Volatility and risk management in agricultural commodity markets

Bian, Siyu

Permalink

https://hdl.handle.net/2142/116036 Copy

Description

Title

Volatility and risk management in agricultural commodity markets

Author(s)

Bian, Siyu

Issue Date

2022-06-29

Director of Research (if dissertation) or Advisor (if thesis)

Serra, Teresa

Doctoral Committee Chair(s)

Serra, Teresa

Committee Member(s)

• Garcia, Philip
• Irwin, Scott
• Couleau, Anabelle

Department of Study

Agr & Consumer Economics

Discipline

Agricultural & Applied Econ

Degree Granting Institution

University of Illinois at Urbana-Champaign

Degree Name

Ph.D.

Degree Level

Dissertation

Keyword(s)

• Microstructure Noise
• Volatility Forecasting
• Conditional Value-at-Risk
• Risk Management

Abstract

In this thesis we examine agricultural commodity market volatility in modern electronic futures markets from different perspectives. First, we examine the informational shock of news releases in agricultural futures markets by decomposing observed volatility into efficient and noise components. Second, we investigate if machine learning algorithms can capture nonlinear trends in the informational volatility and generate better volatility forecasting results than simple linear models. Third, we assess the liquidity stress in agricultural futures markets induced by both heightened volatility and unexpected price jumps. The first essay demonstrates the importance of disentangling efficient market price variance from microstructure noise in the context of public announcements. Our contribution to the literature lies in the use of a Markov Chain framework that solves the shortcomings of previously used approaches and the analysis of the impacts of recent public information release policy changes affecting agricultural markets. We show that previous literature, by ignoring noise, has underestimated the informational content of the United States Department of Agriculture (USDA) reports, overlooked the costs of price discovery, and produced biased estimates of the time required for the market to absorb the information. Overall, findings are highly relevant to public policy, particularly USDA, and have implications for market design. The second essay provides a systematic comparison between linear Heterogeneous Autoregressive type models and nonlinear machine learning-based models for forecasting realized volatility. To further assess the need for nonlinear models to produce better forecasting results, we propose a novel hybrid artificial neural network model based on residual learning to separate forecasts into linear and nonlinear components. We use these models to forecast the daily realized volatility in highly traded commodity and, for comparison purposes, stock futures markets. We also implement different tests and forecast accuracy measures to compare among the different models. Findings show that the best (worst) out-of-sample forecast performance measures are usually achieved by using linear (nonlinear) models. The decomposition of hybrid neural network models’ forecasts shows that the hybrid models essentially rely on their linear components to produce forecasts. Therefore, we recommend linear models over nonlinear machine learning algorithms when forecasting realized volatility. The third essay evaluates the impact of extreme price volatility on the funding risk faced by agricultural futures traders. Extreme price movements significantly challenge traders’ ability to comply with the exchange’s margin requirements and impose funding risks on all market participants. We propose a new intra-horizon conditional Value-at-Risk method to measure the funding risk faced by agricultural futures traders. Unlike previous methods, our approach incorporates both asymmetric jumps and seasonal volatility components. Findings suggest that funding risks in agricultural futures markets are notably seasonal and heavily influenced by price jumps. Also, we find that margin requirements in agricultural futures markets are inadequate to cover funding risks faced by clearinghouses in the presence of frequent price return jumps.

Graduation Semester

2022-08

Type of Resource

Thesis

Copyright and License Information

Copyright 2022 Siyu Bian

Owning Collections