Modelling prices, returns and volatility in futures markets for industrial metals

 

Keywords: Industrial metals, Cointegration, Commodities, Empirical finance, GARCH, Applied financial econometrics, Volatility, Cost-of-carry, Risk premium, Market index

Summary: While futures markets for industrially-used metals have existed for a long time, their study increased in the academic finance literature from the 1980s as pricing in the metals industries moved from a producer-list basis to being exchange-referenced. By the early 2000s there was a substantial empirical literature on metals markets. Initially, I set out to evaluate what is known about the main industrially-used non-ferrous metals futures markets by examining the empirical significance of econometric modelling in the area. In the survey “Econometric modelling of non-ferrous metal prices”, which was co-authored with Michael McAleer and published in the Journal of Economic Surveys, we evaluate the empirical modelling in 45 prominent published articles. This was the first systematic evaluation of the empirical literature on metals futures markets and is my most cited article. Published empirical research was evaluated in light of the type of contract examined, frequency of data used, choice of both dependent and explanatory variables, use of proxy variables, type of model chosen, economic hypotheses tested, methods of estimation and calculation of standard errors for inference, reported descriptive statistics, use of diagnostic tests of auxiliary assumptions, use of nested and non-nested tests, use of information criteria and empirical implications for non-ferrous metals. The literature contains a number of conflicting results, for instance, on various aspects of market efficiency. We find that important empirical issues such as overlapping data, structural change, measurement error, correct use of proxy variables, non-stationarity and diagnostic testing are frequently ignored suggesting the empirical conclusions of much of the literature should be interpreted with caution.

An unanswered question in the literature was how futures contracts for metals are priced relative to the price in their spot markets. The two dominant theories for the pricing of commodity futures are the cost-of-carry and risk premium hypotheses. In “The Pricing of Non-ferrous Metals Futures on the London Metal Exchange” which was co-authored with Michael McAleer and published in Applied Financial Economics, we nested these hypotheses within a general model and tested between them in a cointegration framework. The modelling incorporated analysis of structural breaks and showed that structural change is important in determining the which futures pricing model describes the data best. The cost-of-carry model is generally superior, which implies that metals inventories are important in futures pricing. However, there are times when stocks are not important, and during these periods risk premia are important. This usually occurs when a metal’s price is on a long trend downward.

Metals market participants have often been quoted in the financial media as claiming that the futures prices for industrially-used metals traded on the London Metal Exchange have become more volatile over time. They have argued that higher volatility has been a consequence of the move from producer-list pricing to exchange-based pricing and the financialisation of commodity markets (the increased participation of portfolio investors). There is some support in the literature for the claim that producer-pricing regimes produced less volatile prices, however this has been disputed by other research that suggests prices have not become more volatile and the results of work suggesting volatility increased are biased. In “How has volatility in metals markets changed?”, co-authored with Michael McAleer and published in Mathematics and Computers in Simulation, we use rolling GARCH models to analyse the time-varying properties of metals futures volatility estimates and forecasts over a long sample period. We show that markets have not become markedly more volatile over time, although during some periods volatility has been harder to model with standard specifications. The volatility process has become more complex and time-varying, while volatility itself has not tended to increase.

Extreme or outlying observations are relatively common in financial returns and returns for commodity futures are no exception. Metals derivatives are sensitive to macroeconomic news and industrial production shocks in particular, global financial shocks, market specific supply shocks, and events such as the manipulation of the silver market by the Hunt Brothers and the copper market by Sumitomo Corporation. Extreme observations are deleterious to volatility models and their forecasts during uncertain times, when investors most need accurate forecasts. For instance, GARCH estimates become biased, moment conditions are violated and the model tends to over-predict the short-term persistence but under-predict the long-term persistence of volatility induced by extreme returns. In “Forecasting commodity market volatility in the presence of extreme observations”, joint with Michael McAleer (Proceedings of the International Congress on Modelling and Simulation, Volume 3: Socioeconomic Systems), we examine a straightforward and practically implementable method of accommodating extreme returns using a simple trimming regime applied to daily returns more than 2.5 standard deviations from a rolling mean. We find this method effective in improving the real-time volatility forecasts that practitioners use.

In “Related Commodity Markets and Conditional Correlations”, co-authored with Michael McAleer and published in Mathematics and Computers in Simulation, we examine the relationships amongst the volatility processes of industrially-used metals futures prices. We also examine how the volatility properties of London Metal Exchange Base Metals Index (LMEX), a price index for metals markets, reflect or differ from those of its constituent metals. We find that the short-run volatility effects between individual metals are more closely related than are the long-run volatility effects. This is largely because short-run volatility persistence is influenced by general financial market factors that have similar influences on the futures price of each metal. Long-run volatility persistence is metal specific as it is influenced by the fundamentals of each metal, such as consumption, production and inventory, as well as futures market liquidity. The volatility process of the LMEX is more closely related to the short-run rather than the long-run volatility components of its constituents.