Date post: | 02-Oct-2016 |
Category: |
Documents |
Upload: | soyoung-kim |
View: | 213 times |
Download: | 1 times |
INTERNATIONAL CAPITAL FLOWS, BOOM-BUST CYCLES,AND BUSINESS CYCLE SYNCHRONIZATION
IN THE ASIA PACIFIC REGION
SOYOUNG KIM and SUNGHYUN H. KIM∗
This article documents evidence of business cycle synchronization in selectedAsia Pacific countries since the 1990s. We explain business cycle synchronizationby the channel of international capital flows and boom-bust cycles. Using the vectorauto-regression method, we find that most Asian countries experience boom-bustcycles following capital inflows, where the boom in output is mostly driven byconsumption and investment. Empirical evidence also shows that capital flow shocksare positively correlated in the region, which supports the conclusion that capitalmarket liberalization has contributed to business cycle synchronization. (JEL F4)
I. INTRODUCTION
Over the past decade, a number of AsiaPacific countries have liberalized their financialmarkets to foreign capital by reducing restric-tions on inward and outward capital flows.Increased capital flows due to financial integra-tion can generate substantial effects on businesscycles. Large capital inflows due to financialmarket liberalization can generate an initialsurge in investment and asset price bubbles fol-lowed by capital outflows and recession, theso-called boom-bust cycles. In the worst case,the boom-bust cycles can generate a suddenreversal of capital flows and eventually finan-cial crises.1 On the other hand, by allowingdomestic residents to engage in internationalfinancial asset transactions, financial market
∗We thank an editor and two anonymous referees fortheir helpful comments. S.K. acknowledges the support byResearch Settlement Fund for the new faculty of SeoulNational University.S. Kim: Department of Economics, Seoul National Univer-
sity, 599 Gwanak-ro, Gwanak-gu, Seoul, Korea. Phone(02) 880-2689, Fax (02) 886-4231, E-mail [email protected]
S. H. Kim: Department of Economics, Sungkyunkwan Uni-versity, Seoul, Korea, and Department of Economics,Suffolk University, Boston MA, USA. Phone 617-994-4232, Fax 617-994-4216, E-mail [email protected]
1. Although other fundamental domestic problems con-tribute to financial crises, capital account liberalization andthe resulting lending booms sometimes end in twin currencyand banking crises.
opening can reduce the volatility of somemacroeconomic variables such as consumptionthrough risk-sharing.2
What are the macroeconomic effects of capi-tal flows, in particular on business cycle fluctu-ations? Do business cycles become less volatileand more synchronized across countries as thedegree of financial integration increases? Under-standing business cycle implications of capitalflows is also important for analyzing wel-fare implications of financial market liberaliza-tion policies as well as international monetaryarrangements.
This article focuses on the effects of capi-tal flows due to financial market liberalizationon business cycles, in particular co-movementsacross countries.3 We aim to shed some lights
2. Domestic residents can reduce fluctuations in incomestream and consumption by borrowing from abroad duringrecessions or lending to foreign countries during booms.International portfolio diversification enables consumers andfirms to achieve risk-sharing gains by diversifying risksassociated with country-specific shocks.
3. We do not focus on the effects of capital flows onbusiness cycle volatility in this article. See, for example,Buch, Dopke, and Pierdzioch (2005) and Kose, Prasad, andTerrones (2003a, 2003b) on this issue.
ABBREVIATIONS
FDI: Foreign Direct InvestmentGDP: Gross Domestic ProductVAR: Vector Auto-Regression
1
Contemporary Economic Policy (ISSN 1465-7287)doi:10.1111/j.1465-7287.2011.00285.x© 2012 Western Economic Association International
2 CONTEMPORARY ECONOMIC POLICY
on this issue by providing detailed stylized factson capital flows and business cycles in theAsia Pacific region and by empirically analyzingthe relationship between capital flows and busi-ness cycles. For empirical analysis, we adoptthe vector auto-regression (VAR) method. First,we identify the capital flow shocks and exam-ine their effects on cyclical movements of keymacroeconomic variables in each country. Then,we examine whether these effects are consis-tent with the boom-bust cycle theory. By fur-ther analyzing the cross-country correlation ofcapital flow shocks, we can analyze the roleof capital flows in explaining business cyclesynchronization.
Economic theory does not provide a unani-mous prediction for the effects of capital flowson co-movements of business cycles. Financialmarket integration can increase business cycleco-movements as macroeconomic effects of cap-ital flows in different countries follow a similarpattern through various channels of contagionand common shocks.4 However, co-movementsof output can decrease as allocation of capitalbecomes more efficient, allowing production tobecome more specialized.5 Other variables canalso affect the relationship between capital flowsand business cycles, including monetary and fis-cal policies, the nature of underlying shocks inthe economy, etc.6
Using the data of 12 Asia Pacific countries,we find the following stylized facts of busi-ness cycles. First, business cycles in the fiveAsian Crisis countries are highly synchronizedand follow business cycles in Japan, while theydiffer from business cycles in Australia. On theother hand, Greater China, including Hong Kongand Taiwan, shows similar cyclical movementsamong themselves. Second, in general, businesscycles since the 1990s are more synchronizedacross countries than those in the 1980s, whichsupport the view that financial and trade inte-gration increases business cycle synchronizationin Asia.
4. See Kim, Kose, and Plummer (2001) for a detailedexplanation on financial contagion.
5. See, for example, Heathcote and Perri (2004), Imbs(2004), and Kalemli-Ozcan, Sorensen, and Yosha (2001).
6. Another important issue in the literature is trade inte-gration and its impact on business cycles. Trade integra-tion can generate synchronized business cycles if countriesmostly engage in intra-industry trade (production fragmen-tation), while trade integration can decrease the degree ofco-movements if trade promotes inter-industry specializa-tion and countries are subject to industry-specific shocks.See, for example, Frankel and Rose (1998) and Shin andWang (2004).
We find empirical evidence that positive cap-ital flow shocks (capital inflows) affect out-put, consumption, and investment positively inmost countries, which is consistent with theboom-bust cycle theory. In addition, capitalflow shocks are positively correlated across theAsian Crisis countries (except for the Philip-pines). These two results imply that capital flowshocks can explain business cycle synchroniza-tion among the Asian Crisis countries to someextent.
The remaining sections of this article areorganized as follows. Section II provides litera-ture review on the relationship between financialintegration and business cycles. In Section III,we analyze trends and stylized facts of businesscycles in the Asia Pacific region. In particular,we investigate how volatility of business cycleshas changed over time and whether we can findany evidence of business cycle synchronizationin the region. We analyze 12 sample countriesin the Asia Pacific region, including 5 AsianCrisis countries (Indonesia, Korea, Malaysia,the Philippines, and Thailand), China, Singa-pore, Taiwan, Hong Kong, Japan, Australia, andNew Zealand. Section IV provides an empiri-cal analysis of the relationship between capi-tal flows and business cycles. We analyze howcapital flow shocks affect various macroeco-nomic variables and investigate whether capi-tal flow shocks generate boom-bust cycles inthe region. We also analyze the properties ofcapital flow shocks identified in our model. Inparticular, we investigate whether the estimatedcapital flow shocks are driven by exogenouseconomic events and correlated across countries.Section V concludes the paper.
II. LITERATURE REVIEW
This section explains different theories onthe effects of economic integration on thesymmetry of business cycles and documentsempirical studies on this issue.7 Financial mar-ket integration can decrease co-movements ofoutput by increasing industrial specialization(Kalemli-Ozcan, Sorensen, and Yosha 2001).Countries with integrated international finan-cial markets can ensure against country-specific
7. Note that we focus on the effects of financial mar-ket integration on output co-movements, not cross-countryconsumption correlation which is expected to increase asconsumers in different countries receive a similar incomestream through portfolio diversification and consumptionsmoothing.
KIM & KIM: INTERNATIONAL CAPITAL FLOWS AND BOOM-BUST CYCLES 3
shocks through portfolio diversification; there-fore, such countries can afford to have aspecialized production structure. That is, finan-cial market integration allows firms to takefull advantage of comparative advantage andengage in production specialization, which inturn increases the asymmetry of output as longas industry-specific shocks exist.
Empirical analysis confirms a decrease incross-country correlation of output in the 1990s.This can be explained by a decrease in cross-country correlation of productivity shocks com-bined with increased financial market integration(Heathcote and Perri 2002). Degree of financialmarket integration endogenously and positivelyresponds to the correlation of shocks. That is,as productivity shocks become less correlated,potential welfare gains from portfolio diversifi-cation increase, as does the degree of financialmarket integration.
However, countries with liberalized capitalaccounts can be significantly more synchro-nized, even though they are more specialized(Imbs 2004). A large body of literature oncontagion argues that capital flows in differ-ent countries, in particular developing countriesin the same region, are synchronized throughvarious channels of financial contagion includ-ing herd behavior and information asymmetryamong others (Calvo and Mendoza 2000; Men-doza 2001). International investors may classifydifferent countries in a single group and makeregion-based investment decisions. In addition,capital flows can be highly synchronized ifshocks that determine capital flows are posi-tively correlated or spill over across countries,or if developing countries go through a finan-cial liberalization process at the same time. Ascapital inflows have significant effects on busi-ness cycles (so-called “boom-bust” cycles), ifcapital flows are highly correlated and have sim-ilar effects on business cycles, then financialintegration can contribute to synchronization ofbusiness cycles.
III. TRENDS AND STYLIZED FACTSOF BUSINESS CYCLES
This section documents the main characteris-tics of business cycles of the selected countriesin the Asia Pacific region.8 We use the annual
8. See Kim, Kose, and Plummer (2003) for a detailedanalysis of stylized facts of business cycles in Asia and theG-7 countries.
data from the International Financial Statis-tics and examine volatility (measured by stan-dard deviation) and co-movements (measured bycross-country correlation) of output, consump-tion, and investment in these countries.9 Thesample period is from 1980 to 2006 and allthe data are real (by gross domestic product[GDP] deflator) and Hodrick-Prescott filtered(with filtering parameter set at 100). As we areinterested in changes in business cycle statisticsas financial markets are liberalized, we exam-ine business cycles in different sub-sample peri-ods: 1980–1989 and 1990–2006. For the secondperiod, we use the data with and without theAsian Crisis period because the data for thatperiod may distort the statistics.
We focus on two aspects of business cyclesrelated to financial market liberalization andexamine whether the stylized facts derived fromthe data support the theoretical predictions stud-ied in the previous section. First, we investigatehow much the volatility of business cycles haschanged over time. As financial markets developover time, volatility of consumption is likely todecrease through consumption smoothing andrisk-sharing channels unless output volatilityincreases substantially. However, the impact onvolatility of output is more ambiguous as arguedin the previous section. Second, we focus on thedegree to which business cycles in the regionare synchronized and the changes in the degreeof business cycle synchronization over time. Weexpect that business cycles in this region becomemore synchronized due to the region’s tradeintegration and high portion of intra-industrytrade. However, the effects of financial inte-gration on business cycle co-movements areambiguous as argued in the previous section.
A. Volatility of Business Cycles
Table 1 presents volatility of output, rela-tive volatility of consumption, and investmentin four different periods: the whole period, the1980s, and the 1990s with and without theAsian Crisis period. The output volatility is lowwith a standard deviation ranging from 1.73 to2.70 in more developed countries in the region:Japan, Australia, and New Zealand. On the otherhand, less developed countries in the regionexhibit higher volatility: 5.93 in Thailand, 5.03in Indonesia, and 4.34 in Malaysia. Developed
9. Quarterly data are not available for some countries.Later for the VAR estimation, we use quarterly data.
4 CONTEMPORARY ECONOMIC POLICY
TABLE 1Volatility of Business Cycles
1980–2006
1980–1989
1990–2006
1990–2006(Without
Crisis)
Standard deviation of outputKorea 2.76 1.50 3.17 2.35Indonesia 5.03 1.28 6.22 5.63Malaysia 4.34 3.14 4.33 3.73Philippines 3.90 5.49 2.90 2.95Thailand 5.93 3.38 6.51 6.43Japan 2.70 0.98 2.79 2.83China 3.65 3.24 3.36 3.49Singapore 3.94 3.61 3.89 4.13Taiwan 2.43 2.51 2.23 2.24Hong Kong 3.77 2.87 4.13 4.05Australia 1.73 1.87 1.50 1.52New Zealand 2.60 2.23 2.45 2.52
Relative standard deviation of consumptionKorea 1.27 0.72 1.28 1.23Indonesia 1.03 2.37 0.88 0.97Malaysia 1.36 1.38 1.29 1.23Philippines 0.91 0.69 1.08 1.10Thailand 0.84 0.64 0.83 0.78Japan 0.81 0.82 0.91 0.92China 0.98 0.87 0.62 0.61Singapore 0.84 1.05 0.76 0.59Taiwan 1.23 1.51 1.05 1.08Hong Kong 1.05 0.84 1.16 1.18Australia 0.52 0.45 0.68 0.60New Zealand 0.93 0.86 1.10 1.14
Relative standard deviation of investmentKorea 4.47 3.42 4.34 3.56Indonesia 4.13 7.44 3.98 3.96Malaysia 4.23 4.82 4.05 4.09Philippines 4.33 4.55 2.82 2.34Thailand 3.47 3.14 3.38 3.16Japan 2.41 5.60 1.89 1.84China 2.21 2.32 2.36 2.42Singapore 3.40 2.44 3.84 3.67Taiwan 4.90 5.84 4.60 4.59Hong Kong 3.01 4.71 2.55 2.40Australia 4.34 3.57 4.03 4.13New Zealand 3.86 4.07 3.52 3.53
countries tend to have more stable industrialstructures and output streams. Small countriesthat depend on natural resources for their mainproducts tend to have volatile output streamsdue to volatile prices (terms of trade) of pri-mary goods. Moreover, the share of agriculturalactivity is higher and the shares of the indus-try and service sectors are lower in the lessdeveloped countries. The agricultural sector out-put is highly variable as it is heavily affectedby extremely volatile productivity and priceshocks.
Comparison of output volatility in the twoperiods shows mixed results. Five countriesshow significant increases (Korea, Indonesia,Malaysia, Thailand, and Japan), one countryshows a significant decrease (the Philippines),and the remaining countries do not experi-ence significant changes over time. Except forthe Philippines, the five Asian Crisis countriesexhibit higher volatility of output in the 1990scompared to the 1980s. This result is consistenteven when the crisis period is excluded.
According to the consumption smoothingproperty in the inter-temporal current accountmodel, consumption should be less volatile thanoutput (Obstfeld and Rogoff 1996). Countries,when facing positive shocks, lend to foreigncountries to smooth the consumption streamover time, vice versa. However, in the table,we observe that this is not the case in someof our sample countries.10 The table shows thatconsumption volatility is significantly less thanoutput volatility in seven countries, mostly inmore developed countries in the region suchas Japan and Australia. Developed countriescan smooth their consumption by using variousrisk-sharing instruments. As financial marketsdevelop, developing countries should be ableto gain access to the risk-sharing instrumentsand reduce the volatility of their consumptionstream. However, consumption volatility doesnot change much over time and no explicit pat-tern is detected in the table.
Investment is three to four times morevolatile than output in the table, which is thetypical result in other empirical and simulationstudies (Baxter and Crucini 1995; Kim, Kose,and Plummer 2001). Investment volatility inChina, Hong Kong, and Japan is among the low-est with a relative standard deviation of less thanor around three, while investment in the fiveAsian Crisis countries is quite volatile with arelative standard deviation higher than four. Wedo not find any significant changes in investmentvolatility from the 1980s to the 1990s, except forIndonesia and Japan, where investment volatilitysignificantly decreases.
Including the Asian Crisis period in the datafor the 1990s does not significantly change thestatistics for all three variables. In particular,
10. We should note that the volatility of consump-tion changes depending on the specific consumption data.It is known that the volatility of durable goods con-sumption is two to four times higher than that of non-durables consumption (see Backus, Kehoe, and Kydland1995).
KIM & KIM: INTERNATIONAL CAPITAL FLOWS AND BOOM-BUST CYCLES 5
volatility does not change much by including orexcluding the crisis period in the data. In con-clusion, we find that output volatility increasesin the 1990s in many countries and consump-tion smoothing is not realized in some countriesas consumption volatility is higher than outputvolatility.
B. Co-Movements of Business Cycles
To illustrate the degree to business cyclesynchronization across countries, we calculatecross-country correlation of output in Table 2.The first panel shows the results from theentire sample period. A significant and pos-itive correlation exists across most countries,except for Australia. Business cycles of Aus-tralia are negatively correlated with five Asiancountries and two countries show near zerocorrelation. This is not surprising because theindustrial structure of Australia is quite differ-ent from the typical structure in Asian coun-tries. Greater China—China, Hong Kong andTaiwan—shows high positive correlation amongeach other. This can be explained by the fact thatthe three economies are in the same economiczone.11 A high correlation between Malaysiaand Singapore can be explained in the samecontext.
The seven Asian Crisis countries (includ-ing Singapore and Hong Kong) show posi-tive output correlation among each other andtheir business cycles are positively correlatedwith those in Japan. This indicates that Japanhas been leading the business cycles in thatregion. McKinnon and Schnabl (2002) showedthat the yen/dollar exchange rate significantlyaffects business cycles in the East Asian coun-tries through trade and foreign direct investment(FDI) channels. For example, depreciation of the
11. Since its recent economic reform, China hasembarked upon a process of financial and real integrationwith Hong Kong and Taiwan. Even before Hong Kong’sreturn to China’s sovereignty in 1997, it had achieved ahigh degree of integration with the mainland. With respectto trade, for instance, Hong Kong intermediates a lion’sshare of China’s external trade via re-exports and offshoretrade. Regarding financial activity, a substantial amount ofthe international capital (in the forms of foreign direct invest-ment, equity and bond financing, and syndicated loans)financing China’s economic expansion is raised via HongKong. Economic links between China and Taiwan have alsoproliferated since the 1990s. According to official statistics(although the official statistics under-represent the overalleconomic interest of Taiwan in China), China is the largestrecipient of Taiwan’s overseas investment and Taiwan isChina’s third-largest source of foreign direct investment(Cheung, Chinn, and Fujii 2003).
yen in 1995 significantly slowed down the EastAsian export expansion, while yen appreciationaccelerates Japanese FDI into the East Asiancountries. Bayoumi and Eichengreen (1999) findthat the correlation of supply shocks in theregion is especially high for two groups, withJapan and Korea in one group and Indonesia,Malaysia, and Singapore in the other. Loayza,Lopez, and Ubide. (2001) examine common pat-terns in aggregate demand and supply shockswith a different methodology. They find strongco-movements for two groups: Japan, Korea,and Singapore make up one group, and Indone-sia, Malaysia, and Thailand, in the other group.These results indicate that there are two differentbusiness cycles in the region, even though theEast Asian countries show strong co-movementsas a whole.
Comparison of the data in the 1980s and1990s proves that business cycles are more syn-chronized in the 1990s. We examine this prop-erty by comparing the number of negative cross-country correlations of output in the two periods.We observe negative correlation in 17 coun-try pairs during the 1980s, while the numberdecreases to 8 in the 1990s: only Australia dis-plays a negative correlation. From a total of 66pairs, 47 pairs show that correlation increasesfrom the 1980s to the 1990s.12 In fact, corre-lation coefficients are significantly positive inmost of the 47 cases; only two pairs exhibit acorrelation coefficient of less than 0.4.
Empirical results on business cycle co-movements in previous studies were mixed,depending on sample countries and periods.Some document that the correlation of outputdecreases over time, in particular in the 1990s.Heathcote and Perri (2002) showed that outputcorrelation among the United States, Europe,Canada, and Japan dropped from 0.76 to 0.26.On the other hand, Kose, Prasad, and Terrones(2003a), using the data for 21 industrial and 55developing countries, showed that output cor-relation in general increased in the 1990s fromthe previous periods. This is mostly due to theindustrial countries in the sample. The empiri-cal results in this article support the view thatbusiness cycles become more synchronized asfinancial markets are liberalized.13
12. This case is indicated by bold and italic numbersin the table. We do not report the case excluding the crisisperiod, but the results are similar.
13. Another channel that can increase output correlationis through increased intra-industry trade (production frag-mentation).
6 CONTEMPORARY ECONOMIC POLICY
TA
BL
E2
Cro
ss-C
ount
ryC
orre
latio
nof
Out
put
Kor
eaIn
done
sia
Mal
aysi
aP
hilip
pine
sT
haila
ndJa
pan
Chi
naSi
ngap
ore
Taiw
anH
ong
Kon
gA
ustr
alia
1980
–20
06In
done
sia
0.70
Mal
aysi
a0.
480.
83Ph
ilipp
ines
0.39
0.48
0.49
Tha
iland
0.74
0.89
0.76
0.59
Japa
n0.
620.
710.
420.
550.
77C
hina
0.26
0.43
0.15
−0.1
90.
250.
26Si
ngap
ore
0.25
0.53
0.72
0.54
0.59
0.37
0.13
Taiw
an0.
460.
310.
060.
230.
30.
510.
490.
21H
ong
Kon
g0.
740.
690.
390.
460.
640.
690.
520.
420.
69A
ustr
alia
0.06
−0.2
1−0
.43
−0.0
7−0
.16
0.08
0.2
−0.1
90.
430.
22N
ewZ
eala
nd0.
470.
430.
110.
210.
320.
310.
660.
130.
510.
680.
4119
80–
1989
Indo
nesi
a0.
05M
alay
sia
−0.1
30.
53Ph
ilipp
ines
0.26
0.48
0.63
Tha
iland
0.35
0.64
0.57
0.77
Japa
n0.
180.
490.
290.
320.
82C
hina
0.30
−0.4
7−0
.65
−0.7
8−0
.38
0.06
Sing
apor
e−0
.06
0.53
0.99
0.69
0.56
0.23
−0.7
1Ta
iwan
0.80
0.29
−0.3
0−0
.01
0.33
0.33
0.46
−0.2
8H
ong
Kon
g0.
780.
43−0
.01
0.33
0.34
0.14
0.05
0.09
0.74
Aus
tral
ia0.
140.
45−0
.17
−0.1
60.
430.
770.
41−0
.24
0.59
0.29
New
Zea
land
0.68
0.31
−0.1
1−0
.03
0.11
0.05
0.32
−0.0
40.
740.
830.
3019
90–
2006
Indo
nesi
a0.
82M
alay
sia
0.82
0.93
Phili
ppin
es0.
420.
590.
41T
haila
nd0.
830.
950.
880.
49Ja
pan
0.60
0.85
0.70
0.76
0.79
Chi
na0.
510.
830.
750.
520.
850.
81Si
ngap
ore
0.43
0.55
0.57
0.37
0.61
0.56
0.78
Taiw
an0.
380.
550.
610.
460.
480.
640.
680.
77H
ong
Kon
g0.
750.
870.
780.
610.
850.
870.
830.
730.
65A
ustr
alia
−0.1
2−0
.27
−0.3
50.
43−0
.29
−0.1
2−0
.24
0.07
−0.0
9−0
.03
New
Zea
land
0.56
0.70
0.53
0.80
0.72
0.65
0.69
0.49
0.28
0.65
0.28
Not
e:N
egat
ive
coef
ficie
nts
are
italic
ized
.B
old
num
bers
inth
ebo
ttom
pane
lin
dica
teth
atco
rrel
atio
nco
effic
ient
sin
crea
sein
the
seco
ndpe
riod
.
KIM & KIM: INTERNATIONAL CAPITAL FLOWS AND BOOM-BUST CYCLES 7
In conclusion, we can summarize the maincharacteristics of the business cycle co-movements as follows. First, business cycles inAustralia are different from those in the EastAsian countries. Second, business cycles in thefive Asian Crisis countries are highly synchro-nized and follow business cycles in Japan. Third,the countries in Greater China, which encom-passes Hong Kong and Taiwan, show similarcyclical movements. Finally, business cycles ingeneral are more synchronized across countriesin the 1990s than in the 1980s, which support theview that financial integration increases businesscycle synchronization.
IV. CAPITAL FLOWS AND BUSINESS CYCLES:EMPIRICAL STUDIES
In this section, we investigate how cap-ital flow shocks affect the business cycledynamics of the Asia Pacific countries: forexample, whether capital flows generate boom-bust cycles, and whether capital flows helpexplain the synchronization of the businesscycles in the Asian countries. Capital flows,especially after the financial market liberaliza-tion, may increase the volatility of businesscycles by creating boom-bust cycles, in par-ticular fluctuations in investment, consumption,exchange rate, and other asset prices. Further-more, if capital flows are positively correlatedacross countries, due to simultaneous capitalmarket liberalization in Asian countries or dueto the herd behavior of international investors ordue to common shocks, the boom-bust cycles ineach country may imply the synchronization ofthe business cycles.
For empirical methodology, we adopt theVAR estimation method to extract the shocks tocapital flows, to analyze how shocks to capitalflows affect the various macroeconomic vari-ables in each country, and to examine how theshocks to capital flows are correlated acrosscountries.14
A. VAR Model
We assume that the economy is described bya structural-form equation
G(L)y t = et(1)
14. A similar empirical methodology was used in Kim,Kim, and Wang (2003) to analyze the boom-bust cycles inKorea. Tornell and Westermann (2002) also examined theboom-bust cycles by using a sample of 39 countries.
where G(L) is a matrix polynomial in thelag operator L, y t is an n × 1 data vector,and et is an n × 1 structural disturbance vec-tor.15 We assume that et is serially uncorrelatedand var(et ) = �, which is a diagonal matrixwhere the diagonal elements are the variances ofstructural disturbances. That is, structural distur-bances are assumed to be mutually uncorrelated.
We can estimate a reduced-form equation(VAR)
y t = B(L)yt−1 + ut ,(2)
where B(L) is a matrix polynomial in lagoperator L and var(ut ) = �.
There are several ways of recovering theparameters in the structural-form equation fromthe estimated parameters in the reduced-formequation. The identification schemes under con-sideration impose restrictions on contemporane-ous structural parameters only. Let G0 be thecontemporaneous coefficient matrix in the struc-tural form, and let G0(L) be the coefficientmatrix in G(L) without the contemporaneouscoefficient G0. That is,
G(L) = G0 + G0(L).(3)
Then, the parameters in the structural-formequation and those in the reduced-form equationare related by
B(L) = −G−10 G0(L).(4)
In addition, the structural disturbances andthe reduced-form residuals are related by
et = G0ut ,(5)
which implies
� = G−10 �G−1
0 .(6)
In the method proposed by Sims (1980), iden-tification is achieved by Cholesky decompositionof the reduced-form residuals, �. In this case, G0becomes triangular so that a recursive structure,that is, the Wold-causal chain, is assumed. Ina general non-recursive modeling strategy sug-gested by Blanchard and Watson (1986) andSims (1986), maximum likelihood estimates of� and G0 can be obtained only through thesample estimate of �. The right-hand side ofEquation (6) has n × (n + 1) free parameters tobe estimated. As � contains n× (n + 1)/2
15. For simplicity, we present the model without thevector of constants. Alternatively, we can regard eachvariable as a deviation from its steady state.
8 CONTEMPORARY ECONOMIC POLICY
parameters, by normalizing n diagonal elementsof G0 to 1s, we need at least n × (n − 1)/2restrictions on G0 to achieve identification. Inthis generalized structural VAR approach, G0can be any structure (non-recursive). In this arti-cle, recursive modeling is used.
B. Basic Model and Effects on Output
We construct a basic model to examine theeffects of capital flow shocks on output. Thebasic model includes three variables, {CUR,RGDP, CAP}, where CUR is the current account(as the ratio to the trend GDP), RGDP is the logof real GDP, and CAP is the capital account(as the ratio to the trend GDP).16 A constantterm and four lags are assumed. CAP and RGDPare included in the model as they are primaryvariables of interest; we examine the effectsof capital flows or capital account on the realGDP. CUR is included to control the capi-tal account movements that depend on currentaccount movements as some capital accountmovements are often related to the financingof current account imbalances and we are inter-ested in extracting autonomous capital flows.
The basic model uses a recursive struc-ture, in which the ordering of the variables is{CUR, RGDP, CAP}, where the contempora-neously exogenous variables are ordered first.With this ordering, the shocks to capital flowsare extracted by conditioning on the current andlagged CUR and RGDP, in addition to their ownlagged variables. We condition on the current(and lagged) CUR as current account imbal-ances are often financed by capital account. Weexclude such endogenous movements of capitalflows from the shocks to capital flows. In addi-tion, we condition on the current (and lagged)real GDP as changes in the real GDP may affectthe capital account. For example, an increasein the real GDP may attract more capital, andimprove the capital account. We exclude theendogenous movements of capital flows due tothe real GDP changes from the shocks to capi-tal flows as we would like to infer the effects ofcapital flow shocks to real GDP.17
16. We use an exponential trend on the GDP level (ora linear trend on the log level of GDP). When constructingthe ratio, we use all variables in terms of U.S. dollars.
17. Note that the effects of CAP shocks on CUR andRGDP are invariant to the ordering between CUR andRGDP. On the other hand, capital flows might affect CURand RGDP within a quarter, and the CUR and RGDP shocksmay reflect some part of (exogenous) CAP shocks. However,even in such cases, CAP shocks still represent the shocks to
The sample period is 1990–2006, duringwhich capital account was liberalized in theseAsian-Pacific countries (de Brouwer 1999, 2001;Grenville 1998). We consider two samples, onewith the Asian Crisis period and the otherwithout it (dropping 1997:3–1998:2). We relatethe capital flow shocks identified in the modelto the financial market liberalization and theglobal common shocks under a more liberal-ized financial market. If the capital account hadbeen tightly controlled (i.e., China), the shocksto capital flows in our model or autonomouscapital flows would have been very small asthe capital account should have been directedto finance the current account imbalances (notethat our model identifies capital flow shocks,by controlling for the current account move-ment). Therefore, by examining the effects ofautonomous capital account shocks during thesample period, we can infer the consequencesof capital account liberalization.
We use quarterly data for the estimation asmonthly data is not available for most countries.We consider eight countries for which quar-terly data series are available for most of thesample period. They are Korea, Japan, Indone-sia, Thailand, the Philippines, Taiwan, Australia,and New Zealand.18 Data sources are Interna-tional Financial Statistics, Bank of China, andBIS Database.
The impulse responses to CAP shocks over3 years are reported in Figure 1 for the sampleincluding the crisis period and Figure 2 for thesample dropping the crisis period. Dotted linesare one standard error bands. The scale repre-sents percentage changes. At the top of eachcolumn, the country names are denoted. At thefar left of each row, the name of each respondingvariable is reported.
First, we explain the results for the sam-ple including the crisis period. In response topositive CAP shocks, the real GDP tends toincrease in all countries. In New Zealand, thereal GDP decreases in the very short-run butincreases over time. In Taiwan, the increase isnot clear, considering the wide error band. In all
CAP that are not endogenous to CUR and RGDP changesas they do not result from endogenous responses to CURand RGDP, although CUR and RGDP shocks may include(exogenous) shocks to CAP in addition to shocks to CURand RGDP.
18. The estimation period for Thailand is from 1993 asthe data series are available only from 1993. The estimationperiod for the regression including real exchange rate ofThailand and Indonesia is from 1994 as the data series areavailable only from 1994.
KIM & KIM: INTERNATIONAL CAPITAL FLOWS AND BOOM-BUST CYCLES 9
FIG
UR
E1
Eff
ects
ofC
apita
lFl
owSh
ocks
:Sa
mpl
eIn
clud
ing
Cri
sis
Peri
od
KIM & KIM: INTERNATIONAL CAPITAL FLOWS AND BOOM-BUST CYCLES 11
FIG
UR
E2
Eff
ects
ofC
apita
lFl
owSh
ocks
:Sa
mpl
eE
xclu
ding
Cri
sis
Peri
od
KIM & KIM: INTERNATIONAL CAPITAL FLOWS AND BOOM-BUST CYCLES 13
other countries, the increase in real GDP is clear.The positive effect of capital inflows is signifi-cant in most countries, including all crisis coun-tries under consideration, and quite persistent inmany countries. Positive effects are especiallystrong in the four Asian Crisis countries (Korea,Indonesia, Thailand, and the Philippines). InIndonesia, Thailand, and the Philippines, posi-tive effects are different from 0 with more than84% probability for more than 4 years. In Korea,positive effects are different from 0 with morethan 84% probability for more than one and ahalf years. In other countries such as Australiaand Japan, positive effects are different from 0with more than 84% probability for more than1 year. The results for the sample excluding thecrisis period, reported in Figure 2, are not muchdifferent, although the persistence of real GDPresponses tends to be smaller.
C. Effects on Other Macro Variables
We modify the basic model to examine theeffects of capital flow shocks on other macro-economic variables. The modified model uses arecursive structure, in which the ordering of thevariables is {CUR, X, CAP}, where X denotes thevariable of interest. With this ordering, the shocksto capital flows are extracted by conditioning onthe current and lagged CUR and X, in additionto their own lagged variables. We condition onthe current (and lagged) CUR and X as before.First, the current account imbalances are oftenfinanced by capital account, and we would liketo exclude such endogenous movements of cap-ital flows from the shocks to capital flows. Sec-ond, we condition on the current (and lagged) Xas changes in X may affect the capital account.19
We include (real) consumption, (real) invest-ment, and the real effective exchange rate as X.Each variable is used as a log form. To constructreal consumption and real investment, nominaldata are deflated by using a GDP deflator. Notethat an increase in the real exchange rate is areal exchange rate appreciation.20
Figures 3 and 4 report the results for theperiod including and excluding crisis period,respectively. The first two rows report the
19. As in the basic model, we order X before CAP.By doing so, CAP shocks represent the shocks to CAPthat are not endogenous to CUR and RGDP changes sincethey do not result from endogenous responses to CUR andX, although CUR and X shocks may include (exogenous)shocks to CAP, in addition to shocks to CUR and RGDP.
20. For Taiwan, consumer price index is used, as a GDPdeflator is not available.
responses of consumption (“CONS”) and invest-ment (“INV”). In all countries, at least one ofthe two variables (consumption and investment)increases. For the full sample period (Figure 3),both consumption and investment increase inall countries but in Taiwan, Indonesia, and thePhilippines only the investment increases. Forthe sample period excluding the crisis period(Figure 4), one or two variables increase inall countries. From this analysis, we can inferthat the increase in output following capitalflow shocks is mostly due to the increase inconsumption and investment because the cur-rent account negatively responds to capital flowshocks (Figures 1 and 2).
The third row reports the real effectiveexchange rate (“RER”). As we expected, realexchange rate appreciation is observed in mostcountries. In particular, a clear exchange rateappreciation is found in the four Asian Cri-sis countries, even in the sample excluding theAsian Crisis period. For the other four coun-tries, real exchange rate appreciation is foundin New Zealand and Australia, but not in Japanand Taiwan.
D. Properties of Estimated Capital FlowShocks
The validity of the VAR results in the pre-vious section depends on the identification ofshocks, whether capital account shocks repre-sent exogenous changes in capital flows, forexample, due to capital account liberalization ordue to abrupt changes in the behavior of interna-tional investors as in the financial crisis or due toglobal common shocks. In this part, we exam-ine whether the estimated capital flow shocksactually represent such shocks by plotting cumu-lative capital flow shocks for each country andrelating them to economic events occurred.
Figure 5 plots identified cumulative capitalaccount shocks in each country.21 For AsianCrisis countries, we observe positive capital flowshocks during the period 1994–1996 whenthese countries actively embarked on financialmarket deregulation and opening (de Brouwer1999; Furman and Stiglitz 1998; Kim, Kose,and Plummer 2001, 2003). For example, Koreaallowed non-residents to directly purchase stocksof Korean companies up to 3% per individ-ual in 1992 and this share increased to 23% inMay 1997. As a result, the external debt in these
21. We plot cumulative capital flow shocks becausecapital account shocks themselves are very volatile.
14 CONTEMPORARY ECONOMIC POLICY
FIG
UR
E3
Eff
ects
ofC
apita
lFl
owSh
ocks
onV
ario
usM
acro
econ
omic
Var
iabl
es:
Sam
ple
Incl
udin
gC
risi
sPe
riod
16 CONTEMPORARY ECONOMIC POLICY
FIG
UR
E4
Eff
ects
ofC
apita
lFl
owSh
ocks
onV
ario
usM
acro
econ
omic
Var
iabl
es:
Sam
ple
Exc
ludi
ngC
risi
sPe
riod
KIM & KIM: INTERNATIONAL CAPITAL FLOWS AND BOOM-BUST CYCLES 19
crisis-hit countries increased dramatically for3 years from 1994 to 1996.
This period also coincides with low worldinterest rate and the appreciation of yen. Yenappreciation increased Japanese overseas directinvestment in East Asia. Low interest rates inthe industrial countries including Japan pro-duced the portfolio flows to the East Asianeconomies. On the other hand, the graphs shownegative capital flow shocks during the cri-sis period 1997–1998 as large current accountdeficits turned into surpluses.
Australia and New Zealand recorded per-sistent current account deficits throughout the1990s. For Australia, we observe positive capitalflow shocks from the mid-1990s when the coun-try persistently marked current account deficits.For New Zealand, the capital inflows continueduntil 1997 and the capital account reversed intodeficits during 1998–2000. In contrast, Taiwanexperienced current account surpluses and netcapital outflows before the Asian Crisis. Thus,for Taiwan, we observe negative capital flowshocks in 1995–1996.
Finally, we observe that capital flows intomost of these countries in the mid-2000s. Wealso observe that capital flows out of some coun-tries at the end of the sample period, in whichthe global financial crisis started.
E. Synchronization of Capital Flowsand Business Cycles
In the previous parts, we have shown that apositive shock to capital flows increases output
in most countries, and the increase in out-put is mostly due to a boom in consumptionand investment. These findings, especially forthe case of the full sample including the cri-sis period, are consistent with the “boom-bust”cycle theory following the financial market lib-eralization. In our model, a big surge in capitalinflows after the financial market liberalizationcan be captured as a positive shock to capitalflows, and such a positive shock leads to a boom.Later, when capital flows are reversed, capitaloutflows can be captured as negative shocks tocapital flows in our model, and such a negativeshock leads to a bust stage.
However, the evidence alone is not enoughto support the hypothesis that capital flowshocks or the financial market liberalization pro-cess increases business cycle synchronizationin the Asia Pacific region. Only when cap-ital flow shocks are highly correlated acrosscountries in the region, can they increase co-movements of business cycles. Otherwise, cap-ital flow shocks may not contribute to businesscycle synchronization.
In this regard, Table 3 reports the cross-country correlations of the capital flow shocksidentified in our model, for the periods with andwithout the crisis. For the sample period with thecrisis, the correlation is positive in most cases.Negative correlations are found only in 6 outof 28 cases. Therefore, as shown in the previ-ous section, as capital flow shocks have similareffects on business cycles across countries, wecan conclude that capital flow shocks contribute
TABLE 3Cross-Country Correlation of Cumulative Capital Flow Shocks
Korea Indonesia Philippines Thailand Japan Taiwan Australia
Including the crisis periodIndonesia 0.41Philippines 0.13 0.14Thailand 0.00 0.16 0.06Japan 0.29 −0.07 −0.13 0.04Taiwan 0.02 0.13 −0.02 0.05 0.04Australia 0.27 0.02 −0.24 0.02 0.19 0.00New Zealand 0.17 0.18 −0.12 0.20 0.27 −0.21 0.20
Excluding the crisis periodIndonesia 0.09Philippines 0.04 −0.12Thailand −0.04 0.10 0.00Japan 0.23 −0.02 −0.14 −0.02Taiwan −0.10 0.13 −0.06 −0.05 0.03Australia 0.39 0.18 −0.24 −0.02 0.19 0.01New Zealand 0.23 0.29 −0.10 0.17 0.27 −0.21 0.20
20 CONTEMPORARY ECONOMIC POLICY
to business cycle synchronization among the cri-sis countries. For the sample period without thecrisis, there are more cases of negative correla-tions, but same conclusion holds. In addition, theaverage of positive correlations is 0.17 but theaverage of negative correlations is −0.09. Thatis, the case of positive correlation is strongerthan the case of negative correlation.
We suggest two possible reasons to explainwhy capital flow shocks are positively corre-lated. First, the timing of financial market liber-alization in those countries was similar, and eachcountry experienced a boom-bust cycle after theliberalization. Thus, the financial market liberal-ization process itself contributes to the synchro-nization of the business cycles. Second, givensome extent of openness in the financial markets,contagion through financial channels contributedto similar capital flows in these countries. Dueto information cascade, international investorsclassify these countries in the same group andapply a single investment decision for the wholegroup. Combined with herd behavior, financialcontagion contributed to the synchronization ofcapital flows and eventually of business cycles.
V. CONCLUSION
The relationship between financial integra-tion and co-movements of business cycles isnot unambiguous, both theoretically and empiri-cally. In this article, we first document businesscycle synchronization in a number of the AsiaPacific countries and explain the phenomenonby examining financial market liberalization andcapital flows. We find that business cycle syn-chronization among the Asian Crisis countries inthe 1990s can be at least partially explained bysynchronization of capital flows and the ensu-ing boom-bust cycles after the financial marketliberalization. That is, business-cycle synchro-nization increased mostly because of commoncapital inflow shocks, rather than via direct con-tagion through commercial capital flows. There-fore, the results imply that financial market liber-alization is likely to synchronize business cyclesacross a group of countries. This is an interestingfinding as recent studies using data from devel-oped countries often conclude the opposite.
Understanding the effects of capital flowson business cycle co-movements has impor-tant implications on various issues. First, poten-tial welfare gains from international risk-sharinghighly depend on the degree of business cycle
synchronization across countries. When coun-tries follow similar business cycles, it is lessefficient to share risks across countries. Iffinancial market liberalization and capital flowsincrease business cycle co-movements, thenpotential welfare gains from financial marketliberalization would be lower than the levelcalculated from the existing level of businesscycle co-movements. Therefore, potential wel-fare gains from financial market liberalizationmight be over-estimated.
Second, the findings of this article can pro-vide implications on financial market liberaliza-tion policies. In implementing financial marketliberalization policies, policymakers should con-sider the effects of the speed and sequencing ofsuch policies on business cycles and eventuallyon welfare. Finally, our results have implicationson regional monetary and financial integrationin terms of optimum currency area criteria. Forexample, one of the conditions for an optimumcurrency area is the presence of similar busi-ness cycle movements in the potential candidatecountries.
When most emerging East Asian countriesstarted to liberalize their financial markets inthe early 1990s, only a few regional risk-sharing channels such as trade channels existed.Although Japan still remained an importantsource country for external financing before thecrisis, Western investors outside the region alsoplayed an important role. Since the crisis, how-ever, most East Asian countries have becomenet providers of international capital due totheir current account surpluses. While receivinginflows of foreign direct and portfolio invest-ment on a net basis, these countries have repaidlarge sums of bank loans for the past severalyears. Looking to the future, whether countriesin the Asia Pacific region have similar patternsof capital flows will be an empirical question.However, until a regional risk-sharing mecha-nism for integrating the financial markets in theregion is fully developed, most East Asian coun-tries are likely to become more integrated intothe global financial markets.
REFERENCES
Backus, D., P. Kehoe, and F. Kydland. “International Busi-ness Cycles: Theory and Evidence,” in Frontier ofBusiness Cycle Research, edited by T. Cooley. Prince-ton, NJ: Princeton University Press, 1995, 331–56.
Baxter, M., and M. Crucini. “Business Cycles and AssetStructure of Foreign Trade.” International EconomicReview, 36, 1995, 821–54.
KIM & KIM: INTERNATIONAL CAPITAL FLOWS AND BOOM-BUST CYCLES 21
Bayoumi, T., and B. Eichengreen. “Is Asia an OptimumCurrency Area? Can It Become One?” in ExchangeRate Policies in Emerging Asian Countries, editedby S. Collinon, J. Pisani-Ferry, and Y. C. Park. NewYork: Routledge, 1999, 318–36.
Blanchard, O., and M. Watson. “Are Business Cycles AllAlike?” in The American Business Cycle: Continuityand Change, edited by R. Gordon. Chicago: Universityof Chicago Press, 1986, 123–56.
Buch, C., J. Dopke, and C. Pierdzioch. “Financial Opennessand Business Cycle Volatility.” Journal of Interna-tional Money and Finance, 24, 2005, 744–65.
Calvo, G., and E. Mendoza. “Rational Contagion and theGlobalization of Securities Markets.” Journal of Inter-national Economics, 51, 2000, 79–113.
Cheung, Y., M. Chinn, and E. Fujii. “China, Hong Kong,and Taiwan: A Quantitative Assessment of Real andFinancial Integration.” China Economic Review, 14,2003, 281–303.
de Brouwer, G. J. Financial Integration in East Asia. Cam-bridge: Cambridge University Press, 1999.
. “East Asian Financial Markets: Current Status andFuture Development,” in Regional Financial Arrange-ments in East Asia, edited by Y. H. Kim and Y. Wang.Seoul, Korea: Korea Institute for International Eco-nomic Policy Press, 2001.
Frankel, J., and A. Rose. “The Endogeneity of the OptimumCurrency Area Criteria.” The Economic Journal, 108,1998, 1009–25.
Furman, J., and J. Stiglitz. “Economic Crises: Evidenceand Insights from East Asia.” Brookings Papers onEconomic Activity, 2, 1998, 1–135.
Grenville, S. “Capital Flows and Crises.” Reserve Bank ofAustralia Bulletin, December 1998, 16–31.
Heathcote, J., and F. Perri. “Financial Globalization andReal Regionalization.” Journal of Economic Theory,119, 2004, 207–43.
Imbs, J. “Trade, Finance, Specialization and Synchroniza-tion.” Review of Economics and Statistics, 2004, 86,723–34.
Kalemli-Ozcan, S., B. Sorensen, and O. Yosha. “EconomicIntegration, Industrial Specialization, and the Asym-metry of Macroeconomic Fluctuations.” Journal ofInternational Economics, 55, 2001, 107–37.
Kim, S., S. H. Kim, and Y. Wang. “Macroeconomic Effectsof Capital Account Liberalization: The Case of Korea.”Review of Development Economics, 8, 2003, 624–39.
Kim, S. H., A. Kose, and M. Plummer. “Understanding theAsian Contagion: An International Business Cycle Per-spective.” Asian Economic Journal, 15, 2001, 111–38.
. “Dynamics of Business Cycles in Asia: Differencesand Similarities.” Review of Development Economics,7, 2003, 462–77.
Kose, A., E. Prasad, and M. Terrones. “How Does Global-ization Affect the Synchronization of BusinessCycles?” American Economic Review, 93, 2003a,57–62.
. “Financial Integration and Macroeconomic Volatil-ity.” IMF Working Paper No. 03–50, InternationalMonetary Fund, 2003b.
Loayza, N., H. Lopez, and A. Ubide. “Co-Movements andSectoral Interdependence: Evidence for Latin America,East Asia and Europe.” IMF Staff Papers, 48, 2001,367–95.
McKinnon, R., and G. Schnabl. “Synchronized BusinessCycles in East Asia: Fluctuations in the Yen/DollarExchange Rate and China’s Stabilizing Role.” IMESDiscussion Paper Series No. 2002-E-13, Bank ofJapan, 2002.
Mendoza, E. “Credit, Prices and Crashes: Business Cycleswith a Sudden Stop.” NBER Working Paper No. 8338,2001.
Obstfeld, M., and K. Rogoff. Foundation of InternationalMacroeconomics. Cambridge, MA: MIT Press, 1996.
Shin, K., and Y. Wang. “Trade Integration and BusinessCycle Co-Movements: The Case of Korea with OtherAsian Countries.” Japan and the World Economy, 16,2004, 213–30.
Sims, C. “Macroeconomics and Reality.” Econometrica, 48,1980, 1–48.
. “Are Forecasting Models Usable for Policy Analy-sis?” Federal Reserve Bank of Minneapolis QuarterlyReview, 10-1, 1986, 2–16.
Tornell, A., and F. Westermann. “Boom-Bust Cycles inMiddle Income Countries: Facts and Explanation.”NBER Working Paper 9219, 2002.