In a recent New York Fed Staff Report (Climate Stress Testing), Hyeyoon Jung, Robert Engle and Richard Berner estimate the exposure of large global banks to climate transition risk. This type of climate risk shock could arise from policies aimed at reducing greenhouse gas emissions, such as the introduction of a carbon tax or regulatory pressure on banks to avoid loans to fossil-fuel firms. The paper estimates that some large U.S. banks would need to raise tens of billions of dollars in capital if there were a severe climate-related shock, and label this estimate “CRISK.” Probably because of the magnitude of the estimated levels of capital shortfalls of large U.S. banks, the paper has received a fair amount of press coverage. Our objective in this post is to explain why it is unlikely that the paper is measuring banks’ exposure to climate-related transition risks, and instead is mixing correlation between banks’ and fossil fuel firms’ stocks with causation. As a result, the paper needs to tighten the identification of the impact of changes in climate risk on the value of bank equity.
A key driver of the results of the paper is an estimate of how banks’ equity valuations respond to declines in the equity values of fossil fuel firms. That response is referred to as the “climate beta” of each bank. This climate beta is estimated from a regression of the bank’s stock return on the return on the S&P 500 and on the return on a portfolio of fossil fuel stocks. The portfolio of fossil fuel stocks is labeled the “stranded asset portfolio” in the paper, but it is no more nor less than a portfolio of fossil fuel stocks. The authors interpret the regression coefficient—that is, how much bank stocks moved up or down when fossil fuel stocks moved up or down (or down and up, as we discuss below)—on the portfolio of fossil fuel stocks as measuring the bank’s exposure to transition-related climate risk.
One would expect that banks with large amounts of loans in the fossil fuel industry will be more sensitive to climate risk on average and will have a positive climate beta. However, in the case that a bank holds a large amount of loans in the renewable energy sector, the bank’s climate beta could be negative. (p. 7)
For example, a stock portfolio that consisted of one-quarter of fossil fuel firms and three-quarters the S&P 500 would have an estimated climate beta of 0.25 using the technique employed in the paper.
However, the estimates of the banks’ climate betas vary far too much over time to be a reliable measure of the composition of the banks’ portfolios. The estimates of each U.S. large bank’s climate betas are shown in Figure 27 from the appendix to the paper, which is shared here. The estimates are the extremely thin black lines. The thick blue lines are six-month moving averages. As the figure makes clear, the estimated climate betas exhibit large and rapid swings. For example, the climate beta of Morgan Stanley (MS) appears to move from about –0.4 on one day in 2008 to about –1.0 a few days later and then quickly back up to –0.4 (see a blow-up of the MS panel reprinted further below). Such moves are also seen in the climate betas of other banks.
Figure 27: Climate Beta of U.S. Banks
Going by the authors’ interpretation of the structural meaning of the climate betas quoted, Morgan Stanley had 40 percent of its assets invested in green energy in 2008, shifted to 100 percent in green firms over a few days, and then cut back to 40 percent. Over the next few years, MS evidently continued to make large high-frequency changes in its exposures to the fossil fuel industry and to green energy firms. This outcome appears highly implausible.
Instead, it seems likely that rather than uncovering large high-frequency swings in the composition of each bank’s portfolio of assets, the climate betas are measuring something other than the banks’ exposure to climate risks. For example, the authors note, “. . . we observe a common spike in the year 2020 as banks’ exposures to the transition risk rose substantially due to a collapse in energy prices.” While it is clear why a bank with a high exposure to the fossil fuel industry would experience a bad stock return when energy prices fell, it is not clear why a bank’s exposure to the fossil fuel industry would go up because energy prices fell.
An Alternative Interpretation of The Results
A more plausible explanation for the increase in climate betas during the peak of the pandemic in 2020 is simply that energy and bank stocks both took an extraordinary beating when COVID-19 hit. Even after controlling for the general decline in the S&P 500, the drop in bank stocks and energy stocks was outsized. The drubbings happened not because banks were exposed to energy firms and dragged down by their performance, but rather because both sectors were highly exposed to the COVID-19 shock that forced the economy to effectively shut down. In addition, as shown in Figure 27 from the paper, climate betas have since declined relative to the peak levels reached during the early stages of the pandemic, as the more typical and largely uncorrelated movements of bank and fossil fuel stocks (after controlling for the S&P) roll into the sample.
In our view, the estimated climate betas are not especially meaningful under the current methodology used in the paper. Most of the time, the coefficients of bank stock returns on fossil fuel stock returns are about zero; however, sometimes they are negative and sometimes positive, depending on unrelated shocks that move bank and energy stocks around. As shown in Panel A of Figure 1, in the spring of 2020 the world was hit by a macroeconomic shock unprecedented in its speed and ferocity. Fear of massive loan defaults and of a global depression led both bank and energy stocks to tank. The co-movement between bank stocks and energy stocks was very high at the onset of the COVID event. Between Jan. 1 and March 23, 2020, bank stocks fell 49 percent and energy stocks declined 61 percent, while the S&P 500 only declined 31 percent. By the end of June 2020, both bank and energy stocks were down 40 percent relative to Jan. 1, while the S&P 500 was nearly back to the levels registered at the start of that year.
The estimated climate beta in the regressions conducted in the paper jumped because of this co-movement. As that jump rolls into the moving averages reported there, the moving averages trend up sharply, as shown in Figure 2 (from the paper). Using those temporarily high coefficients, the authors get newsworthy estimates of how much further bank stocks would fall if fossil fuel stock prices collapsed. (The six-month moving averages of the climate betas end at a sample high in Figure 2 from the Report’s text but decline noticeably in Figure 27 from the paper’s appendix, because the period plotted in Figure 2 ends sooner than in the period plotted in Figure 27.)
We can demonstrate the reasonableness of our alternative explanation by also inspecting the path of bank and energy stock returns during the 2007–2008 global financial crisis. As can be seen in the paper’s Figure 2, at the end of 2007 and in early 2008, large banks’ climate betas were mostly all negative and in almost all cases reached their lowest value over the entire sample. The authors’ interpretation of the results quoted above is that those negative betas reflected heavy but temporary investments in the renewable energy sector. Our interpretation is that there was a shock at the time that moved both bank and energy stocks in the opposite direction. Indeed, as shown in Panel B of Figure 1, we see that between Jan. 1, 2007, and June 30, 2008, bank stocks fell 57 percent and fossil fuel stocks rose 49 percent, while the S&P 500 was little changed.
Recently, the Financial Stability Oversight Council published a report on climate-related financial risk. The report states that because of the impact of climate change in the U.S. economy, economic adjustments must be made to reduce greenhouse gas emissions, and those adjustments present banks with risks as well as opportunities. The econometric methodology developed by Jung, Engle and Berner (2021) can be helpful in assessing climate-related transition risks by combining market data with information on banks’ own exposures to fossil-fuel firms. However, as our analysis demonstrates, the methodology does not allow for a clear separation between the effect climate risks and other developments have on bank equity valuations. The authors of the New York Fed Staff Report would need to tighten the identification of climate risks in their methodology so that “CRISK” can be interpreted as measuring banks’ exposure to climate-related transition risk.
 To be precise, the analysis uses the return on an arbitrage portfolio that is long fossil-fuel stocks and short the S&P 500. But because the S&P 500 is also included in the regression, the regression coefficient on the arbitrage portfolio is identical to that on the fossil fuel stocks (though the coefficient on the S&P 500 will change).