Economic complexity is a measure of productive capabilities indirectly by looking at the mix of sophisticated products that countries export. The economic complexity index proposed a proxy for diversity and ubiquity of products in the export basket.

This study seeks to determine if economic complexity can influence the inequality measured by the Gini index in some selected sub-Saharan African countries.

The need for the study emanates from the notion that that economic complexity can reduce income inequality hence it is imperative to investigate this relationship in the sub-Saharan African region where most countries produce few sophisticated goods that are also labour-intensive. Inadequate literature within the African continent has also contributed to the formulation of this study.

This study employed the autoregressive distribution lag (ARDL) model to analyze a panel data set, which includes eight sub-Saharan African countries for the period 1994–2017.

We found that economic complexity can reduce income disparities.

Sub-Saharan African countries should shift their productive capabilities and resources from primary to sophisticated products in the manufacturing and services sector to increase economic complexity and reduce inequality.

The study makes an important contribution to the debate about the relationship between economic complexity and income inequality in the sub-Saharan African context and it is envisaged that it will inform the actions of the decision-makers to drive future productivity and prosperity in the region.

Sub-Saharan Africa (SSA) remained the second most disproportionate continent following Latin America and inequalities have grown over time, as shown by the Gini index (World Bank

Further, a rapid economic expansion in relation to economic complexity has focused on developing a country’s export sophistication (Bhorat, Steenkamp & Rooney

Despite the fact that Africa is a naturally resource-rich continent, there is a need for structural transformations (McMillan & Rodrik 2014). Monga (

Despite the fact that SSA has advanced in some economic indicators like commodity prices and financing conditions in the past two decades, it has been slow on economic diversification (Yellapragada

Hartmann et al. (

Hartmann et al.’s (

In order to explore the idea predicting that economic complexity can influence the economy further, the study aligns itself with the theoretical views of neo-Schumpeterian economics and evolutionary authors. Complexity ideas have been drawn from a unifying approach for evolutionary and neo-Schumpeterian streams. This pluralism is reflected in the fact that several authors such as Robert, Yoguel and Lerena (

In economic complexity there are knowledge capabilities, and a country with limited knowledge capabilities can be challenged to produce complex products (Mariani et al.

In addition to the measure of knowledge, diversity and ubiquity seem to be crucial. Hence, an economy is regarded as complex if it exports different sophisticated products. Cristelli et al. (

Countries producing more sophisticated products like Japan, Germany and the United States of America (USA) are the largest growing economies in the world that enjoy rapid economic growth. According to the ECI (2018), Japan is rated number one followed by Germany, whilst US is in the seventh place. The irony of the ECI is that according to the data from the International Monetary Forum (IMF; 2018), US has the largest economy, with a GDP amounting to 20.4 trillion, Japan is positioned in the third place with a GDP amounting to 5.1 trillion and followed by Germany in the fourth position by a GDP of 4.2 trillion, all measured in American dollars. This might be because of what products are regarded as complex. Felipe et al. (

Most of the African countries are found at the bottom section of the economic complexity rankings list of 129 countries. The ECI of the chosen SSA countries are ranked in the Atlas of Economic complexity (2014) as follows: South Africa (55), Kenya (73), Tanzania (95), Cote d’Ivoire (99), Ghana (100), Mozambique (109), Cameroon (121) and Nigeria (122). Countries found in this bottom section were described by Desjardins (

On the other hand is the fact that the SSA economies have the most unequal income distributions patterns, as compared to that of the world (Erkan & Yildirimci

The study by Lee and Vu (2019) examined if economic complexity can influence income inequality using an ordinary least squares (OLS) regression analysis and generalised method of moments (GMM) estimator. The study used panel data of 96 countries from 1980 to 2014. The results of the cross-country OLS regression analysis indicated that economic complexity can predict income inequality. The findings from the GMM estimator indicated that there is a positive relationship between economic complexity and income inequality, and this is in contrast with the OLS estimates and Hartmann et al. (

Inspired by Hartmann et al. (

The study made use of a panel data set, which includes eight SSA countries that ranked highest to lowest in the region on the Atlas of Economic Complexity, for the period 1994–2017 (South Africa, Kenya, Tanzania, Cote d’Ivoire, Ghana, Mozambique, Cameroon and Nigeria). Besides the challenges with availability of data, these countries were selected to observe how countries that ranked the highest perform compared to the ones ranked lowest. Data on ECI was obtained from the Atlas of Economic Complexity by Hausmann et al. (

Where, ∞ represents a constant parameter, LGINI is the logarithm of Gini index, ECI, LGFCF is the logarithm of gross fixed capital formation, FDI is foreign direct investment and LGDPPC is the logarithm of gross domestic product per capita, measured in purchasing power parity (US$) as a measure of economic growth in the region. Variables that are out of range are standardised for meaningful and robust results (Brooks

It is expected that economic complexity should have a negative effect on the Gini index (Hartmann et al.

The following econometric procedures will be undertaken to determine if economic complexity can influence inequality in some selected SSA countries.

The panel unit root tests are helpful in choosing the model which is suitable for the data used in the study. Before testing for cointegration (long run relationships), the study employed several approaches to test for unit root in the panel data, namely, the Fisher-Augmented Dickey-Fuller (ADF), Im, Pesaran and Shin (

The IPS test was chosen because it relaxes the restrictive assumption of the LLC test (Pesaran

The Kao cointegration test extends the Engle-Granger framework to test cointegration in panel data. The Kao test follows the basic approach of the Pedroni panel cointegration test, although the Kao test specifies the cross-section specific intercept and homogenous coefficients on the initial stage regressors.

The Fisher cointegration test uses the results of the individual independent tests (Fisher 1932). The panel cointegration was pioneered by Maddala and Wu (

The ARDL models are standard least squares which incorporate lags of both independent and depended variables as regressors (Pesaran & Pesaran

The inverse root of the Autoregressive model (AR) characteristic polynomial is used to test for stability of the model. According to Agung (

This article followed all ethical standards for a research without direct contact with human or animal subjects

The first step towards achieving our goal was to determine the order of integration of the variables through unit root testing. It was found that the three tests (LLC, IPC, Fisher-ADF) show a mixture of levels (

After determining the order of the integration, the next step was to establish the presence of cointegration in the system. The Kao panel cointegration test was employed and the results with the deterministic trend specification of individual intercepts are presented in

The next step was to find the long run and short run estimates in the system, and the results are presented in

Panel autoregressive distribution lag long-run estimates.

Variable name | Coefficients | Standard error | Probability | |
---|---|---|---|---|

ECI | −0.006338 | 0.003427 | −1.849480 | 0.0671 |

LGFCF | 0.184012 | 0.011395 | 16.14835 | 0.0000 |

FDI | 0.006588 | 0.000773 | 8.527436 | 0.0000 |

LGDPPC | −0.017070 | 0.010962 | −1.557234 | 0.1224 |

ECI, economic complexity index; FDI, foreign direct investment; LGDPPC, logarithm of gross domestic product per capita; LGFCF, logarithm of gross fixed capital formation.

Autoregressive distribution lag short run analysis results.

Variable names | Coefficients | Standard error | Probability | |
---|---|---|---|---|

ECT(−1) | −0.303740 | 0.155029 | −1.959251 | 0.0527 |

D(ECI) | −0.007905 | 0.015066 | −0.524672 | 0.6009 |

D(ECI(−1)) | 0.007864 | 0.006126 | 1.283714 | 0.2020 |

D(LGFCF) | −0.070737 | 0.024906 | −2.840187 | 0.0054 |

D(LGFCF(−1)) | 0.015471 | 0.020946 | 0.738591 | 0.4618 |

D(FDI) | −0.001165 | 0.001006 | −1.158768 | 0.2491 |

ECI, economic complexity index; FDI, foreign direct investment; LGFCF, logarithm of gross fixed capital formation.

In order to capture the impact and also the relationship between Gini-coefficient and its regressors, the coefficients

The long run equation indicates that if ECI increase by a percentage, GINI coefficient (LGINI) will decrease by 0.006, and has a significant negative influence on dependent variable. These findings concur with Hartmann et al. (

Finally, the remaining macroeconomic variables demonstrate a significant positive influence on the dependent variable because the implication is that if gross fixed capital formation (LGFCF) and FDI go up by 1%, GINI coefficient will increase by 0.184 and 0.007, respectively. With regards FDI, the results concur with the findings of Kaulihowa and Adjasi (

After cointegration, the next step was the estimation of the short run analysis. The purpose was to use the error correction model to determine both the speed of adjustment and also the short run estimates for each variable on the Gini coefficient. Speed of adjustment is reflected by how much percentage equilibrium will be restored in the next period and the results are presented in

Based on

In order to investigate whether the long run relationships established earlier are stable over the proposed period of study, the inverse root of AR characteristic polynomial was employed and the results are presented in

Stability reliability test.

The study intended to determine if economic complexity can influence inequality in some selected SSA countries. This was achieved by determining whether economic complexity is a significant and negative predictor of income inequality in the region. The panel autoregressive distributive lag (ARDL) model was found to be a suitable approach for the analysis after the results of the three tests showed a mixture of orders of integration (

The long run analysis indicated that even though it was found to be statistically significant at the 10% level, ECI together with GDP per capita have the expected negative influence on the Gini coefficient. These findings are in cohesion with the findings of Cristelli et al. (2016) and as such the views of Yellapragada (

Sub-Sahara African countries should also expand their productive capabilities from primary sector to manufacturing and services sector, to produce and export more complex and sophisticated products. This allows for a far better ECI and thus reduces inequality. Further, negotiating an increased minimum wage with employers by government and trade unions may help bridge the income gap between rich and poor. These policies must be implemented by not only the selected SSA countries, but also the members of the AfCFTA which envisaged stimulated export sophistication across the continent.

Furthermore, gross fixed capital formation and FDI have a positive effect on the Gini-coefficient and they are significant at 1% level. With regards to FDI, these results concur with the findings of Kaulihowa and Adjasi (

The authors have declared that no competing interests exist.

All authors contributed equally to this work.

This research received no specific grant from any funding agency in the public, commercial or not-for-profit sectors.

Data sharing is not applicable to this article as no new data were created but obtained from reliable sources mentioned in the article.

The views and opinions expressed in this article are those of the authors and do not necessarily reflect the official policy or position of any affiliated agency of authors.

Summary of panel unit root test.

Variables | Tests | Tests equations | ||
---|---|---|---|---|

LGINI | LLC | Individual and intercept | 0.5299 | 0.0000 |

Individual, intercept and trend | 0.9649 | 0.0000 | ||

None | 0.9167 | 0.0000 | ||

IPS | Individual and intercept | 0.6349 | 0.0000 | |

Individual, intercept and trend | 0.9478 | 0.0001 | ||

Fisher-ADF | Individual and intercept | 0.8805 | 0.0000 | |

Individual, intercept and trend | 0.9833 | 0.0008 | ||

None | 0.9966 | 0.0000 | ||

ECI | LLC | Individual and intercept | 0.0026 | - |

Individual, intercept and trend | 0.0171 | - | ||

None | 0.0397 | - | ||

IPS | Individual and intercept | 0.0010 | - | |

Individual, intercept and trend | 0.0035 | - | ||

Fisher-ADF | Individual and intercept | 0.0012 | - | |

Individual, intercept and trend | 0.0082 | - | ||

LGFCF | LLC | Individual and intercept | 0.3095 | 0.0000 |

Individual, intercept and trend | 0.2108 | 0.0000 | ||

None | 0.1770 | 0.0000 | ||

IPS | Individual and intercept | 0.3237 | 0.0000 | |

Individual, intercept and trend | 0.0890 | 0.0000 | ||

Fisher-ADF | Individual and intercept | 0.4277 | 0.0000 | |

Individual, intercept and trend | 0.1236 | 0.0000 | ||

None | 0.6858 | 0.0000 | ||

LGDPPC | LLC | Individual and intercept | 0.5776 | 0.0004 |

Individual, intercept and trend | 0.6216 | 0.0003 | ||

None | 1.0000 | 0.0022 | ||

IPS | Individual and intercept | 0.9985 | 0.0014 | |

Individual, intercept and trend | 0.6750 | 0.0200 | ||

Fisher-ADF | Individual and intercept | 0.5776 | 0.0035 | |

Individual, intercept and trend | 0.6216 | 0.0223 | ||

None | 1.0000 | 0.0493 | ||

FDI | LLC | Individual and intercept | 0.1305 | 0.0000 |

Individual, intercept and trend | 0.7326 | 0.0000 | ||

None | 0.0077 | - | ||

IPC | Individual and intercept | 0.0054 | - | |

Individual, intercept and trend | 0.0519 | - | ||

Fisher-ADF | Individual and intercept | 0.0135 | - | |

Individual, intercept and trend | 0.0912 | - | ||

None | 0.1152 | 0.0000 |

ADF, Augmented Dickey-Fuller; ECI, economic complexity index; FDI, foreign direct investment; IPC, Im, Pesaran and Shin tests; LGDPPC, logarithm of gross domestic product per capita; LGFCF, logarithm of gross fixed capital formation; LGINI, logarithm of Gini index; LLC, Levin, Lin and Chu tests.

Kao panel cointegration test results.

Variable | ||
---|---|---|

ADF | −1.819179 | 0.0344 |

Residual variance | 0.000220 | - |

HAC variance | 0.000200 | - |

ADF, Augmented Dickey-Fuller; HAC, Heteroskedasticity- and autocorrelation-consistent.

Johansen Fisher panel cointegration with linear deterministic trend test.

Hypothesised number of CE(s) | Fisher stat. (trace test) | Probability | Fisher stat. (max-eigen test) | Probability |
---|---|---|---|---|

None | 197.1 |
0.0000 | 149.1 |
0.0000 |

At most 1 | 75.57 |
0.0000 | 53.33 |
0.0000 |

At most 2 | 36.61 |
0.0024 | 27.14 |
0.0400 |

At most 3 | 22.24 | 0.1356 | 17.86 | 0.3319 |

At most 4 | 26.81 |
0.0436 | 26.81 |
0.0436 |

Note: *, **, and *** indicate that the

CE, cointegrating equation.

Johansen Fisher panel cointegration test with no deterministic trend test.

Hypothesized number of CE(s) | Fisher stat. (trace test) | Probability | Fisher stat. (max-eigen test) | Probability |
---|---|---|---|---|

None | 165.6 |
0.0000 | 125.7 |
0.0000 |

At most 1 | 63.34 |
0.0000 | 43.04 |
0.0003 |

At most 2 | 34.33 |
0.0049 | 24.24 |
0.0844 |

At most 3 | 24.18 |
0.0856 | 23.75 |
0.0952 |

At most 4 | 13.43 | 0.6412 | 13.43 | 0.6412 |

CE, cointegrating equation.

Johansen Fisher panel cointegration test with Quadratic deterministic trend.

Hypothesised number of CE(s) | Fisher stat. (trace test) | Probability | Fisher stat. (max-eigen test) | Probability |
---|---|---|---|---|

None | 214.0 |
0.0000 | 143.2 |
0.0000 |

At most 1 | 91.31 |
0.0000 | 59.85 |
0.0000 |

At most 2 | 46.47 |
0.0001 | 28.11 |
0.0307 |

At most 3 | 30.32 |
0.0164 | 19.85 | 0.2270 |

At most 4 | 39.84 |
0.0008 | 39.84 |
0.0008 |

CE, cointegrating equation.