In the Beginning Was the Word: LLM-VaR and LLM-ES

37 Pages Posted: 20 Jan 2025 Last revised: 20 Jan 2025

See all articles by Daniel Traian Pele

Daniel Traian Pele

Bucharest University of Economic Studies; Romanian Academy - Institute for Economic Forecasting

Vlad Bolovaneanu

Bucharest University of Economic Studies

Min-Bin Lin

affiliation not provided to SSRN

Rui Ren

Humboldt University of Berlin

Andrei Theodor Ginavar

Bucharest University of Economic Studies

Bruno Spilak

Bucharest University of Economic Studies

Alexandru Victor Andrei

Bucharest University of Economic Studies

Filip Toma

Bucharest University of Economic Studies

Stefan Lessmann

School of Business and Economics, Humboldt-University of Berlin

Wolfgang Karl Härdle

Blockchain Research Center Humboldt-Universität zu Berlin; Charles University; National Yang Ming Chiao Tung University; Asian Competitiveness Institute; Academy of Economic Studies, Bucharest

Date Written: November 01, 2024

Abstract

This study introduces LLM-VaR and LLM-ES, novel approaches utilizing general-purpose large language models (LLMs) for zero-shot forecasting of Value at Risk (VaR) and Expected Shortfall (ES). Using the LLMTime framework, these methods process financial time series data encoded as numerical strings, providing a flexible, assumption-free alternative to traditional risk estimation models such as GARCH and EWMA. Our empirical analysis reveals that LLMs perform effectively within a short-term historical context, particularly in highly volatile markets like cryptocurrencies. However, as the historical context lengthens, the accuracy of LLM-based methods diminishes, with conventional models proving superior for capturing long-term dependencies. These findings highlight the potential of LLMs as adaptable tools for risk assessment over recent historical windows, while underscoring the continued importance of traditional models for robust, long-term financial risk management.

Keywords: Value at Risk, Expected Shortfall, GPT, LLM-VaR, LLM-ES, Large Language Models

Suggested Citation

Pele, Daniel Traian and Bolovaneanu, Vlad and Lin, Min-Bin and Ren, Rui and Ginavar, Andrei Theodor and Spilak, Bruno and Andrei, Alexandru Victor and Toma, Filip and Lessmann, Stefan and Härdle, Wolfgang Karl, In the Beginning Was the Word: LLM-VaR and LLM-ES (November 01, 2024). Available at SSRN: https://ssrn.com/abstract=5104383 or http://dx.doi.org/10.2139/ssrn.5104383

Daniel Traian Pele (Contact Author)

Bucharest University of Economic Studies

Piata Romana nr. 6
Bucharest
Romania

Romanian Academy - Institute for Economic Forecasting ( email )

Calea 13 Septembrie nr. 13
Bucharest, 050711
Romania

Vlad Bolovaneanu

Bucharest University of Economic Studies ( email )

Min-Bin Lin

affiliation not provided to SSRN

Rui Ren

Humboldt University of Berlin ( email )

Unter den Linden 6
Berlin, AK Berlin 10099
Germany

Andrei Theodor Ginavar

Bucharest University of Economic Studies ( email )

Bruno Spilak

Bucharest University of Economic Studies ( email )

Tache Ionescu street, no. 11, sector 1
Bucharest
Romania

Alexandru Victor Andrei

Bucharest University of Economic Studies ( email )

Filip Toma

Bucharest University of Economic Studies ( email )

Tache Ionescu street, no. 11, sector 1
Bucharest
Romania

Stefan Lessmann

School of Business and Economics, Humboldt-University of Berlin ( email )

Unter den Linden 6
Berlin, Berlin 10099
Germany

Wolfgang Karl Härdle

Blockchain Research Center Humboldt-Universität zu Berlin ( email )

Unter den Linden 6
Berlin, D-10099
Germany

Charles University ( email )

Celetná 13
Dept Math Physics
Praha 1, 116 36
Czech Republic

National Yang Ming Chiao Tung University ( email )

No. 1001, Daxue Rd. East Dist.
Hsinchu City 300093
Taiwan

Asian Competitiveness Institute ( email )

Singapore

Academy of Economic Studies, Bucharest ( email )

Bucharest
Romania

Do you have a job opening that you would like to promote on SSRN?

Paper statistics

Downloads
190
Abstract Views
621
Rank
339,381
PlumX Metrics