Explorations in modeling and forecast assessment of energy derivatives

Abstract

Amid economic and geopolitical uncertainty, energy commodities have gone through abrupt shifts in supply and demand dynamics during the last decade. This has caused unprecedented price volatility. To manage excessive price fluctuation, a firm may prefer to hedge its production or consumption. When done well, a firm can efficiently use capital to manage its business risks rather than having a war chest to protect against fluctuating prices. However, depending on the composition of market participants and systematic factors, hedging instruments like exchange-traded futures and options can carry risk premia in their prices. Knowledge of expected premia will help a firm in taking informed hedging decisions. A commodity trading firm must use a correctly specified term structure model for futures for its spot trading, derivative pricing, and hedging operations. Inability to capture crucial aspects of price dynamics may lead to hedging errors, and arbitrage while pricing exotic derivatives. This thesis contributes to these broad issues in two ways. First, we conduct an empirical study on short-dated contracts to find the reliability of option-implied distribution as a density forecast of U.S. crude oil and natural gas. Bias in option-implied density forecast indicates the presence of risk premia, which may exist due to unhedgeable risk factors. This study helps in understanding the supply and demand of risk capital in energy complex after the financialization of commodity markets. Second, we develop a novel way of modeling commodity futures term structure by using string shocks as a noise source for future convenience yield process in the continuous semimartingale framework. We obtain the drift of future convenience yield process under no-arbitrage restriction, and derive closed-form formula for the European call option written on futures contract. The model covers some of the important aspects that are missing in earlier formulations which results into easier calibration and better option pricing.