Monetary Policy Strategy in Rwanda

4.1.2 Methodology

4.1.2.1 Data

The econometric analysis of the version of Taylor rule retained for Rwanda will be undertaken using quarterly data during 1997 (Q1) - 2001(Q4), simply because the monetary authorities actually began to carry out monetary policy in an independent way from 1997. However, data for the variables after 2001(Q4) are not available.

Real GDP, Index of Consumer Prices, monetary stock aggregate and nominal exchange were obtained from the Central Bank of Rwanda and have been all transformed in logarithm form, except the Index of Consumer Price. In addition, the inflation rate is calculated as the change over four quarters of the seasonally adjusted harmonised Index of Consumer Prices and the inflation gap has been taken as the difference between the observed inflation and the inflation target. The Inflation target is not constant and was obtained from IMF and Rwanda (1995-2002) and the National Bank of Rwanda. Potential output is estimated based on the Hodrick-Prescott Filtering Process and the output gap is expressed as (Y-Y^*), where Y is the output and Y^* is the potential output. The monetary stock aggregate variable has been de-trended using the HP filter (see Pesaran and Pesaran, 1997). The nominal exchange rate reported is in terms of Rwandan Francs per US Dollar because of the extensive use of US Dollars to dominate international transactions (Republic of Rwanda, 2000: 369).

By using this data, the focus will be on estimating the model (4) using Microfit 4.0 and by checking whether the estimated parameters of the regression are meaningful to interpretation.

4.1.2.2 Time series properties of the data

Prior to carrying out the model, it is necessary to examine the time series properties of the variables included in it. This allows one to determine whether or not the regression is spurious. For this purpose stationarity of the data set is checked by using a simple appropriate test named Dickey- Fuller. The lag length used in the test is determined using the AKAIKE (AIC) and the Schwartz Bayesian Criterion (SBC) mainly. According to this criterion, the model to be preferred should have the highest AKAIK or the highest SBC.

Tables (4.1) and (4.2) present the integration test results for variables in their level form and in first difference respectively.

TABLE (4.1): UNIT ROOT TEST-LEVELS OF VARIABLES

Variables	Trend	Constant	ADF (t)	Lag
Monetary stock aggregate (M)	Yes	Yes	-3.8241^**	2
Inflation gap (IG)	Yes	Yes	-4.1838^**	1
Output gap (YG)	No	No	-5.1630^**	4
Exchange rate (EX)	Yes	Yes	-1.5882	2

Note: ADF critical values:

* Significant at the 1% level

** Significant at the 5% level

TABLE (4.2): UNIT ROOT TESTS OF THE FIRST DIFFERENCE

Variables	Trend	Constant	ADF (t)	Lag
DM	Yes	Yes	-9.0445^*	0
DIG	No	No	-3.7920^**	0
DYG	No	No	-3.7141^*	0
DEX	Yes	Yes	-16.3773^*	0

Note: ADF critical values:

* Significant at the 1% level

** Significant at the 5% level

The results reported in Table (4.1) indicate that all the variables are stationary in levels except for the nominal exchange rate. The unit root for the variables, which are stationary, is rejected at all the conventional significance levels (the null hypothesis states that the time series has unit root and the alternative is that the time series does not have unit root).

From table (4.2) when the variables are transformed to their first differences, the ADF test rejects the null hypothesis about unit root at the all-conventional level of significance for all the variables except the inflation gap which rejects the unit root at the 5% level of significance. Therefore, all the variables are first - difference stationary. Overall, it can be concluded that all the variables in the model (equation 4), including the exchange, rate can be treated as I (0) because the exchange rate expressed in the model relates the change in the nominal exchange rate. Consequently, the Ordinary Least Squared (OLS) analysis by the feedback rule in which the monetary stock aggregate reacts to the inflation gap, the output gap and the exchange rate or all of them will provide non - spurious results.

Having the monetary stock aggregate (M1) as the dependent variable, one expects that the monetary stock aggregate will increase if inflation is below target, output is below the output gap, that is, the coefficient of YG and IG are expected to carry a negative sign. Regarding the change in the nominal exchange rate, we expect that the response of M to DEX would be negative.