A novel risk management framework for natural gas markets

• Author: Nikos C. Papapostolou,Panos K. Pouliasis,Alexander A. Kryukov,Ilias D. Visvikis
• DOI: http://doi.org/10.1002/fut.22067
• Published date: 01 March 2020
• Date: 01 March 2020

J Futures Markets. 2020;40:430–459.wileyonlinelibrary.com/journal/fut430

|

© 2019 Wiley Periodicals, Inc.

Received: 30 December 2018

|

Accepted: 29 September 2019

DOI: 10.1002/fut.22067

RESEARCH ARTICLE

A novel risk management framework for natural gas

markets

Panos K. Pouliasis

1

|

Ilias D. Visvikis

2

|

Nikos C. Papapostolou

1

|

Alexander A. Kryukov

1

1

Cass Business School, Faculty of Finance,

City, University of London, London, UK

2

School of Business Administration,

Department of Finance, American

University of Sharjah, Sharjah, United

Arab Emirates

Correspondence

Panos K. Pouliasis, Cass Business School,

City, University of London, 106 Bunhill

Row, London EC1Y 8TZ, UK.

Email: p_pouliasis@city.ac.uk

Abstract

This paper examines dynamic hedges in the natural gas futures markets for

different horizons and explores the gains from devising risk management

strategies. Despite the substantial progress made in developing hedging models,

forecast combinations have not been explored. We fill this gap by proposing a

framework for combining hedge‐ratio predictions. Composite hedge ratios lead

to significant reduction in portfolio risk, whether spot prices are partially

predictable or not. We offer insights on hedging effectiveness across seasons,

backwardation‐contango conditions and the asymmetric profiles of long‐short

hedgers. We conclude that forecast combinations better reconcile realized

performance with the hedging process, mitigating model instability.

KEYWORDS

dynamic futures hedging, forecast combination, natural gas

JEL CLASSIFICATION

C32; C53; G11; G32; L95

1

|

INTRODUCTION

Adverse price trends and sharp fluctuations not only affect profit margins, but also impact a company’s probability of

default or even alter the incentives of investing (e.g., infrastructure and transportation)—reducing investment in favor

of lower risk projects. Such business challenges that are directly linked to, inter alia, production/purchasing costs,

earnings, and credit availability, create the need for coherent risk management practices. For oil and gas projects, where

cash flows are almost entirely generated by oil and gas sales, price volatility increases the incentive to mitigate such

effects. Effective natural gas hedging strategies are relevant in reducing price volatility for investors, traders, producers,

and commercial users in the sector. Moreover, hedging policies constitute a key theme for policy‐makers and regulators

to consider alternative reforms and mitigate deficiencies (e.g., transaction costs, poor liquidity, and transparency) in the

current market design. To add, with the Paris Agreement in 2015 and its predecessor the Kyoto Protocol in 1997, there

is increasing interest in energy investments with low emissions, such as natural gas. Therefore, given the broad

economic and financial impact of natural gas volatility, it is vital to study natural gas risk management strategies.

One crucial parameter of futures‐based hedging is the hedge ratio, that is the number of futures contracts to buy or

sell for each unit of the underlying asset on which the hedger bears risk. Earlier studies (e.g., Ederington, 1979) derive

hedge ratios that minimize the variance of the spot/future portfolio based on the principles of portfolio theory. The

Optimum Hedge Ratio (OHR) is typically found by regressing the returns to holding the physical asset on the returns to

holding the hedging instrument. However, the regression approach has several shortcomings. For example, it omits

cointegration between futures and spot prices which might lead to biased OHR forecasts, particularly in the long run

(Lien, 1996). Moreover, Bollerslev (1990) and Kroner and Sultan (1993), among others argue that this approach

implicitly assumes constant risk throughout time as new market information arrives.

Therefore, a number of hedging models have been developed. Engle and Kroner (1995) and Kroner and Sultan

(1993) apply multivariate Generalized Autoregressive Conditional Heteroscedasticity (GARCH) models and derive

time‐varying hedge ratios from the conditional second moments. GARCH models are popular due to their ability to

capture some of the salient features of financial time series, such as volatility clustering, nonlinear dependence, and

thick tails (see Pouliasis, Papapostolou, Kyriakou, & Visvikis, 2018). A popular alternative is the Markov Regime

Switching (MRS) modeling framework, introduced by Hamilton (1989); see also Alizadeh, Nomikos, and Pouliasis

(2008) for an extension to regime switching in a cointegrated GARCH process for hedging energy commodities.

Switching models overcome the limitation of constant parameters, offering better model fit and, thus, are able to

improve on the hedging ability.

The consensus from the literature is that while dynamic OHRs tend to outperform static hedges, the alleged gains

are market specific, though occasionally, the benefits are minimal (Lien & Tse, 2002). Ghoddusi and Emamzadehfard

(2017) examine the hedging effectiveness of the Henry Hub natural gas future contract for different physical positions

and find that cointegration and time‐varying volatilities only marginally effect hedging ability. Overall, studies on the

hedging efficiency of natural gas futures are scant, while results indicate that gas futures are the least effective contacts

compared to other commodities, thus offering a rich experimental context for our tests (see Cotter & Hanly, 2012;

Hanly, 2017).

1

In this paper, we argue that a more effective hedge may be available in the form of a “composite hedge”,by

exploiting the information content of various models. Model combinations have been used extensively. Yet, most of the

works focus on price (Baumeister & Kilian, 2015) or volatility forecasting (Patton & Sheppard, 2009). We fill this gap by

considering a variety of models which feature prominently in the hedging literature, and their combination in the

decision‐making process to minimize operating cash flow variability. To this end, this paper investigates the hedging

effectiveness of New York Mercantile Exchange (NYMEX) Henry Hub and Intercontinental Exchange (ICE) National

Balancing Point (NBP—virtual trading hub) contracts; the most liquid and mature futures markets in the sector.

Our contributions extend in several directions. We first build on a diverse set of models, each targeting a specific

feature of natural gas price movements (seasonality, cointegration, time‐varying volatility/correlation and regime

shifts). From the estimated pool of models—given their theoretical pros and cons and mixed results of their empirical

performance (e.g., see Lien & Tse, 2002)—this paper considers a model combination approach in hedging decisions.

This way, decision‐makers do not rely on some particular econometric specification when estimating the OHR which

might not fully reflect the risk inherent in price movements.

2

As a number of studies on the predictive power of

individual hedging models report mixed results,

3

implementing the proposed framework acts also as an insurance tool

mitigating the undesirable effects of structural breaks and model misspecification, thus, leading to improved forecasting

(see, inter alios, Baumeister & Kilian, 2015; Patton & Sheppard, 2009; Pesaran & Timmermann, 2007). Therefore, we

provide a flexible framework in the hedging process under model uncertainty. Based on the work of Caldeira, Moura,

Nogales, and Santos (2017), we put forward an approach to combine OHR forecasts from candidate models. The

combination weights are directly linked to the decision‐making problem of an investor who wishes to minimize

portfolio risk; each model is given importance proportional to its actual performance.

In addition, we evaluate the hedging ability of different OHR prediction models in terms of variance reduction. The

analysis is executed both in‐and out‐of‐sample and the results are validated on a statistical basis using Hansen’s (2005)

reality check and Politis and Romano (1994) bootstrap simulation methods. This way we provide robust evidence on the

potential gains of the proposed forecast combination hedging strategies taking into account transaction costs and utility

performance fees. We also address the issue of downside risk by examining whether the effects of mean‐variance OHRs

1

Natural gas markets exhibit properties that distinguish them from other markets. Such are, for example, the region‐specific nature of natural gas with restricted access to export markets, as well as the

difficulty in storing and transporting gas which creates high basis risks (see Hanly, 2017). Moreover, natural gas price trajectories and the performance of hedges differ not only from other traditional

assets, such as stocks and bonds, but also from most commodities, a result of the inherent relatively high volatility; see also Pouliasis & Papapostolou (2018) on the role of natural gas in asset allocation.

2

For individual models it is reasonable to discard some modeling features of market dynamics to warrant a parsimonious structure. However, depending on the source of market shocks, ignoring

relevant information in the formulation process might be costly.

3

Gagnon and Lypny (1995) provide evidence in support of GARCH models. In contrast, Lien and Tse (2002) support the traditional regression approach. GARCH models exhibit few limitations. For

example, the observed nonnormalities in return distributions are more pronounced than those implied by GARCH; the model fails to reproduce time variability in higher moments unless explicitly

modeled, and a strong degree of persistence is imputed to volatility which may be due to structural breaks (Lamoureux and Lastrapes, 1990).

POULIASIS ET AL.

|

431