Hello!
I am a fifth-year PhD student in finance at Copenhagen Business School, and I am currently on the job market.
My primary research area is empirical asset pricing with a particular focus on how to use machine learning in financial settings. I also have industry experience as a power/energy trader. For more information about me or my research, check out my CV, or reach out:
E-mail: kh.fi@cbs.dk
Working Papers:
Abstract:
This paper proposes a new type of modelling framework that use machine learning techniques to estimate the parameters of structural models: Deep Structural Models (DSMs). I implement a DSM with a simple Merton (1974) model as a foundation, and show that the DSM jointly estimates expected equity returns and (co)variances with higher predictive power than leading benchmark models. I form long-short and mean variance efficient portfolios with significantly higher average excess returns, alphas, and Sharpe ratios, compared to those formed on the basis of a state-of-the-art machine learning model. Economically, the DSM suggests that systematic risk compensation is the largest contributor to the average expected equity return of firms, while mispricing is the primary driver of the dispersion of expected returns. Finally, the DSM provides evidence that firm leverage is the main reason for an increased equity premium during economic recessions.
With Peter Feldhütter and Arthur Krebbers
R&R at the Journal of Financial Economics
Abstract:
We examine the pricing of sustainability-linked bonds (SLBs), where the cash flows depend on the bond issuer achieving one or more Environmental, Social and Governance (ESG) goals. Investors are willing to accept a 4-7bps lower yield due to the bond's ESG label, providing evidence of investors caring about environmental impact. Furthermore, we find the average probability of meeting the target is 73% so firms set ESG targets that are easy to reach. We find that the SLB market is efficient: the prices of SLBs depend strongly on the size of the potential penalty and there is no evidence of mispricing. Finally, our results suggest that SLBs serve as financial hedges against ESG risk.
Abstract:
A simple decomposition of the expected returns of merger arbitrage trades, whose individual parts are modelled by modern machine learning techniques, lead to better proxies for expected returns than realized merger arbitrage returns. These decomposed expected return estimates yield large economic gains for merger arbitrage investors in terms of both absolute and risk-adjusted returns. Furthermore, the decomposed expected return estimates grant better financial insight into the evolution of the merger arbitrage market, as well as the risk premium associated with each M&A deal.
Machine Learning and Financial Stability in Credit Markets
With Peter Feldhütter, Tarun Ramadorai, and Ansgar Walther
Draft Coming Soon
Abstract:
In this paper, we assess the performance of state-of-the-art machine learning models for predicting firm-level default probabilities across the entire term structure. Our analysis reveals that models based on decision tree algorithms consistently outperform traditional statistical models across all time horizons. Further, we investigate the implications of adopting these machine learning models within the banking sector for predicting defaults, with a specific focus on their impact on financial stability. Our findings suggest that constructing loan portfolios based on these advanced predictive models, does not decrease financial stability, although the resulting loan portfolios are much less diversified.