IV. PROSPECTS FOR U.S. EXPORT GROWTH
The framework associating foreign growth and U.S. export growth can be applied to explain the historical patterns of U.S. export growth as well as future prospects for U.S. export growth. Both U.S. export growth and growth in different foreign regions have varied considerably in the past. Therefore, a decomposition of aggregate U.S. export growth into growth attributable to each foreign region helps better explain why U.S. export growth has changed over time.
Similarly, various foreign regions are expected to grow at different rates in the future. And recent projections of a global growth slowdown pertain differently to different regions. For example, more substantial growth slowdowns are forecast in Europe and Asia than in Canada and Mexico. The framework relating U.S. export growth and each foreign region’s economic growth can help predict changes in future U.S. export growth.
Decomposition of past U.S. export growth
The performance of U.S. exports was markedly different during the recent expansion, recession, and recovery periods. U.S. exports grew at an average annual rate of 8.4 percent during the expansion prior to the recent recession (from the first quarter of 2004 through the third quarter of 2007). Exports then declined at an average annual rate of 3.3 percent during the recession (from the fourth quarter of 2007 through the second quarter of 2009) and subsequently increased at an average annual rate of 9.0 percent during the recovery (from the third quarter of 2009 through the fourth quarter of 2011; Table 3, bottom row). The static model in Section II slightly overestimated average annual U.S. export growth during the 2004-2007 expansion at 9.1 percent but accurately predicted average export growth in both the recession and recovery.
A decomposition of U.S. export growth based on the static analysis shows variation in the importance of different regions to U.S. export growth over the past several years. During the 2004-2007 expansion, the framework shows 3.2 percentage points — or 35 percent — of the predicted 9.1 percent average growth in U.S. exports were attributable to growth in Europe (Table 3, column 1).[29] Another 2.2 percentage points of average U.S. export growth were attributable to Asia, while the share of U.S. export growth attributable to other regions was smaller but still positive.
However, the framework implies that the changes in U.S. export growth attributable to Europe and Asia differed both in magnitude and in direction during the 2007-2009 recession. In particular, 2.6 percentage points of the 3.3 percent average decline in U.S. exports were attributable to economic contraction in Europe (Table 3, column 2).
An additional 1.6 percentage points of the reduction in average export growth were attributable to negative growth in Canada and Mexico. On the other hand, Asia maintained positive average growth during the period, partially offsetting the negative U.S. export growth attributable to other regions by 1.1 percentage points.[30]
The slow economic recovery in Europe was associated with a smaller increase in U.S. export growth during the post-2009 recovery than during the 2004-2007 expansion that preceded the recession. Indeed, 2.4 percentage points of average U.S. export growth were attributable to European growth during the recovery, which was less than the comparable 3.2 percentage points during the expansion. It also was slightly less than the 2.5 percentage points attributable to growth in Asia during the recovery. Average growth in U.S. exports attributable to growth in Canada, Mexico, and the rest of the world, taken together, totaled 3.6 percentage points (Table 3, column 3).
In addition to growth in different regions, depreciation of the U.S. dollar against foreign currencies increased the pace of U.S. export growth. During the expansion and recovery, the U.S. dollar depreciated against foreign currencies, increasing foreign demand for U.S. goods and services and thus adding to U.S. export growth (Table 3, row 6). In contrast, during the recession, the U.S. dollar appreciated, most likely due to an increase in demand for safe-haven assets.[31] The appreciation of the dollar was associated with a decline in demand for U.S. exports and thus a reduction in U.S. export growth.
The global growth slowdown and future U.S. export growth
The analysis presented in this article has distinct implications regarding the prospects for both future U.S. real export growth and, in turn, future U.S. real GDP growth. In the past two years, the IMF has consistently revised downward its estimates for growth prospects across different regions. These downward revisions are illustrated in a comparison of IMF growth forecasts in successive years. From September 2011 to October 2012, the IMF forecast for average, annual, global growth over the 2013-2016 period was lowered by 0.5 percentage point (Table 4)[32] Of particular note are Asia and Europe, for which average growth was revised down by 1.0 percentage point and 0.4 percentage point, respectively. In contrast, the expected growth in Canada and Mexico essentially has been unchanged. Based on the framework of analysis presented in this article, three conclusions may be drawn.
First, the overall effect on U.S. exports of the projected growth slowdown across different regions is most noticeable in the near term. Specifically, U.S. export growth is projected to be 2.0 percentage points lower in 2013 and 0.9 percentage point lower in 2014 than previously estimated. The expected reduction in U.S. export growth diminishes to 0.4 percentage point and 0.2 percentage point in 2015 and 2016, respectively (Chart 6).[33]
Second, the reduction in U.S. export growth from 2013 to 2016 is primarily attributable to the growth slowdowns in Europe and Asia (Chart 6). In particular, a 1.1-percentage-point reduction in U.S. export growth is attributable to weaker growth in Europe in 2013. The reduction attributable to Asia’s growth slowdown in 2013 is smaller at 0.5 percentage point. However, because Asia’s growth slowdown is expected to persist, the associated reduction in U.S. export growth continues through 2016, while growth reductions attributable to Europe dissipate beyond 2013. The large change in future U.S. export growth attributable to European growth stems from a combination of two facts: U.S. export growth is most sensitive to growth in Europe (Table 1, column 2), and growth in Europe is expected to be significantly slower over the next few years (Table 4).
Third, downward revisions in U.S. export growth attributable to Canada, Mexico, and the rest of the world are small in 2014 through 2016. This is primarily because of the relatively small revisions in these regions’ forecasted growth over the medium term.
To summarize, the overall reduction in future U.S. export growth associated with expected slowdowns in foreign countries’ economic growth is significant and varies widely across regions. Most of the growth reduction in U.S. exports is attributable to slower growth in Europe and Asia, while the parts of the reduction attributable to Canada and Mexico are much smaller. Because U.S. exports account for about 14 percent of total U.S. GDP, the expected slowdown in foreign growth suggests lower U.S. real GDP growth both in the near term and in the medium term. In particular, weaker export growth is expected to reduce the contribution of exports to U.S. real GDP growth over the next four years. The contribution of exports to U.S. real GDP growth is projected to be reduced by 0.4 percentage point in the 2013-2014 period and by 0.1 percentage point in the 2015-2016 period (Chart 7).
V. CONCLUSION
U.S. export growth and economic growth across regions of the world are closely related. Numerous studies have documented that when aggregate foreign economic growth declines, U.S. export growth tends to decline as well. But the relationship between U.S. export growth and different regions’ economic growth has been a somewhat neglected area of research.
This article fills the gap by examining how U.S. export growth is related to real GDP growth in different regions and what leads to the differences in these relationships. Two approaches, one emphasizing the static relationships (a regression analysis) and another incorporating dynamic interactions (a vector autoregression), provide analysis of twenty-seven years of data on U.S. export growth, real GDP growth in different regions, and exchange rates.
The analysis shows U.S. export growth is most closely associated with growth changes in Europe, followed by growth changes in Canada and Asia. The close relationship between U.S. export growth and European growth results from Europe having the largest share of world GDP and the largest share of U.S. export goods.
The analysis can be applied not only to assess U.S. export growth in the past but also to forecast future U.S. export growth. Based on projections of slower economic growth in foreign regions, particularly in Europe and Asia, a 0.4-percentage-point reduction in the contribution of annual U.S. real export growth to U.S. real GDP growth is projected in the 2013-2014 period.
Original study from the Kansas City Federal Reserve
About the Author
Jun Nie is an economist at the Federal Reserve Bank of Kansas City. Lisa Taylor is a research associate at the bank.
Footnotes
1The shares of export goods and GDP growth rates in this section are calculated
using annual data, while in later sections, the data used in the analysis are
quarterly. The use of annual data in place of quarterly data does not significantly
change the statistics in this section.
2Data on export services by destination are limited before 1992.
3Developed countries are those classified as “advanced economies” by the
IMF, while developing countries are those categorized as “emerging and developing
economies” by the IMF.
4See Constessi and Li for a similar comparison of U.S. export shares by destination
in 2000 and 2011. Earlier studies include Schmidt.
5On average across the 1985-2011 period, Asia, Europe, Canada, and Mexico
together accounted for 87 percent of all U.S. export goods. Since 1992, the
shares of export services to Canada, Mexico, and Asia have declined while the
shares to Europe and the rest of the world have increased.
6The simple correlation between the share of U.S. export goods to developing
countries and the real GDP growth rate for developing countries from 1985
to 2011 was 0.58.
7A handful of studies focus on the income elasticity of exports for different
countries or for particular export sectors. Houthakker and Magee estimate the
income elasticity of U.S. exports separately for different destinations as well as for
various commodity classes, while Shane, Roe, and Somwaru focus on identifying
the effects of changes in exchange rates and foreign income on U.S. exports of
agricultural goods and their subcomponents.
8Appendix A provides a detailed description of the data sources and the definition
of each variable used in the regressions. Appendix B provides more details
on the regression models considered in this section. In a check of the robustness
of the analysis, factors such as lags of U.S. real GDP growth and oil prices were
added to the regression model. Including these variables did not change the main
results, however, and most of the regression coefficients on these variables were
not statistically significant.
9In theory, U.S. export growth also can be related to factors other than relative
prices and incomes, such as trade policies and technological advancements.
However, these factors are hard to measure, and because these factors change
infrequently, including them in the model is unlikely to significantly alter the
results. For more studies on U.S. export growth, relative prices, and foreign incomes,
see Crane, Crowley, and Quayyum; Chinn; Goldstein and Khan (1978);
Marquez; and Bussière, Chudik, and Sestieri. In this literature, both relative prices
(measured as ratios of price indices in different countries) and real exchange
rates are used (see Cardarelli and Rebucci; and Crane, Crowley, and Quayyum).
10These estimates are consistent with the income and exchange rate elasticities
found in the existing literature. For a summary of previous estimates, see Crane,
Crowley, and Quayyum; Goldstein and Khan (1985); and Marquez.
11Aggregate foreign real GDP growth and real GDP growth in Europe, Asia,
and the rest of the world are defined as a weighted average of the quarter-overquarter
annualized real GDP growth of countries in the region. The weights are
based on GDP. For more details, see Appendix A. An analysis using trade weights
in place of GDP weights generates qualitatively similar results, but the significance
of some estimates changes.
12The sum of the five regions’ coefficients need not match the foreign GDP
growth coefficient in column 1 (Table 1) because foreign growth has been disaggregated,
changing the regression model.
13The F-test cannot reject the joint hypothesis that the coefficients are the
same for Canada, Mexico, Europe, and Asia. However, as will be shown later, after
controlling for the economic size of each region, significant differences have been
found in the relationships between regions’ GDP growth and U.S. export growth
(Table 1, column 3).
14The overall sensitivity needs not equal the coefficient in column 2 because
including the time-varying share of world GDP changes the regression model.
15From 1985 to 2011, Mexico’s average share of world GDP was 0.015, Asia’s
share was 0.207, and the rest of the world’s share was 0.046. Mexico’s overall sensitivity
is 0.2 percentage point, Asia’s is 0.5 percentage point, and the rest of the
world’s is 0.04 percentage point.
16Again, data on U.S. service exports to different regions generally are not
available at quarterly frequency since 1985 for all countries in the sample. The
shares of U.S. export goods to different regions are used as an approximation to
the shares of total U.S. exports to different regions. Crane, Crowley, and Quayyum
study the relationship between foreign growth and U.S. exports of goods and of
services separately.
17From 1985 to 2011, Canada’s average share of U.S. export goods was 0.218,
Europe’s share was 0.230, Asia’s share was 0.214, and the rest of the world’s share
was 0.054. Canada’s overall sensitivity is 0.5 percentage point, Europe’s is 1.0
percentage points, Asia’s is 0.4 percentage point, and the rest of the world’s is 0.1
percentage point.
18In fact, the export growth elasticities of Mexico and Canada after controlling
for their shares of world GDP are much larger than the corresponding export
growth elasticity for Europe. However, Mexico’s and Canada’s large elasticities are
offset by their small shares of world GDP, weakening their overall importance to
U.S. export growth.
19A joint hypothesis test based on the F-statistic rejects the hypothesis
that export growth elasticity with respect to a region’s economic growth after
controlling for the share of world GDP is the same across Canada, Mexico,
Europe, and Asia at the 10-percent significance level.
20The distance from the United States for each region is defined as the average
distance, in miles, from Washington, D.C., to the capital of each country included
in the region. See Appendix A for a list of countries included in each region.
21Transportation costs tend to increase with distance. For recent studies
on transportation costs and U.S. export growth, see Alessandria and Choi; and
Hummels. Note that the export growth elasticities controlling for the share of
U.S. export goods are similar for Canada, Mexico, Europe, and Asia. Formally,
the hypothesis that the elasticities for Canada, Mexico, Europe, and Asia are the
same cannot be rejected by the F-test (Table 1, column 4). This is expected because
trade shares are correlated with distance from the United States and thus (at
least partially) controlled for in the model.
22As will be made clear, the short term means the first four quarters following
a growth change, while the long term refers to the fifth through eighth quarters
following a growth change.
23In practice, when using a VAR with four lags as in this analysis, the effect
of an initial growth increase will dissipate after about 10 periods (or quarters).
24The VAR approach has not been used widely in this literature. Ahearne and
others apply this approach to study exports of Asian emerging economies. The
estimation of the simple VAR and details of the implementation of the simulation
are described in Appendix C.
25Mexican GDP growth has a negative correlation with U.S. export growth
in the first four quarters.
26This negative correlation is most likely related to the fact that Mexican
GDP growth changed at a slightly different pace than other regions’ GDP growth
in the sample period.
27Consistent with this result, Hooper, Johnson, and Marquez find that the
income elasticity of exports is smaller in the long run than in the short run.
28In this simulation, the U.S. export growth rate is compared with its historical
average. See Appendix C for further details.
29This share is based on the predicted average annual export growth reported
in Table 3, not the actual average annual export growth. That is, 35 percent is the
result of dividing growth attributable to European growth, 3.2 percentage points,
by the predicted average annual export growth rate in the expansion, 9.1 percent.
30Compared with Europe during the same period, Asia experienced a less
severe and shorter period of negative growth that was offset by positive growth.
31During the downturn, international investors increased their demand for
U.S. dollars, either because they preferred the high liquidity of the dollar in general
or because they needed dollars to purchase other U.S. assets that were considered
less risky than foreign assets.
32The IMF publishes annual real GDP forecasts in its World Economic Outlook.
The growth forecasts for Canada and Mexico were taken directly from the
IMF’s database, while the regional forecasts for Europe, Asia, and the rest of the
world were based on the countries included in the sample as listed in Appendix A.
The regional forecasts were constructed by weighting each country’s growth rate
by its share of the region’s GDP, based on purchasing-power-parity valuation. The
World Bank also forecasts global and regional growth. These forecasts are similar
to those of the IMF, although more pessimistic in the near term. IMF forecasts
were used for the purposes of this study to make possible forecasting growth for
the regions as defined in the analysis over the forecast period through 2016 (further
than the World Bank’s forecast horizon).
33These results are similar to the revisions made by the IMF to their forecasts
of real U.S. export growth in 2013 through 2016. On average, the revisions implied
by this analysis are slightly larger than those of the IMF (-0.9 percentage
point versus -0.7 percentage point, respectively). The IMF had smaller downward
revisions in 2013, 2014, and 2016 and the same downward revision in 2015. In
addition, it is also interesting to study why a growth slowdown is expected to take
place in the next few years and whether it is related to a growth slowdown in the
United States. Such questions are not in the scope of this analysis.
34In practice, the simulation length, N, is set to be eight periods (quarters).