Information about Asia and the Pacific Asia y el Pacífico
Chapter

8 Financial Liberalization and Money Demand in ASEAN Countries: Implications for Monetary Policy

Editor(s):
John Hicklin, David Robinson, and Anoop Singh
Published Date:
July 1997
Share
  • ShareShare
Information about Asia and the Pacific Asia y el Pacífico
Show Summary Details
Author(s)
Robert Dekle and Mahmood Pradhan 

Monetary developments in four of the member countries of the Association of South East Asian Nations (ASEAN)—Indonesia, Malaysia, Singapore, and Thailand—since the early 1980s must be assessed in the context of a remarkably successful economic performance that has contributed to the rapid development of domestic financial markets. The extent of financial liberalization—interest rate deregulation and greater competition in banking markets, as well as the liberalization of restrictions on cross-border capital flows—has been considerably greater than in many other developing countries. It would be surprising if these structural changes in financial markets and the associated rapid growth did not affect the relationship between money, economic activity, and inflation. In many industrial countries that underwent substantial episodes of financial deregulation and financial innovation during the early and mid-1980s, there were significant shifts in the orientation of monetary policies. Several countries found it difficult to retain intermediate targets and moved more toward explicit targets for final objectives, typically inflation.

This paper examines how financial market changes in the four countries have affected money demand behavior and seeks to draw the implications for monetary policy. The core of the paper assesses whether money demand equations are relatively stable and predictable—an important prerequisite for operating a policy framework centered on monetary targets. The results of this exercise caution against excessive reliance on monetary aggregates to gauge monetary conditions. Similar to the experience of many industrial countries, ongoing changes in financial markets suggest that policy actions need to be based on a wider set of monetary and real sector indicators. The paper also discusses—in the context of increasing integration of financial markets and substantial foreign capital inflows—the feasibility of alternative policy frameworks, including nominal exchange rate targets and inflation targets, although this topic does not stem directly from the empirical work on which this paper is based.

Impact of Financial Liberalization on Money Demand

Measures to promote competition among financial institutions will generally tend to lower transaction costs, and technological advances such as the introduction of automatic teller machines and credit cards may cause money demand to respond more rapidly to interest rate changes, thereby increasing the interest elasticity of money demand. More generally, measures that promote financial market development could result in the introduction and deepening of markets for new and more attractive assets, such as money market paper, stocks, and bonds. At the same time, they may cause gradual portfolio shifts away from monetary assets, possibly reducing the predictability of money demand. In practice, a failure to allow for changes in money demand following financial reform could result in monetary policy that is tighter or looser than is planned before the reforms are implemented.

The conventional money demand equation expresses the demand for real money balances (M/P) as a function of a scale variable, usually the level of real income (Y); and an opportunity cost variable, usually the rate of interest on an alternative asset i:

where ε is an error term representing money demand shocks. Instability of this error term will weaken the relationship between money holdings and income and interest rates. The potential instability in money demand will affect the coefficients, mainly b, and c, but also the intercept term a.

In the four countries considered here, financial liberalization since the mid-1970s has involved deregulating deposit rates and introducing or deepening alternative monetary instruments, bonds, and equities (Table 1). The liberalization of interest rates has been the most important feature of financial reform in these countries. Except in Singapore, real interest rates were sometimes negative before the reform, as in other previously financially “repressed” economies. In Indonesia, after the 1983 reform, time deposit rates more than doubled and real interest rates remained positive, even during subsequent high-inflation years. In Malaysia, deposit rates increased following the 1978 liberalization, ending the era of financial repression. Nominal and real rates increased markedly between 1988 and 1993, raising the differential between the money market rate and the London interbank offered rate (LIBOR) and inducing inflows of foreign capital. In Singapore, the liberalization of interest rates was complete by 1975, and the extremely open nature of the economy made it difficult for the government to pursue an independent monetary policy. The relatively low levels of both the nominal and real rates in Singapore during most of the 1980s mirrored U.S. interest rate trends. In Thailand, despite financial repression until the mid-1980s, real rates moved to positive levels from the early 1980s onward as inflation subsided. Until the 1989 liberalization measures, however, time deposit rates in Thailand moved in discrete steps as deposit rates were controlled by the authorities.

Table 1.Financial Liberalization
IndonesiaMalaysiaSingaporeThailand
Interest rate liberalizationControls on deposit and lending interest rates lifted in 1983.In 1978, deposit and lending rates liberalized. In the mid-1980s, lending rates of all banks pegged to the lending rates of the two “leading” banks. In 1991, lending rates again liberalized.Domestic interest rate cartel abolished, and deposit and lending rates liberalized in 1975.Ceilings on all time deposit rates removed in 1989-90, and those on lending rates removed in 1992.
Bank deregulation and competitionIn 1988, relaxation of entry requirements of domestic and joint-venture banks. Total number of banks rose from 111 in 1989 to about 240 in 1994, when authorities curtailed granting of new licenses.Deregulation since 1989 has removed barriers between different types of financial institutions and allowed finance companies to participate in the interbank market and merchant banks to issue nonnegotiable CDs; Labuan offshore center introduced in 1990; new entry of banks in domestic market remains restricted.Since late 1960s, free entry, subject to standards set by the Monetary Authority of Singapore (MAS). Today, highly competitive market with close to 150 domestic commercial banks and close to 40 foreign banks with full domestic privileges.Since late 1980s, liberalization of permissible activities and asset holding requirements of commercial banks. Now, commercial banks may hold a greater variety of assets and engage in activities such as trading securities and underwriting debt instruments. Entry of foreign banks through Bangkok International Banking Facilities (BIBF), an offshore banking center, liberalized in 1993.
Financial market developmentDeepening money markets since the mid-1980s introduction of Bank Indonesia paper (SBIs and SBPUs). Growing commercial paper market since the early 1990s. Small corporate bond market, and no government bond market. Rapid recent growth in the stock market, owing to improvement in market infrastructure and supervision of the Jakarta Stock Exchange by Bapepam, the regulatory agency.Since 1979, growing markets in CDs and bankers acceptances. Government bond market, although large, declining relative to GDP since 1988. Since the 1990 establishment of a credit rating agency, the corporate bond market has grown. Stock market capitalization relative to GDP highest among the four countries.Rapid growth in the money markets since 1975, as duties abolished on CDs, bills of exchange, and promissory notes. Large bond market dominated by Asian dollar bonds (98% of bond market capitalization); small domestic government bond issues mainly to absorb Central Provident Fund and Post Office Deposits. Stock market has grown rapidly since the 1973 delinking from the Malaysian stock exchange.Between 1979 and 1990, the market comprised mainly repos; since 1990, CDs, commercial bills, and promissory notes have grown. Traditionally, small outright trading in government bonds; corporate bond issuance severely restricted until 1992, but has since grown with introduction of rating bureau and Bond Dealers Club. Stock market boomed after establishment of Securities and Exchange Commission in 1992.
Management and supervisionIn early 1990s, imposition of new rules on capital adequacy and restrictions on commercial bank involvement in the equity and commerical paper markets.In 1989, the Banking and Financial Institutions Act placed all banking institutions under central bank supervision and strengthened prudential regulation based on the Basle limits capital framework.MAS sets minimal capital and licensing standards for banks. In early 1990s, following guidelines of the Bank for International Settlements (BIS) single customer lending set to 25% of a bank’s capital and a minimum Tier-I capital adequacy ratio of 12% imposed.In early 1990s, central bank applied BIS guidelines on asset quality and capital adequacy to both commercial banks and finance companies. It introduced measures to improve the quality of securities and finance companies by encouraging the merger of those companies that were not sufficiently competitive.
Capital account and opennessOpen capital account since late 1960s.Restrictions on capital inflows are limited (although restrictions on short-term inflows were temporarily introduced in 19941. Capital outflows above certain amounts are subject to central bank approval.Highly open capital account since 1978, when all foreign exchange controls abolished.In 1991, most restrictions on capital outflows were eliminated. The BIBF was established in 1993, providing foreign currency loans to domestic and foreign businesses.

Generally, in the four ASEAN countries, the liberalization of interest rates preceded the development of money and bond markets, although the money markets developed much faster than the bond markets. With the exception of Thailand, short-term money markets developed rapidly, soon after the liberalization of interest rates. In Thailand, the money market, comprising mostly repurchase agreements (repos), started to develop in 1979, a full decade before the liberalization of deposit interest rates.

The development of bond markets in the four countries has been hampered by strong government fiscal positions in Malaysia, Singapore, and Thailand, the “balanced-budget” rule in Indonesia, and, until recently, restrictions on corporate bond issues and the absence of bond rating agencies. In Indonesia, bond market development has also been hindered by the paucity of institutional investors. While still small, the Malaysian corporate bond market has grown since the establishment of a credit rating agency in 1990. The Singapore bond market is the largest in the region, but is dominated by foreign bonds—about 98 percent of the capitalization is in the form of Asian dollar bonds. In Thailand, corporate bond issuance was severely restricted until 1992, but has since grown with the establishment of a credit rating agency and the Bond Dealers Club.1

The development of the equity markets in these countries has been rapid and has closely tracked their impressive overall economic performance. The stock market in Malaysia has a long history, dating back over a hundred years, and market capitalization relative to GDP is the highest among the four countries. The stock exchange of Singapore was established in 1973, when it was formally delinked from the exchange in Malaysia. It has grown rapidly since that time and is now comparable in size to the major stock markets in the world. The Thai stock exchange-established in 1974—experienced only modest growth initially, but grew rapidly in the mid-1980s. In Indonesia, since the early 1990s, the improvement in market infrastructure and supervision of the Jakarta Stock Exchange by Bapepam, the regulatory agency, has aided the growth of the equity market, with market capitalization increasing from $81 million in 1986 to $67 billion at the end of 1995.

The financial market reforms and financial developments described above may change the velocity of broad money—in principle, in either direction.2 Reforms that increase the number of banks and spur institutional and technological advances, such as credit cards and electronic transfers of deposits or cash machines, can raise the velocity of broad and narrow money, as these developments make it easier to convert money into money substitutes. However, as noted by Bordo and Jonung (1987), in many developing countries, the velocity of broad money may decline over time because of the increasing monetization of the economy or financial deepening. Further more, as interest rates on time deposits are liberalized, private agents may shift their assets from currency and demand deposits to time deposits, raising the velocity of narrow money, but lowering the velocity of broad money.

For Indonesia, Malaysia, and Thailand, there has been a marked secular decline in the velocity of broad money (Figure 1). In Singapore, broad money velocity has declined since 1985, which is somewhat surprising, given the fall in both nominal and real time deposit rates and the boom in the Singapore Stock Exchange. The velocity of narrow money has been considerably more volatile, particularly in Indonesia and Thailand, although, except in Malaysia, there has not been a trend decline in the velocity of narrow money.

Figure 1.Velocity of Monetary Aggregates

Source: IMF, International Financial Statistics.

Evolving Monetary Policy Framework

Financial liberalization can affect the choice of targets of monetary policy and the variables that central banks monitor to gauge monetary conditions. In a developing financial market, interest rates tend to be set through administrative controls, and the central bank usually targets quantity variables such as broad money. Following financial liberalization, the stability of monetary aggregates may be reduced. Central banks presiding over relatively advanced financial markets often resort to monitoring price variables such as exchange and interest rates. In many industrial countries, broad money targets are effectively used as monitoring ranges, with very few central banks attempting to strictly adhere to monetary targets or to base policy actions entirely on deviations of actual money growth from projected growth.

In each of the four countries, the role of monetary targets in the conduct of monetary policy has been reduced in recent years. This process took place earliest in Singapore, which since the early 1980s has focused primarily on managing the exchange rate as its principal monetary instrument (previously, it monitored a variety of intermediate targets, including the monetary base, interest rates, and loan growth, as well as exchange rates).3 However, in recent years the process has also been apparent in the other three countries. In Malaysia, the emphasis of monetary policy shifted during the 1980s from Ml to M2 and then to M3; in recent years, policies have focused more on short-term interest rates, although money and credit aggregates are still monitored (Table 2). Similarly, the Bank of Thailand has shifted its policy emphasis from M2 toward commercial bank credit to the private sector and domestic interest rates (Tivakul, 1995). Indonesia, while continuing to set money and credit targets, has in practice given increased weight to interest rates and exchange rates.

Table 2.Monetary Policy in the 1990s
IndonesiaMalaysiaSingaporeThailand
Decision-making processThe authorities monitor broad money, credit aggregates, and reserve money. In addition, they monitor the real value of the rupiah against major trading partner currencies.The short-term operating target is the one-month interbank rate, while money and credit growth and the exchange rate are all monitored.The nominal exchange rate is managed to maintain low inflation. There are no money, credit, or interest rate targets.The baht is pegged to an undisclosed basket of currencies. The short-term operating target is the interbank rate; an overall target for private credit is set in the credit plan.
Main instrumentsOpen market operations involving SBIs, and SBPUs, reserve requirements, and foreign exchange operations.Reserve requirements, direct lending and borrowing from the interbank market, sales of Bank Negara Malaysia bills, shift of government and provident fund deposits to Bank Negara Malaysia.Foreign exchange operations.Repurchase operations, sales of Bank of Thailand bonds, and the credit plan.
Recent reforms in monetary instrumentsNo major changes since introduction of SBIs in February 1984 and SBPUs in February 1985.Introduction in 1993 of Bank Negara Malaysia bills.No major changes in recent years.Reintroduction in 1995 of Bank of Thailand bonds.

The shift away from formal monetary targeting has occurred for several reasons. In Singapore—and increasingly in a number of other countries as well—it has reflected the growing difficulties in seeking to target simultaneously the exchange rate and monetary aggregates in increasingly open economies.4 But it has also reflected concerns that the demand for money may have become more unstable as financial liberalization has accelerated and, thus, is a less reliable guide for policy formulation.

The instruments of monetary policy have depended on the maturity and depth of financial and capital markets and on the flexibility of interest rates. There has been greater reliance on open market operations to affect short-term interest rates as financial markets have developed and, in general, a move away from achieving broad money targets by limiting bank lending through moral suasion or changes in reserve requirements.

Since the 1980s, Indonesia, Malaysia, and Thailand have tried to introduce or intensify the use of open market operations (Table 2). The absence in the early 1980s of government debt instruments in these countries meant that the shift to open market operations was accompanied by the issuance of the central banks’ own debt instruments. To date, however, only Indonesia has a short-term paper market of sufficient depth to conduct traditional open market type operations. Normally, when tightening monetary conditions, Bank Negara Malaysia raises reserve requirements or borrows directly from the interbank market, and the Bank of Thailand sells repos or Bank of Thailand paper.

Singapore’s monetary policy, in contrast, is implemented through foreign exchange operations, with the Monetary Authority of Singapore (MAS) selling foreign exchange for Singapore dollars to achieve a steady appreciation of the nominal exchange rate. Although treasury bills are auctioned and yields are competitively determined, the MAS does not carryout traditional open market operations. Official exchange rate intervention is able to exert a stronger influence on the nominal exchange rate because various regulations, such as limits on bank lending in Singapore dollars, have prevented the Singapore dollar from being widely held by foreigners.

Empirical Estimates of Money Demand Equations

The estimation of money demand has a long history, but cointegration techniques have begun to be applied only recently. In conventional money demand equations, such as (1), if M/P, Y, and i are cointegrated, then, in the long run, movements in these variables will be closely related. If some shock drives the long-run relationship between money, real income, and the opportunity cost of money out of equilibrium, real money balances will adjust over time, such that these variables move together again. Thus, the existence of a cointegrating relation means that, in the long run, the economy will return to some stable relationship between money, income, and the opportunity cost of money. Without a proper understanding of the structural parameters of the long-run money demand equation, policymakers may react to an adverse shock to real income, for example, by easing monetary conditions excessively, leading to inflation that is higher than targeted.

To estimate long-run real money demand relationships, we use the Johansen (1988) full information maximum likelihood method. A necessary condition for the existence of a stable long-run relationship is that there be a cointegrating vector containing money, income, and interest rates. The test for this is whether the “maximal eigenvalue” or “trace eigenvalue” statistics from the Johansen procedure are above the relevant critical values, in which case we can reject the hypothesis of no cointegration. In principle, there may, of course, be more than one cointegrating vector between these variables. In such cases, given the issue that is of immediate interest, we focus only on the vector that has money on the left-hand side (normalized on M/P), although in practice, this problem did not arise in any of the countries considered here. Details of the estimation procedure are provided in Appendix I, and variable definitions and data sources are provided in Appendix II.

Results

The estimation results, presented in Table 3, by and large, do not provide strong evidence of stable relationships.5 We find stable demand equations with reasonable coefficients for real narrow and broad money only in Malaysia (and even here, coefficients on key variables are statistically insignificant). Overall, these results suggest that it is difficult to obtain stable real money demand functions using only the conventional determinants—real income and interest rates. Alternative specifications have not been explored because the focus here is on the relatively narrow question of whether there is a stable relationship between money, income, and interest rates that could provide the basis for a particular monetary policy framework.6

Table 3.Estimates of Real Money Demand Elasticities
IndonesiaMalaysiaSingaporeThailand
NarrowBroadNarrowBroadNarrowBroadNarrowBroad
GDP1.511.391.181.560.881.201.01.26
(4.98)(7.68*)(4.93*)(4.52*)(17.35*)(17.35*)(8.62*)(0.78)
Time deposit rate10.01-0.076-0.079
(0.32)(16.53*)(15.89*)
Call money-broad money return1-0.017-0.068-1.0
(4.34)(3.74)(7.2*)
Broad money return10.028
(6.21*)
Foreign return1-0.0033-0.021
(0.75)(0.75)
Dummy 19830.380.38
(2.39)(18.96*)
Dummy 1988-0.090.52
(-0.12)(21.88*)
1974–951976–951975–951978–95
Maximal eigenvalue statistic2, 318.131.221.4*17.618.923.219.513.0
Trace eigenvalue statistic2, 333.14625.632.6*22.635.726.420.1
Stable?NoNoYesYesNoNoNoNo
Note: Estimated by Johansen’s (1988) method, with one lag. All variables except for interest rates are in logarithms; χ2 tests for statistical significance are in parentheses; * denotes significance at 5 percent level.

Semielasticity.

Eigenvalue tests for the null hypothesis of no cointegrating vectors. The eigenvalue statistics are adjusted for degrees of freedom (Reimers, 1992).

Critical values for Indonesia, which include two dummy variables, are simulated. For the other countries, critical values are from Osterwald-Lenum (1992).

Note: Estimated by Johansen’s (1988) method, with one lag. All variables except for interest rates are in logarithms; χ2 tests for statistical significance are in parentheses; * denotes significance at 5 percent level.

Semielasticity.

Eigenvalue tests for the null hypothesis of no cointegrating vectors. The eigenvalue statistics are adjusted for degrees of freedom (Reimers, 1992).

Critical values for Indonesia, which include two dummy variables, are simulated. For the other countries, critical values are from Osterwald-Lenum (1992).

Summary of Nominal Money Demand Results

Before estimating real money demand, we estimated nominal money demand equations of the form M = a + bY + ci + dP + ε to test if the coefficient on the log price level (d) is equal to one. If d is unity (price homogeneity)—a doubling of the price level will double nominal money demand—this would then allow us to estimate the real demand for money. The results are shown in Appendix I. We reject the assumption that d is equal to one for nominal narrow money in Singapore and for nominal broad money in Malaysia, Singapore, and Thailand. However, to estimate real money demand, we impose the restriction that d equals one because the rejection could be a result of sample-specific factors.7 Over the long run, price illusion is unlikely to exist; rather, the rejection most likely reflects ongoing changes in financial markets and money-holding behavior among private sector agents.

Real Narrow Money

In Indonesia, Singapore, and Thailand, we are unable to find a stable relationship between real narrow money8 and its conventional determinants—real GDP, and an opportunity cost variable (typically the time deposit rate). This is perhaps not surprising for Indonesia and Thailand, which have experienced substantial financial reform since the 1980s. (In Indonesia, including dummy variables to capture the effects of the 1983 and 1988 financial liberalization episodes does not help in achieving stability.) For Singapore, the freeing of interest rates and other major reforms were almost completed by the beginning of our sample, 1975.9 Thus, the instability of narrow money demand is probably related more to the financial innovations that were common to all international financial centers in the 1980s—the greater use of credit cards, electronic transfers, and the introduction of mutual funds with checking accounts—enabling Singaporeans to economize on narrow money holdings. The difficulty in finding stable money demand functions in a number of industrial countries, such as Australia, the United Kingdom, and the United States, over the 1980s and the early 1990s is often attributed to similar, albeit more widespread, institutional and technological innovations. It is noteworthy that Malaysia, where reforms have been less extensive than in Indonesia and Thailand and where financial markets are less developed than in Singapore, is the only country of the four with a stable narrow money demand function.

Previous research on the stability of narrow money demand in the four ASEAN countries is limited, but, in general, has had more success in finding stability. The differences between the earlier work and the results reported here can be attributed partly to different sample periods and partly to differences in specification and estimation techniques. However, as explained in more detail in Appendix I, some previous studies have not corrected the test statistics for the small sample size and may therefore have erroneously rejected the null hypothesis of no cointegration. Using data only up to 1989, Tseng and Corker (1991) found that real narrow money demand equations were stable for Indonesia, Malaysia, and Singapore, but unstable for Thailand. On the basis of a very different specification, Hataiseree (1994) found that real narrow money, real income, and nominal interest rates were cointegrated for Thailand.10 Using estimation methods somewhat different from that adopted here, Price (1994) and Arize (1994) found stability for narrow money in Indonesia and Singapore.11

Real Broad Money

The estimated real broad money equations are unstable for Indonesia, Singapore, and Thailand, but not for Malaysia (Table 3).12 For Malaysia, the elasticity of real broad money with respect to real income is higher than that for real narrow money, but the opportunity cost semielasticity, although reasonable, is statistically insignificant.13 For Indonesia, Malaysia, and Thailand, we use the difference between the call money rate and the return on broad money as the opportunity cost of broad money.14 For Singapore, given the openness of its capital market, we include the foreign return along with the return on broad money. Although the coefficient estimates are all reasonable, we fail to achieve cointegration for Indonesia, Singapore, or Thailand.

It is somewhat surprising that the results for real broad money are not better than those for real narrow money. The freeing of time deposit rates should mainly cause a shift from one component of broad money to another, from narrow money to quasi-money. These instabilities in real broad money demands may therefore reflect the growth of money alternatives, such as stocks and money market instruments. Equity markets grew very rapidly in the 1980s in Indonesia and Thailand, and firms and individuals in these countries, as a result, may have changed their money holding behavior. In contrast, in Malaysia, the equity market was well entrenched by the beginning of our sample period.

Previous research on the stability of real broad money demand in these four countries is, again, limited. However, consistent with our results, the earlier work has had greater difficulty in finding stability for real broad money than for real narrow money. Among the four countries, using the period up to 1989, Tseng and Corker (1991) found broad money stability only for Indonesia. Hataiseree (1994) and Arize (1994), using specifications and estimation methods different from ours and those of Tseng and Corker, found stability for Thailand and Singapore.

Policy Implications and Conclusions

The empirical results of the previous section, although preliminary, have an important bearing on the feasibility of framing monetary policy around targets for monetary aggregates. Monetary targeting to control inflation depends on the stability and predictability of money demand. Only then can monetary authorities have a reasonable degree of confidence that, if actual money growth is above target, there is likely to be upward pressure on prices and, consequently, a need for some policy actions to tighten monetary conditions. If money demand behavior is not predictable, however, monetary authorities face the difficulty of not knowing whether “excess” money growth reflects an underlying shift in the private sector’s desire to hold money balances, or whether the actual money holding is temporarily above what private agents would wish to hold over the long term.

During the 1980s, many industrial countries, such as Canada, the United Kingdom, the United States, and a number of countries in continental Europe, faced similar policy dilemmas. Following deregulation of financial markets in the late 1970s and early 1980s, a number of these countries experienced rapid growth of financial markets that was also spurred by continuing advances in underlying transaction technologies. Money demand instability effectively implied that money growth rates were poor predictors of future inflation and output trends.15 The dilemma for policymakers is to decide to what extent policy actions should be constrained by preannounced targets. If money growth rates are not good leading indicators of future inflation, then it may be preferable to abandon them as intermediate targets or, as many industrial countries did, to downgrade them to one of a set of variables that policymakers monitor regularly.16

Some of the ASEAN countries that are currently faced with similar uncertainty regarding money growth must make the potentially difficult judgment about how much emphasis to place on intermediate targets. If money targets are announced, but policy actions are not seen to be based on money growth because specific episodes of “excess” money growth are judged not to indicate inflation pressures, there is a risk that the credibility of policies could be undermined. Against this concern, policymakers must also weigh the reduced effectiveness of policies if money growth and inflation are not closely related. Indeed, as discussed in the section Evolving Monetary Policy Framework, a number of countries have reduced the emphasis on strictly adhering to monetary targets.

But moving away from a framework based on monetary targets raises the question of whether there is an alternative yardstick by which monetary conditions can be assessed. If there is no single variable that can be used as an intermediate target, either because of an unstable relationship with economic activity or because those that are closely related to the state of the economy cannot be directly influenced by central bank actions, then the assessment of monetary conditions and policy actions will necessarily be based on monitoring a range of indicators. In practice, of course, all central banks monitor a wide set of variables, including some real sector variables that can be influenced only indirectly. The challenge for policymakers is to ensure that, in the absence of an explicit intermediate target, the central bank’s resolve to maintain low inflation continues to be viewed as credible. When the assessment of monetary conditions is based on a range of indicators, there is always a risk that policy inaction will be seen as a weakening in the anti-inflation stance. While policies must demonstrate consistency, transparency of the monetary policy decision-making process is also important to provide more information to market participants about the rationale for policy actions.

It is sometimes argued that if countries cannot pursue money-based disinflation strategies, they can simplify the operation of monetary policy by adopting an exchange rate target. In terms of the decision-making process, an exchange rate target is perhaps the most simple to operate; central banks are only required to maintain a fixed rate with respect to either a basket of trading partner currencies or a single major foreign currency. However, the benefits of fixed exchange rates are strongest for countries that lack credibility and have a history of relatively high inflation. In the four ASEAN countries in our study, which have maintained macroeconomic stability and relatively low inflation rates, the benefits of a pegged exchange rate in terms of stability must be weighed against other considerations, including the greater difficulties of managing capital inflows and other real shocks. In particular, reduced monetary autonomy weakens the ability to control inflation.

An alternative approach, when there is no suitable intermediate target variable that can be predictably influenced by policy and that has a close relationship with inflation, is to target inflation directly. The instruments available to the monetary authorities and the objectives of monetary policy—maintaining low inflation—are exactly the same under both approaches. Indeed, when the operation of monetary policy in countries that have explicit inflation targets is compared with that in countries that frame policy decisions around intermediate targets, the difference may be more semantic than economic. Policy objectives are in most cases specified in terms of price stability.

If targeting inflation is not very different from targeting intermediate variables, what then are the benefits of moving toward explicit inflation targets, and, with respect to the ASEAN countries, what would be required to pursue this approach? The empirical analysis of money demand behavior in this paper establishes only that strict adherence to intermediate targets on monetary aggregates may not be desirable in these four countries, but does not provide enough evidence to determine whether inflation targets would be beneficial and also whether they would be feasible in these countries. Further research on the underlying determinants and variability of inflation would be necessary to address these issues. Nevertheless, it is helpful to consider the benefits of inflation targeting in a general context and the key ingredients of this approach.

In countries that have adopted inflation targets, formulating explicit medium-term price objectives has helped fill an important gap following the abandonment of monetary targets and, in some cases, exchange rate targets. When monetary policy assessments are based on a range of indicators, or when policy is framed around intermediate targets, there may be a tendency for the policy framework to lack an explicit forward looking element. Moreover, private agents may find it difficult to gauge the policy stance when the authorities’ actions are based on a complex feedback rule. This is essentially a presentational problem that can undermine credibility. Inflation targets help get around this presentational problem by forcing the authorities to base policy actions on their forward-looking assessment of inflation. Furthermore, central banks find it easier to justify’ policy actions by making public their assessment of future inflation, which also enhances the credibility of policies.

Adopting inflation targets is not, however, costless. By definition, forward-looking assessments are subject to wide margins of uncertainty. Forecasting errors in inflation projections typically tend to be relatively large.17 This in effect gives rise to a trade-off between flexibility and credibility. To ensure that targets are met, central banks may define a relatively large inflation target band, but this approach will not enhance the credibility of policies. Thus, with respect to Indonesia, Malaysia, Singapore, and Thailand, an important prerequisite is to model the inflation process and to examine both the magnitude and the source of forecast errors. If, for example, an economy is subject to frequent supply or structural shocks, actual inflation may deviate significantly from the target range. In such circumstances, although it may not be appropriate for monetary authorities to tighten conditions, they could lose credibility as a result. Although the economic structure of the four ASEAN countries is diverse, sector-specific shocks may still have strong economywide effects, and these issues need to be investigated in greater detail to establish the desirability of an explicit inflation-targeting approach.

Appendix I. Money Demand Estimation

The estimates reported in Table 3 in the text are derived from the Johansen maximum likelihood tests for cointegration between money, prices, real income, and interest rates. To ascertain the order of integration of these variables (i.e., to determine whether the levels are stationary or whether their first differences are stationary), Table 4 presents Augmented Dickey-Fuller (Dickey and Fuller, 1979) statistics for unit root tests on log levels of money, prices, and income, and on the level of interest rates. These test statistics suggest that most of these variables are integrated of order one (I(1)), although for some variables—such as narrow money and interest returns on broad money in Indonesia and real GDP and the foreign rate of return in Thailand—the ADF statistics indicate that their first differences are not stationary. However, some of these time-series properties are likely to reflect the relatively small sample period; it is difficult to accept in an economic sense that these variables would be I(2) in the long run. Moreover, univariate tests of this kind are typically of low power compared with stationary alternatives. The analysis in this paper, therefore, treats all variables as I(1).18

Table 4.Testing for a Unit Root: ADF Statistics
LGDPTIMERETCMRFORLRNMLRBMLNMLBMLCPI
Indonesia
Order of integration
I(1)-1.40-2.48-1.45-2.79-1.27-2.75-1.86-2.41-3.32-4.03*
I(2)-4.49*-3.84*-3.061-4.32*-3.13*-4.02*-3.78*-2.961-4.04*-3.11*
Malaysia
Order of integration
I(1)-1.79-2.24-1.91-3.81*-1.76-0.97-1.62-1.90-0.76-2.61
1(2)-4.37*-4.00*-4.41*-4.62*-3.78*-4.00*-3.82*-3.77*-3.632-4.28*
Singapore
Order of integration
I(1)-2.31-1.83-3.07-2.17-2.65-2.56-2.96-2.38
I(2)-3.67*-3.273-4.05*-5.35*-4.18*-5.48*-4.44*4.51*
Thailand
Order of integration
I(1)-2.28-2.92-2.64-2.69-1.72-2.46-1.83-2.08-3.26-2.42
I(2)-2.984-2.85*5-4.26*-3.425-2.586-3.411-3.352-3.632-4.50*-3.72*
Notes: * denotes rejection at the 5 percent level. LGDP is log of real GDP, TIME is time deposit rate, RET is broad money return, CMR is call money or other money market return, FOR is foreign interest rate (LIBOR plus expected currency appreciation), LRNM is log of real narrow money, LRBM is log of real broad money, LNM is log of nominal narrow money, LBM is log of nominal broad money, and LCPI is log of consumer price index. The stationarity tests included a constant, a trend term, and up to four lags. For any variable x the ADF statistic tests the null hypothesis of a unit root in x (order of I(1)) against the alternative of a stationary root. For a null order of I(2), the ADF statistic tests for a unit root in the first difference of x.

Critical value is -3.63.

Critical value is -3.69.

Critical value is -3.79.

Critical value is -3.66.

Critical value is -3.83.

Critical value is -3.29.

Notes: * denotes rejection at the 5 percent level. LGDP is log of real GDP, TIME is time deposit rate, RET is broad money return, CMR is call money or other money market return, FOR is foreign interest rate (LIBOR plus expected currency appreciation), LRNM is log of real narrow money, LRBM is log of real broad money, LNM is log of nominal narrow money, LBM is log of nominal broad money, and LCPI is log of consumer price index. The stationarity tests included a constant, a trend term, and up to four lags. For any variable x the ADF statistic tests the null hypothesis of a unit root in x (order of I(1)) against the alternative of a stationary root. For a null order of I(2), the ADF statistic tests for a unit root in the first difference of x.

Critical value is -3.63.

Critical value is -3.69.

Critical value is -3.79.

Critical value is -3.66.

Critical value is -3.83.

Critical value is -3.29.

Cointegration

Tables 58 report the estimates and the associated test statistics for cointegration between money, prices, real income, and the opportunity cost variables for Indonesia, Malaysia, Singapore, and Thailand. The number of cointegrating vectors (r), is determined by two likelihood ratio tests. In the first test, based on the maximal eigenvalue, the null hypothesis is that there are at most r cointegrating vectors against the alternative of r + 1 cointegrating vectors. The second test is based on the trace of the stochastic matrix where the null hypothesis is that there are at most r cointegrating vectors against the alternative hypothesis that there are r or more cointegrating vectors.

Table 5.Indonesia: Cointegration Analysis of Money Demand, 1974–95
I.Nominal Narrow Money
Hypothesesr=0r≤1n≤2r≤3
L(max)25.017.711.33.6
95% critical value57.236.918.37.1
L(trace)57.732.614.93.6
95% critical value99.555.323.07.1
Coefficients
LGDPTIMELCPI1983D1988D
13.40.211.62.1-2.3
Weak exogeneity test statistics
LGDPTIMELCPI
χ2(1)0.71.210.7*
Statistics for testing the significance of a given variable
LGDPTIMELCPI1983D1988D
χ2(1)1.13.82.010.760.08
Statistic for testing whether coefficient on LCPI = 1
χ2(1)0.25
II.Real Narrow Money
Hypothesesr=0r≤1r≤2
L(max)18.110.64.9
95% critical value36.918.37.1
L(trace)33.115.04.5
95% critical value55.323.07.1
Coefficients
LGDPTIME1983D1988D
1.510.0110.38-0.09
Weak exogeneity test statistics
LGDPTIME
χ2(3)0.08*3.73*
Statistics for testing the significance of a given variable
LGDPTIME1983D1988D
χ2(3)4.980.322.390.12
III.Nominal Broad Money
Hypothesesr=0r≤1r≤2r≤3
L(max)35.621.39.57.9
95% critical value57.236.918.37.1
L(trace)74.338.717.57.9
95% critical value99.555.323.07.1
Coefficients
LGDPCMR-RETLCPI1983D1988D
0.760.0360.790.831.01
Weak exogeneity test statistics
LGDPCMR-RETLCPI
χ2(1)2.401.6710.51
Statistics for testing the significance of a given variable
LGDPCMR-RETLCPI1983D1988D
χ2(1)7.17*6.0812.48*19.10*20.27*
Statistic for testing whether coefficient on LCPI = 1
χ2(1)0.83
IV.Real Broad Money
Hypothesesr=0r≤1r≤2
L(max)31.28.92.9
95% critical value36.918.37.1
L(trace)46.014.92.9
95% critical value55.323.07.1
Coefficients
LGDPCMR-RET1983D1988D
1.39-0.0170.380.52
Weak exogeneity test statistics
LGDPCMR-RET
χ2(1)0.093.45
Statistics for testing the significance of a given variable
LGDPCMR-RET1983D1988D
χ2(1)7.68*4.3418.96*21.88*
V.Real Broad Money (foreign interest rate)
Hypothesesr=0r≤1r≤2
L(max)25.011.22.6
95% critical value36.918.37.1
L(trace)38.913.82.6
95% critical value55.323.07.1
Coefficients
LGDPFOR-RET1983D1988D
1.16-0.190.800.32
Weak exogeneity test statistics
LGDPFOR-RET
χ2(1)12.27*0.0004
Statistics for testing the significance of a given variable
LGDPFOR-RETFOR1983D1988D
χ2(1)7.98*10.35*21.43*20.81*6.27
Notes: The vector autoregression includes one lag on each variable. The statistics L(max) and L(trace) are Johansen’s maximal eigenvalue and trace eigenvalue statistics for cointegration. The weak exogeneity and significance test statistics are evaluated under the assumption that rank = n and are therefore asymptotically distributed as χ2(n). Critical values are simulated, because standard tables do not exist in the presence of step dummy variables. * denotes significance at the 5 percent level. All monetary aggregates are expressed in logs. 1983D and 1988D are dummy variables that take on values of unity after 1983 and 1988.
Notes: The vector autoregression includes one lag on each variable. The statistics L(max) and L(trace) are Johansen’s maximal eigenvalue and trace eigenvalue statistics for cointegration. The weak exogeneity and significance test statistics are evaluated under the assumption that rank = n and are therefore asymptotically distributed as χ2(n). Critical values are simulated, because standard tables do not exist in the presence of step dummy variables. * denotes significance at the 5 percent level. All monetary aggregates are expressed in logs. 1983D and 1988D are dummy variables that take on values of unity after 1983 and 1988.
Table 6.Malaysia: Cointegration Analysis of Money Demand, 1976–95
I.Nominal Narrow Money
Hypothesesr=0r≤1r≤2r≤3
L(max)20.416.65.51.4
95% critical value27.121.014.13.8
L(trace)44.023.67.01.4
95% critical value47.229.715.43.8
Coefficients
LGDPTIMELCPI
1.66-0.100.45
Weak exogeneity test statistics
LGDPTIMELCPI
χ2(1)4.79*2.460.0004
Statistics for testing the significance of a given variable
LGDPTIMELCPI
χ2(1)1.14.62*0.03
Statistic for testing whether coefficient on LCPI = 1
χ2(1)0.25
II.Real Narrow Money
Hypothesesr=0r≤1r≤2
L(max)21.4*2.41.7
95% critical value21.014.13.8
L(trace)25.64.11.7
95% critical value29.715.43.8
Coefficients
LGDPTIME
1.18-0.076
Weak exogeneity test statistics
LGDPTIME
χ2(1)20.71*6.8*
Statistics for testing the significance of a given variable
LGDPTIME
χ2(1)4.93*16.53*
III.Nominal Broad Money
Hypothesesr=0r≤1r≤2r≤3
L(max)29.7*15.35.70.03
95% critical value27.121.014.13.8
L(trace)50.8*21.15.80.03
95% critical value47.229.715.43.8
Coefficients
LGDPCMR-RETLCPI
0.35-0.073.29
Weak exogeneity test statistics
LGDPCMR-RETLCPI
χ2(1)2.43.12.72
Statistics for testing the significance of a given variable
LGDPCMR-RETLCPI
χ2(1)1.416.05*11.91*
Statistic for testing whether coefficient on LCPI = 1
χ2(1)11.4*
IV.Real Broad Money
Hypothesesr=0r≤1r≤2
L(max)17.612.22.8
95% critical value21.014.13.8
L(trace)32.6*15.02.8
95% critical value29.715.43.8
Coefficients
LGDPCMR-RET
1.56-0.068
Weak exogeneity test statistics
LGDPCMR-RET
χ2(1)4.52*3.32
Statistics for testing the significance of a given variable
LGDPCMR-RET
χ2(1)1.693.74
V.Real Broad Money (foreign interest rate)
Hypothesesr=0r≤1r≤2r≤3
L(mix)22.213.211.30.3
95% critical value27.121.014.129.7
L(trace)47.124.511.60.3
95% critical value47.229.715.43.8
Coefficients
LGDPRETFOR
1.55-0.0044-0.0092
Weak exogeneity test statistics
LGDPRETFOR
χ2(1)2.441.450.05
Statistics for testing the significance of a given variable
LGDPRETFOR
χ2(1)9.38*0.072.25
Notes: The vector autoregression includes one lag on each variable. The statistics L(max) and L(trace) are Johansen’s maximal eigenvalue and trace eigenvalue statistics for cointegration, adjusted for degrees of freedom (Reimers, 1992). The weak exogeneity and significance test statistics are evaluated under the assumption that rank = n and are therefore asymptotically distributed as χ2(n). Critical values are from Osterwald-Lenum (1992). * denotes significance at the 5 percent level. All monetary aggregates are expressed in logs.
Notes: The vector autoregression includes one lag on each variable. The statistics L(max) and L(trace) are Johansen’s maximal eigenvalue and trace eigenvalue statistics for cointegration, adjusted for degrees of freedom (Reimers, 1992). The weak exogeneity and significance test statistics are evaluated under the assumption that rank = n and are therefore asymptotically distributed as χ2(n). Critical values are from Osterwald-Lenum (1992). * denotes significance at the 5 percent level. All monetary aggregates are expressed in logs.
Table 7.Singapore: Cointegration Analysis of Money Demand, 1975–95
I.Nominal Narrow Money
Hypothesesr=0r≤1r≤2r≤3
L(max)30.8*10.74.70.0028
95% critical value27.121.014.13.8
L(trace)46.315.54.80.0028
95% critical value47.229.715.43.8
Coefficients
LGDPFORLCPI
0.62-0.0171.62
Weak exogeneity test statistics
LGDPFORLCPI
χ2(1)3.95*0.0887.56*
Statistics for testing the significance of a given variable
LGDPFORLCPI
χ2(1)7.90*11.34*22.63*
Statistic for testing whether coefficient on LCPI = 1
χ2(1)15.06*
II.Real Narrow Money
Hypothesesr=0r≤1r≤2
L(max)18.93.60.0055
95% critical value21.014.13.8
L(trace)22.63.70.0055
95% critical value29.715.43.8
Coefficients
LGDPFOR
0.88-0.0033
Weak exogeneity test statistics
LGDPFOR
χ2(1)5.17*0.85
Statistics for testing the significance of a given variable
LGDPFOR
χ2(1)17.35*0.75
III.Nominal Broad Money
Hypothesesr=0r≤1r≤2r≤3r≤4
L(max)37.9*23.99.98.10.1
95% critical value33.527.121.014.13.8
L(trace)79.9*42.018.18.20.1
95% critical value68.547.229.715.43.8
Coefficients
LGDPRETFORLCPI
5.42-0.520.13-9.50
Weak exogeneity test statistics
LGDPFORRETLCPI
χ2(1)0.687.62*0.4226.8*
Statistics for testing the significance of a given variable
LGDPFORRETLCPI
χ2(1)0.6731.7*10.9*7.56*
Statistic for testing whether coefficient on LCPI = 1
χ2(1)26.75*
IV.Real Broad Money
Hypothesesr=0r≤1r≤2r≤3
L(max)23.27.94.30.1
95% critical value27.121.014.13.8
L(trace)35.712.44.50.1
95% critical value47.229.715.43.8
Coefficients
LGDPRETFOR
1.20.028-0.021
Weak exogeneity test statistics
LGDPRETFOR
χ2(1)5.17*1.200.85
Statistics for testing the significance of a given variable
LGDPRETFOR
χ2(1)17.35*6.21*0.75
Notes: The vector autoregression includes one lag on each variable. The statistics L(max) and L(trace) are Johansen’s maximal eigenvalue and trace eigenvalue statistics for cointegration, adjusted for degrees of freedom (Reimers, 1992). The weak exogeneity and significance test statistics are evaluated under the assumption that rank = n and are therefore asymptotically distributed as χ2(n). Critical values are from Osterwald-Lenum (1992). * denotes significance at the 5 percent level. All monetary aggregates are expressed in logs.
Notes: The vector autoregression includes one lag on each variable. The statistics L(max) and L(trace) are Johansen’s maximal eigenvalue and trace eigenvalue statistics for cointegration, adjusted for degrees of freedom (Reimers, 1992). The weak exogeneity and significance test statistics are evaluated under the assumption that rank = n and are therefore asymptotically distributed as χ2(n). Critical values are from Osterwald-Lenum (1992). * denotes significance at the 5 percent level. All monetary aggregates are expressed in logs.
Table 8.Thailand: Cointegration Analysis of Money Demand, 1978–95
I.Nominal Narrow Money
Hypothesesr=0r≤1r≤2r≤3
L(max)22.318.26.10.74
95% critical value27.121.014.13.8
L(trace)47.4*25.16.90.74
95% critical value47.229.715.43.8
Coefficients
LGDPTIMELCPI
1.130.00930.67
Weak exogeneity test statistics
LGDPTIMELCPI
χ2(1)5.28*2.620.026
Statistics for testing the significance of a given variable
LGDPTIMELCPI
χ2(1)4.26*0.243.98*
Statistic for testing whether coefficient on LCPI = 1
χ2(1)1.63
II.Real Narrow Money
Hypothesesr=0r≤1r≤2
L(max)19.56.00.86
95% critical value21.014.13.8
L(trace)26.46.90.86
95% critical value29.715.43.8
Coefficients
LGDPTIME
1.00-0.079
Weak exogeneity test statistics
LGDPTIME
χ2(1)5.78*0.88
Statistics for testing the significance of a given variable
LGDPTIME
χ2(1)8.62*15.89*
III.Nominal Broad Money
Hypothesesr=0r≤1r≤2r≤3
L(max)25.313.010.00.22
95% critical value27.121.014.13.8
L(trace)18.523.310.30.22
95% critical value47.229.715.43.8
Coefficients
LGDPCMR-RETLCPI
1.560.361.06
Weak exogeneity test statistics
LGDPCMR-RETLCPI
χ2(1)4.68.12*8.65*
Statistics for testing the significance of a given variable
LGDPCMR-RETLCPI
χ2(1)6.55*6.31*7.37*
Statistic for testing whether coefficient on LCPI = 1
χ2(1)3.49
IV.Real Broad Money
Hypothesesr=0r≤1r≤2
L(max)13.06.80.4
95% critical value21.014.13.8
L(trace)20.17.10.4
95% critical value29.715.43.8
Coefficients
LGDPCMR-RET
1.26-1.00
Weak exogeneity test statistics
LGDPCMR-RET
χ2(1)7.46*1.00
Statistics for testing the significance of a given variable
LGDPCMR-RET
χ2(1)0.787.2*
V.Real Broad Money (foreign interest rate)
Hypothesesr=0r≤1r≤2r≤3
L(max)19.012.39.50.24
95% critical value27.121.014.13.8
L(trace)41.122.19.80.24
95% critical value47.229.715.43.8
Coefficients
LGDPRETFOR
1.090.23-0.035
Weak exogeneity test statistics
LGDPRETFOR
χ2(1)2.013.610.05
Statistics for testing the significance of a given variable
LGDPRETFOR
χ2(1)0.316.10*0.63
Notes: The vector autoregression includes one lag on each variable. The statistics L(max) and L(trace) are Johansen’s maximal eigenvalue and trace eigenvalue statistics for cointegration, adjusted for degrees of freedom (Reimers, 1992). The weak exogeneity and significance test statistics are evaluated under the assumption that rank = n and are therefore asymptotically distributed as χ2(n). Critical values are from Osterwald-Lenum (1992). * denotes significance at the 5 percent level. All monetary aggregates are expressed in logs.
Notes: The vector autoregression includes one lag on each variable. The statistics L(max) and L(trace) are Johansen’s maximal eigenvalue and trace eigenvalue statistics for cointegration, adjusted for degrees of freedom (Reimers, 1992). The weak exogeneity and significance test statistics are evaluated under the assumption that rank = n and are therefore asymptotically distributed as χ2(n). Critical values are from Osterwald-Lenum (1992). * denotes significance at the 5 percent level. All monetary aggregates are expressed in logs.

The critical values for the trace and maximal eigenvalue statistics are from Osterwald-Lenum (1992), except for Indonesia (see below). Miyao (1996) shows, on the basis of simulations of U.S. money demand equations, that there are substantial size distortions in the Johansen (1988) procedure. Using conventional critical values, the Johansen test tends to reject the null hypothesis of no cointegration too often (Reimers, 1992). To partly address this size problem, we apply a simple small sample correction to the eigenvalues by multiplying both eigenvalue statistics by T - n * m, instead of T, where T is the sample size, n is the number of endogenous variables, and m is the number of lags (m = 1 in all the estimates reported in Tables 58).

For Indonesia, two (0,1) dummy variables are included to capture the effects of the major financial reforms in 1983 and in 1988. The corresponding critical values are simulated because published critical values are not available for the Johansen procedure when the estimation includes dummy variables.19 These critical values are simulated through the following sequence. First, 22 random observations—equal to the sample size—are simulated, corresponding to each of our endogenous variables. These variables are regressed on a constant and two dummy variables, and the residuals from this regression are used to form the sample moments that asymptotically converge to the standard Wiener processes involved in the expressions for the Johansen procedure. Using these expressions, we form the approximate limiting distributions of the maximal and trace eigenvalue statistics; 10,000 replications are generated to approximate the limiting distribution from which we can find the 5 percent critical values.

With the above small sample eigenvalue corrections and critical values, it is more difficult to reject the null of no cointegration. The estimates of the long-run cointegrating vector are reported in Tables 58, including in cases where we cannot find cointegration. It should be noted, however, that a number of other studies using the Johansen procedure with limited samples do not correct the critical values for the sample size, and this may lead to rejecting the null hypothesis (no cointegrating vectors) too often. If, as in earlier studies, we use conventional asymptotic critical values, the null of no cointegration can be rejected far more often in the present estimates as well. However, this procedure is clearly not valid, and we therefore conclude that in most cases conventional money demand equations do not cointegrate.

Our results imply that the null hypothesis of no cointegration can be rejected only for Malaysia. Although some nominal money demand equations do cointegrate, equations with real money on the left-hand side do not, and, moreover, a number of coefficients have the wrong sign, suggesting that these behavioral relations are poorly determined. As a result, for most countries, it would not be valid to proceed further either with testing for exogeneity of right-hand-side variables or with modeling the short-run adjustment processes.

In Indonesia, the null of no cointegration cannot be rejected for any of the money demand specifications, including when the differential between the foreign return and the domestic return on broad money (FOR-RET) is used as the opportunity cost variable.

In Malaysia, we can reject the null hypothesis of no cointegration for real narrow money, nominal broad money, and real broad money. For real narrow money, the coefficients on LGDP and TIME are significant and of reasonable magnitude—a 1 percent increase in GDP raises the demand for real narrow money by 1.18 percent. For nominal broad money, surprisingly, the constraint of unit price homogeneity is rejected, and the coefficient on LGDP is well below unity. In the real broad money equation, although the coefficients are reasonable, none is statistically significant. When the foreign interest rate is substituted for the money market rate, it has the predicted negative effect, but the own rate does not exert the predicted positive effect. Moreover, in both the narrow money and broad money equations, income and interest rate variables are not weakly exogenous, suggesting that the relationship between these variables may not be unidirectional—in a statistical sense, movements in income, for example, may lead to movements in monetary aggregates.

In Singapore, only the equations for nominal narrow and broad money cointegrate. For nominal narrow money, the coefficient on the foreign interest rate appears reasonable, while the coefficient on real GDP is rather small. The hypothesis that the coefficient on LCPI is unity can be rejected at a very high level of significance. For nominal broad money, the coefficients are not plausible—both the domestic and the foreign returns have the wrong signs.

In Thailand, only the equation for nominal narrow money cointegrates with plausible coefficient signs and magnitudes, and the test for unit elasticity on LCPI cannot be rejected.

Appendix II. Data Construction and Sources

The opportunity cost of holding narrow money is proxied by the rate of return on time deposits, while the opportunity cost of holding broad money is proxied by the money market rate less the time deposit rate weighted by the share of quasi-money in broad money.

For Singapore, where domestic residents have access to a large Eurodollar market, the opportunity cost of narrow and broad money is proxied by the three-month dollar LIBOR minus (plus) the expected depreciation of the Singapore dollar vis-à-vis the U.S. dollar. The expected rate of exchange rate depreciation is proxied by the five-year moving average of actual exchange rate changes. For consistency, similar foreign interest rate variables are included in the empirical money demand equations for the other ASEAN countries.

All data are taken from the IMF’s International Financial Statistics. Interest rate data are from lines 60b and 601. Data on narrow money and on broad money (quasi-money), with the exception of Indonesia since 1988, are from International Financial Statistics, lines 34 and 35. For Indonesia, post-1988 data on monetary aggregates are from Bank Indonesia. Data on nominal and real GDP and consumer price indices for all countries are from International Financial Statistics, lines 99b, 99b.p, and 64, respectively.

References

    ArizeA.C.1994“A Re-examination of the Demand for Money in Small Developing Economies,”Applied EconomicsVol. 26 (March) pp. 21728.

    • Search Google Scholar
    • Export Citation

    BordoMichael and LarsJonung1987The Long-Run Behavior of the Velocity of Money: The International Evidence (Cambridge, United Kingdom; New-York: Cambridge University Press).

    • Search Google Scholar
    • Export Citation

    DickeyDavid and Wayne A.Fuller1979“Distribution of the Estimates for Autoregressive Time Series with a Unit Root,”Journal of the American Statistical AssociationVol. 74 (July) pp. 42731.

    • Search Google Scholar
    • Export Citation

    EricssonNeil and SunilSharma1996“Broad Money Demand and Financial Liberalization in Greece,”IMF Working Paper 96/62 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

    FisherPaulSuzanneHudson and MahmoodPradhan1993“Divisia Measures of Money,”Bank of England Quarterly BulletinVol. 33 (May) pp. 24055.

    • Search Google Scholar
    • Export Citation

    GoodhartCharles1989“The Conduct of Monetary Policy,”Economic JournalVol. 99 (June) pp. 293346.

    HataisereeRungsun1994“The Relationship Between Monetary Aggregates and Economic Activities: Some Thai Evidence Using a Cointegration Approach,”Papers on Policy Analysis and Assessment No. 2537 (Bangkok: Bank of Thailand).

    • Search Google Scholar
    • Export Citation

    JohansenSoren1988“Statistical Analysis of Cointegration Vectors,”Journal of Economic Dynamics and ControlVol. 12 (June) pp. 23154.

    • Search Google Scholar
    • Export Citation

    MiyaoRyuzo1996“Does a Cointegrating M2 Demand Relation Really Exist in the United States,”Journal of Money Credit and BankingVol. 28 (August) pp. 36580.

    • Search Google Scholar
    • Export Citation

    Osterwald-LenumMichael1992“A Note with Quantiles of the Asymptotic Distribution of the Maximum Likelihood Cointegration Rank Test Statistics,”Oxford Bulletin of Economics and StatisticsVol. 54 (No. 3) pp. 46172.

    • Search Google Scholar
    • Export Citation

    PillHuw and MahmoodPradhan1994“Monetary Aggregation: A Reconciliation of Theory and Central Bank Practice,”IMF Working Paper 94/188 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

    PriceSimon1994“The Demand for Indonesian Narrow Money: Long-Run Equilibrium, Error Correction, and Forward-Looking Behavior,”Journal of International Trade and Economic DevelopmentVol. 3 (July) pp. 14763.

    • Search Google Scholar
    • Export Citation

    ReimersH.-E.1992“Comparisons of Tests for Multivariate Cointegration,”Statistical PapersVol. 33 pp. 33559.

    StevensGlenn and GuyDebelle1995“Monetary Policy Goals for Inflation in Australia,” in Targeting Inflation, a Conference of Central Banks on the Use of Inflation Targets Organised by the Bank of England, 9–10 March 1995ed. by A.Haldane (London: Bank of England).

    • Search Google Scholar
    • Export Citation

    TivakulAroonsri1995“Globalization of Financial Markets in Thailand and Their Implications for Monetary Stability,”paper presented at the SEACEN-IMF Seminar ManilaMay 19–21.

    • Search Google Scholar
    • Export Citation

    TsengWanda and RobertCorker1991Financial Liberalization Money Demand and Monetary Policy in Asian Countries IMF Occasional Paper No. 84 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
Note: Robert Dekle is an Economist, and Mahmood Pradhan is a Senior Economist, in the IMF’s Asia and Pacific Department. The authors thank John Hicklin and David Robinson for extensive advice and help on earlier drafts of the paper. They are also extremely grateful to John McDermort for help with the empirical section of the paper.
1For a more detailed discussion of the factors that contributed to the development of bond markets in Thailand, see Chapter 9 in this volume.
2The velocity of money is defined as nominal income divided by the quantity of nominal money.
3The shift in policies in Singapore also reflected recognition of the significant role of the exchange rate in a small open economy.
4Indonesia widened its exchange rate band to 8 percent in September 1996, thereby increasing monetary autonomy.
5All money demand equations in this paper are estimated on annual data.
6There may, of course, be a stable relationship between money and some components of consumer price indices.
7There are several reasons why the statistical tests may reject unit price homogeneity over our sample period. First, as an economy grows, the basket of goods in the consumer price index (CPI) may become less relevant for firms and households that are increasing their broad money holdings, and second, technological progress may have changed the relationship between nominal money and prices.
8Real narrow money is defined as currency plus demand deposits divided by the CPI.
9In Singapore, where domestic residents have substantial scope for investing in dollar-denominated assets, instead of the time deposit rate, as the opportunity cost variable, we include a variable that represents the rate of return that domestic residents can earn on dollar assets. Dollar asset returns are approximated as LIBOR minus the expected depreciation of the U.S. dollar against the Singapore dollar.
10Hataiseree appended a goods market equation (the investment-saving relation) to the money demand equation.
11Both studies used the error-correction specification to model the short-run dynamics. We would have pursued a similar procedure had we been more successful in finding stable long-run relationships.
12Real broad money is defined as nominal broad money (narrow money plus quasimoney, time and saving deposits) divided by the CPI.
13The demand for broad money depends on the desire to hold money as an asset, in addition to holding it for transaction purposes. Given that wealthier agents accumulate more assets, we would expect the elasticity for broad money to be higher than that for narrow money.
14The return on broad money is equal to the time deposit rate times the share of quasimoney in broad money.
15For an extensive survey of financial innovation and the implications for monetary policy in industrial countries, see Goodhart (1989).
16Some central banks in industrial countries—for example, the United Kingdom and the United States—have periodically published and monitored weighted monetary aggregates, such as the Divisia index, where monetary assets are assigned weights that reflect differences in the transaction services provided by different components of monetary aggregates. The Bank of England currently publishes a Divisia broad money aggregate. For details of how this index is constructed and how it compares with conventional simple-sum aggregates, see Fisher, Hudson, and Pradhan (1993). See also Pill and Pradhan (1994) for a discussion of why such indices, despite their strong theoretical foundation, especially in periods of rapid financial changes, have failed to gain widespread acceptance among policymakers.
17Stevens and Debelle (1995) find that in Australia 95 percent confidence intervals around one-year-ahead inflation projections are about 5 percentage points.
18A number of other authors also use this assumption when faced with ambiguities about the time-series properties of variables. See, for example, Ericsson and Sharma (1996).
19We are grateful to John McDermott for simulating these small sample critical values.

    Other Resources Citing This Publication