Physica A 387 (2008) 2561–2564www.elsevier.com/locate/physa

Exchange rate regimes, saving glut and the Feldstein–Horiokapuzzle: The East Asian experience

Secil Kaya-Bahce, Erdal Ozmen∗

Middle East Technical University, Department of Economics, 06531, Ankara, Turkey

Received 13 September 2007; received in revised form 24 October 2007Available online 9 January 2008

Abstract

This paper investigates whether the recent experience of the emerging East Asian countries with current account surpluses isconsistent with the “saving glut” hypothesis and the Feldstein and Horioka puzzle. The evidence suggests that the saving retentioncoefficients declined substantially in most of the countries after an endogenous break date coinciding with a major exchange rateregime change with the 1997–1998 crisis. Exchange rate flexibility appears to be enhancing financial integration. The results areconsistent with an “investment slump” explanation rather than the “saving glut” postulation.c© 2008 Elsevier B.V. All rights reserved.

PACS: 89.65.Gh

Keywords: Capital mobility; Current account imbalance; East Asia; Exchange rate regimes; Parameter stability; Structural breaks

1. Introduction

The development of substantial domestic saving-investment gaps has been at the centre of debate in internationalmacroeconomics with the recent experience of large current account (CA) imbalances of a number of countriesincluding the US. The recent CA surpluses of emerging East Asian countries have drawn special attention as theyhave been a major source of finance of the growing US CA deficits. According to the “saving glut” explanation ofthe global imbalance expounded by Bernanke [1], excess domestic saving over investment (CA surplus) especially inEast Asia indeed has caused the sustainability of the US CA deficits.

The “saving glut” and the global imbalances are closely related to the Feldstein and Horioka puzzle (Feldstein andHorioka [2], hereafter FH) that the saving (S)-investment (I) relationship has been persistently strong in spite of policyregime changes towards flexible exchange rates and capital mobility. Coakley et al. [3] provides a recent survey of theFH literature. FH consider the following equation:

It = γ0 + γ1St + ut . (1)

In (1), u is a disturbance term and γ1 is the ‘saving retention coefficient’. In a financial autarky, investment can only befinanced by domestic saving causing γ1 = 1. With capital mobility, investment can be financed by the worldwide pool

∗ Corresponding author. Tel.: +90 312 210 3044; fax: +90 312 210 7964.E-mail address: ozmen@metu.edu.tr (E. Ozmen).

0378-4371/$ - see front matter c© 2008 Elsevier B.V. All rights reserved.doi:10.1016/j.physa.2008.01.016

2562 S. Kaya-Bahce, E. Ozmen / Physica A 387 (2008) 2561–2564

of saving and domestic saving can be a source of overseas investment, thus the value of γ1 decreases. Consequently,γ1 is an indicator of the extent of capital mobility in the FH sense. A γ1 of zero (one) indicates perfect (the absence ofany) capital mobility.

The growing global imbalances may be interpreted as reflecting an increase in international financial integrationin the FH sense. However, such an interpretation does not make a clear distinction between CA deficits and surplusespersisting under different exchange rate (ER) regimes. Furthermore, CA imbalances, γ1 and thus the FH puzzle maynot be invariant to the prevailing ER regime (Sarno and Taylor [4], Ozmen and Parmaksız [5]). Fixed ER regimesare often associated with CA deficits due to a real appreciation of domestic currency. A (credible) fixed ER regime,according to Razin and Rubinstein [6] p. 122, “provides a less risky environment for investors and the country may beable to attract more external funds to complement more domestically funded investment”. Consequently, γ1 may beexpected to be higher under a fixed ER regime. ER flexibility, on the other hand, can act as a shock absorber [7] or adisciplining device on CA deficits [8] by allowing exchange rates to adjust to CA disequilibrium.

The experience of most of the East Asian countries with CA deficits under fixed ER regimes until 1997–1998 andsurpluses thereafter under ER flexibility may be consistent with the argument that the FH puzzle is not invariant toER regime changes. The following section investigates this issue empirically. To this end, we first test whether thereis no endogenous break in γ1 for the emerging East Asian countries. We then proceed with the investigation of thepostulation that the shifts in γ1 can be explained by ER regimes. The results are discussed also in the context of the“saving glut” arguments.

2. Empirical results

We start with the estimation of the conventional FH equation:

invt = γ0 + γ1savt + ut (2)

where sav = gross national savings (% of GDP) and inv = gross capital formation (% of GDP). The annual data(1970–2005) for sav and inv are from the World Bank World Development Indicators. Table 1 reports the OLSestimates of (2) for each of the emerging East Asian countries — Hong Kong, Indonesia, S. Korea, Malaysia,Philippines, Singapore and Thailand. The results suggest that all the countries, except Thailand and Indonesia, canbe interpreted as financially open in the FH sense. The statistical insignificance of γ1 for Malaysia, Philippines andHong Kong supports the view that they are financially integrated. However, a negative (Singapore) or an insignificantγ1 may be consistent also for a country where domestic savings tend to finance investments abroad due to the lack ofsufficient domestic financial intermediation and/or a declining equity home bias.

All these results, however, should be interpreted with a caution as γ1 may not remain stable during the period.To test whether there is no endogenous break in γ1, we consider Andrews–Quandt test for parameter stability [9,10].Table 1 reports the values of the Andrews–Quandt SupF statistics which are the maximum of the individual ChowWald-F breakpoint tests estimated sequentially between two dates τ1 = ηT + 1 and τ2 = T − ηT − 1, where T is thesample size and η(0.10) is the trimming parameter. The table also reports the Hansen [11] p-values for SupF and thecorresponding estimated break date (TB) for γ1. The results suggest that the Andrews–Quandt test is maximum andsignificant at 1998 for Indonesia, S. Korea, Malaysia and Thailand and at 1999 for Philippines, which are indeed thecountries severely hit by the 1997–1998 financial crisis. For Singapore the Chow Wald-F test for γ1 is maximum at2001 albeit a break occurring at 1998 also appears to be data-acceptable with the test yielding 63.2 (p = 0.00). ForHong Kong, γ1 is found to be stable during the period.

To estimate the shift in the saving retention coefficient γ1, we consider the following equation:

invt = γ0 + γ1savt + α(D∗t savt ) + ut (3)

where D = 1(t > TB), 1(.) is the indicator function. The results presented by Table 1 suggest a significant downwardshift in γ1 after TB for all the countries. According to the values of γ1 for the pre-break sample, all the crisis-hitcountries, except S. Korea to a certain extent, can be classified as financial autarky in the FH sense. However, thisinterpretation may not be compatible with the fact that these countries often experienced excessive capital inflowsunder fixed ER regimes before the crisis [12]. A high domestic S-I correlation may be consistent also with a policy ofCA targeting in a financially open economy with fixed exchange rates [5]. Consequently, the high values of γ1 may

S. Kaya-Bahce, E. Ozmen / Physica A 387 (2008) 2561–2564 2563

Table 1Conventional and augmented FH equations

γ0 γ1 α β R2 DW SupF TB

Indonesia 3.02 (0.80) 0.77 (6.00) 0.51 0.81 16.6 [0.00] 19984.89 (1.55) 0.74 (6.99) −0.18 (−4.07) 0.68 0.938.05 (2.65) 0.89 (8.87) −0.04 (−5.08) 0.73 1.25

S. Korea 18.26 (6.46) 0.43 (4.66) 0.39 0.91 14.3 [0.00] 199816.77 (6.91) 0.51 (6.31) −0.14 (−3.78) 0.57 1.0016.78 (7.35) 0.62 (7.31) −0.02 (−4.50) 0.62 0.95

Malaysia 20.35 (3.77) 0.22 (1.46) 0.06 0.37 40.6 [0.00] 19981.00 (0.21) 0.90 (6.05) −0.39 (−6.37) 0.58 0.7619.52 (4.13) 0.07 (0.47) 0.03 (3.38) 0.30 0.91

Philippines 19.75 (5.05) 0.13 (0.81) 0.02 0.29 85.1 [0.00] 19991.96 (0.69) 1.00 (7.86) −0.51 (−9.22) 0.73 0.9511.46 (2.37) 1.00 (2.72) −0.05 (−2.60) 0.19 0.46

Thailand 7.56 (1.45) 0.80 (4.42) 0.35 0.44 69.7 [0.00] 19981.84 (0.56) 1.09 (9.75) −0.36 (−8.35) 0.79 0.841.54 (0.39) 1.21 (7.61) −0.12 (−5.26) 0.65 1.40

Hong Kong 16.65 (1.96) 0.31 (1.19) 0.04 0.38 7.55 [0.12] 2002Singapore 49.66 (8.92) −0.32 (−2.40) 0.15 0.25 64.7 [0.00] 2001

47.18 (14.3) −0.20 (−2.48) −0.40 (−8.04) 0.71 0.7339.93 (8.46) 0.78 (2.47) −0.07 (−4.92) 0.51 0.53

Notes: t-ratios in parentheses. Bold faces denote significance at the 5% level. Hansen [11] p-values for the SupF tests are given in brackets.

be interpreted supporting endogenous policy actions towards CA targeting, rather than capital immobility under fixedER regimes.

All the countries, except Hong Kong and Malaysia, adopted more flexible ER regimes after the crisis. Hong Kongretained its managed ER regime during the period. Malaysia moved to a more rigid ER regime accompanied withrestrictions on capital flows. The saving retention coefficient may not be invariant to ER regime changes. To investigatethis issue, we estimate the following equation

invt = γ0 + γ1savt + β(ERR∗t savt ) + ut (4)

where ERR = de facto exchange rate regime classification by Reinhart and Rogoff [13]. Reinhart and Rogoff [13]classify de facto regimes on a 1–14 scale, with higher values denoting more flexible exchange rates. A similarclassification is provided by the IMF’s “Classification of Exchange Rate Regimes and Monetary Policy Framework”home page. The ERR data are from [13] and the IMF home page for the 1970–2001 and 2002–2005 samples,respectively.

The results reported by Table 1 strongly suggest that the interaction variable (ERR∗t savt ) is significant for every

country which experienced an ERR change. The saving retention coefficient significantly declines with ERR flexibilityfor all of the countries except Malaysia. The positive interaction variable coefficient for Malaysia may not be surprisingas Malaysia’s adoption of a more rigid ERR after the crisis was accompanied with capital controls. ERR flexibilityappears to be enhancing financial integration for all the countries except Malaysia.1

3. Concluding notes

This paper investigated whether the recent experience of the emerging East Asian countries is consistent withthe “saving glut” hypothesis and the FH puzzle. The evidence suggests a substantial decline in the saving retentioncoefficients after the estimated endogenous break date coinciding with a major ER regime change with the 1997–1998crisis. Consistent with the view that a CA targeting policy is most plausible under managed exchange rates, the

1 As discussed by [14] and elsewhere, the benefits of ERR flexibility increase with the degree of capital mobility. Capital controls, on the otherhand, may lead to a decrease in the saving retention coefficient as they tend to reduce the supply of capital, raise the cost of financing and increasethe financial constraints for firms [15]. The Malaysian evidence after the crisis may thus better be interpreted as reflecting the impact of capitalcontrols under a fixed ERR rather than ERR flexibility per se.

2564 S. Kaya-Bahce, E. Ozmen / Physica A 387 (2008) 2561–2564

investment-saving relationship is found to be high under fixed ER regimes for all of the countries except Hong Kong,Singapore and Malaysia. Hong Kong and Singapore appear to be financially integrated through the period. In all thecountries, except Malaysia which imposed capital control after the crisis, γ1 declines with ERR flexibility. In the FHcontext, this supports that flexible ER regimes enhance financial integration.

The “saving glut” view contends that it is the excess saving over stable investment causing CA surpluses in theemerging East Asian countries. This may not be compatible with the recent East Asian data as domestic savingremained relatively stable whilst investment declined sharply in almost all of the crisis-hit countries [12]. Thedownward shift in the saving retention coefficient after the estimated endogenous break date provides an importantinsight for the explanation of the recent CA surpluses of the East Asian countries. The CA imbalances appear toreflect basically investment dynamics. Consequently, “an investment slump” [16,17] under relatively stable savingrates rather than the “saving glut” may be a better interpretation of the recent East Asian experience.

References

[1] B.S. Bernanke, The global saving glut and the US current account deficit, in: Remarks at the Homer Jones Lecture, St. Louis, Missouri, April14, 2005.

[2] M.S. Feldstein, C.Y. Horioka, Domestic saving and international capital flows, Economic Journal 90 (1980) 314–329.[3] J. Coakley, A.M. Fuertes, F. Spagnolo, Is the Feldstein–Horioka puzzle history? The Manchester School 72 (5) (2004) 569–590.[4] L. Sarno, M.P. Taylor, Exchange controls, international capital flows and saving-investment correlations in the UK: An empirical investigation,

Weltwirtschaftliches Archiv 134 (1) (1998) 69–98.[5] E. Ozmen, O.K. Parmaksız, Policy regime change and the Feldstein–Horioka puzzle: The UK evidence, Journal of Policy Modeling 25 (2003)

137–149.[6] A. Razin, Y. Rubinstein, Evaluation of currency regimes: The unique role of sudden stops, Economic Policy (January) (2006) 119–152.[7] S. Edwards, Thirty years of current account imbalances, current account reversals, and sudden stops, IMF Staff Papers 51 (2004) 1–49.[8] E. Ozmen, Macroeconomic and institutional determinants of current account deficits, Applied Economics Letters 12 (2005) 557–560.[9] D.W.K. Andrews, W. Ploberger, Optimal tests when a nuisance parameter is present only under the alternative, Econometrica 62 (1994)

1383–1414.[10] D.W.K. Andrews, Tests for parameter instability and structural change with unknown change point, Econometrica 61 (1983) 821–856.[11] B. Hansen, Approximate asymptotic P-values for structural change tests, Journal of Business and Economic Statistics 15 (1997) 60–67.[12] G. Corsetti, P. Pesenti, N. Roubini, What caused the Asian currency and financial crisis? Japan and the World Economy 11 (1999) 305–373.[13] C.M. Reinhart, K.S. Rogoff, The modern history of exchange rate arrangements: A reinterpretation, Quarterly Journal of Economics CXIX

(1) (2004) 1–48.[14] S. Edwards, Capital controls, capital flow contractions, and macroeconomic vulnerability 26 (5) (2007) 814–840.[15] K.J. Forbes, The microeconomic evidence on capital controls: No free lunch, NBER working paper no. 11372.[16] J. Felipe, K. Kintanar, J.A. Lim, Asia’s current account surplus: Savings glut or investment drought? Asian Development Review 23 (1) (2006)

16–54.[17] M.D. Chinn, H. Ito, Current account balances, financial development and institutions: Assaying the world ‘savings glut’, Journal of

International Money and Finance 26 (2007) 546–569.