A new empirical study confirms that export price changes explain a substantial part of commodity currency fluctuations, particularly at high frequencies. More importantly, country-specific export price indices help predicting commodity countries’ future exchange rate dynamics. The predictive power appears to be most robust over a horizon of one month.
Kohlscheen , Emanuel, Fernando Avalos and Andreas Schrimpf (2016), “When the Walk is not Random: Commodity Prices and Exchange Rates”, BIS Working Papers No 551, March 2016.
This post ties in with systematic value generation through information efficiency, particularly through macro trends (view summary here).
The below are excerpts from the paper. Headings, links and cursive text have been added.
The relevance of commodity export prices
“In the short-run, higher commodity prices lead to an increased supply of foreign exchange in the markets of commodity exporters, as a result of increased export revenues – causing an appreciation of the domestic currency. In the medium to long-run, this effect might then be compounded by ensuing foreign direct investment, as a result of more attractive investment prospects in the local commodity sector.”
“Price variation of key export commodities is often seen as a reasonably good proxy for terms of trade movements, as export price variation typically trumps the variation in import prices – which tends to be more dependent on more rigidly priced manufactured goods.”
Calculating export price indices
“We…construct country-specific commodity export price indices at daily frequency for 11 commodity-exporting countries… We were able to associate quoted prices at daily frequency with a total of 83 UN Comtrade 3-digit commodity groups. 26 referred to metals, 36 to agricultural commodities, 11 to livestock and 10 to energy. Price information was collected from Datastream and from Bloomberg… The weight of each commodity in each country basket was chosen so as to match the share of export revenues in total commodity export revenues in the respective country between 2004 and 2013… The sample period covers the time span between 2 January 2004 and 28 February 2015.”
“We show how the information that is contained in these indices clearly improves the predictive performance of exchange rate models for all 11 of the commodity exporters that we study. In addition, the indices provide more prompt information about the direction in which equilibrium exchange rates may be moving.”
The concurrent impact of export prices
“We provide extensive evidence that there is a distinct commodity-related driver of exchange rate movements, even at relatively high frequencies….Variation in commodity prices has an effect on nominal exchange rates at high frequency that goes beyond the impact of global risk appetite…Commodity prices explain a significant part of the variation of the exchange rate that is orthogonal to risk.”
“We run some simple panel regressions to explore the contemporaneous relation between [changes in] exchange rates and commodities prices… The estimated coefficient indicates that a 10% increase in the price of the commodities that are exported by a country in our sample is associated with a 2.1% appreciation of the respective currency – on average… the information of commodity price variation alone, explains more than 23% of the variation in the USD exchange rate in the cases of Australia and Canada. On the other hand, this explanatory power was only about 3% for Peru.”
The predictive power of export prices
“We find that commodity prices predict exchange rate movements of commodity exporters up to two months ahead when the analysis is based on in-sample panel regressions.”
“The monthly horizon stands out as being the one in which forecasting performance is more robust across countries. Commodity prices emerge as significant in-sample predictors for 10 of the 11 countries. With the notable exception of South Africa, in-sample exercises suggest that exchange rates are at least to some extent predictable.”
“[Commodity price-based export price indices] also tend to deliver better predictive accuracy than standard models based on interest rate differentials (carry).”
The “out-of-sample” predictive power of export prices
“We use a rolling window of fixed length to estimate the [elasticity between export price changes and subsequent exchange rate change]… which are then used to produce an out-of-sample forecast. The window is then rolled forward one period at a time to produce the coefficient estimates for the subsequent period…Our procedure is able to capture the long-term variations in the sensitivity of exchange rates to commodity prices that may result for instance from secular changes in the share of commodities in the total exports of a country – or changes in FX intervention policies. The use of out-of-sample forecasts for performance evaluation also diminishes the risks associated with data mining.”
“Out-of-sample estimations also show that simple linear predictive models based on our commodity price indices tend to have superior predictive performance for exchange rates when compared to random walk benchmarks.”
“The results show that the information on commodity price variation clearly leads to a one-step [day] ahead prediction performance that beats both benchmark random walk models.”
“We checked whether this link is also evident at lower frequencies… for most cases the relation is also found to be important at lower frequencies. [However] in most cases, lengthening the window in which price variations are measured has the effect of weakening the relation somewhat in this out-of-sample exercise.”