Chapter

II. Can Asia Decouple? Investigating Spillovers from the United States to Asia

Author(s):
International Monetary Fund. Asia and Pacific Dept
Published Date:
April 2008
Share
  • ShareShare
Show Summary Details

With the U.S. economy slowing and possibly facing a recession, the question arises whether Asia will be able to decouple. Views on this question range from those who think the impact on Asia will be minimal, cushioned by continued strong growth in China and India, to those who think Asia has made itself very vulnerable to U.S. shocks through a combination of growing dependence on external demand for its products and rising domestic financial imbalances, notably asset bubbles.17 The impact of a U.S. slowdown on Asia matters not just for the region but for the world, given that in recent years Asia has been growing faster than any other region and has contributed close to half of total global growth. To shed light on this issue, this chapter uses a variety of statistical methods—correlation analysis, regressions and vector autoregressions (VARs), simulations of dynamic stochastic general equilibrium models, and recession-event studies—to estimate spillovers from the United States to Asia, how these spillovers have evolved over time, and how they vary across countries in Asia.

In a globalized world, one should not simply look at direct spillovers from the United States to Asia. This chapter controls for how U.S. shocks may affect Asia through their indirect effects on Europe and other parts of the world. This is particularly important because Europe has supplanted the United States as the main trading partner for many countries in the region. Moreover, given the ongoing turmoil in many credit and money markets worldwide, but particularly in the United States, the chapter looks at how financial stress may amplify real sector spillovers.

Spillovers from the United States have had a moderate effect on Asia on average over the past 15 years, but the evidence suggests that the impact could be substantial now. Over this relatively long period, the analysis suggests that a 1 percentage point slowdown in the United States has led to a ¼ percentage point slowdown in Japan, and to a ¼–½ percentage point average slowdown in emerging Asia, with substantial variation across countries in the region. Yet, there are reasons to believe that the current U.S. slowdown could have a significantly larger impact than suggested by these estimates:

  • Long-sample estimates find virtually no spillovers from the United States to China and India, two countries with large weights in the regional aggregate. However, long-sample regressions can be problematic for such countries, which are experiencing rapid structural change. Indeed, the chapter finds evidence that spillovers have grown in recent years for these and other countries, consistent with rising trade and financial integration with the United States.

  • Model simulations of U.S. demand shocks that also assume realistic declines in global confidence—a likely possibility going forward—result in growth falling by 0.8 percentage point in Asia for a 1 percentage point decline in the United States.

  • The 2001 recession in the United States had a large impact on Asia, with the roughly 1¾ percentage point decline in the output gap in the former resulting in a 1¼ percent decline in the latter.18 While many commentators have pointed to the fact that the 2001 recession was concentrated on electronics, a key export for the region, and hence may provide an upper-bound estimate of spillovers, it is worth noting that the current slowdown in the United States is expected by the IMF to be deeper and more prolonged than the 2001 recession, and is accompanied by substantial stress in money and credit markets.

That said, while Asia has clearly not decoupled, it may continue to enjoy solid growth. The results suggest that the current U.S. shock is likely to have a significant impact on regional growth, but the region carries considerable momentum, and it would likely take a sharper-than-currently-envisaged slowdown in the United States to derail such momentum. Nevertheless, in some countries in Asia where spillovers are estimated to be high but where growth is currently trailing the regional average, the impact of the U.S. slowdown is likely to be felt more noticeably.

Trade and Financial Exposure to the United States

Over the past two decades, Asian export growth has been driven by growth in intraregional trade. Intraregional exports now account for 41 percent of total emerging Asia exports, versus 23 percent in 1986. Growth in intraregional trade has owed much to growing trade with China, which accounts for almost 60 percent of intraregional trade growth over the past 20 years. Moreover, this contribution has been gaining momentum in recent years.

However, trade exposure to industrial countries has increased substantially over time. Most intraregional trade in Asia is occurring within vertically integrated regional supply chains that, by and large, ship intermediate goods (depending on the sophistication of the source country) that are then assembled in China into final goods for shipment to industrial countries.19 Using highly disaggregated SITC data comprising more than a thousand categories of goods, we computed measures of indirect exposure to the United States and the European Union that account for shipments of intermediate and capital exports used as inputs to goods assembled in all third countries and then reexported to the United States and the European Union for final consumption.20 While direct trade exposure to the United States, measured as exports to the United States as a share of GDP, has increased only modestly for Asia as a whole and has declined in four Asian countries over the past 15 years, total exposure including indirect exposure has increased for all countries but one, and has increased by larger margins than direct exposure for the region (Table 2.1).21 Similar patterns can be observed for a group of 15 countries belonging to the European Union (EU-15), which has surpassed the United States as the main trading partner for many countries in the region.

Table 2.1Export Exposure to Industrial Countries(In percent of GDP)
Exposure to the U.S.Exposure to the EU-15
DirectTotal1DirectTotal1
19942006199420061994200619942006
Japan2.53.43.04.41.42.22.03.5
Australia0.91.11.62.11.51.92.23.1
New Zealand2.83.03.74.03.73.64.85.1
China5.69.67.612.23.97.76.011.7
India1.72.42.03.12.63.13.34.5
Hong Kong SAR16.714.820.021.812.615.716.924.7
Korea4.95.16.18.72.75.03.98.2
Singapore23.917.331.930.817.020.125.535.7
Taiwan POC10.49.912.915.55.27.17.913.6
Indonesia3.33.54.55.63.43.74.86.4
Malaysia18.022.725.031.711.413.818.325.4
Philippines8.88.09.812.03.77.15.012.5
Thailand7.010.58.915.15.28.77.514.7
Vietnam1.415.22.818.56.615.08.620.8
Asia17.79.010.013.35.88.28.313.6
Industrial Asia22.12.52.83.52.22.63.03.9
Emerging Asia29.210.812.015.96.89.79.816.2

Includes indirect exposure through exports of intermediate and capital goods via third countries. See Appendix for calculations of indirect trade exposure.

Arithmetic nonweighted average.

Sources: UN COMTRADE Database; and IMF staff calculations.

Includes indirect exposure through exports of intermediate and capital goods via third countries. See Appendix for calculations of indirect trade exposure.

Arithmetic nonweighted average.

Similarly, financial integration with the rest of the world, particularly the United States, has increased dramatically. Both foreign assets held by Asian residents as well as domestic liabilities held by foreign residents have risen as a share of GDP, with the former outpacing the latter as Asia has built a substantial net foreign asset position. Perhaps more relevant for the purposes of this chapter, U.S. holdings of Asian portfolio securities (both debt and equities) and Asian country holdings of U.S. portfolio securities (same) have also increased dramatically (Table 2.2).22 Much of the increase on the asset side has taken place in debt securities, reflecting a large accumulation of international reserves placed in U.S. treasuries.23

Table 2.2.Financial Exposure to the United States(In percent of GDP)
U.S. Holdings of Asian

Portfolio Securities
Asian Holdings of U.S.

Portfolio Securities
Dec-94Dec-06Dec-94Jun-06
Japan2.513.04.425.0
Australia6.820.42.615.0
New Zealand10.59.32.912.7
China0.32.22.328.8
India5.52.5
Hong Kong SAR12.642.214.861.3
Korea1.412.41.214.2
Singapore8.635.842.9129.2
Taiwan POC0.219.413.139.8
Indonesia1.23.71.03.4
Malaysia11.59.26.810.5
Philippines3.17.93.37.9
Thailand3.15.74.48.2
Vietnam0.14.1
Asia15.113.38.325.9
Industrial Asia16.614.23.317.6
Emerging Asia14.613.110.028.2

Arithmetic nonweighted average.

Sources: U.S. Department of the Treasury, Treasury International Capital System; CEIC Data Company Ltd.; Haver Analytics; and IMF, Information Notice System, and staff calculations.

Arithmetic nonweighted average.

It appears that Asia’s exposure to the United States has risen sharply over the last 15 years. Has this translated into a larger synchronization between the United States and Asian business cycles, or has the emergence of autonomous domestic demand allowed the region to gradually decouple from the U.S. cycle? This is addressed in the next section.

Asia’s Growth and Financial Cycles: Are They Synchronized with the United States?

Growth in Asia now appears to be substantially more correlated with the U.S. growth cycle than in the early 1990s (Table 2.3). The average correlation of growth rates with the United States has increased from 0.1 percent in the pre-Asian-crisis 1990s, to 0.4 percent since 2000. However, it is worth noting that China’s GDP growth has been largely uncorrelated with U.S. growth even since 2000.24

Table 2.3.Growth Correlation with the United States
1990-962000-07
Japan−0.060.41
Australia0.740.38
New Zealand0.280.23
China0.08
India0.14
Hong Kong SAR0.160.61
Korea−0.320.30
Singapore0.310.62
Taiwan POC0.240.61
Indonesia0.060.05
Malaysia−0.260.52
Philippines0.280.47
Thailand−0.200.47
Vietnam0.20
Asia10.110.36
High trade exposure1, 20.090.48
Low trade exposure1, 20.140.24
Western Hemisphere countries1, 30.34
Canada and Mexico10.410.49

Arithmetic nonweighted average.

Countries are ranked according to our measure of total trade exposure to the United States as of 1994.

Includes Argentina, Brazil, Canada, Chile, and Mexico.

Sources: CEIC Data Company Ltd.; and IMF staff calculations.

Arithmetic nonweighted average.

Countries are ranked according to our measure of total trade exposure to the United States as of 1994.

Includes Argentina, Brazil, Canada, Chile, and Mexico.

Moreover, a country’s trade exposure to the United States appears to affect the degree to which its growth is correlated to that of the United States. Those economies with higher trade exposures—Singapore, Malaysia, or Taiwan Province of China, for example—have tended in recent years to have higher correlations. Indeed, the rank correlation coefficient between trade exposure and growth correlation over 2000–07 is quite high, close to 0.5.

Growth synchronization between Asia and the United States is high by international standards. Over the past seven years, the seven Asian countries in the sample with the highest trade exposure to the United States have been as synchronized with the United States as Canada and Mexico, two countries with deep economic ties with their neighbor (Table 2.3). Similarly, Asia as a whole has, on average, been as synchronized with the U.S. growth cycle as a group comprising Argentina, Brazil, Canada, Chile, and Mexico.

Growth correlations have fallen in the past two years, but it is too early to draw much comfort from this. Correlations over the past 18 months have been 0.19 on average, compared with 0.36 for 2000–07. However, 18 months is too short a period to draw robust inference. Indeed, correlations over the 18 months prior to the 2001 U.S. recession were also relatively low (0.23), and, as will become clear in subsequent sections, Asia did not decouple in 2001.

Financial correlations with the United States also seem to have increased in recent years. Stock markets in Asia now seem to be moving more in tandem with U.S. markets than was the case in the 1990s. On average across Asia, the correlation between monthly returns in each country’s main stock index and monthly returns in the S&P 500 has increased from 0.29 between 1990 to 1996 to 0.45 between 2000 and 2007 (Table 2.4). Not surprisingly, the regional financial centers of Hong Kong SAR and Singapore exhibit some of the highest correlations. That stronger correlations are a result of greater financial integration with the rest of the world is suggested by the fact that countries with deeper stock links with the United States also exhibit stronger return correlations. For instance, the seven countries in the sample with the highest holdings of U.S. equity securities exhibited an average correlation of 0.56 over 2000–07, versus 0.34 for the bottom seven.

Table 2.4.Correlations in Stock Market Returns(Main stock market index of a country with U.S. S&P 500)
1990-962000-07
Japan0.260.52
Australia0.520.71
New Zealand0.49
China0.08
India−0.010.45
Hong Kong SAR0.350.69
Korea0.120.59
Singapore0.490.61
Indonesia0.260.43
Malaysia0.370.30
Philippines0.340.45
Thailand0.380.43
Vietnam10.10
Asia20.290.45
High financial exposure2, 30.350.57
Low financial exposure2, 30.230.34

Data begin in September 2000.

Arithmetic nonweighted average.

Countries are ranked according to their holdings of U.S. equity securities as a share of their respective gross domestic products. Rank correlations in terms of U.S. holdings of Asian equity securities are not as high.

Sources: Bloomberg LP.; and IMF staff calculations.

Data begin in September 2000.

Arithmetic nonweighted average.

Countries are ranked according to their holdings of U.S. equity securities as a share of their respective gross domestic products. Rank correlations in terms of U.S. holdings of Asian equity securities are not as high.

The descriptive statistics examined so far, while informative, do not measure the size of the spillovers to Asia from U.S. growth, and do not formally control for other factors that affect Asian economic performance, notably growth in countries other than the United States. We now turn to more comprehensive econometric estimates.

Estimating U.S. Spillovers

Regression Analysis

Growth in the United States matters for Asia, and over the past 15 years it appears to have mattered substantially more than growth in Europe or intraregional growth (Table 2.5). When looking at panel regressions for Asian countries, U.S. growth appears to significantly affect growth in the region, and the magnitude of the coefficient implies that a 1 percentage point slowdown in the United States would result in a 0.6 percentage point slowdown in Asia.25 While the impacts from the EU-15 and Asian intraregional shocks are also positive, they are statistically insignificant and economically small. The importance of the United States for Asia stands out in contrast with the observed patterns in other regions. For instance, in the EU-15, the magnitude of the coefficient on U.S. growth is less than half of that for Asia, while EU-15 countries are strongly affected by European intraregional shocks.

Table 2.5.Growth Spillovers Among Regions
Dependent variable: quarterly growth of countries

(1991-1996 & 2001-2007)
All countriesBy regions
Asia2EU-15Western Hemisphere3
Explanatory variables1
Growth in U.S.0.40***0.61***0.29***0.39
Growth in EU-1540.31**0.060.81***0.36
Growth in Asia40.24**0.15−0.010.24
Number of observations1,922597714357

Other regressors include country fixed effects, growth of the terms of trade, and controls for the Argentine crisis of 2001 -02, the Mexican crisis of 1995, and German reunification of 1991.

Asia includes Australia, China, Hong Kong SAR, Indonesia, Japan, Korea, Malaysia, New Zealand, the Philippines, Singapore, Thailand, and Taiwan POC.

Western Hemisphere includes Argentina, Brazil, Canada, Chile, Ecuador, Mexico, and Peru. Rest of the world, not mentioned above, includes Switzerland, Israel, Norway, South Africa, and Turkey.

Does not include the country of the dependent variable.

Source: IMF staff estimates.Note: ***, **, and * indicate significance at the 1, 5, and 10 percent levels, respectively.

Other regressors include country fixed effects, growth of the terms of trade, and controls for the Argentine crisis of 2001 -02, the Mexican crisis of 1995, and German reunification of 1991.

Asia includes Australia, China, Hong Kong SAR, Indonesia, Japan, Korea, Malaysia, New Zealand, the Philippines, Singapore, Thailand, and Taiwan POC.

Western Hemisphere includes Argentina, Brazil, Canada, Chile, Ecuador, Mexico, and Peru. Rest of the world, not mentioned above, includes Switzerland, Israel, Norway, South Africa, and Turkey.

Does not include the country of the dependent variable.

Moreover, spillovers from the United States appear to have grown in recent years, in particular for China and India. Looking at country-specific regressions for the period 1991–2007, the coefficient on U.S. growth for China and India is small—or negative—and insignificant. However, these countries have experienced dramatic structural change over the period, and some of the changes suggest that spillovers from U.S. growth have increased: trade and financial exposure is now much higher; and there is evidence that the demand elasticity of exports has gone up in recent years, as has value-added in the export sector.26 Indeed, when the regressions are reestimated over 2001–07, estimated spillovers from the United States to China and India are significantly larger, although limited degrees of freedom prevent a tight estimation of the coefficients (Table 2.6).27 For the region as a whole, intra-Asia growth also appears to be relatively more important now than in the past, particularly if one excludes the years 2001–02 (the bust of the information technology, or IT, bubble in the United States, which had a big impact on Asia).

Table 2.6Recent Growth Spillovers Among Regions
Dependent variable: quarterly growth of countries
Asia2JapanChinaIndia
Explanatory variables1
1991-2007
Growth in U.S.0.61***0.330.07−0.18
Growth in EU-1530.060.52−0.87−0.72
Growth in Asia30.15−0.060.73**0.77
Number of observations597676442
2001-07
Growth in U.S.0.74**0.330.801.52
Growth in EU-1540.110.201.241.96
Growth in Asia40.370.21−0.26−1.30
Number of observations324272727
2003-074
Growth in U.S.0.58*
Growth in EU-1530.19
Growth in Asia30.75*
Number of observations228

Other regressors include country fixed effects, growth of the terms of trade, and controls for the Argentine crisis of 2001-02, the Mexican crisis of 1995, and German reunification of 1991.

Asia includes Australia, China, Hong Kong SAR, Indonesia, Japan, Korea, Malaysia, New Zealand, the Philippines, Singapore, Thailand, and Taiwan POC.

Does not include the country of dependent variable.

Results for Japan, China, and India are not available, as we cannot secure the satisfactory degree of freedom for the regressions.

Source: IMF staff estimates.Note: ***, **, and * indicate significance at the 1, 5, and 10 percent levels, respectively.

Other regressors include country fixed effects, growth of the terms of trade, and controls for the Argentine crisis of 2001-02, the Mexican crisis of 1995, and German reunification of 1991.

Asia includes Australia, China, Hong Kong SAR, Indonesia, Japan, Korea, Malaysia, New Zealand, the Philippines, Singapore, Thailand, and Taiwan POC.

Does not include the country of dependent variable.

Results for Japan, China, and India are not available, as we cannot secure the satisfactory degree of freedom for the regressions.

Spillovers from U.S. growth are also positively related to a country’s trade exposure to the United States. Regressing the country-specific coefficients on U.S. growth on a country’s global and U.S. trade exposure, its financial exposure, and other control variables, shows that the more exposed to trade, in particular with the United States, a country is, the more affected its growth cycle is by U.S. growth (Table 2.7).28 Financial exposure also has explanatory power, although the coefficients are not as consistently significant as those for trade exposure, and adjusted R-squares are smaller.

Table 2.7Globalization and Spillovers(Cross-country regression)
Dependent variable: estimated coefficients from country spillover regressions1
Explanatory variables
Trade openness
Trade openness to the world0.0027**[0.141]
Direct exposure to the U.S.0.0350***[0.166]
Total exposure to the U.S.0.0297***[0.176]
Financial openness
Financial exposure to the world0.0003[0.010]
Financial exposure to the U.S.0.0080**[0.113]
Index of openness in capital account transactions0.0557[-0.005]
Fixed exchange rate−0.0083[-0.026]
Fixed exchange rate x Index of openness in capital account transactions0.3129*[0.058]

Numbers in brackets are adjusted R-squares. Number of observations is 40, which includes the countries in Table 2.5 plus India and Vietnam. Since the dependent variable is an estimated coefficient from a first-stage regression, statistical significance should only be seen as indicative.

Source: IMF staff estimates.Note: ***, **, and * indicate significance at the 1, 5, and 10 percent levels, respectively.

Numbers in brackets are adjusted R-squares. Number of observations is 40, which includes the countries in Table 2.5 plus India and Vietnam. Since the dependent variable is an estimated coefficient from a first-stage regression, statistical significance should only be seen as indicative.

The next section looks at results from VARs, to examine the dynamic response of growth in Asia to U.S. shocks.

Vector Autoregression Analysis

The estimated spillovers from VARs are broadly consistent with those from the previous section. While the impact of U.S. growth is short lived and typically peaks after 2 to 4 quarters, it is sizable and significant for most of the Asian countries, ranging from 0 to 0.9 percentage point (Table 2.8).29 For instance, for the highly exposed NIEs, the average annual impact of a 1 percentage point decline in U.S. growth is 0.5 in the baseline VAR. In general, the estimated impact is larger for countries with a higher trade exposure and stronger financial linkages with the United States, as well as those with higher growth-on-growth correlations. For instance, the average spillover effect for the countries with the highest correlation in 2000–07—Singapore, Taiwan Province of China, Hong Kong SAR, and Malaysia—is almost ½ percentage point in the baseline VAR. Spillovers from the United States to China and India are found to be negligible, however, most likely due to the long sample used—1995–2007—to estimate the VARs.30 For both countries, the estimated impulse response is small across all orderings and statistically insignificant. Because the VAR is estimated over a period in which both economies have gone through important changes, the impulse responses are likely to underestimate the current sensitivity to U.S. shocks.

Table 2.8Impact of 1 Percentage Point Decline in U.S. Growth(In percentage points)
Baseline VAR

(1991–2007)
Augmented VAR1

(1991–2007)
Japan0.10.2
Australia0.10.5
New Zealand0.00.3
China20.00.0
India20.00.0
Hong Kong SAR0.40.8
Korea0.20.1
Singapore0.60.9
Taiwan POC0.60.9
Indonesia0.00.4
Malaysia0.20.7
Philippines0.00.4
Thailand0.00.5

Includes financial conditions index.

Sample period is 1995–2007.

Source: IMF staff estimates.

Includes financial conditions index.

Sample period is 1995–2007.

Spillovers are stronger when financial conditions in the United States are accounted for, suggesting that financial linkages have become an important channel for the transmission of shocks. The estimated impact of U.S. shocks on Asian countries generally increases when we augment the baseline VAR with an index proxying for financial conditions in the United States with the exceptions of Korea, China, and India (Table 2.8).31 As in the baseline case, the results vary considerably across countries, but the estimated impact of U.S. shocks typically increases by a factor of two or more once U.S. financial conditions are taken into account. Again, countries with a higher financial exposure on average experience larger spillover effects.

Evidence from the VARs suggests, again, that spillovers from the United States have grown stronger over time. For those economies with a long sample available (Japan, Singapore, Hong Kong SAR, and Taiwan Province of China), the estimated impact of U.S. shocks is larger for the subsample 1996–2007 than for the earlier subsample, 1980–95 (Figures 2.1 and Figures 2.2). This is consistent with the finding that the direct and indirect trade exposures to the United States as well as financial linkages have increased, amplifying the spillover effects of U.S. growth on Asia.

Figures 2.1Impact of 1 Percentage Point Decline in U.S. Growth: Singapore

(VAR impulse response function, in percentage points)

Source: IMF staff estimates.

Figures 2.2Impact of 1 Percentage Point Decline in U.S. Growth: Taiwan Province of China

(VAR impulse response function, in percentage points)

Source: IMF staff estimates.

We now turn to simulations using the IMF Global Economy Model (GEM), to try to gauge spillovers in specific scenarios that more realistically capture real and financial conditions present currently. Model simulations will also help estimate the potential contribution of countercyclical policies in mitigating a U.S. slowdown.

Model Simulations

The two simulated scenarios attempt to replicate a shock to aggregate demand in the United States, and a similar U.S. demand shock accompanied by global decline in confidence, respectively. In the first scenario, we simulate a protracted slowdown (lasting 4 quarters) of 1 percentage point in the United States brought about by a decline in private investment and private consumption. In the second scenario, the slowdown in the United States, while similar in magnitude to that in the first scenario, affects consumer and business confidence (and hence spending) globally in addition to its direct impact on trade. In the current context, this could be interpreted as proxying for continued stress in financial markets. It is assumed that the additional decline in consumption and investment spending is proportional to each region’s trade exposure to the United States. For Japan and emerging Asia, these confidence shocks are equivalent in size to 25–35 percent of the shocks in the United States. Finally, both scenarios assume a 20-basis-point increase in risk premiums, smaller than we have observed so far in the current environment.32

According to the model, aggregate demand shocks in the United States induce nontrivial slowdowns in Asia, even when global confidence does not decline (Figure 2.3). The mechanisms at play are mainly related to the trade channel—with weaker economic activity in the United States reducing demand for imported goods from Japan, emerging Asia, and the rest of the world. At the same time, other mechanisms also influence these results, including relative price changes. The lower import demand in the United States improves the U.S. current account deficit by roughly ½ percent of GDP, supported by a depreciation of the U.S. dollar in both nominal and in real effective terms. In Japan and emerging Asia, currencies appreciate in nominal and real effective terms, with changes in the relative prices of their exported goods playing a key role in the magnitude of the real appreciation. These relative price changes reinforce the negative impact of the U.S. slowdown on current account surpluses in Japan and emerging Asia, which fall by ¼ percent and ½ percent of GDP, respectively. Growth falls by 0.2 percentage point in Japan and 0.4 percentage point in emerging Asia—an order of magnitude comparable with the findings elsewhere in this chapter. With lower demand and more appreciated currencies, CPI inflation declines in all regions. Monetary policies at home and in the United States are loosened to bring inflation back on target, gradually offsetting the effects of the U.S. slowdown.

Figure 2.3GEM Simulation: Spillover from U.S. Slowdown

(Without confidence effect)
(Without confidence effect)
(Without confidence effect)
(Without confidence effect)

Source: IMF staff estimates.

Not surprisingly, spillovers are significantly larger when confidence declines globally. In the second scenario, consumer and business confidence—and therefore spending—in Japan and emerging Asia are affected, and to a greater extent than in other regions because of Asia’s larger trade exposure to the United States. Overall, the spillover effects are roughly twice as large as in the first scenario—growth declines by 0.7 percentage point in Japan and 0.8 percentage point in emerging Asia (Figure 2.4)—suggesting that demand shocks in the United States that are accompanied by financial disruptions affecting confidence could have significant effects on the region.

Figure 2.4GEM Simulation: Spillover from U.S. Slowdown

(With confidence effect)
(With confidence effect)
(With confidence effect)
(With confidence effect)

Source: IMF staff estimates.

Model simulations are also a useful tool to estimate the potential contribution of countercyclical policies in mitigating the effects of a U.S. slowdown. The GEM model incorporates monetary and fiscal policies explicitly, and the effects of these policies can be measured by comparing the effect of a given U.S. slowdown assuming policy reactions in the region, and assuming no policy reactions. In particular, when we simulate the first scenario shock33 assuming no policy reaction for four quarters, the output contraction in emerging Asia is twice as large, implying that the cumulative impact of no policies is approximately equivalent to a ½ percentage point additional slowdown (Figure 2.5). These numbers are simply indicative, in particular since countries in the region do not necessarily conduct monetary policy exactly as assumed in the model.

Figure 2.5GEM Simulation: Contribution of Countercyclical Policies

(Response of emerging Asia GDP growth to U.S. slowdown)

Source: IMF staff estimates.

Because spillovers from U.S. growth may be highly nonlinear, it is important to focus on periods of particular stress in the United States to complement the analysis. The next section looks at spillovers during recent U.S. recessions.

Event Study: Impact of U.S. Recessions on Asia

Lack of data prevents one from drawing much inference on the impact of U.S. recessions on Asia in the 1980s and 1990s. On average, Asian economies appear to have suffered relatively little compared to the United States during the 1980, 1981–82, and 1990–91 U.S. recessions (Table 2.9).34 However, quarterly GDP data for Asian economies are relatively scant for these three episodes, and the sample estimates are thus based on a small number of observations: six for the 1980s, and eight for the 1990s. Also, while the average Asian economy suffered a relatively mild decline in the output gap, this masks wide variations across countries. Finally, the structure of Asian economies has changed substantially since the 1980s and early 1990s, and hence it is questionable how much information can be drawn from these three recessions.

Table 2.9Impact of U.S. Recessions1(In percent)
19801981-821990-91
United States−2.45−2.72−1.83
Asia2−0.76−0.40−0.48
Strongest impact−6.34−2.78−2.35
Mildest impact1.971.751.52

Measured as the average change in the output gap during the recession relative to the four quarters preceding the recession. Potential output is estimated using the Hodrick-Prescott filter.

Includes arithmetic nonweighted average of Australia, Hong Kong SAR, Japan, Korea, Singapore, and Taiwan POC for recessions in the 1980s. New Zealand and the Philippines are added for the recession in 1990-91.

Sources: CEIC Data Company Ltd.; and IMF staff calculations.

Measured as the average change in the output gap during the recession relative to the four quarters preceding the recession. Potential output is estimated using the Hodrick-Prescott filter.

Includes arithmetic nonweighted average of Australia, Hong Kong SAR, Japan, Korea, Singapore, and Taiwan POC for recessions in the 1980s. New Zealand and the Philippines are added for the recession in 1990-91.

The 2001 recession had a substantial impact on Asian economies, in particular for those with larger trade exposure to the United States. Drawing on the full sample of 14 countries, output gaps declined by an average of 2 percent during the recession, very close to the estimated decline in the United States (Table 2.10). Moreover, (1) while there is a large cross-country variation, almost all countries appear to have suffered negative changes in output gaps during the 2001 recession, in contrast to previous U.S. recessions; and (2) there is a high (more than 0.7) rank correlation between the measure of total trade exposure to the United States and the change in the output gap during the recession. The reasons why the 2001 recession had such a large impact on Asia have been discussed at length, and include the facts that the shock was concentrated on electronics, which is a key export for Asia; that Europe and Japan were not providing support for the global economy before and during the recession; and that domestic demand in Asia was still recovering from the 1997–98 financial crisis.35 While these facts may suggest that the 2001 recession provides an upper-bound estimate of spillovers, it is worth noting that the current U.S. slowdown is expected to be deeper and more protracted and, unlike the 2001 recession, is being accompanied by significant stress in money and credit markets around the world.

Table 2.10Impact of 2001 U.S. Recession1(In percent)
Hodrick-Prescott

Filter
Baxter-King

Filter
United States−1.90−1.89
Japan−1.41−1.49
Australia−0.70−0.92
New Zealand−0.04−0.18
China−0.42−1.53
India0.16−0.73
Hong Kong SAR−2.84−3.39
Korea−0.94−1.01
Singapore−7.80−7.72
Taiwan POC−5.54−5.59
Indonesia0.480.61
Malaysia−3.41−3.51
Philippines−1.17−1.63
Thailand−0.98−1.15
Vietnam2−0.60
Asia3−1.80−2.17
High trade exposure3, 4−3.19−3.83
Low trade exposure3, 4−0.41−0.75

Measured as the average change in the output gap during the recession relative to preceding four quarters.

Owing to short span of data, the deviation from trend could not be calculated using Baxter-King filter.

Arithmetic nonweighted average.

Countries are ranked according to our measure of total trade exposure to the U.S. as of 1994.

Sources: CEIC Data Company Ltd.; and IMF staff calculations.

Measured as the average change in the output gap during the recession relative to preceding four quarters.

Owing to short span of data, the deviation from trend could not be calculated using Baxter-King filter.

Arithmetic nonweighted average.

Countries are ranked according to our measure of total trade exposure to the U.S. as of 1994.

It appears that Asian countries availed themselves of countercyclical policies during the 2001 recession.36 Looking at the relationship between changes in fiscal policy, changes in nominal and real monetary policy rates, and changes in the output gap in the region during the 2001 recession, there is a clear (and statistically significant) relationship between policy variables (notably fiscal policy) and outcomes, suggesting that those countries most affected by the U.S. recession made use of countercyclical tools when needed (Figures 2.6 and 2.7).37 Improved macroeconomic frameworks in Asia have created room for countercyclical policies in the event that the global outlook deteriorates sharply or spillover effects are larger than expected (see Chapter I).

Figures 2.6Change in Output Gap and Change in Fiscal Policy in Asia

(During 2001 U.S. recession)

Source: IMF staff estimates.

1 Annual change in fiscal balance as a percentage of GDP between 2000 and 2001.

Figures 2.7Change in Output Gap and Change in Monetary Policy in Asia

(During 2001 U.S. recession)

Source: IMF staff estimates.

1 Difference between the highest and lowest rates over the period July 2000 - December 2002.

Conclusions

The impact of the current U.S. slowdown on Asia could be significant. While spillovers from the United States to Asia have, on average, been modest over the past 15 years, the evidence suggests that they have increased over time. Moreover, the simulations and recession-event study indicate that spillovers can be substantially larger under specific circumstances (Table 2.11).38 These latter results should be given an important weight in light of the potential severity of the current slowdown as well as ongoing financial stress. This being said, Asia has considerable growth momentum, suggesting that concerns about growth are largest in the most trade-exposed countries in the region39 and in those where growth is currently least robust.

Table 2.11Summary of Results: Impact of a U.S. Slowdown on Asia1(In percentage points)
VAR with Financial VariablesCross Country RegressionsGEM BasecaseGEM with Confidence Effects2001 Recession
Japan0.20.30.20.70.7
Australia0.50.70.4
New Zealand0.30.90.0
China0.00.10.2
India0.0−0.2−0.1
Hong Kong SAR0.81.01.5
Korea0.10.10.5
Singapore0.91.14.1
Tawan POC0.91.22.9
Indonesia0.40.2−0.3
Malaysia0.70.51.8
Philippines0.40.60.6
Thailand0.51.00.5
Asia20.20.30.30.80.6
Emerging Asia20.20.20.40.80.5
Emerging Asia2 (excl. China and India)0.50.51.1

Scaled to 1 percentage point.

Weigthed average using nominal GDP at market exchange rates.

Source: IMF staff estimates.

Scaled to 1 percentage point.

Weigthed average using nominal GDP at market exchange rates.

Appendix

Measuring Indirect Trade Exposure to the United States

To measure possible spillovers from the United States through third countries, the chapter calculates the following two indices using trade flow data from the UN Comtrade database (at 5-digit SITC levels, equivalent to more than a thousand categories of goods):

where X(i,j) denotes country i’s exports to country j, and MNonfinal(j) denotes country j’s nonfinal good imports. The measures take into account all possible js (i.e., all possible indirect routes to the United States). The first measure, the share of the sum of country i’s exports to third countries, weighted by these countries’ direct exposure to the United States, is a broad proxy for indirect exposure. The second measure tries to take more directly into account growing triangular trade, in which intermediate/capital goods are flowing into third countries, with final products then shipped to the United States. As the results from the two measures are quite similar, the chapter reports results using the first measure (which is more intuitive) in the main text.

Correlations

Correlations between growth rates in Asian countries and in the United States are based on the three-year rolling correlation of the four-quarter moving average of quarter-on-quarter GDP growth. Rolling correlations of year-on-year growth do not materially change the results.

Regressions

The panel regressions in this chapter cover 38 countries, emerging and industrial, complemented by panel regressions for countries in each region (Asian countries, 15 countries belonging to the European Union, and countries in the Western Hemisphere) for the period 1991–2007. The fixed-effects panel regressions have growth in the country as the dependent variable, and growth in the United States, growth in the 15 countries of the European Economic Community (EU-15), growth in Asia (REO definition), terms of trade changes, and various control dummies as explanatory variables. Regional growth aggregates were computed using 2000 market exchange rate–based GDPs as weights. Needless to say, regressions exclude the country in the left-hand side from the regional grouping in the right-hand side when estimating the regressions.

The chapter also estimates country-specific versions of the panels, and then regresses the resulting country-specific U.S. growth spillovers on the country’s trade exposure, financial exposure, and other control variables. Trade exposure is measured as the sum of exports and imports as a share of GDP, as well as direct and total trade exposure to the United States as previously defined in this chapter. Financial exposure is measured as total assets and liabilities as a share of GDP using the Lane and Milesi-Ferretti (2007) data set, as well as financial exposure to the United States as previously defined. The other control variables include an index of capital account openness by Chinn and Ito (2007), among others.

Vector Autoregressions

The VAR model estimated in this chapter (ignoring exogenous variables) assumes that the global linkages can be represented by

where yt=[gtUSgtESgtROWvgtJAPANgti] is the n X 1 data vector containing the quarter-on-quarter GDP growth for the United States, the European Union, Japan, rest of the world (ROW), and Asian country i; k is a vector of constants, Bi is an n x n matrix of coefficients (i = 1, …, p), and ut is the vector of “structural” shocks. ROW in the baseline specification consists of a simple average of growth in Australia, Canada, and Switzerland. The results are unaffected if a larger set of countries (Australia, Canada, Denmark, Norway, New Zealand, Sweden, and Switzerland) is considered. Given their diversity, shocks to this aggregate are likely candidates for a global shock. But at the same time, given their relatively small size, they are a reasonable proxy for the rest of the world as shocks to this aggregate are unlikely to have significant contemporaneous effects on the other major regions included in the VAR.

The role of financial conditions in propagating spillovers is assessed by augmenting the baseline VAR with a financial conditions index (FCI). The FCI is calculated as the average of the S&P500 return and the Chicago Board Options Exchange Volatility Index (VIX), such that a 1 percentage point increase in volatility (VIX) or a decline in stock returns of the same magnitude represent an equal deterioration of financial conditions in the United States.40

The generalized impulse response functions of Pesaran and Shin (1998) are estimated along with the average of impulse response functions (IRFs) from different recursive orderings as in Bayoumi and Swiston (2007).41 The ordering does not influence the statistical significance of the spillover effects much and the statistical significance is broadly in line with the effects obtained with generalized impulse response functions (which do not depend on the ordering the variables).42

The reduced-form model is estimated by OLS and results based on the generalized impulse response functions are reported in the main text. The model is estimated in first differences (quarter-on-quarter growth rates) and the lag structure is determined according to the Bayesian information criteria. The 16–84 percent error bands (roughly one standard deviation for a normal distribution) for the impulse response functions are calculated by Monte Carlo simulations with 1,000 replications. In the case of the recursive orderings, the standard errors of the impulse response functions are averaged across different orderings.

Global Economy Model (GEM)

In the version of the GEM used here, the world economy consists of five regions: the United States, the euro area, Japan, emerging Asia, and the rest of the world. All regions are assumed to have a flexible exchange rate regime with the monetary authorities targeting inflation (the monetary rule is forward looking, with the policy rate depending on its lagged value, the neutral rate, and the expected inflation gap). With regard to fiscal policy, there are lump-sum taxes, capital income taxes, and endogenous labor income taxes. The government adjusts lump-sum taxes in a smooth manner to stabilize the debt-to-GDP ratio over the medium term. Fiscal policy matters in the short run because a subset of consumers is subject to liquidity constraints, and in the longer run through an ad hoc link between government debt and net foreign assets. There are financial intermediation costs (risk premiums) for accessing the international bond market but there is no distinction between gross and net positions, and hence limited scope for valuation changes, which have important wealth effects (see Lane and Milesi-Ferretti, 2007). See Faruqee and others (2005,2006) for a fuller description of the model.

Recession-Event Study

Potential output is estimated using two standard methodologies, the Hodrick-Prescott and the Baxter-King filters, and common parameters for the filters across countries (lambda = 1600 for the HP filter). While computations of the output gap could be improved by tailoring the parameters of the filters to each specific country and by using country-specific information such as the dynamics of inflation and unemployment, applying a common and systematic methodology across countries avoids perceptions that the data were mined to generate the desired cross-sectional results.

    Other Resources Citing This Publication