Banks (and central bankers) on the verge of a (climatic) nervous breakdown
By Antonio Cabrales Goitia, Chair Professor of Economics (UC3M)
Last month I attended a very nice conference organized by our CSI Getafe–Leganés colleagues Pilar Perales and David Ramos on “Banking And Finance In Stressed Times: Climate, Resilience And Exit,” and I had to moderate a panel in which a group of central bankers told us about their experience about this crucial topic for the future. In the following lines I will tell you what I understood. I will first summarize each of the interventions, and then I will finish with a few thoughts they prompted on me. All the participants started with the usual very emphatic disclaimer that what they said represented their personal views and not those of their employer.
The first speaker, Carmelo Salleo, from the ECB explained to us a key tool for central banks’ risk management, stress tests, and how they work for climate change. First, they think of a (usually limited) number of scenarios about what can happen and their implications. To be specific, the baseline is an orderly transition with early and effectively implemented policies and limited costs from transition and physical risk. Then, another one where delayed policies are implemented, and with high costs from transition and average costs from physical risk. Finally, another scenario with no new policies implemented (only current policies), very limited costs from transition but extremely high costs from physical risk. I personally think that a combination of the second and third are more likely, to be honest.
Then, they get data from millions of firms and banks in the world, their current exposures to physical and transition risks and their exposures to one another, to understand the possible contagion possibilities. Naturally, the highest transition risk comes from the highest emitting sectors: mining, electricity, manufacturing. The physical risk hazards are very heterogeneous across countries. In Europe the South is more subject to wildfire, the North to floods. A few results emerge. The risk seems to be concentrated in a few institutions but for those it can be quite large. Also, new models are needed to understand how it can spread through contagion and macroeconomic impact. Finally, most countries are subject to higher transition than physical risk, however, a few countries are more vulnerable to high physical risk, Spain, together with Portugal and Greece, being some of them.
The second speaker, Miguel Molico, from the Bank of Canada, focused on assessing system risks. The Bank of Canada also does scenario analysis, but this presentation was more explicit on some of the limitations. For example, there is a lack of geographic and (granular) sectoral information on counterparties. It is also challenging to map financial institutions’ exposures to climate sensitive sectors and industries, there is limited counterparty emissions data and there is a lack of standardized approaches to assess risks over a long horizon.
He explained a very detailed pilot work they have done with the impact on the residential loan portfolio of banks. This is useful for various reasons. First, properties are used as collateral, and are immobile, thus susceptible to destructive climate events. Then, real-estate secured lending is amortized over decades and so more affected by climate change. Finally, it is significant portion of the lender balance sheets. This exercise is useful because it can solve some of the data problems, which are smaller, and can be used to tune up the methodology.
The final presentation was by Fernando Linardi of the Bank of Brazil. He shared the results of stress tests exercises done for Brazil. For physical risks, they find that, as in Europe, the risk is concentrated on a few firms that are more highly exposed to droughts and floods. For transition risks, they find that credit exposures to firms in GHG intensive sectors are small compared to the credit outstanding.
But perhaps the most interesting aspect of this last presentation was the description of the challenges. As the other presenters, he emphasized data gaps or how to translate climate related events into bank’s financial losses. But I found significant three aspects he brought to the table. First, climate related events may be subject to severe non-linearities (a.k.a. tipping points). Then, that firms’ and banks’ can adaptation measures, so balance sheets are dynamic. Finally, in the same vein, we need to learn how to model the effect of technological change, innovation and policy.
And this brings me to my own reflections on the problem. I think the current models are of course well-intentioned and probably the best that can be done, but they are sorely lacking to address the problem. Trying to understand a dynamic problem with very strong externalities and backward and forward loops with a static matrix of cross firm exposures is extremely naïve.
As the situation unfolds, the banks and firms, not to mention governments and regulators, are going to react to one another’s actions. This could, on one dimension mitigate effects, as banks will start cutting exposures to high emitting firms. This makes them become more isolated and thus less prone to create contagion, and also their activity will subside. On the other hand, this could happen very quickly and more many firms at the same time, so the disruption in their output and employment could create a large and sudden stop in macroeconomic activity, which would cause major problems for the economy.
All of the above mentioned problems indicate that the right tool to understand the problem is a dynamic stochastic general equilibrium (DSGE) model for the economy, and one of the more difficult ones, since it cannot concentrate on steady states of the problem. The transitional dynamics are crucial, and it needs to consider the heterogeneity of the agents involved. So, strictly speaking a HANK (heterogeneous agents New Keynesian) DSGE model. The Central Banks do have economists capable of churning out DSGE HANK models. I wonder why the regulators are not using those, instead of these more traditional stress test approaches, which are totally OK for short run risk management. Maybe some of our readers will know and be kind enough to enter the discussion at this point.