An Analysis of the Variance and Distribution of Commodity Price Changes

A method of jointly estimating the time-dependent variance of daily commodity price changes and their distribution is presented. The data are copper spot prices (1966-74) and sugar futures prices (1961-73), for London contracts. Much of the leptokurtosis observed in the price change distributions is shown to result from the mixinq of non-normal distributions whose variances differ substantially. There are important consequences for conventional autocorrelation tests, which falsely assume a constant variance. The usefulness of the logarithmic transformation of prices is assessed statistically and it is found that the transformation does help to equalise the variance of price changes.