Information from successive changes in gold and silver prices during phases of the business cycles

Abstract

This paper concerns the behavior of gold and silver prices on a daily, weekly and monthly time span during January 1970 to December 1989. The methodology consists of extracting the predictive power of time series of changes in past prices for obtaining optimal forecasts for next-period changes in prices. Optimizations are made in the context of information theory via minimizing the degree of diversity between the actual and predicted changes in prices. This methodology has merit in that it does not rest on, generally speaking, unacceptable assumptions regarding the shape of the distribution, stationarity of variance or its existence. The behavior of gold and silver prices are studied during peak to trough and trough to peak of the business cycles over 1970–1989. It is generally shown that information contained in past prices of gold and silver does not allow one to predict next-period changes in prices in the short run. However, longer-term predictions are possible. This study further reveals that as the length of the time interval expands, gold prices exhibit a higher degree of dependency on past prices than silver.