Comparative analysis of asymmetric impact of cashew nuts and cocoa beans exports on Ghana’s economic growth: a non-linear ARDL approach

Abstract

The study applied the newly developed Non-linear Autoregressive Distributed Lag (NARDL) model by Shin et al. on annual data for Ghana from 1970 to 2019. Evidence of a long-run asymmetric cointegration relationship exists between cashew nuts exports, cocoa beans exports, and economic growth. Findings further revealed both cashew nuts and cocoa beans exports have a positive impact on Ghana’s economic growth in the long-run in support of the export-led growth theory. However, the impact of cocoa bean exports is greater than that of cashew nuts exports on the economic growth of Ghana. The Granger causality test revealed the existence of a unidirectional (one-way) causality running from economic growth to both the cashew nuts and cocoa beans exports. Government and policy makers should therefore continue to encourage and promote agricultural export growth to spur the economic development of Ghana. Also, policies and measures geared towards exchange rate stability by the government and Bank of Ghana should be promulgated and implemented to propel economic growth and export expansion in Ghana.

Keywords:

• Asymmetric impact
• cashew nuts exports
• cocoa beans exports
• economic growth
• non-linear autoregressive distributed lag

REVIEWING EDITOR:

• Christian Nsiah, Baldwin Wallace University, United States

SUBJECTS:

• Economics
• Political Economy
• Finance

1. Introduction

The contributions of export activities are integral to the economic development of nations and firms at large (Lages & Montgomery, 2004). An increase in a country’s export of goods and services can reduce unemployment, improve the balance of payments, and reduce pressure on external borrowing (Chenery & Strout, 1968). The export activities influence the amount of foreign exchange reserves which is used to purchase manufactured goods, capital goods, and technology. The level of imports of a country can shape public perceptions of a national competitiveness, enhance societal prosperity, and help national development (Hinson & Sorensen, 2006). In the light of this, for firms or nations to grow and enhance their levels of competitiveness in the international market, it is imperative for them to expand their exporting activities (Adu-Gyamfi & Korneliussen, 2013; O’Cass & Julian, 2003). Ghana is no exception in terms of revenue generation from production and export of cash crop products with a focus on cashew and cocoa. This sector provides 30% export earning to Ghana’s GDP. Similar to the cashew sector, which also provides export earnings to Ghana’s GDP added to the job creation opportunities in the Ghanaian economy.

According to Hinson and Sorensen (2006), Ghana is heavily dependent on its export sector as an engine of growth within an increasingly competitive global market. The notable feature of external trade in Ghana is dominated by traditional exports such as cocoa beans, gold, and log of timber, which account for about 70% of total exports (MoTI, 2012). In 2012, the Bank of Ghana reported that income received from exports stood at US$ 13.5 billion. For instance, the traditional export commodities such as gold and cocoa beans accounted for about US$ 5.6 billion and US$ 2.8 billion dollars, respectively. Also, in 2020, gold export receipts increased to US$ 6.80 billion, from US$ 6.23 billion in 2019. Cocoa beans exported amounted to US$ 1.48 billion, an increase of 2.0% year-on-year. The value of ‘other exports’, which is made up of non-traditional exports, electricity, and other minerals (aluminum alloys, bauxite, diamond, and manganese), was estimated at US$ 2.30 billion, compared to US$ 2.49 billion in 2019, representing a decline of 7.6%. The value of merchandise exports for the year was estimated at US$ 14.47 billion, representing a decline of 7.6% compared to US$ 15.67 billion for 2019 (BoG, 2020). To this end, Ghana recorded a surplus trade balance of US$ 2.04 billion (3.0% of GDP), compared with US$ 2.26 billion (3.4% of GDP) in 2019. This phenomenon has made government and private officials see the importance of an export-led economy (Hinson & Sorensen, 2006) since the sector is seen as the engine of economic growth.

The export sector in Ghana is usually a combination of the traditional and non-traditional exports (NTEs). However, in December 2010 when oil exports from the Jubilee Field commenced, the country added a third sector, which is the oil sector (GEPA, 2014). While the traditional export sector dominates the sector with 70% of exports comprising of cocoa beans, minerals, unprocessed timber, and electricity, non-traditional export sector consists of handicrafts, garments, horticulture, and many others, which contribute over 85% to employment in the manufacturing sector and 70% to the GDP (MoTI, 2012). The non-traditional exports consist of all other exports apart from the traditional exports, which comprises of handicrafts, garments, food crops, horticulture, furniture, among others (Hinson & Sorensen, 2006). The Non-traditional Export sector is categorized into three main sub-sectors: Agriculture products, Processed/Semi-Processed products, Industrial Art and Craft (Handicrafts). Although, Services exports are also mainstreamed into NTEs portfolio (GEPA, 2019).

Statistics from GEPA showed that the non-traditional agricultural exports sector has been experiencing a tremendous increase in real values, amounted to US$ 2.813 billion in 2018 and US$ 2.899 billion in 2019. This reflects an increase of 3.10% over 2018 earnings. 2018 performance for instance, was primarily due to a 43.84% increase in cashew nuts, a component of non-traditional agricultural export from US$262.95 million in 2017 to US$378.21 million in 2018. Over the last five years, that is from 2015 to 2019, Ghana’s Non-Traditional Exports grew at an annual average rate of 2.97% and contributed about 18% to the total merchandise exports of Ghana in 2019. The percentage contribution of non-traditional exports (NTEs) to total national exports in 2015, 2016, 2017, 2018, and 2019 were 20, 23, 18, 19, and 18%, respectively (GEPA, 2019).

Against this backdrop, the government of Ghana developed a policy framework and has put in measures to revamp the export performance by investing and diversifying into non-traditional exports through the Economic Recovery Programme in 1993 (Hinson & Sorensen, 2006). In furtherance to this, the government of Ghana set out modality to assist in the development of the non-traditional sector, known as the Non-Traditional Export Strategy.

However, NTEs have not been able to achieve the policy target of replacing traditional exports. A greater chunk of Ghana’s export revenue still comes from cocoa, gold, and timber, and earnings from NTEs such as cashew exports still have less export value as compared to those from traditional sources such as cocoa beans. Overall, the contribution from oil, gold, and cocoa, corresponds to over 80% of Ghana’s total exports (Okudzeto et al., 2014).

The trade balance in our national accounts is still deficit and this is believed to be a major cause of increasing fiscal deficit and macroeconomic volatility. The policy direction should be to reduce the magnitude of the trade deficit through alternative source of export earnings. The statistics indicate that the performance of NTEs against traditional exports is still below expectations.

Despite the above-mentioned importance of the agriculture sector contribution to the traditional and non-traditional export growth, this study explores and compares the direct impact of cashew nuts and cocoa beans exports to the economic performance of Ghana. In view of this, the study investigates the potential asymmetric impact of cashew nuts and cocoa beans exports on the Ghanaian economy since the country has a comparative advantage in the production of cashew and cocoa.

2. Empirical review

A huge body of literature is available on the role of exports in economic growth. In view of the important of the subject and the wide divergence in theoretical positions, several empirical studies have been conducted using the ELGH theory to study the effect of entire export on economic growth. However, the findings of these previous studies on cross-country and country studies revealed inconclusive results on the nexus between export and economic growth and did not provide any strong evidence either for or against the ELGH.

Earlier studies such as Balassa (1978), Chenery and Strout (1968), Heller and Porter (1978), Kormendi and Meguire (1985), Michaely (1977), and Tyler (1981) analyzed the relationship between economic growth and exports by employing a simple correlation coefficient technique and concluded that growth of exports and economic growth were highly positive correlated.

The second group of researchers like Balassa (1985), Feder (1983), Ram (1987), Sprout and Weaver (1993), Ukpolo (1994), and Voivodas (1973) used regression techniques to examine the relationship between export growth and economic growth, taking into account the neo-classical growth accounting equation. They found that the coefficient of the export variable was positive and highly significant.

The third batch of researchers like Ahmad and Kwan (1991), Bahmani-Oskooee et al. (1991), Darrat (1987), Holman and Graves (1995), Jin and Yu (1995), Jung and Marshall (1985), Khan and Saqib (1993), and Kunst and Marin (1989) examined the causality relationship between the growth of export and economic growth using Granger causality test. The studies concluded that there existed some evidence of causality relationship between exports and economic growth.

Although, previous studies depict a positive relationship between total exports and economic growth, neglecting the contribution of the primary agricultural exports to total exports. It is therefore reasonable to question whether the same holds for all the primary agricultural exports. However, empirical research into the relationship between agricultural exports and economic growth has not been given serious attention until recently.

Studies in this regard include that of Johnston and Mellor (1961), who discussed the role of the agriculture sector in the process of economic development in many ways. They emphasized that expanding agricultural exports were the main source of rising incomes and increasing foreign exchange earnings.

Levin and Raut (1998) investigated the effect of primary commodity and manufactured exports on economic growth. The exports of primary commodity included both agricultural products and other that is metals and oil products. The study found that manufacturing exports were the main source of economic growth and the exports of primary products had a negligible effect.

Ekanayake (1999) analyzed the causal relationship between economic growth and export growth by employing the error correction and co-integration models. The author used the time series data of eight Asian developing countries spanning the period from 1960 to 1997. The results of the study concluded that there was a bi-directional causality between export growth and economic growth in all the developing countries included in the analysis except Malaysia. There existed strong evidence for long-run Granger causality in all countries.

Dawson (2005) estimated the contribution of agricultural exports to economic growth in least developed countries. The author used the two theoretical models in his analysis. The first model was based on agricultural production function, including both agricultural and non-agricultural exports as inputs. The second model was a dual economy model that is agricultural and non-agricultural where each sector was sub divided into exports and non-export sectors. Fixed and Random effects were estimated in each model using a panel data of sixty-two less developed countries for the period 1974 to 1995. The study provided evidence from less developed countries that supported the theory of export led-growth. The results of the study highlighted the role of agricultural exports in economic growth. The study suggested that the export promotion policies should be balanced.

Aurangzeb (2006) investigated the relationship between economic growth and exports in Pakistan based on the analytical framework developed by Feder (1983). The author tested the applicability of the hypothesis that the economic growth increased as exports expanded by using time series data from 1973 to 2005. The findings of the study revealed that the export sector had significantly higher social marginal productivities. Hence the study concluded that an export-oriented and outward looking approach was needed for high rates of economic growth in Pakistan.

Sanjuán-López and Dawson (2010) estimated the contribution of agriculture exports to economic growth in under developed countries. They estimated the relationship between Gross Domestic Product and agrarian and non-agrarian exports. Panel co-integration technique was used in analyzing the data set of 42 under developed countries. The results of the study showed that there existed long-run relationship and the agriculture export elasticity of GDP was 0.07. The non-agriculture export elasticity of GDP was 0.13. Based on the empirical results, the study suggested that the poor countries should adopt balanced export promotion policies but the rich countries might attain high economic growth from non-agricultural exports.

Faridi (2012) studied the contribution of agricultural exports to economic growth in Pakistan. The author estimated the relationship between Gross domestic product (GDP) and agricultural and non-agricultural exports for Pakistan employing Johansen co-integration technique for the period 1972–2008. The findings of the study showed that the agricultural exports have a negative and significant effect on economic growth while non-agricultural exports elasticity was 0.58. Moreover, there is bidirectional causality in agricultural exports and real GDP. It is suggested that non-agricultural exports should be promoted.

Ehinomen and Daniel (2012) applied ARDL and Granger causality test to find out if there exists a causal and long-run relationship between export and economic growth in Nigeria. Examining the real GDP, exchange rate, export values, gross capital formation, labour force population, and imports values, annually time series data was employed from 1970 to 2010 which revealed the existence of a unidirectional relationship between economic growth and export in Nigeria and also the existence of long-run correlation between export and economic growth.

Noula et al. (2013) applied Cobb Douglas production function with Engle-Granger two step approach and Vector Correction Model (VECM) to explore the impact of agricultural exports on economic growth in Cameroon between the period of 1975–2009, which revealed a mixed effect of agricultural exports on economic growth, with cocoa export having a negative and insignificant impact on economic growth. However, coffee and banana exports have a positive relationship and significant to the economic growth of the country.

Rahman and Hossain (2014) applied Johansen & Juselius cointegration test, Vector Auto-Regressive (VAR) approach joined with Granger causality test to study the role of agriculture in the economic development of Bangladesh. A long-run correlation between economic growth and agricultural was uncovered, and the existence of unidirectional causality was revealed moving from agriculture to economic development based on the Granger causality test in Bangladesh. The Vector Auto-Regressive (VAR) results also confirmed that changes in agricultural GDP react more critically to economic growth and therefore suggested that, to stimulate economic development in Bangladesh, effort should be made on boosting the agricultural sector.

Njimanted and Aquilas (2015) examined the effect of timber export on economic growth in Cameroon by applying Johansen Cointegration and Vector Error Correction Model through an annual time series spanning a period of 1980–2014. The findings showed that timber exports contained an insignificant affirmative impact on economic growth in both long-run and short-run. To these researchers, it was incumbent on the Cameroonian government to encourage increased consumption of locally-manufactured wood products, encourage the establishment of locally based wood processing industries, and restrict imported manufactured wood products.

Yifru (2015) examined the effect of disaggregated agricultural exports on economic growth of Ethiopia using the vector error correction model, Johansen cointegration test, and Granger causality test. The study discovered a negative and insignificant correlation between pulses export and economic growth, although both oilseeds and coffee export were discovered to have a significant and positive influence on economic growth. It was further found that there exists a bi-directional correlation among coffee and oilseeds export, and economic growth, while there was unidirectional causality moving from pulses export to economic growth.

Bashir et al. (2015) studied the export-led growth theory in Pakistan using the vector error model, Granger causality tests, and cointegration on yearly time series data for the period of 1972 to 2012 in Pakistan. The study found that, in both long-run and short-run there is a significant positive relationship between exports and economic growth. Hence, the study concluded that there is an existence of export-led growth theory in Pakistan and therefore made a suggestion to the government to put in place incentives such as export bonuses, export credit guarantee schemes, and export financing, to encourage Pakistani exporters.

Edeme et al. (2016) evaluated the impact of agricultural exports on the economic growth of fifteen ECOWAS countries using panel data for the period 1980–2013. They employed time series variables, such as labour force participation rate, capital stock, agricultural exports, non-agricultural exports, inflation, and economic growth and came out with the following findings: The results of the fixed-effect model showed that agricultural exports have not impacted significantly on the economic growth of ECOWAS countries such as Côte d’Ivoire and Nigeria with respect to the Republic of Benin, which was the selected baseline. The study also analyzed the country combined effect of the agricultural exports and found that it was significant but the rate of impact was weak.

Ouma et al. (2016) examined the relationship of the agricultural trade by means of economic growth in East African Community (EAC) from the period 2000 to 2012. Vector Error Correction Models (VECM) and bi-variate Vector Auto-Regressive (VAR) were employed in the study for which the empirical results revealed that East African countries (EAC) member states have different and mix results. Kenya and Rwanda exhibited a unidirectional link between agricultural export and economic growth while Uganda, Burundi, and Tanzania show no relationship.

Bakari and Mabrouki (2017) investigated the effect of agricultural exports on economic growth in South-Eastern Europe Countries using Panel Data for the period 2006 to 2016 and was tested using correlation analysis and the static gravity model. The results revealed that agricultural exports had a positive effect on economic growth. These results also indicated that agricultural exports were a provenance of economic growth in South Eastern Europe Countries. For this reason, it was very important to refine investment in agricultural sector and create more effective agricultural trade openness policies.

Siaw et al. (2018) examined the correlation between agricultural exports and economic growth. The study was carried out in Ghana at the disaggregate level using the Autoregressive Distributed Lag (ARDL) model with yearly time series data spanning from 1990Q1 to 2011Q4. Both the long-run and the short-run results revealed that cocoa export had a positive and significant impact on economic growth while the export of pineapple and banana had a negative effect on economic growth even though pineapple export was not significant in both long run and short run. In addition, the study found unidirectional causality running from banana to economic growth, a bi-directional causal relationship between cocoa export and economic growth, and no causal relationship was found between economic development and pineapple export in Ghana.

Urriola et al. (2018) analyzed and quantified the short- and long-run impact of agricultural exports–both traditional and non-traditional products–on economic growth of Peru using an annual time series data from 2000 to 2016. A Vector Autoregressive (VAR) model, Johansen Co-integration test, and Granger Causality test were employed for the data analysis. The findings revealed that in the short run, traditional agricultural exports had a positive but non-significant effect on economic growth while non-traditional agricultural exports had a positive and significant effect on economic growth. Moreover, the co-integration result showed that there was a long-run relationship between the studied variables and a unidirectional causality relationship between the determinant variables and economic growth.

Okyere (2020) investigated the impact of export and import on economic growth of Ghana using a data set from 1998 to 2018. The study employed the simulation technique of co-integration and error correction. The results revealed that there was no significant causal relationship between imports in international trade and Ghana’s GDP growth. Exports were however revealed to have a significant causal relationship with Ghana’s GDP growth, such as cocoa. There was no causal relationship between exchange rate and inflation rate. However, there exists a causal relationship between GDP, exchange rate, and inflation rate.

Dastjerdi et al. (2021) investigated the factors affecting pistachio exports in Iran. The study employed an Autoregressive distributed lag (ARDL) model using a data set from 2001 to 2019. The results indicated that economic growth, appreciation of the exchange rate, and bank facilities remained positive and had a significant effect, while liquidity growth had a significant negative effect on pistachio export.

Ikechukwu (2021) empirically analyzed the effects of international trade on the economic growth of Nigeria from 1981 to 2018 using the Ordinary Least Squares (OLS) technique. The Johansen co-integration test was conducted and the results confirmed the absence of long-run equilibrium. Findings from the study revealed that exchange rates in the country had a negative and insignificant relationship with economic growth. However, the several trade policies in Nigeria have been seen to retard growth in the economic prosperity of Nigeria’s economy since the impact is negative and significant on GDP growth.

3. Materials and methods

3.1. Theoretical framework

This study aims to establish an economic model that will provide a deeper understanding of the export-led growth hypothesis (ELGH) by analyzing the impact of cashew nuts and cocoa beans exports on the economic growth of Ghana. The basis for formulating the research will be (Erhieyovwe & Onokero, 2013). They used three variables in their model: GDP as dependent variable, production and exchange rate as regressors. This research follows a similar pattern by adding more variables to examine the contribution of cashew nuts and cocoa beans exports on the Ghanaian economy. In line with the above framework, we estimated the growth model for our study according to the stated objectives as follows: (1) $${\mathit{\text{RGDP}}}_t=\mathit{\text{f}}({\mathit{\text{CASHEW}}}_t,{\mathit{\text{COCOA}}}_t,{\mathit{\text{EXC}}}_t,{\mathit{\text{FDI}}}_t,{\mathit{\text{GCF}}}_t,{\mathit{\text{INF}}}_t)$$(1)

To discard the differences in the measurement units, we applied the natural logarithm on both sides of Equation (1)(1) $${\mathit{\text{RGDP}}}_t=\mathit{\text{f}}({\mathit{\text{CASHEW}}}_t,{\mathit{\text{COCOA}}}_t,{\mathit{\text{EXC}}}_t,{\mathit{\text{FDI}}}_t,{\mathit{\text{GCF}}}_t,{\mathit{\text{INF}}}_t)$$(1) , and the general form of the econometric model is outlined as follows: (2) $$\mathrm{\Delta }\mathit{\text{Ln}}{\mathit{\text{RGDP}}}_t={}_0+{\mathrm{\beta }}_1{\mathit{\text{LnCASHEW}}}_t+{\mathrm{\beta }}_2\mathit{\text{Ln}}{\mathit{\text{COCOA}}}_t+{\mathrm{\beta }}_3\mathit{\text{Ln}}{\mathit{\text{EXC}}}_t+{{\mathrm{\beta }}_{4}L\mathit{\text{nFDI}}}_t+{{\mathrm{\beta }}_{5}L\mathit{\text{nGCF}}}_t+{{\mathrm{\beta }}_{6}L\mathit{\text{nINF}}}_t+{\mathrm{\epsilon }}_t$$(2)

Where, ${\mathit{\text{RGDP}}}_t$ is the real Gross Domestic Product in Ghana, ${\mathit{\text{CASHEW}}}_t$ is the export value of raw cashew nuts, ${\mathit{\text{COCOA}}}_t$ is the export value of raw cocoa beans, ${\mathit{\text{EXC}}}_t$ is the exchange rate of the local currency to US$ dollar, ${\mathit{\text{GCF}}}_t$ is the gross capital formation, ${\mathit{\text{FDI}}}_t$ is the foreign direct investment and ${\mathit{\text{INF}}}_t$ is the annual consumer price index, ${\beta }_0$ is the constant term, ${\beta }_1$- ${\beta }_6$ are the elasticity coefficients of the regressors and ${\mathrm{\epsilon }}_t$ is the error term.

3.2. Empirical models specification

To trace both the short-run and long-run effects of cashew nuts and cocoa beans exports on economic growth, the above Equation (2)(2) $$\mathrm{\Delta }\mathit{\text{Ln}}{\mathit{\text{RGDP}}}_t={}_0+{\mathrm{\beta }}_1{\mathit{\text{LnCASHEW}}}_t+{\mathrm{\beta }}_2\mathit{\text{Ln}}{\mathit{\text{COCOA}}}_t+{\mathrm{\beta }}_3\mathit{\text{Ln}}{\mathit{\text{EXC}}}_t+{{\mathrm{\beta }}_{4}L\mathit{\text{nFDI}}}_t+{{\mathrm{\beta }}_{5}L\mathit{\text{nGCF}}}_t+{{\mathrm{\beta }}_{6}L\mathit{\text{nINF}}}_t+{\mathrm{\epsilon }}_t$$(2) was illustrated in a compact form as: (3) $${\mathrm{\Delta }L\mathrm{n}\mathit{\text{RGDP}}}_t={\mathrm{\alpha }}_0+{\mathrm{\alpha }}_1{\mathit{\text{LnCASHEW}}}_t+{{\mathrm{\alpha }}_{2}L\mathit{\text{nCOCOA}}}_t+{\sum }_{j=3}^N{\mathrm{\alpha }}_j\mathit{\text{Ln}}{\mathit{X}}_t+{\mathrm{\epsilon }}_t$$(3)

Where X is a vector of the control regressors comprising of EXC, FDI, GCF, and INF; t refers to time; ${\alpha }_0$ to ${\alpha }_j$ are the model coefficients to be estimated.

The ECM of the Linear ARDL of Equation (2)(2) $$\mathrm{\Delta }\mathit{\text{Ln}}{\mathit{\text{RGDP}}}_t={}_0+{\mathrm{\beta }}_1{\mathit{\text{LnCASHEW}}}_t+{\mathrm{\beta }}_2\mathit{\text{Ln}}{\mathit{\text{COCOA}}}_t+{\mathrm{\beta }}_3\mathit{\text{Ln}}{\mathit{\text{EXC}}}_t+{{\mathrm{\beta }}_{4}L\mathit{\text{nFDI}}}_t+{{\mathrm{\beta }}_{5}L\mathit{\text{nGCF}}}_t+{{\mathrm{\beta }}_{6}L\mathit{\text{nINF}}}_t+{\mathrm{\epsilon }}_t$$(2) is illustrated below: (4) $${\mathrm{\Delta }\mathit{\text{LnRGDP}}}_t={\mathrm{\beta }}_0+{\sum }_{i=1}^{p1}{\mathrm{\beta }}_{1i}\mathit{\text{Ln}}{\mathrm{\Delta }\mathit{\text{RGDP}}}_{t-i}+{\sum }_{i=0}^{p2}{\mathrm{\beta }}_{2i}{\mathrm{\Delta }\mathit{\text{CASHEW}}}_{t-i}+{\sum }_{i=0}^{p3}{\mathrm{\beta }}_{3i}{\mathit{\text{Ln}}\mathrm{\Delta }\mathit{\text{COCOA}}}_{t-i}+{\sum }_{i=0}^{p4}{\mathrm{\beta }}_{4i}\mathit{\text{Ln}}{\mathrm{\Delta }\mathit{\text{EXC}}}_{t-i}+{\sum }_{i=0}^{p5}{\mathrm{\beta }}_{5i}{\mathit{\text{Ln}}\mathrm{\Delta }\mathit{\text{FDI}}}_{t-i}+{\sum }_{i=0}^{p6}{\mathrm{\beta }}_{6i}{\mathit{\text{Ln}}\mathrm{\Delta }\mathit{\text{GCF}}}_{t-i}+{\sum }_{i=0}^{p7}{\mathrm{\beta }}_{7i}{\mathit{\text{Ln}}\mathrm{\Delta }\mathit{\text{INF}}}_{t-i}+{\mathrm{\lambda }}_1\mathit{\text{Ln}}{\mathit{\text{GDP}}}_{t-1}+{\mathrm{\lambda }}_2{\mathit{\text{LnCASHEW}}}_{t-1}+{\mathrm{\lambda }}_3\mathit{\text{Ln}}{\mathit{\text{COCOA}}}_{t-1}+{\mathrm{\lambda }}_4{\mathit{\text{LnEXC}}}_{t-1}+{\mathrm{\lambda }}_5{\mathit{\text{LnFDI}}}_{t-1}+{\mathrm{\lambda }}_6{\mathit{\text{LnGCF}}}_{t-1}+{\mathrm{\lambda }}_7{\mathit{\text{LnINF}}}_{t-1}+{\mathrm{\epsilon }}_t$$(4)

Following Shin et al. (2014), the asymmetric long-run relationship with only cashew nuts and cocoa beans exports entering the equation asymmetrically (introducing their partial sum of positive and negative changes) can be written as: (5) $${\mathrm{\Delta }\mathit{\text{LnRGDP}}}_t={\mathrm{\beta }}_0+{\mathrm{\beta }}_1^{+}{\mathit{\text{LnCASHEW}}}_t^{+}+{\mathrm{\beta }}_2^{-}{\mathit{\text{LnCASHEW}}}_t^{-}+{\mathrm{\beta }}_3^{+}{\mathit{\text{LnCOCOA}}}_t^{+}+{\mathrm{\beta }}_4^{-}{\mathit{\text{LnCOCOA}}}_t^{-}+\mathrm{\varphi }{\mathit{\text{Ln}}\mathit{X}}_t+{\mathrm{\epsilon }}_t$$(5)

Where X is a vector of the control variables consisting of EXC, INF, GCF, and FDI, ϕ measures the effects of the control variables in the equation and ${\mathit{X}}_t$ is a vector of regressors entering Equation (5)(5) $${\mathrm{\Delta }\mathit{\text{LnRGDP}}}_t={\mathrm{\beta }}_0+{\mathrm{\beta }}_1^{+}{\mathit{\text{LnCASHEW}}}_t^{+}+{\mathrm{\beta }}_2^{-}{\mathit{\text{LnCASHEW}}}_t^{-}+{\mathrm{\beta }}_3^{+}{\mathit{\text{LnCOCOA}}}_t^{+}+{\mathrm{\beta }}_4^{-}{\mathit{\text{LnCOCOA}}}_t^{-}+\mathrm{\varphi }{\mathit{\text{Ln}}\mathit{X}}_t+{\mathrm{\epsilon }}_t$$(5) symmetrically.

CASHEW and COCOA in Equation (5)(5) $${\mathrm{\Delta }\mathit{\text{LnRGDP}}}_t={\mathrm{\beta }}_0+{\mathrm{\beta }}_1^{+}{\mathit{\text{LnCASHEW}}}_t^{+}+{\mathrm{\beta }}_2^{-}{\mathit{\text{LnCASHEW}}}_t^{-}+{\mathrm{\beta }}_3^{+}{\mathit{\text{LnCOCOA}}}_t^{+}+{\mathrm{\beta }}_4^{-}{\mathit{\text{LnCOCOA}}}_t^{-}+\mathrm{\varphi }{\mathit{\text{Ln}}\mathit{X}}_t+{\mathrm{\epsilon }}_t$$(5) are decomposed into their partial sum components with the focus being on the positive and negative partial sums as follows; (5a) $$\mathrm{\Delta }{\mathit{\text{CASHEW}}}_t={\mathit{\text{CASHEW}}}_0+{\mathit{\text{CASHEW}}}_t^{+}+{\mathit{\text{CASHEW}}}_t^{-}$$(5a) ${\mathit{\text{CASHEW}}}_0$ refers to no change in CASHEW series, ${\mathit{\text{CASHEW}}}_t^{+}$ is the partial sum of positive changes in the CASHEW variable given by: (5b) $${\mathit{\text{CASHEW}}}_t^{+}={\sum }_{i=1}^t\mathrm{\Delta }{\mathit{\text{LnCASHEW}}}_t^{+}={\sum }_{i=1}^t\mathrm{\text{max}}(\mathrm{\Delta }{\mathit{\text{LnCASHEW}}}_t,0)$$(5b) ${\mathit{\text{CASHEW}}}_t^{-}$ is the partial sum of negative changes in the CASHEW variable given by: (5c) $${\mathit{\text{CASHEW}}}_t^{-}={\sum }_{i=1}^t\mathrm{\Delta }{\mathit{\text{LnCASHEW}}}_t^{-}={\sum }_{i=1}^t\mathrm{\text{min}}(\mathrm{\Delta }{\mathit{\text{LnCASHEW}}}_t,0)$$(5c) (5d) $${\mathrm{\Delta }\mathit{\text{COCOA}}}_t={\mathit{\text{COCOA}}}_0+{\mathit{\text{COCOA}}}_t^{+}+{\mathit{\text{COCOA}}}_t^{-}$$(5d) ${\mathit{\text{COCOA}}}_0$ refers to no change in COCOA series, ${\mathit{\text{COCOA}}}_t^{+}$ is the partial sum of positive changes in the COCOA variable given by: (5e) $${\mathit{\text{COCOA}}}_t^{+}={\sum }_{i=1}^t\mathrm{\Delta }{\mathit{\text{LnCOCOA}}}_t^{+}={\sum }_{i=1}^t\mathrm{\text{max}}(\mathrm{\Delta }{\mathit{\text{LnCOCOA}}}_t,0)$$(5e) ${\mathit{\text{COCOA}}}_t^{-}$ is the partial sum of negative changes in the COCOA variable given by: (5f) $${\mathit{\text{COCOA}}}_t^{-}={\sum }_{i=1}^t\mathrm{\Delta }{\mathit{\text{LnCOCOA}}}_t^{-}={\sum }_{i=1}^t\mathrm{\text{max}}(\mathrm{\Delta }{\mathit{\text{LnCOCOA}}}_t,0)$$(5f) ${\mathit{\text{CASHEW}}}_0$ and ${\mathit{\text{COCOA}}}_0$ are subsumed into the constant term are the threshold values (Threshold of zero) for the partial sum of changes in the CASHEW and COCOA variables (Eberhardt & Presbitero, 2013; Shin et al., 2014).

Again, following Shin et al. (2014) the ECM from Equation (5)(5) $${\mathrm{\Delta }\mathit{\text{LnRGDP}}}_t={\mathrm{\beta }}_0+{\mathrm{\beta }}_1^{+}{\mathit{\text{LnCASHEW}}}_t^{+}+{\mathrm{\beta }}_2^{-}{\mathit{\text{LnCASHEW}}}_t^{-}+{\mathrm{\beta }}_3^{+}{\mathit{\text{LnCOCOA}}}_t^{+}+{\mathrm{\beta }}_4^{-}{\mathit{\text{LnCOCOA}}}_t^{-}+\mathrm{\varphi }{\mathit{\text{Ln}}\mathit{X}}_t+{\mathrm{\epsilon }}_t$$(5) can be specified as follows: (6) $${\mathrm{\Delta }\mathit{\text{LnRGDP}}}_t={\mathrm{\beta }}_0+\mathrm{\rho }{\mathit{\text{LnRGDP}}}_{t-1}+{\mathrm{\beta }}_1^{+}{\mathit{\text{LnCASHEW}}}_{t-1}^{+}+{\mathrm{\beta }}_2^{-}{\mathit{\text{LnCASHEW}}}_{t-1}^{-}+{\mathrm{\beta }}_3^{+}{\mathit{\text{LnCOCOA}}}_{t-1}^{+}+{\mathrm{\beta }}_4^{-}{\mathit{\text{LnCOCOA}}}_{t-1}^{-}+\mathrm{\beta }{\mathit{\text{Ln}}\mathit{X}}_{t-1}+{\sum }_{i=1}^{p-1}{\mathrm{\delta }}_i{\mathrm{\Delta }\mathit{\text{LnRGDP}}}_{t-i}+{\sum }_{i=0}^{q-1}({\mathrm{\theta }}_i^{+}\mathrm{\Delta }{\mathit{\text{LnCASHEW}}}_{t-i}^{+}+{\mathrm{\theta }}_i^{-}\mathrm{\Delta }{\mathit{\text{LnCASHEW}}}_{t-i}^{-})+{\sum }_{i=0}^{m-1}({\mathrm{\psi }}_i^{+}\mathrm{\Delta }{\mathit{\text{LnCOCOA}}}_{t-i}^{+}+{\mathrm{\psi }}_i^{-}\mathrm{\Delta }{\mathit{\text{LnCOCOA}}}_{t-i}^{-})+{\sum }_{i=0}^{n-1}{\mathrm{\phi }}_i\mathrm{\Delta }{\mathit{\text{Ln}}\mathit{X}}_{t-i}+{\mathrm{\epsilon }}_t$$(6)

All variables retain their descriptions as given above. ρ is the autoregressive parameter and the asymmetric long-run parameters of the positive and negative partial sum coefficients of the lagged levels of CASHEW is given by ${\beta }_1^{+}$ = −${\widehat{\beta }}_1^{+}$/$\widehat{\mathrm{\rho }}$ and ${\beta }_2^{-}$ = −${\widehat{\beta }}_2^{+}$/$\widehat{\mathrm{\rho }}$ and that of COCOA is given by ${\beta }_3^{+}$ = −${\widehat{\beta }}_3^{+}$/$\widehat{\mathrm{\rho }}$ and ${\beta }_4^{-}$ = −${\widehat{\beta }}_4^{+}$/$\widehat{\mathrm{\rho }}$ β is a vector of symmetric long-run parameters of the control variables. ${\sum }_{i=1}^{p-1}{\mathrm{\delta }}_i$ is the symmetric short-run parameters of the lagged difference of the dependent variable and ${\sum }_{i=0}^{n-1}{\mathrm{\phi }}_i$ is the symmetric short-run parameters of the control variables. ${\sum }_{i=1}^{q-1}{\mathrm{\theta }}_i^{+},$ ${\sum }_{i=1}^{q-1}{\mathrm{\theta }}_i^{-};$ ${\sum }_{i=1}^{m-1}{\mathrm{\psi }}_i^{+},$ ${\sum }_{i=1}^{m-1}{\mathrm{\psi }}_i^{-}$ represent the asymmetric short run coefficients of the lagged difference of cashew nuts and cocoa beans exports. Δ is the difference operator while ‘p’, ‘q’, ‘m’, and ‘n’ are the maximum (optimal) lag lengths of the dependent and independent variables, respectively.

3.3. Variables and data source

Secondary data on export values of Cashew Nuts and export values of Cocoa Beans for Ghana were used for the study. Also, secondary data on Exchange Rate, Real Gross Domestic Product (GDP), Gross Capital Formation, Consumer Price Index (Inflation), and Foreign Direct Investment for Ghana were employed in the study. The study uses a 50-year annual data set over the period from 1970 to 2019. The main reason for limiting the time span of data was the availability of data, hence our choice for this time frame. Data on export values of cashew nuts and cocoa beans were accessed from the Food and Agriculture Organization of the United Nations (FAOSTAT) and the rest of the variables were obtained from the World Development Indicators (WDI) database of the World Bank.

3.4. Stationarity/unit roots test

A very important exercise in time series studies is to explore the stationarity properties of the series used in the study. This is to ensure that the regression result(s) is not spurious. This occurs when the regression results have high R-squared and the Durbin-Watson statistic close to 0 or 4 but with insignificant coefficients of the independent variables (Gujarati, 2004).

Basically, the ARDL (Pesaran et al., 2001) and NARDL (Shin et al., 2014) approaches to cointegration do not require testing for unit roots in the series. However, it is imperative to be certain that none of the variables is non-stationary or has a unit root even at first difference or is an I (2) process. In this regard, the Dickey-Fuller Generalized Least Square DF-GLS (1992) and KPSS (1992) unit root tests were used to check for stationarity of the variables to be sure that none of the variables was integrated of order 2 in which case the Bounds testing procedure breaks down (Pesaran et al., 2001).

3.5. Tests for cointegration

According to Pesaran et al. (2001), to arrive at a long-run cointegration relationship, if the F-statistic computed is greater than the Upper Bound Critical (UBC) values constructed by Pesaran et al. (2001), then the null hypothesis is rejected for the alternative. But, if the F-static computed is lesser than the Lower Bound Critical (LBC) values of Pesaran et al. (2001), then the null is not rejected. And if the F-statistic computed falls between the UBC and LBC, then it becomes inconclusive (Pesaran et al., 2001). However, whereas Pesaran et al. (2001) focus on the significance of the F-Statistic, Banerjee et al. (1998) prefer the significance of the ECM term along with the value lies between −1 and 0 for the confirmation of long-run equilibrium relationship to check and test the speed of adjustment back to equilibrium after a shock occurs.

The null hypothesis of no cointegration was tested against the alternative hypothesis of the existence of cointegration relationship as follows:

($H_{0\mathrm{ }}:$ ρ = ${\beta }_1^{+}$ = ${\beta }_2^{-}$ = ${\beta }_3^{+}$ = ${\beta }_4^{-}$ = 0)

($H_{1\mathrm{ }}:$ ρ = ${\beta }_1^{+}$ ≠ ${\beta }_2^{-}$ ≠ ${\beta }_3^{+}$ = ${\beta }_4^{-}$ = 0)

3.6. Ganger causality test

The concept of Granger causality is based on the principle that two variables X and Y can have effects on one another. The existence of cointegration between X and Y implies Granger causality. Furthermore, since cointegration between variables does not automatically imply causality between them, the evidence of causality between the variables must be provided by Granger causality analysis. X is said to granger-causes Y if both current and lagged values of X can be better used to predict another variable Y even when the lagged forms of Y are taken into consideration. Such a test reveals the direction of causation. This direction, if it exists, can be unidirectional, bidirectional, or independent (no causality). A simple Ganger causality test involving two variables, agricultural exports, and RGDP can therefore be specified as: (7) $${\mathit{\text{LnRGDP}}}_t={\mathrm{\alpha }}_0+{\sum }_{j=1}^p{\mathrm{\alpha }}_j{\mathit{\text{LnRGDP}}}_{t-j}+{\sum }_{j=1}^q{\mathrm{\phi }}_j{\mathit{\text{LnAEXP}}}_{t-j}+{\mathrm{\epsilon }}_t$$(7) (8) $${\mathit{\text{LnAEXP}}}_t={\mathrm{\beta }}_0+{\sum }_{j=1}^p{\mathrm{\beta }}_j{\mathit{\text{LnAEXP}}}_{t-j}+{\sum }_{j=1}^q{\mathrm{\delta }}_j{\mathit{\text{LnRGDP}}}_{t-j}+{\mathrm{\nu }}_t$$(8)

Where, ${\mathit{\text{LnRGDP}}}_t$ represents the natural log of real gross domestic product (as a measure of economic growth) and ${\mathit{\text{LnAEXP}}}_t$ represents the natural log of agricultural exports (cashew nuts and cocoa beans exports). ${\mathrm{\epsilon }}_t$ and ${\mathrm{\nu }}_t$ are serially uncorrelated white noise error terms; the coefficients α, φ, β, and δ 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 hypotheses to be tested are:

$H_0:$ ${\alpha }_j$ = 0, j = 1…. q, Agricultural Export growth does not cause GDP Growth.

$H_1:$ ${\beta }_j$ = 0, t = 1…. q, GDP growth does not cause Agricultural Export Growth.

4. Results and discussions

4.1. Descriptive statistics

Descriptive statistics of the variables used in the study from 1970 to 2019 as shown in Table 1 below showed the mean, standard deviation, minimum and maximum. These were converted into natural logarithms for ease of interpretation of parameters (in terms of percentages or elasticities) and the possibility of reducing the problem of multicollinearity and heteroscedasticity.

Table 1. Descriptive statistics of the parameter variables from 1970 to 2019.

	RGDP	CASHEW	COCOA	EXC	FDI	GCF	INF
Mean	23.560	7.350	13.280	−3.250	0.200	2.710	2.980
Median	23.400	6.100	13.060	−2.230	0.510	2.750	2.890
Maximum	24.770	13.800	14.710	1.650	2.250	3.430	4.810
Minimum	22.850	3.400	12.150	−9.190	−3.090	1.320	−0.900
Std. Dev.	0.600	3.120	0.680	3.850	1.520	0.520	0.950
Coef. of Var.	2.550	42.450	5.120	−118.460	7600	19.190	31.880
Skewness	0.620	0.600	0.590	−0.420	−0.490	−0.620	−1.060
Kurtosis	2.050	1.970	2.300	1.660	2.170	2.720	7.170

Note: All variables are log transformed. Coefficient of Variation (CV) is computed as CV = $\mathrm{\sigma }/\overline{X}$*100 where $\mathrm{\sigma }$ is the standard deviation; and $\overline{X}$ is the mean.

Source: Author’s Computation (2021).

Exports of cocoa beans generate the highest earnings for Ghana’s economic growth, followed by cashew nuts exports. This means that on average, Ghana earned approximately US$13.28 million/ton annually from cocoa bean exports. On the other hand, Ghana generated US$ 7.35 million/ton annually from cashew nuts exports.

The RGDP value showed an average of US$ 23.56 for the period under study and gives an indication that on average, Ghana’s GDP grew at (approximately US$ 23.56) annually from 1970 to 2019. The exchange rate (EXC) also, showed a mean value of −3.25% and indicates that, on average, the annual exchange rate of Ghana depreciated at (approximately 3.25%) from 1970 to 2019. The highest standard deviation was recorded by the exchange rate (EXC) variable and indicates the largest spread over the period under study, while Investment (GCF) is the minimal spread over time.

Skewness is a measure of the probability distribution of a real-valued random variable about its mean. A normal distribution is symmetrical at point 0. If the value is >0 it is positively skewed but if it is <0, it is negatively skewed. Except for RGDP, CASHEW, and COCOA, all the other variables were negatively skewed.

Kurtosis measures the peakness or flatness of the distribution of the series and if the kurtosis is above 3, the distribution is peaked or leptokurtic relative to the normal and if the kurtosis is <3, the distribution is flat or platykurtic relative to normal. Kurtosis for Inflation (INF) was above 3, hence leptokurtic relative to normal, and the rest of the variables were below 3, hence the distributions were platykurtic relative to normal.

4.2. Pairwise correlation analysis

A correlation analysis was carried out which gives information on the strength and direction of the relationship among the variables used in the study and to ascertain whether or not there are multicollinearity problems. Multicollinearity is when independent variables (predictors) are highly correlated making it difficult to unbundle the individual effects of the predictors on the dependent variable. Multicollinearity leads to an increase in standard errors; hence, coefficient estimates turn out to be statistically insignificant because of smaller t-ratios (Gujarati, 2004). Table 2 below presents the correlation matrix, and there is only one set of highly correlated independent variables (CASHEW & RGDP) with a correlation coefficient of 0.9531. However, the estimation technique used –NARDL is able to correct potential multicollinearity problems (Pesaran et al., 2001; Shin et al., 2014).

Table 2. Pairwise correlation analysis for the NARDL model.

Variables	RGDP	CASHEW	COCOA	EXC	FDI	GCF	INF
RGDP	1
CASHEW	0.953***	1
COCOA	0.876***	0.850***	1
EXC	0.881***	0.832***	0.678**	1
FDI	0.788***	0.729***	0.627***	0.630***	1
GCF	0.707***	0.647***	0.467***	0.765***	0.697***	1
INF	−0.527***	−0.586***	−0.399**	−0.381**	−0.479***	−0.395**	1

Note: All variables are log transformed. *** and ** indicates stationarity at significance levels of 1%, and 5% levels, respectively.

Source: Author’s Computation (2021).

4.3. Stationarity/unit root test results

The stationarity of the variables was concluded based on the outcome of both the DF-GLS and KPSS at constant ad trend technique. The results are presented in Table 3.

Table 3. Results of DF-GLS and KPSS unit root tests (intercept option).

Variable	DF-GLS test		KPSS test
	(1%: −2.59, 5%: −1.94, 10%: −1.61)		(1%: 0.739, 5%: 0.463, 10%: 0.347)
	Levels	First difference IO	Levels	First difference IO
RGDP	−0.780	−4.410*** I (1)	0.881	0.685*** I (1)
EXC	−0.170	−1.780* I (1)	0.897	0.194*** I (1)
GCF	−1.440	−6.700*** I (1)	0.619***	0.104 I (0)
INF	−2.620***	−5.540 I (0)	0.517***	0.266 I (0)
FDI	−1.500	−4.650*** I (1)	0.602***	0.165 I (0)
CASHEW	−0.740	−8.160*** I (1)	0.837	0.185*** I (1)
COCOA	−0.380	−1.820* I (1)	0.700***	0.091 I (0)

Note: All variables are log transformed. *** and * indicates stationarity at significance levels of 1% and 10% levels, respectively.

Source: Author’s Computation (2021).

From Table 3 above, the test statistic failed to reject the null hypothesis of a unit root in level data for all the variables except the inflation rate (INF) variable. However, the null hypothesis was rejected at 1% significance level after their first differences for RGDP, GCF, FDI, and CASHEW variables and at 10% significance level for the EXC and COCOA variables using DF-GLS test. It must be noted that, in the case of CASHEW and INF variables, a lower lag length (e.g. with only 0 lag was included). Also, the null hypothesis was rejected at 1% significance level of the INF variable after level testing using DF-GLS test. For the KPSS test, the null hypothesis was rejected at 1% significance level at the level for GCF, INF, FDI, and COCOA variables. However, for RGDP, EXC, and CASHEW variables, the null hypothesis was rejected at the 1% significance level after their first difference using the KPSS test. The results from the test statistic, however, revealed that not all the series were integrated in the same order I (1). Some were integrated at order zero I (0) in levels and the rest were integrated at order one I (1) after the first difference. None of the variables was integrated of a higher order that is I (2). To this end, the study adopted the Non-linear ARDL model to investigate a possible cointegration of the variables along with important diagnostic checking since none of the variables was stationary beyond order one.

4.4. Long-run relationship (cointegration) test results

Stationarity tests revealed that not all the variables were integrated in the same order. For this reason, the Johansen test for cointegration could not be used for this study due to the mixed levels of integration of the variables at levels [I (0) and first difference I (1)]. Consequently, the F-Bounds test developed by Pesaran et al. (2001) and Shin et al. (2014) were employed.

To test and investigate the potential asymmetric impact of cashew nuts and cocoa beans exports on economic growth in Ghana, the NARDL model was estimated and the Bounds test for long-run relationship was conducted. Table 4 below contains the results of the cointegration test.

Table 4. Results of NARDL bounds cointegration test under an asymmetric framework.

Model		K	F-statistic
RGDP		12	7.670
	LBC I (0)	UBC I (1)
10%	1.830	2.940
5%	2.060	3.240
1%	2.540	3.860

Source: Author’s Computation (2021).

From Table 4 above, the F-statistic of Bounds test of NARDL asymmetric framework was 7.67. This calculated value exceeds that of the upper bound critical value of 3.86 at the 1% level of significance thereby indicating cointegration relationship. More importantly, the ECM term represented by ECT (−1) was −0.8390 as shown in Table 3 was statistically significant at 1% and also lied between −1 and 0. This implied that the speed of adjustment towards equilibrium was 84% annually when a shock occurred.

Pesaran et al. (2001) focused on the significance of the F-statistic while Banerjee et al. (1998) preferred the significance of the ECM term value which should lie between −1 and 0 for the confirmation of long-run equilibrium relationship. Both criteria were satisfied by NARDL framework in Table 5 below.

Table 5. Short-run NARDL estimates of the impact of cashew nuts and cocoa beans exports on economic growth.

Variable	Coefficient	Std. Error	Prob.
D(LCASHEW_POS)	0.016***	0.004	0.001
D(LCASHEW_POS(−1))	−0.016***	0.005	0.002
D(LCOCOA_POS)	0.110***	0.020	0.000
D(LCOCOA_NEG)	0.049*	0.026	0.076
D(LEXC_POS)	0.057***	0.016	0.002
D(LEXC_NEG)	−135.166***	10.969	0.000
D(LFDI_NEG)	0.067***	0.012	0.000
D(LFDI_NEG(−1))	−0.089***	0.015	0.000
D(LINF_NEG)	−0.001	0.005	0.883
D(LINF_NEG(−1))	−0.028***	0.006	0.000
ECT(−1)	−0.839***	0.068	0.000

Note: All variables are log transformed. *** and * indicate significance levels of 1% and 10% levels, respectively.

Source: Author’s Computation (2021).

4.5. Short-run asymmetric results analysis

It is imperative to emphasis that the interpretations of the coefficients for the negative partial sums (decrease/decline) are not conventional. That is, the signs of the coefficients for negative changes in cashew nuts and cocoa beans exports under the NARDL framework are interpreted in the opposite.

From the short-run results in Table 5 below, positive changes in cashew nuts and cocoa beans exports have a significant impact on economic growth in different magnitudes. For instance, on the average, a 1% increase or shift of cashew nuts exports in the current period in the short-run would increase annual economic growth by approximately 0.02%, all things being equal.

More specifically, positive changes of cashew nut exports in the short-run (with a lag) have a negative significant impact on economic growth. This is attributed to the low price that farmers and exporters are likely to receive with a potential glut in the market. This therefore allows for importers and buyers to negotiate for low prices with a corresponding potential for supplier and farmers to hold on to cashew nuts. This is supported by Cassama (2022) who concluded that there is a negative effect on the unit price on the export volumes when supplies are greater than demand and confirms the findings of the results. On the average, a 1% increase in cashew nuts exports in the short-run would decrease the annual economic growth rate by approximately 0.02%, all things being equal. The findings are compatible with empirical studies of Sanjuán-López and Dawson (2010) who found non-traditional agricultural exports to exert a positive significant effect on the economic growth of least developed countries, and Urriola et al. (2018) who also found that non-traditional agricultural exports have a positive significant impact on the economic growth of Peru in short-run. However, the finding deviates from the findings of Urriola et al. (2018) who found non-agricultural exports to have a positive significant impact on economic growth in Peru in the short-run with a lag term.

In the case of cocoa bean exports, on the average, a 1% increase in exports of cocoa beans in the short-run would increase the annual growth rate by approximately 0.11% all things being equal. The finding conforms to that of Okyere (2020) and Siaw et al. (2018) who both revealed a positive relationship between cocoa exports and economic growth in Ghana. Also, Shashi and Vigneri (2011) suggested an evidence of a positive association between the cocoa sector and the growth rate of the Ghanaian economy. However, contrary to positive changes in cocoa bean exports, negative changes in exports of cocoa beans also have significant impact on economic growth in the short-run. On the average, a 1% decrease in exports of cocoa beans in the short-run would decrease the annual growth rate by approximately 0.05% holding the rest of the regressors constant. The finding also agrees with that of Noula et al. (2013) who both revealed that cocoa export exerted a negative and significant impact on economic growth in Cameroon in the short-run.

With respect to the control variables, it is interesting to note that all the positive and negative changes have a significant impact on economic growth (except the negative changes of inflation in the current period) in the short-run. Thus, for the positive changes in exchange rate, on the average, a 1% appreciation of the local currency in the short-run would increase the annual economic growth rate by approximately 0.06% ceteris paribus. Also, for the negative changes in the exchange rate, a 1% depreciation of the local currency in the short-run would increase the annual economic growth rate by approximately 135.17% all other things being equal. The finding disagrees with the findings of Ikechukwu (2021), who found that the exchange rate in Nigeria exerted a negative and insignificant relationship with economic growth.

Furthermore, on the average, a 1% decrease in foreign inflows within the current period in the short-run would decrease economic growth by approximately 0.07% all other things being equal. This finding is in consistent with studies like Xu (2000) who found FDI to have a negative effect on economic growth in some provinces of China, and Jude and Levieuge (2015) who found for a panel of developing countries that FDI on its own does not promote growth. Also, on the average, a 1% decrease in foreign inflows (with a lag) term in the short-run would increase economic growth by approximately 0.09% all other things being equal. This result contradicts the findings of Carbonell and Werner (2018) who found FDI to exert a negative insignificant impact on growth in Spain.

In addition, a 1% decrease in inflation in the contemporaneous term in the short-run on average, would increase economic growth by approximately 0.0007% all other things being equal. It is however not significant on economic growth. However, on the average, a 1% decrease in inflation (with a lag) term in the short-run would increase annual economic growth by approximately 0.03% all other things being equal. This finding deviates from the findings of Faridi (2012) who found that inflation exerted a negative insignificant relationship with GDP in Pakistan and that of Salifu (2019) who found inflation to exert a negative significant impact on economic growth in Ghana.

4.6. Long-run asymmetric results

The long-run asymmetric impact of cashew nuts and cocoa beans exports in Table 4 above, is derived by taking the negative ratio of the coefficients of the positive and negative changes to the coefficient of the lagged dependent variable and by applying the Wald test to determine the significance of the coefficients.

From the long-run results in Table 6 below, positive changes in cashew nuts exports have a significant impact on economic growth. It is clear that on average, a 1% increase or shift of cashew nuts exports in the long-run is likely to increase annual economic growth by approximately 0.05%, ceteris paribus. This is consistent with the short-run results estimated earlier and findings were also compatible with other studies like Sanjuán-López and Dawson (2010) and Faridi (2012) who both found non-traditional agricultural exports to exert a positive significant effect on the economic growth of the least developed countries and Pakistan, respectively in the long-run.

Table 6. Long-run NARDL estimates of the impact of cashew nuts and cocoa beans exports on economic growth.

Variable	Coefficient	Std. error	Prob.
LCASHEW_POS	0.051***	0.012	0.000
LCASHEW_NEG	0.002	0.011	0.863
LCOCOA_POS	0.203***	0.033	0.000
LCOCOA_NEG	0.183***	0.054	0.003
LEXC_POS	0.163***	0.035	0.000
LEXC_NEG	−146.123***	38.876	0.001
LGCF_POS	0.105*	0.053	0.060
LGCF_NEG	−0.079*	0.042	0.073
LFDI_POS	0.003	0.010	0.772
LFDI_NEG	0.260***	0.030	0.000
LINF_POS	−0.017	0.018	0.375
LINF_NEG	0.005	0.028	0.870

Note: All variables are log transformed. ***^, **, and * indicate significance levels of 1, 5, and 10% levels, respectively.

Source: Author’s Computation (2021).

However, contrary to the positive changes, negative changes in cashew nuts exports do not have significant impact on economic growth. Thus, on average, a 1% decrease in cashew nuts exports is expected to decrease annual economic growth by approximately 0.002% in the long-run, ceteris paribus. This finding is not consistent with the findings of Sanjuán-López and Dawson (2010) who found non-traditional agricultural exports to exert a positive significant effect on the economic growth of least developed countries, and Urriola et al. (2018) who also found that non-traditional agricultural exports have a positive and significant impact on the economic growth of Peru in the long-run.

In the case of cocoa bean exports, both the positive and negative changes had a significant impact on economic growth. Consistent with the short-run results, cocoa bean export was found to have a positive impact on growth. Thus, on the average, a 1% increase in cocoa bean exports in the long-run would increase annual economic growth by approximately 0.20%, all things being equal. Cocoa is basically an international commodity with a high demand, and it is advantageous for countries to produce more cocoa towards increasing its GDP. Thus, cocoa export is directly inclined to influence economic growth in Ghana. Empirical results by Okyere (2020) and Siaw et al. (2018) revealed a positive relationship between cocoa exports and economic growth in Ghana. Also, Shashi and Vigneri (2011) showed evidence of a positive association between cocoa sector and growth rate in the Ghanaian economy. Also, on the average, a 1% decrease in cocoa bean exports in the long-run would decrease annual economic growth by approximately 0.18%, ceteris paribus. This finding is line with the findings of Noula et al. (2013) who discovered that, in Cameroon, cocoa export negatively relates to economic growth.

With respect to the control regressors, both the positive and negative changes do have significant impacts on economic growth in Ghana in the long-run. However, not all the variables have significant effects on economic growth as evidenced by their p-values. For instance, on the average, a 1% increase in the local currency in the long-run, would increase annual economic growth by approximately 0.16%, all things being equal. Also, a 1% depreciation of the local currency in the long-run would increase economic growth by approximately 146.12%, all things remain constant and confirms findings by Baylie (2011), Henneberry and Khan (2000), Siaw et al. (2018), and Yifru (2015), respectively. However, it contradicts studies by Adjei (2019) and Mwinlaaru and Issac (2019), who both found the exchange rate to exert a negative significant effect on growth in Ghana in the long-run.

In the same vein, for the gross capital formation, a 1% increase in investment in the long-run would increase annual economic growth by approximately 0.10%, ceteris paribus. Also, a 1% decrease in investment in the long-run would increase economic growth by approximately 0.08%, holding the rest of the regressors constant. However, an increment in domestic investment over the years has not been able to spur the desired boost in economic growth in Ghana. This finding is in line with that of Bakari and Mabrouki (2017), Njimanted and Aquilas (2015), and Siaw et al. (2018) who established a positive and significant correlation among investment and economic growth in Ghana, South Eastern Europe Countries, and Cameron, respectively.

Also, a 1% increase in foreign inflows will increase economic growth by approximately 0.003% in the long-run but this was not significant on economic growth. Findings also revealed that a 1% decrease in foreign inflows in the long-run would decrease economic growth by approximately 0.26% all other things being equal. These findings are consistent with studies by Xu (2000) who found FDI to have a negative impact on economic growth for selected provinces in China, Jude and Levieuge (2015) found that in several developing countries that FDI on its own does not promote growth and Carbonell and Werner (2018) found FDI not to have a significant impact on growth in Spain. The crowding out of domestic firms and lack of absorptive capacity specifically human capital, sound institutions, and developed financial systems are often noted as being the underlying reasons for the insignificant or negative effects of FDI on growth especially for least developed countries.

Finally, inflation does not have any impact on economic growth either positively or negatively and this was confirmed by the 1% increase in inflation, which led to a decline in the annual growth rate by 0.02% in the long run. On the other hand, a 1% decrease in inflation in the long-run would decrease annual economic growth by approximately 0.005%, ceteris paribus. The finding confirms that of Faridi (2012) who found that inflation exerted a negative and insignificant relationship on GDP in Pakistan and Salifu (2019) who found inflation to exert a significant but negative impact on economic growth in Ghana.

Results from the comparative analysis of the short-run and long-run asymmetric impact of cashew nuts and cocoa beans exports on Ghana’s economic growth, it can be concluded that both cashew nut exports and cocoa beans exports have significant impacts on economic growth in the short and long-runs. However, cocoa bean exports had a greater impact on Ghana’s economic growth compared to cashew nuts exports as evidenced in the short and long run coefficients. In the case of cashew nuts exports, the coefficients of the positive asymmetric effects on economic growth were 0.02 and −0.02, respectively for the short-run, and 0.05 for the long-run. Also, in the case of cocoa bean exports, the coefficients of the positive and negative asymmetric effects on economic growth were 0.11 and 0.05 for the short-run, and 0.2 and 0.18 for the long-run.

4.7. Granger causality test

To draw the further inference, a Granger causality also known as block exogeneity Wald test is conducted under both the linear ARDL and NARDL approaches. Results of the pairwise Granger causality test among all the export variables and economic growth are presented in Table 7 below.

Table 7. Results of Granger causality test.

Null hypothesis	F-statistic	Prob.	Results
LCASHEW → LRGDP	0.482	0.621	Unidirectional
LRGDP → LCASHEW	6.433	0.004
LCOCOA → LRGDP	0.743	0.482	Unidirectional
LRGDP → LCOCOA	3.296	0.047

Source: Author’s Computation (2021).

Results of the Granger Causality test as shown in Table 7 above indicate that there is a strong evidence of a one-way Granger causality between cashew nuts exports, cocoa beans exports, and economic growth (RGDP) in Ghana. This implies that cashew nuts and cocoa beans exports have a unidirectional impact on economic growth. Thus, RGDP Granger causes both cashew nuts and cocoa beans exports in Ghana and this is attributed to the fact that these products are generally exported as raw materials rather than value-added products, and a higher gross domestic product increases investment in these sectors. Similar findings were found for Urriola et al. (2018) in Peru, Ouma et al. (2016) in Uganda, Tanzania, and Burundi, Yifru (2015) in Ethiopia, and Faridi (2012) in Pakistan.

5. Conclusion

Evidence of decelerating trend in the growth of cashew nuts exports relative to an accelerating trend in cocoa beans exports was recorded from 1970 to 2019. There was also evidence of a long-run symmetric and asymmetric cointegration relationship between cashew nuts exports, cocoa beans exports, and economic growth and these products had a positive and significant impact on Ghana’s economic growth with greater economic impacts.

There period under review showed evidence of both short-run and long-run asymmetric impact of cashew nuts exports on economic growth compared to only short-run asymmetric effects of cocoa bean exports. Exchange rate fluctuations impact negatively on cashew nuts exports in the short-run but the exchange rate continuous to determine the price of cocoa beans which translates into the growth of the cocoa exports in the long-run. Also, real GDP continuous to play a significant role in the performance of cashew nuts exports but exchange rate, FDI, and inflation affect the performance of cocoa beans exports. There is a unidirectional (one-way) causal relationship between economic growth, cashew and cocoa beans exports. Exports of both cashew nuts and cocoa beans contribute positively to Ghana’s economic growth in the long-run in line with the agricultural export led-growth. Government and policy makers should continue to encourage and promote agricultural export growth to spur the economic development of Ghana. Also, value-added cocoa beans and cashew nuts will attract more earnings to farmers, exporters, and the government leading to exchange rate stability.

The government through the Ministry of Trade and Industry (MoTI) should continue to encourage and promote traditional and non-traditional agricultural exports growth since the two sectors are integral to the economic development of Ghana. The government through the Ministry of Food and Agriculture (MoFA) should invest in improved agricultural technologies specifically improved varieties, pest and disease spraying to improve production, storage, branding, and packaging to make NTAEs attractive in the international market.

Acknowledgements

This manuscript was culled out of the postgraduate thesis of the Lead Author and no part of this study was funded by any agency or organization.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

Data will be made available upon reasonable request.

Additional information

Notes on contributors

Ibrahim Abdul-Karim

Ibrahim Abdul-Karim is a holder of a Master of Philosophy (MPhil) degree in Agricultural Economist with focus on international trade and time series forecasting.

Osman Tahidu Damba

Osman Tahidu Damba is a PhD holder in Agricultural Economics and currently a Senior Lecturer with the Department of Agricultural and Economics at the University for Development Studies (UDS), Ghana. He has researched in the area of agribusiness, climate change and international trade. He has published in peer reviewed journals supported by grants from different development agencies.