SEM217: Nick Gunther, Infima: NLP-based Information Extraction Applied to MBS Offering Documents
Tuesday, April 30th @ 11:00-12:30 PM, 648 Evans Hall and Zoom
Large Language Models ("LLMs") have impressed researchers and observers with their success at classification, translation, text generation and other standard NLP tasks. Starting with word2vec in 2013 and accelerating to contemporary transformer models such as BERT, researchers have continued to discover exciting new applications and improve existing ones.
Because financial data is largely numerical and NLP’s prominent successes arose in unrelated areas, applications to finance have generally lagged outside a few bespoke areas, such as sentiment analysis for stock price prediction.
SEM217: Jose Menchero, Bloomberg: Evaluating and Comparing Risk Model Performance
Tuesday, January 23rd @ 11:00-12:30 PM, 648 Evans Hall and Zoom
Factor risk models are used for three primary purposes: (1) predicting portfolio volatility, (2) portfolio optimization, and (3) decomposing risk and return into factor and idiosyncratic components. In this paper, we propose explicit tests or “horse races” to evaluate and compare the performance of competing risk models along these three basic dimensions.
SEM217: Zachary Feinstein, Stevens Institute of Technology: Implied Volatility of the Constant Product Market Maker in Decentralized Finance
Tuesday, January 30th @ 11:00-12:30 PM, via Zoom
Automated Market Makers (AMMs) are a decentralized approach for creating financial markets by allowing investors to invest in liquidity pools of assets against which traders can transact. Liquidity providers are compensated for making the market with fees on transactions. The collected fees, along with the final value of the pooled portfolio, act as a derivative of the underlying assets with price given by the pooled assets. Following this notion, we study the implied volatility constructed from the constant product market maker.
SEM217: Michael Bauer, Eric Offner, Glenn Rudebusch: The Effect of U.S. Climate Policy on Financial Markets: An Event Study of the Inflation Reduction Act
Tuesday, February 6th @ 11:00-12:30 PM, 648 Evans Hall and Zoom
The Inflation Reduction Act of 2022 (IRA) represents the largest climate policy action ever undertaken in the United States. Its legislative path was marked by two abrupt shifts as the likelihood of climate policy action fell to near zero and then rose to near certainty. We investigate equity price reactions to these two events, which represent major realizations of climate policy transition risk. Our results highlight the heterogeneous nature of climate policy risk exposure. We find sizable reactions that differ by industry as well as across firm-level measures of greenness such as environmental scores and emission intensities. While the financial market response to the IRA was economically significant, it did not lead to instability or financial stress, suggesting that transition risks posed by climate policies even as ambitious as the IRA may be manageable.
SEM217: Ruimeng Hu, UC Santa Barbara: Directed Chain Generative Adversarial Networks for Multimodal Distributed Financial Data
Tuesday, February 13th @ 11:00-12:30 PM, Zoom
Real-world financial data can be multimodal distributed, and generating multimodal distributed real-world data has become a challenge to existing generative adversarial networks (GANs). For example, neural stochastic differential equations (Neural SDEs), treated as infinite-dimensional GANs, are only capable of generating unimodal time series data. In this talk, we present a novel time series generator, named directed chain GANs (DC-GANs), which inserts a time series dataset (called a neighborhood process of the directed chain or input) into the drift and diffusion coefficients of the directed chain SDEs with distributional constraints. DC-GANs can generate new time series of the same distribution as the neighborhood process, and the neighborhood process will provide the key step in learning and generating multimodal distributed time series. Signature from rough path theory will be used to construct the discriminator. Numerical experiments on financial data are presented and show a consistent outperformance over state-of-the-art benchmarks with respect to measures of distribution, data similarity, and predictive ability. If time permits, I will also talk about using Signature to solve mean-field games with common noise.
SEM217: Aaron Yoon, Northwestern Kellogg: Return on Environmental Investments: First Evidence from Chinese Regulatory Filings
Tuesday, February 20th @ 11:00-12:30 PM via Zoom
From ESG ratings and current public disclosures on ESG in the U.S., investors cannot disentangle how much firms invest in ESG and the associated ESG performance. To overcome this problem, we use a setting in China, in which regulators mandate firms to disclose the actual amount of environmental investment in annual reports. We examine whether and how environmental investments relate to emission reductions, and how the market reacts to this information. We find that firms that invest more in the environment exhibit greater emission reduction, while changes in environmental ratings from ESG raters fail to predict emission reduction. For firms that have a greener narrative, are less financially constrained, and face higher environmental demand from investors, we observe a more positive relation between investments and reductions. In addition, this positive relation is also strengthened if local governments have more environmentally friendly policies. We find that the market on average penalizes firms that make more environmental investments, but the amount of emission reduction moderates this penalty. Overall, the findings highlight the decision usefulness of newly released information on environmental investment. It has implications for regulators across the globe that are considering mandating certain ESG-related information.
SEM217: Alicia Montoya & Thomas Phillips, Swiss Re: Quantum Cities™ - Data-centric Approaches for Resilient, Sustainable Cities
Tuesday, February 27th @ 11:00-12:30 PM, via Zoom
More than 56% of today's global population lives in cities, set to reach 70% by 2050. 80% of global GDP is generated in cities. Cities consume two-thirds of global energy and account for more than 70% of greenhouse gas emissions.Much of this urban growth is happening in densely populated and rapidly urbanising river plains and coastlines in developing countries, where 89% of the world’s flood-exposed people live. Cities therefore play an important role in tackling climate change, both due to the efficiencies dense urban environments offer in terms of climate mitigation and adaptation measures, but also due to their exposure to growing disaster risks. Cities also increasingly depend on global, interconnected supply chains, operating in volatile, climate-challenged environments— not to mention the constant risk of earthquakes. These increased risks and interdependencies mean risk events can quickly propagate globally, driving large and sometimes catastrophic losses, and leaving cities with potentially disastrous supply gaps. Starting with (entangled) urban data and usingphysics-based, hybridised machine learning approaches,re/insurers can nowmodel risks like urban flood as well as business interruption risk propagation in supply chains so as to predict, prevent, and manage risks to reduce losses and enable urban resilience.
SEM217: Shachar Kariv, UC Berkeley, Ever Since Allais and Ellsberg
Tuesday, March 5th @ 11:00-12:30 PM, 648 Evans Hall and Zoom
The Allais critique of expected utility theory (EUT) has led to the development of theories of choice under risk that relax the independence axiom, but which adhere to the conventional axioms of ordering and monotonicity. Unlike many existing laboratory experiments designed to test independence, our experiment systematically tests the entire set of axioms, providing much richer evidence against which EUT can be judged. Our within-subjects analysis is nonparametric, using only information about revealed preference relations in the individual-level data.
SEM217: Jaeyeon Lee, UC Berkeley Haas: Agency CMBS: The Rise of ARM Share
Tuesday, March 19th @ 11:00-12:30 PM, 648 Evans Hall and Zoom
This paper investigates the increase in adjustable-rate mortgage (ARM) credit volume shares in the government-sponsored enterprise(GSE) commercial mortgage-backed securities(CMBS) market since the global financial crisis. We find evidence of borrowers exploiting prepayment options embedded in ARM contracts during low policy rate periods, which coincide with upside collateral value cycle. Moreover, we find that ARM borrowers, who are landlords, pass-through interest rate risks to renters during federal funds rate hike periods while they do not pass-through declines in interest rates. Through a stylized two-period model, we show that the two options triggered in two different monetary policy regimes can affect mortgage choice among borrowers in the GSE CMBS market.
SEM217: Sunil Wahal, Arizona State University: R&D, Innovation, and the Stock Market
Tuesday, April 2nd @ 11:00-12:30 PM, 648 Evans Hall and Zoom
We investigate the relation between inventive input (R&D), inventive output (the economic value of patents, EVP), firm-level profitability and asset growth, and stock returns. Current R&D and EVP forecast future profitability. Neither forecast future asset growth. Factor models motivated by q-theory and the dividend discount model fail to price R&D and EVP correctly, leaving large alphas on the table. But model failure is due to design specifics, not economic underpinnings: using cash-based operating profitability to measure expected profitability resurrects both models.
SEM217: Anders Christjansen, Morten Larsen, Kalle Mickelborg, Innovation Centre Denmark
Tuesday, April 9th @ 11:00-12:30 PM, 648 Evans Hall and Zoom
SEM217: Markus Pelger, Stanford: Shrinking the Term Structure
Tuesday, April 16th @ 11:00-12:30 PM, 648 Evans Hall and Zoom
We propose a new framework to explain the factor structure in the full cross section of Treasury bond returns. Our method unifies non-parametric curve estimation with cross-sectional factor modeling. We identify smoothness as a fundamental principle of the term structure of returns. This implies factors that are investable portfolios and correspond to unique spanning basis functions with decreasing order of smoothness. These reflect and thus explain the slope and curvature shapes frequently encountered in PCA. In a comprehensive empirical study, we show that the first four factors explain the time-series variation and risk premia of the term structure of excess returns. Cash flows are covariances as the exposure of bonds to factors is fully explained by cash flow information. The fourth factor, which captures complex shapes of the term structure premium, substantially reduces pricing errors and pays off during recessions.
SEM217: Baeho Kim, Korea University:Long-History PCA under Dynamic Factor Model with Weaker Loadings (Joint work with Robert M. Anderson and Donghan Ryu)
Tuesday, April 23rd @ 11:00-12:30 PM, 648 Evans Hall and Zoom
The accurate estimation of the covariance matrix and its inverse (= precision) matrix of asset returns significantly shapes the range of admissible portfolio compositions and the potential magnitude of associated losses, given the portfolio manager's predetermined risk budget. In this study, we propose Long-History PCA (LH-PCA), which uses longer data histories (T_L), such as 1500 trading days or six years, to predict daily portfolio risks. We show that LH-PCA consistently estimates dynamic factor models with essentially arbitrary variable volatility structures and time-varying loadings with heterogeneous strengths. By using a longer data history, the excess dispersion bias of minimum-variance optimized portfolios can be effectively mitigated by reducing the finite-sample correlations between factor returns and idiosyncratic returns, particularly in the presence of weaker factors. Combined with Responsive Covariance Adjustment (RCA) using a short half-life (T_S) of 40 days, our proposed approach offers substantial improvements in risk prediction for minimum-variance portfolios compared to alternative specifications, including traditional approaches using a medium horizon (T_M) of one or two years, in both simulations and empirical studies.