View of The Economic Uncertainty with Money Demand: Empirical Evidence for Selected Developed and Developing Countries

(1)

The Economic Uncertainty with Money Demand: Empirical Evidence for Selected

Developed and Developing Countries

Pei-Tha Gan*1_{, Kok-Jing Yee}2_{, Norasibah Abdul Jalil}3_{, Norimah Rambeli@Ramli}4

1,2,3,4_{Department of Economics, Faculty of Management and Economics,}

Universiti Pendidikan Sultan Idris (Sultan Idris Education University), 35900 Tanjong Malim, Perak, Malaysia gan.pt@fpe.upsi.edu.my*1

Article History: Received: 10 November 2020; Revised: 12 January 2021; Accepted: 27 January 2021;

Published online: 05 April 2021

Abstract: An important element of the empirical research on demand for money is that many published papers focus on common issues of economic activity rather than on economic uncertainty to identify a stable money demand function. Therefore, by encompassing economic uncertainty in money demand study cannot be used to disprove its precision simply because of model misspecification in general use and an intricate process of data collection. This paper uses a sample of four selected developed countries and five selected developing countries. By using the dynamic heterogeneous panel co-integration test of autoregressive distributed lag, this paper examines the money demand relationship that considering economic uncertainty and two additional control variables, namely exchange rate and inflation. The paper’s findings suggest that the economic uncertainty is an important explanatory indicator about unknown economic events that may assist in fine-tuning money demand stability, and that the exchange rate and the inflation rate roles as well as the income and the interest rate roles remain significant in the process of central bank’s monetary decision making, which eventually help to enhance money demand controls for a sensible monetary transmission mechanism.

Keywords: Dynamic heterogeneous panel, Monetary aggregates, Money demand, Optimal Economic Uncertainty

JEL Classification: C23 ∙ E51 ∙ F41 ∙ D81

1. Introduction

A steady-state money demand function enables the controls of monetary aggregates for a sensible monetary transmission mechanism, which should in turn have an anticipated impact on economic indicators, for instance, income, interest rate and inflation (Friedman, 1959; Laidler, 1977; Sriram, 1999). Although the performance of numerous studies on money demand has rarely been disaster, the role of money has been reduced due to the cause of close substitute liquid-financial-assets (i.e., economic agents increase their liquidity without depending exclusively on money) (Gurley and Shaw, 1961). In line with this drawback, failure to select an appropriate money demand framework and appropriate variables may lead to poor results in the estimation of money demand (Sriram, 2001).

An important element of the empirical research on demand for money is that many published papers focus on common issues (i.e., scale variable, opportunity cost variables and external variable)1_{rather than on economic}

uncertainty2_{to identify a stable money demand function. Therefore, by encompassing economic uncertainty in}

money demand study cannot be used to disprove its precision simply because of model misspecification in general use and an intricate process of data collection. For example, with respect to the scale variable and opportunity cost variables, Ewing and Payne (1999) suggest that income and interest rate are not sufficient to explain the reaction function of money demand. Bahmani-Oskooee and Bohl (2000) find that money demand is co-integrated with income and the interest rate; however, the study’s results are limited to the unstable parameter estimates. Tang (2002) and Tang (2004) argue that there is the problem of potential bias if the individual variable, namely real gross domestic product (GDP) or real gross national product (GNP), is used as an individual determinant in the money demand function. On the other hand, Valadkhani (2008) argues that the inclusion of inflation due to financial innovation is essentially to explain why inflation weakens people’s desire to hold money. Maravic and Palic (2005) suggest that demand for money is co-integrated with inflation but the demand for money can be unstable.

With respect to the external variable, i.e., exchange rate, Oskooee and Shin (2002) and Bahmani-Oskooee and Rehman (2005) examine the stability of money demand in selected Asian developing countries. They find that money demand is co-integrated with the exchange rate, but the parameter estimate for exchange

1

The scale variable is income; opportunity cost variables are interest rate and inflation; the external variable is exchange rate. 2

(2)

rate is unstable. In contrast, by applying the autoregressive distributed lag (ARDL) procedure, Dagher and Kovanen (2011) examine the outcome of the real effective exchange rate on the demand for money. Their evidence supports a negative long-run relationship between the demand for money and the real effective exchange rate. With respect to economic uncertainty, Atta-Mensah (2004) examines the effects of economic uncertainty on money demand in Canada; economic uncertainty is an index that is computed by using the generalized autoregressive conditional heteroscedasticity (GARCH) technique. Using co-integration analysis, he suggests that the money demand and economic uncertainty have short- and long-run relationships. Bahmani-Oskooee and Xi (2011) determine relationships between money demand and economic uncertainty through the ARDL procedure; the GARCH technique is employed on the volatility of real income to construct the data on economic uncertainty. They conclude that money demand and economic uncertainty have relationships in the both short-run and long-run. A recent study by Ӧzdermir and Saygili (2013) employed 12 different measures of uncertainty on the money demand function in Turkey. By applying vector error correction model (VECM), they find that there exists a robust long-run relationship between money demand and economic uncertainty.

The motivation for this study is the fact that the growth of money contains essential information on unforeseen future economic events (Bernanke, 2006). Solans (2003) suggests that the banking sector creates the substance of the credit money that depends in part on the products of financial innovation; on the stage of financial development, it can cause economic uncertainty that eventually reduces the ability to predict economic agents’ behavior hold money. Bernanke (2007) corroborates the above findings that economic uncertainty can cause unpredictable changes in money demand. Furthermore, empirical data in developed and developing countries typically indicate increases in monetary aggregates, i.e., narrow money (𝑚1) and broad money (𝑚2), in the aftermath of the crisis, compared to the pre-crisis period (see International Monetary Fund (IMF), International Financial Statistics (IFS)3_{; the economic conditions change from pre-crisis period to post-crisis}

period (i.e., in the aftermath of the crisis) may imply increasing negative economic uncertainty. Undoubtedly, in conditions of negative economic uncertainty, economic agents prefer to hold less risky assets that can cause money demand to rise (Atta-Mensah, 2004). Therefore, in the presence of economic uncertainty, it seems natural to examine the money demand function to help promote a stable demand for money. As stressed by the Bank for International Settlements (2001), the monetary aggregate contains useful information for economic stability.

The paper’s aim is to examine the money demand (i.e., narrow money and broad money) relationship that considering economic uncertainty and two additional control variables, namely exchange rate and inflation, such that the economic uncertainty is an important explanatory indicator about unknown economic events that may assist in fine-tuning money demand stability, and that the exchange rate and the inflation rate roles remain significant in the process of central bank’s monetary decision making; variables in the convention of money demand are also included in the study, namely income and interest rate. In doing so, a stable money demand function should help to enhance money demand controls for a sensible monetary transmission mechanism. This paper uses a sample of four selected developed countries including Canada, Japan, Switzerland and the United States (US), and five selected developing countries, including Indonesia, Malaysia, the Philippines, Singapore, and Thailand. For the empirical investigation process in examining the relationship between money demand and its determinant (i.e., economic uncertainty, exchange rate, and inflation), this paper uses the dynamic heterogeneous panel co-integration test of ARDL proposed by Pesaran et al. (1997, 1999).

The rest of the paper is structured as follows. Section 2 describes both the theoretical specifications and empirical specifications of the money demand function. Section 3 discusses the empirical findings of this study, and Section 4 presents the paper’s conclusions.

2. Model and Econometric Methodology

Keynes (1936) proposes that the motives for demanding money are transactions, precautionary, and speculative. The transactions motive results from people’s desire to hold money to finance expenditures for goods and services, which depends on income. The precautionary motive results from people’s desire to hold money against an uncertain future. The speculative motive results from people’s desire to hold less money against a higher rate of return in alternative investments. The inputs for the general money demand function are given by the following equation:

𝑚𝑡= 𝑓(𝑦𝑡, 𝑖𝑡) (1)

3

To avoid a monotonous presentation, a line chart can be used to illustrate country’s monetary aggregates in economic upheavals (e.g., Asian financial crisis (1997), dot-com bubble (2000), sub-prime housing crisis (2007), global financial crisis (2008) and European debt crisis (2009)) and in periods of recession to further ensure this point.

(3)

Where 𝑚𝑡 is the real demand for money, 𝑦𝑡 is the real income (also known as the scale variable) and 𝑖𝑡 is the nominal interest rate (also known as the opportunity cost variable). Subsection 2.1, theoretical model, augments the general money demand function by encompassing the economic uncertainty; additionally, exchange rate and inflation are also encompassed in the augmented money demand function.

Theoretical Model

The present paper introduces the economic uncertainty by augmenting the standard money demand function with some modifications. Among other researchers, this standard function is used by Bahmani-Oskooee and Rehman (2005), and Özdemir and Saygili (2013). This paper re-specifies the standard money demand function to include the economic uncertainty and two control variables (i.e., the inflation and the real exchange rate). The inputs for the standard money demand function are given by the following equation:

𝑚𝑔_𝑡 = 𝛼 + 𝛽1𝑦_𝑔𝑡− 𝛽2𝑟_𝑔𝑡− 𝛽3𝜋_𝑔𝑡− 𝛽4𝑒_𝑔𝑡− 𝛽5𝑈𝑡+ 𝜇𝑡 (2)

From the above equation, mg is the real money demand gap, yg is the real income gap, rg is the real interest rate gap, πg is the inflation gap and eg is the real exchange rate gap (i.e., these variables are expressed in terms of gap with their equilibrium/potential value). Uis the economic uncertainty (Note that U is not expressed in terms of gap because it is at an equilibrium/potential value). αis the constant term. β1,β2, β3, β4 and β5 are coefficients. μtis a shock in money demand that surpasses the one that is imputable to the real income gap, the real interest rate gap, the inflation gap, the real exchange rate gap and the economic uncertainty.

Eq. (2) is an augmented money demand function. The negative and positive signs indicate the relationships between money demand and its determinants. The inclusion of real income implies that when people’s real income increases, their desired transactions increase, and thus, the demand for money increases. The inclusion of the real interest rate implies that when real interest rate increases, the opportunity cost of holding money increases, and thus, people will want to hold less money. The inclusion of inflation captures the effect of upward price movement of goods and services on the demand for money. Inflation causes a loss in purchasing power, which reduces people’s desire to hold money. The inclusion of the exchange rate implies that depreciation in the domestic currency would increase the demand for money; this assumption underlies the monetary approach (Carbaugh, 2008). The reason may lie in the fact that a depreciation of the domestic currency induces an expansion of the domestic balance of trade, which results in larger amounts of money being needed for transactions that eventually increase people’s desire to hold money. The inclusion of the economic uncertainty captures the effect of unknown economic events on the demand for money (Gan, 2019). Negative economic uncertainty can possibly magnify bad economic conditions resulting from income falls, inflation decreases, exchange rate depreciation, and decline in the long-term interest rate, which increases people’s desire to hold money.

Methodology

Because the number of the cross-sectional and time-series observations is relatively large, the indecoroussupposition of non-stationarity and the slope parameters homogeneity are noteworthy glitches in many panel data models. This paper uses the dynamic heterogeneous panel co-integration test of ARDL proposed by Pesaran et al. (1997, 1999). It can examine money demand relationships in both short run and long run via Eq. (2) as specified in subsection 2.1. The advantage of using ARDL approach is this approach allows the variables to be I(1) and I(0). The inputs for the empirical specifications of the ARDL dynamic heterogeneous panel cointegration method are given by the following equation:

∆𝑚𝑔_𝑖𝑡= 𝛼𝑖+ ∑ 𝛿̂𝑖𝑗∆𝑚_𝑔𝑖, 𝑡−𝑗 𝑝−1 𝑗=1 + ∑ 𝛿̂1𝑖𝑗′ ∆𝑦_𝑔𝑖, 𝑡−𝑗 𝑞−1 𝑗=0 + ∑ 𝛿̂2𝑖𝑗′ ∆𝑟_𝑔𝑖, 𝑡−𝑗 𝑞−1 𝑗=0 (1) + ∑ 𝛿̂3𝑖𝑗′ ∆𝜋_𝑔𝑖, 𝑡−𝑗 𝑞−1 𝑗=0 + ∑ 𝛿̂4𝑖𝑗′ ∆𝑒_𝑔𝑖, 𝑡−𝑗 𝑞−1 𝑗=0 + ∑ 𝛿̂5𝑖𝑗′ ∆𝑈𝑖,𝑡−𝑗 𝑞−1 𝑗=0 (2) +𝛾̂ (𝑚𝑖 _𝑔𝑖, 𝑡−1− 𝛽̂1𝑖𝑦𝑔𝑖𝑡− 𝛽̂2𝑖𝑟𝑔𝑖𝑡− 𝛽̂3𝑖𝜋𝑔𝑖𝑡− 𝛽̂4𝑖𝑒𝑔𝑖𝑡 −𝛽̂5𝑖𝑈𝑖𝑡) + 𝜇𝑖+∈𝑖𝑡 (3)

From the above equation, ∆ is the first difference, 𝑚𝑔 is the real money demand gap, 𝑦𝑔 is the real income gap, 𝑟𝑔 is the real interest rate gap, 𝜋𝑔 is the inflation gap, 𝑒𝑔 is the real exchange rate gap and 𝑈𝑡 is the economic uncertainty. 𝛽̂ is the long-run coefficient; 𝛾̂ is the error correction coefficient; 𝑡 = 1,2, … , 𝑇 represents the time 𝑖 periods; 𝑖 = 1,2, … , 𝑁 represents the number of countries; 𝜇𝑖 is the group-specific effect.

𝛾̂𝑀𝐺= 1 𝑁∑ 𝛾̂𝑖 𝑁 𝑖=1 , 𝛽̂𝑀𝐺= 1 𝑁∑ 𝛽̂𝑖, 𝛿̂𝑀𝐺 𝑁 𝑖=1 = 1 𝑁∑ 𝛿̂𝑖 𝑁 𝑖=1 (4)

(4)

𝛾̂𝑃𝑀𝐺= 1 𝑁∑ 𝛾̂𝑖 𝑁 𝑖=1 , 𝛽̂𝑃𝑀𝐺= 𝛽̂𝑖∀𝑖, 𝛿̂𝑃𝑀𝐺 = 1 𝑁∑ 𝛿̂𝑖 𝑁 𝑖=1 (5)

From the panel ARDL model, the MG estimator relies on single time-series regression in each country and averaging long-run coefficients. The PMG estimator allows the short-run coefficients to vary across groups while the long-run coefficients are constrained to be the similar.

Given Eq. (3), the steady-state equilibrium in country can bedefined as follows:

𝛼𝑖+ 𝛾̂𝑀𝐺(𝑚_𝑔𝑖∗− 𝛽̂1𝑀𝐺𝑦_𝑔𝑖∗− 𝛽̂2𝑀𝐺𝑟_𝑔𝑖∗− 𝛽̂3𝑀𝐺𝜋_𝑔𝑖∗− 𝛽̂4𝑀𝐺𝑒_𝑔𝑖∗− 𝛽̂5𝑀𝐺𝑈𝑖∗) = 0(𝟔) 𝛼𝑖+ 𝛾̂𝑃𝑀𝐺(𝑚_𝑔𝑖∗− 𝛽̂1𝑃𝑀𝐺𝑦_𝑔𝑖∗− 𝛽̂2𝑃𝑀𝐺𝑟_𝑔𝑖∗− 𝛽̂3𝑃𝑀𝐺𝜋_𝑔𝑖∗− 𝛽̂4𝑃𝑀𝐺𝑒_𝑔𝑖∗− 𝛽̂5𝑃𝑀𝐺𝑈𝑖∗) = 0 (7)

where the 𝛾̂ is also known as the speed of adjustment term that is anticipated to be significantly negative and different from 0. If the cointegrating vector of the money demand function is expected to be significantly from 0, then there is evidence of a long-run relationship.

3. Data and Results Data

This study uses a sample of four selected developed countries, including Canada, Japan, Switzerland and the US, and five selected developing countries, including Indonesia, Malaysia, the Philippines, Singapore, and Thailand. The quarterly data from 1994 quarter one to 2020 quarter oneare collected from two different databases, namely the Bank for International Settlements (BIS) Statistics and the IMF, IFS. The variables’ descriptions are as follows:

 Consumer price index (CPI): The CPI data is gathered from IFS (Note that inflation is obtained by calculating the log of the CPI in the first difference.

 Economic uncertainty: Because the economic uncertainty is not available in reality, this paper applies the grid search algorithm application proposed by Gan (2014) to compute the optimal economic uncertainty index (OEUI); the OEUI represents economic uncertainty.

 Real exchange rate: The real effective exchange rate (REER) is used as the real exchange rate. Data on REER is taken from BIS.

 Real income: The Gross Domestic Product (GDP) represents the income. Data on nominal GDP is obtained from IFS. Real income can be obtained through a fraction between nominal GDP and CPI.

 Real interest rate: The money market rate (MMR) represents the interest rate. Data on MMR is obtained from IFS. The real interest rate is the difference between the nominal interest rate and inflation.

 Real money demand: The quarterly series of the nominal money stocks, i.e., nominal narrow money (𝑀1) and nominal broad money (𝑀2), are taken from IFS. 𝑀1 (𝑀2) is divided by the CPI to construct 𝑚1, i.e., real narrow money demand (𝑚2, i.e., real broad money demand).

The real money demand gap, 𝑚𝑔𝑡 time series is calculated with the difference between the logged time series

of current real money demand and potential real money demand, which is then multiplied by 100 (Note that this application is applied to the calculation of the real narrow money demand gap (𝑚1𝑔𝑡) and the real broad money

demand gap (𝑚2𝑔_𝑡)). The differences between the logged time series of the REER and the potential REER is calculated as a percentage point change in the exchange rate and can be used to compute the real exchange rate gap, 𝑒𝑔𝑡 time series. The real income gap is the deviations of the real income from the potential real income. The

calculation of the real income gap, 𝑦𝑔𝑡 time series is the difference between the logged time series of the current

real income and the potential real income, which is then multiplied by 100. The difference between the current inflation and the potential inflation is calculated as the inflation gap, 𝜋𝑔_𝑡 time series. The difference between the current real interest rate and the potential real interest rate is calculated as a percentage point change and can be used to compute the real interest rate gap, 𝑟𝑔_𝑡time series. The potential real money demand, potential real exchange rate, potential real income, potential inflation and potential real interest rate are computed by using the Hodrick-Prescott (HP) filter with a smoothing parameter, 𝜆 , equals 1600; the potential means equilibrium. Among others, the HP filter is also applied by Taylor (1999), Combes et al. (2011) and Gan (2014).

(5)

4. Discussion of Results

Preceding to the estimation of the narrow and broad money demand functions, this paper determines the level of stationarity for the real money demand gap (i.e., 𝑚1𝑔_𝑡and 𝑚2𝑔_𝑡), real income gap (𝑦𝑔_𝑡), real interest rate gap (𝑟𝑔𝑡), inflation gap (𝜋𝑔𝑡), real exchange rate gap (𝑒𝑔𝑡) and economic uncertainty (𝑈𝑡) by using Im et al. (2003),

Levin et al. (2002) and Maddala and Wu (1999) panel unit root tests. The results show that 𝑚1𝑔𝑡,𝑚2𝑔𝑡, 𝑦𝑔𝑡, 𝑟𝑔𝑡,

𝜋𝑔𝑡, 𝑒𝑔𝑡 and 𝑈𝑡 follow random walks, and concluded that these variables

are I(1), i.e., integrated of order one.

Table 1. Tests of the panel unit root

Variables Methods

Im et al. (2003) Levin et al. (2002) Maddala and Wu (1999)

𝑚1𝑔𝑡 Level 0.607 (1) 0.862 (1) 22.80 (1) First Difference -18.09 (1) *** -16.62 (1) *** 242.1 (1) *** 𝑚2𝑔_𝑡 Level -0.334 (1) -0.979 (1) 20.40 (1) First Difference -15.35 (1) *** -16.57 (1) *** 208.9 (1) *** 𝑦𝑔_𝑡 Level 0.829 (3) -0.470 (2) 11.19 (3) First Difference -19.05 (1) *** -18.89 (1) *** 236.7 (1) *** 𝑟𝑔𝑡 Level -1.213 (24) -0.136 (2) 22.51 (15) First Difference -20.29 (1) *** -19.49 (1) *** 257.2 (1) *** 𝜋𝑔𝑡 Level -1.101 (10) 1.478 (5) 19.01 (10) First Difference -26.72 (1) *** -24.37 (1) *** 249.3 (1) *** 𝑒𝑔𝑡 Level 1.049 (1) 1.114 (1) 15.84 (1) First Difference -16.14 (1) *** -14.75 (1) *** 219.4 (1) *** 𝑈𝑡 Level -1.027 (22) -0.769 (5) 25.66 (20) First Difference -22.19 (1) *** -22.12 (1) *** 281.4 (1) ***

Notes: *** implies the null hypothesis of unit root is rejected at the 1% significance levels. The parenthesis is the lag lengths selection based on the Schwarz criterion.

The dynamic heterogeneous panel co-integration test of ARDL suggested by Pesaran et al., (1997, 1999) are employed to examine the long-run relationship between the real money demand gap (i.e., real narrow money demand gap and real broad money demand gap) and its determinants (i.e., real income gap, real interest rate gap, inflation gap, real exchange rate gap and economic uncertainty). The panel ARDL model selects the lag order based on the Schwarz’s Bayesian criterion. To test the null hypothesis of slope homogeneity for long-run coefficients, the Hausman’s test is applied; the rejection of the null hypothesis suggests that the model is heterogeneous, and thus, the MG estimation is consistent compared to the PMG estimation.

For the 𝑚1 demand function, Table 2 reports the results of panel co-integration test (i.e., the MG and PMG estimates) using a sample of the four selected developed countries and five selected developing countries. According to the Hausman’s test statistic of 1.07 ( 𝑝-value=0.957), the PMG results are more appropriate

(6)

compared to the MG results, i.e., pooling sample data allow for long-run slope coefficients to be equal across countries; imposing long-run homogeneity reduces the standard errors of the long-run coefficients. The long-run coefficients of the PMG estimates of determinants of the 𝑚1 demand function, i.e., the coefficient of the real income gap (0.134), the coefficient of the inflation gap 9.197), the coefficient of the real interest rate gap 3.415), the coefficient of the real exchange rate gap 0.302) and the coefficient of the economic uncertainty (-2.417), have the expected sign and are statistically significant.

Furthermore, from the PMG estimates, the error correction coefficient is negative and statistically significant different from zero. Hence, we conclude that a long-run relationship exists between 𝑚1𝑔_𝑡 and its determinants (i.e., 𝑦𝑔𝑡, 𝜋𝑔𝑡, 𝑟𝑔𝑡, 𝑒𝑔𝑡 and 𝑈𝑡) and that the deviation from the long-run equilibrium influences the short-run

dynamics of the variables; Table 3 shows the individual countries’ results of the PMG estimates and these short-run coefficients have the expected sign and are statistically significant (Note that Table 3 does not include the individual countries’ results of the MG estimates because they are inappropriate compared to the PMG estimates).

Table 2. Panel co-integration estimations of the 𝑚1 demand function (four selected developed countries and five selected developing countries)

ARDL MG estimates PMG estimates Hausman’s test (7, 9, 0, 4, 14)

Coef. s.e. t-ratio Coef. s.e. t-ratio H p-value

Long-run coefficients 𝑦𝑔_𝑡 1.118 1.179 0.95 0.134 0.075 1.78* 1.07 0.957 𝜋𝑔𝑡 -11.70 4.506 -2.60*** -9.197 1.462 -6.29*** 𝑟𝑔_𝑡 -6.377 4.001 -1.59 -3.415 0.432 -7.90*** 𝑒𝑔𝑡 -0.577 0.346 -1.67* -0.302 0.138 -2.19** 𝑈𝑡 -3.772 4.989 -0.76 -2.417 1.455 -1.66*

Error correction coefficients

EC -0.417 0.053 -7.84*** -0.272 0.044 -6.17***

Notes: The lag order of the panel ARDL model is selected by using the Schwarz’s Bayesian criterion. *, ** and *** represent 10%, 5% and 1% significance levels, respectively.

Table 3. PMG estimates’ of individual results of the 𝑚1 demand function (four selected developed countries and five selected developing countries).

Countries EC coef. 𝑦𝑔𝑡 coef.

a _𝜋 𝑔𝑡 coef. a _𝑟 𝑔𝑡 coef. a _𝑒 𝑔𝑡 coef. a _𝑈 𝑡 coef.a Canada -0.427 (-4.81)*** 2.368 [1] (2.35)** 5.033 [3] (4.33)*** 2.710 [5] (2.99)*** -1.238 [0] (-1.90)* -11.61 [3] (-1.92)* Japan -0.221 (-4.59)*** 0.875 [1] (1.98)** 1.566 [4] (2.39)** 1.850 [5] (3.84)*** 0.937 [6] (2.08)** 1.686 [7] (2.85)*** -0.468 [1] (-3.82)*** -0.340 [3] (-1.86)* -7.213 [2] (-1.73)* -13.29 [3] (-2.70)*** -11.65 [4] (-2.12)** -18.59 [5] (-3.19)*** -17.27 [6] (-3.72)*** -13.75 [7] (-3.61)*** -14.83 [8] (-4.26)*** -6.291 [9] (-2.52)**

(7)

Switzerland -0.134 (-1.75)* -2.186 [9] (-1.74)* -2.869 [0] (-2.10)** -0.344 [4] (-1.89)* US -0.119 (-2.11)** -0.975 [3] (-1.65)* -1.954 [1] (-2.17)** -2.327 [3] (-1.85)* -0.859 [12] (-1.83)* -0.787 [14] (-1.82)* Indonesia -0.319 (-4.24)*** -0.270 [2] (-1.70)* -1.798 [1] (-1.66)* -3.486 [2] (-2.34)** -1.670 [4] (-2.08)** -0.999 [6] (-1.70)* -1.217 [7] (-1.84)* -3.049 [10] (-3.81)*** Malaysia -0.404 (-5.40)*** -1.259 [0] (-2.35)** Philippines -0.366 (-6.52)*** -2.038 [1] (-1.70)* -1.971 [12] (-1.88)* Singapore -0.090 (-4.48)*** 0.416 [4] (2.22)** -0.952 [4] (-2.08)** -1.088 [5] (-2.41)** -0.164 [1] (-1.99)** -2.416 [6] (-2.74)*** Thailand -0.368 (-5.95)*** 2.700 [1] (2.52)** 2.826 [5] (3.63)*** -0.364 [2] (-2.07)** -9.787 [1] (-1.86)* -9.378 [5] (-3.06)***

Notes: a_{The table of results report only variable that has the expected sign and is statistically significant. ( )}

represents the t-ratio and [ ] represents the lag order. *, ** and *** represent 10%, 5% and 1% significance levels, respectively.

To examine the robustness of panel dataset results for both developed and developing countries, this study splits the initial panel dataset into two panel datasets, i.e., the panel dataset for developed countries (i.e., Canada, Japan, Switzerland and the US) and the panel dataset for developing countries (i.e., Indonesia, Malaysia, the Philippines, Singapore and Thailand). The idea is that the financial system in the developed countries is more developed than in the developing countries (Didier and Schmukler, 2015). The financial markets in developing countries are under developed, for example, in that they lack financial sector instruments and payment

(8)

technologies and that many transactions are restricted to narrow money (Kumar et al., 2013). Estimating the panel datasets gives the results of panel co-integration test (i.e., the MG and PMG estimates) for developed countries and developing countries, which are reported in Table 4 and Table 5, respectively. According to the Hausman’s test statistic of 1.17 ( 𝑝-value=0.948 ) in Table 4 and the Hausman’s test statistic of 1.01 (𝑝-value=0.962) in Table 5, the results suggest that the PMG estimator is more appropriate compared to the MG estimator, i.e., pooling sample data allow for long-run slope coefficients to be equal across countries; imposing long-run homogeneity reduces the standard errors of the long-run coefficients. From Table 4 and Table 5, the long-run coefficients of the PMG estimates of determinants of the 𝑚1 demand function have the expected sign and are statistically significant (Note that, in Table 4, the coefficient of the real income gap (10.24), coefficient of the inflation gap (-2.036), coefficient of the real interest rate gap (-7.619), coefficient of the real exchange rate gap (-1.320) and coefficient of the economic uncertainty (-39.21); in Table 5, the coefficient of the real income gap (0.171), coefficient of the inflation gap (-4.747), coefficient of the real interest rate gap (-2.216), coefficient of the real exchange rate gap (-0.673) and coefficient of the economic uncertainty (-2.665)). Moreover, each error correction coefficient of the PMG estimator in Table 4 and Table 5 is negative and statistically significant different from zero. Hence, we conclude that a long-run relationship exists between 𝑚1𝑔_𝑡 and its determinants (i.e., 𝑦𝑔𝑡, 𝜋𝑔𝑡, 𝑟𝑔𝑡, 𝑒𝑔𝑡 and 𝑈𝑡) and that deviation from the long-run equilibrium influences the short-run dynamics

of the variables. The individual countries’ results of the PMG estimates for developed countries and developing countries are reported in Table 6 and Table 7, respectively; these short-run coefficients have the expected sign and are statistically significant (Note that Table 6 and Table 7 do not include the individual countries’ results of the MG estimates because they are inappropriate compared to the PMG estimates).

Table 4. Panel co-integration estimations of the 𝑚1 demand function (four selected developed countries).

ARDL MG estimates PMG estimates Hausman’stest (0, 0, 0, 6, 28)

Long-run coefficients 𝑦𝑔_𝑡 5.237 2.891 1.81* 10.24 0.587 17.45*** 1.17 0.948 𝜋𝑔_𝑡 -7.018 3.630 -1.93* -2.036 0.169 -12.02*** 𝑟𝑔𝑡 -13.43 7.327 -1.83* -7.619 0.483 -15.77*** 𝑒𝑔𝑡 -0.735 0.562 -1.31 -1.320 0.051 -25.98*** 𝑈𝑡 -28.56 14.76 -1.94* -39.21 2.035 -19.26***

EC -1.022 0.303 -3.37*** -0.428 0.257 -1.67*

Notes: The lag order of the panel ARDL model is selected by using the Schwarz’s Bayesian criterion. *, ** and *** represent 10%, 5% and 1% significance levels, respectively.

Table 5. Panel co-integration estimations of the 𝑚1 demand function (Five selected developing countries).

(9)

𝑦𝑔𝑡 1.329 0.326 4.08*** 0.171 0.051 3.34*** 1.01 0.962

𝜋𝑔_𝑡 -9.602 2.183 -4.40*** -4.747 0.949 -5.00***

𝑟𝑔𝑡 -1.016 0.754 -1.35 -2.216 0.366 -6.06***

𝑒𝑔_𝑡 -0.795 0.675 -1.18 -0.673 0.174 -3.88***

𝑈𝑡 5.160 5.038 1.02 -2.665 1.030 -2.59**

EC -0.306 0.060 -5.11*** -0.314 0.088 -3.58***

Notes: The lag order of the panel ARDL model is selected by using the Schwarz’s Bayesian criterion. *, ** and *** represent 10%, 5% and 1% significance levels, respectively.

Table 6.PMG estimates’ of individual results of the 𝑚1 demand function (four selected developed countries). Countries EC coef. 𝑦𝑔𝑡 coef.

a _𝜋 𝑔𝑡 coef. a _𝑟 𝑔𝑡 coef. a _𝑒 𝑔𝑡 coef. a _𝑈 𝑡 coef.a Canada -1.155 (-35.66)*** -3.117 [0] (-12.31)*** -1.391 [0] (-3.22)*** -0.612 [3] (-23.75)*** -0.132 [4] (-3.24)*** -0.466 [5] (-12.79)*** -31.06 [1] (-12.15)*** -41.64 [2] (-15.09)*** -48.45 [3] (-17.01)*** -56.19 [4] (-18.17)*** -66.91 [5] (-19.72)*** -69.33 [6] (-19.15)*** -66.21 [7] (-18.32)*** -70.63 [8] (-20.02)*** -71.07 [9] (-20.25)*** -71.78 [10] (-20.87)*** -62.47 [11] (-18.74)*** -70.96 [12] (-22.24)*** -72.06 [13] (-22.11)*** -66.13 [14] (-21.65)*** -60.99 [15] (-20.40)*** -72.19 [16] (-24.15)*** -66.36 [17] (-22.90)*** -60.09 [18] (-22.30)***

(10)

-59.78 [19] (-23.86)*** -49.11 [20] (-22.55)*** -45.55 [21] (-24.05)*** -30.67 [22] (-19.45)*** -22.58 [23] (-16.07)*** -31.63 [24] (-21.69)*** -15.11 [25] (-12.27)*** -7.217 [26] (-10.87)*** -6.430 [27] (-13.69)*** -6.795 [28] (-16.10)*** Japan -0.172 (-4.52)*** 2.020 [0] (3.37)*** -0.474 [1] (-3.23)*** -0.621 [2] (-3.22)*** -1.104 [5] (-4.74)*** -6.774 [1] (-3.67)*** -8.107 [4] (-3.31)*** -7.602 [5] (-3.63)*** -3.510 [7] (-2.01)** -7.217 [8] (-4.11)*** -8.299 [9] (-5.01)*** -6.521 [10] (-3.51)*** -7.033 [11] (-3.64)*** -8.417 [12] (-3.61)*** -7.146 [13] (-3.35)*** -5.285 [14] (-3.30)*** -10.34 [15] (-4.04)*** -6.355 [16] (-3.28)*** -13.14 [17] (-5.38)*** -9.822 [18] (-3.49)*** -9.762 [19] (-3.44)*** -12.20 [21] (-4.03)*** -8.001 [22] (-2.81)*** -16.12 [23] (-4.51)*** -11.77 [24] (-3.62)***

(11)

-9.207 [25] (-2.51)** -9.171 [27] (-3.65)*** -11.51 [28] (-6.30)*** Switzerland 0.016 (2.20)** -2.020 [0] (-2.16)** -0.446 [1] (-2.20)** -2.103 [1] (-4.34)*** -5.006 [16] (-6.83)*** -2.212 [17] (-3.35)*** -2.954 [19] (-4.13)*** -2.880 [20] (-3.21)*** -4.698 [21] (-5.45)*** -4.421 [23] (-4.42)*** -5.165 [25] (-6.26)*** US -0.399 (-4.21)*** -1.750 [0] (-3.33)*** -0.706 [2] (-4.87)*** -0.332 [4] (-2.39)** -0.207 [5] (-1.82)* -0.398 [6] (-3.76)*** -16.09 [2] (-7.61)*** -8.420 [3] (-3.77)*** -5.144 [4] (-2.15)** -9.179 [5] (-4.01)*** -7.514 [6] (-3.24)*** -8.600 [7] (-5.18)*** -5.676 [8] (-4.07)*** -7.267 [9] (-3.21)*** -11.99 [10] (-6.99)*** -13.59 [12] (-7.37)*** -8.447 [13] (-3.70)*** -10.82 [14] (-3.50)*** -8.308 [15] (-3.81)*** -14.28 [16] (-4.73)*** -12.88 [17] (-3.74)*** -13.05 [18] (-4.25)*** -12.67 [19] (-3.92)*** -15.45 [20] (-4.55)*** -8.241 [21] (-2.44)**

(12)

-7.445 [22] (-3.01)*** -13.59 [23] (-5.49)*** -9.159 [24] (-3.42)*** -12.25 [25] (-5.77)*** -8.427 [26] (-3.65)*** -7.263 [28] (-4.12)***

Notes: a _{The table of results report only variable that has the expected sign and is statistically significant. ( )}

Table 7. PMG estimates’ of individual results of the m1demand function (five selected developing

countries).

a _𝜋 𝑔𝑡 coef. a _𝑟 𝑔𝑡 coef. a _𝑒 𝑔𝑡 coef. a _𝑈 𝑡 coef.a Indonesia -0.373 (-5.37)*** 0.567 [2] (1.69)* 1.093 [3] (2.87)*** 2.340 [5] (5.06)*** 1.407 [6] (2.90)*** 0.838 [8] (2.36)** -0.343 [1] (-3.09)*** -0.330 [9] (-3.02)*** -7.021 [1] (-2.99)*** -11.37 [2] (-3.14)*** -16.36 [3] (-4.51)*** -16.08 [4] (-4.77)*** -15.56 [5] (-4.03)*** -14.48 [6] (-3.62)*** -11.17 [7] (-3.39)*** -12.50 [8] (-5.20)*** -9.047 [9] (-4.20)*** Malaysia -0.094 (-1.95)* -0.829 [1] (-1.94)* -1.284 [2] (-2.63)*** -1.207 [3] (-2.14)** -1.425 [4] (-2.22)** -1.795 [5] (-2.97)*** -1.997 [6] (-3.97)*** -1.777 [1] (-1.88)* -2.223 [2] (-1.80)* -3.472 [3] (-2.76)*** Philippines -0.353 (-4.37)*** -0.228 [2] (-1.67)* -0.344 [10] (-3.58)*** -1.995 [2] (-1.87)* -4.988 [10] (-7.63)***

(13)

Singapore -0.590 (-6.49)*** -1.991 [5] (-2.12)** -1.313 [6] (-1.74)* -0.744 [7] (-2.83)*** -0.481 [8] (-2.07)** -0.499 [10] (-2.00)** -7.320 [9] (-2.11)** Thailand -0.159 (-6.61)*** 0.420 [1] (3.72)*** 0.322 [2] (2.13)** 0.385 [7] (2.21)** 0.506 [8] (3.11)*** -0.751 [2] (-1.90)* -0.926 [5] (-2.53)** -0.413 [6] (-3.04)*** -2.662 [2] (-2.55)** -3.759 [6] (-3.75)*** -2.069 [8] (-2.87)***

On the other hand, for the 𝑚2 demand function, Table 8 reports the results of panel co-integration test (i.e., the MG and PMG estimates) using a sample of four selected developed countries and five selected developing countries. According to the Hausman’s test statistic of 0.08 ( 𝑝-value=0.999 ), the PMG results are more appropriate compared to the MG results, i.e., pooling sample data allow for long-run slope coefficients to be equal across countries; imposing long-run homogeneity reduces the standard errors of the long-run coefficients. The long-run coefficients of the PMG estimates of determinants of the 𝑚2 demand function, i.e., the coefficient of the real income gap (0.181), coefficient of the inflation gap 0.947), coefficient of the real interest rate gap (-0.471), coefficient of the real exchange rate gap (-0.300) and coefficient of the economic uncertainty (-3.174), have the expected sign and are statistically significant. Furthermore, from the PMG estimates, the error correction coefficient is negative, statistically significant different from zero. Hence, we conclude that a long-run relationship exists between 𝑚2𝑔_𝑡 and its determinants (i.e., 𝑦𝑔_𝑡, 𝜋𝑔_𝑡, 𝑟𝑔_𝑡, 𝑒𝑔_𝑡 and 𝑈𝑡) and that the deviation from the long-run equilibrium influences the short-run dynamics of the variables; Table 9 shows the individual countries’ results of the PMG estimates and these short-run coefficients have the expected sign and are statistically significant (Note that Table 9 does not include the individual countries’ results of the MG estimates because they are inappropriate compared to the PMG estimates).

Table 8. Panel co-integration estimations of the 𝑚2 demand function (four selected developed countries and five selected developing countries).

Long-run coefficients

𝑦𝑔𝑡 -0.632 0.461 -1.37 0.181 0.060 3.03*** 0.08 0.999

𝜋𝑔𝑡 -12.64 6.330 -2.00** -0.947 0.201 -4.71***

𝑟𝑔_𝑡 -0.582 1.160 -0.50 -0.471 0.156 -3.01***

(14)

𝑈𝑡 3.027 3.600 0.84 -3.174 0.896 -3.54***

EC -0.339 0.062 -5.50*** -0.270 0.062 -4.34***

Table 9. PMG estimates’ of individual results of the 𝑚2 demand function(four selected developed countries and five selected developing countries)

Countries EC coef. 𝑦𝑔_𝑡 coef.a 𝜋𝑔_𝑡 coef.a 𝑟𝑔_𝑡 coef.a 𝑒𝑔_𝑡 coef.a 𝑈𝑡 coef.a

Canada -0.325 (-5.37)*** 1.216 [3] (2.59)** 2.661 [5] (5.48)*** 0.534 [7] (2.03)** 1.059 [9] (4.35)*** -0.979 [1] (-2.53)** -1.200 [4] (-2.31)** -0.708 [6] (-1.66)* -0.209 [5] (-2.89)*** -11.68 [5] (-4.48)*** -3.277 [9] (-2.39)** -1.791 [14] (-2.80)*** Japan -0.649 (-8.92)*** 0.317 [1] (2.73)*** 0.472 [3] (3.00)*** 0.306 [4] (1.69)* 0.415 [5] (2.52)** 0.428 [6] (2.36)** 0.379 [9] (2.28)** -0.561 [4] (-3.54)*** -0.595 [5] (-3.51)*** -0.277 [6] (-2.05)** -0.188 [1] (-4.14)*** -0.132 [3] (-2.83)*** -0.078 [5] (-1.79)* -4.703 [1] (-6.40)*** -6.397 [2] (-5.50)*** -8.004 [3] (-6.38)*** -5.138 [4] (-3.86)*** -5.652 [5] (-3.90)*** -2.601 [6] (-1.72)* -2.670 [7] (-1.75)* Switzerland -0.011 (-0.52) 0.185 [1] (2.87)*** 0.179 [5] (2.00)** 0.206 [7] (2.76)*** 0.330 [9] (5.76)*** 0.200 [10] (3.26)*** -1.119 [1] (-7.22)*** -1.455 [2] (-7.38)*** -1.099 [3] (-4.67)*** -0.763 [4] (-3.27)*** -0.603 [5] (-3.33)*** -0.361 [6] (-3.09)*** -0.491 [7] (-3.55)*** -0.703 [9] (-5.45)*** -0.752 [10] (-5.76)*** -0.219 [11] (-2.40)** -0.203 [12] (-2.36)** -0.249 [14] (-3.03)*** US -0.142 (-2.66)*** 0.187 [1] (1.75)* 0.241 [2] (1.80)* 0.483 [6] (2.78)*** -0.890 [1] (-2.57)*** -0.166 [1] (-2.16)** -1.838 [2] (-2.15)** -0.893 [14] (-2.57)*** Indonesia -0.218 (-2.40)** -2.452 [2] (-1.74)* -1.959 [3] (-1.89)* Malaysia -0.313 (-3.92)*** 0.676 [7] (3.30)*** -2.662 [3] (-2.74)*** -0.487 [1] (-1.93)* -14.82 [7] (-3.38)***

(15)

-1.967 [4] (-1.89)* -0.516 [3] (-1.76)* -0.602 [4] (-2.02)** Philippines -0.189 (-3.15)*** -2.559 [1] (-2.32)** -2.144 [2] (-2.16)** -2.477 [12] (-2.26)** Singapore -0.422 (-11.56)*** 0.197 [1] (1.66)* -1.418 [1] (-4.27)*** -1.367 [2] (-3.59)*** -0.985 [3] (-2.14)** -0.925 [4] (-2.38)** -0.712 [5] (-2.17)** -0.363 [1] (-4.55)*** -0.155 [2] (-2.04)** -0.253 [3] (-2.39)** -0.366 [6] (-2.59)*** -2.814 [1] (-3.47)*** -4.359 [2] (-4.18)*** -2.721 [3] (-2.86)*** -2.074 [4] (-2.29)** -2.768 [5] (-3.10)*** -2.378 [6] (-2.86)*** -2.470 [7] (-3.15)*** -1.540 [8] (-2.05)** Thailand -0.158 (-1.71)* 1.262 [5] (2.03)** 0.734 [8] (2.70)*** -1.378 [2] (-2.34)** -1.313 [3] (-2.73)*** -2.208 [8] (-1.82)* -2.770 [14] (-2.52)**

To examine the robustness of panel dataset results for both developed and developing countries, this study splits the initial panel dataset into two panel datasets, i.e., the panel dataset for developed countries (i.e., Canada, Japan, Switzerland and US) and the panel dataset for developing countries (i.e., Indonesia, Malaysia, the Philippines, Singapore and Thailand). The idea is that the financial system in the developed countries is more developed than in the developing countries (Didier and Schmukler, 2015). Financial markets in developing countries are underdeveloped, for example, they lack financial sector instruments and payment technologies, and many transactions are restricted to narrow money (Kumar et al., 2013). By estimating the panel datasets, the results of panel cointegration test (i.e., the MG and PMG estimates) for developed countries and developing countries are reported in Table 10 and Table 11, respectively. According to the Hausman’s test statistic of 0.32 (𝑝-value=0.997) in Table 10 and the Hausman’s test statistic of 0.44 (𝑝-value=0.994) in Table 11, the results suggest that the PMG estimator is more appropriate compared to the MG estimator, i.e., pooling sample data allows for long-run slope coefficients to be equal across countries; imposing long-run homogeneity reduces the standard errors of the long-run coefficients. From Table 10 and Table 11, the long-run coefficients of the PMG estimates of determinants of the 𝑚2 demand function have the expected sign and are statistically significant (Note that in Table 10, the coefficient of the real income gap (1.375), coefficient of the inflation gap (-3.314), coefficient of the real interest rate gap (-1.223), coefficient of the real exchange rate gap (-0.247) and coefficient of the economic uncertainty (-10.21); in Table 11, the coefficient of the real income gap (0.259), coefficient of the inflation gap (-6.220), coefficient of the real interest rate gap (-1.172), coefficient of the real exchange rate gap (-1.010) and coefficient of the economic uncertainty (-4.193)). Moreover, each error correction coefficient of the PMG estimator in Table 10 and Table 11 is negative and statistically significant different from zero. Hence, we conclude that a long-run relationship exists between 𝑚2𝑔_𝑡and its determinants (i.e., 𝑦𝑔_𝑡, 𝜋𝑔_𝑡, 𝑟𝑔_𝑡, 𝑒𝑔_𝑡 and 𝑈𝑡) and that the deviation from the long-run equilibrium influences the short-run dynamics of the variables. The individual countries’ results of the PMG estimates for developed countries and developing countries are reported in Table 12 and Table 13, respectively; these short-run coefficients have the expected sign and are statistically significant (Note that Table 12 and Table 13 do not include the individual countries’ results of the MG estimates because they are inappropriate compared to the PMG estimates).

(16)

Table 10. Panel cointegration estimations of the 𝑚2 demand function (four selected developed countries). ARDL MG estimates PMG estimates Hausman’s test (0, 0, 0, 11, 23)

Long-run coefficients 𝑦𝑔𝑡 -0.046 1.196 -0.04 1.375 0.758 1.81* 0.32 0.997 𝜋𝑔𝑡 -5.958 2.628 -2.27** -3.314 0.631 -5.25*** 𝑟𝑔𝑡 -3.007 1.209 -2.49** -1.223 0.436 -2.80*** 𝑒𝑔𝑡 0.071 0.184 0.39 -0.247 0.105 -2.35** 𝑈𝑡 -3.297 6.277 -0.53 -10.21 4.399 -2.32**

EC -0.343 0.224 -1.53 -0.292 0.177 -1.65*

Table 11. Panel co-integration estimations of the 𝑚2 demand function(Five selected developing countries).

Long-run coefficients 𝑦𝑔𝑡 -0.738 1.003 -0.74 0.259 0.098 2.63*** 0.44 0.994 𝜋𝑔𝑡 -8.167 4.439 -1.84* -6.220 0.944 -6.59*** 𝑟𝑔𝑡 -2.401 3.036 -0.79 -1.172 0.353 -3.32*** 𝑒𝑔𝑡 -0.282 0.368 -0.76 -1.010 0.173 -5.83*** 𝑈𝑡 4.617 4.392 1.05 -4.193 1.790 -2.34**

EC -0.231 0.053 -4.36*** -0.220 0.047 -4.63***

Table 12. PMG estimates’ of individual results of the 𝑚2 demand function(four selected developed countries).

a _𝜋 𝑔𝑡 coef. a _𝑟 𝑔𝑡 coef. a _𝑒 𝑔𝑡 coef. a _𝑈 𝑡 coef.a Canada -0.782 (-7.36)*** -0.210 [3] (-2.82)*** -4.678 [13] (-2.09)**

(17)

-0.337 [5] (-4.48)*** -0.148 [6] (-2.52)** -0.232 [7] (-3.44)*** -0.292 [9] (-4.38)*** -0.135 [10] (-2.02)** -0.262 [11] (-5.05)*** -4.014 [14] (-1.97)** -3.876 [17] (-2.07)** Japan -0.115 (-3.01)*** 0.308 [0] (5.56)*** -1.812 [0] (-2.69)*** -2.342 [0] (-3.52)*** -0.025 [1] (-1.92)* -0.052 [4] (-3.70)*** -0.279 [14] (-2.33)** -0.776 [16] (-3.96)*** -0.654 [18] (-3.25)*** -0.650 [20] (-3.06)*** -0.731 [22] (-3.49)*** Switzerland 0.032 (1.13) -5.925 [0] (-6.07)*** -4.943 [0] (-5.14)*** -1.274 [7] (-3.65)*** -1.482 [8] (-4.75)*** -0.840 [9] (-2.88)*** -0.435 [10] (-1.81)* -2.794 [5] (-1.84)* -4.175 [6] (-2.52)** -7.812 [7] (-4.33)*** -7.395 [8] (-4.53)*** -3.450 [9] (-1.99)** -3.685 [10] (-2.87)*** -2.567 [19] (-3.40)*** -3.209 [20] (-3.57)*** -3.527 [21] (-4.49)*** -1.405 [22] (-1.75)* -2.890 [23] (-3.61)*** US -0.303 (-5.12)*** -0.508 [0] (-2.26)** -0.578 [0] (-3.00)*** -1.134 [13] (-1.70)* -1.564 [14] (-2.27)** -1.075 [16] (-1.68)* -1.472 [17] (-2.05)** -2.862 [20] (-4.14)*** -1.299 [23] (-2.03)**

(18)

Table 13. PMG estimates’ of individual results of the 𝑚2 demand function (five selected developing countries).

a _𝜋 𝑔𝑡 coef.a 𝑟𝑔𝑡 coef. a _𝑒 𝑔𝑡 coef. a _𝑈 𝑡 coef.a Indonesia -0.203 (-6.36)*** 0.883 [1] (6.16)*** 0.635 [3] (3.81)*** 0.419 [4] (2.65)*** 0.522 [5] (3.11)*** -0.233 [3] (-2.07)** -0.106 [1] (-2.54)** -5.787 [1] (-4.38)*** -7.958 [2] (-4.52)*** -9.874 [3] (-6.39)*** -8.754 [4] (-5.07)*** -9.344 [5] (-5.67)*** -4.910 [6] (-3.26)*** -3.263 [7] (-2.48)** -1.526 [9] (-1.78)* -0.811 [13] (-1.81)* -0.904 [17] (-3.41)*** Malaysia -0.093 (-2.61)*** 0.275 [1] (2.99)*** 0.235 [2] (1.93)* 0.322 [3] (2.49)** -1.786 [1] (-3.31)*** -1.447 [1] (-2.69)*** -0.420 [7] (-1.96)** -0.460 [8] (-2.46)** -0.226 [2] (-2.58)** -1.312 [1] (-2.16)** -1.569 [2] (-2.13)** -2.181 [3] (-2.22)** -0.763 [14] (-1.82)* Philippines -0.159 (-2.76)*** -1.771 [1] (-2.38)** -1.455 [1] (-1.86)* -0.707 [4] (-1.99)** -0.634 [6] (-1.90)* -4.513 [3] (-4.83)*** -2.750 [6] (-3.71)*** -1.416 [8] (-1.73)* -2.286 [15] (-2.67)*** Singapore -0.278 (-4.35)*** 0.320 [5] (2.10)** 0.582 [7] (3.84)*** -0.959 [5] (-2.54)** -0.638 [1] (-2.50)** -10.40 [7] (-3.57)*** -0.731 [16] (-1.96)* -1.097 [17] (-3.19)*** Thailand -0.366 (-7.07)*** 0.199 [4] (1.99)** 0.329 [6] (2.79)*** -4.511 [4] (-2.72)*** -4.583 [5] (-3.24)*** -4.679 [6] (-3.84)*** -5.714 [7] (-5.32)*** -3.891 [8] (-4.27)*** -2.780 [9] (-3.26)*** -2.473 [10]

(19)

(-3.62)*** -2.384 [11] (-3.82)*** -2.321 [12] (-4.11)*** -1.745 [13] (-3.85)*** -1.619 [14] (-4.01)*** -1.003 [15] (-2.78)*** -0.798 [16] (-2.78)***

From the results described above, a few points are worth noting. First, a long-run relationship exists between real money demand (i.e., real narrow money demand and real broad money demand) and its determinants (i.e., economic uncertainty, exchange rate, inflation, interest rate and income) and that the short-run dynamics are influenced by deviations of the current equilibrium value from the long-run equilibrium values. This finding suggest that (i) in conditions of negative economic uncertainty, economic agents prefer to hold less risky assets, which eventually causes the demand for money to rise (Atta-Mensah, 2004), (ii) a depreciation of the domestic currency induces an expansion of the domestic balance of trade that eventually increases people’s desire to hold money for transactions (Carbaugh, 2008), (iii) inflation causes a loss in purchasing power that eventually reduces people’s desire to hold money, (iv) an increase in the interest rate causes the opportunity cost of holding money to increase, which eventually decreases people’s desire to hold money, and (v) people’s real income increase causes their desired transactions to increase, which eventually causes the demand for money to rise. Second the finding suggests that the real narrow and broad money demand functions in both developed and developing countries are stable along with their determinants (i.e., economic uncertainty, exchange rate, inflation, interest rate and income); despite the fact thatthere is a discrepancy status in the financial system development between developed and developing countries.

A few policy implications can be drawn from the above discussion. The findings propose that the economic uncertainty is an important explanatory indicator about unknown economic events that may assist in fine-tuning money demand stability, and that the exchange rate and the inflation rate roles as well as the income and the interest rate roles remain significant in the process of central bank’s monetary decision making, which eventually help to enhance money demand controls for a sensible monetary transmission mechanism. Although there is a discrepancy in the development of the financial system between developed and developing countries, this paper suggests that the central banks consider the stability of the demand for money (i.e., demand for narrow money and demand for broad money) that would enable them to forecast and fine-tune the economic conditions (e.g., recession and inflation).

5. Conclusions

The paper’s aim is toexamine the money demand (i.e., narrow money and broad money) relationship that considering economic uncertainty and two additional control variables, namely exchange rate and inflation; variables in the convention of money demand are also included in the study. The results gained from the estimated narrow and broad money demand functions via the panel co-integration method support the hypothesis that a stable money demand can help to enhance money demand controls for a sensible monetary transmission mechanism. Thus, the economic uncertainty is an important explanatory indicator about unknown economic events that may assist in fine-tuning money demand stability. Moreover, the exchange rate and inflation rate roles as well as the income and the interest rate roles remain significant in the process of central bank’s monetary decision making. In line with this inference, this study extends Keynes’s (1936) and Gan’s (2019) theory of money demand that the precautionary motive demand for money may help to insure against economic uncertainty, i.e., unknown economic events.

This paper has somelimits. First, it includesonly nine countries and five determinants (i.e., economic uncertainty, exchange rate, inflation, interest rate and income), and is restricted to narrow and broad money

(20)

demand functions. For future research, potential areas of investigations may involve the inclusion of other control variables (e.g., capital mobility, financial development, and other asset prices) in the money demand function to augment the range of the study; however, containing excessive control variables in the function would inevitably confuse this measure). Moreover, one can replicate the study’s procedure to other countries and/or measure other broader monetary aggregates demand functions. Second, future investigation may consider the best level of money demand measure via a normative analysis that can help to achieve the best monetary transmission mechanism (Note that the present study is a positive analysis that studies the reaction function of money demand, i.e., concerns analysis of money demand behavior). Finally, future investigation may also consider other issues of uncertainty of the money demand measures, namely the design and the analysis methods’.

6. Acknowledgments

The authors feel thankful to the Ministry of Higher Education Malaysia for financial aid; the grant number of the 'Fundamental Research Grant Scheme’ is 2012-0018-106-02.

References

1. Atta-Mensah, J., 2004.Money demand and economic uncertainty. Bank of Canada Working Paper No. 2004-25.

2. Bahmani-Oskooee, M., Bohl, M.T., 2000. German monetary unification and the stability of long-run German money demand function. Economics Letter, 66 (2), 203-208.

3. Bahmani-Oskooee, M., Rehman, H., 2005. Stability of the money demand function in Asian developing countries. Applied Economics, 37 (7), 773-792.

4. Bahmani-Oskooee, M., Shin, S., 2002.Stability of the demand for money in Korea. International Economic Journal, 16 (2), 85-95.

5. Bahmani-Oskooee, M., Xi, D., 2011.Economic uncertainty, monetary uncertainty and the demand for money in Australia. Australia Economics Papers, 50 (4), 115-128.

6. Bank for International Settlements. 2001. 71st Annual Report. Basel, Switzerland: BIS.

7. Bernanke, B.S., 2006. Monetary aggregates and monetary policy at the Federal Reserve: a historical perspective. Speech at the Fourth ECB Central Banking Conference, Frankfurt, Germany. http://www.federalreserve.gov/newsevents/speech/bernanke20061110a.htm. Accessed 08 April 2014. 8. Bernanke, B.S., 2007.Monetary policy under uncertainty.Speech at the 32nd Annual Economic Policy

Conference, Federal Reserve Bank of St. Louis.

http://www.federalreserve.gov/newsevents/speech/bernanke20071019a.htm. Accessed 08 April 2014. 9. Bloom, N., Kose, M.A., Terrones, M.E., 2013.Held back by uncertainty. Finance and Development,

50(1), 38-41.

10. Carbaugh, R.J., 2008.International economics 12th_{edition.Cengage Learning, United States.}

11. Combes, J., Kinda, T., Plane, P., 2011. Capital flows, exchange rate flexibility, and the real exchange rate. IMF Working Paper WP/11/9.

12. Dagher, J., Kovanen, A., 2011.On the stability of money demand in Ghana: a bound testing approach.IMF Working Paper WP/11/273.

13. Didier, T., Schmukler, S.L., 2015. Financial development in Asia: beyond aggregate indicators. In: Chen, E.K.Y., Dobson, W. (eds) Financial development and cooperation in Asia and the Pacific. Routledge, New York, 143-169.

14. Ewing, B.T., Payne, J.E., 1999.Long-run money demand in Chile. Journal of Economic Development, 24 (2), 177-190.

15. Friedman, M., 1959. The demand for money: some theoretical and empirical results. Journal of Political Economy, 67 (4), 327-351.

16. Gan, P.T., 2014. The optimal economic uncertainty index: a grid search application. Computational Economics, 43 (2), 159-182.

17. Gan, P.T., 2019. Economic uncertainty, precautionary motive and the augmented form of money demand function. Evolutionary and Institutional Economics Review, 16 (2), 397–423.

18. Gurley, J.G., Shaw, E.S., 1961.Money. In: Harris, S.E. (ed) American economic history. McGraw-Hill, Washington, D.C., 101-129.

19. Im, K.S., Pesaran, M.H., Shin,Y., 2003.Testing for unit roots in heterogenous panels. Journal of Econometrics, 115 (1), 53-74.

20. Keynes, J.M., 1936.The general theory of employment, interest, and money. Macmillan, London. 21. Kumar, S., Webber, D.J., Fargher, S., 2013. Money demand stability: a case of Nigeria. Journal of

(21)

22. Laidler, D.E.W., 1977. The demand for money: theories and evidence. 2nd_{edn.Harper & Row, London.}

23. Levin, A., Lin, C., Chu, C.J., 2002. Unit root tests in panel data: asymptotic and finite-sample properties. Journal of Econometrics, 108 (1), 1-24.

24. Maddala, G.S., Wu, S., 1999.A comparative study of panel data unit root tests and a new simple test. Oxford Bulletin of EconomicsandStatistics, 61 (S1), 631-652.

25. Maravic, J., Palic, M.,2005.Econometric analysis of money demand in Serbia. National Bank of Serbia Working Paper, 1-22.

26. Özdemir, K.A., Saygili, M., 2013. Economic uncertainty and money demand stability in Turkey. Journal of Economics Studies, 40 (3), 314–333.

27. Pesaran, M.H., Shin, Y., Smith, R.P., 1997.Pooled estimation of long-run relationships in dynamic heterogeneous panels.Cambridge Working Papers in Economics 9721, Faculty of Economics, University of Cambridge.

28. Pesaran, M.H., Shin, Y., Smith, R.P., 1999. Pooled mean group estimation of dynamics heterogeneous panels. Journal of the American Statistical Association, 94 (446), 621-634.

29. Solans, E.D., 2003.Financial innovation and monetary policy. Speech at 38th SEACEN Governors Conference and 22nd Meeting of the SEACEN Board of Governors on ‘Structural Change and Growth Prospects in Asia - Challenges to Central Banking’.

30. http://www.ecb.europa.eu/press/key/date/2003/html/sp030213.en.html. Accessed 26 November 2014. 31. Sriram, S.S., 1999. Survey of literature on demand for money: theoretical and empirical work with

special reference to error-correction models. IMF Working Paper WP/99/64.

32. Sriram, S.S., 2001. A survey of recent empirical money demand studies. IMF Staff Papers 47 (3), 334-365.

33. Tang, T.C., 2002. Demand for M3 and expenditure components in Malaysia: assessment from bounds testing approach. Applied Economics Letters, 9 (11), 721-725.

34. Tang, T.C., 2004. Demand for broad money and expenditure components in Japan: an empirical study. Japan and the World Economy, 16 (4), 487-502.

35. Taylor, J.B., 1999.A historical analysis of monetary policy rules. In: Taylor, J.B. (ed) Monetary Policy Rules. University of Chicago Press, Chicago.

36. Valadkhani, A., 2008. Long-run and short-run determinants of the demand for money in the Asian-Pacific countries: an empirical panel investigation. Journalof Economics and Finance, 9 (1), 77-90.