Predicting white metal prices by a commodity sensitive exchange rate

We show that the exchange rate of South African Rand has statistically and economically significant out of sample forecasting power for palladium and platinum prices, and to a lesser extent for silver prices. We argue that this finding is consistent with the implications of the present value models of exchange rate determination. We utilize a recently developed method for testing for instabilities in predictive power, which markedly affects overall conclusions. For the palladium market, the predictive power is also robust to using cross exchange rates. Importantly, no reverse causality is detected from white metals to the exchange rate. We also employ asymmetric impulse response functions and show that positive and negative exchange rate shocks do not have materially different impacts on white metals.