Subsample stability, change detection and dynamics of oil and metal markets: A recursive approach
MetadataShow full item record
CitationKhan, A. I., Shahbaz, M. ve Napari, A. (2023). Subsample stability, change detection and dynamics of oil and metal markets: A recursive approach. Resources Policy, 83,1-16. https://dx.doi.org/10.1016/j.resourpol.2023.103601
The analysis of historical price data for patterns and using such patterns for predictions and policy recommendations has become ubiquitous in the existing economics literature. These predictions and recommendations are premised on the stability of the statistical properties and inter-variable dynamics for which a single regime or few number of regimes can capture. This, however, is a strong assumption with serious repercussions if violated. In this study, the appropriateness of the stability assumption is questioned using various recursive regressions to test stability, consistency of stationarity and stability in inter-variable dynamics between crude oil, gold, silver, and platinum prices. Using monthly data sourced from the World Bank Commodity Price Data (Pink Sheet) from January 1, 9960 to March 2022, our empirical analysis found level prices of oil, gold, and platinum to be consistently non-stationary with rare exceptions. The level price of silver however is found to be inconsistent with multiple regime switches while the logged series of all variables yielded non-stationarity. The default is stationarity for all the variables when price series are logged differenced and/or differenced for oil, silver, and platinum. Differenced gold prices resulted in inconsistent stationarity with multiple regime changes. Even if rare, the stationarity of all the variables is dependent on time and sample size due to the inconsistence in the stationarity verdict. On the bi-variate relationship in the long run, only level silver prices are found to be cointegrated with oil while logged silver prices are inconsistently cointegrated with logged oil prices. Also, in the short-run, only log of oil prices is found to Granger cause log of silver prices. It is thus recommended that researchers and policy makers be tempered in extrapolating statistical findings in general and the price and interprice dynamics of oil, gold, silver and platinum into the future.