Journal Issue
Share
Article

Shorter Papers and Comments: Wealth, Financial Liberalization, and the Demand for Money in Japan

Author(s):
International Monetary Fund. Research Dept.
Published Date:
January 1990
Share
  • ShareShare
Show Summary Details

This paper provides explanations for two features of broad money demand in Japan that have not been fully accounted for in the empirical literature. First, the long-term trend decline in the income velocity of broad money is explained by conventional wealth effects in a money-demand function. Second, the sharp acceleration in the decline of velocity after 1985 is attributed to a combination of rapid growth in demand for private financial assets and to the falling opportunity cost of holding money. The process of financial liberalization is identified as an important indirect influence on the more recent developments in velocity. The paper does not, however, support the view expressed by some researchers that there has been a shift in money-demand behavior in recent years.

The paper is organized as follows. Section I describes trends in broad money during 1970–88 and provides background information for the subsequent empirical analysis. Sections II and III describe, respectively, the specification and estimation of a money-demand function that is appropriate for the analysis of a broad monetary aggregate. Section IV then describes how the estimated money-demand function can be used to explain broad money growth in the period 1986–88. Conclusions are summarized in Section V. The Appendix provides data definitions and sources and a listing of important deregulation measures taken during 1985–88.

I. Background: Trends in Broad Money, 1970–88

The focus of this paper is M2 plus certificates of deposit (CDs), hereafter denoted as broad money. The aggregate consists of cash plus demand and time deposits in the banking system, with CDs included since their introduction in May 1979. Broad money is one of the most closely watched indicators of monetary conditions in Japan. Indeed, reflecting its importance in the formulation of monetary policy, the monetary authorities have announced one-quarter-ahead growth forecasts for broad money since July 1978. Understanding the determinants of broad money demand, therefore, has significant policy implications.

Over the last 20 years—apart from a brief period in the first half of the 1970s—the growth of broad money has consistently outstripped nominal income growth. The resulting trend decline in the income velocity of broad money has averaged over 2 percent a year in this period (Figure 1, upper panel). Reflecting this trend, most empirical studies have estimated the income elasticity of broad money demand to be greater than unity.1 These studies implicitly suggest that the trend in velocity is a structural phenomenon. However, little reason is offered as to why this should be so.

Figure 1.Broad Money, 1970–88

(In percent)

Source: Author’s estimates.

The rate of decline in velocity accelerated to about 4½ percent a year in the period 1986–88. The reasons for this accelerated decline are not fully explained in most empirical studies.2 Indeed, some researchers (for example, Bank of Japan (1988)) have concluded that there was a structural break in money-demand behavior in this period. The structural break is attributed to the process of financial liberalization, which gained momentum in the second half of the 1980s.

This paper contends that developments in the income velocity of broad money in the 1970s and 1980s can be explained by a money-demand function that includes a combination of conventional wealth effects and an appropriate measurement of the opportunity cost of holding money. The motivation for including wealth as well as income in the demand function stems from the need to model more carefully broad money’s dual function as both a store of value and as the means of payment for current transactions. The trend decline in the income velocity of broad money can then be explained by the more rapid historical growth rate of wealth, compared to income, which increased the relative importance of the portfolio motive for holding broad money at the expense of the transactions motive. The comparatively slower rate of growth of demand for broad money to satisfy transactions needs would, at the same time, account for a gradual historical decline in the share of broad money in private financial portfolios, despite a steady decline in the opportunity cost of holding broad money (see Figure 1, upper panel; and Table 1).

Table 1.Average Return on Financial Assets, 1971–88(In percent)
Asset1971–751976–801981–851986–88
Return on broad moneya2.72.52.71.9
Three-month Gensaki rate9.07.16.74.3
Government bond yields7.97.87.54.5
Memorandum item:
GNP deflator (percent change)10.25.82.20.9

Author’s calculations; see Appendix for more details.

Author’s calculations; see Appendix for more details.

The importance of wealth effects on broad money demand might be inferred from developments in the structure of broad money during the 1970s and 1980s. In particular, demand for quasi-money—which is more likely to be held for portfolio investment as opposed to transactions purposes—grew very rapidly, so that by the end of 1988 it accounted for 71 percent of broad money compared with 62 percent at the beginning of 1970 (see Table 2). Correspondingly, narrow money (M1)—which is a subcategory of M2 plus CDs—grew much less rapidly than broad money. Indeed, the income velocity of narrow money demonstrated little marked trend in the 1970s and 1980s.

Table 2.Composition of Broad Money, 1970–88
Broad Money (M2 + CDs)1970:11979:11985:21988:4
(In trillions of yen; end period)
46.6177.6299.8419.7
Cash and currency3.914.021.031.5
Deposit money13.952.666.380.3
Quasi-money28.8111.0202.4297.5
Time deposits with
liberalized interest
ratesa85.5
Money market certificatesa3.129.5
Certificates of deposit10.010.4
(In percent of total broad money)
Cash and currency8.37.97.07.5
Deposit money29.929.622.119.1
Quasi-money61.862.567.570.9
Time deposits with
liberalized interest rates20.3
Money market certificates1.07.0
Certificates of deposit3.32.5
Memorandum item:
Share of broad money with
market-related interest ratesb4.329.8
Sources: Bank of Japan, Economic Statistics Monthly; and author’s estimates.

Data from banking accounts of city banks, sogo banks, and shinkin banks.

Money market certificates plus certificates of deposits plus time deposits with liberalized interest rates.

Sources: Bank of Japan, Economic Statistics Monthly; and author’s estimates.

Data from banking accounts of city banks, sogo banks, and shinkin banks.

Money market certificates plus certificates of deposits plus time deposits with liberalized interest rates.

Sectoral developments in broad money in the period 1986–88 would also be consistent with the importance of the portfolio investment demand for money. In this period, the unusually rapid growth of broad money was led by corporate demand for quasi-money at a time when the size of corporate balance sheets was expanding very rapidly. The expansion of corporate balance sheets can, in turn, be attributed to a sharp increase in corporate wealth and the relatively new phenomenon in Japan of corporate financial engineering (often called “zaitech”). Furthermore, corporations benefited most from financial liberalization in this period, which focused on the deregulation of interest rates on deposit-type instruments for large-scale investors (see Appendix for more details). Such financial liberalization contributed to furthering the historical trend decline in the opportunity cost of holding broad money (see Figure 1, lower panel).

II. Specification of the Demand-for-Money Function

The specification of the demand function for a broad monetary aggregate such as M2 plus CDs should take into account the aggregate’s dual nature: it is held both to finance transactions and as a store of value. It is not clear, therefore, that commonly estimated functions, which link money demand to income and interest rates, would adequately capture the portfolio demand for broad money. Instead, the store of value characteristic of broad money provides a strong argument for including a measure of wealth in the demand function in order to capture portfolio allocation decisions among financial assets.

The money-demand function should also take into account the fact that a large proportion of broad money is interest bearing. In typical theoretical specifications, money demand depends negatively on the opportunity cost of holding money. This latter variable should, therefore, be defined as the return on competing assets net of the average return on holding broad money.

Theoretical and empirical specifications of money-demand functions are surveyed exhaustively elsewhere (for example, Laidler (1985)). In short, most specifications relate money demand to scale variables, such as income and wealth, opportunity-cost variables, and, sometimes, to other miscellaneous factors. In this paper, the money-demand function is specified in real terms, assuming a log-linear functional form. In the long run, the money-demand function can be written as

where M denotes money, P is a price index, W is wealth, Y is income, and R is the opportunity cost of holding money (relative asset returns). The a’s are elasticities to be estimated, and A0 is a constant. However, because of adjustment costs and because expected developments in the independent variables are partly adaptive to recent outcomes, a dynamic version of equation (1) is required for estimation.3 The dynamic specification chosen here is of the error-correction variety (see Davidson and others (1978)), in which changes in money demand depend on both changes in and the levels of the independent variables, plus lagged dependent variables. The error-correction model is a more complex form of partial adjustment (Nickell (1985)), whereby deviations from long-run desired money holdings (as represented by equation (1)) lead to changes in money demand to restore equilibrium. A major benefit of the error-correction specification is that it permits estimation of long-run relationships between the levels of economic variables that are often lost by the first-differencing of variables needed to circumvent the econometric problem of nonstationarity.4

III. Estimation Results

Quarterly seasonally adjusted data were used to estimate the demand-for-money function.5 For stock variables (money, wealth), period average values were used. Wealth was proxied by the positive financial assets of the private nonfinancial sectors. This variable is appropriate for capturing private portfolio allocation decisions among financial assets, but it does not directly capture some potentially interesting portfolio allocation decisions between financial and real assets (notably, land). Income was proxied by gross national product (GNP), and the GNP deflator was chosen as the relevant price variable.6 The opportunity-cost variable was proxied by the three-month average Gensaki rate minus the average return on holding broad money. In principle, the opportunity-cost variable should include a vector of returns on many diverse financial assets. In practice, potential multicollinearity between asset returns restricted the number of asset returns that could be included in the equation.

The demand-for-money function was estimated over the period 1970–88 by instrumental variables in order to provide some protection against potential biases arising from simultaneity. The initial dynamic specification assumed that changes in real money demand depended on a lagged dependent variable, current and one-period lagged values of first differences in the independent variables, and one-period lagged values of the levels of the independent and dependent variables.7 Some of these terms had insignificant coefficients and were dropped from the final equation. Data-supported restrictions on the coefficients of some of the lagged-level terms were also imposed. The estimation results were as follows:

Instrumental variables estimation, 1970:4–1988:48

where D log is the first difference of the log of the variable, 2 is the coefficient of determination adjusted for degrees of freedom, DW is the Durbin-Watson statistic, and SE is the equation standard error. The figures in parentheses are t-statistics.

AUTO(5) is the X2 test statistic for (up to) fifth-order autocorrelation (Godfrey (1978)); the critical 95 percent confidence level is in parentheses. This test is passed, and suggests that there is no evidence of residual autocorrelation. STAB(n) is the X2 test statistic for parameter stability in the last n periods of the data sample (Hendry (1980)); the critical 95 percent confidence level is in parentheses. As discussed in Section IV below, this test is passed for n = 16 and n = 36.

The estimation results support the hypothesis that both income and wealth are important determinants of the demand for broad money in Japan.9 Both variables have significant positive coefficients in the estimated equation. There are also significant negative coefficients on the opportunity-cost variables. According to the dynamics of the equation, there are no immediate effects on money demand of a change in financial wealth, but, instead, money demand changes gradually over a fairly long period to restore desired portfolio shares. By contrast, changes in income and interest rates have significant short-run effects on money demand in addition to their long-run effects.

The terms defined as the difference between the levels of money and wealth and between the levels of money and income give the dynamics of equation (2) their error-correcting properties. These terms arose from restrictions on the more general specification of the equation described above. To bring out more clearly the nature of the error-correction mechanism, the estimated equation can be rewritten as

where M* is defined by the relationship

which corresponds to the long-run money-demand function (compare with equation (1)). Thus, as the level of money holdings rises above the desired long-run level for a given level of income, wealth, and interest rates, the error-correction mechanism acts to reduce money-demand growth.

The long-run money-demand function has the property that the income and wealth elasticities of money demand sum to one.10 This implies that, for a given level of the opportunity cost of holding money, if income and wealth grow at the same rate, the demand for broad money will also grow at this rate. As a consequence, the long-run income velocity of broad money can be written as

From equation (5) it is clear that, in the long run, the income velocity of money is only affected by the relative return on money and competing financial assets, and by differential growth rates between income and wealth. This result provides empirical support to the view, proposed in Section I, that the trend decline in the income velocity of broad money over the last two decades has been primarily due to the faster growth of wealth compared to income. As a corollary, the result provides an explanation for the high income elasticity of broad money demand estimated by other researchers.

IV. The Demand for Broad Money, 1986–88

Equation (2) explains the growth of broad money during 1986–88 quite well, and can account for the sharp acceleration in the decline of money velocity that occurred (Figure 2). Furthermore, formal parameter stability tests find no evidence of a behavioral break in this period, or, indeed, during the 1980s in general.11 This perhaps surprising result, bearing in mind the rapid pace of financial deregulation in this period, contrasts with the findings of some researchers.

Figure 2.Income Velocity of Broad Money, 1980–88

(GNP/M2 + CDs, in percent)

Source: Author’s estimates.

The parameter stability can be traced mainly to the inclusion of wealth in the equation, but appropriate specification of the opportunity cost of holding broad money also helped to reduce out-of-sample prediction error. In this regard, the equations’ parameters were no longer stable if wealth was omitted from the equation, while omitting the return on broad money from the opportunity-cost variable led to a notable increase in the out-of-sample prediction error in the period 1986–88. Thus, the parameter instability after 1985 of money-demand equations that omit wealth (for example, those reported in Suzuki and others (1988), and Bank of Japan (1988)) most likely results from misspecification. An implication of this result is that it does not appear that financial liberalization caused a break in money-demand behavior after 1985,12

The accelerated decline in the income velocity of broad money during 1986–88 can be attributed to two factors: the rapid expansion in the size of private financial balance sheets in this period and the declining opportunity cost of holding money. Simulations of the equation reveal that, of these two factors, the former was the most important. For example, if financial asset demand had grown at the same rate as GNP, as opposed to nearly twice as fast, growth of broad money would have been about 2½ percent a year slower. At the same time, if the opportunity cost of holding money had remained unchanged, instead of declining by about 2 percentage points, the growth of broad money would have been about 1¼ percent a year slower. It might be noted that about one fourth of this 2 percentage point decline in the opportunity cost of holding money can be directly attributed to financial liberalization, which raised the share of deposits paying a market-related return. The remainder of the decline can be attributed to different movements in regulated and unregulated interest rates.

These two simulations can account for the more rapid decline in money velocity in 1986–88. Some caution is warranted in interpreting these results too strictly, however, because of the partial equilibrium nature of the simulations.

V. Conclusions

A conventional money-demand function can describe both secular trends during the 1970s and 1980s and recent developments in broad money growth in Japan quite well. However, the money-demand function must properly take into account the store of value characteristics of broad money by including a measure of wealth, and adjust the opportunity-cost variable for the rate of return on money. When these refinements are made, it does not appear that money-demand behavior has changed significantly during the recent period of financial liberalization in Japan. Instead, the acceleration in the decline of the income velocity of broad money during 1986–88 can be attributed to wealth effects that caused a rapid expansion in demand for all assets, and a narrowing gap between the rate of return on broad money and competing assets. Both of these effects, however, were in part related to the process of financial liberalization.

APPENDIX Data Sources and Definitions and Deregulation Measures Taken From 1985–88

The data sources used were Nihon Keizai Shimbun, Inc.’s Nikkei Telecom: Japan News and Retrieval (Tokyo); the Bank of Japan’s Economic Statistics Monthly (Tokyo); and the International Monetary Fund’s International Financial Statistics (Washington). All data, except for interest rates, were seasonally adjusted. For money and wealth, seasonal adjustment was carried out by the author using the XII program.

The definitions of most of the data described in the main text, with the exception of financial wealth and the opportunity cost of holding money, do not require further elaboration. Financial wealth was defined from flow-of-funds data as the average of beginning- and end-period stocks of the sum of the total positive financial assets of the personal and corporate sectors. The opportunity cost of holding money was defined as the three-month Gensaki rate minus the average return on holding money. The latter was defined as a weighted average of the interest rate on three-month certificates of deposit and the guideline three-month deposit rate. The weight on the CD rate was equal to the share of liberalized time deposits and banking sector money market certificates plus CDs in broad money. The weight on the guideline deposit rate was equal to the share of quasi-money, excluding liberalized time deposits, money market certificates, and CDs, in broad money.

Important interest rate and related financial deregulation measures taken in Japan from 1985–88 are detailed in Table 3.

Table 3.Interest Rate and Related Financial Deregulation, 1985–88
DateMeasure
March 1985Introduction of money market certificates (MMCs) for mutual banks and credit associations;

minimum denomination ¥ 50 million, maturity 1–6 months.
April 1985Introduction of MMCs for other banks; minimum denomination ¥ 50 million, maturity 1–6 months.
Reduction of minimum denomination of CDs from ¥ 300 million to ¥ 100 million; minimum maturity of CDs

shortened from 3 months to 1 month.
October 1985Decontrol of interest rates on deposits of ¥ 1 billion or more with maturity of 3–24 months.
Enlargement of the ceiling on CD issues from 100 percent to 150 percent of each bank’s net worth.
Enlargement of the ceiling on MMC issues from 75 percent to 150 percent of each bank’s net worth.
April 1986Decontrol of interest rates on deposits of ¥ 500 million or more.
Maximum CD maturity lengthened from 6 to 12 months and ceiling on CD issues enlarged from 150 percent

to 200 percent of each bank’s net worth.
Enlargement of the ceiling on MMC issues from 150 percent to 200 percent of each bank’s net worth.
May 1986Enlargement of the ceiling on deposit insurance from ¥ 3 million to ¥ 10 million.
September 1986Decontrol of interest rates on deposits of ¥ 300 million or more.
Reduction of the minimum amount of MMCs from ¥ 50 million to ¥ 30 million and enlargement of the ceiling

on MMCs from 200 percent to 250 percent of each bank’s net worth.
Enlargement of the ceiling on CD issues from 200 percent to 250 percent of each bank’s net worth.
April 1987Decontrol of interest rates on deposits of ¥ 100 million or more.
Reduction of the minimum denomination of MMCs from ¥ 30 million to ¥ 20 million; maximum maturity

of MMCs lengthened from 12 to 24 months; enlargement of the ceiling on MMC issues from 250 percent to

300 percent of each bank’s net worth.
Enlargement of the ceiling on CD issues from 250 percent to 300 percent of each bank’s net worth.
October 1987Minimum maturity of large-denomination time deposits shortened from 3 months to 1 month.
Minimum denomination of MMCs reduced from ¥ 20 million to ¥ 10 million.
Ceilings on issues of MMCs and CDs removed.
April 1988Reduction of the minimum denomination of large time deposits and CDs from ¥ 100 million to ¥ 50 million.
Range of CD maturities widened from 1–12 months to 2 weeks-2 years.
November 1988Decontrol of interest rates on deposits of ¥ 30 million or more.
Source: Bank of Tokyo, Tokyo Financial Review, March 1989.
Source: Bank of Tokyo, Tokyo Financial Review, March 1989.
REFERENCES

    Bank of JapanRecent Growth of Money Stock,Special Paper No. 161 (Tokyo: The Bank of Japan, Research and Statistics DepartmentFebruary1988).

    BennettAdamWealth, Expenditure, and the Demand for Money: The Case of France,IMF Working Paper 87/26 (Washington: International Monetary Fund1987).

    CuthbertsonKeithThe Demand for Ml: A Forward-Looking Buffer Stock Model,Oxford Economic Papers (Oxford) Vol. 40 (March1988) pp. 11031.

    DavidsonJ.E.H.DavidHendryFrankSrba and StephenYeoEconomic Modeling of the Aggregate Time Series’ Relationship Between Consumer Expenditure and Income in the United Kingdom,The Economic Journal (London) Vol. 88 (1978) pp. 66192.

    GodfreyL.G.Testing Against General Autoregressive Moving Average Error Models when the Regressors Include Lagged Dependent Variables,Econometrica (Evanston, Illinois) Vol. 46 (1978) pp. 12931301.

    GriceJ. and AdamBennettWealth and the Demand for £M3 in the United Kingdom, 1963–78,Manchester School of Economic and Social Studies Vol. 52 (September1984) pp. 23971.

    HendryDavid F.“Predictive Failure and Econometric Modeling in Macroeconomics: The Transactions Demand for Money” in Econometric Modelinged. by PaulOrmerod (London: Heinemann1980).

    HendryDavid F.“Predictive Failure and Econometric Modeling in Macroeconomics: The Transactions Demand for Money” in Econometric Modelinged. by PaulOrmerod (London: Heinemann“The Encompassing Implications of Feedback Versus Feedforward Mechanisms in Econometrics” Oxford Economic Papers Vol. 40 (March1988) pp. 13249.

    HyllebergSvend and Gray ham E.MizonCointegration and Error Correction Mechanisms,The Economic Journal Vol. 99 (1989Supplement) pp. 11325.

    LaidlerDavid E.W.The Demand for Money (New York: Harper and Row1985).

    NickellStephenError Correction, Partial Adjustment and All That: An Expository Note,Oxford Bulletin of Economics and Statistics Vol. 47 (1985) pp. 11929.

    SuzukiY.AkioKuroda and HiromichiShirakawaMonetary Control Mechanism in Japan,Bank of Japan Monetary and Economic Studies Vol. 6 (November1988) pp. 127.

    TobinJamesA General Equilibrium Approach to Monetary Theory,Journal of Money Credit and Banking Vol. 1 (1969) pp. 1529.

    UedaKazuoFinancial Deregulation and the Demand for Money in Japan,Discussion Paper No. 66 (Osaka: Osaka University Faculty of EconomicsJuly1988).

    YoshidaTomooOn the Stability of the Japanese Money Demand Function: Estimation Results Using the Error Correction Model,Bank of Japan Monetary and Economic Studies Vol. 8 (January1990) pp. 148.

Robert Corker, an economist in the Asian Department, holds degrees from Oxford University and the University of Warwick.

The author would like to thank Bijan Aghevli, Jorge Márquez-Ruarte, and Daniel Citrin for their helpful comments.

For example, Yoshida (1990) estimates the income elasticity to be 1.4; Suzuki and others (1988) estimate the elasticity at around 1.5, and Bank of Japan (1988) suggests the elasticity lies in the range 1–1.2.

Ueda (1988) is an exception.

In principle, it would be preferable to separate out the dynamics arising from expectations formations and adjustment costs. Cuthbertson (1988) details one such attempt, but the econometric demands are substantial and may not ultimately improve the specification (see Hendry (1988)).

For more discussion on this point, and other issues concerning the econometric properties of error-correction models, see Hylleberg and Mizon (1989).

See Appendix for data definitions and sources.

Experimentation with other income variables (for example, aggregate demand) and price deflators (for example, the consumer price index) generally produced less satisfactory estimation results.

Unlike Yoshida (1990), no attempt was made to “nest-down” from a general specification that included higher-order lagged variables. The more parsimonious specification adopted here nevertheless allows money demand to respond in a fairly complex dynamic fashion to a change in the independent variables.

The instruments used were real government consumption, lagged changes in the yen-dollar exchange rate, average government bond yields, the U.S. Treasury bill rate, and U.S. real GNP.

The result is, of course, not unique to Japan. See, for example, Bennett (1987) and Grice and Bennett (1984) for similar results for other countries.

The equation, therefore, conforms to the theoretical money-demand function described in Tobin (1969). In the unrestricted version of the equation, the elasticities had summed to 0.92. A standard F-test did not reject the restriction that the sum of the elasticities was unity: F (1, 65) = 0.38 (3.99).

Although the sharp acceleration in the decline in velocity dates from the beginning of 1986, the statistical test STAB(16) tested for parameter stability during 1985–88 in order to examine more generally possible behavioral changes during this period’s progressive deregulation of interest rates. The test STAB(36) tested for parameter stability during 1980–88.

The result is in broad agreement with Ueda (1988). Yoshida (1990) also presents estimates of a money-demand function that has stable parameters in this period, but that does not include wealth as an explanatory variable. However, Yoshida’s equation has an unexplained long-run (implicitly structural) income elasticity of 1.4.

Other Resources Citing This Publication