Money Demand in Rwanda: A Cointegration Analysis 2008-2015

Abstract

This paper estimates the demand for money in Rwanda over the 2008Q3 to 2015Q4 period via unit root and cointegration methods. Utilizing the Johansen cointegration methodology, it establishes that a long-term relationship exists among the included variables. The paper also estimates an error correction model (ECM) as well as a vector error correction model (VECM), extending previous analyses by testing for Granger causality among the variables. It finds that the narrow definition of money, M1, serves as a relatively better measure of the money aggregate than M2, and M3. The long-term interest rate (LKRR) also seems to provide relatively better results than the short-term rate (LRR, and LTR) when we use broad money definition, M2. Both the ECM and VECM for M1, narrow definition of money estimates showed the expected signs, in the ECM model as expected LM1 and LGDP were positively related while LM1 and LKRR, LRR, and LTR were negatively related. The adjustment coefficient in the ECM showed that about 79.75 % of disequilibrium is corrected in each quarter. Impulse response functions suggest that the traditional money demand function, which places LM1 as its ‘dependent’ variable while including income and interest rates as its regressors, was stable with little responses in the specific case of Rwanda over the period under review.