124 - IMF eLibrary - International Monetary Fund

OCCASIONAL PAPER

Edited by Ulrich Baumgartner and Guy Meredith, with a staff teamcomprising Juha Kahkonen, Kenneth Miranda,

Garry J. Schinasi, and Alexander W. Hoffmaister

INTERNATIONAL MONETARY FUND

124

Analytical Studies

Saving Behavior and theAsset Price "Bubble" in Japan

Washington DC

April 1995

©International Monetary Fund. Not for Redistribution

© 1995 International Monetary FundReprinted June 1995

Library of Congress Cataloging-in-Publication Data

Saving behavior and the asset price "bubble" in Japan : analytical studies /Ulrich Baumgartner, Guy Meredith ; with a staff team comprising JuhaKahkonen . . . [et a l ] .

p. cm. — (Occasional Papers, ISSN 0251-6365 ; no. 124)Includes bibliographical references.ISBN 1-55775-462-41. Saving and investment—Japan. 2. Stocks—Prices—Japan.

3. Financial crises—Japan. 4. Japan—Economic conditions—1945-1989. I. Baumgartner, Ulrich, 1946- . II. Meredith, Guy.III. Kahkonen, Juha. IV. Series: Occasional paper (InternationalMonetary Fund); no. 124.HC465.S3S25 1995332.63'222'0952—dc20 95-10141

CIP

Price: US$15.00(US$12.00 to full-time faculty members and

students at universities and colleges)

Please send orders to:International Monetary Fund, Publication Services

700 19th Street, N.W., Washington, D.C. 20431, U.S.A.Tel.: (202) 623-7430 Telefax: (202) 623-7201

Internet: [email protected]

recycled paper


Contents

Preface

I. OverviewUlrich Baumgartner and Guy Meredith

II. Does Japan Save Too Much?Kenneth Miranda

How Much Does Japan Save?Recent Trends in National SavingJapanese Saving in an International Context

How Much Should Japan Save?"Golden Rule" of AccumulationDynamic EfficiencyMarginal Productivity

ConclusionsReferences

III. Japan's Capital FlowsJuha Kahkonen

Recent Developments in the Capital AccountLong-Term Capital FlowsShort-Term Capital FlowsFlow of Financial Resources to Developing Countries

Foreign Direct InvestmentConcept and MeasurementMain DevelopmentsWorldwide PerspectiveDeterminants of Japanese Outward

Foreign Direct InvestmentCharacteristics of Japanese Overseas SubsidiariesEffects of Japanese Foreign Direct Investment

on Other EconomiesEffects on Japan

References

IV. Demographic Change and Household Saving in JapanGuy Meredith

Previous Studies on Demographics and SavingEvidence Supporting the Life-Cycle ModelChallenges to the Life-Cycle Model

Measurement of Saving in Household DataSimulation Using a Life-Cycle ModelConclusions

Pagevii

1

4

445568

121414

16

1619232324252628

2932

323334

36

363638383941

iii©International Monetary Fund. Not for Redistribution

CONTENTS

Appendix 4-1. Effects of Demographic Change on SavingWith and Without Social Security

Appendix 4-2. A Life-Cycle Model of Household Behaviorfor Japan

References

V. Alternative Long-Run ScenariosGuy Meredith

Population Dynamics and Output GrowthAlternative Fiscal ScenariosImplications for the Saving-Investment BalanceReferences

VI. Movements in Asset Prices Since the Mid-1980sJuha Kahkonen

Developments Since the Mid-1980sCauses of Asset Price Inflation and Deflation

Role of FundamentalsChanges in the Financial EnvironmentWas There a Bubble?

Impact of Asset Price Movements on the EconomyReferences

VII. Asset Prices, Financial Liberalization, and Inflationin JapanAlexander W. Hoffmaister and Garry J. Schinasi

Key Developments and IssuesAsset Price and Balance Sheet AdjustmentsAlternative Explanations and the Role of Monetary FactorsCasual Empirical Evidence and Unresolved Issues

A Multiequation Model of Land Prices and InflationA Vector-Autoregression ModelEmpirical Results

ConclusionsAppendix 7-1. Vector-Autoregression Modeling

Time-Series PropertiesLag LengthStructural Breaks

References

Tables

2-1. Saving Rates of Major Industrial Countries2-2. Optimal Gross Saving Rates Under Alternative Parameter Values2-3. Private Sector Dynamic Efficiency2-4. Aggregate Dynamic Efficiency2-5. Return to Capital

3-1. Summary of the Capital Account3-2. Flows of Securities

42

4344

46

46474950

51

51545456565862

63

63646566676768757676767677

57

101113

1820

iv


V

3-3. Long-Term Capital Outflow by Region and Type3-4. Long-Term Capital Inflow by Region and Type3-5. Net Long-Term Capital Outflow by Region and Type3-6. Net Flow of Financial Resources to Developing Countries

and Multilateral Agencies from Japan3-7. Foreign Direct Investment by Region and Industry

4-1. Summary of Studies on Demographics and Saving4-2. Income, Consumption, and Saving of Retired Households4-3. Effect of a Demographic Transition on Aggregate Saving

6-1. Asset Price Developments6-2. Consolidated Balance Sheet of Household Sector6-3. Consolidated Balance Sheet of Nonfinancial Incorporated Enterprises

7-1. Selected Indicators of Financial Balance7-2. F-Tests for Lagged Values of Explanatory Variables:

Broad Money (M2 + CDs)7-3. F-Tests for Lagged Values of Explanatory Variables:

Credit Consistent with National Accounts7-4. Variance Decompositions for the Land Price and Consumer Price

Equations: Monetary Variables First in the Ordering7-5. Variance Decompositions for the Land Price and Consumer Price

Equations: Monetary Variables Last in the Ordering7-6. Unit Root Tests7-7. Lag Length Tests7-8. Likelihood Ratio Tests for Structural Breaks

Charts

2-1. Gross and Net National Saving Rates2-2. Gross National Saving Rates of Major Industrial Countries2-3. Dynamic Efficiency

3-1. Current Account Balance and Capital Flows3-2. Foreign Assets and Liabilities3-3. Outward Foreign Direct Investment3-4. Japanese and World Direct Investment Outflows3-5. Direct Investment Outflows Among Five Major Industrial Countries3-6. Direct Investment Inflows Among Five Major Industrial Countries3-7. Direct Investment Income

5-1. Long-Term Demographic and Output Projections5-2. Long-Term Fiscal Projections: Pre-Reform Program Parameters5-3. Long-Term Fiscal Projections: Pre-Reform Program Parameters

Versus Pension Reform5-4. Long-Run Saving-Investment Balances: Pre-Reform Program

Parameters Versus Pension Reform

6-1. Movements in Asset Prices6-2. Stock Price Developments6-3. Land Price Developments6-4. Asset Prices and "Fundamentals"6-5. Actual and Predicted Stock Prices6-6. Actual and Predicted Land Prices6-7. Household Saving and Net Wealth

212224

2628

374043

535961

65

71

72

72

73767777

459

17172729293034

4748

49

50

52525455575860

Contents


The following symbols have been used throughout this paper:

... to indicate that data are not available;

— to indicate that the figure is zero or less than half the final digit shown, or that the item doesnot exist;

- between years or months (for example, 1991-92 or January-June) to indicate the years or monthscovered, including the beginning and ending years or months;

/ between years or months (for example, 1991/92) to indicate a crop or fiscal (financial) year.

"Bil l ion" means a thousand million; "trillion" means a thousand billion.

Minor discrepancies between constituent figures and totals are due to rounding.

The term "country," as used in this paper, does not in all cases refer to a territorial entity that is a stateas understood by international law and practice; the term also covers some territorial entities that arenot states, but for which statistical data are maintained and provided internationally on a separate andindependent basis.

vi

CONTENTS

7-1. Urban Land and Stock Prices7-2. Total Private Nonfinancial Sector Debt7-3. Household Sector Balance Sheet7-4. Money, Debt, and Inflation7-5. Asset Prices and GDP Deflator7-6. Predictions of Land Price Inflation Using Private Credit

as the Financial Aggregate7-7. Predictions of Consumer Price Inflation Using Private Credit

as the Financial Aggregate7-8. Dynamic Simulations Using Two-Equation Subsystem and

M2 + CDs as the Financial Aggregate7-9. Dynamic Simulations Using Two-Equation Subsystem and Private

Credit as the Financial Aggregate: Urban Land Prices7-10. Dynamic Simulations Using Two-Equation Subsystem and Private

Credit as the Financial Aggregate: Consumer Prices7-11. Impulse Response to M2 + CDs7-12. Impulse Response to Private Credit

6464646767

69

70

71

73

747575


Preface

This Occasional Paper is based on research papers prepared in connection with the IMF'srecent Article IV consultations with Japan. The authors are grateful for the cooperation ofthe Japanese authorities in the preparation of these studies. They would also like to thankBijan B. Aghevli for his encouragement and support of the project, and Daniel Citrin,Yusuke Horiguchi, Kenji Okamura, Chikahisa Sumi, and Christopher Towe for their valuablecomments. Thanks are also due to Harish Mendis and Toh Kuan for research assistance,and to Elizabeth Elliott, Prabha Job, and Tammy Shear for secretarial support. JamesMcEuen of the External Relations Department edited the paper for publication and coordi-nated production. The views expressed here, as well as any errors, are the sole responsibilityof the authors and do not necessarily reflect the opinions of the Government of Japan,Executive Directors of the IMF, or other members of the IMF staff.

vii


This page intentionally left blank


I OverviewUlrich Baumgartner and Guy Meredith

This volume brings together various analytical stud-ies the IMF staff has undertaken on the Japanese

economy, focusing on two areas of particular interestfor both longer-term economic performance and recentcyclical developments. The first is Japan's saving behav-ior; the second is the remarkable swing in asset pricesthat occurred in the late 1980s and early 1990s. As regardssaving, Japan for many years has had the highest nationalsaving rate among the major industrial countries. Whileits domestic investment rate has also been higher thanthat of other major industrial countries, the excess ofsaving over investment has been reflected in Japan'ssignificant current account surplus since the early 1980s.Japan's saving performance raises several issues. DoesJapan save "too much" on economic grounds? How hasJapan's saving been channeled abroad? What effect willan aging population have on future saving patterns ofthe private and public sectors? Answers to these questionsare the focus of Sections II through V of this volume.

Section II considers whether Japan's saving can becharacterized as excessively high. To address this issue,Kenneth Miranda invokes three criteria—based on"golden rule," dynamic efficiency, and marginal produc-tivity conditions—to shed light on whether a country isover- or undersaving. The golden-rule approach allowscalculation of the optimal level of saving based on param-eters for productivity and population growth, capital de-preciation, the capital share in output, and the social rateof time preference. In practice, the rate of time preferencecannot be observed directly but is inferred from observedsaving data and the other parameters. The evidence isconsistent with a rate of time preference ranging from1/2 of 1 percent to 2 percent a year, suggesting that theassumption of abnormal rates of time preference is notneeded to explain Japanese saving. The dynamic effi-ciency approach, in contrast, provides a nonparametrictest of whether a country is oversaving: if investment,on average, exceeds the share of profits in output, thenlong-run consumption possibilities could be increasedby reducing saving. For Japan, the investment rate hasaveraged less than the share of profits in gross nationalproduct (GNP) since the mid-1970s, indicating that theeconomy has operated in a dynamically efficient range.Finally, marginal productivity tests assess whether thereturn to capital has exceeded alternative estimates of

the opportunity cost of capital: again, results suggest thatcapital is unlikely to have been overaccumulated. A l l ofthese indicators of excess saving are subject to caveats,but they provide a broad range of evidence against theproposition that Japanese saving has been inefficientlyhigh.

Section III, by Juha Kahkonen, analyzes developmentsin Japan's capital flows since 1980. Japan has been theworld's largest exporter of capital over this period, de-spite having the highest domestic investment rate amongmajor industrial countries. Trends in Japan's capital ac-count since 1980 can be divided into three phases. In thefirst half of the 1980s, the liberalization of external capitalflows coincided with a surge in long-term capital out-flows, especially of portfolio investment. In the secondhalf of the decade, further steps toward financial liberal-ization—both external and domestic—as well as thesurge in asset prices led to a further rise in long-termcapital outflow, financed in part by short-term externalborrowing. The share of foreign direct investment (FDI)in Japan's capital exports increased sharply during thisperiod because appreciation of the yen made overseasinvestment more attractive. These developments werereversed in the early 1990s. The collapse of the asset pricebubble and the introduction of Bank for InternationalSettlements capital-adequacy guidelines led to a sharpretrenchment by financial institutions—short-term for-eign liabilities were repaid by sales of long-term assets,causing long-term capital exports to fall. At the sametime, the economic downturn reduced corporate profitsand led to a sharp rise in excess domestic capacity, reduc-ing the attractiveness of FDI. The section continues witha more detailed exploration of the composition of Japan'sFDI and its effects on other countries. Both structuralfactors (such as evolving comparative advantage) andmacroeconomic developments have played importantroles in driving FDI. The benefits to other countriesconsist not only of the usual gains from internationalintegration, but also of positive spillovers in the form ofnew technology and organizational skills.

Sections IV and V deal with the potential impact ofan aging population on Japan's saving rate. In SectionIV Guy Meredith assesses the effect on saving behaviorat the household level, and in Section V he presents asimulation analysis of the impact on aggregate public

1


and private saving. Population aging is an important issuefor Japan because of the projected size of the demo-graphic shift in coming years: during 1990-2020, Japanis expected to experience the sharpest increase in its "old-age dependency ratio" (the ratio of the elderly to theworking-age population) among the major industrialcountries. As discussed in Section IV, if households be-have in accordance with life-cycle consumption theory,private saving will fall as the share of the elderly—whoare net dissavers—in the population rises in relation tothat of other segments of the population. The magnitudeof this effect has been questioned by some observers,however, because some microeconomic data suggestthat the saving rate of the elderly is not significantlydifferent from that of working-age households. Theanalysis in this section shows that, when properly mea-sured, the saving rate of the elderly is indeed well belowthat of working-age households. Furthermore, shifts indemographic structure will generate changes in the savingrates of individual population cohorts through their effecton public pension contribution and benefit rates. Theoverall impact of population aging on the aggregate sav-ing rate is quantified through simulations of a life-cyclemodel of household behavior. The results suggest thatthe household saving rate will decline significantly as aresult of population aging, although the drop may notbe as large as some studies using macroeconomic datahave indicated.

Section V provides a long-run simulation analysis ofthe effects of population aging. Particular emphasis isplaced on the impact of rising pension benefits and in-creased health care on public sector saving. In this con-text, Japan is often regarded as having a relatively healthyfiscal situation, in large part because of the substantialsurplus in the social security system. The analysis shows,however, that, in the absence of pension reform, a socialsecurity deficit would emerge by early in the next decade;by 2025, the deficit on social security could amount toalmost 15 percent of gross domestic product (GDP).Combined with the government deficit on operations notrelated to the social security system, this implies a sub-stantial financing "gap" that will have to be filled by acombination of pension reform, higher taxes, and reducedspending on other government operations—or a combi-nation of these—if an explosive rise in government debtis to be avoided. Part of this gap could be filled by theenactment of a tax reform package that would offset theloss in revenues from the 1994 income tax cut. In addi-tion, the implementation of recent reforms to the pensionsystem will play an important role in addressing the fiscalimbalance. Nevertheless, simulations show that furthermeasures will need to be taken to put Japan's fiscalposition on a sustainable footing. From the point of viewof the overall economy, a sharp drop in public saving,combined with a demographically induced decline inprivate saving, would cause a sharp swing in the externalsurplus in coming years.

The second area analyzed in this volume is asset pricedevelopments, during both the 1987-90 "bubble" periodand the subsequent economic downturn. The boom inJapan's equity and land prices during the bubble wasaccompanied by an extraordinary surge in private spend-ing, especially on business investment. Output rose wellabove its supply capacity, leading to inflationary pres-sures in product markets, while the current account sur-plus declined sharply in relation to GDP. In the event,both the inflated level of asset prices and the boom inspending were abruptly reversed in the early 1990s, caus-ing one of the longest and deepest recessions in postwarhistory. These developments and their effects are ana-lyzed in Sections VI and VII.

Section VI, by Juha Kahkonen, describes the evolutionof asset prices during and after the bubble period, assessesthe role played by various factors in driving asset pricesduring this episode, and provides estimates of their effecton real activity. On the basis of equations summarizingthe behavior of equity and land prices in the periodpreceding the bubble, it is it is shown that changes in"fundamentals"—such as economic growth rates, mone-tary policy, and corporate earnings—can explain at leastpart of the surge in asset prices during the late 1980s.Greater risk taking, owing to changes in the financialenvironment, and distortions in Japan's land tax systemmay also have played contributing roles. But there re-mains an unexplained component of asset price move-ments, consistent with the view that speculative forcescarried prices beyond the level consistent with underlyingdeterminants. In terms of private spending, an importantchannel was the effect on investment of increased marketvaluation of the capital stock, which could have raisedcapital spending by about 10 percent in the late 1980s.Higher land prices may also have eased corporate bor-rowing constraints by raising collateral values. The effecton private consumption was smaller, since Japanesehouseholds are not large holders of corporate equities,and changes in land prices are estimated to have only aminor effect on consumption. Similarly, the surge inland prices does not appear to have driven the boom inresidential construction.

In Section VII, Alex Hoffmaister and Garry Schinasiexamine the relationship between macroeconomic vari-ables and asset price inflation in the 1980s. The focus ison land price inflation, rather than on stock price move-ments; although stock prices are difficult to model empiri-cally, land prices generally move systematically in re-sponse to changes in economic fundamentals. Severalrelated questions are investigated: (1) whether there wasa structural break in the way monetary factors affectedasset prices in the 1980s; (2) whether monetary factorscontributed to asset price inflation in important waysin the 1980s; (3) factors responsible for the divergentbehavior of asset prices and consumer prices; and (4)whether there is support for the view that the effects ofmonetary factors were "concentrated" in asset markets

2

I OVERVIEW


rather than in goods markets. Strong evidence is foundfor a shift in the relationships between monetary factorsand land prices in the early 1980s: in particular, theparameters imply a much more important channel oftransmission in the second part of the sample than in thefirst. A key conclusion is that monetary shocks led tomore asset price inflation (and less consumer price infla-

tion) in the 1984-93 period than during 1970-83. This"regime shift"—which is largely attributed to the effectsof financial liberalization—made it difficult to interpretthe factors underlying the rise in asset prices in the late1980s, perhaps explaining why policymakers did notfully perceive the implications of allowing the bubble tocontinue as long as it did.

3

Overview


II Does Japan Save Too Much?Kenneth Miranda

J apan's high saving rate relative to those of other indus-trial countries gives rise to the question of whether

Japan is saving "too much." This section utilizes theconditions on optimal steady-state saving behavior de-rived from neoclassical growth theory to examinewhether Japan saves too much (or too little)—thus as-sessing the optimality of its national saving behavior.

The section is organized as follows: the first part pro-vides a brief review of recent trends in Japanese savingbehavior and a comparison with other major industrialcountries; following an outline of three testable condi-tions derived from neoclassical growth theory for as-sessing the sufficiency of saving, the second part exam-ines the empirical evidence for each of these approaches;the last part draws some conclusions from the analysis.

How Much Does Japan Save?

Saving behavior is important because it helps to deter-mine the evolution of future consumption opportunities.As such, saving can be viewed as the portion of currentincome that allows a nation to raise its future consumptionopportunities—that is, to raise its standard of living.

Japan's postwar saving behavior can be viewed in thislight. Saving has enabled Japan to increase its stock ofassets rapidly (physical and financial as well as domesticand international), increase worker productivity, achieverapid rates of economic growth, and raise its standardof living.

Recent Trends in National Saving

Chart 2-1 shows Japan's national saving rate, bothgross and net of depreciation, for the period 1975-93.Both of these measures indicate a negative trend throughthe first part of the 1980s, with levels in the early 1980sthat are about 2 percentage points of GNP lower thanthose recorded at the beginning of the period. Since 1983,however, there has been a moderate increase in savingrates.

Gross saving in Japan declined from about 33 percentof GNP in 1975 to 30 percent in 1983, before recoveringbeginning in the mid-1980s. Preliminary data for 1993suggest that gross saving rebounded to 33 percent of

Chart 2-1. Gross and Net

GNP. For net saving, the secular decline through 1983is more marked (from 20 percent of GNP in 1976 to16 percent of GNP by 1983) because of rapid rates ofdepreciation charges over the period. Chart 2-1 alsoshows disaggregated saving data. As can be seen, theupturn in the saving rate in the latter half of the 1980swas due primarily to an increase in general govern-ment saving.

Sources: Japan, Economic Planning Agency (EPA), Annual Report |on National Accounts (Tokyo, various issues); and IMF staffestimates.

1Includes public enterprise sector savings.

4

National Saving Rates(In percent of GNP)


How Much Should Japan Save?

Source: IMF, World Economic Outlook (Washington, various issues).

Japanese Saving in an International Context

Table 2-1 and Chart 2-2 show Japanese gross savingin comparison with other major industrial countries forthe most recent ten-year period. As can be seen, thereare wide disparities in saving behavior across countries.Japan is clearly the highest saver among the major indus-trial countries, and by a large margin. The other countriesare clustered in terms of their saving rates, with theUnited States and the United Kingdom at the low endof the spectrum.

Chart 2-2. Gross National Saving Ratesof Major Industrial Countries(In percent of GNP)


The preceding indicates that Japan's saving rate ishigh, both in absolute terms and relative to other coun-tries. In saving, a country chooses to forgo current con-sumption in order to increase future consumption oppor-tunities. Clearly, a balance must be struck between thecosts of forgoing consumption today and the benefits ofincreased future consumption. Too little saving will besuboptimal in that a low level of capital formation willresult in a low level of sustainable consumption. Toomuch saving, however, can also be suboptimal becausepresent and future consumption opportunities are forgonein favor of building and maintaining the stock of capital.1

Neoclassical growth theory provides at least three sepa-rate, but interrelated, conditions by which to assesswhether an appropriate balance between consumptionand saving (investment) is struck—that is, whether acountry is saving too much or too little.2 These are (1)the (modified) "golden-rule" criterion; (2) the dynamic

1 In such a circumstance, a society puts itself on a path toward capitalsaturation or overaccumulation, thereby driving the marginal productof capital toward zero.2 In a neoclassical growth model, the economy converges to a steady-

state balanced-growth path (with a constant capital-labor ratio, con-stant net rate of return to capital, and a constant rate of growth). Theconstant long-run rate of growth wil l equal the (exogenous) rate ofpopulation growth plus the (exogenous) pace of technological prog-ress. Although the long-run rate of growth is not affected by theeconomy's saving rate, the saving rate determines the steady-statecapital-labor ratio as well as the level of consumption per capita.Thus, the "choice" of the saving rate has important implications forsteady-state consumption opportunities. See Solow (1956), Hahn andMatthews (1964), Jones (1976), and Phelps (1961 and 1966) for anoverview of neoclassical growth models and golden rules. Also, seeEvans (1992) for a discussion of the neoclassical growth model andthe adequacy of saving in the United States.

5

Table 2-1. Saving Rates of Major industrial Countries(In percent of GNP)

CanadaFranceGermanyItalyJapanUnited KingdomUnited States

1984

21.118.821.721.130.817.616.7

1985

20.118.722.020.431.617.715.1

1986

19.019.823.820.131.916.613.4

1987

20.219.323.419.732.116.713.6

1988

20.620.824.219.933.216.714.3

1989

20.121.625.619.733.516.514.1

1990

17.421.324.319.433.715.812.9

1991

15.820.622.018.334.414.612.3

1992

15.019.721.917.534.113.711.6

1993

15.318.220.718.233.213.212.6

Average1984-93

18.519.922.919.432.815.913.7

Source: IMF, World Economic Outlook (Washington, variousissues).


II DOES JAPAN SAVE TOO MUCH?

efficiency condition; and (3) the marginal productivityapproach.

Briefly, the golden-rule criterion allows an optimalsaving rate to be calculated. If the actual saving rate isless than (greater than) the golden-rule saving rate, thena country is saving too little (too much) from the perspec-tive of maximizing the level of sustainable consumption.In practice, the actual saving rate can be compared witha range of golden-rule saving rates, derived by varyingthe values of the key parameters (given uncertaintiessurrounding their exact values). The second condition,the dynamic efficiency criterion, compares profits withinvestment. So long as profits exceed investment, thencapital (the sum of past saving decisions) has not beenoveraccumulated. The criterion is a variant of the com-monly cited proposition that, at the golden rule, all profitsare saved (that is, invested). The third criterion, the mar-ginal productivity approach, suggests that capital has notbeen overaccumulated so long as the net marginal productof capital exceeds the growth rate of the economy or arelevant social discount rate. It is a variant of anothercommonly cited result of neoclassical growth theory: atthe golden rule, the net marginal product of capital isequal to the steady-state growth rate.

"Golden Rule" of Accumulation

In neoclassical growth models, a commonly cited crite-rion for choosing among saving rates is the golden rule,which maximizes per capita consumption through time.The golden-rule criterion suggests that the highest sus-tainable level of per capita consumption is attained whenthe net rate of return on capital equals the sum of thepopulation growth rate and the rate of technological prog-ress. If allowance is made for time preference, then amodified golden-rule proposition emerges under whichthe welfare-maximizing steady-state path is characterizedby the condition that the rate of return on capital minusthe rate of pure time preference equals the sum of thepopulation growth rate and the rate of technological prog-ress. In the steady state, a saving rate above or belowthis welfare-maximizing rate will be suboptimal.3

The modified golden rule allows one to calculate thesaving rate and the capital-output ratio that would be ap-proached in the long run. Briefly, it can be shown that only

two equations need to be parameterized to solve for theoptimal steady-state capital-output ratio and saving rate:

K/Y = a/(8 + p + g + n) (2-1)

S/Y = (n + g + 8)K/Y =a[(n + g + 8)/(p + n + g + 8)], (2-2)

where K, Y, S, a, 8, p, g, and n are the capital stock,output, saving, capital's share of output, rate of deprecia-tion, social rate of time preference, exogenous rate oftechnical progress, and growth rate of the labor force,respectively.4 Equation (2-1) defines the capital-outputratio consistent with the modified golden rule, and equa-tion (2-2) defines the steady-state saving ratio. Fromequation (2-2), it is easy to see that if the pure rate oftime preference (p) is zero, then the optimal steady-statesaving rate is equal to capital's share of output (a)—commonly interpreted in the literature as the propositionthat, in the golden-rule steady state, all profits aresaved (invested).

Parameter Values

To make the two equations above operational, it isnecessary to evaluate the likely values of the parametersin the case of Japan.

Capital's Share of Output (a). National accounts datain Japan suggest that capital's share of national incomeis about 35 percent. Estimates based on an aggregateproduction function approach, however, indicate an evenhigher share (of about 40 percent).5 For analytical pur-poses, the work that follows will consider values of 0.35and 0.40.

Rate of Depreciation (8). Rates of economic deprecia-tion of 7 percent (low) and 9 percent (high) are consid-ered. This range is based on the recent shift towardrelatively shorter-lived assets (in part owing to increasingspeeds of technological obsolescence) and takes into ac-count the behavior of the capital goods deflator and thecomposition of investment (public, private fixed, and resi-dential).

Rate of Technical Progress (g). Assuming that most"catch-up" effects have already occurred, the prospectiveexogenous rate of technical progress (multifactor produc-tivity growth) is assumed to be about 1 percent a year.6

3 Note that, in the short run, a saving rate even higher than the long-run value implied by equation (2-2) below can be justified. This wouldbe the case if the economy were approaching the steady state frombelow (that is, from a capital-output ratio lower than the long-runcapital output). In such an instance, saving might be higher than thelong-run rate in response to a rate of return on capital greater thanthe long-run rate. Saving would eventually fall to its long-run levelas the marginal productivity of capital was driven to its long-run rate.In this regard, Christiano (1989) argued that trends in Japan's savingbehavior have been associated with efforts to rebuild the capital stock(and the attendant high rates of return to capital) after World War II.

4 See Evans (1992) for a full description of the methodology and aderivation of the conditions. This section draws heavily from thissource.

5 See, for example, estimates of the aggregate production functionin Japan (1993).

6 Alexander (1994) has noted that, for many sectors of the Japaneseeconomy, significant productivity gaps exist between Japan and theUnited States, suggesting that either technological catch-up is not asyet complete (or that further capital deepening is necessary). A nassumption of slightly higher multifactor productivity growth, how-ever, has a minimal effect on the golden-rule saving rates calculatedbelow, all other things held constant.

6



This value is consistent with the multifactor productivitygrowth rate embodied in estimates for potential outputgrowth. A lower-bound value of 0.5 percent is alsoconsidered.

Growth Rate of the Labor Force (n). Although demo-graphic projections for the total population and for theprime working-age group (aged 15-64) suggest very littlegrowth, a modest rise in the labor force participation rate(especially for women, but also reflecting a gradual boostin the retirement age) is expected. Accordingly, a long-run rate of labor force growth of 1/2 of 1 percent isutilized.

A final issue that must be considered is the social rateof time preference.7 Unfortunately, the social rate of timepreference is not an observable variable. Rather thanassigning an arbitrary value, the approach adopted in thissection is to provide a benchmark optimal saving rateand capital-output ratio on the assumption that the socialrate of time preference is zero. The revealed social rate

7 The social rate of time preference is intended to reflect a society'sevaluation of the relative desirability of consumption at differentpoints in time. There is an extensive literature on the choice of theappropriate social rate of time preference, but the little agreementthat has emerged suggests that the social rate of time preference islow, i f not zero. See Jones (1976, Chapter 9) for a brief summary ofthe academic debate; see also Sen (1967) and Arrow and Lind (1970)for specific approaches to the issue.

of time preference can then be deduced from the devia-tions of the actual from the optimal rate of saving.8

Implied and Actual Saving Rates

Table 2-2 provides illustrative calculations for the opti-mal saving rates and capital-output ratios implied by theranges of parameter values outlined above. As can beseen, the golden-rule saving rates are 35 percent and 40percent (equal, of course, to the assumed capital shares).The long-run capital-output ratios range between 3.3 and5.0. These results provide a benchmark by which to judgethe optimality of recent Japanese saving behavior. Duringthe 1984-93 period, the gross national saving rate inJapan averaged 33 percent of GNP, with a high of 34percent in 1991 and a low of 31 percent in 1984. In thesame period, the capital-output ratio averaged 2.9, witha period-ending high of 3.2.9

These figures suggest that Japan's saving rate andcapital-output ratio are currently both lower than thelong-run steady-state rates implied by a modified golden

8 The subsection on dynamic efficiency, below, provides an approachto the question of oversaving that is independent of the social rateof time preference.

9 Capital-output ratios are derived by cumulating real net investmentflows (public, private fixed, residential, and inventory investment)and dividing by real GNP.

7

Table 2-2. Optimal Gross Saving Rates Under AlternativeParameter Values1

Source: IMF staff calculations.

Capital Share2

35

40

Rate of Technical Progress2

0.51.0

0.51.0

0.51.0

0.51.0

Gross Saving Rate2 Capital-Output Ratio

3535

3535

4040

4040

8 = 7

8 = 9

8 = 7

8 = 9

4.44.1

3.53.3

5.04.7

4.03.8

1The calculations assume a labor force growth rate of 1/2 of 1 percent annually and a social rate of timepreference of 0 percent.

2In percent.



rule under a range of reasonable parameter values. Onthis basis, Japan cannot be said to be oversaving. Indeed,the calculations suggest that Japan could increase itssustainable level of consumption by raising its savingrate by 2 percent of GNP, if capital's share is 35 percent,and by 7 percent of GNP, if capital's share is 40 per-cent. Note, however, that the deviation between actualand optimal saving rates would be consistent with anonzero social rate of time preference. The deviationsuggests that, with a capital share of 35 percent andallowing the other parameters to take on their range ofvalues, the revealed social rate of time preference is about1/2 of 1 percent. If the capital share is 40 percent (and,again, allowing the other parameters to take on theirrange of values), then the revealed social rate of timepreference is 2 percent.

Dynamic Efficiency

An alternative approach to the question of saving be-havior that is also derived from neoclassical growthmodels is the dynamic efficiency criterion. This criterionallows a judgment of whether an economy has overaccu-mulated capital—that is, oversaved—on the basis of eas-ily observable economic variables. Moreover, while oneis left, under the modified golden rule, with a plausiblerange of saving rates that can be deemed optimal, thedynamic efficiency criterion is a one-way test: it allowsone to assess whether the current capital stock is toohigh—and thus, whether overaccumulation (oversaving)has occurred.

Conceptual Underpinnings

According to the literature on optimal economicgrowth, an economy is said to be dynamically efficientif it invests less than the return to capital. In the case ofdynamic efficiency, the economy has not overaccumu-lated capital in the sense that the marginal product ofcapital exceeds the rate of growth of the economy. Inthe case of dynamic inefficiency, the reverse is true—implying that a "society would be reducing its consump-tion merely to support the growth of a capital stock whichis so large that diminishing returns have robbed it of itscapacity to support its own growth and leave a surplusfor extra consumption" (Solow (1970, p. 28)). Thus, ina dynamically inefficient economy, it is possible for soci-ety to go on a consumption binge, thereby reducing thestock of (overaccumulated) capital. Thereafter, consump-tion per capita could be higher because resources do nothave to be devoted to maintaining an inefficiently largecapital stock.

More formally, the condition for an economy to bedynamically efficient can be derived from the steady-state relationship among consumption (C), output (Y),the capital stock (K), and the rate of economic growth (u):

C = Y - uK. (2-3)

Equation (2-3) says that steady-state consumption isequal to steady-state output, less the proportion of outputthat must be invested each period for the capital stockto grow at the same rate as output.10 Differentiating equa-tion (2-3) with respect to the capital stock gives

dC/dK = dY/dK - u = MPK - u. (2-4)

Equation (2-4) implies that steady-state consumption canbe raised by increasing the capital stock as long as themarginal product of capital (MPK) exceeds the rate ofeconomic growth. If this is the case, the economy is saidto be dynamically efficient: it has not overaccumulatedcapital. But if the marginal product of capital is lessthan the rate of economic growth, more capital must bereinvested to maintain a constant capital-output ratio thancapital produces at the margin. Consumption is thus re-duced to maintain the capital stock at an inefficientlyhigh level.11

A final step is needed to implement a test for dynamicefficiency. Multiplying both sides of equation (2-4) bythe capital stock results in the following measurable con-dition for testing whether an economy is in the dynami-cally efficient region:

MPK(K) - u(K) = MPK(K) - I > 0. (2-5)

That is, an economy is dynamically efficient if the returnto capital exceeds investment.12 The intuition behind thisresult is straightforward: if the return to all past invest-ments exceeds current investment, then part of the returnis being consumed by society. The capital sector is con-tributing to consumption opportunities. If this were notthe case, then society would never partake in the fruitsof earlier sacrifices and would perpetually not only rein-vest all the returns from past investment, but also depresscurrent consumption to further add to the capital stock:the capital sector is a drain on consumption opportunities.

Previous Findings

Abel, Mankiw, Summers, and Zeckhauser (1989; here-after, AMSZ) used equation (2-5) to test for dynamicefficiency in a cross section of countries, including Japan.In their approach, they showed that equation (2-5) holdsif production technology exhibits constant returns to scale

10 If the capital stock grows at a rate other than u, the capital-outputratio would be changing, thus violating the steady-state assumption.For simplicity, depreciation is ignored. It is, however, straightforwardto generalize equation (2-3) by adding the depreciation rate to u.

11 The golden-rule growth path is defined as the point wheredC/dK = 0, or where the marginal product of capital equals the rateof economic growth. At this point, the associated capital-output ratiomaximizes sustainable consumption. Such a point is consistent withan assumption of a zero rate of social time preference.

12 Phelps (1966) noted that, in a world of uncertainty, overaccumula-tion may be optimal, so that a reserve of capital could be consumedin the event of an earthquake, war, or other probabilistic phenomena.In this sense, a strict nonnegativity criterion might not be the appro-priate test of dynamic efficiency.

8



and there are no monopoly profits, so that capital earnsa competitive return. Equation (2-5) also requires that asteady state has been achieved. As a result, less weightcan and should be given to results pertaining to anyindividual year; rather, investment rates and returns tocapital should be examined for longer (five-year or ten-year) periods (over which it can be assumed that thecapital-output ratio is fairly constant). In addition, A M S Znoted that their analysis takes no account of investmentin human capital and may be misleading in its treatmentof the return to land.13 Finally, A M S Z ignored the roleof public investment in assessing dynamic efficiency(see below).

In examining Japan for the period 1960-84, A M S Zfound that the economy is dynamically efficient. Thegross profit rate exceeded the investment rate throughoutthe period by an average of slightly more than 10 percent-age points of GNP, although the difference between therates shows a marked secular decline. In the early 1980s,the average difference falls to about 8 1/4 percent of GNP,compared with an average difference of 13 3/4 percentagepoints in the decade of the 1960s. A M S Z concluded that,despite Japan's high rate of capital accumulation andtradition of low real interest rates, the criterion for dy-namic efficiency is comfortably satisfied.

Recent Evidence

The latter part of the 1980s saw a remarkable boomin investment activity, and it is therefore relevant torevisit the question of dynamic efficiency in Japan. Thefollowing sections first examine dynamic efficiency alongthe lines of the A M S Z approach, but extending the resultsthrough 1992, and then expand the A M S Z results byconsidering two other large sources of capital accumula-tion in Japan: public and foreign investment.

Extending the AMSZ Results

Table 2-3 presents profit and investment rates for theprivate sector based on national accounts data from Ja-pan's Economic Planning Agency (EPA) for the period1976-92.14 Profits are calculated as the sum of operatingsurpluses of the nonfinancial corporate sector, the finan-

13 To take human capital accumulation into account, it is necessaryto add investment in human capital to the gross investment figureand to make an estimate of the proportion of wage compensation thatis due to human capital formation. This proportion could then beadded to the gross profit figure. Although this would be an interestingextension, it is beyond the scope of the present analysis.

For the treatment of land, A M S Z pointed out that part of the grossprofit is a return to land, and that as a result the return to capital (thegross profit rate) may be overstated. They noted, however, that re-search in this area has not produced conclusive results (see A M S Z ,p. 9).

14 A M S Z used data from the Organization for Economic Cooperationand Development (OECD) for their calculations.

Chart 2-3. Dynamic Efficiency(In percent of GNP)

cial sector, and the unincorporated nonfinancial enterprisesector, plus the capital consumption allowances of thesesectors. The operating surplus of the unincorporated non-financial enterprise sector has been adjusted downward,since a large part of the operating surplus may actuallyreflect a return to labor.15 The adjustment factor used is65 percent, on the assumption that the return to labor inthis sector does not differ significantly from other esti-mates of labor's share in income. The profit rate is theratio of gross profits to GNP. The investment rate isthe sum of gross fixed capital formation plus inventoryaccumulation in each of the sectors, relative to GNP.

The data indicate that Japan's private sector has beendynamically efficient over the period (Chart 2-3, toppanel). On average, over the 1976-92 period the profitrate exceeded the investment rate by 7.2 percentagepoints of GNP, with a minimum of 4.2 percent of GNP

15 Because the EPA ' s national accounts data consolidate the house-hold sector with unincorporated nonfinancial enterprises and includethe imputed rent for the flow of services from owner-occupied housingwithin the (consolidated) operating surplus, this amount is netted forpurposes of calculating the wage component of the operating surplus.It is, however, included in the calculation of gross profits.

9

Sources: Japan, EPA, Annual Report on National Accounts (Tokyo,various issues); and IMF staff estimates.



Table 2-3. Private Sector Dynamic Efficiency

Year

19761977197819791980198119821983198419851986198719881989199019911992

Five-year average1976-801981-851986-901988-92

Ten-year average1976-851983-92

Period average1976-92

Gross Profits/GNP

(1)

30.530.331.531.432.432.031.631.531.832.332.632.432.933.032.832.230.5

31.231.832.732.3

31.532.2

31.9

Investment/GNP(2)

26.325.024.425.825.824.823.822.222.623.022.623.125.026.227.126.824.9

25.523.324.826.0

24.224.4

24.7

Dynamic EfficiencyIndicator

(1) - (2)

4.25.47.15.66.67.27.89.39.29.3

10.19.27.96.85.75.45.6

5.88.67.96.3

7.27.9

7.2

Sources: Japan, Economic Planning Agency (EPA), Annual Report on National Accounts (various issues); and IMFstaff estimates.

in 1976 and a m a x i m u m of 10.1 percent of G N P in 1986.For the decade ending in 1985, the average differencewas 7.2 percent of G N P ; and for the decade ending in1992, the average difference was 7.9 percent of G N P .

Role of Public and Foreign Investment

A broader approach to the question of dynamic effi-ciency is provided by including public as well as foreigninvestment (saving).16 Foreign investment (defined as thecurrent account balance less capital transfers) is includedbecause it ultimately represents a claim on capital, evenif such claims are not located domestically. Moreover,

such investment (saving), whether located domesticallyor abroad, represents forgone consumption opportunities,a key issue in deciding whether overaccumulation isoccurring. Finally, at a technical level, G N P includesfactor incomes from abroad, and thus the return to capitalcalculated therefrom includes returns to domestic andforeign investment. Government investment is includedfor two reasons: it is as germane to the question ofoveraccumulation as private investment; and public in-vestment m a y well be used by the Government to bringthe economy closer to a golden rule. That is, it m a ywell be used to supplement suboptimal private sectorinvestment, brought about by the disincentive effectsinherent in government tax and regulatory policies as wellas higher than socially optimal rates of time preference inthe private sector (see Evans (1992); Foley and Sidrauski(1971); and Atkinson and S a n d m o (1980)).

Table 2-4 presents profit and investment rates for thisexercise. After public and foreign investment were in-cluded, the test indicated that Japan has been dynamicallyefficient over the period. O n average, the profit rate has

16 A n alternative test, between the A M S Z approach and the one usedhere, was also considered: foreign investment (saving) was added toprivate gross fixed capital formation and inventory accumulation toarrive at a broader measure of private sector accumulation, and netproperty income from abroad was included in deriving the gross rateof profit—thereby providing an augmented test for private sectordynamic efficiency. The inclusion of these flows does not alter thebasic conclusions on private sector dynamic efficiency.

10



Year

19761977197819791980198119821983198419851986198719881989199019911992




Gross Profits/GNP(1)

30.930.832.132.133.032.632.532.432.833.533.833.834.334.634.233.732.2

31.832.834.233.8

32.333.5

32.9

Investment/GNP(2)

32.532.432.631.631.231.730.629.830.831.631.932.133.233.533.734.434.1

32.130.932.933.8

31.532.5

32.2

DynamicEfficiencyIndicator(1) - (2)

- 1 . 6- 1 . 6- 0 . 5

0.61.90.91.92.72.11.91.91.71.2I.I0.5

- 0 . 7- 1 . 9

- 0 . 31.91.30.0

0.8I.I

0.7

Sources: Japan, EPA, Annual Report on National Accounts (various issues); and IMF staff estimates.

exceeded the investment rate by 0.7 percentage pointsof GNP over the 1976-92 period. For the decade endingin 1985, the average difference was 0.8 percent of GNP;and for the decade ending in 1992, the average differencewas 1.1 percent of GNP.

In examining the results, two features are apparent.First, the average difference, whether for the entire periodor for the identified decades, is positive but small.17 Sec-ond, for some years of the period under investigation,the difference between the gross profit rate and the invest-ment rate is negative. One possible interpretation of thesefeatures is that Japan, while dynamically efficient, maybe close to an inefficient region. There are, however, a

17 In a policy sense, the size of the difference between the grossprofit and investment rates is less germane when the difference issmall. It is more relevant when the difference is large, since thismay suggest that, although the economy is efficient in that it is notoveraccumulating, it does not rule out that the economy may beunderaccumulating.

number of reasons, related in part to data and measure-ment issues, for questioning this interpretation.

Although the average difference between the grossprofit rate and the investment rate is relatively small, itis probably substantially understated. The inclusion ofpublic investment, which averaged 8 percent of GNPover the 1976-92 period, results in a bias toward dynamicinefficiency because national accounts data do not makean imputation for the flow of services yielded by thepublic capital stock, and hence the overall gross profitrate in the economy may well be understated.18 Moreover,

18 Of course, part of the flow of services may indeed be captured,to the extent that public services are priced (toll roads, airport taxes)or to the extent that the flow of unpriced or underpriced services arecaptured in firms' profitability or workers' wages. The degree ofcomplementarity among public capital, private capital, and labor in-puts then becomes an issue. Still, to the extent that a large part ofthe service flow from public investment is not captured and imputed,the difference between gross profit and investment may be signifi-cantly biased downward—that is, away from dynamic efficiency.

11

Table 2-4. Aggregate Dynamic Efficiency



the period under investigation includes subperiods of twomajor recessions (1976-78, when the economy was stillin a recovery phase from the first oil shock; and 1991—92, the onset of the current recession). In general, cyclicalconditions can adversely affect measures of dynamicefficiency—both because gross profits tend to fall andbecause declines in private investment may well be com-pensated by increases in countercyclical public invest-ment expenditure measures. This consideration is particu-larly relevant for the two subperiods in which thedifference between the gross profit rates and investmentrates are negative. Finally, as noted earlier, because ofthe underlying steady-state assumption, that the measureof dynamic efficiency in any one year shows a negativedifference cannot be taken as conclusive evidence ofinefficiency. Rather, averages over longer periods aremore germane. That cyclical conditions also influencethe measure of dynamic efficiency bolsters the argumentfor looking at averages over five- or ten-year periods.

In summary, the results indicate that Japan is dynami-cally efficient; that is, it has not overaccumulated capital.This is true whether narrow (private sector) or broaderconcepts (including government capital and foreign in-vestment) of accumulation are considered.

Marginal Productivity

A third approach to the question of over- or undersav-ing involves calculating the marginal productivity of cap-ital, and comparing it either with a measure of the oppor-tunity cost of savings or with the growth rate of theeconomy. The former comparison is really a cost-benefitapproach (or deadweight loss approach),19 wherein a de-mand price (or marginal social benefit) is compared witha supply price (or opportunity cost variable). The lattercomparison is related both to the golden rule and todynamic efficiency.

Cost-Benefit Approach

The cost-benefit approach is a direct test of whetheran economy is oversaving or not. If the marginal productof capital exceeds the opportunity cost of saving, thenan incremental unit of saving will increase intertemporalconsumption opportunities. The problem then becomesone of estimating the marginal product of capital andfinding a measure of the opportunity cost of saving.

The marginal product of capital can be computed fromnational accounts data and a measure of the capital-outputratio. Assuming that aggregate output is governed by aCobb-Douglas production function with constant returns

to scale, gross profits as a share of output are equal tocapital's share of output:

(gross profits/Y) = a= [f'(k)(K)]/Y = r(K/Y). (2-6)

Dividing equation (2-6) by the capital-output ratio yieldsthe marginal product of capital, r:

(gross profits/Y)/(K/Y) = a/(K/Y)= f'(k) = r. (2-7)

In turn, subtracting depreciation yields a net return tocapital:

R = r - 8. (2-8)

This (net-of-depreciation) return can be compared witha variable for the opportunity cost of saving.

As regards the opportunity cost of saving, variousauthors have considered alternative approaches. Some(for example, Harberger (1972) and Feldstein (1977))adjusted the return to capital for corporate and personaltaxes to find the net-of-tax return received by investors.20

The supply price of saving derived from this exercise isthe market's revelation of investors' rate of time discount.Although such an approach is useful in measuring thedeadweight loss associated with capital market distor-tions, it is less useful in addressing the issue of oversav-ing—since, by construction, the demand price will ex-ceed the supply price. Others (for example, Feldstein(1977) and Boskin (1986)) have used cardinalist utilityapproaches to derive a theoretical supply price. Underthese approaches, assumptions about the elasticity of mar-ginal utility with respect to consumption and the marginalrate of substitution between consumption at differentdates are made to derive a "planner's" rate of time prefer-ence, which is a function of the rate of consumptiongrowth. The planner's rate of time preference can in turnbe augmented to take into account individuals' myopiaand the probability of death in order to find the privaterate of time preference. Either of these rates, dependingon the situation, can then be compared with the returnto capital to establish whether too much or too littlesaving is taking place.

The approach taken here is along lines of the latterapproach: thus, so long as the calculated return to capitalis greater than a range of values for the "planner's" orsocial rate of time preference, oversaving cannot be saidto have occurred. Table 2-5 presents data for alternativemeasures of the return to capital under alternative depreci-ation rates of 7 percent and 9 percent. The first measureis an estimate of the return to private capital; the secondis the return to private and public capital. Again, becauseof cyclical and other influences, both annual and periodaverages are provided.

19 That is, to the extent that the demand price differs from the supplyprice, the deadweight loss of the distortion (as well as the theoreticallevel of saving and investment in an undistorted market) can becalculated on the basis of estimated values for the interest elasticitiesof saving and investment.

20 Alternatively, the average net-of-tax return to savers can be com-pared directly with the net return to capital, attributing the differenceto a tax wedge.

12



Table 2-5. Return to Capital

Sources: Japan, EPA, Annual Report on National Accounts (various issues); and IMF staff estimates.

As can be seen, the first measure averages 10 percentover the full period at the depreciation rate of 7 percent.If the higher rate of depreciation is used, the average is8 percent. Thus, for values of the social rate of discountas high as even 2 percent, the data suggest that capitalaccumulation in Japan has not driven the stock of capitalto the point that the return to capital is below the relevantopportunity cost. The second measure, although smaller,also exceeds reasonable values of the social rate of timepreference. Again, recall that this measure may be biaseddownward because the national accounts make no impu-tation for the flow of services from the government capitalstock. Indeed, all the caveats discussed in the section ondynamic efficiency apply equally to this measure.

Growth Rate Approach

The return to capital can also be compared with thereal rate of economic growth, reflecting the standard

result from growth theory that, at the golden rule, R =g.21 Thus, if R > g, then the capital stock has not reachedits golden-rule level, and the economy cannot be said tohave overaccumulated.22 For this reason, Table 2-5 alsoincludes a column on the real rate of economic growth.The private sector return to capital exceeds the real rateof economic growth for both of the assumed rates of

21 This can easily be seen in the derivation for the dynamic efficiencycriteria above.

22 Note that a number of authors have compared the growth ratewith the risk-free real interest rate, rather than with the net return tocapital. In general they have found that the real risk-free rate is lowerthan the real growth rate, which would in turn suggest that capitaloveraccumulation has occurred. Other authors, however, note that thereturn on equities has substantially exceeded the real growth rate ofthe economy and that the test for capital overaccumulation basedsolely on the risk-free rate of return may not be sufficiently rigorous(see, for example, Blanchard and Fischer (1989) and Boskin (1986)).

Year

19761977197819791980198119821983198419851986198719881989199019911992




Private SectorReturn to Capital

8 = 7

10.710.411.010.911.110.610.210.010.210.610.410.110.510.29.68.97.4

10.810.310.29.3

10.69.8

10.2

8 = 9

8.78.49.08.99.18.68.28.08.28.68.48.18.58.27.66.95.4

8.88.38.27.3

8.67.8

8.2

Aggregate EconomyReturn to Capital

8 = 7

5.35.05.45.35.34.94.74.54.75.05.04.95.25.24.94.63.6

5.24.85.14.7

5.04.8

4.9

8 = 9

3.33.03.43.33.32.92.72.52.73.03.02.93.23.22.92.61.6

3.22.83.12.7

3.02.8

2.9

Growth Rateof Real G N P

4.24.85.05.63.53.43.42.84.35.22.64.36.24.84.84.31.4

4.63.84.64.3

4.24.1

4.2

13



depreciation. For the aggregate economy, the return tocapital exceeds the growth rate of the economy at adepreciation rate of 7 percent, but it is lower than thegrowth rate of the economy at a depreciation rate of 9percent. Once again, the same caveats apply. Using amodified golden-rule result does not alter these findingsfor reasonable values of the social rate of time prefer-ence.23 In sum, comparing rates of return with growthrates of the economy suggests that the golden-rule levelof capital stock is unlikely to have been exceeded.

ConclusionsThis section has used conditions from neoclassical

growth theory to investigate Japanese saving behavior.More specifically, three separate but interrelated testswere conducted to determine whether the flow of savingor the stock of accumulated capital could be deemed"excessive" from an economic viewpoint. The modifiedgolden-rule criteria, although perhaps the weakest of thethree tests (because of the uncertainty surrounding param-eter values), suggests that Japan's recent saving behaviorand capital-output ratio are lower than those consistentwith maximizing the level of sustainable consumptionover the long run. Tests of dynamic efficiency indicatedthat for the period 1975-92 Japan has not overaccumu-lated capital. The result holds for both the private sectoras well as the aggregate economy (that is, includingpublic and foreign investment). A final test relating tothe marginal productivity of capital showed that the netrates of return to capital have generally exceeded opportu-nity costs, unless the depreciation rate is assumed to berelatively high. Again, this was true both for the privatesector and for the aggregate economy.

Still, it is important to recognize several caveats withrespect to the work that has been presented. First, asregards the golden-rule approach, there is some uncer-tainty about the relevant parameter values. In addition,the social rate of time preference is an unobservablevariable. Second, there may be a bias introduced in thetests for dynamic efficiency as well as those for themarginal productivity because national accounts data donot make an imputation for the service flows generatedby public investment. Finally, to the extent that Japan isstill in the process of capital deepening, technologicalcatch-up, or even postwar reconstruction, the assumptionthat the economy is at or near a steady state may not bejustified. Notwithstanding these caveats, the three sets oftests, whether looked at individually or together, suggestthat neither Japan's flow of saving nor its stock of accu-mulated capital can be considered excessive from theperspective of maximizing sustainable consumption.

23 Under a modified golden rule, i f the rate of return to capital lessthe rate of time preference exceeds the growth rate of the economy,then the economy cannot be said to have overaccumulated.

ReferencesAbel, Andrew B., Gregory N. Mankiw, Lawrence H. Sum-

mers, and Richard J. Zeckhauser, "Assessing DynamicEfficiency: Theory and Evidence," Review of EconomicStudies, Vol. 56 (January 1989), pp. 1-20.

Aghevli, Bijan, and others, The Role of National Saving inthe World Economy: Recent Trends and Prospects, IMFOccasional Paper 67 (Washington: International Mone-tary Fund, 1990).

Alexander, Arthur J., "Japan as Number Three: Long-TermProductivity and Growth Problems in the Economy,"Report 17a (Washington: Japan Economic Institute, April29, 1994).

Arrow, Kenneth, and Robert Lind, "Uncertainty and the Evalu-ation of Public Investment Decisions," American Eco-nomic Review, Vol. 60 (June 1970), pp. 364-78.

Atkinson, A.B., and A. Sandmo, "Welfare Implications ofthe Taxation of Savings," Economic Journal, Vol. 90(September 1980), pp. 529-49.

Blanchard, Olivier Jean, and Stanley Fischer, Lectures onMacroeconomics (Cambridge, Massachusetts: MIT Press,1989).

Boskin, Michael, "Theoretical and Empirical Issues in theMeasurement, Evaluation, and Interpretation of PostwarU.S. Saving," in Savings and Capital Formation: ThePolicy Options, edited by F. Gerard Adams and SusanM. Wachter (Lexington, Massachusetts: LexingtonBooks, 1986).

Cass, D., "Optimum Growth in an Aggregative Model ofCapital Accumulation," Review of Economic Studies,Vol. 32 (1965), pp. 233-40.

, "On Capital Overaccumulation in the AggregativeNeoclassical Model of Economic Growth: A CompleteCharacterization," Journal of Economic Theory, Vol. 4(1972), pp. 200-23.

Christiano, Lawrence J., "Understanding Japan's Saving Rate:The Reconstruction Hypothesis," Federal Reserve Bankof Minneapolis Quarterly Review, Vol. 13 (Spring1989).

Evans, Owen, "National Savings and Targets for the FederalBudget Balance," in The United States Economy: Perfor-mance and Issues, edited by Y. Horiguchi and others(Washington: International Monetary Fund, 1992).

Feldstein, Martin, "Does the United States Save Too Little?"American Economic Review, Vol. 67 (February 1977),pp. 116-21.

Foley, Duncan K., and Miguel Sidrauski, Monetary and FiscalPolicy in a Growing Economy (London: Macmillan,1971).

Hahn, F.H., and R.C.O. Matthews, "The Theory of EconomicGrowth," Economic Journal, Vol. 64 (December 1964),pp. 779-902.

Harberger, Arnold C., Project Evaluation: Collected Papers(New York: Macmillan, 1972; Chicago: University ofChicago Press, 1976).

Japan, Economic Planning Agency (EPA), Economic Surveyof Japan, 1992-1993 (Tokyo, July 1993).

Jones, Hywel G., An Introduction to Modern Theories ofEconomic Growth (New York: McGraw-Hill, 1976).

Phelps, Edmund. S., "The Golden Rule of Accumulation: AFable for Growthmen," American Economic Review, Vol.51 (1961), pp. 638-43.

14


, Golden Rules of Economic Growth: Studies of Effi-cient and Optimal Investment (New York: Norton, 1966).

Sen, A . K . , "Isolation, Assurance, and the Social Rate of Dis-count," Quarterly Journal of Economics, February 1967,pp. 112-24.

Solow, Robert, " A Contribution to the Theory of Economic

Growth," Quarterly Journal of Economics, Vo l . 32(1956), pp. 65-94.

, "Comment on the Golden Rule," Review of EconomicStudies, Vo l . 29 (June 1962), pp. 255-57.

, Growth Theory: An Exposition, (Oxford:Clarendon, 1970).

References

15


III Japan's Capital FlowsJuha Kahkonen

S ince the early 1980s, Japan has been the world'slargest exporter of capital. Despite having the highest

investment rate of all major industrial countries, Japanhas invested less at home than it has saved, transferringpart of its saving abroad and, as a consequence, runningcurrent account surpluses. From modest amounts in 1980,outflows of portfolio capital and foreign direct investment(FDI) have since surged; Japan has also become theworld's largest provider of development assistance. Thissection discusses the developments in Japan's net capitaloutflows, with particular attention paid to the determi-nants and impact of Japan's FDI.

The first part provides an overview of capital flowsand Japan's international asset position since 1980. Threephases can be distinguished. In the first half of the 1980s,Japan's long-term capital outflows, especially portfolioinvestment, grew rapidly in line with the current accountsurplus, helped by the liberalization of capital controls.During the second half, the rise in net long-term outflowsaccelerated, with direct investment gaining importance,but these outflows were partially offset by large net bor-rowing of short-term capital. These developments weredriven by deregulation of domestic financial markets,further capital account liberalization, rising domesticasset prices, and a sharp appreciation of the yen. So farin the 1990s, net long-term outflows have been wellbelow the level reached in the latter half of the 1980s,and Japan's rising current account surpluses have beenaccompanied by reductions in short-term liabilities. Thechanged pattern of Japan's capital flows reflects the com-pletion of the stock adjustment following relaxation ofcapital controls, the collapse of the asset price "bubble,"and banks' response to the introduction of Bank of Inter-national Settlements (BIS) capital adequacy guidelines.

The second part of the section focuses on FDI. Japan'soutward FDI gained high visibility in the second half ofthe 1980s, with Japan becoming the largest investor inthe world (in flow terms) and with the stock of overseasinvestment growing fivefold between 1985 and 1992.The pattern of Japanese FDI also changed, from resource-oriented investment in developing countries toward theacquisition of real estate and financial institutions in themajor industrial countries. Although much of Japan's FDIcan be seen as part of a process of evolving comparativeadvantage, the same macroeconomic factors that pro-

vided incentives for other types of long-term capital out-flows (such as the yen appreciation, the extraordinaryrise in asset prices, and liberalization of foreign ex-change controls) were also important determinants of thesurge in Japan's FDI in the late 1980s. The rise in Japan'sFDI can be seen as beneficial to the world economy; notonly has it provided potential for the usual gains frominternational integration to be realized, but there are, inaddition, possible positive spillovers into the host econo-mies in the form of new technology and organizationalskills. Although FDI, like domestic investment, has ef-fects on the structure of output and employment,any permanent impact on total employment is likely tobe small. Similarly, because FDI is unlikely to have asignificant influence on the underlying determinantsof saving and investment, Japanese FDI will probablyhave only temporary effects on the current accountposition.

Recent Developmentsin the Capital Account

Japan's capital flows have been the counterpart of thecurrent account balance (Chart 3-1). The current accountsurplus rose rapidly until 1986-87, reaching a peak of4 1/4 percent of GDP in 1986. (In U.S. dollars, the peakcame in 1987, at $87 billion.) Thereafter, the surplusdeclined steadily, to a low of 1 1/4 percent of GDP ($36billion) in 1990. In the early 1990s, the surplus againrose sharply, stabilizing at about 3 percent of GDP in1992-93. (In U.S. dollars, an all-time high of $131 billionwas recorded in 1993.)

In most years, the combined net outflow of short-termand long-term capital closely matched the current accountsurplus (Table 3-1). The exceptions were the years1986-88 and 1993, when a significant part of the surpluswas channeled to official reserves, owing to interventionby the Bank of Japan in the foreign exchange market.In particular, in 1987 the monetary authorities absorbedabout one half of the current account surplus.

Developments in Japan's capital account since 1980can be divided into three markedly different phases. Inthe first half of the decade, long-term capital flows (grossand net) grew rapidly, albeit from a low base, while

16


Recent Developments in the Capital Account

Chart 3-1. Current Account Balance

Source: Nikkei Telecom.

short-term flows were relatively small and stable, owingto limits on banks' open short positions in foreign curren-cies. Overall, Japan's net foreign assets rose from 1 per-cent of GDP in 1980 to 6 percent of GDP in 1984 (Chart3-2). The buoyancy of long-term flows was largely theresult of the liberalization of capital controls, starting withthe revision of the Foreign Exchange and TransactionsControl Law in December 1980, which, among otherthings, abolished the system of prior approval for foreignsecurities investment. As a result, both the flow of port-folio capital and the share of securities in foreign assetsand liabilities increased markedly during the period. Anextensive system of monitoring and administrative guid-ance remained in place, however, and ceilings on pur-chases of foreign securities were an effective constrainton an even faster reallocation of Japanese investors'portfolios.

In the second half of the 1980s, gross long-term capitaloutflows continued their rapid rise, with foreign direct

Chart 3-2. Foreign Assets and Liabilities

Source: Japan, Economic Planning Agency (EPA), Annual Reporton National Accounts (Tokyo, 1993).

investment becoming an important component of capitalexports, and net long-term outflow also reached recordlevels. In contrast to developments earlier in the decade,however, there was a large net inflow of short-term capi-tal. Reflecting these partially offsetting flows, Japan'snet foreign asset position remained stable at around 10percent of GDP in 1985-89. A combination of factorswas responsible for the pattern of capital flows observedduring this period. Dismantling of capital controls againplayed an important role. This time the liberalizationmeasures not only stimulated long-term capital outflowsbut also enabled banks to import substantial amounts ofshort-term capital, since limits on open short positions in

and Capital Flows

1Net inflow of short-term and long-term capital, excludingerrors and omissions.

17


III JAPAN'S CAPITAL FLOWS

1980 1981 1982 1983

Current account balance

Net long-term capitalBy type of capital

Net direct investmentNet securities

BondsEquities and other

Net loansOther

By asset or liabilityAsset

Direct investmentSecurities


LoansOther

LiabilitiesDirect investmentSecurities


Loans

Other

Errors and omissions

Basic balance1

Short-term capitalBankNonbank

Overall balanceIncrease in reservesOther2

-10 .7

2.3

-2.19.4

- 2 . 8-2.1

-10 .8- 2 . 4- 3 . 8

- 2 . 6-2.113.10.3

13.1

- 0 . 2

—

-3.1

-11.5

- 1 . 9- 5 . 0

3.1

-13 .44.9

-18.3

4.8

- 9 . 7

- 4 . 74.4

- 5 . 3-4.1

-22 .8- 4 . 9- 8 . 8

-5.1-4.1

13.10.2

13.2

- 0 . 2

-0.1

0.5

- 4 . 4

- 1 . 3

2.3

- 5 . 73.2

- 8 . 9

6.8

-15 .0

-4.12.1

- 1 . 03.2

-8.1- 4 . 9

-27 .4- 4 . 5- 9 . 7-6.1- 3 . 7- 7 . 9- 5 . 2

12.40.4

11.95.06.8

- 0 . 2

0.3

4.7

- 3 . 4

- 6 . 6- 5 . 0- 1 . 6

-10 .0-5.1- 4 . 9

20.8

-17 .7

- 3 . 2- 1 . 9

-10.18.3

- 8 . 5- 4 . 2

-32 .5- 3 . 6

-16 .0-12.5

- 3 . 5- 8 . 4- 4 . 4

14.80.4

14.12.4

11.8—0.2

2.1

5.2

- 3 . 5- 3 . 6

0.0

1.61.20.4

1984

35.0

-49 .7

- 6 . 0-23 .6-23 .3

- 0 . 3-12 .0-8.1

-56 .8- 6 . 0

-30 .8-26 .8

- 4 . 0-11 .9-8.1

7.1—7.23.53.7

-0.1

—

3.7

-10 .9

13.317.6

- 4 . 3

2.41.80.5

1985

49.2

-64 .5

- 5 . 8-43 .0-49 .0

5.9-10 .5

- 5 . 2

81.8

- 6 . 5-59 .8-53 .5

-6 .3-10 .4

- 5 . 217.30.6

16.74.5

12.2-0.1

—

4.0

-11 .4

9.910.8

- 0 . 9

- 1 . 50.2

- 1 . 7

1986

85.8

-131.5

-14.3-101.4-95.1

-6 .3-9 .3- 6 . 5

-132.1-14.5

-102.0-93 .0

- 9 . 0-9 .3- 6 . 4

0.60.20.5

-2.12.7—

-0.1

2.5

-43.2

56.958.5

- 1 . 6

13.715.7

- 2 . 0

1987

87.0

-136.5

-18 .4-93 .8-66 .2-27.6-16.3

- 8 . 0

-132.8-19.5-87 .8-72 .9-14 .9-16.2

- 9 . 4- 3 . 7

1.2-6.1

6.7-12 .8-0.1

1.3

- 3 . 9

-53 .4

95.771.823.9

42.339.23.0

1988

79.6

-130.9

-34 .7-66 .7

-107.440.8

-15.3-14.3

-149.9-34 .2-86 .9-85 .8-I.I

-15 .2-13.5

19.0- 0 . 520.3

-21 .641.9

-0.1

- 0 . 8

2.8

-48 .5

64.044.519.5

15.516.2

- 0 . 7

1989

57.2

-89.2

-45.2-28 .0-91 .7

63.6- 4 . 7

-11.3

-192.1-44.1

-113.2-94.1-19.1-22 .5-12.3

102.9-I.I85.12.4

82.717.8

1.0

-22 .0

-54.1

29.48.6

20.8

-24 .7-12 .8-11 .9

1990

35.8

-43 .6

-46.3- 5 . 0

-12 .06.9

16.9- 9 . 2

-120.8-48 .0-39 .7-29 .0-10 .7-22.2-10 .9

77.21.8

34.717.017.739.1

1.7

-20 .9

-28 .7

7.8-13 .6

21.5

-20 .9- 7 . 8

-13 .0

1991

72.9

37.1

-29.441.0

-47 .087.925.00.4

-121.4-30 .7-74.3-68.2-6.1

-13.1-3 .3158.5

1.4115.321.294.038.1

3.7

- 7 . 8

102.1

-119.2-93.5-25 .8

-17.1-8.1- 9 . 0

1992

117.6

-28.5

-14.5-26.2-43 .8

17.68.33.9

-58 .0-17.2-34 .4-35.6

1.3- 7 . 6

1.229.52.78.2

- 8 . 216.415.9

2.7

-10 .5

78.6

-80 .0-73 .0

- 7 . 0

- 1 . 4- 0 . 3-I.I

1993

131.4

-78.3

-13.6-62 .7-30.1-32.6

-3 .81.9

-73.6-13 .7-51 .7-29.9-21.8

-8 .2-0.1- 4 . 7

0.1-11.1

-0 .2-10.9

4.31.9

- 0 . 3

52.9

-29.4-15.0- 14.4

23.526.9

- 3 . 5

Source: Bank of Japan, Balance of Payments Monthly (Tokyo, various issues).1 Including errors and omissions.2 Including yen-denominated holdings of foreign monetary authorities (with sign reversed).

foreign currencies were lifted in June 1984.1 Deregulationand structural changes in the domestic financial sectoralso had a major impact. Deregulation increased competi-tion and induced domestic financial institutions to search

1 The measures that helped to increase long-term flows includedpermitting various types of Euroyen transactions, allowing residentpurchases of foreign-currency-denominated certificates of deposit andcommercial paper (both in 1984), and raising the ceilings on purchasesof foreign securities (starting in 1986).

for profit opportunities abroad, while a structural changefrom depository-type financial institutions (such as postalsavings banks) toward securities-oriented institutions(such as insurance companies and investment trusts) in-creased a trend toward foreign securities holding. Thesharp increases in land and equity prices during the bub-ble period in the late 1980s also contributed to the accu-mulation of foreign assets by Japanese residents. Risingasset prices made equity-related bond issues a low-costform of financing, and institutional investors were able

18

Table 3-1. Summary of the Capital Account(In billions of U.S. dollars)



to convert capital gains into income gains by engagingin capital account transactions.2 Finally, the sharp appre-ciation of the yen, especially in 1985-86, stimulated netcapital outflows by providing incentives for FDI.

The 1990s have witnessed a major departure from pastpatterns of capital flows. The gross outflow of long-termcapital has declined dramatically from the record levelin 1989, and the net long-term outflow has been subdued(it even was negative in 1991). Further, having borrowedheavily short term in the latter half of the 1980s, Japanhas become a major net exporter of short-term capital inthe 1990s. These developments are largely the resultof three factors. First, the stock adjustment followingrelaxation of capital controls appears to have been largelycompleted by the late 1980s, with Japanese investorshaving had sufficient time to reach a desired degree ofinternational diversification. Consequently, these invest-ors became more sensitive to the fundamental determi-nants of capital flows, including interest rate differentials(adjusted for expected exchange rate changes) and ex-change risk. The already heavy exposure to foreign cur-rency assets, the associated exchange risk, and reducedinterest differentials in favor of foreign-currency-denomi-nated assets following the tightening of Japanese mone-tary policy from mid-1989 all reduced incentives forcontinued rapid accumulation of foreign assets.

Second, the collapse of the asset price bubble in 1990reduced financial institutions' "hidden" assets (unrealizedcapital gains). Because exchange losses could not beoffset by capital gains to the same extent as before, theinstitutions became reluctant to purchase foreign securi-ties. The bursting of the bubble also depressed net capitaloutflow by attracting foreign buyers of Japanese equityand by discouraging direct investment abroad, especiallyin real estate. Third, concern about capital adequacy, inparticular the need to conform with BIS guidelines, madebanks adjust their international strategy. Although bankshad in the 1980s accumulated large amounts of long-term foreign assets (closely matched by short-term bor-rowing in foreign currencies, to satisfy regulations tokeep banks' open foreign currency positions within anarrow limit), they now had to downsize their interna-tional positions.

2 Life insurance companies—major institutional investors—had ac-cumulated a large amount of unrealized capital gains ("hidden" assets)that they could not, according to regulations, distribute to policyholders. However, income gains from foreign securities (which earnedhigher nominal interest rates than Japanese assets) could be distrib-uted, and capital losses owing to the depreciation of the U.S. dollarcould be offset by capital gains from domestic assets—as long as theprices of domestic assets were high and rising. This explains theseemingly puzzling buoyancy of Japanese purchases of foreign securi-ties even when the U.S. dollar was widely expected to depreciate.See Kawai (1991, p. 20) for further discussion.

Long-Term Capital Flows

Accompanying the increase in the current account sur-plus up to 1987, net long-term capital outflows roserapidly, reaching a high of $137 billion in 1987. Withthe external surplus falling in 1988-90, the deficit inthe long-term capital account also narrowed. In 1991,however, when the current account surplus again startedto rise, there was a net inflow of long-term capital, of$37 billion, for the first time since 1980. In 1992-93,the long-term capital account again recorded net out-flows, but the amounts ($29 billion in 1992 and $78billion in 1993) were well below the average annualoutflow of $110 billion in the second half of the 1980s.

The ups and downs in net outflows of long-term capitalwere the result of a markedly different pattern of growthin assets (foreign assets accumulated by Japanese resi-dents—outflow) and liabilities (Japanese assets acquiredby foreigners—inflow). The outflow grew rapidlythroughout the 1980s, peaking at $192 billion (6 1/2 per-cent of GDP) in 1989, but it declined thereafter andaveraged $66 billion in 1992-93 (see Table 3-3). Bycontrast, the inflow was relatively small—less than $20billion annually—until 1989, when it jumped to over$100 billion. The inflow peaked in 1991 at $159 billion,which exceeded that year's outflow. In 1992-93, how-ever, the inflow returned to the low pre-1989 levels,averaging $13 billion.

Securities have been by far the most important vehiclefor transferring Japan's savings surpluses abroad, ac-counting for 52 percent of the net outflow in 1980-93.Investment by Japanese residents in foreign securitiesgrew more than tenfold between 1982 and 1986, reachinga high of $113 billion in 1989. The rapid rise in securitiesinvestment was almost entirely in the form of bonds(Table 3-2). In the second half of the 1980s, the bulk ofthis rise was the result of increased offshore intermedia-tion of Japanese funds in the form of purchases by Japa-nese investors of equity-related bonds issued by Japanesecompanies in Euromarkets.3 When a major plunge instock prices in Tokyo made equity warrants less desirablein 1990, securities investment abroad also declinedsharply and has remained subdued, averaging $50 billionannually in 1990-93.

Japanese investment in securities other than regularbonds has generally been small. Purchases of foreignstocks and shares have been relatively limited, exceeding$10 billion only twice (in 1987 and 1989). The small

3 Many Japanese companies chose to issue dollar-denominated war-rants (often combined with currency and interest rate swaps) in theEuromarket to raise funds on favorable terms, avoiding the high costof flotation in the more restricted Tokyo market. The bulk of thesewarrants were bought by Japanese institutional investors in the expec-tation of large capital gains in the then-bullish Tokyo stock market.

19



(In billions of US. dollars)

Net securities (+ = outflow)

Japan's investment inforeign securities

Stocks and sharesBondsYen-denominated external

bonds, etc.

Foreign investment inJapanese securities

Stocks and sharesBondsExternal bonds

1980

- 9 . 4

3.8- 0 . 2

3.0

1.0

13.16.55.31.2

1985

43.0

59.81.0

53.5

5.3

16.7- 0 . 7

4.512.9

1986

101.4

102.07.0

93.0

1.9

0.5-15 .8-2.1

18.4

1987

93.8

87.816.972.9

- 2 . 0

-6.1-42 .8

6.730.1

1988

66.7

86.93.0

85.8

- 1 . 9

20.36.8

-21 .635.1

1989

28.0

113.217.994.1

1.2

85.17.02.4

75.7

1990

5.0

39.76.3

29.0

4.5

34.7-13.3

17.030.9

1991

-41 .0

74.33.6

68.2

2.5

115.346.821.247.3

1992

26.2

34.4- 3 . 035.6

1.7

8.28.7

- 8 . 27.6

1993

-62 .7

51.715.329.9

6.4

-11.120.0

- 0 . 2-30.8

Source: Bank of Japan, Balance of Payments Monthly (Tokyo, various issues).

share of foreign stocks in Japanese equity portfolios couldreflect that stocks, compared with bonds, are less stan-dardized and more difficult to manage (Kawai (1991))or that Japanese investors expect higher average returnsin the domestic stock market than in foreign stock markets(French and Poterba (1991)).4 Japanese purchases ofSamurai bonds (yen-denominated external bonds) andShogun bonds (dollar-denominated external bonds issuedin Tokyo) have also been small.

FDI, which had averaged $4 billion in the first half ofthe 1980s, became an increasingly important contributorto the intermediation of the saving surpluses in the latterhalf of the decade. A peak was reached in 1990, whenoutward FDI amounted to $48 billion and accounted for40 percent of the total outflow of long-term capital. Sincethen, however, FDI has declined steadily, falling to $14billion in 1993. (For a more detailed discussion of devel-opments in FDI, see "Foreign Direct Investment,"below.)

Long-term loans extended by Japanese residents (in-cluding official development assistance) were more im-portant than FDI until the mid-1980s, but the situation

4 French and Poterba (1991) studied the reasons for the tendency ofportfolio investors in major industrial countries to hold nearly all oftheir equity in domestic stocks, despite the well-known benefits ofinternational diversification. (They estimate that Japanese investorshad only 1.9 percent of their equity in foreign stocks at end-1989.)After concluding that institutional constraints do not play a majorrole, French and Poterba showed that current portfolio patterns areconsistent with the explanation that investors in each nation expectreturns in their domestic equity market to be several hundred basispoints higher than returns in other markets.

has subsequently been reversed. Other outflows, in-cluding trade credits, have typically been a very smallshare of total outflows.

The Japanese long-term capital outflow has been di-rected mainly to industrial countries (Table 3-3).5 In1988-93, the United States and the European Union (EU)received three fourths of the outflow. The newly industri-alizing Asian economies (Hong Kong, Korea, Singapore,and Taiwan Province of China) were the destination ofless than 2 percent of Japan's long-term capital exports.Although about one half of outward FDI went to theUnited States, over 60 percent of Japanese investmentin securities was directed to the E U , in particular Luxem-bourg and the United Kingdom. Both countries offerEuromarkets where many Japanese firms issued equity-related external bonds. In addition, in London, U.K."gilts" also attract Japanese investors. Loans have beenprimarily channeled to developing countries (included inthe "other" category in Table 3-3).

Long-term capital inflows have mainly been in theform of securities, which have generally been even moredominant than in the case of outflows (Table 3-4). Pur-chases of Japanese securities by foreigners were particu-larly important in 1989, when Japanese residents issuedlarge amounts of overseas bonds, and in 1991, whensharply lower equity prices in Japan led foreign investorsto buy a record-high amount of stocks in Tokyo, inaddition to continued large purchases of Japanese bondsby nonresidents. With inward foreign investment having

5 Offshore intermediation of Japanese funds makes it difficult toestablish the actual destination of some types of outflows.

Table 3-2. Flows of Securities

20



Table 3-3. Long-Term Capital Outflow by Region and Type

Year

Total 198819891990199119921993

United States 198819891990199119921993

European Union 198819891990199119921993

Of which: United Kingdom 198819891990199119921993

Newly industrializing economies1 198819891990199119921993

Other 198819891990199119921993

Total —United States —European Union —

Of which: United Kingdom —Newly industrializing economies1

Other —

Total

149.9192.1120.8121.458.073.6

61.557.116.435.022.833.9

55.585.558.049.526.330.6

16.317.18.6

19.418.218.2

2.84.45.12.7

- 1 . 2- 1 . 3

30.145.141.234.210.010.4

100.031.742.713.71.8

23.9

DirectInvestment Securities

In billions of U.S. dollars

34.244.148.030.717.213.7

19.021.225.615.28.96.8

5.89.7

11.08.03.43.2

2.94.25.64.71.91.7

2.13.42.61.00.60.3

7.49.78.86.54.33.5

86.9113.239.774.334.451.7

36.226.7

-16 .215.68.5

22.0

42.667.739.541.027.425.9

10.7II.12.0

14.717.116.8

- 0 . 20.3—0.60.20.1

8.318.616.317.2

- 1 . 83.8

In percent of total flow in 1988-93

100.051.421.811.25.3

21.4

100.023.261.018.10.3

15.6

Loans

15.222.522.213.17.68.2

2.85.03.22.84.85.7

1.84.75.2—

- 3 . 41.4

0.9I.I0.9—

- 0 . 5-0.1

0.3—2.41.4

- 0 . 9-0.1

10.312.811.49.07.1I.I

100.027.410.92.53.6

58.1

Other

13.512.310.93.3

- 1 . 20.1

3.54.33.81.50.6

- 0 . 5

5.43.32.30.6

- 1 . 20.1

1.80.60.10.1

-0 .3-0.1

0.60.70.1

- 0 . 3-I.I- 1 . 6

4.14.04.71.50.42.0

100.033.827.35.6

- 4 . 443.3

Source: Bank of Japan, Balance of Payments Monthly (Tokyo, various April issues).1 Hone Kong, Korea, Singapore, and Taiwan Province of China.

21


Ill JAPAN'S CAPITAL FLOWS

Table 3-4. Long-Term Capital Inflow by Region and Type

Source: Bank of Japan, Balance of Payments Monthly (Tokyo, various April issues).1Hong Kong, Korea, Singapore, and Taiwan Province of China.

Total

United States

European Union

Of which: United Kingdom

Newly industrializing economies1

Other

TotalUnited StatesEuropean Union

Of which: United KingdomNewly industrializing economies1

Other

Year

198819891990199119921993

198819891990199119921993

198819891990199119921993

198819891990199119921993

198819891990199119921993

198819891990199119921993

—

Total

19.0102.977.2

158.529.5

- 4 . 7

2.23.34.7

16.7- 4 . 4

-25 .0

21.498.735.788.719.612.5

22.697.434.880.616.415.6

- 5 . 26.9

32.947.412.42.4

0.5-6.1

3.85.71.95.4

100.0- 0 . 672.469.925.3

3.0

DirectInvestment Securities


- 0 . 5-I.I

1.81.42.70.1

- 0 . 6- 1 . 5

0.6-0.1

0.80.5

0.10.3I.I0.61.3

-I.I

0.10.10.21.00.1

0.10.1

0.10.2

0.1

0.80.50.4

20.385.134.7

115.38.2

-11.1

2.94.63.7

16.6- 5 . 8

-25 .5

21.492.230.186.217.313.0

22.791.030.278.614.715.3

- 5 . 2- 4 . 6- 1 . 3

11.3-2.1- 1 . 7

1.2- 7 . 0

2.21.1

- 1 . 23.1

In percent of total flow in 1988-93

100.0- 6 . 754.334.412.240.2

100.0- 1 . 4103.1100.0- 1 . 4- 0 . 3

Loans

-0.117.839.138.115.94.3

0.20.50.20.6

6.24.61.80.90.6

6.24.41.80.70.3

11.534.036.114.33.8

-0.1

100.01.3

12.211.786.6

Other

- 0 . 81.01.73.72.71.9

-0.1

-0.1

—

—

—

- 0 . 70.91.73.72.71.9

100.0-1.1

101.1

22



been insignificant, loans have been the only other im-portant source of long-term capital inflow, particularlysince 1989. European countries have been the source ofthe bulk of inward securities investment, and the newlyindustrializing economies have accounted for the largestshare in loans to Japan.

With few exceptions, the regional pattern of net long-term capital flows is different from the pattern of tradeflows, which is to be expected under a multilateral pay-ments system and free capital movement (Table 3-5). Inthe case of the United States, the recent large bilateraltrade surpluses in favor of Japan have been by and largematched by net exports of long-term capital from Japanto the United States. However, the United Kingdom,having had approximately balanced trade with Japan in1988-93, and the newly industrializing economies, hav-ing run large trade deficits with Japan, were major sourcesof net inflow of long-term capital to Japan.6 By contrast,the developing countries had a trade surplus with Japanand also were important recipients of Japanese long-termcapital outflows.

Short-Term Capital Flows

The fluctuations in short-term capital flows have beeneven sharper than those in long-term flows (see Table3-1). The net inflow, which had been negligible in thefirst half of the 1980s, surged in the second half to anaverage of $51 billion, peaking in 1987 at $96 billion.This increase mainly reflected the activities of privatebanks, which accumulated net foreign liabilities of$194 billion in 1985-89. Following a small net inflowin 1990, the short-term capital account turned to a largeoutflow position of $119 billion in 1991 and has remainedin deficit since. Banks already had become net exportersof short-term capital in 1990, and by end-1993 they hadreversed all of the massive run-up in net foreign liabilitiesduring the second half of the 1980s. By contrast, almosthalf of the $85 billion increase in the nonbank privatesector's net short-term liabilities in 1987-90 remainsoutstanding.

Flow of Financial Resourcesto Developing Countries

Although the bulk of Japan's large current account sur-pluses has been intermediated to other industrial countriesthrough private sector financial institutions, especiallyinstitutional investors, a significant portion of the sur-pluses has also been transferred to developing countries,both by the Government and the private sector. Indeed,by 1987 net financial flows from Japan to developing

6 The regional pattern of capital flows is distorted by the existenceof offshore markets, especially those in London, Luxembourg, HongKong, and Singapore.

countries were the largest in the world, and since 1989Japan has ranked first in official development assistance.

Since the mid-1980s, Japan has taken several initia-tives to stimulate the flow of financial resources to thedeveloping countries. The most comprehensive initiativeshave been two multiyear plans—a capital recycling planfor 1987-92 amounting to $65 billion, and a "Funds forDevelopment" initiative for 1993-97 with a target of$120 billion. The recycling plan pulled together officialdevelopment assistance, other official flows, and privateflows into one package, with official flows intended toserve as a catalyst for increased private flows, particularlyto heavily indebted countries. Financing under that planincluded direct bilateral lending, cofinancing with multi-lateral lending agencies, parallel lending with the IMF,and funds allocated to highly indebted countries underthe strengthened debt strategy. The new $120 billion planconsists of $70 billion in untied official funds and $50billion in other financing, including loans from theExport-Import Bank of Japan and international tradeinsurance.

In recent years, Japan's total net financial flows todeveloping countries have averaged $20 billion annually(Table 3-6). About half of the total has been accountedfor by official development assistance, which takes theform of contributions to the multilateral financial institu-tions and bilateral loans, grants, and technical assistance.Other official flows, including loans by the Export-ImportBank of Japan, have been about 10 percent of the total.The remaining part has originated from the privatesector, including loans from commercial banks, directinvestment, and commercial bank cofinancing with theExport-Import Bank and with multilateral developmentinstitutions.

Japan's net disbursements of official development as-sistance amounted to about $11 billion a year in 1991-93.Although Japan remains the world's largest supplier ofnonmilitary aid, its ratio of official assistance to GNP(0.26 percent in 1993) is slightly below the OECD'sDevelopment Assistance Committee average and fallswell short of the United Nations target of 0.7 percent.The target for 1993-97 has been set at $70-75 billion,enhancing Japan's position as the largest donor over themedium term but implying little change in the aid-to-GNP ratio. The new plan is the first implemented by theGovernment since an Official Development AssistanceCharter was adopted in 1992. The Charter outlines fourprinciples that mandate more stringent conditions forforeign aid. They stipulate that Japan must pay closeattention to the following issues: environmental concerns;restraint in military expenditures and weapons develop-ment; democratization and human rights; and the foster-ing of market-oriented economies.

The regional allocation of Japan's official assistanceis well diversified. Although Asia receives the bulk ofJapan's bilateral aid—with Indonesia, China, the Philip-pines, India, and Thailand accounting for over 40 percent

23



of the total—Africa, the Middle East, and Latin Americaare also major recipient areas (Egypt, Jordan, Turkey,and Peru have recently been among the top ten recipientsof aid). Japan is now the leading donor in some 30developing countries, not only in Asia but also in partsof Africa.

Foreign Direct Investment

Japanese outward FDI surged in the second half ofthe 1980s, with the stock of overseas investment growingalmost fivefold between 1985 and 1992 (Chart 3-3). Thisspectacular increase prompts a number of questions:

• Was the increase in FDI particular to Japan, or was itpart of a worldwide spurt in FDI? Is Japan's FDIunusually large by international standards?

• What is the regional and sectoral distribution of Japa-nese FDI, and how has the pattern changed over the

years? To what extent was the rapid growth in thesecond half of the 1980s attributable to investment inreal estate and financial institutions in the UnitedStates?

• What have been the main factors responsible for therapid growth in FDI? Is the increase a natural part ofthe process of industrial restructuring, or did it occurin response to government intervention (tax policy,trade barriers) or because of macroeconomic factors(yen appreciation, asset price bubble)?

• What are the salient characteristics of Japanese over-seas subsidiaries? Do Japanese affiliates differ fromthose involved in other countries' FDI, with respect toprofitability, repatriation rates, imports, and destinationof production?

• What have been the main effects of Japanese FDIon other economies? Has Japanese investment createdjobs and promoted growth in the host countries, andhas it been a drain on these countries' trade accounts?

Table 3-5. Net Long-Term Capital Outflow by Region and Type(In billions of US. dollars)

24

Total

United States

European Union

Of which: United Kingdom

Newly industrializing economies1

Year

198819891990199119921993

198819891990199119921993

198819891990199119921993

198819891990199119921993

198819891990199119921993

Total

130.989.243.6

-37.128.578.3

59.353.911.718.427.258.9

34.1-13 .2

22.2-39 .2

6.718.1

-6 .3-80.3-26.1-61 .2

1.82.6

8.0-2 .5

-27.7-44 .7-13.5

- 3 . 7

DirectInvestment

34.745.246.329.414.513.6

19.622.825.015.38.16.2

5.79.49.97.32.04.3

2.94.15.54.40.91.6

2.13.32.61.00.50.1

Securities

66.728.0

5.0-41 .0

26.262.7

33.322.1

-19.8- 1 . 0

14.347.4

21.2-24.6

9.4-45.3

10.112.8

-12 .0-79.8-28.2-63.9

2.31.5

5.04.91.3

-10 .72.41.8

Loans

15.34.7

-16.9-25.0

-8 .33.8

2.84.82.82.64.25.7

1.8- 1 . 4

0.6- 1 . 8-4 .3

0.9

0.9-5 .1-3 .5- 1 . 8- 1 . 2- 0 . 4

0.3-11.5-31.7-34 .7-15.2

-3 .9

Other

14.311.39.2

- 0 . 4- 4 . 0-1 .9

3.54.33.81.50.6

-0 .5

5.43.32.30.6

- 1 . 20.1

1.80.60.10.1

-0 .3-0 .1

0.60.70.1

-0 .3- I . I-1.6



• Has FDI had a significant impact on Japan's economy,especially the external surplus? Are Japanese exportsand FDI complementary, or does FDI generate in-creased imports from overseas subsidiaries? Does in-come from direct investment provide a significant con-tribution to foreign exchange receipts?

These issues will be addressed in the remainder of thissection, preceded by a discussion of the concept andmeasurement of FDI and a review of main developmentssince the 1950s.

Concept and Measurement

Conceptually, FDI refers to an investment made by aforeign resident to acquire a controlling interest in a host-country company (IMF (1992, p. 24)). It is the motiveof the investment—corporate control—that distinguishesFDI from portfolio investment, which is simply the estab-

lishment of a claim on an asset for the purpose of realizinga return without being involved in management.

In practice, capital inflows designated as direct invest-ment are distinguished from portfolio investment solelyon the basis of percentage of foreign ownership. At pres-ent, the international standard (also followed by Japan)is a cross-border holding of 10 percent or more of thevoting power in an incorporated enterprise (or a similarinterest in an unincorporated enterprise). Although the10 percent ownership criterion is arbitrary, it is unlikelyto be an important source of measurement errors (Grahamand Krugman (1991, pp. 7-11) and IMF (1992, pp. 24-25)). For example, foreign companies with operations inthe United States own on average an 80 percent share oftheir affiliates, suggesting that for most of these affiliatesforeign ownership is both clear in practice and accuratelyrecorded in balance of payments statistics (Froot (1991)).FDI data, however, tend to understate actual foreign con-trol because they do not include investment by foreign

25

Table 3-5 (concluded)

Other

TotalCurrent accountLong-term capital account

United StatesCurrent accountLong-term capital account

European UnionCurrent accountLong-term capital account

Of which: United KingdomCurrent accountLong-term capital account

Newly industrializing economies1Current accountLong-term capital account

OtherCurrent accountLong-term capital account

Year

198819891990199119921993

_

—

_

—

_

—

—

_

—

_

—

Total

29.651.137.428.5

8.15.0

82.455.6

45.338.2

21.34.8

0.5-28.3

30.8-14.0

-15.026.6

DirectInvestment

7.49.78.85.73.93.1

Securities

7.125.614.216.1

- 0 . 60.7

Average, 1988-932

_

—

_

—

_

—

—

_

—

—

_

—

_

—

—

_

—

—

_

—

Loans

10.312.811.49.07.1I.I

_

—

—

—

_

—

—

_

—

Other

4.83.13.0

- 2 . 2- 2 . 2

0.2

_

—

—

—

_

—

—

_

—

Source: Bank of Japan, Balance of Payments Monthly (Tokyo, various April issues).1Hong Kong, Korea, Singapore, and Taiwan Province of China.2Current account surplus in favor of Japan, and net long-term capital outflow from Japan.



Table 3-6. Net Flow of Financial Resources to Developing Countriesand Multilateral Agencies from Japan1

affiliates that is financed by selling securities to unrelatedparties (domestic or foreign). Furthermore, official statis-tics record only the book value of the FDI and do notallow for increases in the value of foreign-controlledassets.

There are two main sources of data on Japanese FDI.The Bank of Japan compiles monthly balance of pay-ments data, with no breakdown by industry but with someaggregated breakdown by host country made available onan annual basis. The Ministry of Finance publishes annualflow and stock data on FDI on a notification basis, witha detailed breakdown by industry and host country. Thesedata tend to overstate the actual amount of FDI becausesome investment announced is not actually undertakenand because implementation may follow notification witha considerable lag. In recent years, the annual inflows ofFDI as reported by the Ministry of Finance have beenabout 50 percent larger than those recorded in the balanceof payments by the Bank of Japan (see Chart 3-3). In

addition to these two main data sources, the Ministry ofInternational Trade and Industry (MITI) conducts ques-tionnaire surveys on Japanese enterprises' activitiesabroad. These surveys provide information on the marketdestination of Japanese FDI and on cross-border intrafirmtrade. The following discussion mainly uses balance ofpayments data, so that international comparisons can bemade. Ministry of Finance data are used in analyzingthe regional and sectoral composition of FDI and aresupplemented by results from MITI surveys.

Main Developments

Japanese postwar FDI was conducted only on a smallscale until the early 1970s, in part because of stringentgovernment regulations imposed because of the weakbalance of payments. According to balance of paymentsdata, annual outflows rose slowly, to an average of $150million in the second half of the 1960s, and at end-1970

26

Official development assistanceBilateral

LoansGrants and other

Multilateral

Other official flowsExport creditDirect investmentMultilateral

Private flowsExport creditDirect investmentOther bilateralMultilateral

Grants by privatevoluntary agencies

Total resource flows

Official development assistance

Total resource flows

1987

7.55.23.02.22.2

- 1 . 8- 2 . 0

0.3

14.7I.I7.44.41.9

0.1

20.5

0.31

0.86

1988

9.16.43.52.92.7

- 0 . 6- 1 . 8

1.4- 0 . 2

12.80.28.22.81.6

0.1

21.4

0.32

0.75

1989 1990 1991


9.06.83.73.12.2

1.6- 1 . 2

1.90.9

14.30.7

11.31.30.2

0.1

25.6

9.16.83.83.02.3

3.4- 1 . 0

4.20.3

6.3

8.1- 2 . 6

0.7

0.1

18.7

In percent of GNP

0.31

0.84

0.31

0.63

11.08.95.53.32.1

2.6-0 .5

3.20.1

11.20.65.06.2

- 0 . 7

0.2

24.9

0.32

0.73

1992

11.28.44.63.82.8

3.30.12.1I.I

1.5- 1 . 0

2.82.8

- 3 . 0

0.2

16.2

0.30

0.44

Est.1993

11.3

0.26

Source: Japanese authorities.1Calendar years; Development Assistance Committee (OECD) basis.



Chart 3-3. Outward ForeignDirect Investment

Sources: Bank of Japan, Balance of Payments Monthly (Tokyo,various issues); and Japan, Ministry of Finance, Zaisei Kinyu TokeiGeppo (Tokyo, various issues).

the stock of FDI was about 2 percent of that year's GDP.During this period, Japanese FDI took two principalforms. One was the acquisition of raw materials (espe-cially mining products) to supply manufacturing indus-tries in resources-short Japan. The other was foreigninvestment in labor-intensive manufacturing activities inthe nearby Asian countries. Overall, almost two thirds ofJapanese FDI in the 1950s and 1960s went to developingcountries (Table 3-7).

In the early 1970s, Japanese FDI surged, rising fivefoldbetween 1971 and 1973 before stabilizing at around$2 billion for the remainder of the decade. The largeincrease was induced in part by an improvement in Ja-pan's current account position, which turned into a sur-

plus. This led not only to the easing of governmentregulations on FDI but also to a policy to promote it:restrictions on FDI were eased in four steps in 1969-72;the Export-Import Bank of Japan lowered interest rateson loans for foreign investment; and, in order to reducethe risks of FDI, tax provisions for overseas investmentloss reserves were revised (Komiya and Wakasugi (1991,p. 51)). As regards the pattern of Japanese FDI in the1970s, the share of mining declined substantially, andlarge investments were made in the U.S. distributionsector to support the marketing of exports of automobilesand other durable goods. Nevertheless, developing coun-tries continued to host the bulk of Japanese FDI.

During the 1980s, Japan's FDI grew spectacularly. Inthe first half of the decade, overseas investment increasedbriskly, tripling to $6.5 billion (1/2 of 1 percent of GDP)between 1980 and 1985. Even more extraordinary growthwas experienced in the second half of the decade. Japa-nese FDI during the four-year period 1986-89 exceededthe cumulative overseas investment from all postwaryears combined. By the late 1980s, Japan's FDI outflowhad become the largest in the world, and a peak outflowof $48 billion (1 1/2 percent of GDP) was reached in1990. At that time, the annual flow was 20 times largerthan a decade before, and the stock of FDI stood at$302 billion (10 percent of GDP) compared with$48 billion (4 percent of GDP) in 1980.

The unprecedented growth in FDI in the late 1980swas accompanied by a substantial change in the patternof overseas investment. During the 1980s, the share ofJapanese FDI in manufacturing declined sharply, whereasthe tertiary sectors, which had earlier accounted for lessthan half of the total, gained a combined share of morethan 70 percent. Overseas investment increased particu-larly fast in finance and insurance, transport, and realestate, all of which had played only a minor role inJapanese FDI before the 1980s. Regionally, the share ofdeveloping countries (not only in Africa and LatinAmerica but also in Asia) declined sharply, whereas in-dustrial countries absorbed over two thirds of JapaneseFDI, with the United States alone receiving over 40percent of the investment.

The boom in Japanese FDI came to an end in the early1990s, with the annual outflow declining steadily fromthe peak of $48 billion (1 1/2 percent of GDP) in 1990 to$14 billion (1/3 of 1 percent of GDP) in 1993. Althoughthe outflows were still large in absolute terms, relativeto GDP they marked a return to the level first reachedin the early 1970s. The regional and sectoral pattern ofFDI remained broadly the same as in the years of rapidgrowth. The latest surveys of FDI intentions suggestthat future areas of growth in Japanese overseas invest-ment will be machinery, chemicals, and automobiles,with Asian countries (especially China and the membersof the Association of South-East Asian Nations, ASEAN)likely to receive an increasing share of total JapaneseFDI.

27



Table 3-7. Foreign Direct Investment by Region and Industry(In percent of tola!)

By Region

Source: Japan, Ministry of Finance, Taigai chyokusetsu-toshi no kyoka todokede zisseki (Tokyo, various issues).

Worldwide Perspective

The recent surge in FDI was not simply a Japanesephenomenon (Chart 3-4). World FDI rose sharply in thesecond half of the 1980s, both in absolute terms (to apeak of $238 billion in 1990) and relative to output (froma long-term average of 1/2 of 1 percent of GDP to over1 percent of GDP in 1989-90). As in Japan, world FDIhas been on a declining trend so far in the 1990s. Therapid growth of world FDI in the second half of the1980s can be attributed in part to easing of restrictionson capital movements, deregulation of financial markets,and a lagged adjustment to exchange rate misalignments.

The growth in Japan's overseas investment in the1980s was, however, faster than that in the other majorindustrial countries (Chart 3-5). In absolute terms, Japanhad been the fourth largest direct investor in the 1970s(behind the United States, the United Kingdom, and Ger-many), accounting on average for 5 percent of worldFDI. In the early 1980s, Japan moved to the top three,and during 1989-91 it was the world's leading direct

investor, with a share of as much as 20 percent of thetotal. Even during these peak years, Japan's overseasinvestment relative to the size of the economy was lowin comparison with other major industrial countries (ex-cept the United States). Moreover, despite large recentFDI outflows, Japanese overseas subsidiaries produce arelatively small share of the parent companies' output incomparison with affiliates of other industrial countries.In 1989, overseas production accounted for only 5 percentof the total sales of Japanese manufacturing corporations,compared with 15-20 percent for the United States anda number of European countries.

Although Japanese overseas investment flows havelong been at levels comparable with other industrial coun-tries, Japan remains an outlier in terms of inward FDI,with an insignificant share of the world total FDI goingto Japan (Chart 3-6). Nevertheless, in contrast to develop-ments in outward investment, Japanese inward FDI hasrisen steadily in recent years, from an annual average of$200 million in the 1980s to an average $1 1/2 billion in1990-93.

28

Period

1951-601961-701971-751976-801981-851986-901991-921951-92

1951-601961-701971-751976-801981-851986-901990-921951-92

United States

30.718.622.026.634.846.342.142.0

Manufacturing

44.924.733.336.725.125.529.927.4

All

17.321.328.127.320.412.416.315.5

Agriculture,forestry,

andfishery

2.52.72.02.70.70.40.80.7

Asia(newly

industrializingeconomies)

4.65.0

10.210.28.76.95.47.1

Mining

30.431.825.115.49.92.13.05.0

Europe

I.I19.315.29.5

13.921.221.719.6

LatinAmerica

30.015.317.516.620.110.98.0

12.0

By Industry

Commerce

11.310.714.815.615.48.3

12.010.6

Finance andinsurance

3.99.48.05.5

17.924.312.819.7

Middle East

19.88.45.26.21.50.2I.II.I

Transport

12.55.25.65.7

Oceania

0.77.65.37.83.66.17.56.2

Realestate

0.11.75.4

19.318.815.8

Other

0.48.56.16.05.72.93.33.6

Other

7.120.916.722.413.114.917.115.1

Total

100.0100.0100.0100.0100.0100.0100.0100.0

Total

100.0100.0100.0100.0100.0100.0100.0100.0



Chart 3-4. Japanese and WorldDirect Investment Outflows

Chart 3-5. Direct Investment OutflowsAmong Five Major Industrial Countries

Source: IMF, International Financial Statistics (IFS) (Washington,various issues).

Determinants of Japanese OutwardForeign Direct Investment

There are two main approaches to explaining FDI:micro (industrial organization) theories, and macro (cost-of-capital) theories.7 The industrial-organization viewemphasizes that firm-specific intangible assets—such aspatents, brand names, superior technology, and organiza-tional skills—may under certain circumstances make itprofitable for a firm to internalize the rents on theseadvantages through overseas investment, rather thanthrough licensing or exports. According to this view, thepresence of transaction costs in markets for intermediategoods, the desire to keep technological secrets, and at-tempts to circumvent trade restrictions are examples of

7 For surveys of alternative theories of FDI, see Agarwal (1980),Kojima and Ozawa (1984), Jones and Neary (1986), Graham andKrugman (1991, pp. 35-38 and Appendix B), and Lizondo (1991).Empirical studies are surveyed in Mann (1993) and United Nations(1992).

possible reasons for direct investment. The cost-of-capitalview, by contrast, holds that if an investment is madeby a foreign firm, rather than a domestic company, itmay be because the foreign firm has a lower cost ofcapital and therefore requires lower returns.8 In that case,incentives for FDI can be provided by the liberalizationof capital markets, exchange rate movements, and

8 This characterization of the macroeconomic view is that of Grahamand Krugman (1991, p. 35). These authors also describe the micro-economic view as suggesting that when foreign rather than domesticfirms make the investment it is because they expect higher returns—that is, the investment is expected to be more profitable in foreignhands.

29

Source: IMF, IFS (Washington, various issues).



Chart 3-6. Direct Investment InflowsAmong Five Major Insdustrial Countries cries

macroeconomic policies (especially monetary and taxmeasures), among other things.

Lower costs of production in the host country com-pared with the source country—a factor often mentionedin popular discussions—alone cannot be a reason forFDI. If, say, labor is relatively cheap in the host country,it is cheap for domestic and foreign firms alike. Hence,foreign firms, rather than domestic ones, will carry out theinvestment more profitably only if they possess specificadvantages, such as those mentioned above under themicro and macro approaches.

How well do the micro and macro approaches explaindevelopments in Japan's FDI? The distinction between

the industrial-organization and cost-of-capital explana-tions has important implications (Graham and Krugman(1991, p. 38)). If overseas investment is motivated purelyby industrial-organization considerations, Japan's FDIhas little to do with the transfer of Japan's surplus saving.However, if macroeconomic factors are behind Japan'soverseas investment, future FDI flows are to some extentlinked to prospects for the current account. Most studiesagree that both approaches are relevant to explaining therapid growth of Japan's overseas investment in the 1980sand its subsequent decline.9

Proponents of the microeconomic approach point outthat much of Japanese FDI has been in response to indus-trial restructuring and evolving comparative advantage.As mentioned above, in the 1950s and 1960s, the bulkof Japanese overseas investment was in the areas ofnatural resource acquisition and labor-intensive manufac-turing. The FDI in natural resource extraction representeda form of backward vertical integration by Japanese usersof raw materials or by Japanese trading companies havingclose links to them. Similarly, the investment in manufac-turing in nearby Asian countries represented a transferof production abroad in a sector where Japan was losingcomparative advantage. In the 1970s, the heavy invest-ments in the distribution sector to support the marketingof consumer durables, especially automobiles, were aform of forward vertical integration by Japanese manu-facturers. As noted by Caves (1993), these investmentswere in accordance with the standard theory of FDI basedon transaction costs, with controlled-distribution subsidi-aries displacing arm's-length distributors. As regards themore recent period, empirical studies (such as Kogutand Chang (1991) and Hennart and Park (1991)) haveestablished a positive relationship between the rapidgrowth of research and development expenditure by Japa-nese firms and Japan's FDI, consistent with the industrial-organization view that overseas investment takes placeto arbitrage intangible assets (specific advantages) accu-mulated by firms in the source country.10

The industrial-organization literature on FDI also re-gards circumventing trade restrictions (existing or pro-spective) as an important explanation for Japanese FDI.Increased protection reduces a foreign company's netrevenues from exporting, raising the relative profitabilityof foreign investment and turning the exporter into a

9 See, for example, Barrell and Pain (1993), Caves (1993), Froot(1991), Georgiou and Weinhold (1992), Graham and Krugman (1991),Mann (1993), and Thomsen (1993).

10 Kogut and Chang (1991) also found Japanese FDI to the UnitedStates to increase with the intensity of research and development ofthe industries in the United States. The positive effect of U.S. researchand development on Japanese FDI could indicate that Japanese firmscome to acquire intangible assets and not just to exploit those alreadyin their possession.

Source: IMF, IFS (Washington, various issues).

30



direct investor.11 As surveyed by Caves (1993, p. 290),there is a wealth of evidence from econometric and casestudies that trade restrictions in the United States havehad a significant positive effect on FDI from Japan. Inparticular, voluntary export restraints in the automobilesector and the 1986 semiconductor trade agreement havebeen shown to induce sizable direct investment flowsfrom Japan. As for Japanese FDI to the E U , the evidenceis less clear-cut (see Thomsen (1993, pp. 306-09)). Manystudies, including Barrell and Pain (1993) and Heitgerand Stehn (1990), have found protection in the E U orthe prospect of a single European market (or both) tohave been an important determinant of Japanese FDI tothe E U ; however, some (notably Nicolaides and Thomsen(1991)) have suggested that preparation for the singlemarket affected mainly the timing, rather than the long-run level, of direct investment in the EU. Besides theUnited States and the EU, trade restrictions also moti-vated some of Japan's FDI in other Asian countries, theexports of which were not subject to the trade barriersaimed at Japanese exporters.

Although industrial-organization considerations maywell explain a major part of Japan's trend FDI and someof the increase in Japanese FDI in the 1980s, most ana-lysts agree that macroeconomic factors are also neededto explain the increase.12 Indeed, the purchase of propertyabroad (which was responsible for 24 percent of theincrease in Japanese FDI in 1986-90) and part of theinvestment in banking and finance, though recorded asdirect investment, are clearly more akin to portfolio in-vestment and therefore unrelated to industrial-organiza-tion motives.13 Among the macroeconomic determinants,three factors seem to have played key roles: the apprecia-tion of the yen; the extraordinary rise in asset prices;and liberalization of foreign exchange controls. Albeitpotentially important, changes in tax policy and the busi-ness cycle do not appear to have had a significant effect.

Between the fall of 1985 and late 1988, the yen appreci-ated by 42 percent in real effective terms. This strengthen-ing of the yen brought about a sharp decline in the costof production and investment in host countries relativeto the cost in Japan, raising the profitability of directinvestment. Exchange rate appreciation also workedthrough other channels to increase FDI. For example,with a stronger yen, Japanese firms, whose book values

11 The "tariff-jumping" or "defensive" FDI hypothesis that firmsinvest abroad to avoid trade barriers was introduced by Mundell(1957). He used a traditional Heckscher-Ohlin trade model to showthat when one country imposes a tariff on its importable (capital-intensive) good, this wil l generate an inflow of capital from the othercountry that wil l substitute for trade.

12 See Barrell and Pain (1993), Caves (1993), Froot (1991), Georgiouand Weinhold (1992), Graham and Krugman (1991), Komiya andWakasugi (1991), and Thomsen (1993).

13 See Tabata (1990) for a discussion of Japanese banks' FDI to theUnited States.

rose compared with those of foreign companies, wereable to collateralize assets to finance new investmentmore easily than were their competitors in countries withdepreciated currencies (Froot and Stein (1991)). Thisenabled Japanese investors to pay higher prices thanliquidity-constrained bidders in host countries. Anotherpossible channel, often mentioned in popular discussionsbut implying irrational behavior, is that Japanese invest-ors may have focused on the comparatively low pricesof physical assets in host countries with depreciated cur-rencies, neglecting the question of whether the economicreturns were equivalent (Graham and Krugman (1991,p. 46)). Whatever the transmission mechanism, empiricalstudies, such as those by Caves (1989), Froot and Stein(1991), and Mann (1993), have typically found a signifi-cant positive relationship between exchange rate appreci-ation and outward FDI.

In the second half of the 1980s, the Japanese economyexperienced an asset price bubble during which landprices doubled and stock prices tripled.14 The spectacularprice increases were in part driven by an easy monetarypolicy, and they enabled Japanese companies to increasetheir overseas investment both through lower cost ofcapital and through additional liquidity. The increasedliquidity was created by Japanese banks, which couldcount 45 percent of their large, unrealized capital gainson stockholdings as capital, enabling them to lend morethan competitors elsewhere. The borrowers, includingcompanies engaged in FDI, could in turn use high-pricedland and equities as collateral to increase their investment.As noted by Graham and Krugman (1991, p. 47), to theextent that the asset price increases represented a bubble,the lower cost of capital can be interpreted as a subsidyto Japanese firms investing abroad that was paid by thosewho bought the overpriced stocks and land.

Liberalization of capital controls, described in the firstpart of this section, also contributed to the spurt of FDIin the 1980s. Although Japanese enterprises in manufac-turing, commerce, and services were generally free toundertake FDI even before the revision of foreign ex-change laws in December 1980, after the revision finan-cial and insurance firms also became free to invest abroad.In the event, FDI in finance and insurance came to ac-count for 31 percent of the increase in Japanese FDI in1986-90.

Changes in corporate taxation, particularly in theUnited States and Japan, provide another possible (butprobably not a major) explanation for developments inJapan's FDI since 1980. According to the Japanese taxsystem, foreign subsidiaries of Japanese firms pay corpo-rate taxes in the host country. However, when they repa-triate income to their parent company, they are liable to

14 See Section VI of this volume for a description and analysis ofthe asset price bubble.

31



taxation at the Japanese rate, with a credit for taxes paidabroad. Under this system, a cut in the corporate tax rateabroad puts host-country firms, which receive the fullbenefit of the tax cut, at an advantage relative to thesubsidiaries of Japanese firms, whose lower foreign taxliability would be offset by an increased liability in Japan.Thus, the 1981 corporate tax cuts in the United States,notably the introduction of accelerated depreciation,acted as a disincentive to Japanese FDI in the UnitedStates, whereas the 1986 tax reform, which eliminatedthe special investment incentives, removed this biasagainst foreign ownership (Graham and Krugman (1991,pp. 47-49)). Similarly, changes in the corporate tax ratein Japan affect the incentives of Japanese firms to investoverseas. The most significant change in Japan's corpo-rate taxation in recent years was a 4.5 percentage pointcut in the corporate tax rate in 1990, which improvedthe profitability of domestic investment and thereforeshould have discouraged outward FDI. Although theabove-mentioned tax changes in the United States andJapan are likely to have made some contribution to theups and downs of Japanese FDI over the past decade,empirical studies (reviewed in Iwamoto (1990)) typicallyfind taxation to have played only a subsidiary role.

The business cycle has also been mentioned as a deter-minant of FDI (Graham and Krugman (1991, pp. 50-1)).There is no obvious reason for cyclical fluctuations toaffect the share of foreign control in the host country ina systematic way. However, the balance of paymentsmeasure of FDI includes intrafirm financing, which isaffected by economic prospects and financing conditionsin both the source and host countries. Indeed, simpleregressions suggest that world FDI behaves procyclically,rising faster than world output during periods of recoveryand falling faster during recessions. Thus, the economicboom in Japan and other major industrial countries inthe second half of the 1980s could have been responsiblefor some of the rapid increase in Japanese FDI. Giventhat the growth of FDI was many times faster than thatof output, other factors are likely to have played a moreimportant role.

What explains the marked decline in Japanese overseasinvestment after 1990? It is difficult to pinpoint industrial-organization factors that might have played a role, but thereversal of a number of macroeconomic developmentsseems to provide a plausible explanation: the real effec-tive value of the yen depreciated by 22 percent betweenlate 1988 and mid-1990 and, despite subsequent gradualappreciation, remained below the earlier peak until 1993;the asset price bubble collapsed, with land and equityprices returning to trend levels by 1993; the stock adjust-ment to the removal of capital controls was likely tohave been completed by the early 1990s; and the Japaneseeconomy has been in recession since mid-1991. It is stilltoo early to tell whether the marked appreciation of theyen in 1993-94 will induce another burst of JapaneseFDI in the mid-1990s.

Characteristics of JapaneseOverseas Subsidiaries

As discussed above, standard industrial-organizationand cost-of-capital theories seem to go a long way towardexplaining the main developments in Japan's FDI. Someobservers have noted, however, that the overseas affiliatesof Japanese companies differ from those of other coun-tries in several respects. They have a greater propensityto import—for example, Japanese manufacturing subsidi-aries import three times as much per worker as do otherforeign manufacturing affiliates in the United States,which themselves are more import-intensive than domes-tic companies (Graham and Krugman (1991, p. 78)). TheJapanese parent companies import relatively less fromtheir affiliates and strongly direct their interfirm exportsto their foreign distribution subsidiaries. Japanese compa-nies have also displayed a marked (but diminishing) pref-erence to establish their investments by building a newfacility ("greenfield" investment) instead of acquiringcontrol of an existing company (Caves (1993, p. 291)).Furthermore, the affiliates' profitability is low comparedwith that of U.S. firms. The following average profit ratesfor 1983-88 were adapted from Komiya and Wakasugi(1991):15 Japanese-owned overseas subsidiaries, 0.4 per-cent (average for all Japanese corporations, 0.9 percent);U.S.-owned overseas subsidiaries, 5.0 percent (averagefor all U.S. corporations, 2.7 percent).

The differences can probably be attributed primarilyto the more recent origin of Japanese FDI rather than toany difference in the underlying microeconomic behav-ior. Because of the recent rapid growth of Japanese FDI,the average Japanese affiliate has been in operation fora far shorter time than its counterpart from other coun-tries, let alone the average domestic competitor. Thereis little evidence to suggest that the above-mentioneddifferences will persist once Japanese FDI matures.Nevertheless, the possibility that reasons such as corpo-rate group (keiretsu) relationships contribute to these dif-ferences cannot be excluded.

Effects of Japanese Foreign Direct Investmenton Other Economies

The general effects of foreign direct investment arewell known. In the absence of distortions, FDI generatesnet benefits. These benefits include not only the usualgains from trade and international integration—exploita-tion of comparative advantage, a larger and more efficientscale of production, and increased competition—but alsopositive externalities in the host economy in the formof new technology, work force skills, and managementtechniques (Graham and Krugman (1991), Nicolaides(1991), and Georgiou and Weinhold (1992)).

15 Profits after taxes, divided by sales.

32



Despite the general presumption that FDI is beneficial,concern has been expressed that foreign investment car-ries costs for the host country. The adverse effects ofFDI are often claimed to include loss of employmentand negative trade balance effects. Critics of FDI arguethat as foreign-owned companies tend to be import inten-sive, their activities will result in reduced demand fordomestic products, which will cost jobs. In turn, advo-cates of FDI point to anecdotal evidence that foreign-owned companies employ a large number of workersand account for a significant share of job creation in manycountries. For example, Japanese-owned manufacturingcompanies employ 100,000 Europeans (Thomsen (1993,p. 302)). In the United States, employment in Japanese-controlled firms increased by 428,000 workers between1977 and 1989 (Graham and Krugman (1991, p. 26)),and in 1987-89 one third of all new manufacturing jobswere the result of Japanese FDI.

Both of these anecdotal views are somewhat mis-leading because they are based on partial equilibriumanalyses. To be sure, foreign investment, like investmentby domestic companies, will cause changes in the distri-bution of employment within the country and acrossindustries, but the net effect on employment is likely tobe small beyond the short run. To the extent that FDIhas no influence on potential output and the natural rateof unemployment, it should not affect the long-run levelof employment. If there is any long-run effect, however,it is likely to be positive: FDI may enhance potentialoutput, and thereby employment, through the externali-ties discussed earlier.

That Japanese-owned firms tend to have a higher pro-pensity to import than do their host-country counterpartshas prompted critics to claim that Japanese FDI worsensthe external position of host countries. This criticismis again based on partial equilibrium considerations: ifproduction by Japanese affiliates crowds out domesticfirms with a lower import content, the sectoral tradebalance will worsen in the short run.16 Standard macro-economic theory suggests, however, that any permanenteffects of FDI on the aggregate current account balanceare likely to be small (Graham and Krugman (1991,pp. 63-4)). By definition, the current account position isthe difference between domestic saving and domesticinvestment. Changes in import propensities, like othermicroeconomic factors, cannot influence the current ac-count unless they affect the fundamental determinantsof saving or investment. However, the higher importpropensity of Japanese affiliates will, if sustained, putdownward pressure on host-country currencies. A depre-ciated exchange rate will compensate for the higher im-

16 Partial equilibrium arguments could also lead to the opposite con-clusion: if an FDI project has been undertaken to replace exports withhost-country production (perhaps to circumvent trade restrictions),the trade balance of the host country wil l improve.

port intensity, encouraging exports, discouraging im-ports, and thereby leaving the overall current accountposition (but not sectoral trade balances) unchanged. Esti-mates for the United States, the main destination forJapan's FDI, suggest that the quantitative importance ofthe increased share of foreign control on the value of thedollar is small (Graham and Krugman (1991, pp. 69-70)).

Effects on Japan

The discussion in the previous subsection underscoresthat, in the long run, overseas investment should havelittle effect on Japan's employment and current accountbalance: adjustments in the domestic labor market willbring aggregate employment to the "natural" level, andexchange rate changes will keep the external position,as determined by saving and investment behavior, un-changed. Much attention in Japan has, however, beenpaid to the influence of growing outward FDI on thetrade balance in the short term and medium run (see,for example, Dai-ichi Kangyo Bank (1994) and JapanResearch Institute (1990)).

A stylized (partial equilibrium) view of the impact ofFDI on the trade balance can be summarized as follows.In the early stages of overseas production, the trade bal-ance is likely to improve, owing to exports of capitalgoods and intermediate inputs from Japan to the newlyestablished subsidiaries abroad. Later, the trade balancestarts to decline, both because of reverse imports (exportsto Japan by subsidiaries abroad) and because goods pro-duced by overseas subsidiaries replace exports in localor third-country markets. The latter negative effect onthe trade balance is in part offset by decreased importsof inputs for the relaced exports. Similar to the J-curveeffect of exchange rate changes, this sequence of move-ments in the trade surplus is sometimes called the "over-seas investment J-curve effect."

In the long run, once (general equilibrium) adjustmentsin the exchange rate in response to FDI flows have takenplace, the presumption is that the trade balance will beworse than in the absence of FDI by an amount equivalentto the repatriated profits of the subsidiaries. An indicationof the magnitude of the likely long-run effect of FDI onJapan's trade account is obtained by looking at recenttrends in overseas direct investment income (Chart 3-7).Between 1978 and 1993, direct investment income fromJapanese-owned companies abroad increased tenfold, toover $8 billion. Income from direct investment has alsorisen relative to the size of the economy, but still amountsto only about 0.2 percent of GDP. The return on FDI(ratio of repatriated profits to the stock of FDI) has,however, declined to about 3 percent in recent years,compared with an average of 6 percent until the mid-1980s. Assuming that this decline, which is likely toreflect the usual lags between investment and profits, isreversed and that direct investment grows in line with

33



Chart 3-7.

Sources: Bank of Japan, Balance of Payments Monthly (Tokyo,various issues); and Nikkei Telecom.

1 Repatriated profits relative to the stock of foreign directinvestment (balance of payments data).

GDP, investment income from FDI could rise to 0.4percent of GDP.

References

Agarwal, Jamuna P., "Determinants of Foreign Direct Invest-ment: A Survey," Weltwirtschafliches Archiv, Vo l . 116,Heft 4 (1980).

Barrell, Ray and Nigel Pain, "Trade Restraints and JapaneseDirect Investment Flows," National Institute of Eco-nomic and Social Research Discussion Paper 43 (London,March 1993).

Caves, Richard E., "Exchange Rate Movements and ForeignDirect Investment in the United States," in The Interna-tionalization of U.S. Markets, edited by D.B. Audretschand M.P. Claudon (New York: New York UniversityPress, 1989), pp. 199-228.

, "Japanese Investment in the United States: Lessonsfor the Economic Analysis of Foreign Investment,"World Economy, Vo l . 16 (May 1993), pp. 279-300.

Dai-ichi Kangyo Bank, "Japan's Overseas Direct Investmentsat a Turning Point," DKB Economic Report, Vo l . 24(February 15, 1994).

French, Kenneth R., and James M . Poterba, "Investor Diversi-fication and International Equity Markets," AmericanEconomic Review, Papers and Proceedings, Vol . 81(May 1991), pp. 222-26.

Froot, Kenneth A. , "Japanese Foreign Direct Investment,"N B E R Working Paper 3737 (Cambridge, Massachusetts:National Bureau of Economic Research, June 1991).

Froot, Kenneth A. , and Jeremy C. Stein, "Exchange Ratesand Foreign Direct Investment: A n Imperfect CapitalMarkets Approach," Quarterly Journal of Economics,Vol . 106 (November 1991), pp. 1191-1217.

Georgiou, George C , and Sharon Weinhold, "Japanese DirectInvestment in the U.S.," World Economy, Vol . 15 (No-vember 1992), pp. 761-78.

Graham, Edward M . , and Paul R. Krugman, Foreign DirectInvestment in the United States, 2nd ed. (Washington:Institute for International Economics, 1991).

Healey, Derek T., Japanese Capital Exports and Asian Eco-nomic Development (Paris: Development Centre of theOECD, 1991).

Heitger, Bernhard, and Jurgen Stehn, "Japanese Direct Invest-ment in the EC—Response to the Internal Market 1992?"Journal of Common Market Studies, Vol . 29 (September1990), pp. 1-15.

Hennart, J.-F., and Y. -R. Park, "Location, Governance, andStrategic Determinants of Japanese Manufacturing In-vestment in the United States," Working Paper (Cham-paign: University of Illinois, 1991).

International Monetary Fund, Report on the Measurement ofInternational Capital Flows (Washington, 1992).

Iwamoto, Yasushi, "Japanese Corporate Tax Policy and DirectInvestment Abroad," Australian National UniversityWorking Papers in Economics and Econometrics, No.204 (Canberra, May 1990).

Japan Research Institute, "Research Study on Influences ofGrowing Foreign Direct Investments on InternationalStructure of Industry and Trade," English summary(1990).

Jones, Ronald W., and J. Peter Neary, "The Positive Theoryof International Trade," in International Trade: Surveysof Theory and Policy, edited by Ronald W. Jones (Am-sterdam and New York: North-Holland, 1986).

Kawai, Masahiro, "Japanese Investment in Foreign Securitiesin the 1980s," Department of Economics Discussion Pa-per 91-37 (Vancouver: University of British Columbia,September 1991).

Kogut, Bruce, and Sea Jin Chang, "Technological Capabilitiesand Japanese Foreign Direct Investment in the UnitedStates," Review of Economics and Statistics, Vol . 73(August 1991), pp. 401-13.

Kojima, Kiyoshi, and Terutomo Ozawa, "Micro- and Macro-economic Models of Direct Investment: Toward a Syn-thesis," Hitotsubashi Journal of Economics, Vol . 25(June 1984), pp. 1-20.

Komiya, Ryutaro, and Ryuhei Wakasugi, "Japan's ForeignDirect Investment," Annals of the American Academy ofPolitical and Social Science, Vol . 513 (January 1991),pp. 48-61.

Lizondo, J. Saul, "Foreign Direct Investment," in InternationalMonetary Fund, Determinants and Systemic Conse-quences of International Capital Flows, Occasional Pa-per 77 (Washington, March 1991), pp. 68-82.

34


Mann, Catherine L. , "Determinants of Japanese Direct Invest-ment in U.S. Manufacturing Industries," Journal of Inter-national Money and Finance, Vo l . 12 (October 1993),pp. 523-41.

Mundell, Robert A. , "International Trade and Factor Mobil-ity," American Economic Review, Vo l . 47 (1957),pp. 321-35.

Nicolaides, Phedon, "Investment Policies in an IntegratedWorld Economy," World Economy, Vol . 14 (June 1991),pp. 121-37.

Nicolaides, Phedon, and Stephen Thomsen, "Can Protec-tionism Explain Direct Investment," Journal of Com-mon Market Studies, Vo l . 29 (December 1991),pp. 635-43.

Sleuwaegen, Leo, "Foreign Direct Investment and Intra-FirmTrade: Evidence From Japan," Institute for EconomicResearch Discussion Paper 9002-G (Rotterdam: ErasmusUniversiteit, 1990).

Tabata, Naoki, "Japanese Banks' Direct Investment to theUnited States," in JCIF Policy Study Series, No. 16 (To-kyo: Japan Center for International Finance, 1990),pp. 25-34.

Thomsen, Stephen, "Japanese Direct Investment in the Euro-pean Community: The Product Cycle Revisited," WorldEconomy, Vol . 16 (May 1993), pp. 301-15.

United Nations, Centre on Transnational Corporations, TheDeterminants of Foreign Direct Investment: A Survey ofthe Evidence, ST/CTC/121 (New York, 1992).

References

35


IV Demographic Change and HouseholdSaving in JapanGuy Meredith

D emographic projections for coming decades indi-cate that, of the major industrial countries, Japan

will experience the most rapid increase in the share ofthe elderly in the total population.1 This demographicshift is expected to cause pressures on the financing ofthe social security system and slower growth in the laborforce and in potential output. Many observers also believethat a consequence of the transition to a more elderlypopulation will be a reduction in the private saving rate,based on the view that retired households save less thanthose of working age. Such a decline would reduce theflow of saving to both Japan and overseas economies,implying a reduction in domestic investment or the exter-nal surplus, or both.

The view that the transition to a more elderly popula-tion will reduce the aggregate saving rate is controversial,however. Broadly speaking, there are two schools ofthought on the issue of saving and demographics. Thefirst, associated with the life-cycle model of householdbehavior, views saving as being motivated by the desireof households to smooth lifetime consumption in the faceof uneven income flows. The saving rate of the working-age population is positive, whereas that of the retiredpopulation is negative. An increase in the ratio of theretired population to the working-age population thuslowers the aggregate saving rate. The second view fol-lows from evidence suggesting that the saving rate ofthe elderly is not significantly lower than that of theoverall population, a phenomenon that is sometimes at-tributed to the existence of an altruistic bequest motivefor saving. An inference drawn from the apparent similar-ity of saving rates of different age groups is that a shifttoward a more elderly population will have little effecton the aggregate saving rate.

This section first reviews the literature on demograph-ics and saving, with a focus on the Japanese experience.As shown in the next part, the aggregate evidence onsaving and demographic structure for Japan and othercountries appears to be consistent with the life-cyclemodel. Some studies using household data, however,seem to contradict predictions of the life-cycle model.In the third part, the household data for Japan are re-

1 A comparison of the outlook for the major industrial countries ispresented in Masson and Tryon (1990).

examined. These data show that retired households inJapan do, in fact, dissave; the rate of dissaving is magni-fied when income is adjusted for social security benefits.Indeed, it appears that an inappropriate treatment of socialsecurity and pension income is a flaw in many household-level studies, both for Japan and other countries. Thefourth part presents simulation results, based on a life-cycle model, that suggest that Japan's private savingrate will decline significantly as a result of demographicfactors. The last part summarizes the results.

Previous Studies on Demographicsand Saving

Evidence Supporting the Life-Cycle Model

Much work has been done on the ability of the life-cycle model to explain aggregate saving.2 As discussedabove, an important prediction of the model is that ashift in the demographic structure toward a higher ratioof elderly households to working-age households willreduce the aggregate saving rate. In addition, life-cyclemodels that include a period of youth preceding workinglife predict that the aggregate saving rate will be nega-tively related to the population share of the young. An-other prediction of the life-cycle model is that savingduring working life will depend on expected income fromsources other than household saving during retirement.For instance, income in the form of benefits from fullyfunded private pension plans is likely to reduce savingby working-age households, excluding contributions topension plans. Unfunded public pension plans also tendto reduce household saving in a life-cycle model, al-though the exact impact is complicated by the implicitrate of return households expect to realize on their contri-butions and the possible earnings-testing of benefits.3

2 For one of the early expositions of the life-cycle model, see Andoand Modigliani (1963).

3 When pension benefits are lump sum and the expected return oncontributions equals the market interest rate, the offset wil l be onefor one. When benefits are earnings-tested, or the rate of return exceedsthe market interest rate, social security may induce households toretire earlier. Higher private saving for a longer retirement period(the "retirement" effect) would then tend to offset, at least partially,the saving "replacement" effect (see Feldstein (1974)).

36


Previous Studies on Demographics and Saving

Table 4-1. Summary of Studies on Demographics and Saving

Note: Lines I, 2, and 3 are cited in Heller (1989); line 5 is from Graham's equation I in Table 2; line 6 is

Evidence based on aggregate data typically supportsthe predictions of the life-cycle model regarding demo-graphics and saving, as indicated by the results shownin Table 4-1. Most of the studies have been based oncross-section data for OECD countries, while others haveused time-series data for Japan only or have pooled cross-section and time-series data for several industrial coun-tries. The methodology typically has involved regressingthe saving rate on variables that capture the demographicstructure of the population and other factors. The demo-graphic structure is summarized by the ratios of the el-derly and the young to the population of working age—that is, the elderly and youth dependency ratios.

In most cases, the demographic effects are both statisti-cally and economically significant. The impact of achange in the elderly dependency ratio typically exceedsthat of the youth dependency ratio: an unweighted aver-age of the estimation results indicates that a 1 percentagepoint rise in the elderly ratio lowers the saving rate by0.86 percentage point, whereas the same increase in theyouth dependency ratio lowers the saving rate by 0.61percentage point. Effects of this magnitude imply a largechange in the aggregate saving rate in response to pro-jected shifts in Japan's demographic structure. Specifi-cally, the elderly dependency ratio is projected to riseby over 20 percentage points from the first half of the

Aggregate cross-section studies

1. Modigliani (1970)

2. Modigliani and Sterling (1983)

3. Feldstein (1980)

4. Horioka(l986)

5. Graham (1987)

6. Koskela and Viren (1989)

7. Horioka(l99l)

8. OECD (1990)

Time-series studies

9. Shibuya (1987)

10. Horioka(l99l)

11. Masson and Tryon (1990)

Average estimated impact

Data Source

—

—

—

21 OECD countries1976-82 average

24 OECD countries, 1975or 1970-80 average

23 countries,1979-83 average

21 OECD countries

14 OECD countries,1980-88 average

1966-83 (Japan)

1956-87 (Japan)

1969-87 pooled3

—

Effect on Saving Rate of1 Percentage Point Risein Demographic Ratio

Youth1

-0.20 (3.7)

-0.13 (1.4)

-0.77 (3.9)

-0.92(4.2)

-0.87(2.9)

-0.73 (1.7)

-0.44 (1.7)

-0.30 (5.1)

-1.10

-0 .61

Elderly2

-0.88 (3.1)

-0.51 (4.3)

-1.21 (2.7)

-1.61 (4.0)

0.12 (0.3)

-0.76 (0.8)

- 1.09 (2.4)

-0.93 (2.4)

-0.34 (3.8)

-1.13(3.7)

-1.10

-0 .86

from Koskela and Viren's equation (6) in Table I; lines 7 and 10 are cited in Horioka (1991); line 9 is asimulated effect based on 1980 values; line I I is the simulated post-1980 impact. Figures in parentheses areestimated t-statistics.

1Ratio of population aged 0-19 to population aged 20-64.2Ratio of population over 64 to population aged 20-64, except line 8 is ratio to total population.3Pooled data for the seven major industrial countries and the small industrial countries.

37


IV DEMOGRAPHIC CHANGE AND HOUSEHOLD SAVING IN JAPAN

Other studies have cast doubt on the applicability ofthe life-cycle model for both Japan and other countries,on the basis of the observed saving behavior of the elderlyin household data. In particular, these studies suggest thatthe saving rates of elderly households are not significantlylower than those of working-age households; that theelderly do not decumulate assets; and that elderly house-holds leave substantial bequests. Hayashi, Ando, andFerris (1988), for instance, found only limited supportfor the life-cycle model in the behavior of elderly single-member households in Japan. For the elderly more gener-ally, they attributed the apparent lack of wealth decumula-tion to the importance of intergenerational bequests. Bos-worth, Burtless, and Sabelhaus (1991) attempted to inferthe effect of demographic changes on Japan's aggregatesaving rate by examining differences in age-specific sav-ing rates. In particular, they held the saving rate of eachage group constant and changed the population sharesof each group to obtain alternative aggregate savingrates.5 The size of the effect on the aggregate saving ratedepended on how different the age-specific saving rateswere; these authors concluded that the differences insaving rates were not large enough to affect the aggregatesaving rate (Bosworth, Burtless, and Sabelhaus (1991,pp. 220-21)).

The (apparent) high saving rate of the elderly hassometimes been explained in terms of a dynastic modelof household behavior, in which the current generationcares about the welfare of the next generation (see Barro(1974)). Such intergenerational altruism can motivateintentional bequests, explaining why the propensity ofelderly households to consume out of wealth might be

lower than that predicted by the life-cycle model. How-ever, although the dynastic model can explain intentionalbequests, it is less clear that it can explain why the savingrates of elderly households would be similar to thoseof working-age households.6 In particular, in both thedynastic and life-cycle models, households will smoothconsumption over time. If income is not similarly smooth,then the household saving rate will vary systematicallywith income in either framework. Continued saving bythe elderly, then, must imply that their incomes do notdecline with age.7 This is not a prediction of either thelife-cycle or dynastic model per se, but rather must reflecteither continued high labor supply by elderly households,or the replacement of labor income by other sources ofincome as households age. These issues are examined inthe next part of this section.

4 Horioka (1991) has presented calculations that yield even largereffects, implying that the household saving rate would become sig-nificantly negative as early as 2010.

5 Appendix 4-1 provides an example of why holding the saving ratesof different age groups constant is inappropriate in the presence ofa social security program that is not fully funded.

6 See Auerbach, Cai, and Kotlikoff (1990) for a dynastic model thatgenerates significant changes in saving rates as a result of demo-graphic transitions.

7 Imperfections in capital markets could limit the ability of house-holds to smooth consumption by borrowing against future earnings.To the extent that such constraints are relevant in the life-cycle model,consumption would tend to be lower in the early stages of life andhigher in the retirement period. This would tend to magnify the gapbetween the saving rates of working-age and elderly households.

8 Interest income on the remaining principal is small relative to therate at which the principal is being run down.

38

Measurement of Savingin Household Data

There are two difficulties with studies of householdsaving that show continued saving by the elderly. Thefirst is that income and wealth are often defined inappro-priately, in that no distinction is made between earnedincome and other sources of income. The second is thathousehold saving is often not observed directly; rather,saving is inferred from hypothetically constructed wealthprofiles of the elderly, and these profiles may be subjectto considerable mismeasurement.

As regards the definitions of income and wealth, itis important to distinguish between earned income andincome from other sources such as social security benefitsand private pension plans. The former yields a flow ofgoods and services to the economy as a whole, while thelatter represents either a transfer from other householdsor the running down of assets that are not included inhousehold balance sheets. Income from fully funded pri-vate pension plans is an example of retirement incomethat largely represents asset decumulation.8 Benefits froma pay-as-you-go social security program represent trans-fers from working-age to retired households—consump-tion out of this income will lower the saving rate for theeconomy as a whole, unless consumption by working-age households falls by an equivalent amount. This will

Challenges to the Life-Cycle Model

1990s to 2020. In the absence of changes in other factorsthat affect saving, these results would suggest a reductionin the household saving rate of over 15 percentage points.4

Fewer studies have examined the impact of socialsecurity on private saving in Japan. Using aggregate time-series data from the period 1966-83, Shibuya (1987)found a significant negative effect from the social securityreplacement ratio. Yamada, Yamada, and Liu (1990) alsofound a significant effect of social security retirementbenefits on both the saving and retirement decisions,using time-series data for the 1951-82 period. The savingreplacement effect dominates the retirement effect, lead-ing to a strong negative overall impact on saving fromthe introduction of the social security system.


Using a Life-Cycle Model

not, in general, be the case when households are forwardlooking. As illustrated in Appendix 4-1, working-agehouseholds will spread the effect of a change in socialsecurity taxes over both current and future consumption.A rise in transfers from the young to the elderly thenreduces the consumption of the young by less than itincreases that of the elderly, causing the aggregate savingrate to fall.

The second difficulty with some household studies isthat they must infer, only on the basis of observationsfor a single period of the wealth of households of differentages, how the wealth of households will change overtime.9 There are several reasons why this is problematic,especially in the case of Japan. One is that the elderlypoor tend to get absorbed in the families of their childrenand thus drop out of the sample of households. Looking,in a given year, at the profile of household wealth byage of the household head then tends to overrepresentrelatively rich elderly households. Bias also arises be-cause the elderly who have been hospitalized on a long-term basis are not included in survey data. Their savingrates are typically very negative, especially when thepublic health care component of their consumption isincluded. Furthermore, adjustments must be made be-cause households of different ages have different lifetimeincomes and, thus, asset wealth owing to productivitygrowth and other factors. Although researchers typicallyattempt to control for these "cohort" effects, the tech-niques are often arbitrary. In general, the power of theresults is weakened by the need to impute key informa-tion.10 Finally, the translation of wealth profiles into sav-ing rates is complicated by unanticipated capital gains andlosses on existing wealth, as well as inter vivos transfers.

The problems of adjusting for unearned income andimputing wealth profiles can be avoided by using directinformation on the consumption and income by source ofelderly households. This permits the distinction betweenearned and unearned income and obviates the need toconstruct artificial age-wealth profiles. Such data werecompiled in Japan for both retired and working elderlyhouseholds in the 1990 Annual Report on the FamilyIncome and Expenditure Survey (FIES; Japan (1990)),as shown in Table 4-2.

Elderly retired households in Japan account for 11 1/2percent of the total households covered in the survey,whereas elderly working households account for 5 per-cent.11 For retired households, living expenditures exceeddisposable income by a significant margin: on average,

these households dissave at an annual rate of 21 percent,even before adjusting for unearned income. In the event,unearned income in the form of social security benefitsaccounts for by far the largest component of income forretired households, averaging over 70 percent of totalincome.12 For elderly households whose head is stillworking (whose average age is significantly lower thanthat for retired households), the situation is quite differ-ent: the saving rate remains positive at 17 1/2 percent—income is almost double that of retired households, whileexpenditures are only 30 percent higher. These data sug-gest that, after retirement, household income drops offmore sharply than consumption, resulting in dissavingfor retired households.13 The drop in earned income, how-ever, is partially buffered by a rise in transfer payments,and this moderates the change in the saving rate thatwould otherwise be observed.

The aggregate saving rate of both retired and workingelderly households is minus 3 1/2 percent, compared withthe 25 percent saving rate of working-age householdscomputed from the 1990 FIES. Assuming that the savingrates of these two groups remain unchanged, it is possibleto calculate the effect of a shift in the demographic struc-ture toward a higher proportion of elderly. Specifically,a rise in the elderly dependency ratio of 1 percentage pointwould lower the aggregate saving rate by 0.2 percentagepoint, holding the saving rate of elderly and working-age households constant.14 While significant in light ofthe size of the projected rise in the elderly dependencyratio, this response is substantially lower than the averageestimated effect obtained using aggregate data shown inTable 4-1. The difference is examined in more detail inthe next part of this section.

It was shown above that there are significant differ-ences in the saving rates of elderly and working-agehouseholds in Japan. Nevertheless, holding the savingrate of each group constant, the impact of a demographicshift on the aggregate saving rate is smaller than thatsuggested by some econometric studies. A more sophisti-cated estimate of the effect of demographic shifts onaggregate saving is presented below. Specifically, a life-cycle model for Japan is simulated using projectedchanges in the age composition of the population. The

9 This is especially true for the United States, where insufficient datahave been available on household consumption to observe householdsaving rates directly.

10 Campbell (1991), for instance, has questioned the approach takenby Hayashi and others (1988).

11 Elderly households are defined as having a household head aged60 years or over.

12 Takayama (1992) observes a similarly high ratio of social securitybenefits to household income in the 1984 National Survey of FamilyIncome and Expenditures.

13 These data are only suggestive because of possible differences inthe characteristics of working and retired households.

14 The initial elderly dependency ratio is 19 percent, and the ratioof the disposable income of working-age to elderly households is 1.8.

39

Simulation Usinga Life-Cycle Model

Simulation



Table 4-2. Income, Consumption, and Saving of Retired Households1

(Monthly average in thousands of yen)

Household incomeTaxes

Tax rate (in percent)Disposable income

Wages and salariesBusiness and homeworkSocial security benefitsOther

Disposable income excluding socialsecurity benefits ("earned" income)

Living expenditures

Saving ratesTotal saving rate (in percent)3

Saving rate as a ratioto earned income4

Ratio of earned income to totalhousehold living expenditures

1988

201.524.011.9

177.427.84.0

143.014.2

51.4

212.9

-20.0

-314.2

24.1

1989

209.023.711.3

185.332.44.6

143.914.0

57.7

220.2

-18.8

-281.6

26.2

1990

225.524.410.8

202.831.54.5

I65.92

12.1

54.8

228.3

-12.6

-316.6

24.0

results indicate an effect on saving exceeding that sug-gested by differences in household saving rates, but lowerthan the average estimate obtained using aggregate data.

The life-cycle model is described in detail in Appendix4-2. It follows in the tradition of Tobin (1967) and White(1978), in that the focus is on the behavior of individualhouseholds, with variables such as the real interest rateand the real growth rate held to be exogenous. The repre-sentative household is forward looking and maximizeslifetime utility subject to a budget constraint. Householdincome consists of after-tax labor income, interest incomeon accumulated assets, social security benefits, and inher-ited wealth. The consumption path is affected by factorssuch as the rate of time preference, the real interest rate,and the intertemporal elasticity of substitution. Uncer-tainty about the time of death causes households to plan toconsume less toward the end of their life span, reflecting alower probability that they will be alive to benefit fromthis consumption.15 Uncertainty about death also gener-

ates a precautionary demand for saving. Because wealthcannot become negative even if the household lives be-yond the average life expectancy, households plan tofinance consumption over a longer life span than theyexpect, on average, to live. This implies that householdstypically die with positive wealth and leave bequests tothe next generation, even though they derive no utilityfrom bequests per se.

Initial values for certain variables were chosen tobe consistent with observed data for Japan during the1970-90 period: in particular, the risk-free real interestrate averaged 3 percent;16 labor-augmented productivitygrowth averaged PA percent; and population growth av-eraged 1 percent. Parameters determining consumptionprofiles were similar to those used in other studies ofhousehold behavior.17 Simulating the model with theseassumptions yielded a household saving rate of 16.7 per-

15 See Hurd (1990) for a discussion of the effect of mortality hazard

16 As measured by the ex post yield on government bonds, adjusted

40

on consumption profiles.

for increases in the GNP deflator.17 The parameterization of the model is discussed in more detail in

Appendix 4-2.

Source: Japan, Management and Coordination Agency (1990, Table 18, p. 114).1Defined as households whose head is jobless and over age 60.2The surge in 1990 is due to a reduction in the payment interval for annuity benefits from three to two

months.3Defined as disposable income less living expenditures as a share of disposable income.4Saving rate out of disposable income, excluding social security benefits.


cent and a ratio of household assets to income (that is,accumulated wealth) of 5.5, compared with observed data(1970-90 average) of 18.0 percent and 5.5, respectively.

The predictions of the model are very similar to theactual data, both for the saving rate and the level ofaccumulated assets. The appropriateness of the life-cyclemodel was also tested by comparing its predictions forthe behavior of elderly households with the observeddata. The household survey data obtained from the 1990FIES were first corrected for imputed rent on owner-occupied housing by multiplying the estimated value ofowner-occupied housing by the rate of return on otherassets in the model. This imputed rent was added toboth the observed income and consumption of elderlyhouseholds, and the saving rate was recalculated on aconsistent basis. For elderly households in 1990, themodel's prediction for the saving rate was — 2 percentcompared with an adjusted observed saving rate of- 2 1/2 percent. Predictions for the share in income ofearned income, social security benefits, and other incomewere 31, 34, and 35 percent, respectively, compared withadjusted survey data of 30, 32, and 38 percent. Again,the predicted and actual series correspond closely. Theconsistency of the predictions with the observed data isreassuring evidence that the life-cycle model can pro-vide a realistic characterization of household savingbehavior.18

The model was then simulated to track projected demo-graphic changes over the period 1990-2030. The elderlydependency ratio is projected to rise from 17.7 in 1990to 46.2 in 2030. The model results for the projectedhousehold saving rate were: 1990, 21.3 percent; 2000,18.7 percent; 2010, 14.7 percent; 2020, 12.0 percent;2030, 10.4 percent—implying a total change during the1990-2030 period of -10.9 percent. Of course, thissimulation indicates only the change in the householdsaving rate attributable to demographic considerations;the actual change in the saving rate will depend on otherfactors. For instance, the decline in saving would tendto be moderated by a rise in the real interest rate iflower saving otherwise would cause a shortage of savingrelative to investment.19

The ratio of the simulated change in the householdsaving rate to the change in the elderly dependency ratioover the 1990-2030 period is about 0.4. This lies betweenthe estimates discussed above: specifically, the ratio of0.2 obtained by changing population shares while holding

18 This evidence supports the original findings of Tobin (1967) forthe United States and contradicts those of White (1978) and otherauthors who have suggested that the life-cycle model is not capableof explaining observed household saving behavior.

19 The saving rate would also be affected by the stance of fiscalpolicy. These simulations assume that the pension system is operatedon a pay-as-you-go basis, implying a future rise in contribution ratesto balance increasing benefits. Section V of this volume discussesalternative long-run simulations in more detail.

the saving rates of the elderly and the working-age popu-lations constant; and the average estimated impact of0.86 derived from aggregate data.

The difference between the simulation results and theimpact when household saving rates are held unchangedis due primarily to the induced effect on the saving ratesof changes in social security taxes. In particular, socialsecurity tax rates are currently low relative to those thatwill be needed to support a higher ratio of elderly toworking-age households in the future (see Section V ofthis volume). As illustrated in Appendix 4-1, the increasein the tax rate will be reflected partly in lower consump-tion of working-age households, and partly in a reductionin their saving rate. This induced effect on the savingrate of working households reinforces that of the shift inweights toward elderly households with a lower savingrate.

It is more difficult to reconcile the predictions of thelife-cycle model with the estimation results obtainedusing aggregate data. One possibility is that the householddata on which the life-cycle model is based may underrep-resent some elderly households that have high rates ofdissaving, such as those in which the head of the house-hold is hospitalized. In this case, the picture provided bythe aggregate data would be more accurate. However,estimation using aggregate data has yielded a consider-able range of estimates. The lack of consensus reflects,in part, that differences in the population structure acrosstime periods and countries tend to be correlated withother macroeconomic variables. The comovement inthese variables makes the estimation results sensitive tothe exact specification of the equations and can lead to"overfitting" the data.20 In any event, it is interesting thatthe effect predicted by the life-cycle model is quite closeto that obtained in the only study based exclusively onaggregate Japanese data since the mid-1960s (Shibuya(1987)).

This paper has analyzed the probable effect on thehousehold saving rate of a shift in Japan's demographicstructure toward a more elderly population. Althoughmany econometric studies using aggregate data have pre-dicted a very large effect on the saving rate, some studiesusing household data suggest that the impact may besmall.

By using more recent and detailed information onthe income and consumption of retired households, theanalysis has shown that the saving rates for the elderly

20 The collinearity of the explanatory variables implies that t-statisticsmay be low, even though the parameters are large in absolute magni-tude. The choice of a specification that yields statistically significantparameters may imply parameter values that lie on the high end ofthe range of values that would be supported by the data.

41

Conclusions

Conclusions



calculated in some household-level studies may be mis-leading. It appears that the elderly do dissave, and thatthe rate of dissaving is very similar to that predicted bya life-cycle model of household behavior. In addition, alarge share of the income of the retired elderly in Japanconsists of social security benefits, which represent trans-fer payments from the working generation as opposedto a net flow of resources to the economy. The existenceof social security benefits and taxes implies that the effecton the aggregate saving rate of changes in populationshares cannot be assessed by holding the saving rates ofelderly and working-age households constant.

Simulations of a life-cycle model indicate that thehousehold saving rate is likely to decline steadily incoming decades as a result of the shift toward a moreelderly population. Although the effect of this demo-graphic shift is significant, it is smaller than that suggestedby some studies using aggregate data, which have sug-gested that the household saving rate may become nega-tive over the long run. The effect is larger, however, thanthe change implied when the saving rates of the elderlyand working-age populations are held constant, becauseof the role played by social security benefits and taxes.

A(SIY) = E Awn , (A4-1)

where Awi is the change in the share of generation i inaggregate income, and si is its saving rate. Demographicshifts alter wi and thus the aggregate saving rate whensi differs across generations. This decomposition is notuseful, however, when the income of some generationis zero because si is undefined. An alternative approachthat avoids this problem defines saving as a share of con-sumption:

A(S/Q = X Aw/st' , (A4-2)

where Awi and si refer, respectively, to the share ofeach generation in aggregate consumption and its saving-to-consumption ratio.

A simple example of the effect of a demographictransition without social security is shown in the upperpanel of Table 4-3. Each household divides its consump-tion equally between two periods; household incomeequals 1 in the first period and 0 in the second; the interestrate is 0. In the initial equilibrium—period 1—it is as-sumed that there are one "young" household and one

"elderly" household. The aggregate wealth of the econ-omy is 1/2, representing the saving of the young to financeconsumption in retirement. The saving rate of the youngin relation to consumption is 1, while that of the elderlyis - 1 . The aggregate saving rate is 0, since each genera-tion has an equal share in total consumption.

In period 2, the demographic structure shifts as thenumber of young households rises to two while the num-ber of elderly households remains at one. Wealth rises,and the economy-wide saving rate becomes positive. Thechange in the latter can be inferred from equation(A4-2), since the share of the young in total consumptionrises to 2/3 In period 3, the number of young householdsreturns to one, while there are now two elderly house-holds. Economy-wide saving is negative because thewealth accumulated in period 2 is consumed by the el-derly households. In period 4, the demographic transitionis complete: saving and wealth have returned to theirinitial levels. An important aspect of this example is that,in each period, the saving-to-consumption ratio of thetwo generations is unchanged at 1 for the young and — 1for the elderly. This allows the effect on the aggregatesaving rate of demographic shifts to be calculated usingthe above decomposition combined with unchanged sav-ing rates.

Calculating the effect of a demographic transition ismore complex in the presence of an unfunded socialsecurity program. In this case, the tax burden on younghouseholds depends on the number of elderly householdsthat must be supported. Changes in the tax burden affectthe saving decision of young households, leading to aninterdependence between their saving rate and the demo-graphic structure. This is illustrated in the lower panelof Table 4-3. The demographic transition is the same asin the first example; now, however, a social securityprogram transfers income of V2 to each elderly household.These transfers are financed on a pay-as-you-go basis bytaxes on young households. The initial equilibrium inperiod 1 is the same as that described above, exceptthat young households do not save because consumptionduring retirement is financed by social security benefits.This lowers aggregate wealth to 0. In period 2, the socialsecurity tax rate on young households falls becausethe ratio of young to elderly households rises. Younghouseholds save a part of this tax cut to finance consump-tion during retirement; thus, their saving rate becomespositive, and economy-wide saving also becomes pos-itive.

In period 3, the tax rate on the young households risessharply to finance social security benefits for elderlyhouseholds. The saving rates of both generations becomenegative because the elderly consume wealth accumu-lated in period 2, while the young finance a part of theirsocial security taxes by going into debt. In period 4, thedemographic structure returns to the initial equilibrium,but the saving rate is raised because of previous demo-graphic changes. This "echo" effect occurs because the

42

Appendix 4-1. Effects of DemographicChange on Saving With and WithoutSocial Security

In simple versions of the life-cycle model, the savingrate of each generation is independent and unaffected bythe demographic structure of the population. The effectof a demographic transition on the aggregate saving ratethen equals the change in the share of each generationin total income multiplied by its saving rate:


Appendix 4-2

Table 4-3. Effect of a Demographic Transition on Aggregate Saving

elderly household must save out of social security benefitsto repay debt incurred in period 3. Period 5 (not shownin the table) is identical to period 1—both wealth andsaving are 0.

Although this example is very stylized, it indicatessome important aspects of the effect of demographicchange on saving in the presence of an unfunded socialsecurity program. The first is that the saving rates ofthe young and elderly vary over time in response todemographic shifts. This means that the approach ofweighting fixed saving rates by changing populationshares is invalid. For instance, in period 1, the savingrates of both generations are zero. Varying the weightsapplied to these saving rates would imply—incorrectly—an unchanged aggregate saving rate in the face of ademographic shift. The second point is that there is nogeneral tendency for the saving rate of the young toexceed that of the elderly. Indeed, in periods 3 and 4,the saving rate of elderly households exceeds that ofyoung households. Nevertheless, the shift toward a moreelderly population in period 3 reduces the aggregate sav-ing rate by the same amount as in the example in whichthere is no social security and the saving rate of theelderly is always below that of the young.

Appendix 4-2. A Life-Cycle Modelof Household Behavior for Japan

The life-cycle model used in this section is similar tothat described by Tobin (1967) for the United States.The principal differences are that the model used heretakes into account social security benefits and contribu-tions, and also allows for precautionary saving, sincethe risk of death is not insured through private annuitycontracts. In addition, the model used here is solvednumerically rather than analytically, allowing more gen-eral specification of some relationships. The economyconsists of overlapping generations of households, eachof whose behavior follows life-cycle principles. A periodin the model is one year in length. Households are as-sumed to become economically active at age 20, and theoldest possible age is 100. Each life-cycle then consistsof 81 periods, and there are 81 distinct ages of householdsliving at any time.

The time of death is uncertain; the probability of dyingat a given age is based on 1990 mortality tables for Japan.Households maximize expected lifetime utility subjectto the constraint that they cannot die with negative wealtheven if they live to the oldest possible age. Thus the

43

Population

ConsumptionIncomeWealth (end of period)SavingSaving/incomeSaving/consumption

ConsumptionIncome

EarnedSocial securityTotal

Wealth (end of period)SavingSaving/total incomeSaving/consumption

Period 1

Young

1

1/21

1/21/21/21

1/2

1-1/2

1/20000

Old

1

1/200

-1/2—

- 1

1/2

01/21/20000

Total

2

11

1/2000

1

1010000

Period 2

Young

2

Old

1

Total

3

Period 3

Young

1

Old

2

Total

3

Without social security

1211

1/21

1/200

-1/2—

- 1

11/2

21

1/21/41/3

1/21

1/21/21/21

100

- 1—

- 1

11/2

11/2

-1/2-1/2-1/3

With unfunded social security program

11/4

2-1/2

11/2

1/41/41/61/5

1/2

01/21/20000

3/4

202

1/41/41/81/7

1/4

1- 1

0-1/4

-1/4—

- 1

11/4

0110

-1/4-1/4

-1/5

11/2

101

-1/4

-1/2-1/2-1/3

Period 4

Young

1

1/21

1/21/21/21

1/2

1-1/2

1/20000

Old

i

1/200

-1/2—

- 1

1/4

01/21/20

1/41/21

Total

1

11

000

3/4

1010

1/41/41/3



present value of planned consumption to age 100 cannotexceed that of lifetime wealth. Households that die beforeage 100 typically die with positive assets and leave abequest; the after-tax value of bequests is divided amonghouseholds aged 20 (the effective tax rate on bequestsis assumed to be 25 percent).

Household wealth at age 20 consists of human wealth,social security wealth, and bequests. Human wealth isthe discounted value of labor income, based on a cross-section profile of household labor income from the 1990FIES. The profile of labor income shifts up over time, onthe basis of assumed productivity growth of 1.7 percent ayear. Growth in the earnings of a specific household thenreflects both movements up the age-earnings profile aswell as the upward trend in the profile itself. Householdspay social security taxes on labor income and receivebenefits beginning at age 60, on the basis of an averagereplacement rate of 40 percent. The tax rate on workersis adjusted to finance benefits on a pay-as-you-go basis.Labor supply is exogenous; thus, labor income is notaffected by changes in the social security tax rate orbenefits.

The consumption path of each household is determinedby maximizing lifetime utility, where the latter is definedas the utility of "discretionary" consumption in eachperiod discounted by a rate of time preference (p); the rateof time preference was set to equal 0 in the simulations,following Tobin (1967). The utility derived from discre-tionary consumption is proportional to the number of"adult-equivalent" members of the household at a giventime; for this purpose, each child under the age of 20represents 1/2 of an adult. The utility function for eachperiod exhibits constant relative risk aversion; assuminga degree of risk aversion of unity implies that utility issimply the natural logarithm of consumption. Discretion-ary consumption at age t is measured as total consump-tion, Ct, less a minimum ("subsistence") level of con-sumption, C:

U(Ct) = log (Ct - C).

The subsistence level of consumption was set to equalabout 1/3 of the total consumption of working-age house-holds. The optimal growth rate of discretionary consump-tion at age t is then

(Ct - C)/(Ct-1 - C) = {(1 + r)/[(l + p)(l + a t ]} ,

where r is the real interest rate (assumed to remain con-stant at 3 percent a year), and at is the mortality hazardat age t. Household saving in each period equals thedifference between income and consumption, where in-come includes the return on accumulated assets.

The equilibrium household saving rate and assets-to-income ratio for an assumed population growth rate of1 percent a year were shown in the main text. The effecton the results of changing the population and productivitygrowth rates, and other parameters, was also examined.Compared with the baseline saving rate of 16.7 percent

and asset-to-income ratio of 5.5, when the populationgrowth rate is raised to 2 percent a year, the saving raterises to 17.2 percent, but the asset-income ratio falls to4.3. If productivity growth is raised to 2.7 percent a year,the saving rate drops to 12.2 percent, and the asset-income ratio falls to 3.0. When the real interest rate isincreased to 4 percent a year, the saving rate jumpsto 20.1 percent, and the asset-income ratio rises to 6.7compared with the baseline. Increasing the social securityreplacement rate by 10 percent, to 50 percent, makes thesaving rate fall to 13.9 percent, and the asset-incomeratio drops to 4.6.

The most notable aspect of these results is that thesaving rate declines in response to higher productivitygrowth. In contrast, some proponents of the life-cyclemodel have maintained that the saving rate should dependpositively on productivity growth because a faster growthrate would shift income to the working generation—whohave high saving rates—thus raising the aggregate savingrate. Such a result, however, depends on the assumptionthat working-age households are not forward looking.When they are forward looking, faster productivitygrowth raises lifetime resources relative to current in-come, lowering the saving rate. This offsets the shift inincome to working-age households: the net impact onsavings is ambiguous a priori. In the model consideredhere, the rise in consumption by the young more thanoffsets the shift in income shares, causing the aggregatesaving rate to fall.

44

ReferencesAndo, Albert, and Franco Modigliani, "The Life Cycle Hy-

pothesis of Savings," American Economic Review, Vol.53 (May 1963), pp. 55-84.

Auerbach, Alan J., Jinyong Cai, and Laurence J. Kotlikoff,"U.S. Demographics and Saving: Predictions of ThreeSaving Models," NBER Working Paper 3404 (Cam-bridge, Massachusetts: National Bureau of Economic Re-search, July 1990).

Barro, Robert J., "Are Government Bonds Net Wealth?" Jour-nal of Public Economics, Vol. 82 (November-December1974), pp. 1095-1117.

Bosworth, Barry, Gary Burtless, and John Sabelhaus, "TheDecline in Saving: Evidence from Household Surveys,"Brookings Papers on Economic Activity: 1 (1991), TheBrookings Institution (Washington), pp. 183-256.

Campbell, David W., "Wealth Accumulation of the Elderlyin Extended Families in Japan and the Distribution ofWealth Within Japanese Cohorts by Household Composi-tion: A Critical Analysis of the Literature," Jerome LevyEconomics Institute Working Paper 63 (New York: BardCollege, September 1991).

Feldstein, Martin, "Social Security, Induced Retirement, andAggregate Capital Accumulation," Journal of PublicEconomics, Vol. 82 (November-December 1974),pp. 905-26.

, "International Differences in Social Security and Sav-ing," Journal of Public Economics, Vol. 14 (October1980), pp. 225-44.


Graham, J.W., "International Differences in Saving Rates andthe Life Cycle Hypothesis," European Economic Review,Vol. 31 (1987), pp. 1509-29.

Hayashi, Fumio, Albert Ando, and R. Ferris, "Life Cycle andBequest Savings," Journal of the Japanese and Interna-tional Economies, Vol. 2 (December 1988), pp. 417-49.

Heller, Peter S., "Aging, Savings, and Pensions in the Group ofSeven Countries: 1980-2025," Journal of Public Policy,Vol. 9 (April-June 1989), pp. 127-53.

Horioka, Charles Y., "Why Is Japan's Private Saving RateSo High?" (unpublished; Washington: InternationalMonetary Fund, June 1986).

, "The Determinants of Japan's Saving Rate: The Im-pact of the Age Structure of the Population and OtherFactors," Economic Studies Quarterly, Vol. 42 (Septem-ber 1991), pp. 237-53.

Hurd, Michael D., "Research on the Elderly: Economic Status,Retirement, and Consumption and Saving," Journal ofEconomic Literature, Vol. 28 (June 1990), pp. 565-637.

Japan, Management and Coordination Agency, Annual Reporton the Family Income and Expenditure Survey, 1990(Tokyo: Office of the Prime Minister, Statistics Bu-reau, 1990).

Koskela, Erkki, and Matti Viren, "International Differencesin Saving Rates and the Life Cycle Hypothesis: A Com-ment," European Economic Review, Vol. 33 (September1989), pp. 1489-98.

Masson, Paul R., and Ralph W. Tryon, "MacroeconomicEffects of Population Aging in Industrial Countries,"Staff Papers (IMF), Vol. 37 (September 1990), pp.435-85.

Modigliani, Franco, "The Life Cycle Hypothesis of Savingand Intercountry Differences in the Saving Ratio," inInduction, Growth, and Trade, edited by W.A. Eltis,M.F. Scott, and J.N. Wolfe (Oxford: Clarendon, 1970),pp. 197-225.

Modigliani, Franco, and A. Sterling, "Determinants of PrivateSavings with Special Reference to the Role of SocialSecurity—Cross-Country Tests," in The Determinants ofNational Saving and Wealth, edited by Franco Modiglianiand Richard Hemming (New York: St. Martin's, 1983),pp. 24-55.

Organization for Economic Cooperation and Development,OECD Economic Surveys: Japan, 1989/90 (Paris, 1990).

Shibuya, Hiroshi, "Japan's Household Saving Rate: An Appli-cation of the Life Cycle Hypothesis," IMF Working Paper87/15 (Washington, March 1987).

Takayama, Noriyuki, The Greying of Japan: An EconomicPerspective on Public Pensions, Economic Research Se-ries, No. 30 (Tokyo: Institute of Economic Research,Hitotsubashi University, 1992).

Tobin, J., "Life Cycle Saving and Balanced Growth," in TenStudies in the Tradition of Irving Fisher, edited by W.Fellner (New York: Wiley, 1967), pp. 231-56.

White, B.B., "Empirical Tests of the Life Cycle Hypothesis,"American Economic Review, Vol. 68 (September 1978),pp. 546-60.

Yamada, Tetsuji, Tadashi Yamada, and Guoen Liu, "Determi-nants of Saving and Labor Force Participation of theElderly in Japan," NBER Working Paper 3292 (Cam-bridge, Massachussetts: National Bureau of EconomicResearch, March 1990).

45

References


V Alternative Long-Run ScenariosGuy Meredith

This section examines the long-run implications ofan aging population for the Japanese economy. De-

mographic issues are of particular importance to Japan,given that the proportion of the elderly in the populationwill rise more rapidly over the next twenty-five yearsthan in any other major industrial country.1 Many observ-ers have pointed to the surpluses in Japan's overall gov-ernment balance prior to the recent economic downturnas indicating a fundamentally healthy fiscal position. Be-cause the apparent health of Japan's fiscal position re-flects large surpluses in the social security accounts, how-ever, the implications of population aging for future socialsecurity receipts and payments are especially relevant.As shown below, the projected sharp future swing in thesocial security balance in the absence of measures tocontain benefits and raise revenues underscores the needto take a forward-looking view of fiscal sustainability.

To summarize the results of simulations undertakento explore various long-term scenarios, the increasingshare of the elderly in the population will exert pressurefor higher spending on both social security benefits andmedical care. Under the structure of the social securityprogram existing prior to the passage of reforms in late1994, the ratio of pension benefits to GDP would risefrom 5 percent of GDP in 1995 to 13 percent in 2020,and medical spending would rise by 2 percentage pointsof GDP.2 In the absence of a compensating increase inpension contribution rates, the social security surpluswould be eliminated shortly after the year 2000. There-after, rising benefits would lead to a primary social secu-rity deficit of 9 percent of GDP by 2020. Together withgrowing debt-servicing payments, unchanged policieswould imply an overall government deficit of 16 percentof GDP by 2020, and a rise in net government debt to150 percent of GDP.

To avoid such a debt explosion, actions must be takento reform the pension system, and to cut general govern-

1 See, for instance, Van den Noord and Herd (1993).2 Long-run scenarios are subject to many uncertainties and provisos

regarding, for instance, productivity growth and inflation. In Japan'scase, however, over four fifths of all public pension benefits areultimately indexed to wage growth, so different assumptions abouteconomic growth would have little effect on the relative magnitudeof future transfers to the elderly. In this case, the burden of an agingpopulation depends more on the projected share of the elderly in theoverall population, which is relatively robust to alternative demo-graphic assumptions over the horizon considered here.

ment spending or raise revenues (or both). The magnitudeof the adjustments needed to achieve a sustainable long-run fiscal position—defined as a stable ratio of debt toGDP—is estimated at 6 1/2 percent of GDP if implementedin 1995. In other words, combined measures to raiserevenues or reduce spending (or both) by this amountwould be needed to put the fiscal position on a sustainabletrack. The longer the required adjustments are postponed,the larger they will ultimately need to be, given theintervening buildup in government debt.

The pension reform plan introduced in late 1994 envis-ages a phased-in rise in contribution rates, reduction inbenefits, and postponement of the eligibility age. Thesesteps will go far in addressing the fiscal imbalance. Eventhe full implementation over the next thirty years of these(ambitious) changes to the pension system would not besufficient to put the fiscal position on a sustainable path,however. Further measures, totaling almost 2 percent ofGDP if implemented in 1995, would be needed to fullyoffset the burden of the aging population. Of course,these conclusions are subject to the many provisos anduncertainties surrounding long-run projections. Never-theless, the baseline projections provide strong warningsignals about the future fiscal situation.

3 These are based on the "middle series" published in 1993 by theMinistry of Health and Welfare. This projection assumes a recoveryin Japan's fertility rate over the longer run to 1.8 births per womanfrom the current level of 1.5. If, instead, the fertility rate were toremain near the current level (as assumed in the Ministry's "low"projection), the rise in the old-age dependency ratio would be evenlarger, although most of the impact would be felt beyond the year2020. Some analysts have expressed concern that even the "low"projection embodies too high a fertility rate, implying an even sharperrise in the old-age dependency ratio beyond about 2015. Beyond2025, the fiscal position would worsen further because of continuingincreases in the old-age dependency ratio. However, because thesevery long-term results are sensitive to demographic assumptions, thefocus here is on the horizon over which the projected populationstructure is relatively robust to alternative assumptions.

46

Population Dynamicsand Output Growth

The upper panel of Chart 5-1 shows the magnitude ofthe rise in the old-age dependency ratio implied by the1993 official population projections.3 The ratio is pro-


Alternative Fiscal Scenarios

Chart 5-1. Long-Term Demographic

Sources: Japanese authorities; and IMF staff estimates.

jected to more than double, from 20 percent to almost45 percent, during 1994-2020. Expressed as the ratio ofthe working population to the elderly, the number ofpeople of working age for each elderly citizen will bemore than halved, from 5 in 1995 to 2 1/4 by 2020. Asshown in the middle panel, growth in the working-agepopulation turns negative beyond 1995, with the pace ofdecline peaking at almost 1 1/2 percent a year in 2015.

The bottom panel of Chart 5-1 illustrates the implica-tions for Japan's potential output growth of slower growthin the working-age population. This projection is basedon the assumptions that total factor productivity growthwill continue at the rate expected for the 1990s; thatthe capital-to-output ratio will increase in line with thedeclining relative price of capital goods; and that theparticipation rate of the working-age population will be

roughly constant beyond the year 2000. Under theseassumptions, potential growth drops gradually from 2 1/2percent over the remainder of the 1990s, to about 1percent in 20.15, before recovering to about 1 1/2 percentin 2025.

Alternative Fiscal Scenarios

The most important direct impact of population agingon the fiscal position will occur through a rise in socialsecurity benefits for the Employees' Pension InsuranceScheme (EPIS) and the National Pension Scheme (NPS).Benefits will increase in relation to GDP, not only be-cause of the growing share of the elderly in the popula-tion, but also because of the maturation of Japan's pen-sion scheme. Assuming that the key parametersdetermining pension benefits remained unchanged at theirpre-reform levels—that is, eligibility at age 60 and fullindexation to wage and price growth4—the ratio of bene-fits to GDP would rise from 5 percent in 1994 to 6 3/4percent by 2000, and then to 13 percent by 2020. Anothersource of spending pressures will come from higher med-ical and health insurance payments, which are expectedto rise by 2 percent of GDP during 1995-2020.5

Under the assumption of pre-reform program parame-ters, then, total spending on social security would risefrom 12 percent of GDP in 1994 to 22 1/2 percent in 2020(Chart 5-2; upper panel). Social security contributions,meanwhile, would remain stable at 9 1/2 percent of GDP.6

The other components of government spending—currentspending (excluding debt-servicing payments), public in-vestment, and other revenues—are assumed to grow inline with nominal GDP. These developments would dra-matically alter the overall fiscal position, as shown inthe middle panel of Chart 5-2. The increase in socialsecurity benefits implies a swing in the primary socialsecurity balance of 10 percent of GDP from 1995 to2020; including debt-servicing costs, the overall socialsecurity balance would shift from the current surplus of3 1/2 percent of GDP to a deficit of 12 percent. Addingto this the deficit on government operations excludingsocial security implies a rise in the overall deficit to 16percent of GDP. The ratio of government debt to GDPgrows slowly over the remainder of the 1990s and thenaccelerates sharply beyond 2000, rising to 150 percentof GDP by 2020.

These developments are clearly unsustainable, and sig-nificant measures must be taken to prevent an explosiverise in debt over the long run. An indication of the size

4 Benefits under the EPIS are currently indexed to annual consumerprice inflation, with a rebasing every five years to reflect the gapbetween average wage growth and inflation. NPS benefits are indexedto consumer price inflation.

5 Based on updates to the estimates provided in Feldman (1985)using revised population projections.

6 Excluding net transfers from the central Government and inter-est receipts.

and Output Projections(In percent)

1 Shaded areas indicate projections.

47


V ALTERNATIVE LONG-RUN SCENARIOS

2Excluding interest receipts and central government transfers.

of the required actions can be obtained by comparingthe discounted present values of government revenuesand spending (in the absence of additional measures).7

The resulting gap is a measure of the "fundamental"

7 Specifically, the present value of primary expenditures plus thelevel of the initial debt stock is subtracted from the present value ofrevenues. Dividing the result by the present value of GDP indicatesthe actions that need to be taken—as a share of GDP, on a present-value basis—to stabilize the debt-to-GDP ratio over the long run andthus to achieve a sustainable fiscal position.

imbalance in the fiscal position, taking both current andfuture revenues and spending into account. Performingthis exercise for Japan using the assumptions describedabove gives an imbalance of 6 1/2 percent of GDP as of1995. In other words, if the required adjustments weremade in 1995, either a permanent increase in revenuesof 6 1/2 percent of GDP or a cut in expenditures of thesame size (relative to the baseline scenario) would beneeded to ensure that the debt stock does not rise withoutbound in the long run. If the adjustments are postponed,their ultimate size would need to be larger, since thedebt stock and thus the debt-servicing burden wouldaccumulate over the interval.8

To address the strains presented by population aging,the Ministry of Health and Welfare has introduced apackage of reforms to the EPIS to be implemented instages during 1995-2020.9 There are three key elementsto this package: (1) an increase in the combined employer/employee contribution rate to the EPIS from the current14 1/2 percent of earnings to 29 1/2 percent by 2020;10 (2)a phased rise in the eligibility age for EPIS benefits from60 to 65 starting in 2000; and (3) a shift to the use of"net" earnings—after payment of social security taxes—to index EPIS benefits. In addition, it is envisaged thatmonthly contributions to the NPS be roughly doubled inreal terms by 2015, implying a rise of about 50 percentin contributions relative to household income.

The combined effect of these measures would be toraise social security contributions to 14 1/2 percent of GDPby 2020 from the 9 1/2 percent share in the absence ofreform (Chart 5-3). As regards benefits, pension pay-ments would be reduced to 20 percent of GDP by 2020from the "unchanged parameters" level of 22 1/2 percent.Together, these reforms would be sufficient to balancethe EPIS and NPS components of the social securityprogram. They would not, however, offset the rise inmedical care costs, or the burden of rising governmentdebt. This is evident in the middle panel of Chart 5-3,which shows that, although the overall deficit wouldnarrow over the rest of this decade, it would start towiden beyond 2005. The subsequent rise in the deficitand the stock of debt in relation to GDP would be moregradual than in the absence of pension reform, but it isapparent that further measures would be needed toachieve a sustainable long-run fiscal position.

On the present-value basis described above, the re-maining imbalance amounts to 2 percent of GDP. The

8 For instance, if the required adjustments were postponed by fiveyears until 2000, they would rise to 7 1/4 percent of GDP; if postponeduntil 2020, they would reach almost 11 percent of GDP.

9 This long-term reform scenario has been constructed by the Minis-try of Health and Welfare in conjunction with the latest five-yearreview of the pension system.

10 The contribution rate would be raised in increments of roughly2 1/2 percentage points every five years, beginning in 1995-96.

48

1Shaded areas indicate projections.Sources: Japanese authorities; and IMF staff estimates.

(In percent of GDP)Pre-Reform Program Parameters1Chart 5-2. Long-Term Fiscal Projections:


Implications for the Saving-Investment Balance

Chart 5-3. Long-Term Fiscal Projections:

2Excluding interest receipts and central government transfers.

ment.11 An alternative approach on the revenue side couldinvolve eliminating the special deduction for pensionincome in the personal income tax system.12 Although itis difficult to assess the precise impact of such a reformin the absence of microeconomic tax data, a hypotheticalrise in the average tax rate on public pension benefits of30 percent would raise about 2 percent of GDP in addi-tional revenues by 2000.

The long-term prospects for Japan's saving and invest-ment depend on the behavior of public and private saving,as well as overall investment. Public saving will be deter-mined by the policy actions taken to address populationaging: as discussed above, pre-reform policies would leadto an explosive rise in the deficit and debt, whereaspension reform moderates the size and timing of the fiscaldeterioration. Private saving will be influenced by theshift in the demographic structure itself, asset accumula-tion, and expectations of future social security benefitsand tax liabilities. Finally, investment will depend on thedesired capital-to-output ratio, the trend growth rate ofoutput, and the depreciation rate of the capital stock.

Operationally, the projection for (gross) public savingis obtained as the sum of the overall government balanceplus the rate of (gross) public investment, based on thefiscal scenarios described above. Private saving is deter-mined residually as the difference between income andconsumption: the latter moves in line with wealth, whichis defined as the discounted value of labor income andsocial security benefits, less the discounted value of taxesand social security contributions, plus the stock of finan-cial and physical wealth. In addition, consumption in-creases as the overall dependency ratio rises, with a re-sponse parameter given by the relationship betweenJapanese demographics and household behavior.13 Thedesired capital-to-output ratio continues to grow in linewith the declining relative price of capital goods. Grossinvestment is then determined by the investment neededto keep the capital stock growing in line with potentialoutput and the desired capital-to-output rate, and by thedepreciation rate (which is assumed to remain at its histor-ical level).

required adjustment could be achieved in the form ofeither revenue or expenditure measures. Assuming thatrevenues are the source of adjustment, one solution wouldbe to alter the consumption tax rate over the longer runto compensate for the rising social costs of an agingpopulation. Beyond the rise in the consumption tax ratein 1997 that is incorporated in the baseline scenario, afurther increase to roughly 10 percent would be neededby the end of the decade to achieve the required adjust-

11 This is based on the authorities' estimate that each percentagepoint rise in the consumption tax generates about 0.4 percent of GDPin additional revenues. An option that would moderate the requiredrise in the tax rate would be to bring small businesses, which arecurrently tax-exempt, into the consumption tax base.

12 In 1990, the tax-exempt threshold of a childless retired couplewhose household head was aged 65 or over ranged from ¥3.1 trillionto ¥4.5 trillion, whereas for a working-age couple the threshold rangedfrom ¥1.7 trillion to ¥2.3 trillion (see Takayama (1992)).

49

BalanceImplications for the Saving-Investment

Pre-Reform Program Parameters VersusPension Reform1

(In percent of GDP)

1Shaded areas indicate projections.Sources: Japanese authorities; and IMF staff estimates.

13 See Section IV of this volume for a discussion of the methodology.


V ALTERNATIVE LONG-RUN SCENARIOS

(In percent of GDP)

The projections for the components of saving andinvestment along with the implied current account bal-ance are shown in Chart 5-4. Under the assumption of pre-reform social security parameters—and no other fiscalinitiatives to deal with population aging—the dramaticrise in public dissaving is reinforced by a downwardtrend in the private saving rate. The investment ratiodeclines moderately until 2015, reflecting slowing growthin potential output, before rising slightly near the end of

the projection horizon. The net result is a swing in theexternal balance from a surplus of slightly over 2 percentof GDP in 2000 to a deficit of 15 percent of GDP by2020. Such a huge deterioration in the external balancewould almost certainly not be "financeable" in reality;14

neither would the government deficits associated withan exploding debt stock find a ready market. Finally,predicting the response of private saving in the faceof unsustainable fiscal policies is problematic when theeventual resolution of the situation is not specified. Butthe very unsustainability of these outcomes serves tounderscore the imbalances that would potentially resultfrom shifts in the Japanese demographic structure.

As shown in the lower panel of Chart 5-4, full imple-mentation of the pension reform proposals sharply re-duces the decline in public saving, which falls from 8 1/2to 4 1/2 percent of GDP from 2000 to 2020. Nevertheless,combined with a 3 percentage point drop in the privatesaving rate over this period and roughly flat investment,the current account balance falls steadily to a deficit ofabout 5 percent of GDP by 2020. Although the declineis smaller than in the previous scenario, the underlyingfiscal position would remain unsustainable, and furtheradjustments would be required to prevent both govern-ment debt and external liabilities from rising withoutbound beyond this horizon. Scenarios that incorporatethe needed fiscal adjustments suggest that the currentaccount deficit would stabilize in a range of 1-2 percentof GDP beyond 2015.

References

Feldman, Robert A., "Japan: Outlook for Social Expenditure,1980-2025" (unpublished; Washington: InternationalMonetary Fund, July 1985).

Takayama, Noriyuki, The Greying of Japan: An EconomicPerspective on Public Pensions, Economic Research Se-ries, No. 30 (Tokyo: Institute of Economic Research,Hitotsubashi University, 1992).

Van den Noord, Paul, and Richard Herd, "Pension Liabilitiesin the Seven Major Economies," Economics DepartmentWorking Paper 142 (Paris: Organization for EconomicCooperation and Development, 1993).

14 Especially because several other large countries face problems

50

Parameters Versus Pension Reform1Investment Balances: Pre-Reform ProgramChart 5-4. Projected Long-Run Saving

1Shaded areas indicate projections.Agency; and IMF staff estimates.

Sources: Japan, Ministry of Finance and Economic Planning

associated with future population aging.


VI Movements in Asset Prices

I n the second half of the 1980s, Japan's stock pricestripled, and land prices doubled. The surge in asset

prices was followed by a collapse in stock prices startingin early 1990 and by a more gradual downturn in landprices from mid-1990 onward. This section analyzes thecauses of the asset price fluctuations and the effects ofthe fluctuations on economic growth.

The sharp movements in Japanese asset prices sincethe mid-1980s have led many observers to conclude thatthere was a speculative "bubble"—a continuous marketovervaluation followed by a collapse. The evidence dis-cussed below suggests that, in addition to this, "funda-mentals" also played a significant role. In part, the in-crease in the prices of real estate and equities reflectedthe growth in the economy: in the second half of the1980s, real GDP increased by 25 percent, and corporateprofits rose by 69 percent. Another part of the asset priceinflation is attributable to easy monetary policy, whichled to a decline in interest rates—ten-year bond yieldsfell from 6.3 percent in 1985 to 4.2 percent in 1987—and to a consequent ballooning of the present values offuture profits and rents. A third (but difficult to quantify)contributing factor was excessive risk taking associatedwith changes in the financial environment. Distortionsin Japan's land tax system also accelerated the rise inasset prices. As regards the decline in asset prices since1990, monetary tightening and measures designed todampen the real estate market played key roles.

The swings in asset prices had a significant directimpact on economic growth, through wealth effects onconsumption and through capital costs on investment.Estimates indicate that asset price increases boosted con-sumption by a cumulative 2-4 percent in the secondhalf of the 1980s, whereas the impact on business fixedinvestment may have been as large as 10 percent. Thedownturn in asset prices since 1990 (which seems tohave brought asset prices close to their trend levels) isestimated to have depressed real spending by roughlythe same amounts in 1990-93. Besides the direct effects,asset prices have had indirect effects on the economythrough money demand, bank profitability, and thefinancial system.

by a sharp decline in stock prices starting in early 1990and a more moderate downturn in land prices from mid-1990 on. Although both stock and land prices had exhib-ited broad swings twice earlier in the postwar era (in thelate 1950s and early 1970s), the strength of the recentasset price boom and the sharpness of the decline wereunprecedented.

The Nikkei 225 stock price index rose at an annualaverage rate of 31 percent between end-1985 and end-1989 (Chart 6-2). During most of that period, the price-earnings ratio, which had averaged 21 during the firsthalf of the decade, remained above 40. The first phaseof the stock price boom saw the Nikkei 225 double fromabout 13,000 in December 1985 to the 26,000 rangein October 1987. In the latter month, "Black Monday"reversed some of the increase, but, although the pace inmost other stock markets leveled off, the Japanese stockmarket recovered quickly, and stock prices started climb-ing with fresh momentum. An all-time high in dailytrading volume was recorded in July 1988, and the priceindex peaked on the last trading day of 1989, with theNikkei 225 standing at almost 39,000. At that time, themarket capitalization of the Tokyo Stock Exchange was$4.1 trillion (about 1.5 times Japan's GDP), comparedwith $0.9 trillion (equivalent to 60 percent of GDP) atthe end of 1985. The size of the Japanese equity markethad already surpassed that of the United States in 1987,and at the end of 1989 the Tokyo Stock Exchange ac-counted for 41 percent of total world equities.1

Land prices in Japan increased at an annual averagerate of 13 percent between end-1985 and end-1990 (Chart6-3). Compared with previous episodes of rapid landprice increases, this boom had special features. First, itwas unusually long, especially in comparison with theone in 1972-73. Second, in sharp contrast to the fairlyuniform rises in land prices during earlier booms, therewere large differences in land price increases with respectto location and land use. Although the price of land in

1 Note, however, that in Japan cross-ownership of equity shares ismore common than in other major countries. French and Poterba(1991) estimated that in the second half of the 1980s intercorporateequity holdings accounted for about half of the total market valueof Japanese stocks, compared with 2 percent in the United States.Excluding cross-holdings, the market capitalization of Japanese stockswas about two thirds of that in the United States, and about 25 percentof the world total, at the end of 1989.

Since the Mid-1980sJuha Kahkonen

Developments Since the Mid-1980sThe dramatic rise in stock and land prices in the second

half of the 1980s (Chart 6-1 and Table 6-1) was followed

51


VI MOVEMENTS IN ASSET PRICES SINCE THE MID-1 980S

Chart 6-1. Movements in Asset Prices Chart 6-2. Stock Price Developments

Publication (Tokyo, various issues). Source: Nikkei Telecom.

Tokyo doubled in 1986-87 alone, land prices outside themajor cities rose by only 38 percent over the entire five-year period. The surge in land prices started in the mixedcommercial-residential districts of central Tokyo and rap-idly spread to Tokyo's residential areas. In 1988, whenland prices in Tokyo started to stabilize, prices in Osakaand Nagoya, the other large metropolitan areas, acceler-ated. Land prices in the three metropolitan areas peakedat end-1990, when they were, on average, twice as highas in 1985. In all areas, commercial land increased invalue faster than residential land, and the appreciationof industrial land lagged well behind both.

The downturn in asset prices began in 1990. A collapseof the Nikkei 225 index by 3 percent on February 21,

1990 started a sharp declining trend in stock prices. Byyear's end, the index had fallen by almost 40 percentfrom the peak, and the price-earnings ratio had declinedto a level roughly in line with its trend level before theboom. The bottom in the stock price cycle was reachedin August 1992, with the Nikkei 225 dipping briefly toabout 14,000 (36 percent of the peak at the end of 1989).The Government's announcement in August 1992 of aneconomic stimulus package, however, helped the stockmarket to recover, and the Nikkei 225 subsequently stabi-lized in the 16,000-18,000 range. In late March 1993,expectations of another stimulus package led to a sharpincrease in stock prices. Following the announcement ofthe second package in April, the increase in stock prices

Sources: Nikkei Telecom; and National Land Agency, Land Price

52


Developments Since the Mid-1980s

Stock pricesNikkei 225

Land pricesNationwide

CommercialResidential

TokyoCommercialResidential









1985

100.0

100.0100.0100.0100.0100.0100.0

13.2

2.65.12.24.1

12.53.0

100.0

100.0100.0100.0100.0100.0100.0

1986

144.1

107.7113.4107.6123.8148.2121.5

44.1

7.713.47.6

23.848.221.5

138.0

103.1108.6103.0118.5141.9116.3

1987

174.8

131.1138.2134.5204.6238.8204.8

21.3

21.721.925.065.361.168.6

160.7

120.5127.1123.7188.2219.6188.4

1988 1989 1990

Index, 1985 = 100

229.0

141.9152.5145.1208.3245.9205.7

293.8

165.5177.9169.8223.3257.7219.2

182.9

184.2200.9188.0239.0268.3233.7

In percent change

31.0

8.310.37.91.83.00.4

28.3

16.616.717.07.24.86.6

- 3 7 . 7

11.312.910.77.04.16.6

Ratio to GDP, 1985 = 100

197.6

122.5131.5125.2179.7212.1177.4

237.6

133.9143.9137.3180.6208.4177.3

138.1

139.0151.6141.9180.4202.5176.4

1991

171.9

175.7192.9177.4218.9249.8212.4

- 6 . 1

- 4 . 6- 4 . 0- 5 . 6- 8 . 4- 6 . 9- 9 . 1

122.2

124.9137.1126.1155.6177.5151.0

1992

134.0

161.0170.9162.0186.3202.3181.4

- 2 2 . 0

- 8 . 4-11 .4

- 8 . 7-14 .9- 1 9 . 0-14 .6

92.4

110.9117.8111.6128.4139.4125.0

1993

138.6

152.0151.6154.4168.8165.3167.3

3.4

- 5 . 6-11 .3

- 4 . 7- 9 . 4

-18 .3- 7 . 8

94.6

103.7103.5105.3115.2112.8114.1

Sources: Nikkei Telecom; and Japan, National Land Agency, Land Price Publication (Tokyo, various issues).

continued, and the Nikkei 225 reached 21,000 points—a level first recorded in March 1987—in mid-May 1993.With the yen appreciating and the recession continuing,the stock price index declined during the remainder of1993, falling to about 16,000 in November 1993 amidsigns of a further deterioration in the economic outlook.In the first half of 1994, the Nikkei climbed to a rangeof 20,000-21,000, reflecting both the adoption of furthereconomic stimulus measures and signs of a bottomingout of the recession.

The rise in land prices came to a halt in mid-1990. Atthe same time, the number of land transactions, whichhad risen dramatically during the boom, fell signifi-

cantly.2 Land prices in large cities started to fall towardthe end of 1990—modestly at first, but later at an increas-ing pace. At the end of 1992, land prices in the threemetropolitan areas were 25 percent below their level atthe end of 1990, and further declines were recorded in1993 before prices started to stabilize in 1994. By con-trast, the price of land outside the large cities has been

2 If the tax revenue from real estate acquisition (adjusted for theincrease in the assessment value of the tax base) is taken as anindicator, land transactions fell by 11 percent in 1990 and by 6 percentin 1991.

Table 6-1. Asset Price Developments(At end of year)

53


VI MOVEMENTS IN ASSET PRICES SINCE THE MID-1980S

Chart 6-3. Land Price Developments

Sources: Nikkei Telecom; and National Land Agency, Land Price

relatively stable since 1990, and the decline in the priceof land for industrial use has been small.

Theoretically, the price of an asset (P) should equalthe discounted present value of the expected earnings(E) on that asset. The discount factor is the rate of returnon a risk-free asset (r) adjusted for the expected growthin earnings (g), taxes (t), and a risk premium (o):

P = El(r - g + t + a). (6-1)

Thus, an increase in the price-earnings ratio of an assetcould reflect one or more of the following factors: lowerinterest rates, improved prospects for earnings, lowertaxation of the asset, and a reduced risk premium. Al-though arbitrage usually keeps actual asset prices at ornear theoretical values determined by the fundamentals,there can be periods of speculative bubbles when marketparticipants rationalize their purchases with the view thatassets can be sold to someone else at an even higher price.

Much of the Japanese asset price inflation and deflationsince 1985 can be explained by movements in the funda-mentals (Chart 6-4). These movements reflect both policyactions and exogenous influences. First, interest rateshave fluctuated widely, suggesting that monetary policyinfluenced movements in asset prices. In the second halfof the 1980s, interest rates declined to historically lowlevels, reflecting a substantial easing of monetary policy.To counter the deflationary impact of the sharp apprecia-tion of the yen after the Plaza accord in September 1985,the official discount rate was halved from 5 percent to2.5 percent—the lowest rate ever—between January1986 and February 1987. Long-term interest rates alsofell, albeit not as sharply. Although the Japanese economystarted a strong recovery in 1987, the easy stance ofmonetary policy was maintained, as events followingBlack Monday in October 1987 (a worldwide collapsein the stock markets and a sharp depreciation of the U.S.dollar) led to a coordinated easing of monetary conditionsin all major industrial countries. In addition, the authori-ties felt external pressure to reduce the current accountsurplus by expanding domestic demand.

In the event, the official discount rate was not raiseduntil May 1989; by August 1990, however, the rate hadbeen increased several times to reach 6 percent. Long-term interest rates did not respond immediately, but byearly 1990 they were back to levels before the boom.Many observers suggest that market expectations of asharp rise in interest rates were the proximate causefor the February 21, 1990 collapse of the Tokyo StockExchange that marked the beginning of the sharp down-turn in stock prices. Prompted by signs of a recession inthe Japanese economy, the official discount rate waslowered several times between July 1991 and September1993, and at the end of 1994 it stood at 1.75 percent,with long-term rates at historically low levels.

Second, expectations of growth in earnings from assetsare likely to have changed since the mid-1980s. Afterbeing almost stagnant in the first half of the 1980s, corpo-rate profits increased by 69 percent in the second halfof the decade, and this buoyancy could have createdexpectations of permanently higher growth opportunitiesand may have contributed to higher stock prices. Al-though rents increased no faster than nominal GDP,expectations of large capital gains are likely to havefueled land price increases. The second half of the 1980salso witnessed deregulation and structural changes in theJapanese economy (the latter in part reflecting the needfor firms to adjust to a permanently stronger yen), which

Role of Fundamentals

and DeflationCauses of Asset Price Inflation

54

Publication (Tokyo, various issues).


Causes of Asset Price Inflation and Deflation

Chart 6-4. Asset Pricesand "Fundamentals'*

Sources: Japan, Economic Planning Agency (EPA), Annual Reporton National Accounts (Tokyo, 1993); and Bank of Japan, Short-TermSurvey of Business Enterprises and Economic Statistics Monthly (Tokyo,various issues).

tionary features are low assessment of property and inher-itance taxes (thus reducing the cost of holding land rela-tive to other assets and making land the preferred assetfor inheritance); high rates of capital gains tax (thusdiscouraging land sales); and lighter taxation of agricul-tural land (thus raising the price of nonagricultural land).4

Although distortionary land taxation can explain thehigh level of land prices in Japan, changes in the taxsystem are needed to explain land price movements. Sincethe mid-1980s, there indeed were such changes. One ofthem—an observed fall in the property tax assessmentratio—contributed to the surge in land prices.5 Most ofthem, however, represented attempts to curb land priceincreases. In 1987, the Government introduced a heavysurcharge on capital gains from short-term trading, witha view to reducing speculative transactions, and reducedthe long-term holding period (subject to a lower rate) topromote the supply of land. In 1990, the Tax AdvisoryCommission submitted to the Prime Minister a reportentitled "Basic Recommendations on the Ideal Frame-work of Land Taxation" (Japan (1990)), and some of therecommendations were included in the 1991 Land TaxBill.6 The new Bill, which came into effect in fiscal years1991 and 1992, was intended to improve efficiency andequity in land use by encouraging the transfer of land toresidential uses and by putting downward pressure onreal estate prices. A key change in the Bill was theintroduction of a land value tax (a tax on land holdings),initially at a rate of 0.2 percent and since 1993 at a rateof 0.3 percent. Other changes included phased increasesin the assessment ratios of property and inheritance taxes,a higher long-term capital gains tax, and removal of manyspecial treatments of agricultural land.

Finally, some researchers have claimed that the riskpremium on Japanese equities declined in the secondhalf of the 1980s. In particular, Ueda (1990) estimatedthe premium to have fallen to about 1/2 of 1 percent in1986-88 from an average of 5 percent during the preced-ing ten years. Possible reasons for the decline in the riskpremium include a change in the distribution of wealthtoward less risk-averse investors (such as large institu-tional investors) and an increase in the wealth of theaverage investor (which would tend to reduce the degreeof relative risk aversion).

may have created widespread optimism about the growthprospects of the economy.

Third, taxation had an impact on land prices. In Japan,land has traditionally been taxed more lightly than otherassets, and this has contributed to Japanese land pricesbeing among the highest in the world.3 The key distor-

3 See Ishi (1991) and Takagi (1989) for descriptions of land taxationin Japan, and Sachs and Boone (1988) and Ito (1992, pp. 408-14)for international comparisons.

4 An econometric study by Ando and others (1989) concluded thattaxing farmland at the same rate as land for housing would be ahighly effective means of raising the supply of land for residentialuses and lowering land prices.

5 Ishi (1991) estimated that, whereas in the early 1980s property taxassessment was on average two thirds of the official valuation (whichin turn is estimated to be only 70-80 percent of the market price),by 1988 the ratio had declined to about one half and by 1991 toslightly over one third.

6 For details of the land tax reform, see Organization for EconomicCooperation and Development (1991, pp. 153-54).



There is widespread agreement that, in addition to thestandard fundamentals discussed above, changes in thefinancial environment in the 1980s contributed signifi-cantly to the sharp rise in asset prices.7 These changeswere brought about by extensive financial liberalizationand innovation, and they were evident in increased lend-ing to real estate and in vigorous financial investmentactivities of households and nonfinancial corporations.

Japan had entered the 1980s with a tightly regulatedfinancial system, but during the decade extensive reformswere instituted: controls on capital movements were dis-mantled; interest rates on deposits were deregulated; andmarkets were introduced for a number of new instruments(such as commercial paper, futures, and options). Thederegulation went a long way toward enhancing competi-tion and improving the efficiency of the financial system,but it also gave rise to two major developments thatfueled asset price inflation in the late 1980s. First, bankand nonbank lending to the real estate sector increasedsharply because of changes in the behavior of both lendersand borrowers. As large manufacturing corporationsgained access to international capital markets and to anincreasingly developed domestic securities market, theirreliance on bank loans declined dramatically, makingsmall and medium-sized enterprises and households pri-mary customers of banks and releasing loanable fundsto nonmanufacturing sectors, especially real estate. Thisshift led to an apparent increase in the riskiness of banks'loan portfolios, but increased competition prevented lend-ing rates from rising sufficiently to compensate for thehigher risk. By end-1989, bank loans outstanding to real-estate-related activities had more than doubled in valuesince 1985 and accounted for almost one fourth of thebanks' total loan portfolio.8 During this period, nonbanks,which are more lightly regulated than banks, were evenmore aggressive than banks in financing real estate trans-actions. Second, the liberalized financial environment,together with the easy monetary conditions prevailing inthe second half of the 1980s, gave nonfinancial corpora-tions opportunities for financial arbitrage and improvedcredit availability for households. As a result, financialinvestment increased sharply and was to a significantextent channeled into the stock market, driving upequity prices.

Besides directly influencing asset prices, the financialmarkets also helped the increases in land and equityprices to reinforce each other. Higher land prices can beperceived by investors as an increase in the market valueof firms, and this raises stock prices. Higher land pricesalso increase the value of land as a collateral, enabling

landowners to obtain more bank loans. Some of the addi-tional loans can be used for further land and equity pur-chases, which will further raise asset prices. Althoughthis description does not establish the ultimate cause forthe asset.price increases, it is suggestive of a multiplierprocess that works through the financial markets onceone asset price starts to increase.

Throughout most of the bubble period, the Japaneseauthorities chose not to intervene in the financial markets.However, as part of a series of special measures to containland prices (reforms in land taxation constituted the mainpart), the Government in April 1990 placed quantitativerestrictions on bank lending to the real estate sector. Oncethe declining trend in land prices became apparent, theserestrictions were lifted at the end of 1991. In May 1991,the Government also amended the legislation regulatingthe lending industry, with a view to strengthening re-straints on nonbank lending for land-related purposes.

Was There a Bubble?

Practically all empirical studies attempting to explainJapanese asset price movements since the mid-1980shave concluded that there was a speculative bubble; thatis, fundamentals alone are unable to explain the sharpincreases in the second half of the 1980s and the subse-quent collapse.9 These studies typically have estimatedan equation such as equation (6-1) using data from theperiod before the bubble and have compared predictionsfor the postsample with actual developments. To shedlight on the relative importance of various fundamentalsand the possible existence of a bubble, the next fewparagraphs describe the results of two simple exercises.

First, an error-correction equation was estimated forstock prices using quarterly data from 1981 to 1985, withcorporate profits and long-term interest rates as dependentvariables (data on risk premiums and taxes on equityholdings were not readily available):

7 See Japan (1993) for a detailed discussion of the impact of thechanged financial environment on asset price inflation.

8 This estimate includes bank loans to nonbanks for real-estate-related purposes.

9 As regards stock prices, representative studies include Bank ofJapan (1993), French and Poterba (1991), Hardouvelis (1988), Hoshiand Kashyap (1990), Ogawa (1993), and Ueda (1990). For land prices,recent studies include Japan (1992), Japan (1993), and the JapanEconomic Research Center (1993).

Alog (PEQ/PROFT)= 0.22 - 0.03 A(RLB - PROFT%)

(3.4) (2.4)- 0.06 (RLB - PROFT%)-1

(3.3)+ 0.49 Alog (PEQ/PROFT) -1

(2.3)- 0.36 log (PEQ/PROFT)-,

Changes in the Financial Environment

56

(6-2)(3.1)

R2 = 0.49 DW = 1.88 h = 0.44,


Causes of Asset Price Inflation and Deflation

where

PEQ = Nikkei 225 stock price indexPROFT = trend of corporate profitsPROFT% = trend growth in corporate profitsRLB = ten-year government bond rateR2 = coefficient of determinationDW = Durbin-Watson statistich = Durbin's h statistic

and the numbers in parentheses are t-statistics. Note thatthis equation implies a long-run elasticity of stock priceswith respect to the interest rate of — 1/6. Thus, a 1 percent-age point drop in the long-run interest rate should leadto a permanent 17 percent increase in stock prices.

Equation (6-2), which explains stock price movementsin the period before the bubble reasonably well, was thenused to predict developments from 1986 on (Chart 6-5).Several conclusions emerge. First, if interest rates hadremained at their 1985 levels and expected profits hadgrown at the historical trend rate, stock prices wouldhave risen at an annual rate of 6 percent—far below theactual rate observed during the bubble period. Second,the simulations imply that monetary policy contributedmore to the collapse in stock prices than it did to thelong rise of those prices: although the contributionof lower interest rates to the boom in stock prices in1986-89 is estimated at 15 percent, higher interest ratesaccounted for as much as 45 percent of the sharp declinein equity prices in 1990.10 Furthermore, lower interestrates since mid-1991 helped to keep stock prices fromfalling even more rapidly than they actually did. Third,albeit difficult to measure, changed expectations aboutcorporate profit growth could potentially explain a majorpart of the boom and the collapse in stock prices. Theblack line in the middle panel of Chart 6-5 indicates howmuch the expected growth rate of corporate profits wouldhave had to differ from the historical trend in order tofully explain actual movements in stock prices. As canbe seen from the chart, the rise in stock prices in 1986—89 would have been consistent with an increase in theexpected profit growth rate from 6 percent to almost 10percent—not an unreasonably large change, given theunderlying growth rate of actual profits at the time andthe increase in the share of profits in GDP. Part of theincrease in corporate profits was cyclical, however, andshould not have affected a rational investor's valuationof equity. Hence, the possibility of a bubble cannot beexcluded.

A similar exercise was carried out for land prices.With potential GDP growth used as a proxy for the

10 Nevertheless, below-trend interest rates (easy monetary policy)had a significant direct impact on the rise in stock prices in1986-89: the equation suggests that in the peak year 1989 stock priceswere 20 percent higher than they would have been had interest ratesremained at levels recorded before the bubble.

Chart 6-5. Actual and PredictedStock Prices

expected increase in real rents, the estimated equationtook the following form (the data were from the period1970-85):

Alog (PLAND/YT) =- 0.40 - 0.0044(RLB - PEXP - QT%)

(2.2) (1.7)+ 0.61 Alog (PLAND/YT)-1

(5.7)- 0.051 log (PLAND/YT)-1

(2.2) (6-3)R2 = 0.41 DW = 1.90 h = -0.89,

Sources: Nikkei Telecom; and IMF staff estimates.

57



where

PLAND = nationwide land price indexYT = growth rate of nominal potential GDPRLB = ten-year government bond ratePEXP = expected inflation rateQT% = growth rate of potential GDP.

Equation (6-3) implies a long-run interest rate elasticityof land prices of —0.12, similar to that estimated forstock prices.

Dynamic simulations over the period 1986-92 suggestthe following conclusions (Chart 6-6). First, as in thecase of stock prices, easy monetary policy (low interestrates) explains about 15 percent of the increase in landprices (over and above the estimated trend increase of8 1/2 percent) during 1986-89. Tighter monetary policyin 1989-90 helped to bring land prices down, but thefall in land prices continued unabated in 1991-92 despitean easing of monetary conditions. Second, the simula-tions do not suggest a significant role for changed expec-tations about rent growth, leaving the bulk of the landprice increase in the second half of the 1980s unexplained(see Chart 6-6). Although this could simply reflect a lowcorrelation of potential GDP growth with the difficult-to-measure expectations about growth in earnings fromland, it could not signal the existence of a bubble.

Note that the two simple exercises undertaken aboveboth suggest that equity and land prices may have re-turned close to their trend levels in 1992: in Charts 6-5and 6-6, the lines indicating the underlying trend growthbased on experience before the bubble cross the linesrepresenting actual developments. There are, however,factors not included in the estimated models that may

Chart 6-6. Actual and PredictedLand Prices

suggest otherwise. In particular, the changes in land taxa-tion contained in the 1991 Land Tax Bill may have madethe decline in land prices deeper and longer-lasting thanequation (6-3) would imply. This could explain why landprices have continued to fall while stock prices havestarted to recover.

Even though the reasons for the sharp fluctuations ofasset prices may not be completely understood, there islittle doubt that these fluctuations contributed to the longboom in the Japanese economy that began in 1986 andto the recession that started in 1991. This part of thesection discusses the effects of asset prices on economicgrowth, mainly through private consumption and invest-ment, which contributed 2 1/2 percentage points and2% percentage points, respectively, to the average GDPgrowth of 5 percent in 1987-90.

The wide fluctuations in asset prices had a direct im-pact on consumption expenditure through the wealth ef-fect. Household net worth increased sharply in the secondhalf of the 1980s, from the equivalent of 5 1/2 times annualhousehold disposable income at end-1985 to almost 8 3/4times such income by end-1989 (Chart 6-7 and Table 6-2). Capital gains on land holdings (which account forabout half of total household assets) were responsiblefor 70 percent of the rise in net wealth over this period,with stock holdings (less than 10 percent of the total)accounting for another 13 percent. Following a turn-around in asset prices, the household net worth positiondeteriorated in 1990 as a ratio of disposable income and,in 1991, also in absolute terms. Although comprehensivedata are not available, net worth is estimated to havefallen further by some 5 percent in 1992, with the ratioof net worth to household income estimated to havedipped to below the 1987 level.

Compared with the sharp swings in asset prices, thedirect effect on consumption appears to be relativelysmall. Estimates of the positive effect during the upswingtypically range between 1/3 of 1 percentage point and 1percentage point a year; on the same basis, the declinein asset prices in 1991-92 is estimated to have resultedin an annual decline in consumption of 1/4 to 3/4 of 1percentage point.11 The low estimates are obtained byapplying typical estimates of the propensity to consumeout of wealth (ranging from 0.03 in MULTIMOD, theIMF's multicountry macroeconometric model, to 0.06 inJapan (1992)) to changes in real financial wealth (whichaccounts for slightly over one third of total householdwealth in Japan). The high estimates are based on move-

11 Real private consumption grew at an average annual rate of5 1/2 percent during 1987-90 and 4 1/4 percent during 1991-92.

Impact of Asset Price Movementson the Economy

Sources: Nikkei Telecom; and IMF staff estimates.

58


Table 6-2. Consolidated Balance Sheet of Household Sector

Total assetsInventoriesNet fixed assetsNonreproducible tangible assets

LandFinancial assets

CurrencyTransferable depositsOther depositsLong-term bondsCorporate sharesNet equity in life

insurance and pensionsOther financial assets

Total liabilitiesLoans by private sectorLoans by public sectorTrade credit

Net worth

Total assetsLandFinancial assetsCorporate shares

Total liabilitiesNet worth


Total liabilitiesNet worth

MemorandumReal net worth

In trillions of yen(1985 prices)

Percent changeDebt-to-asset ratioHousehold disposable income

In trillions of yenPercent change

Ratio to disposable income(in percent)

Gross interest paymentsConsumer debtHousing debtOther

Net interest income

1985

1,433.09.1

171.1693.5664.7559.3

20.333.7

296.450.965.9

80.611.5

195.8113.237.744.9

1,237.2

7.47.59.3

16.46.87.5

6.53.02.50.30.95.6

1,237.25.1

13.7

220.75.3

5.30.52.12.66.0

1986

1,671.78.6

173.2857.4827.1632.5

22.737.0

314.653.596.1

95.113.5

211.0128.040.043.0

1,460.7

16.724.413.145.8

7.818.1

7.33.62.70.40.96.3

1,454.917.612.6

230.34.3

5.30.62.12.65.9

1987

2,024.68.8

184.01,116.71,085.9

715.124.841.3

337.460.4

123.7

112.315.2

238.7150.543.344.9

1,785.9

21.131.313.128.713.122.3

8.64.63.00.51.07.5

1,775.222.011.8

236.72.8

5.30.62.12.74.8

1988 1989

In trillions of yen

2,235.78.8

194.01,216.61,183.2

816.327.444.8

362.359.7

173.4

133.215.5

265.5167.845.951.8

1,970.2

2,591.59.2

212.31,411.31,376.7

958.731.952.9

397.666.2

237.3

155.517.3

294.1194.550.349.4

2,297.4

In annual percent change

10.49.0

14.240.211.210.3

15.916.417.436.910.816.6

Ratio to disposable income

9.04.83.30.7I.I8.0

1,960.410.411.9

247.74.6

5.30.62.22.54.1

9.85.23.60.9I.I8.7

2,247.914.711.3

263.96.5

5.30.72.32.33.7

1990

2,718.19.6

227.41,531.51,495.0

949.632.253.6

438.769.1

162.7

173.619.7

325.1214.6

53.556.9

2,393.0

4.98.6

- 0 . 9- 3 1 . 4

10.54.2

9.75.33.40.61.28.5

2,283.41.6

12.0

280.06.1

6.50.92.92.64.6

1991

2,665.39.8

238.61,413.11,374.41,003.8

33.056.0

477.965.9

161.8

189.419.8

341.1222.7

57.660.8

2,342.2

- 1 . 9- 8 . 1

5.7- 0 . 6

4.9- 2 . 9

9.04.63.40.51.27.8

2,162.0- 5 . 2

12.8

296.15.8

6.9I.I3.12.75.3

1992

2,520.59.4

244.51,273.01,233.8

993.733.258.5

504.664.2

107.3

205.520.5

340.5224.6

61.154.8

2,180.0

- 5 . 4- 1 0 . 2

- 1 . 0- 3 3 . 7

0.2- 6 . 2

8.34.03.30.4I.I7.2

1,987.2- 8 . 1

13.5

304.82.9

6.21.02.92.34.2

Impact of Asset Price Movements on the Economy

Source: Japan, Economic Planning Agency (EPA), Annual Report on National Accounts (Tokyo, 1993).1IMF staff estimates based on land and stock price developments.

59



Chart 6-7. Household Savingand Net Wealths(In percent of disposable income)

Sources: Japan, EPA, Annual Report on National Accounts (Tokyo,1992); and IMF staff estimates.

•Shaded areas indicate recessions as defined by the EPA.

ments in total wealth, including land. Most studies haveargued (and found empirically) that land holdings havelittle or no effect on consumption in Japan: landowninghouseholds m a y not regard unrealized capital gains fullyas changes in their real wealth, and consumers withoutland will have to increase their saving to obtain financingfor costlier housing. A n exception is a study by Dekle(1990), which used time-series and panel data for Japa-nese prefectures and concluded that the rise in land pricesdid have a positive effect on consumption, since the

increased consumption of homeowners was larger thanthe decreased consumption of renters. Thus, the high-end estimates mentioned above cannot be dismissed asimplausible. Nevertheless, it is clear that changes in dis-posable income rather than swings in asset prices haveaccounted for the bulk of the growth in consumptionsince the mid-1980s.

The second half of the 1980s also saw a b o o m inbusiness fixed investment, to which asset prices contrib-uted by lowering capital costs and increasing the firms'net worth that could be used as collateral. Between 1986and 1990, business fixed investment rose by 55 percent.Rapidly rising equity prices enabled large companies toshift from bank loans to issuing n e w stocks and equity-linked bonds at very low interest rates; buyers acceptedthe low rates because of the anticipation that the bondscould be converted into equities at favorable prices. Surg-ing land prices in turn improved the capacity of smalland medium-sized firms to raise bank loans with land ascollateral.12 For the nonfinancial corporate sector as awhole, leverage (debt-to-equity ratio) fell sharply, andliquidity rose along with capital spending, contrary tohistorical experience in Japan and the experience of m a n yother industrial countries (Table 6-3).

With the decline in asset prices since 1990, there havebeen reverse effects on capital spending. Capital costshave risen. The collapse of stock prices has m a d e itdifficult for firms to raise n e w capital through equityissuance, and outstanding issues of equity-linked bondsare being redeemed. These developments have led largecompanies to return to bank loans and straight-bond is-sues, at a significantly higher cost than in the late 1980s.For small and medium-sized companies, the erosion ofcollateral land values has reduced borrowing opportuni-ties, and the liquidity position of the entire nonfinancialbusiness sector has declined to a level that prevailed inthe early 1980s. The debt-equity ratio, which at the endof 1989 was only one third of its 1985 level, almostdoubled in 1990-91 and is tentatively estimated to havereturned to the pre-bubble level in 1992. The businesssector's current financial difficulties and their impact oncapital spending should not, however, be exaggerated.Although the financial position of firms has declinedsharply over the past two to three years, this representsmore a return to a normal situation following overheatingthan a plunge to historical lows.

In contrast to consumption, there are few representa-tive estimates of the impact of asset price movementson business fixed investment, in part because of the diffi-culty of measuring capital costs and separating the effectof asset price changes from that of changes in monetary

12 O g a w a (1993) has discussed recent literature that implies a positiverelationship between the collateral value of the firm and the level ofinvestment. His empirical work found evidence supporting thishypothesis.

60


Impact of Asset Price Movements on the Economy

Table 6-3. Consolidated Balance Sheet of Nonfinancial incorporated Enterprises

Total assetsStocksNet fixed assetsNonreproducible tangible assets

LandFinancial assets

CurrencyTransferable depositsOther depositsShort-term bills and bondsLong-term bondsCorporate sharesNet equity in life

insurance and pensionsTrade creditOther financial assets

Total liabilitiesLiabilities, excluding corporate

sharesShort-term bills and bondsLong-term bondsCommercial paperLoans by private sectorLoans by public sectorTrade creditOther liabilities

Corporate shares

Net worth


Total liabilitiesLiabilities, excluding corporate sharesCorporate shares

Net worth

Memorandum

Operating surplusIn trillions of yenPercent change

Debt-to-equity ratioAssets/profits

LandOther assets

Liabilities/profitsNet worth/profits

1985

1,095.558.9

299.1261.7252.9475.8

2.340.189.20.4

11.788.3

——

204.039.9

758.5

562.51.6

56.2—

273.563.0

157.211.0

196.0

337.0

7.99.2

10.117.08.56.4

15.06.6

49.79.32.0

20.25.1

15.211.38.9

1986

1,238.855.9

312.6327.7318.6542.6

2.544.9

107.51.2

12.7141.1

——

194.138.5

891.2

589.61.7

62.2—

301.464.3

149.010.9

301.6

347.6

13.126.014.059.817.54.8

53.93.1

51.53.61.4

21.36.2

15.111.49.9

1987

1,470.056.2

330.9442.6433.0640.3

2.843.1

138.60.3

13.0181.3

0.1—

222.438.8

1,022.7

655.71.5

68.61.7

327.265.1

175.416.2

367.0

447.3

18.735.918.028.514.811.221.728.7

53.53.71.3

24.18.1

16.012.311.8

1988 1989

In trillions of yen

1,666.758.2

355.2497.3487.6756.0

3.146.4

159.20.2

12.1250.7

1.8—

239.742.8

1,234.6

714.80.9

74.39.3

357.267.2

185.620.3

519.8

432.1

1,912.363.4

394.7592.4582.1861.8

3.535.7

183.7—

11.6339.7

2.8—

231.653.1

1,497.2

770.80.7

85.813.1

395.271.1

179.725.2

726.4

415.1

In annual percent change

13.412.618.138.320.79.0

41.6- 3 . 4

59.010.41.0

24.08.3

15.712.111.9

14.719.414.035.521.37.8

39.7- 3 . 9

63.78.00.8

24.79.1

15.512.112.6

1990

1,983.367.3

437.0674.2663.7804.8

3.639.0

179.30.4

12.5236.0

5.1—

253.475.4

1,350.8

855.20.5

98.515.8

434.773.2

193.938.5

495.7

632.5

3.714.0

- 6 . 6- 3 0 . 5

- 9 . 810.9

- 3 1 . 852.4

68.67.61.3

25.59.7

15.812.513.0

1991

1,969.869.3

479.4617.1606.8803.9

3.748.1

162.3—

13.3230.9

4.2—

268.473.1

1,394.6

907.20.4

111.212.4

458.878.1

204.941.5

487.3

575.2

- 0 . 7- 8 . 6-0 .1- 2 . 2

3.26.1

- 1 . 7-9 .1

69.51.41.4

25.08.7

16.313.012.0

1992

1,830.669.9

503.7553.6543.2703.4

3.750.9

158.60.8

12.0150.7

3.9—

249.273.5

1,248.3

916.80.2

114.312.2

471.382.7

191.544.6

331.5

582.3

-7 .1- 1 0 . 5- 1 2 . 5- 3 4 . 7- 1 0 . 5

1.1- 3 2 . 0

1.2

62.5—2.1

24.17.8

16.313.211.0

Source: Japan, EPA, Annual Report on National Accounts (Tokyo, 1993).

61



conditions. MULTIMOD simulations suggest that a1,000 point rise in the Nikkei 225 index would result inan increase of V2 of 1 percent to 1 percent in businessfixed investment over two years. This implies that thesurge in equity prices in 1986—89 raised investment cu-mulatively by almost 10 percent over 1987-90, account-ing for one third of the acceleration in the growth ofinvestment. On the same basis, the decline in equityprices in 1990-92, which brought the Nikkei 225 tobelow its level of the end of 1985, is estimated to havedepressed investment by a cumulative 10 percent over1991-93.

Residential investment increased by almost 50 percentduring the early part of the boom before stabilizing in1989. Although the boom in housing construction wasdriven by many of the same factors that caused the assetprice inflation—low interest rates in particular—assetprices had an independent influence on residential invest-ment through two channels. Besides the wealth effect,which tended to raise the demand for housing construc-tion during the boom, higher land prices contained thedemand for housing by households that did not own landand stimulated the supply of land for construction byowners of surplus land, making the impact of higher landprices on residential investment ambiguous. As in thecase of business fixed investment, there are few empiricalestimates. There is, however, some evidence that higherland prices had a differential impact on various typesof residential investment: negative for owner-occupiedhousing, and positive for rental housing in Tokyo (Japan(1992, pp. 46-56)). All in all, the surge in asset pricesdoes not appear to have been the driving force behindthe boom in residential construction. Similarly, the col-lapse of land and equity prices is unlikely to have beenthe main contributing factor in the decline in residentialconstruction in 1991-92.

References

Ando, Itaru, and others, "Econometric Analysis of PropertyTax on Land," paper presented at the International Sym-posium on Structural Problems in the Japanese and WorldEconomy, Economic Planning Agency, Tokyo, Octo-ber 1989.

Bank of Japan, Functions of Stock Markets: Implications forCorporate Financial Activities, Special Paper 225 (To-kyo, February 1993).

Dekle, Robert, 'Alternative Estimates of Japanese Saving andComparisons with the U.S.: Can the Capital Gains toLand Be Included in Saving?" Institute for EconomicDevelopment Discussion Paper Series, No. 13 (Boston,Massachusetts: Boston University, December 1990).

French, Kenneth R., and James M. Poterba, "Were JapaneseStock Prices Too High?" Journal of Financial Econom-ics, Vol. 29 (October 1991), pp. 337-63.

Hardouvelis, Gikas A., Evidence on Stock Market SpeculativeBubbles: Japan, United States, and Great Britain, Fed-eral Reserve Bank of New York Research Paper 8810(New York, April 1988).

Hoshi, Takeo, and Anil K. Kashyap, "Evidence on q andInvestment for Japanese Firms," Journal of the Japaneseand International Economies, Vol. 4 (December 1990),pp. 371-400.

Ishi, Hiromitsu, "Land Tax Reform in Japan," HitotsubashiJournal of Economics, Vol. 32 (June 1991), pp. 1-20.

Ito, Takatoshi, The Japanese Economy (Cambridge, Massa-chusetts: MIT Press, 1992).

Japan, Economic Planning Agency (EPA), Economic Surveyof Japan (Tokyo, 1992 and various issues).

Japan, Economic Research Center, Scenario to RenewedGrowth: Japan's Economic Outlook, FY 1993-97 (To-kyo 1993).

Japan, Ministry of Finance, Institute of Fiscal and MonetaryPolicy, The Mechanism and Economic Effects of AssetPrice Fluctuations, Report of the Research Committee(Tokyo, April 1993).

Japan, Tax Advisory Commission, "Basic Recommendationson the Ideal Framework of Land Taxation" (Tokyo,1990).

Ogawa, Kazuo, "Asset Markets and Business Fluctuations inJapan," paper presented at the 10th International Sympo-sium of the EPA, Business Cycle and Financial PolicyCoordination, Tokyo, March 24-25, 1993.

Organization for Economic Cooperation and Development,OECD Economic Surveys: Japan, 1990/91 (Paris, 1991).

Sachs, Jeffrey, and Peter Boone, "Japanese Structural Adjust-ment and the Balance of Payments," Journal of the Japa-nese and International Economies, Vol. 2 (September1988), pp. 286-327.

Takagi, Keizo, "Rise of Land Prices in Japan: The Determina-tion Mechanism and the Effect of Taxation System,"Bank of Japan Monetary and Economic Studies, Vol. 7(August 1989), pp. 93-139.

Ueda, Kazuo, "Are Japanese Stock Prices Too High?" Journalof the Japanese and International Economies, Vol. 4(December 1990), pp. 351-70.

62


VII Asset Prices, Financial Liberalization,and Inflation in JapanAlexander W. Hoffmaister and Garry J. Schinasi

D uring 1986-93, asset prices in Japan moved througha dramatic and broadly symmetric cycle. Land val-

ues doubled, and corporate equity values tripled; theseincreases were followed by almost equal declines. At thesame time, consumer price inflation remained relativelylow—compared with inflation in other industrial coun-tries and with inflation in Japan in the late 1970s andearly 1980s. Because the asset price cycle in Japan hasbeen costly in terms of lost output and financial distress,it would be useful to identify the factors that caused thesedramatic asset price movements. An important unan-swered question is why the sharp increases in asset pricesand private indebtedness were allowed to persist; that is,why these developments were perceived as sustainable,rather than as manifestations of inflationary pressures.

One explanation that has received some attention isthat the asset price inflation could not be forecast exante on the basis of available macroeconomic or otherfundamental economic information. This characterizationhas been likened to a "bubble" phenomenon—a labelthat is adopted here for brevity—and has been interpretedto mean that the dramatic asset price increases werenot closely related to fundamental economic factors. Analternative explanation that has not been examined is thatthere was a "regime switch" in the mid-1980s—the resultof financial market reform and other structural changes—that made it difficult to accurately interpret economicand policy developments at the time. The structuralchanges in financial markets and in the tax treatment ofreal estate and equity investment in the 1980s, for exam-ple, might have been construed, ex ante, as justifying asustainable re-evaluation of asset values relative to abasket of consumption goods. From this perspective, theasset price cycle initially might have been perceived asbeing driven by economic fundamentals; later in the pro-cess, the structural changes made it difficult to properlyassess the stance of macroeconomic policies.

The evaluation of these alternative hypotheses is nota matter of merely academic or historical interest. If theasset price inflation was indeed a bubble, then there waslittle that could or should have been done to prevent oreliminate the sharp fluctuations in the asset price cycle,except to send a strong (monetary) signal early in theprocess that further asset market speculation would ulti-mately be very costly to speculators. By contrast, theregime-switch hypothesis leaves open at least two relatedpossibilities: macroeconomic policy inadvertently fueled

the asset price inflation; and the monetary policy frame-work in use at the time was inadequate for assessing thestance of policy and its influence on the real economy.If this second hypothesis is correct, then lessons may bedrawn from the recent asset price cycle for conductingmacroeconomic policy in the 1990s.

This section examines the relationship between macro-economic variables and asset price inflation in the 1980s.The focus is on land price inflation, rather than on stockprice movements; although stock prices are notoriouslydifficult to model empirically, real estate prices generallymove in response to changes in fundamental economicfactors, including the business cycle and monetary fac-tors. This section also examines several related questions.(1) Was there a structural break in the way monetaryfactors, in particular, affected asset prices in the 1980s?(2) Did monetary factors contribute in important waysto asset price inflation in the 1980s? (3) What accountedfor the divergent behavior of asset prices and consumerprices? (4) Is there evidence supporting the view that theeffects of monetary factors were heavily concentrated inasset markets rather than in goods markets?

The next part of the section briefly reviews the relevantasset price and balance sheet developments and thenexamines in more detail the alternative hypotheses andimportant unresolved issues regarding the asset pricecycle in Japan. This is followed by an attempt to quantifythe extent to which macroeconomic factors, and in partic-ular monetary factors, influenced the land price cycle inthe late 1980s and to provide evidence about the regimeswitch and the other issues. Finally, the empirical resultsare summarized. An appendix briefly examines the time-series properties of the data and other statistical issues.

Key Developments and IssuesBecause the important facts characterizing the asset

price cycle in Japan have been examined in detail inprevious studies, only the salient features of these adjust-ments are highlighted here.1

1Asset price and balance sheet adjustments occurred in other indus-trial countries—including the United States, the United Kingdom,the Nordic countries, Australia, and New Zealand, and have beenexamined in detail in previous issues of the IMF's World EconomicOutlook; see also Schinasi and Hargraves (1993), which synthesizesthis work. Adjustments in Japan have been described in fuller detailin Hoffmaister and Schinasi (1994).

63


VII ASSET PRICES, FINANCIAL LIBERALIZATION, AND INFLATION

Chart 7-1. Urban Land and Stock Prices rices

Sources: Japan Real Estate Institute, Bulletin of Japan Land Prices(Tokyo, various issues); and Nikkei News Service.

Asset Price and Balance Sheet Adjustments

As earlier mentioned, between 1985 and 1990 landprices in Japan doubled, and equity prices tripled. Theaverage price of residential land in large urban centersrose 13 percent annually between 1985 and 1990, andequity prices rose 31 percent annually over the sameperiod (Chart 7-1).

The turnaround in asset prices coincided with a changein the stance of monetary policy in early 1990. From itspeak in 1990, the average price of land in the six largestJapanese cities declined by 36 percent through the endof 1993 (an average annual decline of about 14 percent).Likewise, equity prices (as measured by the Nikkei 225)plunged from a high of nearly 39,000 in February 1990to a low of about 14,000 in August 1992; the stock marketthen rebounded and traded in a price range between18,000 and 21,500 in the first half of 1994.

The sharp asset price adjustments were accompaniedby equally dramatic changes in private sector balancesheets. Total private sector indebtedness expanded quiterapidly in the mid- to late 1980s, both in absolute termsand relative to GDP (Chart 7-2). Liabilities of the house-hold and business sectors expanded markedly (Chart 7-3and Table 7-1, respectively), and in turn this growthwas reflected in the rapid expansion in the banking andnonbank financial sectors.

The balance sheet expansion in the late 1980s lefthouseholds, businesses, and financial institutions unusu-ally vulnerable to the effects of a tightening of monetaryconditions in early 1990. Once asset prices began to

Chart 7-2. Total Private NonfinancialSector Debt1(In percent of GDP, end of period)

Source: Japan, Economic Planning Agency (EPA), NationalIncome Accounts (Tokyo, various issues).

1Total financial liabilities of the private nonfinancial sectors lesstrade credits.

Chart 7-3. Household Sector Balance Sheet(In percent of disposable income)

Source: Japan, EPA, National Income Accounts (Tokyo, variousissues).

64


Key Developments and Issues

Table 7-1. Selected Indicators of Financial Balance1

Household assetsHousehold liabilitiesHousehold net worth

Business assetsBusiness liabilitiesBusiness net worth

Business net interest paymentsBusiness debt-equity ratio

1970-74

5.230.624.61

3.481.911.57

49.322.52

1975-79

4.960.674.29

3.521.901.61

63.062.70

1980-84

6.150.825.32

3.761.941.82

53.632.35

1985

6.490.895.61

3.851.981.88

46.672.01

1986

7.260.926.34

4.192.002.20

44.351.42

1987

8.551.017.54

4.752.122.63

40.811.26

1988

9.031.077.95

5.042.162.88

38.720.98

1989

9.821.128.71

5.382.173.21

40.940.78

1990

9.701.168.54

5.072.232.85

49.141.26

Source: Japan, Economic Planning Agency (EPA), National Income Accounts (Tokyo, various issues).1Household data as a percent of disposable income; the business sector is defined as the nonfinancial corporate sector as a percent of GDP,

except for the debt-equity ratio and business interest payments, which are in percent of available income.

decline, the asset sides of private sector balance sheetsdeteriorated sharply, leaving many private sector agentswith highly leveraged positions. Moreover, the very deepand prolonged recession and the associated decline indisposable incomes and profits made it difficult for house-holds and businesses to meet existing obligations and atthe same time maintain their spending levels. As a result,both consumer expenditures and investment outlays wereconstrained, as an increasing share of income was de-voted to servicing existing debt levels and to reducingdebt to more normal levels.

Reflecting these adjustment efforts, the private sector'sratio of debt to GDP leveled off and then declined slightlythrough the end of 1992 (the last year for which thereare data). Experience in other countries suggests thatfurther downward adjustments are likely to occur in theperiod ahead—in both the household and business sec-tors. Households have reduced debt levels somewhat,and their net wealth has clearly been affected by the assetprice deflation. Similarly, businesses have been moreconscious of their leverage ratios and have pared backtheir debt ratios. The most difficult adjustments in theperiod ahead are likely to be in the financial sector.

Alternative Explanations and the Roleof Monetary Factors

An important unresolved question about the asset pricecycle in Japan is why the sharp increases in asset pricesand private indebtedness were taken for something otherthan a manifestation of inflationary pressures created byoverexpansionary macroeconomic policies.2 One possibleexplanation, which has been supported by empirical stud-ies, is that there was a large, unexplained bubble compo-

nent to the sharp and prolonged rises in both land and eq-uity prices in Japan. For example, simulations below,based on a single-equation model of asset prices and usingdata during the period 1970:Ql-1984:Q4 show that theinformation available through 1985—the beginning of theasset price cycle—was not sufficient to accurately forecastthe sharp rise in land prices. This conclusion holds evenwhen the actual future values of the explanatory variables(other than the lagged dependent variable) over the fore-casting period were used in the simulations. Although thisexercise cannot statistically verify or falsify the existenceof a bubble component in asset prices, it does forcefullyillustrate how difficult it would have been in the mid-1980sto pinpoint the forces driving asset prices in the early stagesof the inflation cycle.

An unexplored alternative explanation is that the assetprice movements were driven by economic fundamentals,but that the confluence of structural changes in financialmarkets and expansionary macroeconomic policies madeit difficult to interpret accurately the sharp and persistentincreases in asset prices. In Japan, there were many sig-nificant developments in the early 1980s—such as thestructural changes in financial markets and the tax treat-ment of real estate and equity investment—that mighthave altered in fundamental ways the relationships be-tween macroeconomic variables and asset prices.3 These

2 For an analysis of this question, see Schinasi (1994).

3 Extensive reform measures since 1984 have included liberalizationof interest rates on deposits; the easing of restrictions on large timedeposits, certificates of deposit (CDs), and money market certificates;and the introduction of markets for commercial paper, futures andoptions, and offshore transactions. Recent legislative reforms havefurther lowered barriers between banking and securities brokerage,permitting banks to establish subsidiaries that provide brokerage ser-vices and allowing securities firms to establish banking subsidiaries.In addition, tax provisions created incentives for the construction ofapartment houses and condominiums, and changes in the capital gainstreatment of real estate transactions encouraged upgrade purchases.

65



structural changes, which could have been perceived asraising the demand for assets, may have provided reason-able rationalization for believing that assets—in particu-lar, real estate and corporate equities—were undervaluedrelative to consumption goods. In the initial stages of theboom in asset values, therefore, it was not unreasonableto continue to maintain the stance of monetary policyand, in effect, acquiesce to the rapid run-up in the relativeprice of real estate values and other asset prices.

In hindsight, a working hypothesis for understandingthe dramatic asset price developments is that the prevail-ing monetary policy framework, together with dramaticchanges in financial structures, monetary transmissionmechanisms, and other important structural changes,made it difficult to distinguish sustainable adjustmentsin asset prices from unsustainable price increases (seeSchinasi (1994)). In the initial stages of the increase inasset prices in Japan, it was presumed that the asset priceincreases were adjustments in relative prices in responseto fundamental structural changes—including asset-market-specific tax reforms. As a result, there was aprolonged period during which inflationary pressures ac-cumulated. Moreover, the combination of real, financial,and institutional structural changes in Japan created anenvironment in which inflationary pressures were chan-neled to, and concentrated and recycled in, asset marketsfor a prolonged period.4

Casual Empirical Evidenceand Unresolved Issues

There is strong support for the concentration hypothe-sis. First, Japan entered the 1980s with a highly regulatedfinancial system, which was then liberalized rapidly: ceil-ings on deposit interest rates were liberalized in the early1980s; Japanese nonbank financial institutions were al-lowed to compete with established banks; foreign finan-cial institutions were allowed to enter the market; andnew financial instruments were allowed. In this new com-petitive environment there were incentives to ventureinto new markets—in particular, real estate markets—as

4 The concentration hypothesis is that in many industrial countries,including Japan, the confluence of macroeconomic policies, financialliberalization, and other structural changes created an environmentin which excess liquidity and credit were channeled to specific groupsactive in asset markets, including large financial and nonfinancialinstitutions, high-income earners, and wealthy individuals. Thesegroups responded to the economic incentives associated with thestructural changes and borrowed heavily to accumulate assets in globalmarkets—such as real estate, corporate equities, art, and preciouscommodities. Apparently, the excess credit was recycled in assetmarkets several times over. The structural changes that occurred inJapan are detailed in the preceding footnote. For a brief but moregeneral discussion of the reasons that inflationary pressures may havebeen concentrated in asset markets, see Schinasi and Hargraves (1993,pp. 18-20).

traditional borrowers shifted into other forms of financialintermediation. All of these changes led to a financialenvironment in which bank balance sheets were adjustingrapidly and in which both bank credit and deposits wereexpanding rapidly.5 During this kind of rapid change—which was occurring in many other industrial countriesas well—it was difficult to judge accurately the stanceof monetary policy.6

Second, growth in both the monetary aggregate,M2 + CDs, and the credit aggregate, total private credit,remained very high throughout most of the 1980s. More-over, growth in both of these financial aggregates ex-ceeded growth in real GDP by a fairly wide margin,representing potential inflationary pressure. The differ-ence between this "money gap" and inflation averaged3 1/4 percent of GDP a year, and the difference betweenthe "credit gap" and inflation averaged 3% percent ofGDP a year in the 1980s (Chart 7-4). To obtain estimatesof the potential "overhang" of the money stock or thestock of credit relative to the national accounts measuresof economic activity, the year-by-year gaps should beaccumulated, and this suggests a considerable overhang.

Third, and in strong support of the concentration hy-pothesis, the dramatic asset price adjustments in the1980s did not initially pass through to goods markets inJapan, as they had in the 1970s (Chart 7-5). In most othercountries that experienced asset price inflation, increasesin the consumer price index (and the GDP deflator) didreflect the underlying inflationary pressures that werepresent, albeit in some cases with a long delay; by con-trast, in Japan inflation (measured by the GDP deflator)did not rise significantly in the 1980s and did not fullyreflect the demand pressures that were present in the realestate and corporate equity markets. This represented adeparture from patterns that prevailed in the 1970s, wheninflationary pressures tended to raise asset prices initiallyand were then transmitted relatively rapidly to goodsprices. In the 1980s, by contrast, conventional measuresof inflation, such as those based on consumer price in-dices or GDP deflators, remained relatively low duringmost of the period in which asset prices were rising atdouble-digit rates. That goods prices did not rise com-mensurately with asset prices provided some confirma-tion for the initial judgment that the asset price increaseswere sustainable.

When viewed together, these factors—structuralchange in the financial sector, money and credit growthin excess of GDP, and a breakdown in the relationshipbetween asset and goods prices—suggest that monetaryfactors played a key role in the asset price cycle in thelate 1980s, that there was a shift in the transmission of

5 For further details see Hargraves, Schinasi, and Weisbrod (1993).6 For example, see Bank for International Settlements (1984 and

1986).

66


Chart 7-4. Money, Debt, and Inflation(In percent)

Chart 7-5. Asset Prices and GDP Deflator(In percent a year)

Sources: For housing price, Japan Real Estate Institute, Bulletinof Japan Land Prices (Tokyo, various issues). For stock price,WEFA, Inc. data base.

1Urban residential land prices in six largest cities.2The series are smoothed using a four-quarter centered moving

average.

Source: IMF staff calculations.1Total financial liabilities of the private nonfinancial sectors less

trade credits.

monetary policy to inflation between the 1970s and the1980s, and that, as a result, inflationary pressures mayhave been concentrated in asset markets.

A Multiequation Model of Land Pricesand Inflation

Although previous studies have tried to examine thecauses of these sharp adjustments in Japan, most of theissues discussed above, with few exceptions, have notbeen examined empirically or tested.7 This part of the

section attempts to provide some empirical evidence onthese important issues.

A Vector-Autoregression Model

To examine these issues empirically, this subsectionprovides estimates of a standard vector autoregression(VAR) (see Sims (1980)). Previous studies have used asingle-equation approach, which requires strong assump-tions about the dynamic relationships among variablesand which ignores the fact that the explanatory variablesare not exogenous. VAR allows for the exploration ofempirical regularities and relationships with a minimumof assumptions; it also allows for the determination ofthe dynamic structure of relationships. A drawback of

7 One exception is Samiei and Schinasi (1994), which demonstratesthat monetary factors have been an important determinant of landprice inflation in Japan during the period 1970-92. In addition, mone-tary factors were more important in the 1980s than in the 1970s—suggesting a shift in the monetary transmission process.

67

A Multiequation Model of Land Prices and Inflation



making fewer assumptions is a loss of specificity aboutthe underlying structural relationships among the vari-ables in the system.8

Land price inflation in Japan is assumed to be deter-mined by several factors within the context of a moregeneral macroeconomic system. These include: monetaryand financial conditions (that is, interest rates and othermeasures of monetary stance); the general condition ofthe Japanese economy (that is, its position in the businesscycle); and inflation. Because the complexity of a VARsystem increases dramatically as the number of variablesin the system increases, the equation system focuses onfive variables, which are allowed to be jointly determined:land price inflation, consumer price inflation, the outputgap, a policy-determined interest rate, and the growthrate of a policy-related financial aggregate (either a mone-tary or credit aggregate).9 The variables are defined asfollows:

r = call money ratem = growth in M2 + CDs, or growth in total

private credity = real GDP output gaprL = land price inflationr = consumer price inflation.

With X't = (rt, mt, yt, rLf, rt)—where a prime (')indicates the transpose operator—the basic model to beestimated is:

4

Xt = Ei = 1 AI Xt-1 + ut,

where the Ai are (5 X 5) matrices of parameters to beestimated. That is, each of the five variables in the equa-tion system is assumed to be determined by four laggedvalues of each variable in the system, including its ownlagged values.10 The errors are assumed to be identicallyand independently distributed, with zero means and con-stant variances and covariances.

The estimated VAR does not provide estimates ofstructural parameters, but it does provide unconstrainedestimates of the relationships among the variables inthe system. The information provided by VAR can berepresented in three useful and related forms: estimatesof the autoregressive parameters (and the associated teststatistics), which relate each variable to the other system

8 A disadvantage of the standard VAR approach is that the underlyingbehavioral relationships that may exist cannot be identified; the esti-mated parameters are reduced-form estimates, and care must be exer-cised in interpreting the implications of the estimated equations.

9 Other variables that were included are a measure of fiscal policyand alternative measures of the monetary stance, such as the exchangerate, but they did not alter the results in important (or even nota-ble) ways.

10 The parameters are estimated using ordinary least squares, whichis in this case the full-information maximum-likelihood estimator.See Appendix 7-1 for the determination of the lag length.

variables; variance decompositions, which provide esti-mates of the share of the variance of each variable thatis explained by the other variables in the system; andimpulse response functions, which indicate how the vari-ables respond to various orthogonal shocks.11 In examin-ing land price inflation, for example, the estimated auto-regressive parameters on monetary factors can be jointlytested for their statistical significance; the variance de-composition can be examined to determine the impor-tance of monetary innovations (that is, monetary policychanges) in determining the variation in land prices; andthe impulse response function can be used to examine thereaction of land price inflation to a monetary innovation.

To construct the variance decompositions and the im-pulse response functions, the variables in the system mustbe ordered according to their contemporaneous exogen-eity (see Sims (1980)). The following ordering is adoptedinitially: the call money rate; either the monetary or creditaggregate; business cycle conditions; property prices;and, finally, general inflationary pressures. Later, tocheck for robustness, the order is changed to see if theempirical results are sensitive to the original ordering ofthe variables.

The initial ordering of variables is not arbitrary, how-ever. The call money rate is assumed to be the mostcontemporaneously exogenous variable in the system be-cause the Bank of Japan uses the call money rate as itsoperating instrument in the short run. This does not meanthat the call money rate is assumed to be unaffected bythe other variables in the system, such as business cycleconditions; but innovations in the call money rate in anygiven quarter affect all of the other variables in the sys-tem, whereas innovations in other variables do not affectthe call money rate in the same quarter. Innovations inthe call money rate are assumed to affect directly thebanks' cost of funds, which in turn affects monetaryconditions by altering the behavior of financial institu-tions and their balance sheets. Monetary conditions arerepresented in the model either by growth on the liabilityside of the banking system (growth in M2 + CDs) orthe asset side of the banking system (growth in privatecredit). Innovations in both the call money rate and thefinancial aggregate are then presumed to affect the gen-eral condition of the economy, as summarized in theoutput gap; and innovations in monetary conditions andthe output gap are presumed to affect land price inflation,and so on.

Empirical Results

Using the estimated VAR parameters, variance decom-positions, and impulse response functions, several issuesare examined. (1) Was there a regime switch in the 1980s

11 In a standard VAR, the orthogonal innovations are obtained fromthe Choleski decomposition of the covariance matrix.

68



compared with the 1970s? (2) What role did monetaryfactors play in the land price inflation in the 1980s?(3) Did the process of consumer price inflation alsochange? (4) Were the effects of monetary factors moreconcentrated in asset prices, and less concentrated inconsumer prices, in the 1980s compared with the 1970s?

Was There a Regime Switch in the 1980s?

Because of the financial changes that occurred in Japanin the 1980s, it is unlikely that the structure of an esti-mated model would remain stable during the period1970-93—this is probably what accounts for the inabilityof models estimated through 1985 to predict asset pricesaccurately in the mid- to late 1980s and early 1990s.Unfortunately, it is difficult to construct variables thatwould allow for precise estimates of the effects of finan-cial liberalization, for example, or changes in the taxtreatment of real estate investment.

The alternative strategy followed here is to split thesample into two periods, with the intention of capturingthe shift in underlying regimes: one period in which thefinancial system in Japan was tightly regulated, and asecond period in which the financial system had moreor less been liberalized. Although it is not possible todetermine with precision when the Japanese economyentered the liberalized regime, the shift is treated as hav-ing occurred in the period 1982-84, for two reasons.First, although liberalization began well before 1982-84,the full effects of financial liberalization would take sometime to take hold. Second, the global economy experi-enced a prolonged recession in the early 1980s. To pre-vent these transitional factors from affecting the statisticaltests, several breakpoints, beginning in 1983 and endingin 1985, were examined.

By comparing the residuals of the model estimatedover the entire sample with those estimated over thevarious subsamples defined by the breakpoints (that is,by performing likelihood ratio tests) it is possible to testthe hypothesis that the model parameters were constantover the entire sample period. The various likelihoodratio tests reject the hypothesis of parameter constancyfor all of the alternative sample periods in the 1980s thatwere tried (see Appendix 7-1). For example, when theliberalized regime was assumed to begin in the first quar-ter of 1984, the F-test of a structural break is significantat the 9 percent level when M2 + CDs is used as thefinancial aggregate and at the 1 percent level when privatecredit is used as the financial aggregate (see Table 7-1).As discussed in the next subsection, this evidence of astructural change in the VAR system is supported byempirical evidence provided in the estimated equationfor land price inflation and is consistent with the resultsreported in Samiei and Schinasi (1994). The remainderof this subsection documents the changes in the esti-mated land price and consumer price inflation equations,which were estimated over two fixed sample periods:

Chart 7-6. Predictions of Land PriceInflation Using Private Creditas the Financial Aggregate(In percent a quarter)

Source: IMF staff calculations.1Urban residential land prices in six largest cities.

the "regulated" regime during 1970:Ql-1983:Q4 and the"liberalized" regime during 1984:Q1-1993:Q4.

To illustrate the nature of the regime switch, Charts7-6 and 7-7 present a comparison of model predictions(static simulations) of land price inflation and consumerprice inflation, respectively, using the estimated landprice and consumer price equations. The top panel ofChart 7-6 presents the within-sample predictions of theland price inflation equation when its parameters areestimated in the first period, 1970:Ql-1983:Q4. Despitethe wide variation in land price inflation in the regulatedregime, the within-sample predictions indicate that themodel describes the movement in land price inflationreasonably well in the 1970s.12 Using the parameters

12 The next subsection presents dynamic simulations of the land priceequation using projected values of the lagged dependent variable.

69



Chart 7-7. Predictions of Consumer PriceInflation Using Private Creditas the Financial Aggregate(In percent a quarter)

estimated in the regulated regime, the model's out-of-sample predictions of land price inflation were generatedfor the liberalized regime, using the actual observationson the explanatory variables, 1984:Q1-1993:Q4, in-cluding the lagged dependent variable (see the middlepanel). The model estimated in the regulated regime failsto predict land price inflation accurately in the liberalizedregime. By contrast, as shown in the bottom panel, thewithin-sample predictions of the land price inflation equa-tion estimated in the liberalized regime tracks actual landprice inflation reasonably well. Note that casual empiri-cism (see Chart 7-6) suggests that there was a fairlywide variation in land price inflation in both regimes,suggesting that the structural break is not attributabledirectly to a shift in the behavior of the land price data.

A similar structural break is apparent in the consumerprice inflation time series (see Chart 7-7). The parametersestimated in the regulated regime describe quite well thewithin-sample movements in consumer price inflation,but they do not provide an accurate description of con-sumer price movements in the liberalized regime, againsuggesting a structural break. The bottom panel of thischart shows that when the model's parameters are re-estimated using the data in the liberalized regime, themodel describes consumer price developments reason-ably well.

The Role of Monetary Factorsin Land Price Inflation

An important feature of the estimated land price equa-tion is that monetary factors (the call money rate andeither the money or credit aggregate) together played animportant role in the determination of land price inflation;moreover, credit and interest rates together were signifi-cantly more important in explaining land price inflationin the 1980s than in the 1970s. This is clear from themarginal significance levels of the estimated coefficients(Tables 7-2 and 7-3), from the implied variance decompo-sitions (Tables 7-4 and 7-5), and from the estimatedimpulse response functions (discussed below).

Variance decompositions, which quantify the percent-age contribution of each variable to the variation in landprice inflation, are shown in Table 7-4. The share of thevariation in land price inflation accounted for by mone-tary factors increases significantly in the second period,regardless of whether M2 + CDs or private credit isassumed to be the relevant financial aggregate. WithM2 + CDs as the financial aggregate, the contributionof monetary factors to the variation in the land priceinflation increases from a share of 50 percent in the 1970sto over 80 percent in the 1980s; with total private creditas the financial aggregate, the contribution of monetaryfactors to the variation in the land price inflation increasesfrom a share of 35 percent in the 1970s to 75 percent inthe 1980s. Note that when M2 + CDs is assumed to bethe financial aggregate, the contribution of the call moneyrate becomes negligible in the second period; in contrast,the contribution of the call money rate remains importantand increases when credit is assumed to be the financialaggregate. Finally, note that the output gap accounted foralmost one fourth of the variation in land price inflation inthe 1970s, regardless of the choice of financial aggregate.In the 1980s, however, the output gap accounted for lessthan 10 percent of the variation in land price inflationbecause of the dominance of monetary factors.

Variance decompositions are sensitive to the orderingof the variables according to their presumed exogeneity.One way of examining the robustness of a variable'scontribution is to alter the order of the variables. Whenthe monetary variables were placed last in the ordering(that is, assuming that monetary factors were affected by

70




Land PriceInflation

Consumer PriceInflation

Call money rate

M2 + CDs

Output gap

Land price inflation

Consumer price inflation

Memorandum

R2 (coefficient of determination)R2 (adjusted R2)ObservationsDegrees of freedomStandard error


contemporaneous innovations in the other variables), thecontribution of M2 + CDs to the variation in land priceinflation declined in the 1980s, whereas that of privatecredit increased substantially in the 1980s (see Table7-5). Thus, the contribution of M2 + CDs to the variationin land price inflation in the 1980s is not robust, but thatof total private credit is. The changes in the contributionof the call money rate and the output gap noted abovein the discussion of the initial ordering of variables arealso robust.

One interpretation of the estimated land price equationsis that there was a change in the transmission of monetaryfactors to real estate markets in the 1980s. Under thisinterpretation, the change in the transmission process ledto two important structural changes: first, bank credit(bank assets) played a more important role in the 1980sthan in the 1970s; second, bank credit played a moreimportant role in determining land price inflation thandid bank deposits (bank liabilities). This may imply thatbank credit became a more important indicator of infla-tionary pressures in the 1980s. If this structural changeis sustained, it will have important implications for theconduct of monetary policy in the 1990s.

To assess further the relative importance of the trans-mission processes of money versus credit, and to examinehow well the estimated equations track the actual pathof land price inflation in the 1980s, Chart 7-8 presents

Chart 7-8. Dynamic Simulations UsingTwo-Equation Subsystem and M2 + CDsas the Financial Aggregate1(Index, 1990:Q1 = 100)

Source: IMF staff calculations.1 Using estimated land price and consumer price equations from

estimates of the five-equation system.2Urban residential land prices in six largest cities.

Table 7-2. F-Tes Lagged Values of Explanatory Variables:Broad Honey (M2 + CDs)(Significance levels are in parentheses)

1970-83

1.50(0.24)

7.10(0.00)

3.20(0.03)

0.60(0.69)

2.60(0.06)

0.880.77

51.0027.00

1.50

1984-92

0.30(0.87)

2.60(0.08)

0.60(0.69)

1.90(0.17)

1.10(0.39)

0.880.69

39.0015.002.60

1970-83

2.10(0.12)

3.50(0.02)

0.90(0.47)

1.20(0.35)

3.60(0.02)

0.870.76

51.0027.000.75

1984-92

2.20(0.12)

1.00(0.45)

1.10(0.40)

0.20(0.93)

1.40(0.29)

0.650.12

39.0015.000.43

71



Table 7-3. F-Tests for Lagged Values of Explanatory Variables:Credit Consistent with National Accounts(Significance levels are in parentheses)

Call m o n e y rate

Private credit

Output gap

Land price inflation

C o n s u m e r price inflation

Memorandum

R2 (coefficient of determination)

R2 (adjusted)

Observations

Degrees of freedom

Standard error

Land Price

Inflation

1970-83

0.70

(0.60)

2.10

(0.11)

2.20

(0.10)

0.30

(0.86)

1.40

(0.27)

0.88

0.64

51.00

27.00

1.90

1984-92

1.20

(0.35)

5.40

(0.01)

1.20

(0.34)

1.80

(0.18)

2.90

(0.07)

0.95

0.80

36.00

12.00

2.00

Consumer Price

Inflation

1970-83

0.80

(0.53)

2.60

(0.06)

1.60

(0.21)

1.90

(0.13)

5.40

(0.00)

0.94

0.74

51.00

27.00

0.79

1984-92

2.60

(0.09)

2.90

(0.07)

1.90

(0.17)

1.00

(0.46)

2.70

(0.08)

0.91

0.53

36.00

12.00

0.47


Table 7-4. Variance Decompositions for the Land Price and ConsumerPrice Equations: Monetary Variables First in the Ordering(Percent of total variance)1

Land price equation

First period

Second period

C o n s u m e r price equation

First period

Second period

Land price equation

First period

Second period

C o n s u m e r price equation

First period

Second period

Interest

Rate

193

1227

Interest

Rate

2227

1825

M2 + CDs

3080

3838

Credit

1448

1535

Output

Gap

226

1310

Output

Gap

238

1419

Land

Price

184

92

Land

Price

28II

1610

Consumer

Prices

108

2724

Consumer

Prices

136

3612

Monetary

Combined

5083

5064

Monetary

Combined

3575

3460

Source: IMF staff calculations.1Percent of the total variance at the twentieth quarter (that is, in the long run).

72©International Monetary Fund. Not for Redistribution


Table 7-5. Variance Decompositions for the Land Price and ConsumerPrice Equations: Monetary Variables Last in the Ordering(Percent of total variance)1

Land price equation

First period

Second period

Consumer price equation

First period

Second period

Land price equation

First period

Second period

Consumer price equation

First period

Second period

Output

Gap

133

912

Output

Gap

2015

1029

Land

Price

2238

1322

Land

Price

3113

199

Consumer

Prices

1237

3148

Consumer

Prices

2013

4318

Interest

Rate

253

168

Interest

Rate

1826

1417

M2 + CDs

2819

3210

Credit

1033

1527

Monetary

Combined

5322

4818

Monetary

Combined

2859

2845

Source: IMF staff calculations.1Percent of the total variance at the twentieth quarter (that is, in the long run).

Chart 7-9. Dynamic Simulations UsingTwo-Equation Subsystem and PrivateCredit as the Financial Aggregate:Urban Land Prices1(Index, 1990:01 = 100)

Source: IMF staff calculations.1Using estimated land price and consumer price equations from

estimates of the five-equation system.2Urban residential land prices in six largest cities.

dynamic simulations of the land price equation usingM 2 + C D s as the financial aggregate, and Chart 7-9presents dynamic simulations for the model using totalprivate credit as the financial aggregate. The lines labeled"regime 1" represent the dynamic simulation of the modelw h e n the parameters are estimated over the regulatedregime, and the lines labeled "regime 2" represent thedynamic simulation of the model w h e n the parametersare estimated in the liberalized regime.

The simulations are "dynamic" in the sense that thepath of inflation is simulated using the forecasted valuesof the lagged dependent variable, but actual data areused for the other explanatory variables.13 These dynamicsimulations provide information about whether the landprice inflation could have been accurately forecasted bythe model had the evolution of the monetary policy vari-ables and the output gap been k n o w n .

13 These simulations use actual data for nonprice variables (the callm o n e y rate; the financial aggregate, either M 2 + C D s or privatecredit; and the output gap), but projected values for the lagged pricevariables. T o examine the explanatory power of the lagged dependentvariable, the simulation was run assuming coefficients of zero onlagged values of land price inflation. For the equation using totalprivate credit, the only difference in the simulation was that landprices overshot the peak for about two quarters—and by about 1 0 -15 percent—and then turned downward and followed a path parallelto the one presented in Chart 7-9; this is what is suggested by thevariance decompositions presented in Tables 7-4 and 7-5, in whichthe contribution of lagged values in land price inflation to the totalvariation of land price inflation was less than 10 percent.

73



Chart 7-10. Dynamic Simulations UsingTwo-Equation Subsystem and PrivateCredit as the Financial Aggregate:Consumer Prices1(Index, 1990:QI = 100)

Source: IMF staff calculations.1Using estimated land price and consumer price equations from

estimates of the five-equation system.

A s the charts show, regardless of the financial aggre-gate used for estimation, the parameters estimated usingdata in the liberalized regime perform very well and aresignificantly better than the parameters estimated usingdata in the regulated regime. Indeed, the simulations ofregime 1 in each case suggest that fundamental macroeco-nomic variables could not explain very well the landprice inflation—implying that it might be a bubble. Alsonote that, although growth in M 2 + C D s performs rea-sonably well as a financial aggregate, growth in totalprivate credit clearly dominates the monetary aggregate,which confirms the information contained in the variancedecompositions.14

The Consumer Price Inflation Equation

A n important additional manifestation of the apparentregime shift was a breakdown in the 1980s in the relation-

ship between consumer price inflation and the other vari-ables in the estimated system. In the 1970s, lagged valuesof M 2 + C D s , consumer prices, and the call m o n e y ratewere fairly important variables in the estimated consumerprice inflation equation, and these determinants togetherexplained a very high share of the overall variation ofconsumer price inflation in the first subsample period.In the 1980s, however, these marginal contributions de-clined, and the explanatory variables together have verylittle explanatory power—as represented by the adjustedcoefficient of determination, R2.

In contrast, w h e n private credit is used as the financialaggregate, the marginal significance of lagged values ofprivate credit and the call m o n e y rate improve, and thatof lagged values of consumer price inflation declines. Inaddition, with private credit as the aggregate, the marginalsignificance of the output gap is greater than w h e n broadmoney is the financial aggregate. Overall, the equationwith private credit holds up fairly well, and monetaryfactors (the call m o n e y rate and private credit) were moreimportant in determining goods-price inflation in the1980s than in the 1970s.15

A s discussed earlier, the hypothesis that there was astructural break in the early to mid-1980s in the modelis illustrated by the static simulations of the consumerprice inflation equation (see Chart 7-7). Chart 7-10 showstwo dynamic simulations for consumer prices. The linelabeled "regime 1" represents a dynamic simulation ofthe consumer price inflation equation using coefficientsestimated in the regulated data regime, whereas "regime2 " uses the model coefficients estimated using the datain the liberalized data regime. These simulations clearlyshow that, given the actual evolution of the call moneyrate, credit growth, and the output gap, the model esti-mated over the regulated regime would predict greaterinflation on average than either the model estimated overthe liberalized data regime or actual consumer pricedevelopments.

The Concentration of Monetary Shocksin Land Prices

Taken together, the "dynamic" simulations for bothprice equations suggest that the concentration hypothesisdiscussed earlier cannot easily be rejected. The five-equa-tion model estimated over the liberalized data regimeclearly suggests that the actual evolution of monetary

14 W h e n full dynamic simulations are run—that is, using the esti-mated forecasted values of all of the explanatory variables, includingestimated values of r, m, y, TTL, and TT—the models do not performvery well. But as the "dynamic" simulations show, this is largelybecause of the errors in forecasting r, m, and y.

15 W h e n the five-equation system was expanded to include the ex-change rate as a third monetary factor, the land price equation wasqualitatively unaffected; in addition, lagged values of the exchangerate m a d e a statistically significant marginal contribution, the callm o n e y rate and the financial aggregate remained significant, and theoverall fit improved. Regarding the estimated consumer price equa-tion, the most important difference was that the contribution of m o n e -tary factors to the variation in consumer price inflation increased.

74


Chart 7-12. Impulse Responseto Private Credit1

Source: IMF staff calculations.1Impulse responses have been normalized by monetary shock.

Thus, a response of 1.0 means that the variable responds 1.0 timesthe standard error of the innovation.

Source: IMF staff calculations.1Impulse responses have been normalized by credit shock. Thus,

a response of 1.0 means that the variable responds 1.0 times thestandard error of the innovation.

variables in the mid- to late 1980s led to more land priceinflation and less consumer price inflation than wouldhave been predicted by the model estimated over theregulated regime.

These results imply that a money or credit shock tothe equation system would lead to a concentration ofinflationary pressure in land prices and not in goodsprices in the liberalized data regime compared with theregulated data regime. This can be examined directly byexamining the impulse response functions implicit inthe estimated land price and consumer price inflationequations. The responses of land price inflation and con-sumer price inflation to a monetary shock and a creditshock are shown in Charts 7-11 and 7-12, respectively.The impulse response functions suggest that both mone-tary and credit shocks were more heavily concentratedin asset markets than in goods markets in the liberalizedregime than they were in the regulated regime.

Conclusions

An important conclusion to be drawn from the preced-ing analyses is that movements in land price inflation inJapan in the 1980s can be explained quite well by arelatively unrestricted multiequation model that includesmonetary factors, the output gap, and consumer priceinflation—with monetary factors being the most im-portant. A second important conclusion is that the hypoth-esis that there was no regime switch in the period1970-93 is strongly rejected by the data. Specifically, afeature of the estimated system of equations is that thedynamic forces that determined land price and consumerprice inflation in the period 1970-83 were quite differentfrom those in 1984-93. In the second period there wasa sizable increase—a doubling—in the contribution ofmonetary factors to the variation in land price inflation,and both land price and consumer price inflation in the

75

Conclusions

Chart 7-11. Impulse Responseto M2 + CDs1



second period were more affected by total private creditthan by M2 + CDs. In addition, the view that inflationarypressures in the 1980s were highly concentrated in assetprices is broadly supported by the estimated model; amonetary shock (or a credit shock) leads to more landprice inflation and to less consumer price inflation whenthe model is simulated using parameters estimated overthe period 1984-93 than when the parameters are esti-mated over the period 1970-83. Thus, a key conclusionis that, although monetary expansion typically led toconsumer price inflation before the mid-1980s, becauseof structural changes it has since tended to manifest itselfin asset price inflation.

In sum, the empirical results suggest that financialderegulation, which took place in the late 1970s andthroughout the 1980s, had an important influence in redi-recting the influence of monetary factors toward assetmarkets.

Appendix 7-1. Vector-AutoregressionModeling

This appendix consists of three brief subsections thatdeal with the time-series properties of the data, the num-ber of lags to include in the VAR model, and evidenceof a structural break in the model.

Time-Series Properties

Before estimating the VAR model, the time-seriesproperties of the different variables were examined toensure efficient estimation. Standard unit root tests wereused to determine the time-series properties of the vari-ables. These unit root tests suggest that all the variablesare stationary (Table 7-6).

The number of lags included in these tests was chosenby "testing down" for the highest significant lag. MonteCarlo simulations suggest that this selection mechanismwill choose the correct number of lags as the number ofobservations increases (see Campbell and Perron (1991)).

The augmented Dickey-Fuller and the Phillips-Perrontests reject the presence of a unit root for all variablesexcept for land inflation and relative land inflation. Twopossible reasons can explain why these tests failed toreject the null hypothesis of a unit root for land inflationand relative land inflation. First, it is well known thatthese tests have low power to discriminate between amoderately high persistent series and a true unit root.Second, the tests could fail to recognize a stationaryseries when the series contains a break, as is the casewith land price series.

Lag Length

Following Lutkepohl (1985), we have calculated theHannan-Quinn and Schwarz test for four VAR models(Table 7-7). The first two models use M2 + CDs as themonetary aggregate, one with land price inflation andthe other with relative land price inflation. The third andfourth models use credit as the monetary aggregate, onewith land price inflation and the other with relative landprice inflation.

The tests suggest that four lags are appropriate for allmodels except the model with credit and land inflation,where three lags are selected. However, for consistencyin the degrees of freedom, we have chosen to estimateall four models with four lags.

Structural Breaks

To test for a structural break in the VAR model, log-likelihood ratio tests were performed. The 120 coeffi-

Output gap

M2 + CDsCredit

Interest rate

Inflation

Land inflation

Relative land inflation

-3.15-2 .87

-1 .58-3.75-2 .32-2.14

-2 .34

-3.14-2 .92-3.24

-2.41

-3 .58-2 .97

-2.71

-3 .29-3 .29-3.29

-3 .29-3 .29-3 .29

-3 .29

-3.03-3.03-3.03-3.03-3.03

-3.03-3.03

76

1Augmented Dickey-Fuller test.2Phillips-Perron test.3Taken from Guilkey and Schmidt (1989, Table 1; n = 100).

Table 7-6. Unit Root Tests

Critical Values3

ADF1 pp2 5 percent 10 percentSeries


Lag

12345678

2345678

(1)

58.0158.1158.5056.46*61.9562.2760.2064.22

61.2958.2755.28*61.3056.9559.4758.3065.15

(2)

M2 +

57.6157.2957.2654.82*59.8759.7457.2460.81

(1) (2)

- CDs

60.2657.1558.9955.34*59.8759.4060.4565.27

Credit

60.8857.4554.04*59.6454.8656.9555.3461.73

59.8457.4560.6056.46*57.7360.1261.5663.05

59.8556.3357.7653.68*57.7856.8857.4961.86

59.4356.6459.3754.80*55.6557.6058.6059.64

1(1) and (2) correspond to the Hannan-Quinn and Schwarztests; an asterisk (*) denotes the lag length selected.

cients ((5 equations X 5 variables X 4 lags) + (4seasonal dummies X 5 equations)) of the unconstrainedmodel were estimated over two different subsamples,while the coefficients of the constrained model were notallowed to vary and were estimated over the entiresample.

Two VAR models were tested for structural breaks(Table 7-8). The first model uses M2 + CDs as themonetary aggregate, while the second uses credit; bothuse land inflation as the asset price. The test results findstatistically significant evidence to support the hypothesisthat there was a break in the model in the early 1980s.

References

Bank for International Settlements, Financial Innovation andMonetary Policy (Basle, March 1984).

, Changes in Money-Market Instruments and Proce-dures: Objectives and Implications (Basle, March 1986).

Campbell, John Y., and Pierre Perron, "Pitfalls and Opportuni-ties: What Macroeconomists Should Know About UnitRoots," in NBER Macroeconomics Annual 1991, editedby Olivier Jean Blanchard and Stanley Fischer (Cam-bridge, Massachusetts: MIT Press, 1991).

I982:Q4 150.43I983:Q1I983:Q2I983:Q3I983:Q4I984:QII984:Q2I984:Q3I984:Q4I985:Q1I985:Q2I985:Q3I985:Q4

144.20144.37143.19141.16141.61139.33140.78141.08145.03148.49146.60160.67

0.030.070.060.070.090.090.110.090.090.060.040.05

—

185.29179.35171.03171.16173.03175.26182.55183.24188.88193.49202.04204.94215.47

—————————————

Guilkey, David K., and Peter Schmidt, "Extended Tabulationsfor Dickey-Fuller Tests," Economics Letters, Vol. 31(1989), pp. 355-57.

Hargraves, Monica, Garry J. Schinasi, and Steven R. Weis-brod, "Asset Price Inflation in the 1980s: A Flow of FundsPerspective," IMF Working Paper 93/77 (Washington,October 1993).

Hoffmaister, Alexander W., and Garry J. Schinasi, "AssetPrices, Financial Liberalization, and the Process of Infla-tion in Japan," IMF Working Paper 94/153 (Washington,December 1994).

International Monetary Fund, World Economic Outlook(Washington, various issues).

Lutkepohl, Helmut, "Comparison of Criteria for Estimatingthe Order of a Vector Autoregressive Process," Journalof Time Series Analysis, Vol. 6 (No. 1, 1985), pp. 35-52.

Samiei, Hossein, and Garry J. Schinasi, "Real Estate PriceInflation, Monetary Policy, and Expectations in theUnited States and Japan," IMF Working Paper 94/12(Washington, January 1994).

Schinasi, Garry J., "Asset Prices, Monetary Policy, and theBusiness Cycle," IMF Paper on Policy Analysis and As-sessment 94/6 (Washington, March 1994).

Schinasi, Garry J., and Monica Hargraves, "'Boom and Bust'in Asset Markets in the 1980s: Causes and Conse-quences," in Staff Studies for the World EconomicOutlook (Washington: International Monetary Fund,December 1993), pp. 1-27.

Sims, Christopher, "Macroeconomics and Reality," Econo-metrica, Vol. 48 (January 1980), pp. 1-48.

References

Table 7-7. Lag Length Tests1

Land InflationRelative Land

Inflation

Table 7-8. Likelihood Ratio Testsfor Structural Breaks

System withM2 3+ CDs System with Credit

Marginal MarginalX2 significance X2 significance

77


OCCASIONAL PAPERS

Recent Occasional Papers of the International Monetary Fund

124. Saving Behavior and the Asset Price "Bubble" in Japan: Analytical Studies, edited by Ulrich Baumgartner andGuy Meredith. 1995.

123. Comprehensive Tax Reform: The Colombian Experience, edited by Parthasarathi Shome. 1995.

122. Capital Flows in the APEC Region, edited by Mohsin S. Khan and Carmen M. Reinhart. 1995.

121. Uganda: Adjustment with Growth, 1987-94, by Robert L. Sharer, Hema R. De Zoysa, and Calvin A.McDonald. 1995.

120. Economic Dislocation and Recovery in Lebanon, by Sena Eken, Paul Cashin, S. Nuri Erbas, Jose Martelino,and Adnan Mazarei. 1995.

119. Singapore: A Case Study in Rapid Development, edited by Kenneth Bercuson with a staff team comprisingRobert G. Carling, Aasim M. Husain, Thomas Rumbaugh, and Rachel van Elkan. 1995.

118. Sub-Saharan Africa: Growth, Savings, and Investment, by Michael T. Hadjimichael, Dhaneshwar Ghura,Martin Muhleisen, Roger Nord, and E. Murat Ucer. 1995.

117. Resilience and Growth Through Sustained Adjustment: The Moroccan Experience, by Saleh M. Nsouli, SenaEken, Klaus Enders, Van-Can Thai, Jorg Decressin, and Filippo Cartiglia, with Janet Bungay. 1995.

116. Improving the International Monetary System: Constraints and Possibilities, by Michael Mussa, MorrisGoldstein, Peter B. Clark, Donald J. Mathieson, and Tamim Bayoumi. 1994.

115. Exchange Rates and Economic Fundamentals: A Framework for Analysis, by Peter B. Clark, LeonardoBartolini, Tamim Bayoumi, and Steven Symansky. 1994.

114. Economic Reform in China: A New Phase, by Wanda Tseng, Hoe Ee Khor, Kalpana Kochhar, DubravkoMihaljek, and David Burton. 1994.

113. Poland: The Path to a Market Economy, by Liam P. Ebrill, Ajai Chopra, Charalambos Christofides, PaulMylonas, Inci Otker, and Gerd Schwartz. 1994.

112. The Behavior of Non-Oil Commodity Prices, by Eduardo Borensztein, Mohsin S. Khan, Carmen M. Reinhart,and Peter Wickham. 1994.

111. The Russian Federation in Transition: External Developments, by Benedicte Vibe Christensen. 1994.

110. Limiting Central Bank Credit to the Government: Theory and Practice, by Carlo Cottarelli. 1993.

109. The Path to Convertibility and Growth: The Tunisian Experience, by Saleh M. Nsouli, Sena Eken, Paul Duran,Gerwin Bell, and Zuhtu Yucelik. 1993.

108. Recent Experiences with Surges in Capital Inflows, by Susan Schadler, Maria Carkovic, Adam Bennett, andRobert Kahn. 1993.

107. China at the Threshold of a Market Economy, by Michael W. Bell, Hoe Ee Khor, and Kalpana Kochhar withJun Ma, Simon N'guiamba, and Rajiv Lall. 1993.

106. Economic Adjustment in Low-Income Countries: Experience Under the Enhanced Structural AdjustmentFacility, by Susan Schadler, Franek Rozwadowski, Siddharth Tiwari, and David O. Robinson. 1993.

105. The Structure and Operation of the World Gold Market, by Gary O'Callaghan. 1993.

104. Price Liberalization in Russia: Behavior of Prices, Household Incomes, and Consumption During the FirstYear, by Vincent Koen and Steven Phillips. 1993.

103. Liberalization of the Capital Account: Experiences and Issues, by Donald J. Mathieson and Liliana Rojas-Suarez. 1993.

102. Financial Sector Reforms and Exchange Arrangements in Eastern Europe. Part I: Financial Markets andIntermediation, by Guillermo A. Calvo and Manmohan S. Kumar. Part II: Exchange Arrangements ofPreviously Centrally Planned Economies, by Eduardo Borensztein and Paul R. Masson. 1993.

101. Spain: Converging with the European Community, by Michel Galy, Gonzalo Pastor, and Thierry Pujol. 1993.

100. The Gambia: Economic Adjustment in a Small Open Economy, by Michael T. Hadjimichael, ThomasRumbaugh, and Eric Verreydt. 1992.

99. Mexico: The Strategy to Achieve Sustained Economic Growth, edited by Claudio Loser and Eliot Kalter. 1992.

78


Occasional Papers

98. Albania: From Isolation Toward Reform, by Mario I. Blejer, Mauro Mecagni, Ratna Sahay, Richard Hides,Barry Johnston, Piroska Nagy, and Roy Pepper. 1992.

97. Rules and Discretion in International Economic Policy, by Manuel Guitian. 1992.

96. Policy Issues in the Evolving International Monetary System, by Morris Goldstein, Peter Isard, Paul R.Masson, and Mark P. Taylor. 1992.

95. The Fiscal Dimensions of Adjustment in Low-Income Countries, by Karim Nashashibi, Sanjeev Gupta, ClaireLiuksila, Henri Lorie, and Walter Mahler. 1992.

94. Tax Harmonization in the European Community: Policy Issues and Analysis, edited by George Kopits. 1992.

93. Regional Trade Arrangements, by Augusto de la Torre and Margaret R. Kelly. 1992.

92. Stabilization and Structural Reform in the Czech and Slovak Federal Republic: First Stage, by Bijan B.Aghevli, Eduardo Borensztein, and Tessa van der Willigen. 1992.

91. Economic Policies for a New South Africa, edited by Desmond Lachman and Kenneth Bercuson with a staffteam comprising Daudi Ballali, Robert Corker, Charalambos Christofides, and James Wein. 1992.

90. The Internationalization of Currencies: An Appraisal of the Japanese Yen, by George S. Tavlas and YuzuruOzeki. 1992.

89. The Romanian Economic Reform Program, by Dimitri G. Demekas and Mohsin S. Khan. 1991.

88. Value-Added Tax: Administrative and Policy Issues, edited by Alan A. Tait. 1991.

87. Financial Assistance from Arab Countries and Arab Regional Institutions, by Pierre van den Boogaerde. 1991.

86. Ghana: Adjustment and Growth, 1983-91, by Ishan Kapur, Michael T. Hadjimichael, Paul Hilbers, JeraldSchiff, and Philippe Szymczak. 1991.

85. Thailand: Adjusting to Success—Current Policy Issues, by David Robinson, Yangho Byeon, and Ranjit Tejawith Wanda Tseng. 1991.

84. Financial Liberalization, Money Demand, and Monetary Policy in Asian Countries, by Wanda Tseng andRobert Corker. 1991.

83. Economic Reform in Hungary Since 1968, by Anthony R. Boote and Janos Somogyi. 1991.

82. Characteristics of a Successful Exchange Rate System, by Jacob A. Frenkel, Morris Goldstein, and Paul R.Masson. 1991.

81. Currency Convertibility and the Transformation of Centrally Planned Economies, by Joshua E. Greene andPeter Isard. 1991.

80. Domestic Public Debt of Externally Indebted Countries, by Pablo E. Guidotti and Manmohan S. Kumar. 1991.

79. The Mongolian People's Republic: Toward a Market Economy, by Elizabeth Milne, John Leimone, FranekRozwadowski, and Padej Sukachevin. 1991.

78. Exchange Rate Policy in Developing Countries: Some Analytical Issues, by Bijan B. Aghevli, Mohsin S. Khan,and Peter J. Montiel. 1991.

77. Determinants and Systemic Consequences of International Capital Flows, by Morris Goldstein, Donald J.Mathieson, David Folkerts-Landau, Timothy Lane, J. Saul Lizondo, and Liliana Rojas-Suarez. 1991.

76. China: Economic Reform and Macroeconomic Management, by Mario Blejer, David Burton, Steven Dunaway,and Gyorgy Szapary. 1991.

75. German Unification: Economic Issues, edited by Leslie Lipschitz and Donogh McDonald. 1990.

74. The Impact of the European Community's Internal Market on the EFTA, by Richard K. Abrams, Peter K.Cornelius, Per L. Hedfors, and Gunnar Tersman. 1990.

73. The European Monetary System: Developments and Perspectives, by Horst Ungerer, Jouko J. Hauvonen,Augusto Lopez-Claros, and Thomas Mayer. 1990.

72. The Czech and Slovak Federal Republic: An Economy in Transition, by Jim Prust and an IMF Staff Team. 1990.

Note: For information on the title and availability of Occasional Papers not listed, please consult the IMF Publications Catalog or contact IMFPublication Services.

79


Date post:	27-Jan-2023
Category:	Documents
Upload:	khangminh22
View:	0 times
Download:	0 times

124 - IMF eLibrary - International Monetary Fund

Documents