GLOBAL IMBALANCES: A SAVING AND CHAPTER II INVESTMENT...

Global saving and investment rates havefallen and current account imbalanceshave widened to unprecedented levels,yet real long-term interest rates remain

low in most countries. How did the global econ-omy arrive at this position? Some have arguedthat the catalyst is the substantial changes thathave taken place in Asia, where saving has risenbut investment has collapsed since the late1990s. According to this view, the swing in thesaving-investment gap—from deficit to largesurplus—in emerging Asia has resulted in anexcess global supply of saving (a global saving“glut”) that has been channeled to the UnitedStates to finance its large current account imbal-ance (Bernanke, 2005). At the same time, thiswould explain the low level of long-term realinterest rates, which is needed to equilibratedesired saving and planned investment on aglobal basis. Others have argued that the sharpdrop in national saving in the United States—reflecting the deterioration in the fiscal positionand the increase in housing wealth—and therecent rebound in investment are at the root ofcurrent account imbalances (see, for example,Roubini and Setser, 2005). Thus, according tothese observers, current global imbalances aremainly the result of policy decisions—both fiscaland monetary—in the United States. By itself,however, this would not explain the low level ofreal interest rates, as a higher demand for netsaving from the United States would lead (every-thing else equal) to higher, not lower, globalinterest rates.

This chapter examines the main factors thathave driven the recent evolution of saving andinvestment across the globe, to shed light onboth existing global imbalances and low realinterest rates. The analysis covers 46 countries(21 industrial and 25 emerging marketeconomies; 5 of which are oil producing) thataccount for over 90 percent of world GDP.1

Specifically, the chapter addresses the followingquestions.• What factors account for recent movements in

saving and investment in industrial, emergingmarket, and oil-producing countries? Arethese changes due to country-specific develop-ments, or do they reflect broader global andregional trends?

• Looking forward, what policies can helpchange existing saving-investment gaps, andlead to a reduction in global imbalances?An important theme running through the

chapter is that the current constellation ofcurrent account imbalances and low realinterest rates is the result of important changesin saving and investment patterns across theworld. In particular, the chapter finds thatunusually low investment rates across the globeare a contributing factor to low real long-terminterest rates. In addition, the chapter alsofinds that the current pattern of externalimbalances largely reflects a series of diverseand unrelated regional developments. As aresult, the unwinding of these imbalances willrequire economic responses across a largenumber of countries.

91

CHAPTER IIGLOBAL IMBALANCES: A SAVING ANDINVESTMENT PERSPECTIVE

The main authors of this chapter are Marco Terrones and Roberto Cardarelli, with support from Enrique Mendoza andChris Otrok. Stephanie Denis provided research assistance.

1An important preliminary consideration is that any analysis of saving and investment is affected by concerns about thequality of the data (Schmidt-Hebbel and Servén, 1999). Saving, for instance, is usually calculated as the difference betweenincome and consumption. Reflecting this, the measures of saving normally do not adjust for changes in net worth due toasset price movements, including house prices. Similarly, there is growing consensus that the measures of investmentshould include expenditure on research and development and education, as well as households’ spending on durablegoods.

Global Saving and Investment:The Current State of Play

The world economy is experiencing changes inboth saving and investment behavior that are hav-ing implications for the configuration of currentaccount imbalances and the level of real interestrates. Global saving and investment (as a percentof GDP) fell sharply in the decade following thefirst oil price shock in the early 1970s, but werethen relatively stable until the late 1990s.2 Morerecently, however, they again declined, hitting his-toric lows in 2002 before modestly recoveringover the past two years (Figure 2.1). These globaltrends mainly reflect developments in the indus-trial countries, where both saving and investmenthave been trending downward since the 1970s. Incontrast, saving in the emerging market and oil-producing economies has risen over this period,while investment, after increasing substantially upto the time of the Asian financial crisis, has sincefallen and remains below the levels of the mid-1990s. As a result of these trends, the industrialcountry share of global saving and investment hasdropped from about 85 percent in 1970 to 70percent at present.

Focusing in more detail on developmentssince 1997—the period when substantial globalcurrent account imbalances have emerged—about two-thirds of the fall in saving rates in theindustrial countries has been due to a reductionin private saving, with falling household savingonly partly offset by higher corporate saving.3

CHAPTER II GLOBAL IMBALANCES: A SAVING AND INVESTMENT PERSPECTIVE

92

1970 75 80 85 90 95 200020

22

24

26

28

1970 75 80 85 90 95 200014

16

18

20

22

24

-1.5

-1.0

-0.5

0.0

0.5

1.0

1970 75 80 85 90 95 20002

3

4

5

6

7

-1.5

-1.0

-0.5

0.0

0.5

1.0

Figure 2.1. Global Saving, Investment, and Current Accounts(Percent of world GDP)

Global saving and investment have been trending downward since the early 1970s. They reached historic lows in 2002, and have recovered modestly since then.

All Countries

04

Industrial Countries

04

Sources: OECD Analytical Database; World Bank, World Development Indicators; and IMF staff calculations. Includes Norway.1

Emerging Market and Oil-Producing Economies1

04

Saving Investment Current account (left scale)

2In this chapter, saving and investment ratios are calcu-lated as U.S. dollar saving and investment divided by U.S.dollar GDP at current exchange rates. Global saving andinvestment should in principle be equal because of therequirement that current account balances sum to zeroacross the globe. In practice, however, this is not the casebecause of statistical problems that give rise to the so-called global current account discrepancy. In addition, inthis study, the sample does not cover the whole world.

3A full offset between corporate and household savingcould be expected as households are the ultimate ownersof corporations. However, the extent to which households“pierce the corporate veil” has not been fully assessed.Poterba (1987) finds that changes in corporate saving areonly partly offset (between 25 to 50 percent) by changesin household saving in the United States.

Indeed, corporate saving has now overtakenhousehold saving as the main source of privatesector saving in industrial countries.4 In contrast,after dipping in the immediate aftermath of theAsian financial crisis, the saving rate in emergingmarket and oil-producing economies hasresumed its secular increase, reaching a record-high level in 2004. A substantial part of thisincrease reflects higher public saving.

These aggregate developments, however, maskconsiderable variation between the countriesand regions. The recent deterioration in savingrates in industrial countries has been particularlymarked in the United States, Japan, and, to alesser extent, the euro area countries (Figure2.2). In Japan and the euro area, this has contin-ued the decline in saving that began in the early1990s, driven by a large drop in public saving inthe former and lower private (household) savingin the latter (Figure 2.3). In the United States,saving has declined sharply since the late1990s—accelerating the secular downwardtrend—driven initially by a drop in private sav-ing and since 2000 by the swing in the budgetfrom surplus to substantial deficit. Furthermore,in both Japan and the United States, corporatesaving has risen substantially, offsetting lowerhousehold saving. In the other industrial coun-tries, saving has been flat in recent years, afterrebounding from the drop in the early 1990s.

Saving rates in emerging market and oil-producing countries have caught up with andlargely overtaken those of industrial countries(when measured against their own GDP; seeFigure 2.4). Particularly remarkable has beenthe very sharp increase in saving in China, espe-cially since 2000 (see Box 2.1). Elsewhere inAsia, saving rates remain high, although theyhave declined since the early 1990s. In otheremerging market countries, saving has risen

GLOBAL SAVING AND INVESTMENT: THE CURRENT STATE OF PLAY

93

1970 75 80 85 90 95 200014

16

18

20

22

24

26

28

-4

-3

-2

-1

0

1

2

1970 75 80 85 90 95 200018

20

22

24

26

28

30

-4

-3

-2

-1

0

1

2 Other IndustrialEuro Area

1970 75 80 85 90 95 200020

25

30

35

40

45

-2

-1

0

1

2

3

4

5

6Japan

1970 75 80 85 90 95 200012

14

16

18

20

22

24

-8

-7

-6

-5

-4

-3

-2

-1

0

1

2United States

The recent sharp drop in saving in the United States and the decline in investment in Japan and the euro area countries have contributed to recent global current account imbalances.

Figure 2.2. Saving and Investment in the Industrial Countries (Percent of each subregion's GDP)

Sources: OECD Analytical Database; World Bank, World Development Indicators; and IMF staff calculations.

Saving Investment Current account (right scale)

4This trend has become more accentuated since 2000as corporations in several industrial countries have soughtto strengthen their balance sheets. In contrast with thepast, however, the financial sector has contributed sub-stantially to the recent increase in corporate saving(JPMorgan Chase & Co., 2005).

sharply in recent years, driven by higher publicsaving in Latin America. Lastly, saving in oil-producing countries has also increased consider-ably, owing to the impact of higher oil prices onpublic saving.

Investment rates have fallen across virtuallyall industrial country regions, although thishas been most noticeable in Japan and theeuro area countries, where they reachedhistoric lows in 2002 (see Figure 2.2). Giventhat investment in these regions started outhigher than elsewhere, this has underpinneda convergence of investment rates acrossindustrial country regions to about 20 percentof GDP in 2004 (although not a convergencein growth rates). Investment rates in theUnited States are broadly unchanged fromtheir levels in 1997, although they remainbelow the peak in 2000. Of course, the declinein the nominal investment ratios over timepartly reflects the fact that capital goods havebecome relatively less expensive—mainly owingto the extensive process of information tech-nology (IT) capital deepening and productivitygrowth in the capital good–producing sectors.5

In volume terms, the fall in average investmentrates in industrial countries has been moremodest.

Investment rates differ substantially acrossemerging market economies (see Figure 2.4).Investment in China has surged since 2000,and stood at 45 percent of GDP in 2004. Withthe exception of China and a handful of othercountries, however, investment rates havefallen in emerging market economies sincethe Asian financial crisis. Indeed, investmentrates in east Asia have declined by more than10 percentage points of GDP since their peakin the mid-1990s and have not rebounded


94

1970 75 80 85 90 95 200010

20

30

40

50

1970 75 80 85 90 95 2000-8

-4

0

4

8

12

16

1970 75 80 85 90 95 20000

5

10

15

20

25

1970 75 80 85 90 95 20004

8

12

16

20

24

1970 75 80 85 90 95 2000-8

-4

0

4

8

12

1970 75 80 85 90 95 200010

15

20

25

30

35

40

Figure 2.3. Saving Trends Across Regions (Percent of each subregion's GDP)

Despite the increase in corporate saving, private saving in most industrial country regions has fallen over the past decade. Public saving in most industrial country regions has declined recently, but it has risen in China and oil-producing countries.


Other industrial

United States

Euro area

Japan

Emerging Market and Oil-Producing Economies

East Asia

Other emerging marketsOil producers

China

1


Private Saving Public Saving

Emerging Market and Oil-Producing Economies1

Private Saving Public Saving

Sources: OECD Analytical Database; World Bank, World Development Indicators; and IMF staff calculations. Includes Norway.

Data unavailable in 2004 for all regions except the United States.

12

Corporate Saving2Household Saving 2

5The shift toward IT capital has also increased the aver-age depreciation rate, an effect that works in the direc-tion of increasing the amount of gross capital formationconsistent with a constant, desired, level of net invest-ment. Indeed, several authors, including Tevlin andWhelan (2003), attribute the U.S. investment boom in thelate 1990s mainly to the rise in capital depreciation.

despite a sharp increase in public investment.Investment in oil-producing countries has alsoremained low despite the recent strength of oilprices.

As is evident from Figures 2.2–2.4, saving andinvestment in the large industrial countries havefollowed broadly similar trends in recent years.There appears to be much more divergence inbehavior, however, among emerging marketcountries. The correlations reported in Table2.1 confirm this. Saving and investment ratiosare strongly correlated in industrial countryregions, although this correlation has declinedin recent years, while there is a lower—but ris-ing—correlation across emerging market coun-tries. There is little correlation between savingin industrial and emerging market countries,although the degree of co-movement of invest-ment rates across all regions has increased overthe past seven years, possibly reflecting theglobal nature of the IT-related productivityshock. The correlation between saving andinvestment rates within each region has beensignificantly positive on average over the periodconsidered, although the strength of this asso-ciation, originally documented by Feldsteinand Horioka (1980), has fallen over time (froman average of 0.6 in 1970–96 to 0.4 during1997–2004).


95

1970 75 80 85 90 95 200014

16

18

20

22

24

26

-6

-5

-4

-3

-2

-1

0

1

2Other Emerging Markets

1970 75 80 85 90 95 200012

16

20

24

28

32

-9

-6

-3

0

3

6

9Oil Producers

1970 75 80 85 90 95 200025

30

35

40

45

50

55

-6

-4

-2

0

2

4

6China

1970 75 80 85 90 95 200016

20

24

28

32

36

-9

-6

-3

0

3

6

9East Asia

The sharp drop in investment in east Asia and increase in saving in the oil-producing countries are two other important developments behind the recent global current account imbalances.

Figure 2.4. Saving and Investment in the Emerging Market and Oil-Producing Economies (Percent of each subregion's GDP)

Sources: OECD Analytical Database; World Bank, World Development Indicators; and IMF staff calculations.

Saving Investment Current account (right scale)

Table 2.1. Average Correlations of Saving andInvestment Ratios

1970–2004 1970–96 1997–2004

Saving_______________________________Across all regions 0.15 0.18 0.22

Between industrial regions 0.58 0.68 0.48Between emerging

market regions 0.03 0.04 0.27Between industrial and

emerging market regions –0.16 –0.19 –0.08

Investment_______________________________Across all regions 0.27 0.22 0.36

Between industrial regions 0.68 0.69 0.53Between emerging

market regions 0.12 0.11 0.30Between industrial and

emerging market regions . . . –0.14 0.24

Source: IMF staff estimates.


96

After averaging some 40 percent of GDPduring the 1990s, China’s gross domesticsaving rate has increased sharply to close to50 percent of GDP over the past five years (seethe figure). This has been accompanied by asmaller—but still substantial—rise in gross capi-tal formation to about 45 percent of GDP, alongwith a widening external current account sur-plus. The present levels of saving and invest-ment are very high, both in terms of China’sown historical experience, and by comparisonwith experience in other advanced or develop-ing countries.1

Analysis of both saving and investment inChina is hampered by a variety of data lim-itations, and the sectoral breakdown pre-sented in the figure should be taken only asbeing broadly indicative of underlying trends.With that caveat, several interesting pointsstand out.• Corporate saving has risen sharply since

2000. Profitability has increased substantiallyin both the state-owned enterprise and—even more—the non-state-owned enterprisesectors, driven by a combination of strongeconomic growth; low interest rates; fallingunit labor costs; reductions in employeebenefits (see below); and—in resourcesectors—rising commodity prices. More gen-erally, the high level of corporate saving mayalso partly reflect the still-underdevelopedfinancial sector in China—including domes-tic bond and equity markets—as well as thelimited access of non-state-owned enterprisesto financial markets (forcing them to financeinvestment primarily through retainedearnings).

• In contrast to recent experiences in manyother countries—see the main text—the rise

in corporate saving has been accompanied bya surge in investment, in both state- and non-state-owned firms.2 Sectorally, this has beenconcentrated in infrastructure; in manu-facturing, especially aluminum, steel, autos,and cement; and in real estate. While Chinahas clear infrastructural needs, particularly inthe underdeveloped western and centralprovinces, and demand growth in many ofthese sectors has been strong, this has raisedconcerns about potential overcapacity, and—if

Box 2.1. Saving and Investment in China

Sources: Modigliani and Cao (2004); and IMF staff calculations.

1982 86 90 94 98 200230

35

40

45

50Gross Saving and Investment

Gross domestic saving

Gross capital formation

1990 94 98 200220

25

30

35

40

1990 94 98 2002

4

8

12

16

20

24

28 Gross CapitalFormation by Enterprise Sector

Gross Saving by Sector

Household

Enterprise

Government

China's Saving and Investment by Sector(Percent of GDP)

Note: The main authors of this box are MarcosChamon and Akito Matsumoto.

1Such comparisons need to be treated with cautiongiven the possibility that GDP is underestimated inChina (Barnett and Brooks, 2005).

2Including majority state-owned shareholding firms,state-controlled firms still accounted for almost 60 per-cent of urban fixed asset investment in 2004 (Barnettand Brooks, 2005).


97

investment is not efficient—a potential furtherbuildup of nonperforming loans. Such con-cerns are underlined by recent research.Manufacturing investment in China is stronglycorrelated with corporate liquidity, suggestingthat expansion considerations, not profitabil-ity, may be driving investment (Barnett andBrooks, 2005). Moreover, state-owned enter-prise profits do not seem to be a factor indetermining lending by state-owned banks(Podpiera, 2005), and interprovincial bankflows appear to favor provinces with a highshare of state-owned enterprises in their out-put at the expense of high-growth provinces(Boyreau-Debray and Wei, 2005), although lessthan was previously the case (Aitken, 2005).

• Household saving has remained broadlyconstant in recent years—following somedecline in the late 1990s—but, at close to25 percent of disposable income, remainsvery high (despite low real interest rates onbank deposits, the dominant vehicle forhousehold savings). This high level of house-hold saving appears in part to reflect demo-graphic developments, including the gradualaging of the population. Losses or uncer-tainty in the future provision of housing,health, education, and pension benefits—which were traditionally provided by state-owned enterprises—may also result in highlevels of precautionary saving. Households’limited access to credit for the purchase ofdurable goods and housing likely furthercontributes to their saving motives.3

• Government saving has also increasedmarkedly, driven by higher revenues. Thisincrease has been used partly to strengthenthe fiscal position, but mainly to finance asubstantial increase in investment, particularlyby local governments.Looking forward, with the investment share

very high, a key medium-term challenge is to

increase the efficiency of investment, accompa-nied by a welfare-enhancing shift in the compo-sition of domestic demand from investmenttoward consumption. To some extent, this willoccur naturally. For example, strong corporateprofits will—with some lag—feed through intowages; slowing growth in commodity prices willgradually lower profit rates in the resource sec-tor; and—over the longer term—populationaging will gradually reduce the household savingrate. But economic policies also have an impor-tant role to play.• First, on the macroeconomic side, recent

administrative measures to slow investmentneed to be accompanied by a tightening ofmonetary policy—both by reducing excessliquidity and by further raising lendingrates—which would be facilitated by the scopefor greater exchange rate flexibility affordedby recent reforms.

• Second, continued efforts to improve gover-nance of state-owned enterprises and maketheir behavior more market oriented (andaccountable to their owners) will be key.One immediate step could be to requireprofitable state-owned enterprises to paydividends, which would reduce corporateincentives for excessive investment. It wouldalso provide the government with resourcesto help fund existing pension and social lia-bilities, thereby reducing incentives for pre-cautionary saving by households (see Box 3.5in the September 2004 World EconomicOutlook).

• Third, financial sector reforms—both tostrengthen the banking sector and to furtherdevelop bond and equity markets—are criticalto improve the intermediation of China’slarge pool of saving and to direct it to themost productive investments (Prasad andRajan, 2005). This would provide alternativevehicles for saving and additional sources offinancing for firms and households, andwould have the added benefit of promotingbanking reforms by exposing state banks todomestic competition.

3These effects are quantified in Chamon and Prasad(2005).

Consistent with the decline in the cross-country correlation between saving and invest-ment, external current account imbalances(relative to domestic GDP) have, on average,increased, and the dispersion across industrial—and, to a lesser extent, emerging market—countries has widened (see Chapter III of theApril 2005 World Economic Outlook). In particular,external imbalances between some major eco-nomic areas—notably the United States, Asia,and oil producers—are at record levels. Interest-ingly, the current constellation of externalimbalances is very different from that in themid-1980s—the last period of large globalimbalances. At that time, the external deficit ofthe United States peaked at slightly above 3 per-cent of GDP in 1987, and was largely matchedby surpluses in a relatively small number ofcountries (particularly Japan and the euro areacountries). In contrast, current account imbal-ances are now dispersed across a much widergroup of countries and involve many emergingmarket and oil-producing economies. Between1997 and 2004, about two-thirds of the increasein the U.S. current account deficit has been bal-anced by higher external surpluses in emergingmarket and oil-producing countries, with therest matched by larger surpluses in industrialcountries (mainly Japan).

The transformation of emerging markets fromnet importers to net exporters of capital inrecent years is difficult to reconcile with the pre-dictions of economic theory (Lucas, 1990), orwith the historical pattern of international capi-tal flows, particularly in the period before WorldWar I when capital flowed from the core coun-tries of western Europe to the new settlements.While some have argued that these develop-ments are the result of policy decisions in

emerging markets—mainly reflecting the desireto accumulate foreign exchange reserves thatcould be used as a buffer in the event of turbu-lence in financial markets (Bernanke, 2005)—they could also reflect the lack of profitableinvestment opportunities in emerging marketeconomies vis-à-vis industrial countries (seeBox 2.2).

What Drives Saving, Investment, and theCurrent Account?

What are the main factors that have been driv-ing recent saving and investment behavior acrossthe globe? This section uses two approaches—econometric analysis and a dynamic factormodel—to investigate this issue.

Econometric Results

Building on a burgeoning literature on thedeterminants of saving and investment, separatedynamic panel models for saving and investmentwere estimated using data for 46 industrial,emerging market, and oil-producing countries(and separately for the industrial and emergingmarket subsamples) over 1972–2004.6 The keyresults of this analysis—which are shown inTable 2.2 and described in more detail inAppendix 2.1—are described below.7

For saving, the estimated equations fit thedata well and indicate the following.• Higher output growth boosts saving. A sustained

1 percentage point increase in per capita out-put growth in industrial countries would overtime lead to an almost 1 percent of GDPincrease in the national saving rate.8 Foremerging market economies, the estimatedimpact is smaller, at !/2 percent of GDP.


98

6See, for instance, Masson, Bayoumi, and Samiei (1995); Edwards (1995); Haque, Pesaran, and Sharma (1999); andLoayza, Schmidt-Hebbel, and Servén (2000), among others.

7These equations have been estimated using Generalized Method of Moments, which controls for the potential endo-geneity of the explanatory variables (see, for instance, Arellano and Bond, 1991). Indeed, the Hansen tests for all reportedregressions suggest that the lagged values of the variables are valid instruments.

8This can be calculated from Table 2.2, column 2, as the ratio of the coefficient on per capita output growth (0.28) overone minus the coefficient of lagged saving (0.7); that is, 0.28/(1 – 0.7) ≈ 1.

• Fiscal consolidation is associated with increasedsaving, as higher public saving is only par-tially offset by adjustments in private savingbehavior (i.e., Ricardian equivalence doesnot hold, consistent with the findings ofother studies). A 1 percent of GDP increasein public saving in industrial countrieswould lead over time to a !/2 percent of GDPincrease in national saving.9 In emergingmarkets, the impact is larger, raising nationalsaving by 0.85 percent of GDP, reflecting inpart the less developed financial markets inthese countries that make it more difficult forhouseholds to smooth consumption overtime.

• Increases in private sector credit are associated witha reduction in saving. This may reflect the factthat households face borrowing constraintsthat are normally relaxed by the process ofderegulation and innovation in financial mar-kets. Private credit is also likely capturingwealth effects associated with the sharpincrease in asset prices, particularly houseprices, which are believed to have driven the

reduction in household saving in a number ofindustrial countries over the last decade, butespecially in the United States (see Faulkner-MacDonagh and Mühleisen, 2004). Theregression results suggest that a 10 percentof GDP increase in credit in industrial coun-tries would lead over time to a reduction of!/2 percent of GDP in the saving rate. Theimpact is again larger in emerging markets, at0.9 percent of GDP, possibly because of thelower levels of financial intermediation andhigher dependency on bank credit in theseeconomies.

• As populations age, this puts downward pressure onsaving. Given that people tend to dissave dur-ing their retirement years, an increase in theelderly dependency ratio—the ratio of thoseaged over 65 to the working age population—should reduce saving (see Box 2.3). This isconfirmed by the regression results, whichsuggest that an increase in the elderly depend-ency ratio of 1 percentage point in the indus-trial countries would over time reduce savingby about 1!/2 percent of GDP. Because of the

WHAT DRIVES SAVING, INVESTMENT, AND THE CURRENT ACCOUNT?

99

Table 2.2. Global Saving and Investment: Panel Regression1

Saving (percent of GDP) Investment (percent of GDP)________________________________ ________________________________Emerging Emerging

Industrial market Industrial marketAll countries economies All countries economies

Lag-dependent variablePercent of GDP saving 0.62 0.70 0.71 . . . . . . . . .Investment . . . . . . . . . 0.76 0.80 0.80

Main determinantsReal per capita GDP growth 0.17 0.28 0.13 0.26 0.33 0.23Real interest rate2 0.01 –0.07 0.01 . . . –0.08 . . .Credit (percent of GDP) –3.47 –1.53 –2.51 –1.36 0.81 –1.64Change in credit (annual percent of GDP) –2.17 –0.94 –7.39 0.08 0.02 0.12Elderly dependency ratio –0.44 –0.43 –0.66 –0.09 –0.04 –0.19Public saving (percent of GDP) 0.27 0.15 0.24 . . . . . . . . .Terms-of-trade growth 0.08 0.06 0.08 . . . . . . . . .

Source: IMF staff calculations.Note: Bold-faced values are statistically significant at the 5 percent level. Values in italics are statistically significant at the 10 percent level.1The estimated effects reported in the text are the long-term effects calculated as the ratio of the estimated coefficients over one minus the

coefficient of the lagged-dependent variable.2In the investment equation, this is the cost of capital.

9De Mello, Kongsrud, and Price (2004) find that the Ricardian offset for 21 OECD countries ranges from !/3 to !/2. Usinga different approach, Gale and Orszag (2004) find a similar result for the United States.


100

In recent years, international capital hasflowed from the developing world to a numberof industrial countries, particularly the UnitedStates. This is hard to reconcile with standardeconomic theory, which holds that—givenrestrictions on mobility of labor—industrialcountries with abundant capital should exportcapital to developing countries, where capital isscarce but labor plentiful, and the return oncapital is expected to be higher. This puzzle wasaddressed specifically by Lucas (1990). This boxprovides some estimates of the realized returnon capital in developing and industrial countriesto see if they shed light on why capital is flowingthe “wrong way.”1

A first look at returns seems to deepen thepuzzle. The table uses national accounts data tocompare the realized return on aggregate capi-tal over the past decade in emerging marketswith that in the G-7 countries. The focus is onemerging markets as a subset of developingcountries since these countries receive gross cap-ital inflows and are thus integrated into interna-tional capital markets. The return on aggregatecapital is defined as the value added paid to cap-ital owners divided by the aggregate capitalstock.2 Across emerging markets this measureaverages 13.3 percent over the past decade,compared with 7.8 percent in the G-7. If returnsare higher, why would capital be flowing fromemerging markets to industrial countries? Ofcourse, higher risk may deter investors frominvesting in emerging markets. However, it mayalso be the case that true average returns to

investors are much lower than what nationalaccounts–based measures suggest. For onething, they may reflect predominantly small andmedium-sized firms that are difficult to invest in.Moreover, in emerging markets they may dispro-portionately reflect state-owned companies, forwhich claims on capital may not be traded.Finally, the quality of the underlying nationalaccounts data may in some cases be poor.

An alternative measure of the return on capi-tal is the internal rate of return on invested capi-tal (calculated by Fama and French, 1999, forpublicly traded, nonfinancial companies in theUnited States). The internal rate of return is thediscount rate that sets the net present value ofcash flows into and out of the corporate sectorequal to zero. It captures the return to aninvestor who buys firms at market value, receivesor covers their subsequent cash flow, and thensells them at market value.3 Compared with sim-ply looking at the performance of equity mar-kets over a certain horizon, this measure is morecomprehensive since it provides a return on allinvested capital, including debt and equity.Compared with the return on aggregate capital,this measure has the advantage that it is basedon publicly traded companies, so that domesticand foreign investors can buy and sell the sharesof these companies.

To calculate the internal rate of return acrosscountries, IMF staff used Worldscope, an inter-national database that covers balance sheetand other information from annual reports of

Box 2.2. Return on Investment in Industrial and Developing Countries

Note: The main authors of this box are RobinBrooks and Kenichi Ueda.

1Klingen, Weder, and Zettelmeyer (2004) investigatea similar question for sovereign debt. They find thatreturns on sovereign debt in emerging markets barelyexceed returns on U.S. treasury bonds.

2Profit income in the system of national accounts isthe value added paid to capital owners. The capitalstock for each country is estimated by cumulating fixedcapital formation over time, using the perpetual inven-tory method with 1951 as the initial year. The initialcapital stock is assumed, but recent estimates for thecapital stock are robust to changes in initial stocks.

3The present value is calculated at the initial year,1994. The actual cash flows are used for 1994 to 2003,and the value of future cash flows is approximated bythe market capitalization in 2003. More formally, theinternal rate of return is the discount rate, r, thatsolves the following equation for each country:

T(2003) Xt – ItT(2003) FSt – FBt V2003V1994 = ∑ –––––– + ∑ ––––––– + –––––––,

t=1(1994) (1 + r)t t=1(1994) (1 + r)t (1 + r)T

where V denotes the sum of the market value of firms;X, the sum of cash earnings; I, the sum of gross invest-ment; FS, the sum of the market value of firms thatexit from the sample; and FB, the sum of market valueof firms that enter the sample.


101

publicly traded firms. These data are subject toseveral caveats. As the table illustrates, coverageof publicly traded firms in emerging markets isnot as good as in the G-7 countries. On averageover the 1994–2003 period, Worldscope covers71 percent of listed firms in emerging markets,according to the S&P Global Stock MarketsFactbook, while it covers 79 percent in the G-7. Inaddition, the stock market is substantially smallerin relation to economic activity in emergingmarkets than in the G-7. On average over thisperiod, market capitalization in percent of GDPis 64 percent in emerging markets, comparedwith 83 percent in the G-7. In other words, theproxy for the return on capital in emerging mar-kets needs to be interpreted with caution, bothbecause the stock market captures a smallershare of productive capital in these economiesand because coverage is less comprehensive.

The table provides the internal rate of returnfor the nonfinancial corporate sector for the pastdecade, measured in local currency and deflatedusing the price index for investment goods. Theinternal rate of return is –4.7 percent acrossemerging markets over the period, while it is2.4 percent on average in the G-7.4 On this

measure, the return on capital for publiclytraded firms in the nonfinancial corporate sec-tor was lower in emerging markets than in theG-7 during 1994–2003.

Of course, the short time horizon may bias theresults against emerging markets, where returnson investment may only be realized over a longerhorizon.5 The short time horizon also carries theadded risk that cyclical and crisis effects maydominate the results; indeed, returns have recov-ered strongly in emerging markets over the pastfew years. Nevertheless, the return on capital—asmeasured by the internal rate of return—forpublicly traded firms in emerging markets hasbeen below that in the G-7 for the 1994–2003period. This is especially striking because thereturn on capital is measured in local currency.It therefore does not incorporate currency risk,which for many emerging markets is substantial.As a result, attainable rates of return for domes-tic and foreign investors may help explain thecurrent direction of international capital flows.Still, further research is needed to understandwhy the return from investing in publicly tradedfirms in emerging market countries has beenlower than in industrial countries, and whetherthis result holds over different time periods.

Estimated Rates of Return on Invested Capital, 1994–2003(Percent, unless otherwise indicated)

Return on Capital Coverage Size1 Internal Rate of Return

Latin America2 12.9 56.0 37.7 –4.7Emerging Asia3 14.7 79.6 97.2 –4.6Other emerging markets4 11.3 70.2 42.5 –4.7

Average 13.3 70.5 64.1 –4.7

G-7, of which: 7.8 78.8 83.2 2.4United Kingdom 7.7 79.0 149.1 2.6United States 9.9 79.8 126.8 8.6

Source: IMF staff calculations.1Percent of GDP.2Argentina, Brazil, Chile, Colombia, Mexico, and Peru.3China*, Hong Kong SAR, India, Indonesia*, Korea, Malaysia*, the Philippines, Singapore*, Taiwan Province of China*, and Thailand.

Asterisk indicates countries where data on return of capital are not available.4Czech Republic*, Hungary*, Israel, Morocco*, Pakistan*, Poland*, Russia*, and South Africa. Asterisk indicates countries for which

data on return of capital are not available.

4This result is consistent with other measures of cor-porate performance, such as the return on assets oran alternate internal rate of return based on acquiringcosts of assets calculated by Fama and French (1999).

5To put this in perspective, Fama and French (1999)calculate the internal rate of return in the UnitedStates for a period of 47 years.


102

The world is in the midst of major demo-graphic transitions in which declining fertilityrates and increasing life expectancy are signifi-cantly altering the age structure of national pop-ulations. The timing and speed of demographicchanges, however, are highly asymmetric acrosscountries. Advanced economies began theirtransition several decades earlier than develop-ing countries; hence the age composition ofpopulations differs greatly between developedand developing countries (see the figure).Ongoing demographic changes will have signifi-cant albeit uncertain impacts on saving, invest-ment, and current account balances in the yearsahead (Bryant, 2004).

How do demographic changes affect saving,investment, and the current account?

Demographic impacts on saving stem fromindividuals smoothing consumption over theirlifetimes while the age distribution of theirincome follows a hump-shaped profile. This“life-cycle” behavior entails dissaving when indi-viduals are young, little saving early in adult life,high saving at the middle and end of the work-ing life, and then low or negative saving afterretirement. Changes in the age composition ofthe population therefore affect aggregate per-sonal saving. In particular, a demographic transi-tion initially increases household saving as itreduces the number of young dependents andincreases the number of working adults, buteventually it reduces saving as a larger portionof the population retires and reaches old age.

There remain, however, some uncertaintiesabout saving behavior in the later stages of thelife cycle. Studies based on macroeconomic datagenerally support the predictions of life-cycleapproaches (for example, an increase in theelderly dependency ratio—which shows thepopulation aged 65 and older as a share of theworking-age (age 15–64) population—reducessaving). Studies based on microeconomic data,

however, have cast some doubt on the extent towhich the elderly dissave (Poterba, 2004). Thismay be because simplified applications of thelife-cycle approach do not adequately take intoaccount the desire of the elderly to leavebequests, or their uncertainties about their life-span after retirement and the financial supportthey will need. Some empirical studies based onhousehold survey data do not adequately incor-porate the public-pension portion of elderlyincomes, and this is why they may appear atodds with life-cycle behavior (Miles, 1999).

With regard to investment, empirical studiesgenerally find that investment is positivelyrelated to the share of the young in the popula-tion (Higgins, 1998). Countries with higheryouth dependency rates—which show the popu-lation aged 0–14 as a share of the working-age

Box 2.3. Impact of Demographic Change on Saving, Investment, and Current Account Balances

Note: The main authors of this box are Ralph C.Bryant and Marc de Fleurieu.

0

10

20

30

40

50

60

Population Structure, 1950–2050(Percent of total population)

Elderly Dependency Ratio Youth Dependency Ratio

Industrial countries

Developing countries


Developing countries

Source: United Nations, World Population Prospects: The 2000 Revision.

1950 70 90 2010 30 50


103

population—face a relatively higher demand forinvestment related to the development ofhuman capital (schools) and to a growing laborforce (infrastructure). As a population ages,however, the labor force grows more slowly andthe level and composition of investment shiftwith the needs of a more elderly population(medical facilities).

The net effect on the saving-investment bal-ance tends therefore to vary during the differentstages of a demographic transition.1 Countrieswith a relatively young population should expe-rience current account deficits, as investmentdemand outstrips domestic saving. As childrenage, fertility rates decline, and life expectancyrises, the ratio of active workers to the total pop-ulation increases, which in turn tends to causesaving to rise faster than investment. Hence, aseconomies go through the middle stages of ademographic transition, they should experiencecurrent account surpluses. Eventually, as theaging of the population continues, the netimpact on the saving-investment balancebecomes ambiguous, reflecting uncertaintyabout the relative effects of rising elderly ratioson saving and investment. Although higherelderly dependency is often associated with anexcess of investment over saving, and hence acurrent-account deficit, this partial correlationhas lower statistical reliability and may not be arobust guide to the effects of rising elderlydependency ratios on current account balances(Higgins, 1998; Bosworth and Keys, 2004).

Demographics and capital flows

According to UN projections, elderly depend-ency ratios in the advanced countries will nearlydouble by 2050. At the same time, working-agepopulations will rise significantly in many devel-oping countries. How will these populationtrends affect capital flows in the future? For

developing countries, an IMF study (see theSeptember 2004 World Economic Outlook) foundthat demographic forces are likely to lead toimproved current account positions over time asmore of the population move through theirhigher saving years. In industrial countries,demographic trends will probably reduce cur-rent account surpluses in Japan and Europeduring the course of the next 50 years. In theUnited States, the demographic transition is lesssteep and demographic forces by themselvesmay have smaller effects on the current accountbalance.

Understanding how demographic change willaffect saving, investment, and net capital flows isfar from complete. There are uncertainties notonly about the demographic projections them-selves, but also about the reactions of privatesaving and investment as the demographic tran-sitions unfold. Households—in both advancedand developing countries—will probablyrespond according to the broad predictions ofthe life-cycle model; but aggregate saving,investment, and net capital flows will also besignificantly influenced by other factors, includ-ing international differences in policies andbusiness-cycle conditions.

What can the role of labor migrations be?

A remaining issue to consider is the possiblerole of labor mobility in the demographicadjustment process. Most macroeconomicmodels assume that labor does not move acrosscountries. This omission could lead modelpredictions to overstate the role of capital flowsin the adjustment process because movementsof labor from regions with rising working-agepopulations to those with rising elderly depend-ency ratios are a possible alternative to capitalflows.2

1Empirical work suggests that, on average, abouthalf of the demographic effects on national saving arematched by changes in domestic investment, with theremainder altering the saving-investment balance(Higgins, 1998; and Helliwell, 2004).

2Also, the assumption of perfect capital mobility andperfect foresight ignores the presence of capitalaccount restrictions and political risk in developingcountries. As a result, model predictions of the magni-tude of demographically induced capital flows to andfrom developing countries could be overstated.

different population characteristics in emerg-ing countries, the elderly dependency ratio isnot found to be a significant explanator of sav-ing behavior in those countries.

• Saving is positively related to improvements in theterms of trade, which are normally expected tobe transitory (the “Harberger-Metzler” effect).The results suggest that a 1 percentage pointincrease in terms-of-trade growth would implyan increase of !/4 percent of GDP in the savingrate.

• Saving behavior does not appear to be affected byrate of return considerations. While saving andreal interest rates are generally expected to bepositively related—with the strength of thisrelationship likely to depend on the size ofhouseholds’ net asset position (see Deaton,1992)—the regression results did not show astatistically significant impact (at the 5 percentlevel). This finding is consistent with previousempirical work.In turn, the estimated investment equation

suggests the following.• Stronger output growth leads to higher investment

rates. This could reflect demand shocks,responses to changes in productivity, or thepresence of financial market imperfections.10

The results suggest that a sustained 1 percent-

age point increase in per capita output growthin the industrial countries would over timelead to a 1.6 percent of GDP increase in theinvestment rate. In emerging markets, theimpact is smaller, at 1.1 percent of GDP.

• Increased availability of credit is associated withhigher investment, given that firms, in part,depend on external finance. The regressionresults, however, suggest that the effects of anincrease in credit on investment, although sta-tistically significant, are modest.

• An increase in the cost of capital is associated withlower investment. Here, the cost of capital ismeasured as the product of the real interestrate and the relative price of capital (invest-ment deflator over GDP deflator). In the indus-trial countries, a 1 percent increase in the costof capital would, over time, lead to a 0.4 per-cent of GDP reduction in the investment rate.11

What do these equations tell us about thefactors that could explain recent movementsin saving and investment? At the outset, theability of both regressions to capture recentdevelopments varies considerably across coun-tries, reflecting both the panel nature of theregressions and the heterogeneity in regional sav-ing and investment behavior (as documented inthe previous section). The results suggest that


104

More research is needed to clarify the netbenefits of migration for both recipient andsending countries. Permanent immigrationtends to have a neutral effect on a recipientcountry’s public saving—immigrants are aslikely to claim pension and healthcare benefitsas national citizens (Fehr, Jokisch, andKotlikoff, 2004). The effects on private savingcould be significant, though dampened by theextent to which migrants send remittances totheir home country. For sending countries, per-

manent emigration is a net loss apart fromremittances.

Government policies inhibit the flow ofpeople across borders. In fact, immigrationpolicies are a more significant determinant ofmigration than the willingness of individuals tomigrate. Large movements of people across bor-ders in the coming decades are thus unlikely—under current policies—to significantly mediatethe macroeconomic effects of asymmetricdemographic transitions.

Box 2.3 (concluded)

10See Blanchard and Fischer (1989). Indeed, both past and future output growth and real interest rates are expected toinfluence investment (see, for example, Romer, 1996).

11See Pelgrin, Schich, and de Serres (2002) for a related result.

two factors are particularly important in explain-ing the decline in saving in industrial countriesover 1997–2004 (Figure 2.5). First is the increasein credit to the private sector, which is likelyapproximating for wealth effects from the sharpincrease in house prices in many countries (butnot Japan). Second is the fall in public saving,which is particularly important in the UnitedStates—where according to the regression esti-mates it accounted for over one-third of the4!/2 percentage point decline in national savingsince 1997—and Japan. Another factor that hasplayed an important role in Japan and the euroarea—but not in industrial countries as a group—is the rise in the elderly dependency ratio.

Turning to emerging markets, the results alsosuggest that two factors have been key drivers ofthe recent increase in saving. First, there hasbeen a sharp increase in public sector saving,particularly in China and the oil-producingcountries (which has more than offset the weak-ening in public saving in east Asia). Second,stronger output growth has boosted saving in allemerging market regions (again, this appearsparticularly important in China, where it likelycontributed to the sharp increase in corporatesaving—see Box 2.1). In contrast, rising oilprices have had a modestly negative effect, withthe boost to saving in oil-producing countriesoffset by the adverse effect elsewhere (particu-larly in parts of Asia).

The investment equation is less successfulthan the saving equation in tracking recentdevelopments. This result is similar to otherrecent studies, which have found that traditionaleconometric models of investment have diffi-culty explaining recent trends.12 The equationoverpredicts investment in both the industrialand emerging market regions, in some cases by


105

-3 -2 -1 0 1 2 3

Saving


Emerging market and oil-producing economies

Figure 2.5. Explaining Saving and Investment Rate Movements Between 1997 and 2004 (Change in percent of GDP unless otherwise noted)

While the recent evolution of saving is largely explained by the economic fundamentals included in the regression analysis, the evolution of investment is not.

1

-4 -3 -2 -1 0 1 2 3 4

Investment


Emerging market and oil-producing economies 1

Source: IMF staff calculations. Includes Norway.1

Real per capita output growth Credit Actual rate

Terms-of-trade growthPublic savingCost of capital

Projected rate

12According to Tevlin and Whelan (2003), there arethree main reasons that account for this failure: (1) thefalling price of computer equipment played a key role inthe investment behavior of the 1990s; (2) capital depreci-ation rates rose significantly during the 1990s; and (3)depreciation is not homogenous across the diverse typesof capital.

large margins. For instance, while the equationpredicts that investment should have increasedin industrial countries—largely as a result of thedecline in the cost of capital—investment in sev-eral key industrial countries, including Japanand the Large Euro countries (see footnote 15),fell. Similarly, the equation fails to explain thedrop in investment in emerging markets, partic-ularly in the east Asian countries. The equationsuggests that the investment accelerator—whereby investment rates and output growthmove in the same direction—has not worked asstrongly as expected in recent years in thesecountries, most likely because corporates havefocused on reducing debt and strengthening bal-ance sheets, rather than on investing in capital(see Box 2.4 for a discussion of investment inAsian emerging markets).13 One conclusion ofthis analysis is that investment appears to bebelow the levels that would usually be associatedwith this stage of the economic cycle, and thismay be an important factor in explaining boththe current low level of real long-term interestrates and the shift of emerging market econo-mies from net importers to net exporters ofcapital.

A Model of Saving, Investment, and theCurrent Account

The econometric analysis provides someimportant insights into what may be driving

recent saving and investment behavior, but it alsohas drawbacks. Most important, each variable isconsidered separately in the analysis, rather thanas part of an integrated economic system. As washighlighted in the previous subsection, savingand investment—and indeed many of theirpotential determinants, including output andinterest rates—are highly correlated, particularlyacross industrial countries but also increasinglyin emerging markets. This suggests that it is veryimportant to be able to capture the interactionsbetween variables and across countries within anintegrated and consistent framework. Oneapproach is to use a multiregion macroeconomicmodel—such as the IMF’s Global EconomicModel (GEM)—that explicitly captures suchinteractions (see Appendix 1.2). An alternativethat is used here is to estimate a dynamic factormodel to examine the extent to which “global”economic conditions have been driving saving,investment, and current account balances acrossregions.14

The dynamic factor model that was estimatedconsiders five variables—real GDP growth, short-term real interest rates, saving rates, investmentrates, and current account balances—and decom-poses them into the following four estimated(unobserved) components (see Appendix 2.1 formore details on the model):15

• A world factor that captures the commonshocks affecting all regions and all variablesof the model. This will reflect major global


106

13There is also evidence that the current low corporate investment in the euro area reflects the high leverage levels ofthese corporations (Jaeger, 2003).

14Several important features of the model are worth stressing. First, while bivariate correlations capture the degree ofcontemporaneous co-movement of the saving and investment ratio for any pair of regions, the global factors estimated inthe model capture all intertemporal (e.g., including leads and lags) cross-country correlations among the variables consid-ered. Second, by estimating the global and country-/region-specific factors simultaneously, the model correctly identifiesthe relative importance of global and region-specific developments. Third, the global factors estimated in this model areindependent of the choice of any particular weighting scheme, contrary to those obtained as cross-country averages of vari-ables, as in Glick and Rogoff (1995) and Dees and others (2005).

15For this exercise, the countries in the sample were divided into 12 regions. These are the United States, Japan, Anglo-Saxon (Australia, Canada, New Zealand, and the United Kingdom), Large Euro (Italy, France, and Germany), Small Euro(Austria, Belgium, Finland, Greece, Ireland, the Netherlands, Portugal, and Spain), Other Industrial (Denmark, Sweden,and Switzerland), East Asia 1 (Indonesia, Korea, Malaysia, the Philippines, and Thailand), East Asia 2 (Hong Kong SAR,Singapore, and Taiwan Province of China), China, Latin America (Argentina, Brazil, Chile, Colombia, and Peru), OtherEmerging Markets (Egypt, India, Israel, Morocco, Pakistan, South Africa, and Turkey), and Oil-Producing Countries (I.R.of Iran, Mexico, Norway, Saudi Arabia, and Venezuela). Before the model was estimated, all time series were linearlydetrended to avoid the possibility of nonstationarity.


107

Investment in emerging Asia fell during theregional financial crises in the late 1990s, andhas since remained at these lower levels (exceptin China). For Hong Kong SAR, Singapore, andTaiwan Province of China, the decline hastaken investment rates to levels not seen in overthree decades, while investment in Indonesia,Korea, Malaysia, and Thailand, after reachinghistorical peaks in the early to mid-1990s, hasreturned to levels comparable to those in themid-1980s.

Corporate investment in emerging Asia hasfallen particularly sharply—the investment tocapital ratio fell by one-half between 1993–96and 1997–2003—although an increase in publicinvestment has offset part of this decline. Thedrop in corporate investment in the regionreflects a sharp decline in the Tobin’s q—as themarket value of the corporations fell sharply rel-ative to the replacement cost of capital—and theefforts of corporations to strengthen their bal-ance sheets and streamline their operations asthe financial and economic environment deteri-orated. In particular, leverage and liquidity haveimproved significantly since 1997, as shown inthe first figure, even if they have not yet reachedtheir pre-1997 levels.1

These developments raise two related ques-tions: Is investment in emerging Asia now toolow? What are the prospects for a rebound ininvestment?

While these are clearly difficult questions toanswer, one way of addressing the first is tocompare the investment and capital-outputratios in each country with estimates of theirlong-run equilibrium (steady-state) levels. Tothe extent that these countries are still in atransition period (that is, their capital-outputratio is below its long-run level), investmentrates should be above their long-run level.Such calculations are shown in the second

figure. In making these calculations, a deprecia-tion rate of 5 percent is used in estimating thecapital stock (the average depreciation rate for

Box 2.4. Is Investment in Emerging Asia Too Low?

0.0

0.2

0.4

0.6

0.8

1.0

0.00

0.05

0.10

0.15

0.20

0.25

0.30Investment and Leverage

Leverage (left scale)

Investment(right scale)

0220009896941992

2

0.00

0.05

0.10

0.15

0.20

0.25

0.30Investment and Liquidity

Liquidity

Investment

0220009896941992

2

0

1

2

3

4

5

0.00

0.05

0.10

0.15

0.20

0.25

0.30Investment and Tobin's q

Tobin's q

Investment(right scale)

0220009896941992

2

Emerging Asia: Investment, Leverage, Liquidity, and Tobin's q(Regional averages of country medians)

1

Source: IMF staff calculations. Excludes China.

Investment-to-capital ratio.

12

(left scale)

Note: The main authors of this box are RobertoCardarelli and Marco Terrones.

1A description of the database is provided inAppendix 2.1.


108

emerging markets in 2004 was about 5 percent,2

although in the future this may be too low anestimate given that depreciation rates havebeen increasing over the past decade owing tothe compositional shift in the aggregate capitalstock toward short-lived assets such as comput-ers and software).3 A further critical issue is thechoice of the steady-state levels of the capital

stock and investment. Here, they are estimatedfrom a standard neoclassical growth model,using conservative estimates of capital ratiosand potential output growth, although alterna-tives were also tried and yielded broadly similarresults.4

In 1996—the year prior to the regional finan-cial crisis—almost all countries were investingrelatively more than in the steady state as theyincreased their capital stock toward its long-runlevel. This does not preclude the possibility thatthese countries could have been overinvestingduring this period as they moved too quicklytoward the long-run level (Sachs and Radelet,1998). Indeed, for some countries, such asMalaysia, the investment rate in 1996 appears tobe consistent with an excessive speed of conver-gence, as reflected in a capital ratio that in2004 is higher than its estimated long-termlevel. For these countries, the fall in investmentis a response to the excess capacity built overthe past decade. In 2004, however, some coun-tries were investing relatively less than in thesteady state despite the fact that their capitalstock was below its long-run level (and, there-fore, were in the “underinvestment” quadrant).This appears to be the case, in particular, forIndonesia, the Philippines, and Thailand. Onthe other hand, there is some evidence of over-investment in China.

Of course such evidence is tentative, andeven if investment rates are too low in somecountries at present it is difficult to know whenthey may rebound. On a positive note, Tobin’sq appears to be starting to recover, which, giventhe close relationship with investment in theregion and the improved balance sheet positionof corporations, suggests that the investmentoutlook may be turning more positive.

Box 2.4 (concluded)

Emerging Asia: Capital/Output and Investment/Output Relative to Steady-State Level

0.4 0.6 0.8 1.0 1.2 1.4 1.60.0

0.5

1.0

1.5

2.01996

Overinvestment

Underinvestment

Capital-to-output ratio

Inve

stm

ent-t

o-ou

tput

ratio


0.4 0.6 0.8 1.0 1.2 1.4 1.60.0

0.5

1.0

1.5

2.02004

Overinvestment

Underinvestment

Capital-to-output ratio

Inve

stm

ent-t

o-ou

tput

ratio

2Based on consumption of fixed capital from theNational Accounts, deflated by the investment deflator.

3A depreciation rate of 7 percent yielded a broadlysimilar outcome to that shown in the second figure(see Appendix 2.1).

4For example, broadly similar results were achievedwhen steady-state values of capital and investmentratios were set as the averages for industrial countriesover an eight-year period (for example, for 2004, theaverage of ratios in industrial countries over 1996–2004). For a description of the methodology, seeAppendix 2.1.

economic events, such as oil price increases orglobal technological progress.

• A factor common to each of the five variablesin the model. For instance, the saving factorcaptures the common shocks affecting savingrates across all regions (reflecting, for exam-ple, the ongoing process of financial innova-tion) but not other variables.

• A region-specific factor that reflects commonshocks affecting the five variables within eachregion. For instance, the process of Europeanintegration may affect all economic variablesin the European countries, but not in otherregions.

• An idiosyncratic term capturing region-specific shocks to each individual variable ineach region.The results from the dynamic factor model

indicate that a high proportion of the variationsin saving and investment rates in industrialcountries are explained by global factors(defined as the sum of the world factor andthe variable-specific factors—see Table 2.3).Indeed, much of the recent cyclical evolutionin saving and investment in these countries—with the important exception of Japan—canbe explained by the global factor, suggestingthat industrial countries have been subject to


109

However, the low level of investment may alsoreflect structural changes in these economies,such as the shift toward less capital and moreskill- and knowledge-intensive type of exports,particularly information technology–relatedproducts and services (Lee, McKibbin, andPark, 2004) and the start of a demographictransition toward an older population structure(see Box 2.3). Both factors suggest that emerg-ing Asian countries could face a slower pace of

capital accumulation in the future than theyhave in the past.5

5Anecdotal evidence suggests that another potential“structural” explanation for the lower investment ratein many emerging Asian countries involves the reloca-tion of production facilities from these countries toChina. Unfortunately, lack of data on bilateral foreigndirect investment flows that distinguish between green-field investment and mergers and acquisitions preventsany quantitative estimate of the phenomenon.

Table 2.3. Variance DecompositionShort-Term Current Account Investment

Output (y) Interest Rate (r) (CA) Saving Rate (S) Rate (I )

Average for all countriesGlobal 37 51 36 39 34

World 18 8 7 24 21Aggregate 19 43 28 15 14

Region plus idiosyncratic 63 48 63 60 65

Average for industrial countriesGlobal 57 80 32 71 59

World 27 10 7 45 35Aggregate 30 70 25 26 23


Average for emerging market and oil-producing countries

Global 17 23 39 7 10World 8 7 8 3 6Aggregate 9 16 32 4 4


Source: IMF staff calculations.

similar shocks in the variables affecting savingand investment (Figures 2.6 and 2.7).16 Subse-quent causality tests suggest that it is the highdegree of co-movement in productivity andasset prices across industrial countries that isdriving this observed co-movement of savingand investment.17 These results are similar tothose from the econometric analysis, suggestingthat common developments in credit/assetprices and output growth/productivity are themost important factors driving the evolution insaving and investment across most industrialcountries.

In emerging market countries, on the otherhand, the global factor has been much lessimportant in explaining movements in savingand investment. Rather, these regions have beenmore likely to experience region-specific savingand investment cycles, a reflection of the largedifferences in economic structures, institutions,and policies between the two groups of coun-tries.18 For example, the rapid acceleration ofsaving in China and oil-producing countries inrecent years is described by the model as a


110

1972 80 88 96 2004-9

-6

-3

0

3

6

9Oil Producers

1972 80 88 96 2004-10

-5

0

5

10

15 China

The global factor plays an important role in explaining fluctuations of saving rates in industrial countries, but fails to account for the recent decline in saving in Japan and large euro area countries. Saving rates in emerging markets have been largely unexplained by the global factor.

1972 80 88 96 2004-6

-4

-2

0

2

4

6

1972 80 88 96 2004-6

-4

-2

0

2

4

6 Anglo-Saxon Large Euro

1972 80 88 96 2004-6

-4

-2

0

2

4

6

8

1972 80 88 96 2004-6

-4

-2

0

2

4

6 United States Japan

1972 80 88 96 2004-10

-8

-6

-4

-2

0

2

4

6 East Asia 1

1972 80 88 96 2004-6

-4

-2

0

2

4

6East Asia 2

GlobalActual 23

Figure 2.6. Saving: Global Factor (Percent of GDP)

1

Source: IMF staff calculations. See footnote 15 in main text for definition of country groupings. Actual time series have been rescaled by subtracting the average of the period. The global factor is the sum of the world and saving factors. For each region, the world and saving factors are multiplied by their factor loadings in the saving equation.

123

16Figures 2.6, 2.7, and 2.8 plot the global factorstogether with actual (not detrended) time series. How-ever, it is worth stressing that, because the model is esti-mated based on detrended data, the common factorsare not able to explain trend developments in the observ-able variables and, in particular, the trend decline ofsaving and investment ratios in Japan and Large Eurocountries, as well as the trend increase in these ratios inChina.

17This is done through a series of bivariate Grangercausality tests. In particular, movements in the global sav-ing factor appear to be related to changes in real houseprices in the United States, while the global investmentfactor appears to be related to total factor productivitygrowth in industrial countries. These results are consis-tent with the existence of a strong co-movement in hous-ing prices across industrial countries, as discussed in “TheGlobal House Boom,” in Chapter 2 of the September2004 World Economic Outlook, and with other studies show-ing the relevance of productivity dynamics in explainingthe co-movement of investment across G-7 countries(Gregory and Head, 1999; and Kose, Otrok, andWhiteman, 2004).

18This result is in line with that of Kose, Otrok, andWhiteman (2003), who find that output and investmentdynamics are much more idiosyncratic in developingcountries than in developed ones.

highly idiosyncratic event. The same is true forthe sharp drop in investment in east Asia afterthe 1997 regional financial crisis, a finding thatis consistent with the results from the panelregression.

The results from the dynamic factor modelreveal a high degree of synchronization betweenthe current account balances in industrial andemerging market economies (as would beexpected given that the movements of regionalsaving-investment gaps are subject to the globalconstraint that saving should be equal to invest-ment).19 On average, the global current accountfactor explains about one-third of the variationin current account balances for both industrialand emerging market economies. This result,however, is largely driven by relatively highshares for a small group of regions (the UnitedStates, Japan, the Large Euro area, and the EastAsia 2 countries).

Interestingly, the global current account fac-tor captures the global imbalances episode inthe mid-1980s better than it does the currentone (Figure 2.8). This is likely because—asnoted in the previous section—the imbalancesin the mid-1980s were largely concentrated in arelatively small number of countries (in particu-lar, the United States, Japan, the Large Euroarea, and the East Asia 2 countries). In contrast,the imbalances are now distributed across alarger number of countries, and appear to bemore a result of region-specific (idiosyncratic)events rather than any single global event(although clearly the recent increase in oilprices has added to the size of the imbalancesover the past two years). This clearly has impor-tant implications for how existing imbalancescan be resolved, with actions needed across abroad group of countries.


111

1972 80 88 96 2004-4

-2

0

2

4

6Oil Producers

1972 80 88 96 2004-10

-5

0

5

10 China

While the fluctuations of investment rates in industrial countries are relatively synchronized, in the emerging market economies they are more the result of idiosyncratic shocks.

1972 80 88 96 2004-6

-4

-2

0

2

4

6

1972 80 88 96 2004-4

-2

0

2

4


1972 80 88 96 2004-10

-5

0

5

10

1972 80 88 96 2004-4

-2

0

2


1972 80 88 96 2004-9

-6

-3

0

3

6

9 East Asia 1

1972 80 88 96 2004-9

-6

-3

0

3

6

9East Asia 2

GlobalActual 2 3

Figure 2.7. Investment: Global Factor (Percent of GDP)

1

Source: IMF staff calculations. See footnote 15 in main text for definition of country groupings.

Actual time series have been rescaled by subtracting the average of the period. The global factor is the sum of the world and investment factors. For each region, the world and investment factors are multiplied by their factor loadings in the investment equation.

123

19As the regions considered in this study account for alarge share of world GDP, their current account balancescannot all move independently of each other. Thedynamic factor model accommodates this constraint, asthe sign of the impact of the global factors on the currentaccount balances differs across regions (it is positive forthree regions and negative for the other nine).

How Can Existing Global Current AccountImbalances Be Reduced?

The results from the previous section indicatethat saving, investment, and current accountbalances have been affected both by factors thatare common across many countries, such as ris-ing asset prices/credit—which appear particu-larly important in industrial countries—and byfactors that are specific to particular regions andcountries.

What do these findings suggest about howexisting global current account imbalancescould be resolved? To address this question,a factor-augmented vector autoregressive(FAVAR) model was estimated that combinesthe estimated factors from the dynamic factormodel with other variables of interest inselected countries (see, for instance, Bernanke,Boivin, and Eliasz, 2005). This approach has theadvantage of yielding a parsimonious modelthat is able to capture global linkages andspillovers.

The particular simulations considered are anincrease in U.S. national savings; an increase ininvestment in Asia and oil-producing countries;stronger real output growth in Japan and theLarge Euro countries; and an increase in realinterest rates in the United States. The results ofthe analysis are as follows.20

• An increase in U.S. national saving rates wouldhave a significant positive effect on the U.S. currentaccount deficit (Figure 2.9). A permanent (overthe three-year horizon considered by theforecasts) 1 percent of GDP increase in theU.S. gross national saving rate would reducethe U.S. current account deficit by about


112

1972 80 88 96 2004-9

-6

-3

0

3

6

9Oil Producers

1972 80 88 96 2004-6

-4

-2

0

2

4 China

The global current account factor captures well the current account imbalances in the mid-1980s. More recently, it fails to account for the recent developments in the United States, East Asia 1 countries, China, and oil-producing countries.

1972 80 88 96 2004-3

-2

-1

0

1

2

3

1972 80 88 96 2004-4

-2

0

2


1972 80 88 96 2004-4

-2

0

2

4

1972 80 88 96 2004-6

-4

-2

0

2


1972 80 88 96 2004-8

-4

0

4

8

12 East Asia 1

1972 80 88 96 2004-10

-5

0

5

10

15East Asia 2

GlobalActual 2 3

Figure 2.8. Current Account: Global Factor (Percent of GDP)

1

Source: IMF staff calculations. See footnote 15 in main text for definition of country groupings.

Actual time series have been rescaled by subtracting the average of the period. The global factor is the sum of the world and current account factors. For each region, the world and current account factors are multiplied by their factor loadings in the current account equation.

123

20The FAVAR model used to forecast the currentaccount imbalances comprises the world factor, theglobal current account factor, the regional currentaccount balance, and the variable of interest (e.g., theU.S. saving rate). The results are presented as differencesbetween the forecast of the current account balanceobtained imposing a specific time path on the variableof interest (conditional forecasts) and the forecastobtained from the unrestricted VAR (unconditionalforecasts).

!/2 percent of GDP after three years.21 Highersaving in the United States—and the associ-ated reduction in domestic demand—wouldalso reduce the current account surplusesin the Large Euro countries and Japan, by!/4 and #/4 percentage point of GDP by 2007,respectively. In the East Asia 2 countries,the projected decline in the current accountsurplus would be much larger—close to 3 percentage points of GDP by 2007—owingto the much higher exposure of these coun-tries to trade with the United States (seeTable 2.4).

• An investment recovery in Asia (excluding China)and oil-producing countries would offer a signifi-cant contribution to the resolution of currentaccount imbalances. An increase in the invest-ment ratio in the East Asian 1 countries(Indonesia, Korea, Malaysia, the Philippines,Thailand)—as firms in this region completetheir process of deleveraging and begin toincrease their capital stock—would have animportant impact on current account imbal-ances.22 For example, a 5 percent of GDPpermanently higher investment rate in thesecountries—which would reverse about one-half of the decline that has occurred sincethe peak in 1996—would reduce the U.S.

HOW CAN EXISTING GLOBAL CURRENT ACCOUNT IMBALANCES BE REDUCED?

113

Figure 2.9. How Would the Current Account React to an Increase in the U.S. Saving Ratio?(Percent of GDP)

1

A 1-percent-of-GDP increase in the U.S. national saving rate would have a significant positive effect on the U.S. current account deficit, while negatively affecting the current account balances of Japan and the Large Euro and East Asia 2 countries.

-1.6

-1.2

-0.8

-0.4

0.0

0.4

0.8

1.2

2004 05 06 07-1.6

-1.2

-0.8

-0.4

0.0

0.4

0.8

1.2United States Japan

2004 05 06 07

-1.6

-1.2

-0.8

-0.4

0.0

0.4

0.8

1.2

2004 05 06 07-10

-8

-6

-4

-2

0

2

4East Asia 2Large Euro

Source: IMF staff estimates. See footnote 15 in main text for definition of country groupings.

1

2004 05 06 07

Baseline Confidence interval

21Model-based analyses of the effect of U.S. fiscal policyon the current account deficit yield similar results. Forexample, Kumhof, Laxton, and Muir (2005) find that apermanent 1 percentage point of GDP increase in theU.S. government saving ratio—increasing the U.S.national saving rate by #/4 percentage point—improves theU.S. current account by almost !/2 percent of GDP on aver-age during the first five years. Moreover, current accountchanges as a ratio to GDP are similar across other regionsof the world. If the duration of the fiscal consolidationeffort is restricted to five years, the current accountimprovement is smaller—about !/4 percentage point ofGDP—similar to that reported in Erceg, Guerrieri, andGust (2005).

22It should be noted that these results are based on amodel estimated over the past three decades, a periodduring most of which several emerging markets, particu-larly the fast-growing economies of southeast Asia andChina, were much smaller and less important for worldtrade. Hence, an analysis based on past data likely under-estimates the role played by adjustments in this group ofcountries today.

current account deficit by about #/4 percent ofGDP after three years (Figure 2.10). An equiv-alent increase in the investment rate in oil-producing countries, as they devote a largershare of their oil revenues to the accumula-tion of capital, has a broadly similar impact onthe U.S. current account deficit.

• An increase in real GDP growth in Japan and theLarge Euro countries would help reduce the U.S.current account deficit. A !/2 percent a yearincrease in real GDP growth in Japan wouldreduce the U.S current account deficit byabout 0.2 percent of GDP after three years,while worsening the current account in Japanby 0.3 percent of GDP and having almost noeffect on the current account of Large Eurocountries (Figure 2.11). A !/2 percent a yearincrease in real GDP growth in Large Eurocountries would have a broadly similar impacton the U.S. current account deficit.

• An increase in real short-term interest rates in theUnited States would have a limited impact on the

current account. A cumulative 2 percentagepoint increase over the next three years wouldhave a very modest impact on current accountbalances in the United States or other coun-tries (about 0.1 percent of GDP by 2007)(Figure 2.12). This is consistent with the highco-movement of interest rates, saving, andinvestment across industrial countries, mean-ing that as interest rates rise in the UnitedStates they are also likely to increase else-where. Correspondingly, saving and invest-ment in other industrial countries will beaffected, with little impact on global imbal-ances. Of course, if higher interest rates havemore of an effect in countries that have beenexperiencing house price booms, the impacton saving and current account balances couldbe larger.Overall, these results suggest that an increase

in U.S saving, achieved most directly through fis-cal consolidation, is likely to have a significantimpact on the current account deficit in the


114

Table 2.4. Average Trade by Region, 1970–2004(Percent of regional GDP)

United States Japan Anglo-Saxon Other Industrial Large Euro Small Euro

United States . . . 2.0 4.0 0.3 1.5 0.7Japan 4.8 . . . 1.9 0.3 1.2 0.6Anglo-Saxon 14.3 2.5 . . . 1.8 6.1 4.2Other Industrial 3.9 1.6 5.5 . . . 17.4 7.4Large Euro 3.0 1.0 3.3 2.8 . . . 7.0Small Euro 3.3 1.1 5.4 2.9 18.3 . . .East Asia 1 11.2 12.9 3.9 0.8 3.6 2.0East Asia 2 36.7 27.9 15.9 4.1 14.2 5.8China 3.3 4.5 1.3 0.3 1.9 0.8Other Emerging Markets 2.7 1.0 1.8 0.8 4.0 1.7Latin America 4.3 1.2 1.0 0.5 2.7 1.4Oil Producers 13.9 3.4 3.5 2.2 5.4 2.9

Other Emerging OilEast Asia 1 East Asia 2 China Markets Latin America Producers

United States 1.0 0.5 0.5 0.5 0.5 1.8Japan 2.7 0.9 1.1 0.5 0.4 1.5Anglo-Saxon 1.2 0.9 0.6 1.1 0.4 1.6Other Industrial 0.7 0.7 0.4 1.1 0.7 3.3Large Euro 0.6 0.4 0.5 1.3 0.6 1.5Small Euro 0.8 0.4 0.4 1.1 0.7 1.9East Asia 1 . . . 5.6 2.1 1.1 0.5 2.0East Asia 2 30.7 . . . 39.1 4.8 1.2 6.4China 2.1 4.9 . . . 0.4 0.4 0.4Other Emerging Markets 0.7 0.5 0.3 . . . 0.3 1.0Latin America 0.4 0.2 0.4 0.4 . . . 1.4Oil Producers 1.3 0.6 0.4 1.0 1.1 . . .

Sources: IMF, Direction of Trade Statistics; and IMF staff calculations.

United States and would reduce surpluses inJapan, the Large Euro countries, and Asia.Stronger growth in Japan and Europe and apickup in investment in emerging market econo-mies would also play a role in addressing currentimbalances. On the other hand, an increase inreal interest rates in the United States wouldhave a limited impact on its external deficit,given the spillovers on saving and investmentsratios in other regions. These results emphasizethe importance of adequately addressing cross-country linkages in any analysis of policies toreduce global imbalances.

ConclusionsGlobal saving and investment are near historic

lows, having fallen markedly since the late 1990s.These trends largely reflect developments inindustrial countries; in emerging markets, savinghas continued to rise, although investment hasnot recovered since its fall in the aftermath ofthe Asian financial crisis. The decline in globalsaving and investment has been due both to fac-tors that have commonly affected a large num-ber of countries—such as increases in credit andasset prices—and to country-/region-specificdevelopments. Most important among these arethe decline in public saving in the United States,demographic changes in Japan and Europe, andthe slump in investment in Asian economies(excluding China) in the aftermath of theregional financial crisis.

The recent paths of saving and investmenthave had significant implications for the distri-bution of current account imbalances across theworld. In particular, the U.S. current accountdeficit has reached unprecedented levels, andlarge surpluses have emerged in other regions.Contrary to the situation in the mid-1980s—thelast period of significant global imbalances—when the imbalances were concentrated amonga relatively small group of countries, the currentsituation involves a much wider set of players,including many emerging market countries.Consequently, the policy response will need toinvolve many more countries, and coordinating

CONCLUSIONS

115

Figure 2.10. How Would Current Account Imbalances React to Changes in the Investment Rate in Asia?(Percent of GDP)

1

A 5-percent-of-GDP increase in the investment rate in East Asia 1 countries would have a relatively significant effect on the U.S. current account deficit.

-1.0

-0.5

0.0

0.5

1.0

1.5

2004 05 06 07-1.0

-0.5

0.0

0.5

1.0


2004 05 06 07

2004 05 06 07-16

-12

-8

-4

0

4

-1.0

-0.5

0.0

0.5

1.0

1.5Large Euro East Asia 1

Source: IMF staff estimates. See footnote 15 in main text for definition of country groupings.

1

2004 05 06 07


this response will require considerable effortsfrom international policymakers.

As the current constellation of externalimbalances reflects a series of diverse and unre-lated regional shocks, a number of economicand policy developments will be required tounwind them. In particular, the results in thischapter indicate that steps to raise saving in theUnited States, boost growth in Japan andEurope, and increase investment in Asia—including completing ongoing financial andcorporate restructuring—and in oil-exportingcountries would all move global current accountimbalances in the right direction.

Finally, the analysis suggests that unusuallylow investment rates for this stage of the eco-nomic cycle have resulted in an excess supply ofsaving that may be contributing to the low levelof real long-term interest rates. This low invest-ment is largely a result of the still-ongoingefforts by corporates in many countries tostrengthen their balance sheets by paying downdebt. Consequently, despite strong corporateprofit growth, investment has generallyremained weak. The evolution of investment istherefore likely to be a critical factor determin-ing long-term interest rates going forward. Areturn of investment to a more normal cyclicalrelationship with growth would likely putupward pressure on interest rates.

Appendix 2.1. Sample Composition, DataSources, Methods, and ResultsThe main authors of this appendix are Marco Terronesand Roberto Cardarelli. Stephanie Denis providedresearch assistance.

This appendix provides details on the datasources, samples, and econometric methods andresults of the study discussed in the main text.

Sample and Data Sources

• The sample used in this chapter comprises thefollowing 46 industrial and emerging marketcountries. Industrial countries: Australia,


116

A 1/2 percent acceleration in real GDP growth in Japan would help reduce the U.S. current account deficit.

Figure 2.11. How Would the Current Account React to Changes in Japan's GDP Growth?(Percent of GDP)

1

2004 05 06 07-4

-3

-2

-1

0

1

2

3East Asia 1

-0.6

-0.3

0.0

0.3

0.6 Large Euro

2004 05 06 07

Source: IMF staff estimates. See footnote 15 in main text for definition of country groupings.1


2004 05 06 07-0.6

-0.4

-0.2

0.0

0.2

0.4

0.6

2004 05 06 07-0.6

-0.4

-0.2

0.0

0.2

0.4

0.6JapanUnited States

Austria, Belgium, Canada, Denmark, Finland,France, Germany, Greece, Ireland, Italy, Japan,the Netherlands, New Zealand, Norway,Portugal, Spain, Sweden, Switzerland, theUnited States, and the United Kingdom.Emerging markets: Argentina, Brazil, Chile,China, Colombia, Egypt, Hong Kong SAR,India, Indonesia, I.R. of Iran, Israel, Korea,Malaysia, Mexico, Morocco, Pakistan, Peru,the Philippines, Saudi Arabia, Singapore,South Africa, Thailand, Taiwan Province ofChina, Turkey, and Venezuela.23 The oil-exporting countries subgroup is defined ascomprising I.R. of Iran, Mexico, Norway, SaudiArabia, and Venezuela. The data are annualand cover 1970–2004.

• Data were taken from a variety of sources,including the Organization for EconomicCooperation and Development’s (OECD)Analytical Database, the Global FinancialDatabase, the IMF’s International FinancialStatistics and the World Economic Outlook,national authorities, the World DevelopmentIndicators from the World Bank, andWorldscope.

• The main series used in the chapter are asfollows.• Saving and investment rates. These series

were constructed using data from theOECD Analytical Database, WorldDevelopment Indicators, and the WorldEconomic Outlook.

• Interest rates. The short-term interest rateseries were mainly obtained from the GlobalFinancial Database and the IMF’s Interna-tional Financial Statistics.

• Investment deflator. This deflator was calcu-lated using the OECD Analytical Databaseand the World Economic Outlook.

• Private credit (by deposit money banks and otherfinancial institutions) to GDP ratio. Obtainedfrom the World Bank’s Financial StructureDevelopment Database.

APPENDIX 2.1. SAMPLE COMPOSITION, DATA SOURCES, METHODS, AND RESULTS

117

2004 05 06 07-0.2

-0.1

0.0

0.1

0.2

0.3

-0.2

-0.1

0.0

0.1

0.2


2004 05 06 07

Source: IMF staff estimates. See footnote 15 in main text for definition of country groupings.1

-0.2

-0.1

0.0

0.1

0.2

0.3

2004 05 06 07-3

-2

-1

0

1

2Large Euro East Asia 1

2004 05 06 07

Figure 2.12. How Would the Current Account React to Changes in the U.S. Real Interest Rate?(Percent of GDP)

1

A 200 basis point increase in real interest rates in the United States would have a minor effect on global current account imbalances.


23I.R. of Iran and Saudi Arabia are not usually includedamong the emerging market economies.

• Elderly and youth dependence ratios. Theseseries were obtained from the World Bank’sWorld Development Indicators.

• Corporate data series used in Box 2.3. Thecorporate data set used in Box 2.3 is based oninformation from nonfinancial publicly tradedfirms and reported in the Worldscope data-base. The data set includes information fornine emerging Asian economies—namely,Hong Kong SAR, India, Indonesia, Korea,Malaysia, the Philippines, Singapore, TaiwanProvince of China, and Thailand.24 The datawere then used to construct ratios of invest-ment (investment expenditures over bookvalue of plant, machinery, and equipment),leverage (short-term debt over market value ofequity), liquidity (cash flow over book value ofmachinery, capital, and equipment), andTobin’s q (market value of equity plus bookvalue of long-term debt over book value ofmachinery, capital, and equipment) by coun-try, using median values.

Dynamic Panel Model of Saving and Investment

Building on existing literature on saving andinvestment, IMF staff estimated a dynamic panelmodel for these two variables using data for the46 countries over the 1972–2004 period.

The model postulates that saving, in any givencountry (i) and year (t), is explained by the fol-lowing factors:

Sit = γSi(t–1) + Xitβ + νi + ηt + uit , (1)(i = 1, . . . ,N ; t = 2, . . . ,T)

where Sit is the saving rate for country i inperiod t; Xit is a matrix comprising informationon the explanatory variables for country i inperiod t; νi is a country-specific effect; and ηt is atime-specific effect (which would capture therest of the world trends). The right-hand-sidevariables included in the saving regression areof two types.

• Past saving rate, Si(t – 1). This term captures theextent to which saving rates are persistent—that is, the extent current saving rates are cor-related with past rates.

• Economic determinants of saving, Xit. The fol-lowing determinants were considered, with theexpected sign of the relationship between sav-ing and each determinant in parentheses: percapita output growth (+); real interest rate(+); credit to the private sector (–); terms-of-trade growth (+); public sector saving (+); andthe elderly and youth dependency ratios (–).For a discussion, see, for instance, Edwards(1995); Higgins (1998); Haque, Pesaran, andSharma (1999); Loayza, Schmidt-Hebbel, andServén (2000); and the May 1995 WorldEconomic Outlook.Similarly, the model postulates that invest-

ment, in any given country (i) and year (t), isexplained by the following factors:

Iit = γIi(t–1) + Xitβ + νi + ηt + uit, (2)(i = 1, . . . , N ; t = 2, . . . ,T)

where Iit is the investment rate for country i inperiod t, Xit is a matrix comprising informationon the explanatory variables for country i inperiod t, νi is a country-specific effect, and ηt is atime-specific effect (which would capture therest of the world trends). The right-hand-sidevariables included in the investment regressionare of two types.• Past investment rate, Ii(t–1). This term captures

the extent to which investment rates are per-sistent. If investment is a highly persistentprocess, higher investment rates today wouldbe associated with higher investment ratestomorrow.

• Economic determinants of investment, Xit.The following determinants were considered,with the expected sign of the relationshipbetween investment and each determinant inparentheses: per capita output growth (+);credit to the private sector (+); the cost of cap-


118

24Because investment in China has behaved differently from that in other countries in the region, China was excludedfrom the analysis.

ital (–)—measured as the ratio of real interestto the relative price of capital;25 and the eld-erly and youth dependency ratios (undeter-mined). For a discussion, see, for instance,Romer (1996); Higgins (1998); Auerbach(2002); and Pelgrin, Schich, and de Serres(2002).These regression equations were estimated

using the Generalized Method of Moments esti-mator with robust errors (to correct for hetero-geneity in the error term).

A Dynamic Factor Model

Dynamic factor models have become increas-ing popular among economists,26 as theydescribe the covariance or co-movementbetween a group of (observable) time series asthe result of the relationship between these vari-ables and a small number of unobservable vari-ables, known as factors. These unobservedfactors can be regarded as indexes of commonactivity—across the entire data set (e.g., globalactivity) or across subsets of the data (e.g., for aparticular region).

To estimate the factors, the dynamic factormodel decomposes each observable variable—forexample, the saving rate for the United States—into components that are common across allobservable variables or common across a subsetof variables. In particular, for each of the 12regions considered in the chapter, the analysisuses five observable variables: real output growth(g), real short-term interest rates (r), ratio of thecurrent account balance to GDP (CA), savingrate (S), and investment rate (I ).27 In the model,there are three types of factors: the common fac-tor (fW), 5 factors specific to each observed vari-able (f i, one per variable ) and 12 region-specificfactors (f j, one per region). So there are 60 (5 ×12) “regression” equations to be estimated, and60 time series to be explained by 18 (1 + 5 + 12)

factors. For example, for the United States theregressions estimated are:

yUSt = ay,US + bW

y,US f Wt + b y

US f yt + bUS

y f USt + εy

t,US

rUSt = ar,US + bW

r,US f Wt + br

US f rt + bUS

r f USt + εr

t,US

CAUSt = aCA,US + bW

CA,US f Wt + bCA

USf USt + bUS

CA f USt + εCA

t,US

SUSt = aS,US + bW

S,US f Wt + bS

US f St + bUS

S f USt + εS

t,US

IUSt = aI,US + bW

I,US f Wt + b I

US f It + bUS

I f USt + εI

t,US.

In this system, fW is the world factor, thecomponent common to all variables in allcountries—that is, every variable depends on thiscommon factor and that dependence variesacross each variable i and country j through theparameter bW

i,j, which is called the factor loading;f i is the global factor for variable i, capturing co-movement across the world in this variable thatis not explained by the world factor; and f US is aU.S.-specific factor, which captures co-movementacross all five variables within the United Statesthat is not captured by either type of global fac-tor. Finally, ε is the “unexplained” idiosyncraticerror.

The model captures dynamic co-movementby allowing the factors (fs) and idiosyncraticterms (ε) to be (independent) autoregressiveprocesses. That is, each factor depends on lagsof itself and an i.i.d. innovation to the variable:

ft = φ(L)ft + ut ,

where φ(L) is a lag polynomial and ut is normallydistributed. All the factor loadings (bs), and lagpolynomials are independent of each other.Because the factors are unobservable, specialmethods must be employed to estimate them;in the chapter, the model is estimated usingBayesian techniques as described in Kose, Otrok,and Whiteman (2003).

To measure the contribution of each factor tothe variation in the observable variables, thevolatility in each aggregate variable has beendecomposed into components due to each


119

25Because of lack of information on taxes, the measure of cost of capital used in this chapter takes into account neitherdifferences in taxes across countries nor tax changes in a given country.

26See, for example, Stock and Watson (2002) and Forni, Lippi, and Reichlin (2004).27Before estimating the dynamic factor model, the observables were detrended.


120

Table 2.5. Variance Decomposition for Country Regions(Percent)

Short-Term Current Saving InvestmentOutput Real Interest Account Rate Rate

Region Factor (y) Rate (r) (CA) (s) (i)

United States Global 82.2 53.3 53.8 83.9 44.5World 7.0 0.2 3.9 11.1 22.7Aggregate 75.2 53.1 49.9 72.8 21.8

Region and idiosyncratic 17.5 46.5 45.9 11.2 54.8

Japan Global 25.2 69.9 60.4 29.3 52.6World 6.0 24.8 2.9 13.9 21.4Aggregate 19.1 45.1 57.6 15.3 31.2


Anglo-Saxon Global 74.8 93.8 4.0 55.3 53.4World 20.1 0.5 0.2 54.8 45.5Aggregate 54.7 93.3 3.9 0.5 7.9


Other Industrial Global 57.5 85.4 18.5 91.3 69.0World 42.6 21.2 18.3 60.3 38.8Aggregate 14.9 64.3 0.1 30.9 30.2


Large Euro Global 29.4 90.5 42.5 81.6 63.2World 23.2 4.4 9.9 68.5 40.8Aggregate 6.2 86.1 32.5 13.0 22.3


Small Euro Global 71.3 86.4 11.5 85.0 69.2World 62.5 7.8 8.2 63.7 46.0Aggregate 8.7 78.6 3.3 21.4 23.2


East Asia 1 Global 2.5 49.1 37.6 9.4 21.2World 0.3 16.5 34.0 0.6 17.1Aggregate 2.2 32.6 3.6 8.8 4.0


East Asia 2 Global 35.7 19.2 97.2 6.5 1.5World 0.1 16.0 0.3 2.5 0.9Aggregate 35.6 3.3 96.9 4.0 0.6


China Global 29.6 3.4 13.3 3.7 14.3World 19.4 0.4 3.7 2.5 6.7Aggregate 10.3 2.9 9.6 1.2 7.6


Other Emerging Markets Global 2.3 9.1 36.2 7.2 21.4World 2.2 3.5 1.5 6.5 8.9Aggregate 0.1 5.6 34.7 0.7 12.5


Latin America Global 7.5 33.3 33.8 3.2 1.3World 4.1 4.7 0.3 0.4 0.3Aggregate 3.4 28.6 33.4 2.7 1.1


Oil Producers Global 23.6 23.1 18.4 14.2 1.7World 23.5 1.8 5.8 7.9 1.2Aggregate 0.2 21.3 12.5 6.3 0.5


Average Global 36.8 51.4 35.6 39.2 34.4World 17.6 8.5 7.4 24.4 20.9Aggregate 19.2 42.9 28.2 14.8 13.6


Source: IMF staff calculations.

factor. Table 2.5 reports the median of the poste-rior distribution of the variance decompositionsfor each region. For example, the variance ofthe saving rate for the United States is

var(SUSt ) = (bW

US)2 var(f Wt ) + (bS

US)2 var(f St )

+ (bSUS)2 var(f US

t ) + var(εSt,US),

and, therefore, the variance in U.S. saving ratesattributable to the world factor is

(bWUS)2 var(f W

t )––––––––––––.var(SUS

t )

Steady-State Level of Investment and CapitalOutput Ratio for Emerging Asia

Steady-state levels of investment rates are esti-mated based on standard neoclassical growthmodels, and thus as a function of the steady-statecapital-output ratio, the depreciation rate, andthe trend growth rate of output:

k* (g + d)i* = –––––––––,

(1 + g)

where i* is the (steady-state) ratio of real grossfixed investment to real output, k* is the (steady-state) capital-output ratio, g is potential outputgrowth, and d is the depreciation rate. Thus, i* isa positive function of k*, g, and d. Capital stockis estimated based on a standard perpetualinventory method

Kt = Kt–1(1 – d) + It.

Two depreciation rates are used: 5 percentand 7 percent. With data on gross fixed realinvestment starting from 1950 (obtained usingthe Penn World Table Version 6.0), the initialestimate of capital stock is obtained assumingthat the country is at steady-state capital-output ratio in 1950. To obtain this ratio, theaverages of k, g, and d over 1950–60 are used.However, different periods and parametervalues are used to test the sensitivity of the capi-tal stock to the choice of its initial value, andthe results show that the guess at the initialcapital stock becomes relatively unimportantdecades later (Easterly and Levine, 2001; and


121

Table 2.6. Investment and Capital Ratios for Emerging Asia

Steady State 1996 2004_______________________________________ ____________________ ____________________Capital- Potential real Investment Investment Capital Investment Capital

output ratio growth rate ratio ratio ratio ratio ratio

Depreciation rate at 5 percent

China 2.5 5.0 23.8 34.4 2.2 45.6 3.0Hong Kong SAR 2.7 4.0 23.1 31.3 2.6 22.4 2.9India 2.5 5.0 23.8 25.4 2.2 23.5 2.4Indonesia 2.5 5.0 23.8 23.3 1.9 21.0 2.4Korea 2.8 4.7 25.5 37.5 2.7 28.6 3.1Malaysia 3.1 5.0 29.5 42.5 3.0 20.5 3.3Philippines 2.7 4.0 23.3 24.0 2.7 17.0 2.6Singapore 2.9 4.5 26.3 38.0 2.7 24.0 3.1Taiwan Province of China 2.5 5.0 23.8 22.5 1.7 20.1 2.0Thailand 3.0 5.0 28.7 41.1 3.0 25.3 3.0

Depreciation rate at 7 percent

China 2.1 5.0 24.3 34.4 2.0 45.6 2.7Hong Kong SAR 2.3 4.0 24.2 31.3 2.2 22.4 2.5India 2.0 5.0 22.9 25.4 1.8 23.5 2.0Indonesia 2.0 5.0 22.9 23.3 1.7 21.0 2.0Korea 2.4 4.7 27.1 37.5 2.4 28.6 2.6Malaysia 2.7 5.0 30.9 42.5 2.7 20.5 2.8Philippines 2.2 4.0 23.4 24.0 2.2 17.0 2.1Singapore 2.5 4.5 27.7 38.0 2.4 24.0 2.6Taiwan Province of China 2.0 5.0 22.9 22.5 1.5 20.1 1.7Thailand 2.6 5.0 29.7 41.1 2.6 25.3 2.4


Klenow and Rodriguez-Clare, 2001, adopt a sim-ilar methodology).

For a given depreciation rate, the steady statecapital stock ratio, k*, for each country is foundas the maximum value of the capital-output ratioon average over long (15- and 20-year) subperi-ods between 1950 and 2004. However, whensuch capital-output ratio is lower than 2.5 (2) ifcalculated with a 5 (7) percent depreciationrate, it is set to this level to minimize the disper-sion of capital ratios across countries and reducepossible measurement errors. Indeed, the steady-state capital-output ratios are in the range of 2 to3.5 (see Table 2.6), a reasonable estimate for thisgroup of countries (the average for industrialcountries over the 1970–2004 period is slightlybelow 3). Potential growth rates are obtainedfrom the World Economic Outlook database, butare capped at 5 percent a year when above thisvalue.

ReferencesAitken, Brian, 2005, “The Historical Dominance of

Banks in China’s Financial System,” IMF WorkingPaper (Washington: International Monetary Fund,forthcoming).

Arellano, Manuel, and Stephen Bond, 1991, “SomeTests of Specification for Panel Data: Monte CarloEvidence and an Application to EmploymentEquations,” Review of Economic Studies, Vol. 58(April), pp. 277–97.

Auerbach, Alan J., 2002, “Investment,” in The ConciseEncyclopedia of Economics, ed. by David R. Henderson(Indianapolis: Liberty Fund, Inc.). Available via theInternet: http://www.econlib.org/library/Enc/Investment.html.

Barnett, Steven, and Ray Brooks, 2005, “What’s DrivingInvestment in China,” IMF Working Paper (Washing-ton: International Monetary Fund, forthcoming).

Bernanke, Ben, 2005, “The Global Saving Glut andthe U.S. Current Account Deficit,” remarks at theSandridge Lecture, Virginia Association ofEconomics, Richmond, Virginia.

———, Jean Boivin, and Piotr Eliasz, 2005,“Measuring the Effects of Monetary Policy: AFactor-Augmented Vector Autoregressive (FAVAR)Approach,” Quarterly Journal of Economics, Vol. 120(January), pp. 387–422.

Blanchard, Olivier, and Stanley Fischer, 1989, Lectureson Macroeconomics (Cambridge, Massachusetts: MITPress).

Bosworth, Barry P., and Benjamin Keys, 2004,“Increased Life Expectancy: A Global Perspective,”in Coping with Methuselah: The Impact of MolecularBiology on Medicine and Society, ed. by Henry Aaronand William Schwartz (Washington: BrookingsInstitution Press), pp. 247–83.

Boyreau-Debray, Genevieve, and Shang-Jin Wei, 2005,“Pitfalls of a State-Dominated Financial System: TheCase of China,” NBER Working Paper No. 11214(Cambridge, Massachusetts: National Bureau ofEconomic Research).

Bryant, Ralph, 2004, “Cross-Border MacroeconomicImplications of Demographic Change,” in GlobalDemographic Change: Economic Impacts and PolicyChallenges, Jackson Hole, Wyoming, August 26–28,2004, a Symposium Sponsored by the Federal ReserveBank of Kansas City, ed. by Gordon H. Sellon, Jr.(Kansas City: Federal Reserve Bank of KansasCity).

Chamon, Marcos, and Eswar Prasad, 2005, “Determi-nants of Household Saving in China,” IMF WorkingPaper (Washington: International Monetary Fund,forthcoming).

Deaton, Angus, 1992, Understanding Consumption,Clarendon Lectures on Economics (Oxford:Oxford University Press).

Dees, Stephane, Filippo di Mauro, Hashem Pesaran,and Vanessa Smith, 2005, “Exploring the Inter-national Linkages of the Euro Area: A Global VARAnalysis,” CESifo Working Paper No. 1425 (Munich:CESifo).

De Mello, Luiz, Per Mathis Kongsrud, and RobertPrice, 2004, “Saving Behaviour and the Effective-ness of Fiscal Policy,” OECD Economics Depart-ment Working Paper No. 397 (Paris: Organizationfor Economic Cooperation and Development).

Easterly, William, and Ross Levine, 2001, “It’s NotFactor Accumulation: Stylized Facts and GrowthModels,” World Bank Economic Review, Vol. 15, No. 2,pp. 177–219.

Edwards, Sebastian, 1995, “Why Are Saving Rates SoDifferent Across Countries? An InternationalComparative Analysis,” NBER Working PaperNo. 5097 (Cambridge, Massachusetts: NationalBureau of Economic Research).

Erceg, Christopher, Luca Guerrieri, and ChristopherGust, 2005, “Expansionary Fiscal Shocks and theTrade Deficit,” International Finance Discussion


122

Paper No. 825 (Washington: Board of Governors ofthe Federal Reserve System).

Fama, Eugene, and Kenneth French, 1999, “TheCorporate Cost of Capital and the Return onCorporate Investment,” Journal of Finance, Vol. 54(December), pp. 1939–67.

Faulkner-MacDonagh, Chris, and Martin Mühleisen,2004, “Are U.S. Households Living Beyond TheirMeans?” Finance and Development, Vol. 41, No. 1,pp. 36–9.

Fehr, Hans, Sabine Jokisch, and Laurence Kotlikoff,2004, “The Role of Immigration in Dealing with theDeveloped World’s Demographic Transition,”NBER Working Paper No. 10512 (Cambridge,Massachusetts: National Bureau of EconomicResearch).

Feldstein, Martin, and Charles Horioka, 1980,“Domestic Saving and International Capital Flows,”Economic Journal, Vol. 90 (June), pp. 314–29.

Forni, Mario, Marco Lippi, and Lucrezia Reichlin,2004, “Opening the Black Box: Structural FactorModels Versus Structural VARs,” CEPR DiscussionPaper No. 4133 (London: Centre for EconomicPolicy Research).

Gale, William, and Peter Orszag, 2004, “BudgetDeficits, National Saving, and Interest Rates,”Brookings Papers on Economic Activity: 2, BrookingsInstitution, pp. 101–210.

Glick, Reuven, and Kenneth Rogoff, 1995, “GlobalVersus Country-Specific Productivity Shocks and theCurrent Account,” Journal of Monetary Economics,Vol. 35 (February), pp. 159–92.

Gregory, Allan, and Allen Head, 1999, “Common andCountry-Specific Fluctuations in Productivity,Investment, and the Current Account,” Journalof Monetary Economics, Vol. 44 (December),pp. 423–51.

Haque, Nadeem, Hashem Pesaran, and Sunil Sharma,1999, “Neglected Heterogeneity and Dynamics inCross-Country Savings Regressions,” IMF WorkingPaper 99/128 (Washington: International MonetaryFund).

Helliwell, John F., 2004, “Demographic Changes andInternational Factor Mobility,” in Global DemographicChange: Economic Impacts and Policy Challenges,Jackson Hole, Wyoming, August 26–28, 2004, aSymposium Sponsored by the Federal Reserve Bank ofKansas City, ed. by Gordon H. Sellon, Jr. (KansasCity: Federal Reserve Bank of Kansas City).

Higgins, Matthew, 1998, “Demography, NationalSavings, and International Capital Flows,”

International Economic Review, Vol. 39 (May),pp. 343–69.

International Monetary Fund, 2004, “People’sRepublic of China: Article IV Consultation—StaffReport, Staff Statement, and Public InformationNotice on the Executive Board Discussion,” IMFCountry Report No. 04/351 (Washington:International Monetary Fund).

JPMorgan Chase & Co., 2005, “Corporates Are Drivingthe Global Saving Glut,” Global Issues (London:JPMorgan Research, June 24).

Jaeger, Albert, 2003, “Corporate Balance SheetRestructuring and Investment in the Euro Area,”IMF Working Paper 03/117 (Washington:International Monetary Fund).

Klenow, Peter, and Andrés Rodriguez-Clare, 1997,“The Neoclassical Revival in Growth Economics:Has It Gone Too Far?” NBER Macroeconomics Annual1997 (Cambridge, Massachusetts: MIT Press),pp. 73–102.

Klingen, Christoph, Beatrice Weder, and JerominZettelmeyer, 2004, “How Private Creditors Fared inEmerging Debt Markets, 1970–2000,” IMF WorkingPaper 04/13 (Washington: International MonetaryFund).

Kose, Ayhan, Christopher Otrok, and CharlesWhiteman, 2003, “International Business Cycles:World, Region, and Country-Specific Factors,”American Economic Review, Vol. 93 (September),pp. 1216–39.

———, 2004, “Understanding the Evolution of theWorld Business Cycle” (unpublished).

Kumhof, Michael, Douglas Laxton, and Dirk Muir,2005, “Consequences of U.S. Fiscal Consolidationfor the Current Account,” Chapter IV in “UnitedStates: Selected Issues,” IMF Country ReportNo. 05/258 (Washington: International MonetaryFund).

Lee, Jong-Wha, Warwick McKibbin, and Yung ChulPark, 2004, “Transpacific Trade Imbalances: Causesand Cures,” Brookings Discussion Papers inInternational Economics, No. 162 (August).

Loayza, Norman, Klaus Schmidt-Hebbel, and LuisServén, 2000, “Saving in Developing Countries: AnOverview,” World Bank Economic Review, Vol. 14(September), pp. 393–414.

Lucas, Robert, 1990, “Why Doesn’t Capital Flow fromRich to Poor Countries?” American Economic Review:Papers and Proceedings, Vol. 80 (May), pp. 92–96.

Masson, Paul, Tamim Bayoumi, and Hossein Samiei,1995, “Saving Behavior in Industrial and

REFERENCES

123

Developing Countries,” Staff Studies for the WorldEconomic Outlook (Washington: InternationalMonetary Fund).

Miles, David, 1999, “Modelling the Impact ofDemographic Change Upon the Economy,”Economic Journal, Vol. 109 (January), pp. 1–36.

Modigliani, Franco, and Shi Larry Cao, 2004, “TheChinese Saving Puzzle and the Life-CycleHypothesis,” Journal of Economic Literature, Vol. 42(March), pp. 145–70.

Pelgrin, Florian, Sebastian Schich, and Alain deSerres, 2002, “Increases in Business InvestmentRates in OECD Countries in the 1990s: How MuchCan Be Explained by Fundamentals?” OECDWorking Paper No. 327 (Paris: Organization forEconomic Cooperation and Development).

Podpiera, Richard, 2005, “Progress in BankingReform,” IMF Working Paper (Washington:International Monetary Fund, forthcoming).

Poterba, James, 1987, “Tax Policy and CorporateSaving,” Brookings Papers on Economic Activity: 2,Brookings Institution, pp. 455–503.

———, 2004, “The Impact of Population Aging onFinancial Markets,” NBER Working Paper No. 10851(Cambridge, Massachusetts: National Bureau ofEconomic Research).

Prasad, Eswar, and Raghuram Rajan, 2005, “China’sFinancial-Sector Challenge,” op-ed article, FinancialTimes (London), May 10.

Romer, David, 1996, Advanced Macroeconomics (NewYork: McGraw-Hill).

Roubini, Nouriel, and Brad Setser, 2005, “Will theBretton Woods 2 Regime Unravel Soon? The Riskof a Hard Landing in 2005–2006,” paper presentedat the symposium of the Federal Reserve Bank ofSan Francisco and the University of CaliforniaBerkeley, “Revived Bretton Woods System: A NewParadigm for Asian Development?” San Francisco,February 4.

Sachs, Jeffrey, and Steven Radelet, 1998, “The Onsetof the East Asian Financial Crisis,” NBER WorkingPaper No. 6680 (Cambridge, Massachusetts:National Bureau of Economic Research).

Schmidt-Hebbel, Klaus, and Luis Servén, 1999, “SavingAcross the World” (Washington: World Bank).

Stock, James, and Mark Watson, 2002,“Macroeconomic Forecasting Using DiffusionIndexes,” Journal of Business and Economic Statistics,Vol. 20 (April), pp. 147–62.

Tevlin, Stacey, and Karl Whelan, 2003, “Explainingthe Investment Boom of the 1990s,” Journal of Money,Credit and Banking, Vol. 35 (February), pp. 1–22.


124

Date post:	12-Aug-2020
Category:	Documents
Upload:	others
View:	0 times
Download:	0 times

GLOBAL IMBALANCES: A SAVING AND CHAPTER II INVESTMENT...

Documents