Climate Change and Long-Horizon Portfolio Choice: Combining Theory and Empirics
“Climate change will affect the value of every single financial asset”, as argued by Bank of England Governor Mark Carney during the Bloomberg Global Business Forum in 2019. “And the question is: Is the financial sector ready for it?” The Dutch central bank also recognizes that financial institutions are exposed to climate risk and encourages pension funds to quantify the risks related to Environmental, Social, and Governance (ESG) factors and to refine their policies for sustainable investing (DNB (2018, 2019)). Dutch pension funds are increasingly focusing on sustainable investment strategies (e.g., APG (2018) and PGGM (2018)), not only to have a positive impact on society, but also because they are aware that environmental factors can pose large investment risks. However, despite their efforts, institutional investors still struggle to fully grasp the potential impact of climate change on the value of their portfolio and are searching for better approaches to quantify these risks (see, e.g., surveys by KPMG (2019) and Krueger, Sautner, and Starks (2019)). In fact, in his 2020 letter to CEOs, Larry Fink writes that “climate change is almost invariably the top issue that clients around the world raise with BlackRock.” An important obstacle in measuring climate risk is the long-term nature of climate change, which creates uncertainty about its long-horizon impact on asset prices and risk. Professor Robert Stambaugh highlights this uncertainty in a New York Times (2009) article by noting that when we “expand the horizon to the next several decades, the possible effects of global warming range from negligible to catastrophic.” Because pension funds have a long-term investment horizon, they should be particularly concerned about these long-term effects. Existing work shows that climate change causes disasters that hurt the economy and financial markets (e.g., Botzen (2018) and Colacito, Hoffmann, and Phan (2019)). However, the frequency and impact of previous disasters may not be very informative for future events in a scenario of prolonged climate change. As a result, historical data may underestimate the long-term risks induced by climate change.
The goal of our research project is to address these issues by using a novel Bayesian approach that supplements historical data on the impact of climate change on long-horizon asset returns and risk with prior information derived from economic theory. Specifically, we elicit prior beliefs from the temperature long-run risk model of Bansal, Kiku, and Ochoa (2019), in which rising temperatures influence asset prices by increasing the likelihood of future climate disasters. By imposing a structure on the relation between temperature change and asset returns, the model provides information about the financial impact of climate change. Our approach incorporates this additional prior information in the modeling of asset return dynamics to obtain more precise estimates of long-term climate risks than existing methods that solely rely on historical data.
In a nutshell, we estimate a Vector Autoregression (VAR) model to analyze the long-horizon dynamics of asset risk and returns, as is common in the strategic asset allocation literature (e.g., Pastor and Stambaugh (2012)). To capture climate risk, we augment the model with various return predictors related to climate change, such as temperature change and carbon emissions. We derive prior beliefs about the VAR parameters from the temperature long-run risk model. Our analysis initially focuses on equity markets in the Netherlands and in the United States, including both the aggregate market indices as well as green and brown stock portfolios. These portfolios are constructed based on observable environmental characteristics such as a firm’s water usage and carbon emissions and the breakdown of its revenues into green and brown activities. Our flexible Bayesian approach can be easily extended to other countries and other asset classes.
What is the impact of climate change on the optimal portfolio choice of long-term investors such as pension funds?
How do portfolios formed using existing methods, such as screening firms based on their carbon footprint, compare to the optimal portfolio constructed using our novel approach?