conditional volatility series generated by this approach can be regarded as a
generalization of the 12-month rolling standard deviation estimator used in Fama (1984).
The generated statistic allows the conditional mean to vary over time (Equation 4),
permits varying weights on the lagged absolute unpredicted changes in returns (Equation
3), and hence is consistent with ARCH models now commonly used in financial markets
research, which accommodate time varying volatility and autocorrelation evident in
financial asset returns. Finally, the testing equation for the relation between conditional
volatilities of precious metal returns and macroeconomic variables can be simply written
as
hM =α + α1(L)hX + u3 (5)
t o t ,t
where M denotes the precious metals considered in the paper and X is the set of key
macroeconomic variables, as mentioned above.
(b) Data
We employ a large set of macroeconomic variables to investigate the underlying causes
of volatility in precious metals markets. Our data include variables that are well known as
usefully accounting for the effects of the business cycle, monetary environment and
financial market sentiment on asset returns. As mentioned in the introduction, linkages
between the macroeconomy and commodity price movements have been documented in
Strongin and Petsch (1996) and Gorton and Rouwenhorst (2006). Additionally, Pesaran
and Timmermann (1995) suggest macro variables could help increase trading results in
equity markets via time strategies and in fact, Vrugt et al. (2004) and Chan and Young
(2006) consider various trading strategies in commodity markets. Our rationale in