^{1}

^{*}

^{2}

^{*}

The study examined the causal relationship between export and economic growth based on monthly data for the period 2010 to 2019. Composite Index Economic Activity (CIEA) was used as a proxy for real GDP (economic growth). Time series econometric techniques were employed to explore long-run and short run relationships as well as the causality between them. The results revealed the existence of long-run relationships; evidence of bi-directional causality and a rapid adjustment to equilibrium between real GDP and exports. Recommendations are that, policy makers should focus on implementing export oriented policies and promote economic growth to achieve sustainable development.

The study takes another look at an age-old debate, “the export and economic growth nexus”. This debate has its source in literature since the 1970s with a large number of empirical studies conducted using either annual time series data or cross-section data. These studies on the relationship between exports and economic growth were carried out with annual export data and annual data on Gross Domestic Product (GDP) or annual economic growth rate, [

Economic growth is usually used as a yardstick in gauging economic performance. This has driven many nations in search of dynamic economic policies aimed at boosting their Gross Domestic Product (GDP) levels in achieving economic growth. Gross Domestic Product (GDP) is only available as a quarterly frequency, hence extra variables are necessary to establish a monthly chronology. The study adopted the Composite Index Economic Activity (CIEA) as proxy for monthly real GDP. The CIEA is a single summary statistic that tracks the current state of the economy [

The primary purpose of this study is to reexamine both the short and long run relationship and causality between exports and economic growth using monthly data set of Ghana. As a contribution to literature, this study used CIEA as a proxy of monthly real GDP and economic growth. The study aims also to help bridge the empirical knowledge gap by examining the relationship between export and economic growth in isolation rather than as part of a bloc of sub-Saharan African countries or developing countries. The results will be helpful for policy-makers since it is the existence of a long-run association between these two variables that may guarantee sustainable economic development.

There are a number of studies which have been carried out to find the relationship between economic growth and exports using annual data. Some studies have shown that export has increased economic growth. On the other hand, some studies have shown little evidence to prove strong relationship between exports and economic growth.

For instance, studies done by Ram [

Afxentiou and Serletis [

A study in Brazil examined the relationship between export and growth for the period 1960-2017 using annual data and found a bi-directional causality in the long run and export-led economic growth in the short run [

Besides, in another comprehensive study by Medina-Smith [

Hailegiorgis [

On the contrary, study carried out by Jung and Marshall [

Similar to the study of Helleiner [

An empirical research carried out by Kibria and Hossain [

Bahmani-Oskooee and Alse [

Another study in Turkey, used quarterly data from 1980 to 2004 to analyse the dynamic relationship between export growth and economic growth. The empirical research conducted showed a long term Granger causality exists from economic growth to export growth. Again error correction analysis confirms bi-directional short run relationship [

Additional study that used quarterly data, was by Lee and Huang [

There are other studies proposing a reverse relationship to the export led hypothesis, that is, economic growth induces trade flows. According to Bhagwati [

In Tanzania Dimoso and Utonga [

New insights into export-growth nexus, have been undertaken by Adebayo [

Ascertaining the direction of causation is important for policy connotations in developing strategies. For Ghana, Attah [

Twumasi-Ankrah and Wiah [

An empirical research by Owusu [

Tuffour [

In 2015, Tetteh [

The object of the study is to explore the dynamics of the relationship between exports and economic growth in Ghana using the monthly data for the period 2011 to 2019. The variables used in this study were total exports and economic growth. Total exports (EXP), approximates the total value of goods and services made in Ghana but sold abroad, while Composite Index Economic Activity (CIEA) approximates monthly real GDP which was used as a proxy for economic growth. The data for the study was obtained from Bank of Ghana.

Cointegration and error correction modelling techniques were the estimation methodology employed. The estimation procedure is as follows: unit root test, cointegration test and the error correction model estimation.

Macroeconomic time series is susceptible to non stationarity which causes regression results to suffer from spurious regression problem [

Δ Y t = α 0 + α 1 t + α 2 Y t − 1 + ∑ j = 1 p α j Δ Y t − j + ε t (1)

The ADF procedure, test for a unit root is conducted on the coefficient of Y t − 1 in the regression. If the coefficient is significantly different from zero, then the hypothesis that Y_{t} contains a unit root is rejected. The rejection of the null hypothesis implies stationarity. The null hypothesis is that the Y_{t} is a non-stationary series ≡ H 0 : α 2 = 0 with the alternative being H α : α 2 < 0 .

If the calculated value of ADF statistics is higher than McKinnon’s critical values, then the null hypothesis (H_{0}) is not rejected and the series is non-stationary or not integrated of order zero, I(0). Failure to reject the null hypothesis leads to conducting the test on the difference of the series, so further differencing is conducted until stationary is reached and the null hypothesis is rejected. If the time series (variables) are non-stationary in their levels, they can be integrated with I(1), when their first differences are stationary.

Engle and Granger [

In the Johansen framework, the first step is the estimation of an unrestricted, closed p^{th} order VAR in k variables. The VAR model as considered in this study is:

Y t = A 1 Y t − 1 + A 2 Y t − 2 + ⋯ + A p Y t − p + B X t + ε t (2)

where Y_{t} is a k-vector of non-stationary I(1) endogenous variables, X_{t} is a d-vector of exogenous deterministic variables, A 1 , ⋯ , A p and B are matrices of coefficients to be estimated, and ε t is a vector of innovations that may be contemporaneously correlated but are uncorrelated with their own lagged values and uncorrelatd with all of the right-hand side variables.

The stated VAR model is generally estimated in its first-difference form as:

Δ Y t = Π Y t − 1 + ∑ i = 1 p − 1 Γ i Δ Y t − i + B X t + ε t (3)

where, Π = ∑ i − 1 p A i − I , and Γ i = − ∑ j = i + 1 p A j

The Johansen approach to cointegration test is based on two test statistics, namely, the trace test statistic and the maximum eigenvalue test statistic.

The trade test statistics can be specified as: τ t r a c e = − T ∑ i = r + 1 k log ( 1 − λ i ) , where λ i is the i^{th} largest eigenvalue of matrix Π and T is the number of observations. In the trace test, the null hypothesis is that the number of distinct cointegration vector(s) is less than or equal to the number of cointegration relations (r).

The maximum eigenvalue test examines the null hypothesis of exactly r cointegrating relations against the alternative of r + 1 cointegrating relations with the test statistics:

τ max = − T log ( 1 − λ r + 1 ) , where λ r + 1 is the (r + 1)^{th} largest squared eigenvalue. In the trace test, the null hypothesis of r = 0 is tested against the alternative of r + 1 cointegrating vectors.

The general form of the VECM is as follows:

Δ X t = α 0 + λ 1 E C t − 1 1 + ∑ i = 1 m α i Δ X t − i + ∑ j = 1 n α j Δ Y t − j + ε 1 t (4)

Δ Y t = β 0 + λ 2 E C t − 1 2 + ∑ i = 1 m β i Δ Y t − i + ∑ j = 1 n β j Δ X t − j + ε 2 t (5)

where Δ is the first difference operator, E C t − 1 is the error correction term lagged one period; λ is the short-run coefficient of the error correction term (−1 < λ < 0); and ε is the white noise. The error correction coefficient (λ) is very important in this error correction estimation as the greater coefficient indicates higher speed of adjustment of the model from the short-run to the long-run.

The error correction term represents the long run relationship. A negative and significant coefficient of the error correction term indicates the presence of long-run causal relationship. If both the coefficients of error correction terms in both the equations are significant, this will suggest the bi-directional causality. If only λ_{1} is a negative and significant, this will suggest a unidirectional causality from Y to X (GDP to exports), implying that GDP drives exports towards long-run equilibrium, but not the other way around. Similarly, if λ_{2} is negative and significant, this will suggest a unidirectional causality from X to Y (exports to GDP), implying that exports drives GDP towards long-run equilibrium but not the other way around. On the other hand, the lagged terms of ΔX_{i} and ΔY_{i} appeared as explanatory variables, indicating a short-run cause and effect relationship between the two variables. Thus, if the lagged coefficients of ΔX_{i} appear to be significant in the regression of ΔY_{i}, this will mean that X causes Y. Similarly, if the lagged coefficients of ΔY_{i} appear to be significant in the regression of ΔX_{i} this will mean that Y causes X.

A number of arguments have been put forward by different studies concerning the potential contribution of exports to economic growth. To solve this complex issue, the study uses the methodology proposed by Granger. Testing for causality between exports and economic growth in the granger sense involves, whether lagged information on economic growth provides any statistically significant information about export in the presence of lagged exports. If not, then economic growth does not Granger-cause exports. A simple Granger causality test involving two variables, exports and GDP can therefore be specified as:

lGDP t = α 0 + ∑ j = 1 p α j lGDP t − j + ∑ j = 1 q β j lEX t − j + ε t (6)

lEX t = b 0 + ∑ j = 1 p b t lEX t − j + ∑ j = 1 q λ j lGDP t − j + ν t (7)

where, lGDP represents natural log of real gross domestic product (as a measure of economic growth) and lEX represents natural log of real exports. ε_{t} and V_{t} are serially uncorrelated white noise error term; the coefficients α, β, b, λ are expressing the short-run dynamics of the model’s convergence to equilibrium; and p and q are lengths for each variable in each equation. The null hypothesis to be tested is:

H 0 : α j = 0 , j = 1 , ⋯ , q , Export growth does not cause GDP growth.

H 1 : b t = 0 , t = 1 , ⋯ , q , GDP growth does not cause Export growth.

The analysis of the data and interpretation of the results of the study are presented in

The commonly accepted Augmented Dickey-Fuller unit root test was adopted to stationary test of exports and CIEA series. The test results are shown in

The major concern of the VAR model is to determine lag intervals for endogenous. The larger the lag intervals for endogenous is the more it can entirely reflect the dynamic nature of the model. However, more parameters will be needed to be estimated to constantly reduce degrees of freedom of the model. Considering

CIEA | EXPORT | |
---|---|---|

Mean | 7.958032 | 1056.313 |

Median | 6.590669 | 1057.532 |

Maximum | 25.15602 | 1439.350 |

Minimum | −5.173353 | 513.5700 |

Std. Dev. | 5.829357 | 221.4420 |

Skewness | 0.628178 | −0.255167 |

Kurtosis | 3.083517 | 2.325697 |

Jarque-Bera | 7.927023 | 3.575628 |

Probability | 0.018996 | 0.167326 |

Sum | 954.9638 | 126,757.5 |

Sum Sq. Dev. | 4043.787 | 5,835,350. |

Observations | 120 | 120 |

Variables | Test Statistics | Critical | Critical | Critical | P-Value |
---|---|---|---|---|---|

Value 1% | Value 5% | Value 10% | |||

CIEA | −1.379462 | −2.586753 | −1.943853 | −1.614749 | 0.1551 |

LNEXP | 0.686152 | −2.584707 | −1.943563 | −1.614927 | 0.8625 |

Variables | Test Statistics | Critical | Critical | Critical | P-Value |
---|---|---|---|---|---|

Value 1% | Value 5% | Value 10% | |||

D(CIEA) | −8.900595 | −2.585226 | −1.943637 | −1.614882 | 0.0000 |

D(LNEXP) | −14.95897 | −2.584707 | −1.943563 | −1.614927 | 0.0000 |

selection of lag intervals for endogenous, the study adopted the lowest AIC value as primary concern. It can be deduced from

As shown in

Cointegration relationship between DCIEA and DEXP has been investigated using the Johansen technique.

Normalized Cointegration Coefficients:

lCIEA t − 1 = 6.440 lEXP t − 1 − 0.047 (8)

The cointegration equation has been normalized for lCIEA to get a meaning from the coefficients as indicated in

Lag | LogL | LR | FPE | AIC | SC | HQ |
---|---|---|---|---|---|---|

1 | −235.4318 | NA | 0.256235 | 4.314086 | 4.411727* | 4.353696* |

2 | −229.6872 | 11.07523 | 0.248319 | 4.282652 | 4.477933 | 4.361872 |

3 | −226.7900 | 5.481050 | 0.253341 | 4.302523 | 4.595445 | 4.421353 |

4 | −221.2978 | 10.19290* | 0.246691* | 4.275635* | 4.666198 | 4.434075 |

5 | −219.3950 | 3.462594 | 0.256314 | 4.313424 | 4.801628 | 4.511474 |

6 | −214.1920 | 9.281050 | 0.250998 | 4.291748 | 4.877593 | 4.529408 |

7 | −210.5568 | 6.353459 | 0.252907 | 4.298321 | 4.981806 | 4.575590 |

8 | −208.6482 | 3.266962 | 0.262972 | 4.336004 | 5.117130 | 4.652883 |

*indicates lag order selected by the criterion; LR: sequential modified LR test statistic (each test at 5% level); FPE: Final prediction error; AIC: Akaike information criterion; SC: Schwarz information criterion; HQ: Hannan-Quinn information criterion.

Hypothesized | Eigen | Trace | Critical Value | Maximum | Critical value |
---|---|---|---|---|---|

Number of cointegrating | Value | Statistics | at 5% | Eigen | at 5% |

Equations | (p-value) | Statistics | (p-value) | ||

None*^{ } | 0.276 | 68.455 | 15.495 (0.000) | 36.815 | 14.265 (0.000) |

At most 1*^{ } | 0.242 | 31.64 | 3.841 (0.000) | 31.64 | 3.842 (0.000) |

injection to the circular flow of income and an increase in their level results in the expansion of the aggregate demand hence national output.

On the premise of the existence of cointegration relationships, Vector error correction model has been established. The error correction coefficient of −2.04

1 Cointegrating Equation(s): | Log likelihood | −244.0608 | |
---|---|---|---|

D(CIEA) | D(LNEXP) | ||

1.000000 | 6.439905 | ||

(3.79410) | |||

D(CIEA, 2) | −2.039081 | ||

(0.34079) | |||

D(LNEXP, 2) | −0.040274 | ||

(0.01264) |

implies high speed of adjustment of the model from the short-run to the long run. Again since it is significant and negative, it indicates the presence of long-run causal relationship which bi-directional. Both coefficients of error correction terms are significant and negative so it can be said that, export drives economic growth (CIEA) and vice versa.

From ^{2} > 0.5, and AIC and SC criteria values are relatively small, which indicates the reasonability of the model estimation. The overall fit of 0.722 indicates 72.2% of the systemic variation in economic growth is explained by the ECM. Again, it can be justified that the lagged variables are jointly significant in explaining the short run variations in economic growth at 5% significant level (F-statistic = 0.000).

The next step after going through the short run dynamics, is to perform the Wald coefficient test to establish the causality of the independent variable (export) on the dependent variable, CIEA, in the short run. It can be seen from

H 0 : C6 = C7 = C8 = C9 = 0

H 1 : C6 = C7 = C8 = C9 ≠ 0

Where C6, C7, C8 and C9 represents export lagged in first order, export lagged in second order, export lagged in third order and export lagged in fourth order. As indicated in

The study examined the adequacy of the specified models with various diagnostic tests including test for serial correlation and normality of residuals. The study used the Cumulative Sum of recursive residuals (CUSUM) test for the serial correlation of the model, which turned out to be stable or valid, see

Error Correction | D(CIEA) | D(LNEXP) | ||
---|---|---|---|---|

CointEq1 | −2.03908 | [−5.98345] | −0.04027 | [−3.18586] |

D(CIEA(−1)) | 0.601354 | [2.05316] | 0.037880 | [3.48648] |

D(CIEA(−2)) | 0.194234 | [0.82449] | 0.027041 | [3.09427] |

D(CIEA(−3)) | 0.036709 | [0.22217] | 0.018883 | [3.08072] |

D(CIEA(−4)) | −0.14518 | [−1.50140] | 0.008714 | [2.42925] |

D(LNEXP(−1)) | 10.71519 | [3.29731] | −0.83776 | [−6.94963] |

D(LNEXP(−2)) | 9.051741 | [2.27158] | −0.52147 | [−3.52780] |

D(LNEXP(−3)) | 6.354194 | [1.66243] | −0.27085 | [−1.91027] |

D(LNEXP(−4)) | 5.165923 | [1.95064] | −0.08055 | [−0.81997] |

C | 0.084444 | [0.23223] | 0.000840 | [0.06230] |

R-squared | 0.722 | 0.6114 | ||

Log likelihood | −311.1184 | 64.42757 | ||

Akaike AIC | 5.633656 | 0.954870 | ||

Schwarz SC | 5.873673 | 0.714852 |

Wald Test: | |||
---|---|---|---|

Equation: Untitled | |||

Test Statistic | Value | df | Probability |

F-statistic | 3.288545 | (4, 104) | 0.0140 |

Chi-square | 13.15418 | 4 | 0.0105 |

The Breusch Godfrey LM test was also applied on the residuals of the model to test for autocorrelation. This tests the null hypothesis of no serial correlation up to lag order 4 for this dataset. From

Granger causality test was used to further test for the causal relationship between the variables. From

The variance decomposition refers to the decomposition of mean square error in to contribution of each variable. Variance decomposition can be applied to determine the dynamic interaction between CIEA (economic growth) and exports. From

F-statistic | 0.761002 | Prob. F(4, 100) | 0.5531 |
---|---|---|---|

Obs*R-squared | 3.367658 | Prob. Chi-Square(4) | 0.4983 |

Dependent variable: D(CIEA) | |||
---|---|---|---|

Excluded | Chi-sq | Df | Prob. |

D(LNEXP) | 13.154 | 4 | 0.0105 |

All | 13.154 | 4 | 0.015 |

Dependent variable: D(LNEXP) | |||

Excluded | Chi-sq | df | Prob. |

D(CIEA) | 13.805 | 4 | 0.0079 |

All | 13.805 | 4 | 0.0079 |

Variance Decomposition of (CIEA): | |||
---|---|---|---|

Period | S.E | D(CIEA) | D(LNEXP) |

1 | 3.820572 | 100 | 0.000000 |

2 | 4.294618 | 99.97959 | 0.020411 |

3 | 4.377726 | 99.84830 | 0.151695 |

4 | 4.396942 | 99.75011 | 0.249890 |

5 | 4.418157 | 99.19380 | 0.806203 |
---|---|---|---|

6 | 4.495305 | 98.80472 | 1.195278 |

7 | 4.498760 | 98.80592 | 1.194079 |

8 | 4.509456 | 98.78660 | 1.213398 |

9 | 4.511096 | 98.76014 | 1.239855 |

10 | 4.513052 | 98.70932 | 1.290678 |

Variance Decomposition of (LNEXP): | |||

Period | S.E | D(CIEA) | D(LNEXP) |

1 | 0.130989 | 10.23278 | 89.76722 |

2 | 0.137332 | 9.339058 | 90.66094 |

3 | 0.139016 | 10.33899 | 89.66101 |

4 | 0.139889 | 11.04553 | 88.95447 |

5 | 0.140248 | 1.44674 | 88.55326 |

6 | 0.140331 | 11.43371 | 88.56629 |

7 | 0.140495 | 11.54385 | 88.45615 |

8 | 0.140628 | 11.6489 | 88.35110 |

9 | 0.140674 | 11.69288 | 88.30712 |

10 | 0.140678 | 11.69733 | 88.30267 |

Cholesky Ordering: D(CIEA) D(LNEXP).

displayed taking the 3^{rd} and the 10^{th} periods as short run and long run period respectively. In the short run shock to CIEA accounts for 99.85% variation of fluctuation in CIEA that is termed own shock, a shock to export can influence 0.15% fluctuation in CIEA, which indicates a short run equilibrium between the two variables. On the other hand, in the long run, the impact of shock on export contributes 88.30% fluctuations in the variation of exports and shock on CIEA contributes 11.70% variation in the fluctuation in exports.

The study specifically aimed at empirically investigating the relationship between exports and economic growth in Ghana during the period 2010 to 2019 using monthly data instead of the well-known annual data. Composite Index Economic Activity (CIEA) was used as a proxy for monthly real GDP (economic growth). To fulfil the objectives, specific null hypotheses were tested investigating whether or not exports contribute to economic growth and vice versa. In addition, the study tested for causality between export and economic growth. The study also confirmed the existence of long-run relationship (cointegration) between exports and economic growth, inferring that indeed export led growth is truly a sustainable way of achieving economic development in Ghana in the long run.

The above findings show that export led growth holds for Ghana, the positive and statistically significant result for exports has important policy implications for the country’s economic growth and developmental agenda. The government should also go ahead in implementing the recently launched Planting for export and rural development achieve sustainable development through export promotion. Ghana needs to diversify her export basket in order to sustain her export led growth agenda, promotion of non-traditional goods. There is also the need to embark on value-addition if the agenda is to be successful. This can be done through investing in technologies to help in processing its primary export commodities to boost export quality and revenue. Therefore, the government needs to create a conducive investment climate for foreign direct investment in the export sector.

Another policy option for the government is to consider providing subsidies to export-oriented producers especially Small and Medium scale Enterprises (SMEs) who drive the economy. Additionally, the government should increase the producer prices of her major export commodities such as cocoa.

Most of the previous studies fail to utilize monthly data to explore the strength of the causal interaction between exports and economic growth. This paper utilized the composite index economic activity as proxy for monthly real GDP analysis in the case of Ghana. Though empirical estimation of this research is solid by utilizing monthly data estimate, further research is required in other countries using monthly data estimates of other variables that influence economic growth.

The authors declare no conflicts of interest regarding the publication of this paper.

Mensah, A.C. and Okyere, E. (2020) Causality Analysis on Export and Economic Growth Nexus in Ghana. Open Journal of Statistics, 10, 872-888. https://doi.org/10.4236/ojs.2020.105051