Stochastic structure of metal prices: Evidence from fractional integration non-linearities and breaks

This paper investigates the long memory properties of the prices series of two major precious metals (gold and silver) and six non-precious metals (aluminium, copper, lead, zinc, tin and nickel) by using a fractional integration modelling framework, while controlling for structural breaks and non-linearities. We use daily data in a range from November 01, 2007 to March 20, 2020. From the results, Copper and Tin exhibit a very small degree of mean reversion. However, if autocorrelation is permitted, the unit root null hypothesis cannot be rejected in any of the series. We also account for structural breaks and non-linearities. We observe that all the series under investigation are exposed to multiple breaks. Results obtained under white noise errors shows some evidence of mean reversion for silver aluminium, copper, tin and zinc in some of the subsamples. However, under the assumption of Bloomfield autocorrelated errors the confidence intervals are so wide that we cannot confirm this evidence in any single case. •This paper investigates the long memory properties of metal prices using a fractional integration modelling framework, while controlling for structural breaks and non-linearities.•We use two major precious metals (gold and silver) and six non-precious metals (aluminium, copper, lead, zinc, tin and nickel).•We find that Copper and Tin exhibit a very small degree of mean reversion.•We also account for structural breaks and non-linearities and observe that all the series under investigation are exposed to multiple breaks.•Results obtained under white noise errors shows some evidence of mean reversion for silver aluminium, copper, tin and zinc in some of the subsamples.•We find under the assumption of Bloomfield autocorrelated errors that the confidence intervals are so wide that we cannot confirm this evidence in any single case.