^{1}

^{*}

^{2}

^{*}

This study examines the short-run and long-run stability properties of money demand in Thailand using the monetary aggregates M1, M2 and M3, for the period from 1993Q1 to 2012Q4. We use the dynamic OLS specification of Stock and Watson (1993) and Ball (2001), and the estimation technique of the Johansen cointegration test to determine the stability of money demand. The results from the Johansen cointegration test reveal that there is only a long-run relationship between M1 money demand and real GDP (a proxy for real income) and interest rate. In the short run, only a change in real GDP affects M1 money holdings. In the long-run both real GDP and an interest rate determine money demand. The short-run instability of M1 money demand makes it difficult for the monetary authorities to use M1 as an intermediate target to control intermediate-run and long-run inflation.

Empirically, researchers have long been searching for explanatory variables that can influence the function of real demand for money. Two of various determinants of real money demand function are real income (or real GDP) and interest rate. Ericsson [

Goldfeld [^{1}. Barnett et al. [

Empirical studies in some advanced economies also give mixed results. Stock and Watson [

In the present study, we use the most recent time series data obtained from the Bank of Thailand during the first quarter of 1993 and the fourth quarter of 2012 to investigate the long-run relationship between M1, M2, and M3 money demands and the two determinants (real GDP and interest rate). We use the model specification of Stock and Watson (1993) and Ball (2001). Our estimation techniques include the dynamic ordinary least squares (DOLS) and Johansen cointegration tests. We find that the DOLS procedure is not applicable for our data set. However, our results from Johansen cointegration test reveal that there only exists a long-run relationship between M1 money demand, real GDP and interest rate. In the short run, only a change in real GDP affects M1 money holding. This latter result implies that it might be difficult for the monetary authority to us monetary aggregates as intermediate targets to pursue intermediate and long-run inflation goals.

This paper is organized as follows. Section 2 describes the data and methodology. Section 3 presents our empirical results and the last section gives concluding remarks.

The stationarity properties of the time series data are crucial in determining a stable relation between macroeconomic variables. In particular, for our purpose, stationarity is important for using the cointegration test proposed by Johansen and Juselious [

We obtained quarterly data on nominal monetary aggregates M1, M2, and M3, as well as data for real GDP, interest rates (saving deposit rate and 10-year government bond yield) and the consumer price index for all items from The Bank of Thailand website (www.bot.or.th). The data were obtained for the period from the first quarter of 1993 to the fourth quarter of 2012.

Theoretically, real money demand is affected by real income (proxied by real GDP, and interest rate. The functional form of multiple regression that is widely used in empirical studies is^{2}:

where

The Johansen cointegration test is a common estimation technique used in estimating the demand functions and determining their stability properties. The test employs the maximum likelihood procedure to determine the existence of cointegrating vectors in non-stationary time series as a vector autoregression (VAR) in the form:

where x is a vector of non-stationary variables,

The short-run dynamics are depicted by the coefficients of the lagged values of the first difference terms in Equation (3)^{3}. Although these coefficients are used in short-run stability tests, the coefficient of the error-cor- rection term

Before we estimate long-run money demand equations, it is necessary to test for the time-series properties of the variables using unit root tests and cointegration tests. These tests determine whether the variables possess properties that allow us to establish a non-spurious relation, and whether the variables possess long-run stability, respectively. Following these tests, we present the estimates of our long-run equilibrium equations in this section.

We first perform the unit root test using the Phillips and Perron [

. Results of unit root test

Variable | PP Test with Constant |
---|---|

(a) Level of Series | |

Real Money Supply (m − p): M1 | −0.059 (0.950) |

M2 | −1.423 (0.567) [11] |

M3 | −0.947 (0.759) [8] |

Real GDP (y) | −0.542 (0.876) |

Interest Rate (r): Saving Deposit Rate | −1.557 (0.499) [3] |

Ten-Year Government Bond Yield | −1.437 (0.560) [1] |

(b) First Difference of Series | |

Δ(m − p): M1 | −21.048 (0.000)*** |

M2 | −11.679 (0.000)*** [11] |

M3 | −11.928 (0.000)*** [3] |

Δy | −10.628 (0.000)*** |

Δr: Saving Deposit Rate | −7.297 (0.000)*** [1] |

Ten-Year Government Bond Yield | −9.827 (0.000)*** [4] |

Note: The number in bracket is the optimal bandwidth determined by the Bartlett kenel. The number in parenthesis is the p-value of rejecting the null hypothesis of unit root. *** denotes significance at the 1 percent level.

The results from PP tests with a constant show that all variables contain a unit root in their levels since the null hypothesis of unit root cannot be rejected. However, the test rejects the null hypothesis of unit root for first differences of all series. We therefore conclude that all series are integrated of order one, or they are I(1), for differences. When integrated, these series might or might not be cointegrated. The Johanson cointegration test can be applied to determine Cointegration of these series.

The Johansen cointegration test is performed using the levels of the three variables in each equation. The VAR(p) model of three variables is used to determine the optimal lag order p. Based upon the Akaike information criterion (AIC), the optimal lag length is four. The results of the Johansen cointegration tests are reported in

These results determine whether all three variables in the VAR(4) model are cointegrated, i.e., exhibit a long-run equilibrium relationship. The likelihood ratio tests, which are asymptotically distributed with three degrees of freedom, show that the trace and maximum eigenvalue statistics are greater than the 5% critical value for M1, but they are lower than the 5% critical value needed for M2 and M3. Therefore, the null hypothesis that real money demand, real GDP, and interest rate are not cointegrated is rejected for M1 money demand, but not for M2 and M3.

Based upon the above-mentioned results of the Johansen cointegration test, only narrowly defined money (M1) should be considered for a test of long-run money demand stability. The estimated long-run relationship between real money demand, real GDP (as a proxy of real income), and interest rate is shown in Equation (4).

[t-statistics are in parentheses. *** denotes significance at the 1% level.]

. Results of Johansen cointegration test

(a) Demand for M1 | ||||
---|---|---|---|---|

Trace Test | ||||

Hypothesis | Eigenvalue | Trace Statistic | 5% Critical Value | Prob. |

None | 0.290 | 35.728 | 29.797 | 0.009 |

At Most 1 | 0.125 | 9.997 | 15.495 | 0.281 |

Maximum Eigenvalue Test | ||||

Hypothesis | Eigenvalue | Max-Eiegen Statistic | 5% Critical Value | Prob. |

None | 0.290 | 25.731 | 21.131 | 0.011 |

At Most 1 | 0.125 | 9.977 | 14.625 | 0.551 |

(b) Demand for M2 | ||||

Trace Test | ||||

Hypothesis | Eigenvalue | Trace Statistic | 5% Critical Value | Prob. |

None | 0.127 | 17.244 | 29.797 | 0.622 |

At Most 1 | 0.091 | 7.229 | 15.495 | 0.551 |

Maximum Eigenvalue Test | ||||

Hypothesis | Eigenvalue | Max-Eigen Statistic | 5% Critical Value | Prob. |

None | 0.127 | 10.016 | 21.132 | 0.743 |

At Most 1 | 0.091 | 7.096 | 14.265 | 0.778 |

(c) Demand for M3 | ||||

Trace Test | ||||

Hypothesis | Eigenvalue | Trace Statistic | 5% Critical Value | Prob. |

None | 0.132 | 17.047 | 29.797 | 0.637 |

At Most 1 | 0.083 | 6.562 | 15.495 | 0.629 |

Maximum Eigenvalue Test | ||||

Hypothesis | Eigenvalue | Max-Eigen Statistic | 5% Critical Value | Prob. |

None | 0.132 | 10.485 | 21.132 | 0.698 |

At Most 1 | 0.083 | 6.433 | 14.265 | 0.558 |

Note: The probability is the p-value provided by MacKinnon et al. [

The estimated coefficient associated with y_{t} is 0.983, indicating that a 1 percent increase in real income will cause real money demand to increase by 0.983 percent. This result is consistent with the theoretical hypothesis of the Quantity Theory of Money that the income elasticity of money should be unitary elastic. The estimated coefficient of r_{t} is −0.170. These results establish evidence of a long-run relationship between money, and income and interest rates.

Next, we address whether there exists a short-run relationship between narrowly-defined money, income and interest rates. We explore these relationships in an error-correction mechanism (ECM), which tests for the short- run dynamics of the money demand function. The results of these short-run dynamics appear in Equation (5).

[t-statistic in parenthesis. ** and * denote significance at the 5% and 10%, respectively.]

R^{2} = 0.596 F = 6.932 S. E. of Regression = 0.046

The above results show that the impact of a change in real income is significantly different from zero, while the coefficients associated with interest rates are not. The coefficient attached to the error correction term

This paper investigates the short-run and long-run stability properties of money demand functions in Thailand over the period of 1993Q1 to 2012Q4 using monetary aggregates. To do so it uses the Johansen cointegration test to determine whether there is a long-run relationship among the variables that comprise money demand. The findings show that cointegration exists for M1 money demand function, but not for M2 and M3 money demand function variables. Therefore, only M1 appears to play a role in the monetary transmission mechanism. Other variables (exchange rate and inflation rate) were initially included in the money demand equations, but these variables played no role as determinants of money demand in Thailand. Therefore, we exclude these variables from our reported estimates. The short-run dynamics show that real GDP is an important factor in M1 money demand, but that the interest rate is not significant. Even though there is a cointegrating relation for M1 money demand, the error-correction mechanism representing the short-run dynamics of money demand show an unstable function. This instability result implies that the use of M1 as an intermediate target to achieve an intermediate and long-run inflation rate goal might be difficult.