Next, we present the summary statistics of the estimated conditional standard deviations
that are used in our primary regression analysis. As reported in Table 2, the highest
conditional standard deviation is for our term structure variable (a mean of 1.001)
followed by dividend yields on the US and World stock indexes (means of 0.915 and
0.770 respectively). There is evidence of excess kurtosis in several of the series,
indicating deviations from the expected normal distribution. It is also important to
determine whether the estimated series contain a unit root, or are nonstationary, since this
will directly affect the regressions. We rely upon conventional Augmented Dickey-Fuller
and Phillips-Perron tests to investigate this issue and report these findings in Table 3. The
results of this investigation suggest that all estimated series can be characterized as
stationary and hence, can be used in standard regression analysis.
(Insert Table 3 about here)
(b) Main Estimation Results
In this section, we report the analysis of the determinants of conditional volatility in the
precious metals markets used in our sample. The findings, which are reported in Table 4,
put forward several points. First, regarding the main research question of the paper, we
find no evidence that the same macroeconomic factors influence the volatility processes
of the commodity price series examined in this paper. This seems to be consistent with
the arguments raised by Erb and Harvey (2006) that individual commodities are too
11