Grant: Discovery Launch Supplement
University of Guelph
$12,500.00
Enhancing the Accuracy and Interpretability of Market Risk Forecasts for Basel 3 Banking Regulation via Deep Learning with Feature Engineering
FY 2024-25
Summary
Beginning 2023, Canadian and global banks are required to implement Basel 3, which imposes a sweeping overhaul on how banks measure market risk and calculate capital requirements. At the center of Basel 3 is the change of the regulatory measure of financial market risk from Value at Risk (VaR) to Expected Shortfall (ES). Conventional models for forecasting VaR and ES typically use a limited number of variables and are unable to fully incorporate the sea of data, complex time-series patterns, and non-linear interactions among variables. Also, it is difficult for those models to spot the variables that are most important for VaR and ES. The Global Financial Crisis proves the necessity of developing more accurate and interpretable risk models.
The main purpose of the proposed research is to address deficiencies of conventional models by designing and developing deep learning (DL) models with feature engineering to improve the accuracy and interpretability of VaR and ES forecasts, which will benefit regulators and risk managers in their more informed decisions, and enable them to more proactively prevent or respond to future financial crises.
For the above purpose, I will achieve the following short-term (focusing on stocks) and long-term (focusing on more sophisticated asset classes) objectives (Obj.).
Obj. 1: Design and develop DL models with feature engineering of stock and accounting data from the CRSP and Compustat databases to capture firm characteristics; and 124 macroeconomic variables sourced from the Federal Reserve.
Obj. 2: Design and develop DL models with feature engineering of high-frequency stock data from the TAQ database.
Obj. 3: Design and develop DL models with feature engineering of stock option data from the OptionMetrics database.
Obj. 4: Design and develop DL models with textual information from the Dow Jones Newswire dataset, SEC filings, and earnings calls.
Obj. 5: Design and develop DL models with integrated variables from Obj. 1-4.
I will leverage the new "FZ" loss function derived from statistical decision theory to estimate parameters and measure variable importance. This approach marks the first integration of DL models with the "FZ" loss function, and big data for forecasting VaR and ES, with a capacity to identify critical drivers; and will significantly impact academia, regulators, and the financial industry.
My work will be shared with European Central Bank regulators as they expressed strong interest in such research during my presentation to them in 2020 and in follow-up discussions; and will be certainly shared with the Bank of Canada. This research will greatly benefit HQPs, who will be trained theoretically and practically in machine learning (ML), financial econometrics, and risk management, fields highly demanded in academia, government, and industry.
This proposal builds on the success of my recent research, reflected in my two publications in Journal of Financial Econometrics and my 5 working papers.
Program
Discovery Launch Supplement
Recipient
Location
Ontario, Canada
Recipient Type
Academia
Type
Grant
Source
NSERC's Awards Data
As published
Application id
DGECR-2024-00185
Name nom
Liu, Fulei
Department département
Economics and Finance
Organization id
24
Institution établissement
University of Guelph
Province en
Ontario
Province fr
Ontario
Country en
CANADA
Country fr
CANADA
Fiscal year exercice financier
2024
Competition year année de concours
2024
Award amount
12500
Program id
DGECR
Program name en
Discovery Launch Supplement
Program name fr
Supplément Tremplin vers la découverte
Group en
DISCOVERY RESEARCH
Group fr
RECHERCHE AXÉE SUR LA DÉCOUVERTE
Committee code
1508
Committee name en
Mathematics and Statistics
Committee name fr
Mathématiques et statistique
Area of application code
1205
Area of application group en
Advancement of knowledge
Area of application group fr
Avancement des connaissances
Area of application en
Mathematical sciences
Area of application fr
Sciences mathématiques
Research subject code
2960
Research subject group en
Applied mathematics
Research subject group fr
Mathématiques appliquées
Research subject en
Mathematical modelling
Research subject fr
Modélisation mathématique
Field of research list names en
Not available
Field of research list names fr
Non disponible
Socioeconomic objective list names en
Not available
Socioeconomic objective list names fr
Non disponible
Application title
Enhancing the Accuracy and Interpretability of Market Risk Forecasts for Basel 3 Banking Regulation via Deep Learning with Feature Engineering
Keywords
financial econometrics; time series; machine learning; artificial intelligence; value at risk; expected shortfall; risk management; deep learning; neural networks; basel 3
Application summary
Beginning 2023, Canadian and global banks are required to implement Basel 3, which imposes a sweeping overhaul on how banks measure market risk and calculate capital requirements. At the center of Basel 3 is the change of the regulatory measure of financial market risk from Value at Risk (VaR) to Expected Shortfall (ES). Conventional models for forecasting VaR and ES typically use a limited number of variables and are unable to fully incorporate the sea of data, complex time-series patterns, and non-linear interactions among variables. Also, it is difficult for those models to spot the variables that are most important for VaR and ES. The Global Financial Crisis proves the necessity of developing more accurate and interpretable risk models.
The main purpose of the proposed research is to address deficiencies of conventional models by designing and developing deep learning (DL) models with feature engineering to improve the accuracy and interpretability of VaR and ES forecasts, which will benefit regulators and risk managers in their more informed decisions, and enable them to more proactively prevent or respond to future financial crises.
For the above purpose, I will achieve the following short-term (focusing on stocks) and long-term (focusing on more sophisticated asset classes) objectives (Obj.).
Obj. 1: Design and develop DL models with feature engineering of stock and accounting data from the CRSP and Compustat databases to capture firm characteristics; and 124 macroeconomic variables sourced from the Federal Reserve.
Obj. 2: Design and develop DL models with feature engineering of high-frequency stock data from the TAQ database.
Obj. 3: Design and develop DL models with feature engineering of stock option data from the OptionMetrics database.
Obj. 4: Design and develop DL models with textual information from the Dow Jones Newswire dataset, SEC filings, and earnings calls.
Obj. 5: Design and develop DL models with integrated variables from Obj. 1-4.
I will leverage the new "FZ" loss function derived from statistical decision theory to estimate parameters and measure variable importance. This approach marks the first integration of DL models with the "FZ" loss function, and big data for forecasting VaR and ES, with a capacity to identify critical drivers; and will significantly impact academia, regulators, and the financial industry.
My work will be shared with European Central Bank regulators as they expressed strong interest in such research during my presentation to them in 2020 and in follow-up discussions; and will be certainly shared with the Bank of Canada. This research will greatly benefit HQPs, who will be trained theoretically and practically in machine learning (ML), financial econometrics, and risk management, fields highly demanded in academia, government, and industry.
This proposal builds on the success of my recent research, reflected in my two publications in Journal of Financial Econometrics and my 5 working papers.