Asymmetric effects of exchange rate changes on the demand for money in Africa

ABSTRACT

In order to account for currency substitution, the exchange rate is included in the demand for money. More recent studies have demonstrated that exchange rate changes could have asymmetric effects on the demand for money or domestic currency. In this paper, we consider the experiences of 18 African countries and show that in most countries, indeed exchange rate changes have short-run asymmetric effects on the demand for money. However, short-run effects translate to long-run asymmetric effects only in a limited number of African countries.

KEYWORDS:

• Money demand
• exchange rate
• asymmetry
• nonlinear ARDL
• Africa

JEL CLASSIFICATION:

• E41

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹ Note that once normalization takes place, $\hat{b}=\frac{{\hat{\rho }}_1}{-{\hat{\rho }}_0},\stackrel{\hspace{0.17em}}{\hspace{0.17em}}\hat{c}=\frac{{\hat{\rho }}_2}{-{\hat{\rho }}_0}\stackrel{\hspace{0.17em}}{\hspace{0.17em}}and\stackrel{\hspace{0.17em}}{\hspace{0.17em}}\hat{d}=\frac{{\hat{\rho }}_3}{-{\hat{\rho }}_0}.$.

² In practice, POS at time t, is the cumulative sum of all changes in LnEX prior to time t (including t itself) where negative changes are replaced by zeroes. Similarly, NEG at time t is the cumulative sum of all changes in LnEX, where positive changes are replaced by zeroes.

³ By meaningful effects we mean significant coefficient that is also supported by cointegration. In addition to the F test for cointegration, we also provide an alternative test under which we use normalized long-run estimates and long-run model 1 and generate the error term. Labelling this error term as ECM, we go back to the error-correction model (2) and replace the linear combination of lagged level variables by ECM_t-1 and estimate the new specification after imposing the same optimum lags. If ECM_t-1 carries a significantly negative coefficient, that will support cointegration. Since the t-test is used for this purpose, this is also known as the t-test for cointegration. Like the F test, Pesaran, Shin, and Smith (2001, 303) provide new critical values.

⁴ As for the long-run effects of the other two variables, income elasticity is significant in 10 models and in most countries it is above unity, implying some diseconomies of scale. Inflation rate elasticity is significantly negative only in five countries.

⁵ Other diagnostics are similar to those of the linear models, i.e., there is lack of serial correlation in most models, most optimum models are correctly specified, estimates are stable in almost all cases, and size of adjusted R² is reasonable.