Aggregation Bias in Tests of the Commodity Currency Hypothesis

According to the commodity currency hypothesis (CCH), a country's commodity-export prices are predicted by its exchange rate. We investigate two types of aggregation biases that might affect CCH tests. First, monthly commodity prices are sometimes averaged across all days of the month, a practice that creates substantial spurious predictability in price changes. Second, in CCH tests commodity prices are often grouped into an index. If all commodity prices do not react equally fast to news, the active goods' prices should lead those of the slower-acting ones and therefore predict the index. If so, the currency's value changes can proxy for these active goods' prices. We find a strong bias from price averaging in monthly returns, while the bias from ignoring predictability among commodities seems weak. When testing the CCH using end-of-period data the supporting evidence is weak at best.