NONLINEAR FORECASTING OF THE GOLD MINER SPREAD: AN APPLICATION OF CORRELATION FILTERS

• Author: Christian L. Dunis,Peter W. Middleton,Andreas Karathanasopoulos,Jason Laws
• Date: 01 October 2013
• DOI: http://doi.org/10.1002/isaf.1345
• Published date: 01 October 2013

NONLINEAR FORECASTING OF THE GOLD MINER SPREAD:

AN APPLICATION OF CORRELATION FILTERS

CHRISTIAN L. DUNIS,

a

JASON LAWS,

b,c

PETER W. MIDDLETON

b,c

*

AND ANDREAS KARATHANASOPOULOS

d

a

Horus PartnersWealth Management Group, Geneva, Switzerland and EmeritusProfessor of Banking and Finance at Liverpool

John Moores University, Hatton Garden, Liverpool, UK

b

University of Liverpool, CIBEF, Liverpool, UK

c

University of Liverpool Management School, University of Liverpool, Chatham Street, Liverpool, UK

d

London Metropolitan University, Holloway Road, London, UK

SUMMARY

This paper models and forecasts the Gold Miner Spread from 23 May 2006 to 30 June 2011. The Gold Miner

Spread acts as a suitable performance indicator for the relationship between physical gold and US gold equity.

The contribution of this investigation is twofold. First, the accuracy of each model is evaluated from a statistical

perspective. Second, various forecasting methodologies are then applied to trade the spread. Trading models

include an ARMA (12,12) model, a cointegration model, a multilayer perceptron neural network (NN), a particle

swarm optimization radial basis function NN and a genetic programming algorithm (GPA).

Results obtained from an out-of-sample trading simulation validate the in-sample back test as the GPA model

produced the highest risk-adjusted returns. Correlation filters are also applied to enhance performance and, as a

consequence, volatility is reduced by 5%, on average, while returns are improved between2.54% and 8.11% across

five of the six models. Copyright © 2013 John Wiley & Sons, Ltd.

Keywords: spread trading; multilayer perceptron neural network; particle swarm optimization; radial basis function

neural network;genetic programming algorithm; correlation filter

1. INTRODUCTION

This investigation evaluates the relationship between gold bullion (physical gold) and US gold mining

equity. Historically, the statistical relationship between the two has displayed strong correlation as the

values of gold mining stocks are predominately determined by the price performance of gold. However,

despite this long-term relationship the spread frequently experiences short-term irregularities. When

this occurs an opportunity arises for investors to profit from trading disparities found in the spread be-

tween US gold equity and physical gold.

Owing to market manipulation in the form of Western governments’monetary policies, better known

as quantitative easing, hedgers and investors are turning to alternative and more lucrative investmentsin

order to offset the effects of inflation and to seek profitable trading strategies respectively. At the same

time they are looking to profit in a volatile and highly unpredictable economic climate. In par ticular,

some investors are investing in gold to protect their purchasing power against rising inflation; however,

with gold rising to record highs in 2011, many are now speculating whether this upward trend is losing

* Correspondence to: Peter W.Middleton, University of Liverpool, CIBEF,Liverpool, UK. E-mail: peter.william.middleton@ gmail.com

Copyright © 2013 John Wiley & Sons, Ltd.

INTELLIGENT SYSTEMS IN ACCOUNTING, FINANCE AND MANAGEMENT

Intell. Sys. Acc. Fin. Mgmt. 20, 207–231 (2013)

Published online 30 September 2013 in Wiley Online Library (wileyonlinelibrary.com)DOI: 10.1002/isaf.1345

momentum. Nevertheless, the ‘Gold Miner Spread’may present investors with another long-term alter-

native investment opportunity as it also offers exposure to gold mining stocks. The market offers two

gold miners’exchange-traded funds (ETFs). The first is the Market Vectors Gold Miners ETF

(GDX), which provides exposure to large-cap

1

gold miners; the second is the Market Vectors Junior

Gold Miners (GDXJ). The latter has exposure to small-cap

2

gold mining companies. Here, the larger

cap exposure is evaluated by trading the spread between GLD (physical gold) and GDX.

In particular, gold miners use this spread to offset inflation as physical gold is extracted and then ex-

changed for ‘paper’. To hedge against inflation, gold miners ‘long’gold and ‘short’US gold equity to

protect purchasing power of the dollar. Among other variables, inflation is just one of the factors affect-

ing Gold Miners’ability to match the price performance of gold as a physical asset, and hedging on the

financial markets becomes vital to offset systematic risk and maintain profitability.

GDX’s underlyingvalue is derived from stocksand American Depository Receiptsof gold mining com-

panies. Following its inception in 2006 the objective of this ETF has been to replicate the price and yield

performance of the NYSEArca Gold Miners Index (GDM). The net asset valuation of GDX is currently

calculated from 31underlying mining stocks, and thusis a general indicator for the performance of large-

cap US Gold Equity stocks. This ETF is weightedbased on market capitalization, withits larger holdings

being in Barrick Gold Corporation (16.51%), Gold Corp Inc. (13.19%), Newmont Mining Corporation

(11.18%), Kinross Gold Corporation (5.76%), AngloGold Ashanti Ltd (5.60%), Agnico-Eagle Mines

Ltd (4.40%), Randgold Resources Ltd (4.39%) and Yamana Gold Inc. (4.35%).

To gain physical gold exposure many investors are drawn to investing in the SPDR Gold Trust GLD

ETF as this is the most liquid commodity ETF offered on the financial markets. GLD is from the State

Street Global Advisor fund family and was first introduced in 2004 as a means to replicate the price of

gold bullion. As a financial instrument it has provided market participants with exposure to price fluc-

tuations in the commodity while removing risks associated with the delivery of the physical commodity

as well as avoiding storage costs.

Pressures to grow and expand production capacity in the mining industry have been challenged by a

shortage of technical expertise and increasing production costs. Many gold miners have undertaken ex-

pensive and wasteful projects which have eroded profitability. As a result, recent years have seen little

value being returned to shareholders, with much of the capital being spent on project-related activities.

Such wasteful spending has created a drag on gold miners’stock performance relative to the physical

commodity that they produce. In addition, the price of gold has also been affected by monetary policies.

Ultimately, fluctuations in the prices of US gold mining stocks as well as gold create short-term dispar-

ities in the spread from which one can profit by trading effective market timing strategies. With high

inflation and the diminishing purchasing power of the US dollar as a world currency, further analysis

of the Gold Miner Spread is warranted.

2. LITERATURE REVIEW

On review of past literature it becomes apparent that few have analysed the advantages of trading the

Gold Miner Spread. Furthermore, the use of nonlinear modelling as a vehicle to forecast next-day

returns is virtually nonexistent within the confines of current academic spread trading literature.

1

Market capitalization greater than $100million with an average tradingvolume of 50,000 sharesduring a 6-month trading period.

2

Includes small to medium-sized mining companies with the majority producing revenues from either gold and/or silver mining.

208 C. L. DUNIS ET AL.

Copyright © 2013 John Wiley & Sons, Ltd. Intell. Sys. Acc. Fin. Mgmt. 20, 207–231 (2013)

DOI: 10.1002/isaf