Exchange rates and commodity prices: Measuring causality at multiple horizons

• We study causality between commodity prices and exchange rates.
• High-frequency data (daily and intra-day) are used, including exchange rates of four commodity-exporting economies (Canada, Norway, Australia, and Chile) and the prices of the dominant export commodities (WTI crude oil, Brent crude oil, gold, and copper).
• Multi-horizon causality measures are computed.
• Causality from commodity prices to exchange rates is stronger than vice-versa.
• The macroeconomic/trade-based mechanism evidently plays a central role in exchange-rate dynamics, despite the financial feature of these markets.

Different causal mechanisms have been proposed to link commodity prices and exchange rates, with opposing implications. We examine these causal relationships empirically, using data on three commodities (crude oil, gold, copper) and four countries (Canada, Australia, Norway, Chile), over the period 1986–2015. To go beyond pure significance tests of Granger non-causality and provide a relatively complete picture of the links, measures of the strength of causality for different horizons and directions are estimated and compared. Since low-frequency data may easily fail to capture important features of the relevant causal links, daily and some 5-minute data are exploited. Both unconditional and conditional (given general stock market conditions and short-term interest rates) causality measures are considered, and allowance for “dollar effects” is made by considering non-U.S. dollar exchange rates. We identify clear causal patterns: (1) there is evidence of Granger-causality between commodity prices and exchange rates in both directions across multiple horizons, but the statistical evidence and measured intensity of the effects are much stronger in the direction of commodity prices to exchange rates, especially at horizon one: the ratios of causality measures in two different directions can be quite high; (2) causality is stronger at short horizons, and becomes weaker as the horizon increases; (3) conditioning on equity prices (the S&P500) does not change the patterns of causality measures found in the unconditional cases; (4) the main results are robust to eliminating U.S.-dollar denomination effects and including a short-term interest rate as the conditioning variable. In contrast with earlier results on the non-predictability of exchange rates, we find that the macroeconomic/trade-based mechanism plays a central role in exchange-rate dynamics, despite the financial feature of these markets.