Quick Clips: Can Machine Learning Help Manage Climate Risks?

With climate change becoming an ever-increasing consideration for investors, we have taken a closer look at ways to hedge against the realization of climate risk. Although there is a growing body of research around ways to reduce carbon exposure in portfolios and even achieve net negative exposure, emissions data is just a small piece of the larger issue for investors. 

This is a long-term issue with no single solution. However, we explore one possible way of managing climate risk: the implementation of a hedge in an investor's portfolio. In this case, we look at language in the news that references climate change and use it as data to help position the portfolio against longer-term risks.  

AQR Head of Machine Learning Bryan Kelly on building a portfolio well positioned against climate risk (0:48):

One of the challenges of constructing a hedge against climate risk is distilling the information on climate change across news sources and separating out the relevant “data.” We think about this in a few ways: investors can either associate increased climate change reporting with a higher climate risk, or they can focus on negative climate news specifically to determine impending risk. In our research, we look at the frequency of the news and its correlation to price returns, quantifying the data set of language.

Interpreting and quantifying climate risk in the media (0:59):

First, we create a vocabulary based on academic climate change research, and then we compare it to a major news source. By looking at what portion of the news source correlates with the climate change vocabulary over time, we can identify periods where there are spikes in climate news. In terms of implementing the hedge, we seek to build a portfolio of assets that perform well during these spikes. The portfolio is constructed to be well positioned against future materialization of climate change. In addition, our research found that many portfolios with the goal of offsetting climate risk are primarily underweight the energy sector, so we constructed an industry-neutral portfolio to maintain diversification benefits. The outcome? We found that if the portfolio is successful in hedging climate risk in the near-term, it has the potential to appreciate in reaction to climate news over the long term, which we view as its ultimate goal. 

How to measure a successful hedge against longer-term climate risk (0:34):