Few economists would dispute the proposition that the value of the dollar affects the US trade balance. But how much does the value of the dollar affect the trade balances of other countries with independent currencies? An influential recent paper by Gita Gopinath, chief economist of the International Monetary Fund, and coauthors (2020) notes that a disproportionate share of global trade is invoiced in US dollars and claims that dollar invoicing causes trade prices to follow movements in the dollar.
The implication is that when the dollar is overvalued, trade undertaken by other countries may be suppressed because the prices of traded goods are artificially high. Moreover, if the authors' finding is correct, monetary policy outside the United States is unable to influence the price of a country's exports relative to prices in the rest of the world. Accordingly, countries would be losing an important channel of monetary transmission to overall economic activity and inflation.
Our new working paper shows that this outsized effect of the dollar is limited to very small bilateral trade flows and is not apparent in larger flows or in the overall exports of a broad sample of advanced and developing countries. For countries that supply most of global trade, including many small and medium-sized economies, the dollar has at most a modest impact on their export prices, and the export channel of monetary transmission continues to operate. These results support the traditional view that countries can have effective independent monetary policies and that the United States does not uniquely set monetary conditions for the world.
The first column of the following table displays the results of one of the key regressions Gopinath et al. use to support the view of a dominant role of the dollar in international trade pricing. The first row shows the effect of a change in the dollar exchange rate on import prices in importer currency. Import prices rise by 78 percent of any change in the dollar's value. Import prices rise by 14 percent of any increase in the producer price index (PPI) in the exporting country, a small but statistically significant effect. Finally, the coefficient of 0.05 on importer currency captures the extent to which changes in the importer's exchange rate against the exporter are not passed through into import prices. It may also be interpreted as the effect of the importer-exporter exchange rate on export prices in exporter currency.
|Additional regressions to Gopinath et al.'s bilateral import price regressions show dollar role is less dominant|
|Bilateral import prices from Gopinath et al. (2020)||Multilateral export prices|
|Unweighted regression||Trade-weighted regression||Cumulative effects,
|Dollar or euro effect||0.20*|
|* p < 0.05, ** p < 0.01|
|PPI = producer price index|
|Note: This table presents regressions of the logarithmic change in import prices (export prices in columns 5 and 6) on logarithmic changes in the explanatory variables. Estimates, except in columns 1 and 3, are the sums of coefficients on the current value and one lag of the independent variables. Columns 1 and 3 report only contemporaneous coefficients. In columns 5 and 6, importer PPI coefficients are restricted to equal importer currency coefficients. In column 6, US and euro area PPI coefficients are restricted to equal dollar or euro exchange rate coefficients. All column regressions include full sets of country and year effects. Standard errors (in parentheses) are clustered by country. For coefficient sums, standard errors are computed from the estimated coefficient covariance matrix. Within R2s refer to the percent of variation explained after removing country effects.|
|Sources: Authors' calculations using data provided by Gopinath et al. (2020) and Gagnon and Sarsenbayev (2021).|
Column 1 strongly suggests that the dollar has a dominant effect on import prices. But these coefficients cover only the immediate impact of exchange rate changes. Column 2 displays the cumulative effects of exchange rate changes after a year has passed. The regression in column 2 also adds the change in the importer producer price index (PPI), which is an important factor in countries with high and variable inflation. Now we see that import prices rise by only 50 percent of any change in the dollar's value. The effects of exporter PPI and importer currency together rise by about as much as that of the dollar falls. Note that importer PPI is also an important factor, with import prices rising by 25 percent of any rise in importer PPI.
Most of the import prices used in the regressions of columns 1 and 2 are based on small bilateral trade flows. For example, half of the observations are based on flows of less than $300 million per year, which is the median flow. Yet the average bilateral trade flow in the dataset is much higher: $2.3 billion per year. Thus, the vast majority of trade occurs in the largest bilateral relationships. Columns 3 and 4 display results of regressions in which greater weight is given to observations with higher trade flows.
Column 3 shows that the impact effect of the dollar in the Gopinath et al. regressions declines to 0.58 when the observations are weighted by the value of trade flows. Column 4 shows that the cumulative effect of the dollar declines to 0 when importer PPI is included in the regression. Import prices rise by 42 percent of any rise in exporter PPI and 56 percent of any rise in importer PPI. Exporters absorb 59 percent of any change in the importer-exporter exchange rate.
The final two columns display results of regressions on multilateral export prices as described in our paper. In the model of our paper, exporters respond to exchange-rate-adjusted foreign prices, which implies that the coefficients on importer currency and importer PPI should be identical. This restriction is imposed in the regressions of columns 5 and 6. The results in column 5 are close to those of column 4, with no effect of the dollar and a somewhat larger exporter PPI effect and smaller importer currency effect. Column 6 considers whether the euro may be a dominant currency for some exporters. When the dominant currency is allowed to be the dollar in Asia and the Western Hemisphere and the euro in Africa and Europe, it is found to have a statistically significant effect, with export prices rising by 20 percent of any change in either the dollar or the euro. Nevertheless, the exporter PPI and importer currency/PPI effects remain larger than the dollar/euro effect.
Figure 1 displays the export price data used in the table columns 5 and 6 around the period of the sharp dollar appreciation of 2014–15. The dollar appreciated against almost all currencies in these years owing to strong US economic performance and the prospect of a US monetary tightening. This appreciation may be taken as exogenous to our 33 exporting countries. As can be seen in the figure, in only four countries (Israel, Jordan, Sri Lanka, and Thailand) can it be argued that the dollar had any appreciable effect on export prices (in 2016) beyond that predicted by exporter and importer PPIs.
Gopinath, Gita, Emine Boz, Camila Casas, Federico Díez, Pierre-Olivier Gourinchas, and Mikkel Plagborg-Møller. 2020. Dominant Currency Paradigm. American Economic Review 110, no. 3: 677–719.
2. A third difference between the regressions of columns 1 and 2 is that the regression in column 1 includes two lags of every variable in addition to the contemporaneous value (although only the contemporaneous coefficient is reported), whereas that in column 2 includes only one lag. The coefficients on second lags are small and they are not statistically significant in the trade-weighted regression of column 4.
3. Flows for each country pair are averages over the regression sample.
4. The model also implies that US PPI should be included in the regression with a coefficient equal to that on the US exchange rate. Given that the changes in US PPI are identical for all exporters, their effects are fully absorbed by the year effects in the regression and it makes no difference whether they are included or not.
5. This regression includes US and euro area PPIs with the restriction that coefficients are equal to those on the respective exchange rate. Neither this restriction nor the restriction on importer PPI coefficients has a statistically significant effect on the sums of the coefficients, but they do increase the dollar/euro effect and reduce the importer currency effect slightly.
6. Figure 1 in our paper displays the US PPI times the dollar exchange rate because that is what our model of competitive behavior calls for. However, Gopinath et al. focus on the effect of the dollar exchange rate by itself and that is what is plotted here. There is little difference between the two figures because US PPI is relatively stable.
The data underlying this analysis are available here [zip].