Pulling back the curtain on bank climate stress tests
The Federal Reserve’s recent announcement of its decision to join the Network for the Greening of the Financial System (NGFS), coupled with a more climate-friendly policy agenda by the incoming Biden administration, makes it a better than even bet that a new set of regulatory stress tests assessing the banking system’s climate resiliency will materialize in the next year or so. Based on current data and models, implementation of such stress tests is another exercise in “feel good” regulatory policy that on the surface appears to be based on empirically rigorous methodologies that will generate highly uncertain outcomes of limited value to banks and policymakers.
Stress tests have been a favorite tool of bank regulators since the great financial crisis of 2008 as a mechanism to assess the industry’s ability to withstand specific economic scenarios. Having had a front row seat on the receiving end of the first regulatory imposed U.S. banking stress test, I am painfully aware that such analyses are fraught with a dizzying array of technical issues associated with the quality and granularity of data used to make assessments, assumptions and models relating to how macroeconomic variables affect bank financial performance and risk exposures over a relatively short (e.g., nine quarters) time horizon.
Significant strides in stress test analysis have been made over the last decade. But the value of the exercise is inherently limited by the specific set of scenarios evaluated. Underscoring this point is the Federal Reserve’s last-minute addition of a COVID-19 stress test imposed on banks this year, as the annual stress test scenarios were not designed to pick up the effect of a pandemic.
Momentum for climate change policies has accelerated with initiatives such as those by The Task Force on Climate-Related Financial Disclosures, the CFTC’s recent report on the financial system’s vulnerability to climate change and the NGFS’s climate scenarios for central banks and supervisors. Developing the data and analytical models to understand the linkages between climate change and the financial system is daunting and crucial; however, we must remain realistic about the validity of any of the proposed analyses and scenarios at this time.
The NGFS, a credible body taking a lead role in developing underlying climate change economic scenarios, acknowledges the significant uncertainty underlying each of its scenarios. Between the three scenarios proposed by the NGFS – orderly, disorderly and hot house – the decline in global GDP by 2100 ranges from -4 percent to -25 percent. This range should come as no surprise given the lengthy time horizon of each scenario. Reducing this to a 30-year horizon, as some regulatory authorities such as the Bank of England propose, still asks too much of the models.
The danger here is that regulatory authorities, policymakers and academics will build a false sense of confidence regarding the underlying science and accompanying models linking projected physical outcomes to financial performance. These models are technically very elegant complex systems of equations with fragile interdependencies among key variables, any number of which could lead to a wide variability in output depending on what data is being fed into the model. More concerning is that the validation of such models is not well known, further casting doubt on the stability of climate change models. After all, our historical economic data has never recorded a hot house climate scenario. The problem is that the endorsement of a set of climate change economic scenarios widely circulated today may lead policymakers to believe that an empirically valid and unassailable process has been used to develop scenarios that one day could very well be used to determine how much capital a bank should hold.
Peeling back the limitations on the current state of climate stress testing should not deter efforts to analytically pin down the long-term effects from climate change. We should acknowledge the severe limitations of any economic model to generate precision at a global level over a multi-decade horizon and begin our efforts at a more localized level to test drive these scenarios.
Specifically, efforts should focus more on understanding the direct linkages between asset performance and climate events such as hurricanes, floods and droughts for individual financial institutions facing these threats in their portfolios and then evolve those relationships over time to better understand indirect transition and macroeconomic effects. Imposing elaborate scenarios based on unproven models on banks risks losing credibility to those stakeholders who have to implement and ultimately make hard money decisions from those climate stress tests.
Clifford Rossi, PhD, is professor-of-the-practice and executive-in-residence at the Robert H. Smith School of Business at the University of Maryland.