A re-examination of the exchange rate – Interest rate differential relationship in Ghana

Abstract

This study revisits the relationship between the exchange rate and interest rate differential in Ghana with a focus on the period in which the country adopted the inflation targeting regime. Using macro-data spanning 2002 to 2019 for Ghana and the United States, we show the nonexistence of the relationship in both the short-run and long-run. Further, we show a positive but slow responsiveness of the exchange rate to interest rate differential shocks from the short-run to the medium term. The long-run results, however, shows a case of a strong and significant response of exchange rate to interest rate differential shocks. We recommend that the Bank of Ghana (BoG) addresses perennial macroeconomic instability, especially on inflation, which has been shown to fuel investment uncertainty and investment insensitivity to interest rate.

1. Introduction

The relationship between the exchange rate and the interest rate has been of prime importance to policymakers, trade players and academics alike. The significance of these variables stems from the fact that exchange rate is a major driver of investment and trade across borders [1]. The interest rate parity condition posits that interest rate differentials could trigger substantial flow of capital across borders, which can have profound impact on exchange rate movement. Additionally, it has been established that a flexible exchange rate regime has implications for commodity arbitrage, financial innovation, and cross-border portfolio movements [2]. The theoretical foundation for the exchange rate and interest rate link stems from the sticky-price and flexible-price models [3]. The theoretical arguments put forward by Ref. [4], and [5] also link the relationship between exchange rate and interest rate differentials to the international parity conditions, expectation and rapid adjustment in capital markets.

In the developing world, understanding exchange rate and interest rate differentials is crucial because it helps in controlling the formation of expectations, and macroeconomic performance. It is in this regard that emerging economies have, among others, resorted to either inflation targeting, interest rate targeting or exchange rate targeting in a bid to ensure macroeconomic stability [6]. This implies that the lack of a clear and comprehensive understanding of the relationship between the two variables poses serious monetary policy concerns for policymakers [7]. This is linked to the argument by Ref. [8] that in emerging economies, while the exchange rate plays a crucial role in economic fundamentals, the short-run interest rate is the typical policy instrument policymakers use to affect currency values. Analysis of the co-movement between these two variables is therefore crucial in a policy sense. Further, exchange rate and interest rate differential in the short-run is expected to deviate substantially from the long-run due to cyclical macroeconomic instability or weak fundamentals.

Conspicuously, the link between the exchange rate and interest rate has not been revisited since Ghana adopted the Inflation Targeting (IT) framework. In May 2007, Ghana officially became the second country in Sub-Saharan Africa after South Africa to adopt the IT framework. The decision by the monetary authorities to adopt the IT regime centred on the notion that: (a) IT minimizes the problem of ‘inflation bias’ that arises under uncertainty; (b) IT by providing a nominal anchor for monetary policy reduces variability, enhances inflation forecasting by reducing the level of expected inflation and/or increasing its predictability [9]. It is therefore expected that in an IT regime, the interest rate is bid down which in effect, can reduce frequent exchange rate variability.

Despite these theoretical implications of IT for inflation, interest rate and exchange rate, comprehensive empirical contributions informing policy as to whether this matters empirically and the extent of the magnitudes are hard to find. However, since Ghana adopted the IT regime, policymakers are still unaware of the relationship. The debate on the two since the adoption of IT is only gleaned from public discourse without empirical evidence. Indeed, the few studies which are in line with our argument do not interrogate the interest rate and exchange rate relationship in Ghana’s IT regime [see e.g., Refs. [10,11]]. First, though [11] employed the autoregressive distributed lag technique to examine the exchange rate and interest rate relationship in Ghana, they do employ the vector error correction technique to inform policy on the responsiveness of Ghana’s exchange rate to the interest rate differential between the country and the United States of America (heareafter: USA). Also, Ref. [10] contributed to the exchange rate and interest rate discussion but focused on the extent to which movements in the two variables impact inflation in the short-run and long-run. In doing so, the authors employed the ARDL and OLS estimation techniques to determine the presence of the Fisher Effect and the International Fisher Effect (IFE). Robust evidence from the study shows that, in the short-run, a 1% increase in the depreciation of the Ghana cedi leads to an increase in the rate of inflation by 0.20%. The authors, therefore, did not proceed to examine how interest rate differential across Ghana and its major trading partner, the United States, affect the former’s exchange rate movements. Besides, these prior contributions have not pointed out whether the relationship between Ghana-USA interest rate differential differs from the short-term to the long-term.

This study, therefore, seeks to fill these voids in the extant scholarship on Ghana. Overall, the study investigates the short-run and long-run relationship between interest rate and exchange rate differentials in the case of Ghana. The specific objectives are captured in what follows. First, the study provides evidence of the existence or otherwise of the exchange rate and interest rate differential relationship in Ghana since the country adopted the IT framework. Second, the study provides evidence as to whether there exist some short-run and long-run linkages between exchange rate and interest rate differential. Third, we provide evidence on the responsiveness of exchange rate to a shock in interest rate differential. To this end, we mine macro-data on Ghana for the period 2002 to 2019 to address the objectives. First, the study finds the nonexistence of the exchange rate-interest rate differential relationship in both the short-run and long-run. Further, we show a positive but slow responsiveness of exchange rate to interest rate differential shocks from the short-run to the medium-term. The long-run results, however, show a case of a strong and significant response of exchange rate to interest rate differential shocks. The study makes four main contributions. First, it cast doubt on the effectiveness of interest rate differential in improving the country’s currency within an IT framework, especially in the short-run. Secondly, by introducing a foreign price, we deepen understanding of the effect of commodity arbitrage on exchange rate movements. This study, therefore, provides an empirical answer to the debate on the practicability or otherwise of the sticky price model and the Dornbusch overshooting argument in Ghana. Thirdly, the study builds on prior empirical contributions1 on Ghana and makes some remarkable contributions to the existing literature on exchange rate movements and investment in IT regimes. Particularly, this study deepens the understanding as to whether shocks to the Ghana-USA interest rate differential have any short- to long-term implications for the country’s exchange rate movement since the IT framework was adopted. Fourthly and foremost, the study informs Ghana’s monetary authorities that, in inflation targeting regime, shocks to the interest rate have a somewhat weak impact on the value of the country’s currency. In the long term, however, interest rate movements have profound implications for the appreciation/depreciation of the GhanaCedi. The rest of the study is organised as follows. In Section 2, we present a brief literature review pertaining to the relationship between exchange rate and interest rate. We provide the methodological foundation of the study in Section 3 while Sections 4, 5 deal with the presentation of the results and the conclusion and policy recommendations.

2. Literature review

2.1. Ghana’s inflation targeting (IT) experience and exchange rate

The data shows that, with the exception of 1992 and 1999, the Bank of Ghana (BoG) missed its inflation targets in the non-inflation targeting (NIT) regimes. Two main reasons arise: (1) IT was not the headline monetary policy regime and as such a commitment to achieving it was not a strict goal; and (2) the failure of the economic recovery programme (ERP) to maintain a sound socioeconomic climate. The clear disparity in the end-of-period inflation and target inflation levels in the NIT regime even after the introduction of the structural adjustment programme (SAP) and ERP raised serious concerns about the internal and external macroeconomic stability of the country2 [12]. In Fig. 1, we provide the trend of average inflation outturn and targets to put our analyses for Ghana into perspective.

Fig. 1.

Trends in inflation targets, outturn and target misses in NIT and it regimes, Ghana, 1990–2020).

Panel A of Fig. 1 shows the target inflation levels and outruns in the NIT regime (1990–2001). Conspicuous in the NIT regime is an inflation outturn of 50.9% in 1995 compared to a target of 30%. Though the rates of inflation slowed down for the period 1996–2000, the outturns remained in the double digits (see Panel A, Fig. 1). The trend of inflation in the NIT regime meant that there was a need for a new regime in the last quarter of 2002. The country consequently switched to an IT regime although an official announcement was made in2007.3 During this era, 90 pesewas was exchanged for a dollar. Panel B of Fig. 1 shows that the story has however not changed significantly for the better since the switch. In Panel B of Fig. 1, it is clear that even in the IT regime, the BoG hardly achieves its inflation targets.4 In fact, with the exception of 2011, 2012 and 2018, inflation targets have proved elusive, raising serious concerns about the credibility of the BoG in managing a full-fledged IT regime.

In fact, inflation rates in Ghana have generally remained within a band of 10–20% instead of the target 5% ±2. Per the trend of inflation in Panel B of Fig. 1, it is difficult to argue that the IT regime, which has been in operation for about two decades,has indeed taken off. A point of a bailout for Ghana’s monetary authorities on this is the end-of-period inflation target band of ±2.5 Panel C of Fig. 1 shows that even with the allowance of the ±2 band, the BoG has missed its targets for most parts of the IT years. In specifics, BoG missed its targets with the ±2 allowance in 2001, 2003, 2004, 2006, 2009, 2013, 2014, 2015, 2016, and 2017. Further, Panel B of Fig. 1 shows an average inflation target of 10.6% since the regime was adopted. This development has culminated in consistent depreciation of the Ghana Cedi against major trading currencies, especially the US dollar. For instance, according to the BoG, the Ghana Cedi has depreciated by almost 300% to the US dollar in the past two decades.

2.2. The exchange rate-interest rate model

The closest theoretical framework for analysing the linkages between exchange rate and interest rate differential is the [3] overshooting model. The model is expressed in Equation (1) as

$$R=R^{*}+\frac{E^e-E}{E}$$ (1)

where in respective terms, we use $R$ and $R^{*}$ to denote the nominal interest rates of the home and foreign countries. Also, $E$ is used to signify the spot barter price of foreign currency in terms of home currency, and $E^e$ is used to capture the expectation of the exchange rate in the future. It is imperative to point out that the introduction of the expected rate of depreciation $\left(E^e-E\right)/E$ leads to interest rate comparability in the same currency. Assuming uncovered interest parity (UIP), investors move capital from a country with low interest to a country with a higher interest rate, leading to the appreciation of the price of the recipient country’s currency. Equation (1) may be rewritten as:

$$R-R^{*}=\frac{E^e}{E}-1$$ (2)

indicating the basis of the short-run inverse connection between the exchange rate and interest rate differential. In this model, a perpetual one-time loose monetary policy in the home country is associated in the short-run with a decrease in $R$ and a rise in $E^e$ and an upsurge in E that is higher than the increase in $E^e$ to ensure that Equation (2) remains in balance given a fixed $R^{*}$. It follows that concerning the short-run and long-run, as prices rise, the real money supply rises, increasing R, which also results in $E$ declining to keep Equation (2) in equilibrium. Hence, we observe an inverse association between $R-R^{*}$ and $E$ simultaneously, though, without ambiguity, there is also a higher likelihood for periods with lower $R-R^{*}$ to be related with E decreasing, assuming perfect foresight of the long-run exchange rate. The former negative relationship is an immediate short-term one, and the latter positive relationship may be conceived as a longer-term one that requires changing prices. Nonetheless, in the true long-run for the overshooting model, shocks to the stock of money have no impact on the interest rate differential, so they cannot induce any long-run relationship between the exchange rate and interest rate differential at that period.

2.3. Brief empirical literature survey

Earlier research which models exchange rate movement as a function of interest rate differential and other macroeconomic variables in advanced countries is evident in the literature.6 [4,13,14], [4,13,14], For example, [15] provides empirical evidence from the Johansen cointegration approach to show a long-run relationship between exchange rate and interest rate differential in the case of Germany and Japan. However, using the generalized method of moment(GMM) technique, [5] found no such evidence either in the short-run or long-run in the USA, Germany, Japan, and the United Kingdom. Moreover, [16] explored the relationship between exchange rate and interest rate differentials at different timescales and provided evidence to conclude that, over a year, the link between the two is negative at shorter time horizons and positive in longer horizons. The difference in these results is plausibly due to the use of different empirical techniques, the degree of capital mobility, the type of monetary framework, and economic structure.

Providing other alternatives, [17] argues that the failure of the sticky price model prediction in recent times is due to the failure to recognize the nonlinearity in the exchange rate adjustment.7 [18], [18], Using the wavelet analysis, [7] revisited the subject matter in Romania. They provided convincing evidence to show that the association between interest rate and exchange rate behaves differently in the short-run and long-run. In related work, [19] use the bivariate structural vector autoregressive (SVAR) approach in examining the possibility of a contemporaneous relationship between interest rate differential and change in the real exchange rate in twelve countries.8 The results show that out of the twelve countries, nine show the expected negative relationship of which there is empirical evidence for just three. Likewise, only three countries show evidence from the impulse response analyses that a positive real interest rate differential shock can generate a negative initial effect on the real exchange rate.

In examining how the exchange rate is linearly related to the interest rate differential, [20] found evidence consistent with UIP along with rational expectations that a positive relationship between the two variables was observable when the authors used the long-maturity bond data as the interest rate. Nonetheless, the relationship was not found to be non-existent when the short maturity bond data was used to capture interest rates. Their results on the direct relationship between exchange rate and interest rate based on the long-maturity bond data is aligned with the evidence by Ref. [10] who employed the medium-maturity bond data to study the aforementioned relationship.

Again, [21] contributed to the debate in the case of Pakistan by interrogating the effect of inflation, interest rate and money supply on exchange rate movements. In their contribution, the authors employed monthly data spanning July 2000 to June 2009 and the Johansen Cointegration, the Granger causality test, the impulse response function, and the vector error correction model (VECM) to point out the short-run and long-run effects. Evidence from the study revealed a strong short-run and long-run relationship between inflation, interest rate and exchange rate volatility. The authors argued that the increase in interest rate resulted from money supply bidding the price level/inflation up.

Furthermore, [22] interrogated the relationship between foreign exchange rate fluctuation and interest rates in the Nairobi securities market. The authors employed data for the period January 2006–December 2010, Johansen’s test for cointegration, and hierarchical regression for the empirical analysis. Compelling evidence from the study showed that interest rate movements and exchange rates have a strong relationship. Further, the study revealed that the interest rate explains the fluctuations in the exchange rate and the performance of the Nairobi security market.

In a study comprising 80 countries, [1] also used macro data stretching over the period 1974–2009 to examine the short-run and long-run relationship between the exchange rate and interest rates. Comprehensive evidence from the study revealed the relationship between the exchange rate and short-term interest rates is non-monotonic. In particular, the study showed that slight increases in the nominal interest rate appreciate the currency, whereas larger increases depreciate the currency. The authors argued that higher interest rates raise money demand, hence a decrease (appreciation) in the exchange rate. Nonetheless, the higher interest rate also bids up the fiscal deficit and decreases output, which tends to depreciate the currency.

Again, [23] contributed to the debate by examining the dynamic relationship between exchange rates and interest rates in India. By applying the asymmetric autoregressive distributed lag model, the authors provided strong evidence to show the presence of an asymmetric reaction of stock prices to changes in the interest rate and exchange rate in the pre-crisis periods. Nevertheless, this asymmetry evidence proved elusive in the post-crisis period. The authors provided further evidence to suggest that a reduction in money supply increases the interest rate, which in turn appears to retard the stock prices.

The subject matter has also been investigated by Ref. [24] in the BRIC-T (Brazil, Russia, India, China and Turkey) countries. The authors utilised monthly data from the beginning of the flexible exchange rate regime in each country to July 2011 for the analysis. The authors analysed the data using the non-linear causality test and frequency domain causality test approaches. Results from the frequency domain causality test results revealed that interest rate movements affect the exchange rate but only in China, and this effect exists only in the long-run. The authors also found evidence that exchange rate shocks lead to short-term fluctuations in interest rates.

In yet a similar contribution by Balassa-Samuelson, the authors provide strong evidence to show that the equilibrium exchange rate is determined both by real shocks such as monetary shocks and productivity innovations, which drive the relative demands and supplies of liquidity [ [[25], [26], [27]]]. Applying [28] cointegration technique, [29] finds a significant long-run relationship between of the U.S. dollar and the Canadian dollar bilateral exchange rate for the period 1974–1994.

3. Data and methods

3.1. Theoretical foundation

Following [13], we specify equation (3) that presents the exchange rate variability as chiefly influenced by interest rate differential, current account differential, and foreign price level.

$${\mathrm{q}}_{\mathrm{t}}=\propto +{\mathrm{\delta }}^{*}{\mathrm{r}\mathrm{d}}_{\mathrm{t}}+{{\mathrm{\omega }}^{*}\mathrm{c}\mathrm{d}}_{\mathrm{t}}+{\lambda p*}_{\mathrm{t}}+\mathrm{\epsilon }$$ (3)

where ${\mathrm{q}}_{\mathrm{t}}$ is the bilateral exchange rate; ${\mathrm{r}\mathrm{d}}_{\mathrm{t}}$ which is the interest rate differential; and ${\mathrm{c}\mathrm{d}}_{\mathrm{t}}$ is the current accounts differential. In addition, the increase in the domestic current account relative to that of the foreign country, results in the appreciation of the local currency.

3.2. Empirical strategy

First, we specify two models with equation (4) as a bivariate model per the thrust of the paper.

$$s_t={\alpha }_0+{\alpha }_2{ii}_t+{\mu }_t$$ (4)

We further specify equation (5) to incorporate current account differential to determine its effect on the exchange rate.

$$s_t={\alpha }_0+{{\alpha }_{1}s}_{t-1}+{\alpha }_2{ii}_t+{\alpha }_3{caca}_t+{\alpha }_4{p*}_t+{\mu }_t$$ (5)

Where $s_t$ is the cedi-dollar exchange rate in IT regime; ${ii}_t$ is the interest rate differential between Ghana and the USA in IT regime; ${caca}_t$ is the current account differential between Ghana and the USA in an IT regime; and ${p*}_t$ is the USA price level.

Second, coming from the background of the theorised short-run and long-run movements of the cedi-dollar exchange rate, the study applied the autoregressive distributed lag technique to equation (4)[see, [30]]. The ARDL is known to provide robust estimates with a smaller number of countries unlike the estimators like the System GMM, which tend to be more useful when the sample is characterised by a large number of countries. More so, the ARDL technique can also be applied to variables that are integrated of order one and order zero (I(1) & I(0)) or a mixture of the two [ [[31], [32], [33]]]. Based on the error correction model, it is feasible to introduce the dynamic regression applying the $ARDL\left(p,q\right)$ as presented in equation (6):

$$\Delta s_t={\delta }_0+\varnothing s_{t-1}+{\alpha }_1{ii}_{t-1}+{\alpha }_2{caca}_{t-1}+{\alpha }_3\mathit{ln}{p*}_{t-i}+{\sum }_{i=1}^{\rho }{\beta }_1\Delta s_{t-i}+{\sum }_{i=1}^{\rho }{\beta }_2\Delta {ii}_{t-i}+{\sum }_{i=1}^{\rho }{\beta }_3\Delta {caca}_{t-i}+{\sum }_{i=1}^{\rho }{\beta }_4\Delta {p*}_{t-i}+{\epsilon }_t\text{,}$$ (6)

where, $\varnothing and{\alpha }_i$ represent the long-run elasticities while ${\beta }_i$ are the short-run elasticities. Per intuition, the study expects an indirect relationship between exchange rate and interest rate differential in both the long-run and short-run. The same is expected of the link between the exchange and current account differential.

3.3. Unit root test

To address possible concerns regarding spurious regression, we follow [34] by subjecting the variables in this study to stationarity tests. The essence of the unit root test is to indicate whether a variable, the mean, variance and auto-covariance are independent of time. To this end, this study employed the Philips-Perron (PP) [35] and Augmented Dickey Fuller (ADF) [36] tests. These tests are similar but for the way in which they correct for the existence of autocorrelation in the residuals. The ADF and PP are premised on the null hypothesis that a variable is non-stationary against the alternative that there is not a unit root.

3.4. Test for cointegration

The next step involves an examination of whether there exists a long-run relationship between the variables, which we establish by applying the Bounds Test technique put forward by Ref. [30]. It is imperative to point out that the cointegration test is contingent on the presence of stationary covariates. This is so as two or more covariates are said to be cointegrated if each of the series taken individually is non-stationary with I(1), while their linear combination is stationary with I(0).

Having said that, the Bounds testing procedure, which essentially involves three steps, is carried out. In the premier step, we estimate Equation (6) where we rely on the Fisher test to examine whether jointly the underlying model is significant in explaining the relationship between exchange rate and interest rate differentials. We, thus, test the hypothesis that:

$$H_0:{\theta }_1={\theta }_2=0$$

$$H_1:{\theta }_1\ne {\theta }_2\ne 0$$

Next, based on two asymptotic critical bounds (i.e., upper bound and lower bound), the existence or otherwise of a long-run relationship is established. Specifically, if the F-statistic from the cointegration test is above the upper critical value, the null hypothesis of no long-run relationship is rejected. Contrariwise, we fail to reject the null hypothesis of no cointegration if the F-statistic falls below the lower critical values. Nonetheless, in a rare case where the F-statistic lies in between the upper and the lower critical values, the result becomes inconclusive.

In the second stage of the ARDL approach, once cointegration is established, the conditional ARDL (p, q₁, q₂, q₃) model for $s_t$ is estimated as in equation (7):

$$\Delta s_t={\delta }_0+\varnothing s_{t-1}+{\alpha }_1{ii}_{t-1}+{\alpha }_2{caca}_{t-1}+{\alpha }_3\mathit{ln}{p*}_{t-i}+{\epsilon }_t$$ (7)

This involved selecting the orders of ARDL (p, q₁, q₂, q₃) model in the variables using Akaike Information Criterion [37]. The third and the last step in the ARDL technique is to estimate an Error Correction Model (ECM), which computes an error correction term – a value signifying the speed of adjustments of the short-run dynamics to the long-run path following a shock to the system.

3.5. Error-correction model

As [30] argue, the ECM is a standard procedure for jointly analysing the association between the long-run and the short-run dynamics of a model. The ECM is anchored in the intuition that though there may be a cointegration among two or more covariates, in the short run, however, there may be a disequilibrium or deviation from the long-run path. The essence of the ECM mechanism, therefore, is to guide policy, for instance, on the proportion of the disequilibrium in the exchange rate that can be corrected in the next period.

Intuitively, the error correction term should be negative and statistically significant to allow for a total reconciliation of the short-run and long-run behaviour. Accordingly, we specify the ECM as in equation (8):

$$\Delta s_t={\delta }_0+\varnothing s_{t-1}+{\alpha }_1{ii}_{t-1}+{\alpha }_2{caca}_{t-1}+{\alpha }_3\mathit{ln}{p*}_{t-i}+{\sum }_{i=1}^{\rho }{\beta }_1\Delta s_{t-i}+{\sum }_{i=1}^{\rho }{\beta }_2\Delta {ii}_{t-i}+{\sum }_{i=1}^{\rho }{\beta }_3\Delta {caca}_{t-i}+{\sum }_{i=1}^{\rho }{\beta }_4\Delta {p*}_{t-i}+{\tau ECM}_{t-1}+{\epsilon }_t$$ (8)

From equation (8), $\tau$ represents the short-run dynamics coefficients of the model’s convergence to equilibrium. ${ECT}_{t-1}$ is the Error Correction term, whose absolute value determines the speed of adjustment of the model to long-run equilibrium when it is shocked.

3.6. VAR model

The final value we provide in this study is informing policymakers of the short-run to long-run response of exchange rate to a standard deviation shock to interest rate differential. To do this, we present three vector autoregression (VAR) models obtained from the general VAR($\rho )$ as seen in Equation (9).

$$y_t=\rho Y_{t-1}+{\omega }_0{x}_t+{\mu }_t$$ (9)

$y_t$ is the $K\times 1$ vector of endogenous variables; $\rho$ is the $K\times Kp$ matrix of coefficients; ${\omega }_0$ is the $K\times M$ matrix of coefficients; $x_t$ is the $M\times 1$ matrix of coefficients; ${\mu }_t$ is the $K\times 1$ vector of white noise innovations, and $Y_t$ is the $Kp\times 1$ matrix of outcome variables. Given a strong responsiveness of exchange rate to interest rate differential, a significant short-run to long-run impulse response is expected for a shock to interest rate differential.

3.7. Data and variable description

The study uses macro-data spanning 2002 to 2019. Data on nominal exchange rates, nominal interest rates, the current accounts, and consumer price indices of the two countries were sourced from the International Monetary Fund’s International Financial Statistics. The nominal exchange rate is captured as the bilateral cedi-dollar rate; the annual inflation rate is proxied by the consumer price indices for Ghana and the United States of America; the currents account is defined as the ratio of the current account to GDP while the treasury bill rates proxied the nominal interest rates.

4. Results and discussion

4.1. Preliminary results

In accordance with the requirement of the ARDL technique, we report results on the series of preliminary tests. In particular, we pay attention to test of stationarity and cointegration.

4.2. Summary statistics

In this section, we provide the descriptive statistics of the variables. The summary statistics show that but for the current account differential, all the variables have a positive average (Table A1). For instance, the average interest rate differential between Ghana and the USA is 19% while the average nominal exchange rate of the cedi to the dollar was 2.2. This former is obtained from the domestic interest rate of 23.289%, compared to 2.277% for that of the USA. Information gleaned from the standard deviations of all the variables also indicates a minimal variability in the series used.

4.3. Unit root tests

In this section, we present results concerning the test of the statistical properties of the variables. The results for these tests, with intercept, are provided in Tables A2 and A3 respectively. While in Table 2, results for the ADF unit root test are presented, results for that of the PP are reported in Table 3. It is imperative to note that the rejection of the null hypothesis of non-stationarity is based on the critical values [38]. From the results, all the variables are integrated at order one permitting the application of autoregressive techniques.

4.4. Test for cointegration

In this section, we provide evidence on the long-run relationship between exchange rate and interest rate differential. From Table A4, the F-statistics that the joint null hypothesis of lagged level variables is zero is rejected at a 5% level of significance. Since the calculated F-statistics of approximately 3.958 exceeds the upper bound’s critical value of 3.69, we conclude that there is evidence of cointegration among the variables.

4.5. Finding and discussion

In this section, the main regression estimates on the exchange rate and interest rate differential relationship are presented. We first pay attention to the presentation of our findings from the ordinary least squares and proceed with that of the autoregressive distributed lag techniques.

It is evident from our ordinary least squares estimate in Table A5 that the interest rate differential is statistically significant in causing an appreciation of the cedi (Column 2). This also implies that in an IT regime, the BoG is likely to affect real variables through its policy rate. This finding is at variance with that of [39] who provided evidence to show that the link is largely non-existent in 9 out of 12 countries examined. However, at a 90% confidence interval, we show that a 1% increase in current account differential leads to an approximate 0.002% appreciation of the cedi. This suggests a more productive Ghanaian economy with growing exports and falling levels of consumption of foreign goods. The second contribution of the paper is in the utilization of autoregressive techniques on the subject matter [see Columns 4–7]. In the short-run, the evidence suggests that the relationship between exchange rate, and interest rate differential and current account differential are even non-existence (Column 4). In Column (5), however, we show that the relationship between exchange rate and interest rate differential on the one hand, and exchange rate and current account differential on the other hand, are not statistically insignificant in IT regime. The result is in contrast with the finding of [7] who found a negative relationship between the two variables in the BRICS countries. The result, therefore, does not support the case of the sticky price argument that an increase in the nominal interest rate of Ghana relative to that of USA results in an appreciation of the Ghanaian cedi. Additionally, there is empirical evidence that in IT regime, previous year’s depreciation of the cedi fuels current’s depreciation by 0.4%. In Column (5), we find that current account differential suggests a 0.68 appreciation of the cedi relative to the dollar if current account balance of Ghana exceeds that of the USA by 1%, albeit statistically insignificant. This result is similar to the result obtained by Ref. [19] which made it clear that exchange rate movement is related to the current account both through the formation expectations about long-run equilibrium real exchange rate and through changes in the risk premium.

In models (6) and (7), we acknowledge the impact of commodity arbitrage on exchange rate movements by introducing the general price level of the USA as a proxy for foreign price. We show that both the short-run and long-run effects that arise in foreign price leads to a depreciation of the cedi relative to the dollar. The contemporaneous effect of a percentage rise in foreign price is 0.86 in the short-run compared to 0.02 in the long-run. This plausibly underscores the slow rate of economic performance as a chunk of the Ghanaian imports is about consumables. In economies like this, commodity arbitrage can increase even as foreign price rises resulting in marginal depreciation of the local currency. This results in a slump in demand for the Ghanaian cedi, causing the depreciation. The reliability of our estimates lies in the series of diagnostics tests the model passes. The results for the tests are presented in Table A6 and Fig. A2.

4.6. The responsiveness of exchange rate to interest rate differential shocks

On the third contribution of the paper, we provide results on the responsiveness of cedi-dollar exchange rate to a standard deviation shock in the interest rate differential in the IT regime. The findings are based on the vector autoregression estimator. Per the focus of the study, we concentrate on the estimates from Columns (1), (3) and (6) in Table A7.

Similar to the results from the ordinary least squared and autoregressive distributed lag techniques, we find evidence of no exchange rate and interest rate differential relationship (see model 1). From model (6), the interest rate differential effect on the exchange rate is present. There is strong empirical evidence that a 1% increase in the interest rate differential induces a 0.8% depreciation of cedi in the very short term. Also, the effect of current account differential is conventional and suggests that a 1% increase leads to a 0.5% appreciation of the cedi.

Supporting this with the impulse response functions, it is evident that from the short-run to the medium-term, the exchange rate does not respond significantly to a standard deviation shock to the interest rate differential. However, in the long-run, an inverse response is evident. Additionally, the exchange rate responds positively but slowly to a standard deviation shock in the current account differential from the short-run to the medium-term. The response of the exchange rate to a standard deviation shock in the current account differential in the long-run is also positive and relatively faster. The impulse response functions are presented in Fig. A3 of the appendices.

4.7. Theoretical and empirical contributions

The theoretical implication of this study is that we do not confirm the argument espoused in the sticky price model, we show that the Dornbusch overshooting argument prevails in Ghana. In terms of the empirics, information gleaned from the ARDL estimates and that of the VAR shows that in both the short-run and medium-term, the nominal exchange rate of Ghana does not respond significantly to a standard deviation shock to interest rate differential (with that of the USA). Nonetheless, in the long-run, an inverse response is present and notable. Also, we point out that Ghana’s exchange rate responds positively but slowly to shocks in the country’s current account differential from the short-run to the medium-term.

Finally, our model passes the diagnostic test of autocorrelation shown by the Cusum and Cusum squared as presented in Fig. A2 [A and B], the model stability test in Fig. A4 (see appendix), and the Langrange-Multiplier test for VAR shown in Table A8.

5. Concluding remarks and policy recommendations

This study set out to examine the relationship between Ghana’s exchange rate and interest differential with that of the USA since the country adopted the IT framework in 2007. To this end, we draw macrodata on the two countries for the period 2002–2019 for the analysis. Our empirical evidence, which is based on the ARDL estimation technique has generated some interesting findings. First, our evidence suggests that in both the short-run and long-run, the interest rate differential between Ghana and USA has no statistically significant effect on exchange rate movements (i.e., appreciation or depreciation of the cedi).

We conclude that the exchange rate and interest rate differential relationship in Ghana since the adoption of the IT framework is non-existent both in the short-run and long-run. There is, however, strong evidence of the effect of current account differential, and foreign price on the barter price of the local currency. Further, we find slow responsiveness of the exchange rate to interest rate differential and current account differential shocks both in the short-run and medium-term. In the long-run, however, we find a clear and strong positive impulse response of the exchange rate to both interest rate differential and current account differential. The finding shows a clear case of unattractive domestic interest rates to foreign investors raising central bank credibility issues even in IT regime. The result also shows the crucial effect of economic performance and foreign prices on the value of the local currency which should be an incentive for policymakers to prioritise pro-growth spending.

We recommend for the attention of the BoG that the insensitivity of exchange rate in the IT regime could be a result of perennial macroeconomic instability, especially inflation could fuel investment risk. Also, we recommend prudent monetary policy management geared towards the reduction of interest rates to boost economic activity. Second, the slow responsiveness of the exchange rate to interest differential could also be explained by the underdeveloped nature of Ghana’s financial market, which may feed into a loss of investor confidence in the country’s short-term and long-term capital/securities. Accordingly, we recommend that the BoG take steps to tighten up the financial market of the country. This can be also enhanced with effective consultation between the Ghana Stock Exchange and their relatively developed counterparts in Kenya and South Africa. Finally, the significant effect of foreign prices on exchange rate movements in Ghana is an indication of the imported inflation, which indeed cannot trigger significant in/outflow of capital between Ghana and the USA. Ghana’s macroeconomic policy managers should consider the main drivers of interest rate fluctuations and uncertainty in the Ghanaian economy. Doing so would go a long way to keep interest rate in check and by extension, stabilise the country’s exchange rate. Further, the Bank of Ghana should step up efforts aimed at stabilising the Ghana cedi by prudent monetary policy supervision, especially, by slowing down on the frequent changes in the policy rate. Prudent fiscal-monetary policy regulation is also imperative to ensure that a loose fiscal policy does not become counterproductive to interest rate stabilisation. Accordingly, we suggest that Ghana’s monetary authorities forecast the short- to long-term transmission and impact of variations in the country’s prime rate to the cedi-dollar exchange rate.

A major limitation of this study is that the analysis is limited to Ghana’s IT period. The study therefore does not elaborate on the exchange rate and interest rate differential linkages before and after the adoption of the IT framework. For future research, we suggest that other researchers incorporate how uncertainties posed by economic and health shocks like the coronavirus pandemic can alter the short-run and long-run exchange rate and interest rate differential relationship.

Author contribution statement

Mark Kojo Armah: Conceptualisation, Methodology, Analysis, Writing of script, Review & Editing, and Proofreading.t

Isaac Kwesi Ofori: Conceptualisation, Methodology, Analysis, Writing of script, Review & Editing, and Proofreading.

Francis Kwaw Andoh: Conceptualisation, Methodology, Analysis, Writing of script, Review & Editing, and Proofreading.

Funding statement

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Data availability statement

Data will be made available on request.

Declaration of interest’s statement

The authors declare no conflict of interest.

Appendix A.

Fig. A2.

Stability test for ARDL model.

Fig. A3.

Response of Exchange rate to interest rate differential shocks.

Fig. A4.

VAR stability graph.

Table A1.

Summary statistics.

Variable	Obs	Mean	Std.Dev.	Min	Max
Ca	18	−5.838	3.730	−12.492	1.332
ca*	18	−2.553	1.533	−5.817	0.150
I	18	23.289	9.270	10.500	45.000
i*	18	2.277	3.140	0.390	5.020
S	18	2.219	1.655	0.720	6.032
Ii	18	19.113	9.428	2.000	39.750
Caca	18	−3.285	3.733	−10.077	5.839
p*	18	2.282	2.465	−0.356	5.490

Note: Std Dev. represents Standard Deviation while Obs stands for Observation; s is nominal exchange rate; ii is interest rate differential; caca is current account differential; p* is foreign inflation rate; i is domestic interest rate; i* is foreign interest rate; ca is domestic current account balance; and ca* is the foreign current account balance.

Table A2.

Results of Unit Root Test with Trend and constant: ADF Test.

Level		First Difference
Variables	ADF-Statistics	Variables	ADF-Statistics
S	5.143[1.000]	s	−3.238[0.0179]**
Caca	−3.621[0.0054]***	$\Delta$ caca	−8.601[0.0000]***
Ii	−2.142 [0.2279]	$\Delta$ ii	−6.296 [0.0000]***
p*	−4.681[0.0001]***	$\Delta$ p*	−14.11[0.0000]***

∗**p < 0.01, **p < 0.05, *p < 0.1; Δ denotes the first difference; P-values in parenthesis.

Table A3.

Results of Unit Root Test with constant and trend: PP Test.

Level		First Difference
Variables	ADF-Statistics	Variables	ADF-Statistics
S	6.247[1.000]	s	−4.731[0.0001]***
Caca	−3.741[0.0036]***	$\Delta$ caca	−8.565[0.0000]***
Ii	−2.161[0.2208]	$\Delta$ ii	−6.296 [0.0000]***
p*	−4.731[0.0001]***	$\Delta$ p*	−16.71[0.0000]***

***p < 0.01, **p < 0.05, *p < 0.1; Δ denotes the first difference; P-values in parenthesis.

Table 4.

Bounds test results for cointegration.

	Critical Value Bound of the F-statistic
K	90% Level		95% Level		99% Level
	I(0)	I(1)	I(0)	I(1)	I(0)	I(1)
7	2.72	3.77	3.23	3.69	4.29	5.61
F-Statistics: F($s)$(S$|\mathrm{i}\mathrm{i},\mathrm{c}\mathrm{a}\mathrm{c}\mathrm{a},\mathrm{p}*$) = 3.958 **

Source: Authors' Computation (2022).

Table A5.

Effect of interest rate differential on exchange rate movements.

Variables	(1)	(2)	(3)	(4)	(5)	(6)	(7)
	OLS	OLS	OLS	ARDL (SR)	ARDL (LR)	ARDL (SR)	ARDL (LR)
Exchange rate (−1)				0.040		−0.861**
				(0.071)		(0.311)
Interest rate differential	0.172**	0.189**	0.037**		−0.812		0.041*
	(0.068)	(0.063)	(0.013)		(1.752)		(0.020)
Current account differential		−0.159*	0.018		0.685		0.015
		(0.081)	(0.017)		(1.522)		(0.022)
Foreign price			0.026***				0.026***
			(0.001)				(0.001)
Constant	−0.213	−1.137	−0.609**	−0.428		−0.583**
	(1.204)	(1.199)	(0.218)	(0.291)		(0.234)
Observations	16	16	16	16	16	16	16
R-squared	0.311	0.466	0.984	0.414	0.414	0.672	0.672

Standard errors in parentheses; ***p < 0.01, **p < 0.05, *p < 0.1; SR is short-run estimates; LR is long-run estimates.

Table A6.

Diagnostic tests for ARDL model.

Test	F/Chi Version	P-Value
Serial Correlation	1.338	0.2473
Normality	0.916	0.175
Heteroscedasticity	15.25	0.3613
CUSUM	–	Stable
CUSUMSQ	–	Stable

***p < 0.01, **p < 0.05, *p < 0.1.

Table A7.

VAR results for exchange rate and interest rate differential relationship.

Variables	(1) nominal exchange rate	(2) interest rate differential	(3) nominal exchange rate	(4) interest rate differential	(5) current account differential	(6) nominal exchange rate	(7) interest rate differential	(9) current account differential	(10) foreign price
Nominal exchange rate (−1)	1.132***	5.342	1.018***	2.502	−0.175	3.827	0.851***	2.135	8.191***
	(0.173)	(4.064)	(0.170)	(4.151)	(2.525)	(5.169)	(0.207)	(2.596)	(2.096)
Nominal exchange rate (−2)	−0.008	−6.488	0.089	−3.654	−0.184	−4.727	−0.172	5.910**	−1.368
	(0.198)	(4.655)	(0.191)	(4.661)	(2.835)	(5.784)	(0.232)	(2.906)	(2.346)
Interest rate differential (−1)	0.0004	0.759***	0.001	0.819***	0.028	0.876***	0.002	−0.035	0.079
	(0.007)	(0.169)	(0.006)	(0.163)	(0.099)	(0.165)	(0.006)	(0.083)	(0.067)
Interest rate differential (−2)	−0.001	0.046	−0.002	−0.021	−0.067	−0.0251	−0.001	−0.086	−0.066
	(0.006)	(0.163)	(0.006)	(0.158)	(0.096)	(0.156)	(0.006)	(0.078)	(0.063)
Current account differential (−1)			−0.016	0.233	0.501***	0.360	−0.008	0.242*	−0.090
			(0.010)	(0.246)	(0.150)	(0.268)	(0.010)	(0.135)	(0.109)
Current account differential (−2),			−0.013	−0.527**	−0.015	−0.524**	−0.010	−0.076	−0.107
			(0.011)	(0.267)	(0.163)	(0.265)	(0.010)	(0.133)	(0.107)
Foreign price (−1)						−0.500	0.0123	0.152	0.968***
						(0.541)	(0.021)	(0.272)	(0.220)
Foreign price (−2)						0.565	−0.001	−0.414*	−0.024
						(0.500)	(0.020)	(0.251)	(0.203)
Constant	0.0376	4.336**	−0.0172	3.968*	−0.432	3.280	−0.050	0.779	−0.049
	(0.0889)	(2.090)	(0.0842)	(2.049)	(1.247)	(2.081)	(0.083)	(1.045)	(0.844)
Observations	18	18	18	18	18	18	18	18	18

Standard errors in parentheses.

***p < 0.01, **p < 0.05, *p < 0.1.

Table A8.

Lagrange-multiplier test for VAR model.

Test	Chi Statistic	P-Value
Serial Correlation	25.6027	0.429
Normality	26.9935	0.356

***p < 0.01, **p < 0.05, *p < 0.1.