VECM Estimation

In finding the short-run causality among variables, the study uses the Wald test in order to compare F-test with F-statistic. Based on the VECM, the result of the Granger causality test is in Table 2.14. First, the estimated coefficient of the error correction term for real simple-sum narrow money (RM1) of both logarithm models and log-log models have proper negative signs and are statistically significant at the 5% level. The error correction term value of the logarithm model, -0.021, implies 2.1% of money demand deviation in the short run would be adjusted in every quarter towards the steady state equilibrium level of money demand in the long run.

61 Table 2.14 Short Run and Long Run VECM Granger Causality Analysis (Real Simple-Sum

Narrow Money-RM1)

F-statistics [p-Values] [t-statistics]

ΔRM1 - 2.157 1.253 4.424 0.744 -0.013

Asterisks (*), (**) and (***) indicate significant at 10%, 5% and 1% levels, respectively

Meanwhile, the error correction term value of the log-log model, -0.013, implies 1.3% of money demand of short-run deviation could be remedied to the long-run equilibrium level of money demand in every quarter. This suggests that the demand for RM1 performs faster in correcting the return of adjustment to equilibrium in the logarithm model than the log-log model. In both logarithm models and log-log models, the nominal exchange rate (LER) does Granger-cause real simple-sum narrow money (RM1) at the 10%

62 level in the short-run with coefficients, 3.8 and 4.4 respectively. The other regressors are not statistically significant at the 10% level.

On the other hand, for logarithm models, real simple-sum narrow money (RM1) does Granger-cause the 1 month deposit rate (IDEP1M) at the 10% level in the short run with coefficient 3.16. It might be because the higher bank reserve as part of real simple-sum narrow money (RM1) influences the determination cost of funds from the bank that is conveyed to savers.

Meanwhile, the real income (LREALGDP) does Granger-cause the nominal exchange rate (LER) at the 5% level in the short run with coefficient 4.2. It is possible that this happened because the higher domestic income of households leads to a higher level of foreign import of goods and services into the country. As for log-log models, the real income (LREALGDP) does Granger-cause the nominal exchange rate (LER) at the 5% level in the short run with coefficient 4.7.

Second, the estimated coefficient of the error correction term for real simple-sum broad money (RM2) of both logarithm and log-log models have proper negative signs and are statistically significant at the 5% level. The error correction term value of logarithm model, -0.025, implies 2.5% of money demand deviation in the short run would be adjusted in every quarter towards the steady state equilibrium level of money demand in the long run (see Table 2.15).

On the other hand, the error correction term value of the log-log model, -0.021, implies 2.1%

of money demand of short-run deviation could be remedied to the long-run equilibrium level of money demand in every quarter. This suggests that the demand for RM2 performs faster in correcting the return of adjustment to equilibrium in the logarithm model than the log-log model. Meanwhile, in logarithm models, the real income (LREALGDP) and nominal exchange rate (LER) do Granger-cause real simple-sum broad money (RM2) at the 5% and 10% level in the short run but with higher coefficients, 22.7 and 5.06, respectively. The other regressors are

63 not statistically significant at the 10% level. As for log-log models, the real income (LREALGDP) and nominal exchange rate (LER) do Granger-cause real simple-sum narrow money (RM2) at the 1% level in the short run but with higher coefficients, 21.3 and 9.6

respectively. The other regressors are not statistically significant at the 10% level.

Instead, for the logarithm models, real simple-sum broad money (RM2) does Granger-cause both the real income (LREALGDP) and the nominal exchange rate (LER) at the 5% level in the short run but with higher coefficients, 6.2 and 7.4, consecutively. It might be because the Table 2.15 Short Run and Long Run VECM Granger Causality Analysis (Real Simple-Sum

Broad Money-RM2)

F-statistics [p-Values] [t-statistics]

ΔRM2 - 21.286 1.254 9.627 0.172 -0.021

Asterisks (*), (**) and (***) indicate significant at 10%, 5% and 1% levels, respectively

64 changes of real simple-sum broad money (RM2) influence the expenditure consumption of households and encourage economic agents to swap more of their local currency into foreign currency (flight to quality). Meanwhile, the real income (LREALGDP) and the nominal exchange rate (LER) do Granger-cause the 1 month deposit rate (IDEP1M) at the 5% level in the short run but with higher coefficients, 14.6 and 35.3, respectively. Thus, the higher actual output and depreciation of local currency lead to the domestic interest rate. As for log-log models, the real simple-sum broad money (RM2) does Granger-cause both the real income (LREALGDP) and the nominal exchange rate (LER) at the 5% level in the short run but with higher coefficients, 6.5 and 7.6, respectively. In addition to that, both the real income (LREALGDP) and the nominal exchange rate (LER) do Granger-cause the 1 month deposit rate (IDEP1M) at the 1% level in the short run but with higher coefficients, 7.0 and 11.1.

Third, the estimated coefficients of the error correction term for real divisia narrow money (RDIVM1) of both logarithm and log-log models have proper negative signs and are statistically significant at the 10% level. The error correction term value of the logarithm model, -0.191, implies 19.1% of money demand deviation in the short run would be adjusted in every quarter towards the steady state equilibrium level of money demand in the long run (see Table 2.16).

The error correction term value of the log-log model, -0.072, implies 7.2% of money demand of short-run deviation could be remedied to the long-run equilibrium level of money demand in every quarter. This suggests that the demand for RDIVM1 performs faster in correcting the return of adjustment to equilibrium in the logarithm model than in the log-log model.

Furthermore, in logarithm models, the 1 month deposit rate (IDEP1M) does Granger-cause real divisia narrow money (RDIVM1) at the 1% level in the short run but with higher coefficient 16.1. The other regressors are not statistically significant at the 10% level. As for log-log models, the real income (LREALGDP) and 1 month deposit rate (LDEP1M) do Granger-cause

65 real divisia narrow money (RDIVM1) at the 5% and 1% level in the short run but with higher coefficients, 8.9 and 9.2 respectively. The other regressors are not statistically significant at the 10% level.

F-statistics [p-Values] [t-statistics]

ΔRDIVM1 - 8.872 9.191 3.083 2.177 -0.072

Asterisks (*), (**) and (***) indicate significant at 10%, 5% and 1% levels, respectively

On the contrary, for logarithm models, real divisia narrow money (RDIVM1) does Granger-cause the 1 month deposit rate (IDEP1M) at the 10% level in the short run but with coefficient 3.6. It might be because the higher bank reserves as part of real divisia narrow money (RDIVM1) influence the determination cost of fund from the bank that is conveyed to savers.

66 Meanwhile, the real income (LREALGDP) and the nominal exchange rate (LER) do Granger-cause the 1 month deposit rate (IDEP1M) at the 1% level in the short run but with higher coefficients, 19.2 and 34.0 respectively. Thus, the higher actual output and depreciation of local currency lead to a higher domestic interest rate.

As for log-log models, the real divisia narrow money (RDIVM1) does Granger-cause the 1 month deposit rate (LDEP1M) at the 10% level in the short run but with higher coefficient 5.8.

In addition to that, both the real income (LREALGDP) and the nominal exchange rate (LER) do Granger-cause the 1 month deposit rate (LDEP1M) at the 10% and 5% level in the short run but with higher coefficients, 5.7 and 9.1.

Fourth, the estimated coefficients of the error correction term for real divisia broad money (RDIVM2) of both logarithm and log-log models have proper negative signs and are statistically significant at the 5% level. The error correction term value of the logarithm model, -0.022, implies 2.2% of money demand deviation in the short run would be adjusted in every quarter towards the steady state equilibrium level of money demand in the long run (see Table 2.17).

The error correction term value of the log-log model, -0.019, implies 1.9% of money demand of short-run deviation could be remedied to the long-run equilibrium level of money demand in every quarter. This suggests that the demand for RDIVM2 performs faster in correcting the return of adjustment to equilibrium in the logarithm model than in the log-log model.

Furthermore, in logarithm models, the real income (LREALGDP) and 1 month deposit rate (IDEP1M) do Granger-cause real divisia narrow money (RDIVM2) at the 1% and 5% level in the short run but with higher coefficients, 35.336 and 6.693 respectively. The other regressors are not statistically significant at the 10% level. As for log-log models, the real income (LREALGDP) and nominal exchange rate (LER) do Granger-cause real simple-sum narrow

67 money (RDIVM2) at the 1% level in the short run but with higher coefficients, 13.337 and 35.572 respectively. The other regressors are not statistically significant at the 10% level.

Table 2.17 Short Run and Long Run VECM Granger Causality Analysis (Real Divisia Money-RDIVM2)

F-statistics [p-Values] [t-statistics]

ΔRDIVM2 - 13.337 4.291 35.572 0.387 -0.019

Asterisks (*), (**) and (***) indicate significant at 10%, 5% and 1% levels, respectively

On the contrary, for logarithm models, real divisia broad money (RDIVM2) does Granger-cause both the real income (LREALGDP) and the nominal exchange rate (LER) at the 1% level in the short run but with coefficients 9.78 and 8.74 respectively. It might be because the higher real divisia broad money (RDIVM2) influences the expenditure consumption of household and

68 encourages economic agents to swap more of their local currency into foreign currency (flight to quality).

Meanwhile, the 1 month deposit rate (IDEP1M), the nominal exchange rate (LER), and 3 months US Treasury Bills (IUSTB3M) do Granger-cause the real income (LREALGDP) at

the 5%, 1% and 10% level in the short run but with higher coefficients, 7.32, 66.64, and 4.72 respectively. Thus, the higher return on the 1 month deposit rate (IDEP1M) and return on holding foreign asset (including foreign currency) with respect to increasing 3 months US Treasury Bills lead to higher spending due to income effect and wealth effect consecutively.

As for log-log models, the real divisia broad money (RDIVM2) does Granger-cause the 1 month deposit rate (LDEP1M) and the nominal exchange rate (LER) at the 10% and 5%

level in the short run but with higher coefficients, 5.3 and 7.3 respectively. In addition to that, the real income (LREALGDP) does Granger-cause both the 1 month deposit rate (LDEP1M) and the nominal exchange rate (LER) at the 1% level in the short run but with higher coefficients, 22.5 and 40.9 respectively. Finally, 3 months US Treasury Bills (IUSTB3M) do Granger-cause the real income (LREALGDP) at the 1% level in the short run but with higher coefficient, 23.8.