The money demand function in a small, open and quasi-monetary economy: the Gambia

Abstract

The open-economy money demand asserts that for its underlying theory to hold, the variables ought to be co-integrated. Co-integrated variables although nonstationary in level, can share a long-term trend that is indeed stationary. However, the open money demand model has mainly been tested in developed and developing nations. This article investigates the co-integrated open-economy money demand in the Gambia where the macro economy is a quasi-monetary system, small (relative to the world market), but very open with a floating exchange rate regime. In the co-integrated space, the Gambian money demand appears to be quite responsive to domestic income, a measure of interest rate and the real exchange rate fluctuations.

Notes

¹ The majority of open money demand studies report insignificant regression coefficients for the nominal exchange rate variable. The controversy between nominal and real effective exchange rates has been addressed by a number of previous studies Bahmani-Oskooee, (1991; McNown and Wallace, 1992; Arize and Shwiff, 1993).

² For a comprehensive coverage, see www.imf.org/external/NP/PFP/gambia/gamtoc.htm

³ The results of Phillips–Perron's integration test are similar to the findings of the ADF-test.

⁴ For brevity, proofs have not been provided.

⁵ The Gambian real M1 data are quite stationary in the sample period.

⁶ Note that there is no trade weighted measure of exchange rate constructed for the Gambia and thus, the measure of world's price published by the IFS is not applicable. Moreover, among the industrialized nations, the United States appears to be the Gambia's largest trade partner.

⁷ It is not uncommon for λ-trace and λ-max to detect cointegration at different significance levels  – the former test has a general alternative hypothesis, whereas the latter tests the number of co-integrating vectors (r) against (r + 1).

⁸ An exception to this rule is when the error term is serially uncorrelated and the cross correlation of the right-hand side variables is not significantly different from zero.

⁹ The general format for the ECM is ΔX_i  = α _i (ln M_{t −1} − ln Y_{t −1} − ln r_{t −1} − ln (P*)(E/P) _{t −1}) + ΣΔX_{t − i} , where X_i (i = 1, 2, 3, 4) refers to the co-integrating variables (ln M, ln Y, ln r, ln (P*)(E/P) and α _i depicts the speed of adjustment coefficients. A meaningful α _i requires the residuals of the ECM to be nearly white noise (Q(pt) = 18.1, probability = 0.3, where pt is the portmanteau test for the multivariate residual autocorrelation).