Working Projects

Carbon emission and asset prices: new evidence from machine learning (with Feng Li)


We estimate a large data panel of carbon emissions by US firms with a machine learning algorithm known as XGBoost. We predict scope 1 carbon emissions of listed firms from 2002 to 2021. This data set has a broad coverage of 4111 firms per year as compared to 1675 firms provided by data vendors. Based on this data set, we examine firms’ carbon risk pricing in the US equity market. The result shows that the carbon premium was insignificant before the Paris Agreement, and the premium turned significantly more negative after the Paris Agreement. This contrasting phenomenon implies a positive shift in investors’ ESG-related preferences and is more pronounced with our estimated data sample, where we provide a flow-based mechanism to explain the change in carbon premium. Overall, this paper complements rather than challenges previous empirical research from both sides and provides researchers with a novel approach to understanding climate finance.

Presentations at CICF 2023, 2023CFRI&CIRF Joint Conference, SBSICF 2023, CMCSR 2023. (Manuscript here, slides here)

Private Responsible Engagements and ESG Performance (with Danting Chang and Guanmin Liao )


Using unique data on institutional investors’ communications with the company management team during corporate visits, this paper reveals the existence and prevalence of institutional investors’ responsible engagement and its impact on firm ESG performances. Using an instrumental variable approach that isolates the increase in responsible engagement during the bad air quality days of the visits, we show that such engagement leads to subsequent higher ESG performance of the engaged companies. Moreover, this facilitating effect is more pronounced when companies are financially constrained, or when institutional investors exert more pressure, and after the Covid-19 pandemic when the optimal value-maximizing responsibility has arguably increased. Taken together, our findings provide novel empirical evidence that institutional investors proactively and responsibly engage with firms through private in-house meetings by voicing their ESG concerns, which serves as an important factor for companies to achieve better ESG outcomes.

Carbon awareness and return co-movement


This paper documents a rise in investors’ carbon awareness in recent years with a novel model-free approach. We show that for companies that produce carbon emissions on a similar scale, the correlation in their stock price returns is higher. This emission-return correlation only became significant after 2012 and has been steadily increasing ever since, whereas it was barely significant before 2012. We provide evidence showing that this co-movement is driven by investor flows as investors purchase green stocks and divest brown stocks, and is stronger when investors pay more attention to environmental news. To address the endogeneity issue, we adopt a state’s emission reduction initiatives that exogenously increase firms’ emission similarity. Overall, this paper examines whether investors care about carbon risk and the pricing of carbon risk from a different perspective.

(Manuscript here)

Work in Progress

Misclassified green patents

Peacock’s Feathers: Strategic disclosure in mutual fund annual reports (Seems like a forever working project)