Abstract
This thesis analyses the financial implications of climate transition risk. It brings new insights to the topic from three perspectives: (i) the prediction of corporate carbon footprints using machine learning techniques, (ii) a bottom-up loan-level stress-test of the exposure of leading United States [U.S.] banks to climate transition risk in their syndicated loan portfolios, and (iii) an empirical analysis of climate distress risk at the corporate level for S&P 500 non-financial firms.
The first essay in Chapter 2 predicts corporate carbon footprints with publicly available data and machine learning techniques. Carbon footprints are a popular proxy for transition risk at firm-level, as they indicate the amount of greenhouse gas emissions associated with firms and thus their contribution to global warming. The need for prediction arises because only a limited subset of firms disclose emissions, while in contrasts there is growing interest from investors and regulators for emissions data. A two-step framework that uses a Meta-Elastic Net learner to combine predictions from multiple base-learners is the best emission prediction approach. It results in an accuracy gain based on mean absolute error of up to 30% versus the existing models. Further, it shows that the prediction accuracy can be improved with additional energy predictors and additional disclosures in particular sectors and regions.
The second essay in Chapter 3 examines banks’ exposure to climate transition risk. A bottom-up, loan-level methodology incorporating climate stress-test based on the Merton distance to default model and IPCC transition pathways is implemented. Estimations of corporate carbon footprints are matched to syndicated loans initiated in 2010-2018 and aggregated to the loan portfolios of the twenty largest banks in the U.S. Descriptive results indicate that banks vary in their climate transition risk due to their considerable exposure to the energy sectors and due to borrowers’ carbon emission profiles from other sectors. Banks’ transition risk profiles are stable over time, save for a temporary (in some cases) and permanent (in others), reduction in their fossil-fuel exposure after the Paris Agreement. Stress-testing results show that the median loss is 0.5% of loan value, extrapolated to be a 4.1% decrease in core capital [CET1] but this is much larger in the 1.5oC scenarios (12%-16% of CET1 capital) and there is significant tail-end risk (62% of CET1 capital). Banks’ vulnerabilities are driven by the ex-ante financial risk of their borrowers more generally, highlighting that climate risk is not independent from conventional risks.
Finally, the third empirical essay in Chapter 4 investigates whether climate risk affects firm-level default (distress) risk. Using the Merton distance to default model, it explores for the presence of climate distress risk in S&P 500 non-financial firms during 2010-2018, employing both corporate carbon footprints and climate risk disclosure in 10K filings. The system generalized method of moments regression shows that climate risk has a negative impact on firms’ distance to default, but this impact is limited to transition risk and the disclosures of this risk in annual filings. Meanwhile, backward-looking measurements of transition risk such as corporate carbon footprint (Scope 1 and Scope 2) or disclosure of physical risk do not have a similar effect in the U.S. context. The essay also indicates that the Paris Agreement temporarily strengthens the negative relationship between climate risk and distance to default in the year 2015. However, this effect is short-lived and fades away in the later years.