Fred Liu

Assistant Professor of Finance

Gordon S. Lang School of Business and Economics, University of Guelph

fred.liu@uoguelph.ca

Research Interests: Artificial Intelligence, Machine Learning, Financial Econometrics, Asset Pricing, Risk Management

Curriculum Vitae

I am an Assistant Professor of Finance at the Gordon S. Lang School of Business and Economics at the University of Guelph, and an Adjunct Research Professor at the University of Western Ontario. I am affiliated with FIRM labs.

Publications

Regulatory Capital and Incentives for Risk Model Choice under Basel 3, Journal of Financial Econometrics, 2021 [Online Appendix]. Presented at the European Central Bank. (Joint with Lars Stentoft)

In response to the Subprime Mortgage crisis, the Basel Committee on Banking Supervision (BCBS) has spent the previous decade overhauling the regulatory framework that governs how banks calculate minimum capital requirements. In 2019, the BCBS finalized the Basel 3 regulatory regime, which changes the regulatory measure of market risk and adds new complex calculations based on liquidity and risk factors. This paper is motivated by these changes and seeks to answer the question of how regulation affects banks' choice of risk-management models, whether it incentivizes them to use correctly specified models, and if it results in more stable capital requirements. Our results show that, although the models that minimize regulatory capital for a representative bank portfolio also result in the most stable requirements, these models are generally rejected as being correctly specified and tend to produce inferior forecasts of the regulatory risk measures. 

Intraday Market Predictability: A Machine Learning Approach, Journal of Financial Econometrics, 2021. Invited by the Chief Editors. (Joint with Dillon Huddleston and Lars Stentoft)

Conducting, to our knowledge, the largest study ever of five-minute equity market returns using state-of-the-art machine learning models trained on the cross-section of lagged market index constituent returns, we show that regularized linear models and nonlinear tree-based models yield significant market return predictability. Ensemble models perform the best across time and their predictability translates into economically significant Sharpe ratios of 0.98 after transaction costs. These results provide strong evidence that intraday market returns are predictable during short time horizons, beyond what can be explained by transaction costs. Furthermore, we show that constituent returns hold significant predictive information that is not contained in market returns or in price trend and liquidity characteristics. Consistent with the hypothesis that predictability is driven by slow-moving trader capital, predictability decreased post-decimalization, and market returns are more predictable during the middle of the day, on days with high volatility or illiquidity, and in financial crisis periods. 

Working Papers

Quantile Machine Learning and the Cross-Section of Stock Returns, 2023.

We investigate the predictability of future stock return quantiles using machine learning models trained on firm characteristics and macroeconomic variables, and find that multi-task neural networks dominate linear, tree-based, and feed-forward neural network models. We introduce a quantile-based risk premium measure, robust to outliers and heteroskedasticity, and demonstrate that it delivers significant predictive and economic gains for investors, outperforming leading machine learning techniques. We construct machine learning skewness and volatility measures, and find a strong positive relationship between conditional skewness and average returns. We develop a portfolio strategy that incorporates machine learning Sharpe ratio forecasts, leading to significant economic gains.

Can AI Read the Minds of Corporate Executives? 2023. (Joint with Nicolas Chapados, Zhenzhen Fan, Ruslan Goyenko, Issam Hadj Laradji, and Chengyu Zhang)

It can. Using textual information from a complete history of regular quarterly and annual filings by U.S. corporations, we train classic machine learning algorithms and large language models, LLMs, to predict future earnings surprises. We first find that the length of MD&A section on its own is negatively associated with future earnings surprises and firm returns in the cross-section. Second, neither sentiment-based nor bag-of-words classic machine learning regression-based approaches are able to learn from the past managerial discussions to forecast future earnings. Third, only finance-objective trained LLMs have the capacity to understand the contexts of previous 10-Q (10-K) releases to predict both positive and negative earnings surprises, and future firm returns. We find significant, and often hidden in the complexity of presentations, positive and negative informational content of publicly disclosed corporate filings, and superior (to human and classic NLP approaches) abilities of more recent AI models to identify it.

Intraday Predictability Everywhere, 2023. Revise and Resubmit at Review of Asset Pricing Studies. (Joint with Lars Stentoft)
With approximately 900 million observations we conduct, to our knowledge, the largest study ever of intraday stock return predictability using machine learning techniques finding consistent out-of-sample predictability across market, sector, and individual stock returns at various time horizons. While linear models have the strongest statistical predictive power, nonlinear models economically dominate them and machine learning intraday long-short portfolios based on their forecasts attain Sharpe ratios of 4 after transaction costs. Predictability is short-lived, highest in the middle of the day, and more pronounced for less liquid firms, which indicates that slow-moving capital is an economic source of mispricing.

Can the Premium for Idiosyncratic Tail Risk be Explained By Exposures to its Common Factor?

Stocks in the highest idiosyncratic tail risk decile earn 7.3% higher average annual returns than in the lowest. I propose a risk-based explanation for this premium, in which shocks to intermediary funding cause idiosyncratic tail risk to follow a strong factor structure, and the factor, common idiosyncratic tail risk (CITR), comoves with intermediary funding. Consequently, firms with high idiosyncratic tail risk have high exposure to CITR shocks, and command a risk premium due to their low returns when intermediary constraints tighten. To test my explanation, I create a novel measure of idiosyncratic tail risk that is estimated using high-frequency returns, and theoretically establish its time-aggregation properties. Consistent with my explanation, CITR shocks are procyclical, correlated to intermediary factors, priced in assets, and explain the idiosyncratic tail risk premium. Furthermore, volume tail risk also earns a premium, follows a strong factor structure, and its common factor is priced. This duality of idiosyncratic tail risk and volume tail risk provides evidence for my risk-based explanation, and further supports the hypothesis that intermediaries' large trades cause idiosyncratic tail risk and volume tail risk from Gabaix et al. (2006).

Work in Progress

Presentations

Teaching (Mean Rating out of 5)

The CFA curriculum now includes Machine learning and Python. In my FIN*4000 and FIN*6200 courses, I teach the fundamental theories, mathematics (by hand), and applications (via Python) of machine learning, artificial intelligence, and NLP.

Grants

As Principal Investigator:

Mitacs Business Strategy Internship, 2023 (15k)

SSHRC Insight Development Grant, 2022-2024 (65k)

Digital Research Alliance of Canada RAC, 2022 (W22: 90k, W23: 54k)

Lang GRA Summer, 2022 (S22: 5k, S23: 5k)

SSHRC Institutional Grant, 2022-2024 (W22: 12k , F22: 3k)

Supervision

Prospective Students

I am a dedicated and active supervisor for graduate and undergraduate students. I supervise students excited about working on a range of topics including machine learning, financial econometrics, risk management, and asset pricing.

I am recruiting committed students at all levels (undergraduate, Master's, and PhD). Candidates can come from any quantitative field at the University of Guelph such as Accounting, Analytics, Computer Science, Economics, Engineering, Finance, Management, Marketing, Mathematics, Physics, Psychology, Natural Sciences, Statistics, etc. Most importantly, students must be fast learners, capable of independent problem solving, and have a strong passion for quantitative finance. 

If you attend the University of Guelph and are interested in applying, please email me your CV, a copy of your transcript, and a brief cover letter with your goals and interests. I will reach out to you if your academic background fits my research program.


Previous Student Training (post-graduation outcome)

Undergraduate:

Alison Xie, 2023 (Current Economics student, Ivey School of Business)

Mahir Mehta, 2023 (Current Management Economics and Finance student, University of Guelph)

Robert Proner, 2023 (Master of Economics, Western University)

Suresh Aubeeluck, (Master of Economics, University of Toronto)

Guillermo Parra, 2023 (Master of Economics, University of British Columbia)

Paul Cotturo, 2022 (Master of Quantitative Finance, University of Waterloo)

Mitchell Riddell, 2022 (Master of Financial Economics, University of Toronto)

Joseph Acernese, 2022 (Current Computer Science student, University of Guelph)