before getting to the specifics, we should emphasize that the current supervisory practice does have uses. For their own well-being, financial intermediaries need to “identify, assess, control and mitigate the inevitable risks arising from climate and environmental crises” (see here). Exercises like the Bank of England’s exploratory stress test and the development of climate scenarios by the Network for the Greening of the Financial System (NGFS) are designed to encourage intermediaries to gauge their climate risk exposures—a necessary first step if firms are to manage these risks. Furthermore, the disclosure rules that authorities are trying to impose on (financial and nonfinancial) firms are designed to help both them and their investors measure and price climate risk.
We now turn to the high-frequency climate stress testing tool developed by Jung, Engle and Berner (JEB). They ask the following simple question: How sensitive is the market value of a financial intermediary’s equity to changes in the value of a fossil fuel portfolio? Combining the answer to this question with the level of the firm’s liabilities yields a measure of vulnerability to climate-related events that JEB can compute daily using publicly available information. They label their measure CRISK.
To understand the appeal of CRISK, we start with the concepts of physical and transition risk associated with climate change. The Basel Committee on Banking Supervision provides helpful definitions:
Physical risks are the economic costs and financial losses resulting from the increasing severity and frequency of extreme climate change-related weather events, longer-term gradual shifts of the climate, and indirect effects of climate change such as loss of ecosystem services.
Transition risks arise from the process of adjustment towards a low-carbon economy.
Importantly, these are risks. The realization—the actual path that we follow over the next 50 years, and the resulting losses—is just one of many possibilities, each with an associated probability. So, we can usefully characterize physical and transition risks as probability distributions over a set of possible futures (that are surely related to one another). Since asset prices depend on the full distribution of possible outcomes, changes in both the location and the shape of these distributions can alter asset prices. Our interest in managing and adapting to these risks focuses on the tails of those distributions associated with significant losses that threaten financial stability.
This description is important for assessing systemic financial concerns because it focuses our attention on the probability distributions regarding future paths, rather than on the smooth and largely predictable nature of climate change itself. Going further, as people learn about the evolution of climate risks, it is useful to focus on the changing perceptions of those probability distributions, rather than just the true distributions of possible outcomes.
Viewed from this perspective, any change in investors’ perceived distribution of either the severity and frequency of extreme climate change-related events or the process of adjustment to a low-carbon world likely will have an impact on asset prices. Moreover, because extreme climate events initially appear as relatively unlikely “tail events,” experiencing one can sharply and permanently alter their perceived likelihood. Through this mechanism, severely adverse climate-related tail events—both physical and transitional—will affect a wide range of asset prices and (potentially) trigger financial stress, even if the underlying process of climate change remains gradual and largely predictable.
This is where CRISK comes in. CRISK is the younger (and still economically smaller) sibling of the NYU V-Lab measure of systemic risk, SRISK. SRISK is an estimate of the capital shortfall conditional on a specified decline of the global equity market over a period of six months. Summing each intermediary’s SRISK leads to an aggregate measure that observers can use to monitor the vulnerability of the financial system as a whole (see Brownlees and Engle).
In what follows, we focus on “Marginal SRISK”, which is the capital shortfall associated solely with the hypothetical global equity market decline. Even in the absence of such an equity market plunge, many institutions have “baseline” capital shortfalls because the market value of their equity already falls short of a leverage ratio benchmark (say, 8% for the U.S. and Asian firms, and 5½% for European institutions). Marginal SRISK excludes this baseline component of SRISK.
Computing Marginal SRISK requires an estimate of the correlation of the value of a financial firm’s equity return with the value of the global equity market conditional on the specified market plunge, say 40%. This is the conditional market beta. Combining this with the level of liabilities from public disclosures yields a measure of each firm’s capital shortfall relative to a regulatory benchmark. Available daily, Marginal SRISK is a measure that provides information similar to—but at far lower cost and far greater frequency than—slow, lagging and resource-intensive supervisory stress tests like the Federal Reserve’s annual Comprehensive Capital Assessment Reviews (CCAR).
Analogous to Marginal SRISK, Marginal CRISK is the answer to the following question: How much does a financial firm’s capital adequacy (relative to a leverage ratio benchmark) change following a 50% decline in the value of a “stranded asset” portfolio? Stranded assets are the existing fossil fuel reserves that are currently valued but would remain unused following a transition to net-zero carbon emissions. Again, as in the case of SRISK, computing CRISK requires estimating the correlation of a firm’s equity returns with the value of the stranded asset portfolio conditional on a plunge in its value—a conditional climate beta. Like Marginal SRISK, it is available daily in real time using publicly available information.
In our view, Marginal CRISK provides valuable information on the resilience of the financial system. Since asset prices are forward looking, changes in perception about the distribution of possible climate outcomes will move prices today. And, those price movements can be large and sudden.
We see two possible sources of such changes in perceptions regarding the probability distribution of climate-related outcomes. The first is related to physical risk. When people see a devastating fire or destructive hurricane, they may revise their views about the frequency and severity of adverse climate-related events. The second concerns policy actions. For example, the election of candidates advocating immediate climate action may foster expectations of significant tax and regulatory changes, causing a sudden fall in the value of fossil fuels (and losses to the stranded asset portfolio).
Before continuing, we note that the methods JEB develop can be applied to any portfolio that serves as a proxy for the impact of climate risk, not just the stranded-asset measure that they choose.
So, what can we learn from current measures of Marginal CRISK? The following chart shows the 11 U.S. financial firms with the largest estimated total capital shortfalls at the end of October 2022. For each firm, we include four important pieces of information. First, the baseline shortfall (in blue) is the level of capital shortfall relative to an 8 percent leverage ratio benchmark. A negative number means that the firm’s capital exceeds the benchmark.
Marginal SRISK plus Marginal CRISK for U.S. Intermediaries ranked by estimated total capital shortfall, October 2022
Source: NYU V-Lab.
Second, Marginal SRISK (in orange) is an estimate of the losses that the firm would face if the global equity market declined by 40% over six months. Third, we plot Marginal CRISK in red. This is the loss associated with a 50% decline in the value of a stranded asset portfolio. Here, six are negative and nine are positive. That said, even the largest (Morgan Stanley) is quite small—just over $3 billion. Fourth, we plot the total (including baseline, marginal SRISK and marginal CRISK) as a black diamond. (For completeness, we also include a gray bar, which reflects a nonlinear adjustment arising from both the correlation between the market beta and climate beta and the fact that returns are logarithmic.)
The simple, reassuring message we take from these calculations is that news regarding the size and severity of climate losses currently has little impact on the health of the financial system. Put differently, a general deterioration of economic and financial conditions creates much more stress on these institutions—eroding capital buffers by far more—than risks associated with the repricing of climate-related stranded assets.
However, the vulnerability of the financial system to climate-related risks can change, so we ought not be complacent. Fortunately, Marginal CRISK allows us—virtually without cost—to monitor such changes over time, as well as to assess the resilience of the financial system to specific events. As an example, we constructed the change in Marginal CRISK at the time of the California Camp Fire, November 2018. The most destructive wildfire in California history, the Camp Fire burned 153,336 acres (62,053 hectares) and more than 18,000 structure over the course of just over two weeks. Given its severity, the Camp Fire catastrophe might well have altered perceptions of the probability distribution of such costly tail events.
The following chart shows the change in Marginal CRISK from end-October to end-November 2018 for the 11 global firms with the largest increase. We plot the U.S. firms in black and non-U.S. firms in gray. The numbers above each bar represent the size of each firm’s capital buffer (equivalent to the negative of the baseline shortfall in the prior chart). In all but three cases, the increase in marginal CRISK is smaller than the capital buffer. That is, for the most part, the financial institutions that were most exposed to climate risk at the time of the Camp Fire had sufficient capital to absorb the losses from the repricing of assets following this severely adverse climate-related event. Again, we find this reassuring.
Largest Increases in Marginal CRISK during the California Camp Fire, November 2018
Source: NYU V-Lab.
Finally, by aggregating Marginal CRISK we can get a picture of emerging climate risks in a country’s financial system. The following chart provides a picture for the United States. Here, we have a time-series for the baseline shortfall and for something that we label CRISK Exposure. The first, in gray, sums the capital shortfalls (relative to an 8 percent leverage ratio benchmark) for those firms with a shortfall. The second, in red, sums the Marginal CRISK for firms that have both positive total CRISK (Baseline + Marginal CRISK) and positive Marginal CRISK. This CRISK Exposure aggregate measures how much the system’s shortfall increases when increases in climate-related risks either expand a firm’s existing shortfall or (for those that do not have a shortfall) consume a firm’s entire capital buffer. We view it as a tool for assessing the extent to which climate-related risks add to systemic vulnerabilities.
Baseline and CRISK Exposure for the U.S. Financial System (end-month), 2001-October 2022
Source: NYU V-Lab and authors’ calculations.
Looking at the chart, we distinguish the period prior to 2019 from the past three years. While there were several instances before 2019 when CRISK Exposure rose significantly—in January 2010, December 2012 and April 2015—they were small relative to both the baseline and to what happened more recently. Thus, it was not until 2019 that movements in fossil fuel portfolios started to have a material impact on the conditional market value of financial firms. Then, things changed dramatically, possibly because investors took a less benign view of climate risk. Not only is there an enormous rise during the course of 2020, but the average level of CRISK Exposure rose from $22 billion in the 2009-2018 period to $144 billion over the period since January 2019. Interestingly, the estimate of CRISK Exposure peaked in November 2020 at nearly $700 billion, and then fell to between $50 to $250 billion over the next 15 months. Following Russia’s invasion of Ukraine in February 2022, CRISK Exposure has virtually disappeared.
The fluctuations in CRISK Exposure depicted in the chart arguably are associated both with changes in the expectation of climate-related losses and with changes in estimates of the uncertainty surrounding that expectation. Several factors can trigger such changes. Starting with physical risk, perceived increases in the frequency and severity of climate-related events (fires, hurricanes, typhoons, and the like) could lead to an increase in CRISK Exposure. Second, financial firms could increase or decrease their sensitivity to climate-related risks. Third, there could be a change in perceptions about the nature of future climate-based regulation (such as the likelihood of a carbon tax, restrictions that reduce fossil fuel demand, or subsidies for carbon-saving technologies).
To conclude, even for someone doubtful that climate-related developments can create financial instability, we now have low-cost tools that can help us assess their influence on the resilience of financial institutions. These tools use financial market prices and are available daily. Accordingly, we hope that intermediaries and their supervisors who monitor systemic risk vulnerabilities will add metrics like Marginal CRISK to their monitoring frameworks. Over time, we expect that modifications of CRISK to include market-based indicators other than the stranded-asset portfolio will make this monitoring effort increasingly helpful.
Acknowledgements: We thank Robert Engle and Richard Berner for very helpful discussions and comments, and Rob Capellini and Brian Reis of the NYU V-Lab for providing the data.