Financial Liberalization and StabilityDemand for Money:Evidence form an Emerging Economy
Aktham Maghyereh
The HashemiteUniversity,Jordan*
Abstract
The main purpose of this paper is to test empirically whether there exists a stable broad function of demand for money in Jordan over the period 1976-2000. Despite the substantial financial market liberalization in the late of 1988, the co integration and error correction methodology shows that the quarterly time-series data confirms that the broad demand for money in Jordan was stable during the period under investigation. The results also show that the inflation rate is the most important variable that explains the demand for money in the Jordanian economy.
Keywords: Financial Liberalization; Stability Demand for Money;Jordanian Economy; Error-correction model.
JEL Classification codes: E5, E52
*Maghyereh: College of Economics & Administrative Sciences, PO Box, 150459, The Hashemite University, Zarqa, Jordan. Tel: +962 (05) 3826600;Fax: +962 (05) 3826613. E-mail address:
- Introduction
A stable money demand function is considered essential for the formulation and conduct of efficient monetary policy. Consequently, a steady stream of theoretical and empirical research papers has been published during the past several decades[1]. In the case of Jordan, few studies (see for example, Marashdeh, 1990, 1994; Shoter and Barghthi, 1998, 1999; and Malawi, 2000) have attempted to determine the factors that affect the demand for money. None of these papers, however,focuses on the stability of the estimated coefficients. Furthermore, these studies have ignored the impact of financial liberalization on the demand for money in Jordan.
The objective of this study is to examine the nature and stability of the demand for money in Jordanduring the period 1976-2000. Information about the stability of the money demand function in Jordan is critical in understanding the effectiveness of monetary policy implemented by the Central Bank of Jordan (CBJ). Moreover, an accurate calibration of the long-run and dynamics and effects of rates of return on the demand for money are important in the design and assessment of the macroeconomic implications of financial liberalization and for the adoption of indirect monetary policy instruments.
Jordan, as a small and open developing economy, has been conducting its monetary policy with a view to promote monetary and macroeconomic stability that supports sustainable growth rates. Until the late eighties, the Jordanian financial system was heavily regulated. Major characteristics of the system included administered interest rates, financial support of the government at below-market rates and the control of foreign exchange transactions. Given that banks dominated the financial intermediaries in the economy and the capital market was underdeveloped, investment opportunities were heavily dependent on government priorities. During this period (1976-1988), the main instruments of monetary policy consisted of reserve requirements, direct credit controls based on credit ceilings, and Treasury Bill auctions.
The above picture alteredafter the economic crisis and devaluation of the Jordanian Dinar in 1988 through a process of financial liberalization that aimed at restoring market conditions in the system. The main steps included de-regulation of the banking sector, gradual dismantling controlson capital movements, removal of credit restrictions imposed on commercial banks and interest rate floatation. To improve the performance of monetary policy, the CBJ started to use indirect monetary policy instruments like the use of certification of deposits in the money market.
When we consider the formulation of the demand for money function in Jordan, the following points are important. First, the underdevelopment of the capital market and the regulated nature of the system during most of the period (1976-2000) indicate the significant role of real assets and hence inflation (rate of return) as determinants of money demand. Second, the devaluation of the Jordanian Dinar in 1988 and the transition period from regulation to liberalization during the last third of the estimation period (1976-2000) raises an interesting question about the stability of demand for money along with the problem of the appropriate independent variables in this function. Finally, the constrained opportunities for financial investments restrict the choice of a representative rate of return on financial instruments.
In this paper, the empirical foundation for the conduct of a stable money demand function in Jordan is evaluatedusing theco integration analysis and error correction model proposed by Johansen (1988, 1995). The empirical results suggest that despite the substantial financial market liberalization since1988, there exists a stable short- and long-run broad money demand function in Jordan.
The paper is structured as follows. Section II provides a theoretical overview and considers the relevant money demand function to be estimated for Jordan. Section III presents the data used in the analysis. Section IV discusses the econometric methodology. Section V estimates the long-run and short-run models and considers whether the estimated results satisfy the theoretical properties of a stable money demand function. Finally, section VI presents the conclusions.
II. Model Specification
There is a diverse spectrum of money demand theories that emphasizetransactional, speculative and precautionary or utility considerations (see Lewis and Mizen, 2000). These theories implicitly address a board range of hypotheses. One significant aspect, however, is that they all share common variables. In general, these hypotheses quantity the demand for money function as a function of a set of important economic variables linking money to the real sector of the economy (Judd and Scadding, 1982). Consequently, the empirical studies of money demand emerge from the theoretical literature and converge to a specification in which real money balances are a function of a scale variable and the opportunity cost of holding money.Thus, the basic model of demand for money can be expressed as (Ericsson, 1998):
where the demand for real balances is a function of the chosen scale variable to represent the economic activity and the opportunity cost of holding money. This specification represents the "desired" or long-run real money demand function and assumes a long-run unitary elasticity of the nominal cash balances with respect to the price level. This formulation implicitly assumes that the function is homogenous of degree in the level of prices.
Given the above general framework, it is important thereafter to determine the variables that explainthe demand for money. In the empirical literature, the scaled variable is used asa measure of transactions related to economic activity. This is usually represented by income, expenditure or a wealth concept and is expected to have a positive relationship with the demand for money. The price variable(consumer price index) is selected to follow closely the chosen scale variable.
One of the most important aspects of modeling the demand for money is the selection of the appropriate opportunity cost variables. The literature shows that the studies which paid inadequate attention to this matter produced poor results. There are two major ingredients; own rate and alternative return on money. The former happens to be very important, especially if the financial innovation has been taking place in an economy (Ericsson, 1998). The latter involves yields on domestic financial and real assets as well as on foreign assets. The yield on real assets is usually proxied by the expected rate of inflation, return on foreign assets or some form of exchange variable.
The theoretical literature provides some guidance in reference to the relationship between demand for money and its’ elements. As the scale variable represents the transaction effects, it is expected to be positively related to the demand for money. The own-rate is expected to be positively related as higher the return on money, less the incentive to hold assets alternative for money. Conversely, higher the returns on alternative assets lower the intensive to hold money, and hence, the coefficients of alternative returns expected to be negative.The expected inflation generally affects the demand for money negatively as agents prefer to hold real assets as hedges during the periods of raising inflation. The foreign interest rates are expected to exert negative influence as increase in foreign interest rates potentially induces the domestic residents to increase their holding of foreign assets which will be financed by drawing domestic money holdings. Similarly, the expected exchange depreciation will also have a negative relationship. An increase in expected depreciation implies that expected returns from holding foreign money increases, and hence, agents would substitute the domestic currency for foreign currency (see Lewis and Mizen, 2000).
Given the above background, we define the following model of money demand in Jordan:
where is the demand for nominal money balances;is the price level;is the real income used as a proxy for the scale variable;is the domestic interest rate; is the foreign interest rate; is the inflation rate; and is the depreciation of exchange rate. The exact definitions of the variables is discussed in the following section.
The theoretical literature does not provide any rationale for the correct mathematical form of the money demand function. There is a consensus, however, that the log-linear version is the most appropriate functional form (Sriram, 2001). While money, scale variable and interest rates variables enter in logarithms, inflation and exchange depreciation enter in levels. Consequently, in this paper the estimates of the coefficients for the scale and interest rates variables directly provide the measure of elasticities, and those of inflation and exchange depreciation show semi-elasticities.
An overview of the Jordanian economy during the late 1980s and the beginning of the1990s, the liberalization of foreign exchange controls and the adoption of more market-oriented interest rate systems, cast some doubt on the validity of equation (2) as an appropriate money demand function. Indeed, these events may cause a structural break in the long-run relationship between real money balance and its elements. In the following analysis, we will include a dummy variable to represent the measures discussed above.
- The Data
Quarterly data for the period 1976:Q1-2000:Q4 are used. A number of important issues arise when considering the appropriate data to be used as proxies for the variables of the model. In this paper, we use broad money (money plus quasi money) reported by the International Monetary Fund Financial Statistics Yearbook.
As far as the choice of the appropriate income scale variable is concerned, quarterly data about GDP are not available before 1992. In other words, it is necessary to use some proxy for income in the money demand function. In this paper, we use the index of industrial production as a surrogate for national income. The correlation coefficient between the natural logarithmic values of the annual data for GDP deflated by the consumer price index and the index of industrial production is 95 percent (1976 to 2000). We would, therefore, contend that the index of industrial production is a good proxy measure for real GDP.
Regarding the price series, the ideal price deflator would be an expenditure - based deflator in which the weights reflect the components of expenditure for which money is used. This deflator is however, not available for most of the period under consideration. Moreover, while most models of demand for money in developed economies use the GDP deflator, in this paper, we choose to use the consumer price index. In an open economy such as Jordan, the GDP deflator is not appropriate since it is constructed as a value-added deflator that excludes imports. The consumer price index deflator avoids this problem since it includes imports and excludes exports.
The actual rate of inflation is used as a proxy for the expected rate of inflation. It is measured by the first difference of the logarithm of the consumer price index. With regards to interest rate, we note that there is only a small number of interest bearing assets held by the private sector. Throughout most of the period under examination, domestic interest rates have been controlled by the authorities. However, all interest rates have generally been adjusted in a consistent manner over the period such that despite the absence of an active market mechanism through which interest rate changes are transmitted, all the main rates of interest have trended to move together. Consequently, using the discount rate (which refers to the rate at which the monetary authorities lend or discount eligible paper for deposit money banks) is a reasonable approximation to the true interest rate. In addition, the choice of the discount rate is determined by the fact that it is available on aquarterlybasis onlysince 1990. The foreign interest rate, which is used as another measure for the opportunity cost of holding money, is proxied by the annualized three-month London interbank offered rate. Finally, the depreciation rate is measured by the depreciation rate of the Jordanian Dinar per U.S. dollar exchange rate.With the exception of the inflation and depreciation rates, all the variables are expressed in logarithm terms. All the series are obtained the IMF publication International Financial Statistics (CD-ROM, December-2001).
Figure 1 shows the growth of real during the period under consideration. As can be seen, a sudden change or jump in the time pattern of the growth of real board money supply has taken place from 1988onward. This could be due to the devaluation of Jordanian Dinar and financial liberalization that started in the later part of 1988. We have therefore,included dummy variables in the empirical analysis to represent the measures discussed above.
The descriptive statistics of all variables are shown in Table 1. As can be seen, all variables, with the exceptionof inflation, do not significantly depart from the normal distribution (values of skewness and kurtosis are 0 and 3 respectively, if the observed distribution is normal). The Jarque-Bera test statistics for normality are also low and insignificant (either at the 1% level) for all variables, except for inflation and exchange rate depreciation. This suggests that the null hypothesis of these variables conforming to a normal distribution cannot be rejected. The inflation and exchange rate depreciation variablesshow significantly high values of kurtosis (5.206 and 84.114, respectively) indicating that for a given level of standard deviation, observations of this variable cluster around a central point with a small number of large outliers. The Jarque-Bera test statistic is highly significant suggesting that the inflation and exchange rate depreciation variablesdepart significantly from the normal distribution.
It is well-known that most economic time series data might have a unit root. Since correct inference depends on the stationarity of the data, it is necessary to determine the integrating properties of the variables used in this study. We test for the quarterly seasonality of the data by regressing the variables under consideration on quarterly dummies.
In order to address the integration properties of the data,we perform the (Augmented) Dickey Fuller (ADF) and Phillips-Peron (PP) unit roots tests with a step-dummy variable to handle the possibility of the structural break discussed above[2].Hubrich (1999) has shown that PP’swith a step-dummy variable is a good alternative of the ADF test in the case of one regime shift (one expected structural break) of the series[3]. The ADF test statistics are calculated with lags up to 12 periods. Moreover, the selection of the appropriate lag is based on the information criteria provided by the Akaike (AIC), Schwarz-Bayesian (SBC) and the Hannan-Quinn (HQC) statistics. In the case of different recommendation provided by the criteria, greater weight is offered to the SBC and HQC statistics in view of the tendency of AIC statistic to overestimate the lag (Dickey and Fuller, 1981). The equation implied by the selection lag is then tested for autocorrelation in the residuals and in no case the hypothesis of non-autocorrelated residuals is rejected. The results are shown in Table 2.
The degree of integration is clear-cut in the cases of the log of real money supply (RM2), log of domestic and foreign interest rates (DR, FR) and the log of industrial production index (PI). The hypothesis that these variables are stationary (i.e. I(1)) is rejected for levels but accepted for first differences, suggesting that these variables are integrated of order one I(1). There is some ambiguity concerning the results of the test for inflation (INF). The hypothesis of stationarity is rejected by the ADF but accepted by the PP test. The hypothesis of stationarity is accepted for the first differences of the variable. In contrast the depreciation rate variable (DEP) seem to be I(0) with both the ADF and PP statistics indicating stationarity. Thus, the DEP variable might be included as an exogenous variable in the long run money demand function. In the other words, the stationarity of DEP implies that this series can be excluded from the co integration analysis without loss of generality.
III. Econometric Methodology
Given that the statistical underpinnings of modern time series analysis require the data to be stationary, and that most macroeconomic series display significant trends, ledmany to first difference time before estimating economic models. Such an approach, however removes much of the long-run characteristics of the data. Engle and Granger (1987) noted that even though economic series may wonder through time, economic theory often provides a rationale for whycertain variables should obey equilibrium constraints. That is, there may exist some linear combination of the variables that, overtime, converges to equilibrium. If the separate economic series is stationary only after differencing but a linear combination of their levels is stationary, then the series are said to be co integrated. However, this test of co integration proposed by Engle and Granger does not distinguish between the existence of one or more co integrating vectors. More importantly, their test relies on a super convergence result and applies an OLS estimates to obtain estimates of theco integrating vector. These OLS estimates in practice will differ with the arbitrary normalization implicit in the selection of the left-hand-side variables for the regression equation and moreover, different arbitrary normalizations can in practice alter the Engle and Granger test results.