RDP 2016-05: Trade Invoicing Currency and First-stage Exchange Rate Pass-through 4. An Invoice-share-weighted Exchange Rate Index

Thus far, we have emphasised the low degree of pass-through for Australian dollar-invoiced imports. We now examine pass-through in more detail for foreign currency-invoiced imports. Our motivation for doing so is that, if currency choice is a sufficient statistic for desired pass-through, invoice currency should be more relevant for modelling import prices than the currency of the country of origin.

Much of Australia's goods imports are invoiced in a third currency, rather than Australian dollars or the currency of origin (Table 3). The most prominent third currency is the US dollar. In 2014, the United States was the source of only 11 per cent of Australia's goods imports, for the two-digit SITC divisions for which we have data, but 57 per cent of imports were invoiced in US dollars. In contrast, 14 per cent of imports come from the euro area, but just 8 per cent of all imports in two-digit SITC divisions for which we have data are invoiced in euros. Similarly, 9 per cent of Australia's imports come from Japan, but only 1 per cent of imports are invoiced in Japanese yen.

Table 3: Country of Origin and Invoice Currency Differ
Average for 2014
  Per cent of imports that come from country Per cent of imports invoiced in currency of country
United States 11.5 56.8
Euro area 14.3 8.3
Japan 8.8 1.3
United Kingdom 2.7 1.1
New Zealand 2.1 1.0

Note: As a proportion of imports for the SITC divisions for which we have data

Sources: ABS; Authors' calculations

Figure 9 compares the import trade-weighted exchange rate index to the invoice-share-weighted geometric average of the five bilateral nominal exchange rates for which we have invoice currency data.[9] Because the US dollar is by far the most common invoice currency among foreign currencies – 83 per cent of all foreign currency-invoiced imports as of the most recent data – the invoice-share-weighted index is highly correlated with the US dollar. Although there is a high degree of co-movement among exchange rates for Australia's trading partners, there have been episodes where the trade-weighted and invoice-share-weighted exchange rates have diverged for a period of time. For example, the import trade-weighted exchange rate did not depreciate as much as the invoice-share-weighted exchange rate during the late 1990s and early 2000s.

Figure 9: Aggregate-level Australian Dollar Exchange Rate Indices
March 1997 = 100
Figure 9: Aggregate-level Australian Dollar Exchange Rate Indices

Sources: Authors' calculations; RBA

To the extent that prices are sticky in their currency of invoice, the invoice-share-weighted exchange rate should better predict short-run changes in import prices for foreign currency-invoiced goods than the import trade-weighted exchange rate. We test this hypothesis at the aggregate level by running the following regression:

where:

  • Inline Equation is the aggregate import price index for the 19 differentiated-goods two-digit SITC divisions for which we have data; we construct the index using the weights of these two-digit SITC divisions in the aggregate import price index
  • Inline Equation is the invoice-share-weighted exchange rate index
  • Inline Equation is the nominal import trade-weighted exchange rate index
  • Inline Equation is the pass-through coefficient at horizon j for the invoice-share-weighted exchange rate index, and analogously for Inline Equation for the spread between the two indices.

For sticky-price goods invoiced in foreign currency, short-run changes in import prices should be more closely correlated with movements in the invoice-share-weighted than the import trade-weighted exchange rate index. This implies Inline Equation >0 for the first couple of quarters over which prices are sticky, with the size of the coefficients declining by horizon. If our hypothesis that there is no additional information in the import trade-weighted index above the invoice-share-weighted index is true, the coefficients on the spread term (Inline Equation) should all be zero.

Figure 10 shows the estimated coefficients Inline Equation and Inline Equation, together with two standard error confidence bands. As predicted, the first coefficient for the invoice-share-weighted index is significantly greater than zero. However, contrary to our hypothesis, the contemporaneous coefficient for the import trade-weighted index is positive and statistically significant. Thus the data do not support the proposition that the changes in the invoice-share-weighted exchange rate index alone are sufficient for modelling pass-through. This may reflect the possibility that, despite restricting attention to less homogenous goods, prices for some foreign currency-invoiced trade are reset frequently, which could make the trade-weighted exchange rate index relevant above and beyond the invoice-share-weighted index.

Figure 10: Pass-through Coefficients
Figure 10: Pass-through Coefficients

Note: Red bars show two standard error bands

Footnote

We rescale the weights for foreign currencies so that they sum to 100. [9]