Macroeconomic Variables and Exchange Rate Fluctuations in Ghana: An ARDL Approach

1. Introduction

There are a variety of factors contributing to exchange rate fluctuations in countries and examples of such factors include terms of trade, external debt, inflation, interest rate, current account deficits, foreign direct investments, political stability, the exchange rate regime and the independence of the central bank. The degree of impact of each of these factors varies and depends on every country’s economic conditions. Some countries are more vulnerable to exchange rate fluctuations than others and those are countries in the process of transition such as Ghana. This paper analyses the impact of these factors on the exchange rate in Ghana. Exchange rate stability protects the purchasing power of low-income households and prevents imported inflation from eroding real wages which have consequences such as growth in unofficial economic activities (see Nchor et al 2016).

Moffett et al. (2017) describe four types of exchange rate in their study: fixed exchange rate, managed floating exchange rate, freely floating exchange rate and pegged exchange rate. They defined fixed exchange rate as that which is controlled by government, using a country’s reserves for a specific time period. They also defined managed floating exchange rate as the rate based on the demand and supply of specific currencies but somehow under the influence of government. Freely floating rate is the exchange rate that is determined by the forces of demand and supply without any government interference. The last but not the least is pegged exchange rate where the domestic currency is pegged to the currency of another country. Ghana practises the managed floating exchange rate regime.

Before the Economic Recovery Programme in Ghana, Ghana practised a fixed exchange rate regime with occasional devaluation, and exchange rationing. By 1988, the country adopted the managed floating exchange rate regime. The national currency, the Cedi (¢) has experienced instability for the most part of this period after 1988. Against this backdrop, one of the most important issues in Ghana’s macroeconomic policy over the years has been that of exchange rate stabilisation. Exchange rate volatility is associated with unpredictable movements in relative prices in the economy and such unpredictability has serious consequences for stable economic growth.

The research question which this paper seeks to answer is: what are the forces behind the movement of the Ghana Cedi on the foreign exchange market? The paper also investigates whether exchange rate movement in Ghana follows any seasonal pattern. The econometric technique of an autoregressive distributed lag model is applied. The choice of this approach is due to the fact that, the modelled variables are integrated of different orders and also cointegrated. Autoregressive distributed lag model (ARDL) has different advantages as compared to traditional cointegration methods. This approach makes it possible to analyse short run and long run relationships of variables regardless of whether the said variables are stationary at level or after first difference. This approach also estimates better properties of small samples (Pesaran & Shin, 1999). The modelled variables include the real effective exchange rate, external debt, current account deficit, inflation, interest rate, political stability, foreign direct investment and terms of trade. Several authors have attempted papers in this area, examples include Bussiere et al. (2010), Lenz and Savioz (2009), Adler and Grisse (2014). The determinants measured and the conclusions each study derives depend on the conditions of the country in question, but some determinants are common causes in most countries.

Exchange rate depreciation remains one of the most challenging macroeconomic problems that the Ghanaian economy faces. Yearly budget policy statements of government highlights the need for the stabilization of the Ghana Cedi in the exchange market. There is a plethora of descriptive literature on the exchange rate in Ghana but investigation of the exact determinants using econometric models has rather been scanty. This is due to the fact that research on the dynamic behaviour of exchange rate is considered technical and challenging. Also, the problem of data availability makes it further harder thus reducing research interest in this subject. This study reviews existing literature on the exchange rate of Ghana and most are qualitative documents prepared by the bank of Ghana. Others are reports which do not quantify the exact impact of factors causing exchange rate behaviour in Ghana. This study thus, focuses on multiplicity of determinants. Knowing the importance of the findings, the study overlooks the cost and obstacles and focuses on producing results that are relevant to policy makers and academia. The econometric approach (ARDL) that this study applies is traditional but rather unique in the context of Ghana’s exchange rate. The study considers the period from 1988-2024 because it is the period within which Ghana adopted the managed floating exchange rate regime. The period before was characterised by fixed exchange rate regimes and disruptions to the smooth flow of the system through government interference hence not suitable for modelling with several structural break issues.

The first chapter of this paper focuses on the introduction which covers the background on the field of exchange rate in general and specifically in Ghana. It also clearly outlines the objective and the structure of the paper. The second chapter deals with the methodology and data issues. It provides explanation on the research questions and the model the paper uses. It also provides more information on the data, data sources and data transformation techniques. The third chapter interprets the results of the paper. This section also contains a discussion of the findings where comparison is made with the findings of existing studies. The last chapter focuses on the conclusion which summarises the findings of the paper.

2. Materials and Methods

The study applies an autoregressive distributed lag model. This approach is popularized in the works of Pesaran and Shin (1999). The main advantage of this approach is that it is applied irrespective of whether the variables are I(0) or I(1). Another advantage is that the ARDL approach does not require symmetry of lag lengths. The main research questions this paper seeks to answer is what determines the exchange rate in Ghana. The paper also seeks to find out whether there is a seasonal behaviour in exchange rate movement in Ghana. The modelled variables include the real effective exchange rate, external debt, terms of trade, interest rate, inflation, foreign direct investment, political stability and current account deficit. Interest rate, political stability and terms of trade could not be maintained in the model due to issues of collinearity. External debt, foreign direct investment and current account deficit are all measured as percentages of GDP. Inflation is measure as a percentage and it is Consumer Price Index inflation. The data is annual and covers the period from 1988 to 2024. The data is obtained from Global Economy data.

2.1. Data Transformation

To ensure that, the analysis is appropriate and devoid of time series issues, this study checks the data for unit roots. The study performs the Augmented Dickey Fuller (Dickey Fuller, 1979) unit root test. The null hypothesis of this test is that variables are non-stationary. Using a significance level of 5%, the null hypothesis is rejected if the probability value (p value) from the test is less than 0.05. The results obtained show that some variables are stationary at level and others stationary after first order differences.

The study then proceeds to test for the optimal lag length for the model using the various information criteria: Akaike (1979), Hannan-Quinn (1979) and Schwarz (1978). Schwarz criterion settles for lag 1 as optimum while Akaike, Hanann-Quinn and the Final Prediction Error criteria settle for lag 2. The study then proceeds to check if there exists a long run relationship among the study variables or whether the variables are cointegrated. The study applies the Johansen test (Johansen, 1988) of cointegration. The results show that variables are cointegrated with a rank of cointegration of 1. Such a condition is appropriate for an Autoregressive Distributed Lag Model (ARDL) as applied in the studies of Atkins and Coe (2002), Nchor et al. (2015) and Abbot et. al (2001). These studies apply ARDL models to investigate exchange rate volatility and its causes in different countries. The basic equation of the ARDL model is given in equation 1.

$${REER}_t={\beta }_0+{\beta }_1{REER}_{t-1}+\dots +{\beta }_a{REER}_{t-a}+{\gamma }_0{EXTD}_t+\dots +{\gamma }_b{EXTD}_{t-i}+{\delta }_0{CRR}_t+\dots +{\delta }_c{CRR}_{t-k}+{\tau }_0{INF}_t+\dots +{\tau }_d{INF}_{t-l}+{\phi }_0{FDI}_t+\dots +{\phi }_e{FDI}_{t-m}+{\epsilon }_t$$

(1)

where the random disturbance term ${\epsilon }_t$ is serially independent. The ARDL model for the variables is estimated as in Equation (2).

$${REER}_t={\beta }_0+{\displaystyle \sum _{i=1}^n}{\beta }_i{∆REER}_{t-1}+{\displaystyle \sum _{j=1}^q}{\gamma }_j{∆EXTD}_{t-j}+{\displaystyle \sum _{l=1}^x}{\delta }_l{∆CRR}_{t-l}+{\displaystyle \sum _{s=1}^y}{\tau }_s∆{INF}_{t-s}+{\displaystyle \sum _{f=1}^w}{\phi }_f{∆FDI}_{t-f}+{\theta }_0{REER}_t+{\theta }_1{EXTD}_t+{\theta }_2{CRR}_t+{\theta }_3{INF}_t+{\theta }_4{FDI}_t+{\mu }_t$$

(2)

In equation 2, expressions with summation signs represent short run dynamics. Long run elasticities are coefficients of the lagged explanatory variables (${\theta }_1$, ${\theta }_2$, ${\theta }_3$, and ${\theta }_4$) multiplied with a negative sign and divided by the coefficient of the lagged dependent variable (${\theta }_0$,). The appropriate number of lags for each variable in the model is detected automatically by the program procedure of the Schwarz criterion. The automated program procedure leaves only significant lags while the insignificant ones are removed. Furthermore, the Schwarz criterion is more suitable for this kind of analysis due to a relatively short estimation period and because it penalizes models with a larger number of independent variables more than the Akaike Criterion does.

${\beta }_0$ is a constant term. ${\gamma }_j$, ${\delta }_l$, ${\tau }_s$, and ${\phi }_f$ are short run coefficients. ${\theta }_0$, ${\theta }_1$, ${\theta }_2$, ${\theta }_3$, and ${\theta }_4$ are long run coefficients. $∆$ represents first order difference. From estimating the model in Equation (2), an F test of the null hypothesis $H_0:{\theta }_0={\theta }_1={\theta }_2={\theta }_3={\theta }_4$=0 is performed to determine if the variables that have long run coefficients are statistically significant. If the independent variables are statistically significant and co-integrated, then a normal error correction model (ECM) is used to estimate the given relationships. This is given in Equation 3

$${∆REER}_t={\beta }_0+\sum _{j=1}{\beta }_j{∆REER}_{t-j}+\sum _{j=1}{\gamma }_j{∆EXTD}_{t-j}+\sum _{j=1}{\delta }_j{∆CRR}_{t-j}+\sum _{j=1}{\tau }_j{∆INF}_{t-j}+\sum _{j=1}{\phi }_j{∆FDI}_{t-j}+\omega Z_{t-1}+{\mu }_t$$

(3)

In equation (3), $\omega$ is the speed of adjustment parameter; a positive $\omega$ indicates a divergence, and a negative $\omega$ indicates convergence. In the ECM, the speed of adjustment parameter and the short run coefficient estimates (coefficient estimates of all lagged first-differenced variables: ${\beta }_j$ , ${\gamma }_j$ , and ${\delta }_j$ , ${\tau }_j$ and ${\phi }_j$) are directly estimated. The long-run elasticity of the real exchange rate with respect to the individual variables is given by: $-\left({\displaystyle \frac{{\theta }_0}{{\theta }_1}}\right)$, $-\left({\displaystyle \frac{{\theta }_0}{{\theta }_2}}\right)$, $-\left({\displaystyle \frac{{\theta }_0}{{\theta }_3}}\right)$, $-\left({\displaystyle \frac{{\theta }_0}{{\theta }_4}}\right)$ respectively.

3. Results

This section discusses the findings of the study. It includes tables, graphs and econometric models. Interpretations are given to each result in a simple and concise manner.

3.1. Model Results

The results from the ARDL model in Table 1 suggest that the exchange rate of the Ghana cedi is determined by current account deficits, external debt, foreign direct investment and commodity terms of trade. Current account deficits and external debt servicing cause the Ghana cedi to depreciate whereas foreign direct investment inflows and terms of trade cause it to appreciate. The results also indicate that lagged exchange rate representing the value of exchange rate from previous period cause the depreciation of the Ghana cedi. The coefficient of determination (R squared) is 0.96 indicating that 96% of the variations in Ghana’s exchange rate is explained by the modelled variables. The error correction term (-0.30) is found to be negative and highly significant, indicating an adjustment of 30% of deviation from equilibrium in one period.

3.2. Discussion

The paper investigates the relationship between macroeconomic variables and the exchange rate in Ghana. The findings show that in the short run, the behaviour of the cedi relative to major international currencies is largely accounted for by external debt, current account deficits, foreign direct investment and past exchange rate. In the long run, it is accounted for by external debts, foreign direct investment and terms of trade. The paper also finds that the depreciation of the Ghana Cedi is not a seasonal phenomenon. These observations are similar to the conclusions in the study of Mumuni and Owusi-Afriyie (2004) who conclude that the depreciation of Ghana’s currency is not seasonal in nature. The relationship between the modeled variables established in this paper is consistent with the findings of Benazic and Kersan-Skabic (2016) who conclude that, the volality of exchange rate in Croatia is caused by external debt and weak the terms of trade. The findings also agree with Calvo et al. (2003) who conclude that high level of external debt was partly responsible for the high swings of the real exchange rate of the Argentine peso in 2002. Finally, the findings of this paper are in sync with Corsetti et al. (1999) that external borrowings are key factors responsible for the East Asian currency crises during the late 1990s.

The paper failed to establish a direct relationship between inflation and the exchange rate which differs from studies such as Engel and Rogers (2001), Kulkarni and Ishizaki (2002), Arghyrou and Pourpourides (2016); Ebiringa and Anyaogu (2014); Nucu (2011), (Necșulescu and Șerbănescu 2013; Namjour et al. 2014) who conclude that rising inflation reduces the competitiveness of a country’s exports thus leading to depreciation of its currency. This finding on current account deficits agrees with the conclusion of Croke et. al (2005) who study current account in industrial countries and find that deficits tend to be associated with real exchange rate depreciations. Lastly, this paper finds a significant impact of past exchange rate on the behaviour of the Ghana cedi and this corresponds with the findings of Yan (2009), Dornbusch (1988) and Eichengreen and Wypolsz (1993) who argue in support of active role of speculation on the exchange rate system volatility.

4. Conclusion

The paper seeks to investigate the relationship between macroeconomic variables and the exchange rate in Ghana. Secondly, the paper clarifies the question of seasonality of exchange rate depreciation in Ghana. An ARDL model is applied. The study variables include the real effective exchange rate, external debt, current account deficit, inflation, foreign direct investment, interest rate, political stability and terms of trade. Variables such as interest rate, political stability and terms of trade are dropped due to issues of collinearity. The paper finds that depreciation of the Ghana cedi is accounted for by external debt accumulation and servicing, current account deficits and past exchange rate. On the other hand, foreign direct investment inflows and improved terms of trade tend to cause the cedi to appreciate. The speed of adjustment parameter is negative and statistically significant implying that deviations from long run equilibrium are corrected. The adjustment is however slow (-0.30) where only 30% of deviations are corrected in a year.

Reducing external debt in Ghana will help to stabilize the Ghana cedi by lowering foreign exchange demand, boosting investor confidence, and protecting the economy from global shocks. A current account deficit happens when a country spends more money on foreign goods, services, and transfers than it earns from selling its own abroad. This will only increase pressure on the Ghana cedi, therefore, focus should be on lower the deficit. Also, improving the terms of trade which is the ratio of Ghana’s export prices to its import prices will help stabilize a the Ghana cedi by increasing foreign currency earnings and reducing the need to drain foreign reserves. Finally, policymakers should focus on increasing FDI inflows because those inflows are in foreign currency and directly increase the demand for the Ghana cedi in the exchange market leading to a strengthened currency.