The Kuznets Curve Hypothesis Checked Out on Up-To-Date Observations in African Countries

Abstract

The purpose of this article is to study empirically whether the Kuznets’ curve hypothesis on inequality and development is present in the economies located in Africa, as well as whether there is a minimum income for this hypothesis to begin to be fulfilled. In order to study this question, a panel of data from 45 countries is available for the period 1975–2019, and these data are analysed through a graphical point of view and through an econometric analysis using the pooled mean group estimator. The results obtained allow us to conclude that there is evidence in favour of Kuznets’ hypothesis and that a minimum level of income is required for it to be significantly observed. As well as the fact that today there is still a palpable heritage of European colonization.

JEL Classification: D63, O10, O11.

Keywords

Kuznets’ curve economic inequality development Africa

Introduction

At present, there are many countries that are at different economic levels, and these can be classified according to different economic variables. In this article, we will focus on the relationship between the level of development and the level of inequality, as Kuznets (1955) proposed in his article. This approach is important in order to be able to design an economic policy that seeks to attack economic inequality, which can generate so much social conflict in the countries that suffer it, as well as erode the effective functioning of the economy. In recent years there seems to have been greater interest in the theory derived from Kuznets’ hypothesis, the Environmental Kuznets Curve (EKC), which covers the environmental approach and has aroused interest in economic literature. In fact, numerous works have been written in this field recently such as those of Aziz et al. (2020); Caglayan-Akay and Kangalli-Uyar (2021); Dogan and Inglesi-Lotz (2020); Ike et al. (2020); Kacprzyk and Kuchta (2020); Purcel (2020); Sarkodie and Ozturk (2020); Usman et al. (2020) among others. However, the same priority is not given to studying the original hypothesis, which is the objective of this work.

Among the world’s poorest countries, the top ten are on the African continent as shown in the United Nations Development Programme (UNDP) 2019 Human Development Report (UNDP, 2019)(, however, as we shall see below, the fact that these countries have low incomes does not mean that they have an equitable distribution of wealth.

There are also at least two major reasons for studying inequality on the African continent: a) the fact that Africa is a continent very rich in natural resources and has great potential for development; and b) the existing debate on the strength of the Kuznets Curve.

First, it is now understood that Africa’s potential has not been exploited for its own development because of the turbulent history of this continent. Africa is notably marked by slavery, colonization and the subsequent struggle for independence of African colonies, a process that generated political instability on the continent aggravated by the Cold War. Since the mid-20th century, this development has also been hindered by African elites and their ability to influence the politics and economy of the various countries that make up the continent. All this historical baggage has led to the fact that African resources have been used for the benefit of other continents or only for a certain segment of the African population.

Second, Kuznets’ hypothesis on inequality and development has long been discussed since it was first put forward in 1955, having both supporters and detractors. This article justifies and claims that the hypothesis is observable in Africa, as well as that these countries still have the heritage of the colonial era in mind. Besides, it is intuited that there is a minimum income to be able to experience the Kuznets’ curve and in some African countries this minimum is not reached. To achieve this, initially a literature review will be conducted on the Kuznets curve to locate where the issue lies. This will be followed by some data on income and inequality in Africa to give insights into where the economy is. At this point, the empirical section will begin with the presentation of the model, as well as the results obtained and a section on robustness to check whether or not these results are sensitive to the variables that have been used. We will finish this work with our own conclusions.

Background

Theorical literature

The Kuznets Curve, as proposed by the original author, indicates that inequality in terms of economic development, as measured by the Gini index and per capita income respectively, describes an inverted U-curve. That is, countries initially have low levels of inequality. Inequality grows as the country develops until it reaches a maximum where new economic forces will appear that will cause a decrease in inequality. The economic forces that initially put upward pressure on inequality are the concentration of savings in the wealthy classes and rural-urban migration. Subsequently, productivity differentials, the generation of new job opportunities and the birth rates of different segments of the population reduce the inequality initially created. This is the hypothesis put forward by Kuznets (1955).

Kuznets himself in the conclusions stated “[t]he paper is perhaps 5 per cent empirical information and 95 per cent speculation, some of it possibly tainted by wishful thinking”. This statement would provoke many economists to analyse his hypothesis, finding both supporters (Barro, 2008; Greenwood and Jovanovic, 1990; Higgins and Williamson, 1999; León, 2021; Lindert, 1986; Martínez-Navarro et al., 2020; Shahbaz et al., 2015) as well as detractors who argued that there was no such relationship between inequality and development (Baymul and Sen, 2019; Castelló-Climent, 2010; Deininger and Squire, 1996,1998; Kanbur, 2017), and even some who claimed to find no relationship (Theyson and Heller, 2015; Thomas, 2015).

We also find extrapolations of the Kuznets hypothesis to other areas. The most popular and mature is known as the EKC, which studies CO₂ emissions in relation to development. This approach began in the 1960s with works such as Hakkert and de Zeeuw (1968), and investigation has continued since then with works such as Cole et al. (1997); Dasgupta et al. (2002); Tsurumi and Managi (2010) and Ahmad et al. (2021). There are also studies that have tried to apply the Kuznets curve to particular cases. These include the Kuznets curve for obesity (Grecu and Rotthoff, 2015); the Kuznets curve for tourism (Raza and Shah, 2017); the Kuznets curve for crime (Buonanno et al. (2017); the Kuznets curve for suicide (Antonakakis and Collins, 2018), and the Kuznets curve for alcohol consumption (Cantarero-Prieto et al. (2019), among other applications.

In conclusion, since its publication in 1955, Kuznets’ hypothesis has been studied by numerous authors and evidence both for and against has been found. In this study, we will demonstrate the hypothesis by focusing on African countries, as well as considering the repercussions of British or French colonization for these countries, in addition to investigating whether these economies, which have the lowest per capita income in the world, actually have enough income to experience Kuznets’ hypothesis.

Empirical literature

As for the theoretical approach to the Kuznets curve and its variants, there is a multitude of studies that delves more specifically into the empirical issues.

The most relevant empirical work in this respect is Piketty (2013). In this contribution Piketty expands on the work done by Kuznets, with three fundamental differences. First, Piketty calculated economic inequality through taxpayers’ tax returns. Second, he based his study on the growth rate for gross domestic product (GDP) instead of using GDP per capita. And finally, his sample was larger than Kuznets’, since in addition to the United States he also took into account France, Germany, Great Britain, Japan and Sweden. In this way, Piketty observed that economic inequality has increased since the 1950s, due in particular to the concentration of income, which is understood as the totality of property rights over capital, land, shares, bonds and so on.

For Piketty, the reason for this increase in inequality is clear: it is due to the fundamental divergence force, that is: r > g, where r is the average rate of return on capital and g is the rate of GDP growth. Thus, Piketty argues that whenever the average rate of return on capital is higher than the GDP growth rate, inequality will be generated. Based on this hypothesis, Piketty contradicts Kuznets by arguing that the reduction in inequality observed by Kuznets in 1955 was due to the reduction in the rate of return on capital together with the simultaneous increase in the rate of economic growth.

Piketty’s theory seems to be well supported by data and empirical analysis to discredit Kuznets’ hypothesis. However, his analysis is more limited than Piketty assumes in his book, since authors such as Acemoglu and Robinson point out that for Piketty’s empirical model to apply to the real case, 100% of the income received by the owners of capital must be saved, which is a starting hypothesis that could never be fulfilled. Likewise, other authors such as Fortin and Lemieux (2016) criticise Piketty’s empirical analysis as it is based on the richest 1% of the population in highly developed countries, while Lumieux points out that the remaining 99% of the population could behave differently. Based on these authors, we can conclude that Piketty’s study is specific to a certain segment of the population, and not a general study.

Focusing on specific studies examining the case of Africa, we find that inequality on the continent has been studied in depth. However, most focus on the EKC (Sarkodie, 2018; Tenaw and Beyene, 2021). In this study, focusing on the inequality hypothesis in relation to development, we have collected the most notable and recent work to contextualise how many countries, methodology and findings the preceding empirical studies have found. This information has been compiled in Table 1.

Table 1.

List of selected empirical studies.

Author	Country	Methodology	Findings
Meniago and Asongu (2018)	48 African countries	Generalized method of moments (GMM)	These South African authors found that in their case an inverted U-shape could be observed in their results according to the financial perspective.
Chancel et al. (2019)	55 African countries	Systematic analysis combining survey, fiscal and national accounts data	They conclude that they are economies with a notable presence of inequality and that this will seem to increase, except in the northeastern part of the continent, where small falls in inequality can be seen.
Adams and Klobodu (2019)	21 sub-Saharan countries	Pooled mean group (PMG) and common correlated effects mean group (CCEMG)	The authors concluded that based on their results there is no support for Kuznets’ hypothesis.
Le et al. (2020)	90 countries including 14 African countries	Panel corrected standard errors model (PCSE) and feasible generalized least squares (FGLS) for robustness check	These Vietnamese authors conclude that there is global evidence of a Kuznets curve between inequality and trade openness.
Baymul and Sen (2020)	48 countries including 12 African countries	Fixed effect estimation and random effects estimation	These authors conclude that Kuznets’ hypothesis is verified in the transformation of an industrial economy to a service economy. However, in the case of very low-income countries, where the realistic possibility of structural transformation may be a shift of workers from agriculture to services, their findings suggest that inequality could increase with further structural transformation.
Demissew Beyene and Kotosz (2020)	12 East African countries	Pooled mean group (PMG)	They conclude that there is evidence of an inverted U-shape between economic activities and environmental degradation, that is, their study reaffirms the Environmental Kuznets Curve (EKC).
Bukhari et al. (2021)	8 South Asian Association for Regional Cooperation (SAARC) countries	Fully modified ordinary least square (FMOLS)	These authors conclude that the level of enrolment at different levels of education affects inequality in different ways, which is why they propose that education policy should be formulated separately for each level of education.
Tchamyou (2021)	48 African countries	Generalized method of moments (GMM)	The results show that financial depth and financial stability are the best channels to reduce inequality. This envisages an inverted U-shaped relationship between inequality and financial development, as other authors have previously shown.
Ghosh and Mitra (2021)	39 countries including 2 African countries	Dimitrescu and Hurlin panel causality methods.	Their key finding is that developed countries show Kuznets curve behaviour in terms of the relationship between tourism and inequality, while developing countries show inverted Kuznets curve behaviour between tourism receipts and income inequality.

In conclusion, it is observed that there is generally evidence in favour of the Kuznets hypothesis within the context studied by each author (financial, environmental, etc.) except in the case of Adams and Klobodu (2019). Furthermore, it can be seen how they employ various methods for their analysis, with the generalised method of moments (GMM) and pooled mean group (PMG) methods among the most widely used. At this point, it can be said that, following the results of the Hausman test, in this study the analysis has been carried out using the PMG estimator technique.

Income and inequality in Africa

Africa is a continent with great disparities between its countries. If we speak in terms of per capita income, we find that the country with the highest income (Equatorial Guinea at US$31,486) exceeds the income of the poorest (Central African Republic with US$619) by 50 times.

Inequality in Africa is difficult to measure, as the calculation of the Gini index requires a variety of data that are not always easy to obtain and are sometimes from an imprecise source, so that different Gini indices can be found for the same year for the same country. In particular, we have used the information from the Standardized World Income Inequality Database (SWIID) as it is the largest and most general database created to date, although we are aware of its limitations as evidence Wittenberg (2015).

We can then see in Table 2 the countries with the highest level of per capita income and in Table 3 the countries with the lowest level of per capita income. Both tables alone allow us to see that there are large disparities in the volume of income in these countries. However, we could also highlight these countries on a political map, as can be seen in Figures 1 and 2.

Table 2.

Top 10 African countries with the highest per capita income.

Country	GDP per capita (US$)
Equatorial Guinea	31,486
Seychelles	24,856
Mauritius	18,852
Botswana	15,015
Gabon	14,334
Algeria	13,328
South Africa	11,949
Namibia	11,741
Tunisia	10,621
Angola	8,397

Source: Table prepared by the authors based on data from Penn World Table 10 (Feenstra, Inklaar and Timmer, 2015)

Table 3.

Top 10 African countries with the lowest per capita income.

Country	GDP per capita (US$)
Central African Republic	619
Liberia	764
Democratic Republic of Congo	836
Niger	906
Sierra Leone	1070
Mozambique	1288
Guinea-Bissau	1354
Togo	1515
Burkina Faso	1561
Mali	1604

Source: Table prepared by the authors based on data from Penn World Table 10 (Feenstra, Inklaar and Timmer, 2015).

Figure 1.

Map showing top 10 countries with the highest per capita income.

Figure 2.

Map showing top 10 countries with the lowest per capita income.

On the other hand, the countries with the lowest levels of inequality are listed in Table 4. The countries of the euro zone, considered to be benchmarks in terms of inequality, have an average level of 30 on Gini Index. Given this fact, the economies with the lowest prevalence of inequality (Table 5) are in a moderately good situation. Likewise, as for per capita income, the countries are represented on a political map in order to demonstrate the geographical distribution of these levels of inequality (see Figures 3 and 4).

Table 4.

Top 10 most economically unequal African countries.

Country	Gini index
Rwanda	43.7^c
Djibouti	43.7^a
Zimbabwe	44.3^b
Mozambique	44.7^c
Comoros	48^a
Algeria	51.3^a
Guinea-Bissau	52.1^a
Burkina Faso	53.3^d
South Africa	56.3^b
Zambia	57.1^d

Observation from 2018.

Observation from 2017.

Observation from 2016.

Observation from 2015.

Source: Table prepared by the authors based on Standardized World Income Inequality Database Version 9.1 data (Solt, 2020).

Table 5.

Top 10 African countries with least economic inequality.

Country	Gini index
Niger	27^b
Mali	31.3^c
Cameroon	32.7^b
Nigeria	32.8^h
Gabon	33.7^g
Cape Verde	33.8^f
Sao Tome	34.2^e
Ghana	34.4^b
Eritrea	34.7^b
Seychelles	35.7^a

Observation from 2018.

Observation from 2017.

Observation from 2016.

Observation from 2015.

Observation from 2014.

Observation from 2013.

Observation from 2012.

Observation from 2010.

Source: Tables prepared by the authors based on Standardized World Income Inequality Database Version 9.1 data (Solt, 2020).

Figure 3.

Map of top 10 most economically unequal countries.

Figure 4.

Map of top 10 countries with least economic inequality.

If we look at the maps together, we see that Maps 1 and 3 highlight countries in the same area, and that Maps 2 and 4 also show quite a few similarities. All these countries also share a common factor historically, among others, which is that they were areas under the control of France and Great Britain, with the British area in particular showing higher levels of income and economic inequality, and the former French regions showing both lower levels of income and inequality. This indicates that, as Amate-Fortes et al. (2015) point out, European colonisation of the African continent has determined the development and economic evolution of Africa, and we can see the legacy of those years in the present.

Finally, if we make a scatter plot in which we have in the abscissa axis the GDP per capita and in the ordinate axis the Gini index, as expected from Kuznets’ hypothesis, we obtain Figure 5. Before we set out the statistical analysis, this scatter plot already shows an inverted U pattern for this set of African economies, because the trend line takes the form of a concave curve.

Figure 5.

Graphic pattern of inequality with respect to development.

A panel data model for inequality changes

Inequality is directly affected by multiple factors. Many researchers have now investigated the issue and concluded that inequality is influenced by level of education, gender, age, geographical location, form of government and so on. However, the contribution that has had the most relevance because of the successive studies it has generated more than half a century after it was first published, the controversy that surrounds it and the topicality of its approach is the contribution of Kuznets (1955). In this research, in order to test this hypothesis, we propose to focus on the following variables: the level of inequality measured by the Gini index and the country’s development measured by the GDP per capita.

Data

The inequality variable was collected from the SWIID V. 9.1 (Solt, 2020). This source of information has the advantage that it maximizes the comparability of data on income inequality and covers a wide range of countries at different points in time. It is therefore the best database for studies that take into account the levels of inequality of different economies, in this case those in Africa.

The development variable has been obtained from Penn World Tables (PWT) 10 (Feenstra et al., 2015) published by the University of Pennsylvania. This database collects numerous data for 182 countries between 1950 and 2017. As with the SWIID, the data obtained through the PWT are fully comparable since they have been obtained through a homogeneous process, with no distortion between the income levels of different countries due to the value of their currencies, inflation or any other factor. The variable collected from the PWT is Real GDP at constant 2011 national prices (in 2011 USD thousands).

On the other hand, the Human Development Index (HDI) variable will be used for subsequent sensitivity analysis. This variable has been obtained from the 2019 Human Development Report (UNDP, 2019). As in the case of the GDP per capita variable, it is fully comparable since it has been calculated by the same body using similar criteria for each country.

Once the Kuznets’ hypothesis has been studied in the strict sense, that is, only using the development and inequality variables, an extended model will be performed that will try to better capture the behaviour of inequality. The variables that will extend the model will be the following:

1. Public spending for each country, measured as a percentage of its GDP, which will give us an approximation of the size of the public sector in each country. This variable has been included following work such as that of Lee (2005) or more recently Dotti (2020). The source from which we have extracted the data is The Heritage Foundation (2001) which offers this information openly for many countries.

2. The percentage of the population with access to electricity is another relevant variable, which has been extracted from the World Bank (2012). We decided to introduce it following the recent work of Marrero and Rodríguez (2019) and that of Sarkodie and Adams (2020). In this case, we understand access to electricity as an incentive to equal opportunities.

3. Finally, the extended model includes two dummy variables reflecting the earlier discussion on the impact of previous colonisation on development and inequality. These variables are “French Colony” and “British Colony” to distinguish whether the country in question was a British or French colony in the past. If the country was a colony of the generated dummy variable, the value will be equal to 1, and in any other case it will be equal to zero.

Table 6 below provides basic descriptive statistics on the variables to be used in this work.

Table 6.

Descriptive table of variables.

Variable	Mean	SD	Min	Max
Gini	45.96	7.34	32.7	67.1
GDP_PC	4291.63	6019.34	134	46.643
HDI	7.81	0.94	4.89	10.75
Public Spending	73.68	18.53	4.6	99.3
Electricity Access	37.44	30.82	0.01	100

Note: All data provided are unweighted averages.

GDP: gross domestic product; HDI: Human Development Index; SD: standard deviation.

Initially, the information collected has been subjected to various tests in order to decide which econometric techniques are the most suitable and whether any processing of these data is required. The results of these tests are shown in Table 7.

Table 7.

Tests performed on the regressions for the total sample.

Autocorrelation test		Heterocedasticity test		Breusch-Pagan Lagrangian multiplier
Delay	t	F	t	Chi²	t
1	0.002	32.33	0.000	198.10	0.000
2	0.026
3	0.087
Im–Pesaran–Shin (IPS) unit-root test
Variable	IPS-level			IPS-first difference
Gini index	–2.2849***	0.0028
GDP_PC	10.6994	1.0000		–2.0383***	0.0001
HDI	–0.0460	1.0000		–3.7180***	0.0000
Public spending	–2.1623	1.0000		–4.6866***	0.0002
Electricity access	–0.3390	1.0000		–7.3125***	0.0000

Source: Prepared by the authors themselves based on Standardized World Income Inequality Database version 9.1 and PWT 10 data.

GDP: gross domestic product; HDI: Human Development Index.

Starting from the left, the first test we find in Table 7 is the Cumby–Huizinga autocorrelation test on three delays. In all three cases, we can see that the p-value is close to zero and therefore there is autocorrelation in these delays – that is, we assume that our model suffers autocorrelation. This is to be expected a priori, since the data on inequality and on GDP per capita have a lot of persistence and the value of 1 year is closely related to that of the previous period.

The following test is White’s heteroscedasticity test. The result of this test is clear – we have a p-value of 0.000 so we reject the hypothesis that the data distribution is homoscedastic. Therefore, our sample has problems of heterogeneity. This result was also to be expected after observing the big differences that exist between the countries of the African continent – as noted before, the richest country (Equatorial Guinea) has an per capita income 50 times higher than the poorest country (Central African Republic).

In addition, we have the result of the Breusch-Pagan Lagrange multiplier test. This test tells us that the individual effects for each country in the sample are relevant, so it is most appropriate to use panel data for the estimation.

Finally, in order to be confident that our estimates will be robust it is important to check the time series properties of the data. For this purpose, we have run the Im–Pesaran–Shin unit-root test on all the variables we use in this study. The results of the unit-root test are presented at the bottom of Table 7. Most of the variables are stationary at first difference, with the exception of the Gini index, which is initially stationary. We are then certain how to make robust estimates.

In conclusion, we know that our sample does not comply with the hypothesis of non-self-correlation and homoscedasticity. Therefore, we have to use an econometric technique that is efficient in spite of not complying with these basic assumptions. The method we will use will be the PMG estimation method, which makes efficient estimates in cases of heteroscedasticity and cross-sectional dependence.

Estimation technique

In this case we have a macro panel in which we have data for 45 countries over the period 1975–2019. After testing the characteristics of the data, we have come to several conclusions which lead us to believe that there may be cross-sectional dependence among the countries in the sample. Given the results of the tests described previously, we believe that this is a reasonable assumption. Authors such as Pesaran (2006) and Eberhardt and Teal (2011) showed that assuming cross-sectional independence could lead to biased results. However, following Kaulihowa and Adjasi (2018) we will address this problem through heterogeneous dynamic panel estimators, since African countries are perfect candidates to have cross-sectional dependence given their heterogeneity and historical precedents as seen in the section on income and inequality in Africa.

The estimation technique to be employed in this analysis will be chosen through the Hausman test of selection between PMG estimation and mean group (MG) estimation. These estimation techniques have been proposed following previous work such as Blackburne and Frank (2007), Kaulihowa and Adjasi (2018) Adams and Klobodu (2019) and Demissew Beyene and Kotosz (2020). These techniques have the ability to make consistent and efficient estimates despite using panel data with heterogeneity and cross-sectional dependence, allowing us to defer to the parameters in the short-run estimates, and to analyse their convergence in the long run.

The Hausman test has the null hypothesis that the differences between the coefficients are not systematic. If the p-value is 0.7430, the null hypothesis of homogeneity cannot be rejected, so the Hausman test indicates that we should use the PMG technique.

According to Pesaran et al. (1999), the estimation of the PMG estimator technique is expressed as follows in ARDL models (p, q, q, . . ., q):

y_{i t} = \sum_{j = 1}^{p} λ_{i j} y_{i, t - j} + \sum_{j = 0}^{q} δ'_{i j} X_{i, t - j} + μ_{i} + ε_{i t}

(1)

where the subscripts i and t denote the individual and the year of the panel. Thus, yt represents the vector of the dependent variable; Xt is the vector of the independent variables; µi captures the fixed effects; $λ_{i j}$ and $δ ’_{i j}$ are parameters to be estimated, and finally $ε_{i t}$ is the stochastic error term. From (1), the equation that defines the error correction model is expressed as follows:

Δ y_{i t} = ϕ_{i} (y_{i, t - 1} + \frac{β_{i}^{'}}{ϕ_{i}} X_{i t}) + \sum_{j = 1}^{p - 1} λ_{i j}^{*} Δ y_{i, t - j} + \sum_{j = 0}^{q} δ_{i j}^{*}' Δ X_{i, t - j} + μ_{i} + ε_{i t}

(2)

where,

ϕ_{i} = - (1 - \sum_{j = 1}^{p} λ_{i j})

β_{i} = \sum_{j = 0}^{q} δ_{i j}

λ_{i j}^{*} = - \sum_{m = j + 1}^{p} λ_{i m} (j = 1, 2, ... p - 1)

δ_{i j}^{*} = - \sum_{m = j + 1}^{q} δ_{i m} (j = 1, 2, ... q - 1)

The main advantage of equation (2) is that its construction allows it to capture both short-run and long-run dynamics. The parameter $ϕ_{i}$ captures the speed of adjustment; for us to establish that this is good, the estimated coefficient must be negative and also statistically significant. Similarly, the long-run relationship is given by $(y_{i, t - 1} + \frac{β_{i}^{'}}{ϕ_{i}} X_{i t})$ . In case there is no long-run relationship, the parameter $ϕ_{i}$ will be zero. Finally, we impose the homogeneity constraint on the long-run coefficient such that $- \frac{β_{i}^{'}}{ϕ_{i}}$ remains the same $\forall_{i}$ to obtain the following expression:

Δ y_{i t} = ϕ_{i} (y_{i, t - 1} + θ X_{i t}) + \sum_{j = 1}^{p - 1} λ_{i j}^{*} Δ y_{i, t - j} + \sum_{j = 0}^{q} δ_{i j}^{*}' Δ X_{i, t - j} + μ_{i} + ε_{i t}

(3)

where,

θ = θ_{i} = \frac{β_{i}^{'}}{ϕ_{i}} \forall_{i}

Model

The model that we are going to estimate follows the one originally proposed by Kuznets (1955). Therefore, we will estimate the country’s level of inequality according to its level of development. After this, we will be able to see whether the trend described by the coefficients follows an inverted U-shaped pattern or not. In this way, the model to be estimated will be the following:

{Gini}_{i, t} = α + β_{1} . {GDP pc}_{i, t} + β_{2} . {GDP pc}_{i, t}^{2} + μ_{i, t}

(4)

In this, we find the Gini variable that refers to the Gini Index and the GDP per capita variable that refers to the country’s GDP per capita. Finally, we find µ is the random disturbance of the model and we also find the subscripts i and t, which indicate the country and the year, respectively, to which the observation refers.

A sensitivity analysis will also be carried out to check whether the results are robust. This analysis will be done by repeating the estimated model, but with another development variable. The variable GDP pc will be replaced by HDI, thus seeing if the results replicate or change. The equation would then look like this:

{Gini}_{i, t} = α + β_{1} . {HDI}_{i, t} + β_{2} . {HDI}_{i, t}^{2} + μ_{i, t}

(5)

If Kuznets’ hypothesis is true, the results we will find should give us a positive value for β₁ and a negative value of β₂ and less than β₁ so that an inverted U-shape is drawn as a function of the progression of the development variables.

Finally, after checking whether the hypothesis of the Kuznets curve is observable, an extended model will be estimated which will try to better capture the behaviour of the inequality. The equation for the extended model is as follows:

\begin{array}{l} {Gini}_{i, t} = α + β_{1} . {GDP pc}_{i, t} + β_{2} . {GDP pc}^{2}_{i, t} + β_{3} . {PublicSpending}_{i, t} \\ + β_{4} . {ElectricityAccess}_{i, t} + β_{5} . {FrenchColony}_{i, t} \\ + β_{6} . {BritishColony}_{i, t} + μ_{i, t} \end{array}

(6)

As explained previously, these variables have been chosen following recent work by other authors on inequality. We expect that the variable of public spending and the variable of access to electricity will have negative coefficients, since we understand that they tend to equal with the opportunities of the population. Likewise, this expanded model, as it includes dummy variables that do not change over time, cannot be estimated through the PMG technique, since when differences are made, the observations cancel each other out. For this reason, the model will be the only one that will be estimated through the Prais–Winsten technique, which has been shown to be robust with results obtained through PMG and is an improved version of the Cochrane-Orcutt estimation, which has been shown to be robust in inequality studies as shown in Martínez-Navarro et al. (2020) due to its properties with data suffering from heteroscedasticity and AR(1).

Results

The estimated model gives a clear result: there is evidence that economic inequality in Africa describes a pattern similar to that described by Kuznets.

The coefficients are favourable to the hypothesis. First, we find a constant term of 36.87. This informs us that there is a high level of inequality in the economies studied, having a base of 36.87 points on the Gini index. Secondly, in the long-run estimation we have obtained a coefficient for GDP per capita of 0.0037 and for GDP per capita² a coefficient of −2.17 · 10⁻⁷, both significantly different from zero. Therefore, for a low level of GDP per capita, the level of inequality will grow until a point is reached at which the increases in GDP per capita cause a decrease in the level of inequality. This relation initially raised by Kuznets in 1955 was found by other authors a posteriori in other studies like Anand and Kanbur (1993), Jovanovic (2017) or Martínez-Navarro (2020). With respect to the parameters obtained in the short-term estimation, these were not found to be significant, so we cannot describe any behaviour in a reasoned way.

Of course, it would be interesting to know approximately at what income level the turning point occurs that leads economies to decrease their levels of inequality. This point can be calculated as the vertex of a parabola. Therefore, the turning point is calculated with the coefficients obtained for GDP per capita and GDP per capita² as follows: $\frac{- β_{1}}{2 . β_{2}}$ . From this expression we obtain the result US$8683.87 m in constant 2011 terms. This result is clearly an approximation, as not all the variables that influence inequality have been taken into account for its calculation. Nevertheless, it is useful to calculate the vertex of the parabola of the estimate because it gives us an approximation of the behaviour of the inequality.

Similarly, if we look at column 3 of Table 8, we find an estimate of the equation in which the variable GDP per capita is replaced by HDI. In this case, in the coefficients we again observe the coefficient for HDI of 104.36 and for HDI ² of −26.97. However, the coefficients obtained in this case for the development variables are not significantly different from zero, so the robustness test shows us that inequality in African countries is indeed high, and that the human development index variable does not capture changes in inequality as well as GDP per capita. This is easy to see because the lack of more explanatory variables has meant that most of the weight of the estimate has fallen on the constant term. Finally, it is worth noting that convergence is indeed found in the models since both have a negative and significant error correction term (ECT), but for the case of the estimation through GDP per capita, a particularly low speed of adjustment is found.

Table 8.

The role of development in inequality.

Regressors	Gini
Regressors	Estimates	t	Estimates	t
Long-run parameters
L.GDPPC	0.0037***	2.80
L.GDPPC 2	–2.17 · 10 – 7*	–0.61
L.HDI			104.36	0.54
L.HDI 2			–26.97	–0.21
Short-run parameters
D.GDPPC	–0.0007	–1.04
D.GDPPC 2	3.2017 · 10 – 7	1.34
D.HDI			28.48	0.99
D.HDI 2			–18.74	–0.85
Constant	36.87**	0.86	26.45***	2.69
ECT	–0.005*	–0.71	–0.401***	–5.35
Observations	956		956

Significant at 10%. **Significant at 5%. ***Significant at 1%.

ECT: error correction term; GDP: gross domestic product; HDI: Human Development Index.

We also think it is as interesting as checking whether Kuznets’ hypothesis is still valid, to know whether the hypothesis requires a minimum income to be observed. Following this line of thought, we have re-estimated equation (4) by dividing the sample in countries with incomes of more or less than US$4291.63 of GDP per capita because this amount is the average of the sample. After this, the same equation has been re-estimated again but differentiating the sample in this case in terms of whether they have incomes of more or less than US$8683.87 of GDP per capita, since this is the inflection point estimated above.

Table 9 shows the data for the estimate, differentiating between countries by whether they are above or below the average GDP per capita. In these, it is noteworthy that in both cases they are similar in terms of what was stated previously, in that the value of β₁ is positive and the value of β₂ is negative and lower than β₁. However, the estimates differ in that only the coefficients in the estimate for countries that are above the average GDP per capita are significant. Therefore, it seems reasonable to indicate that for the Kuznets curve to be observed in a meaningful way, countries are required to reach a minimum level of income, which is to be expected since, if income does not increase to a certain extent, the inflection point that would generate the inverted U shape described by Kuznets is not reached. A high level of adjustment speed is also found, at 87%.

Table 9.

The role of development in inequality below and above the average GDP.

Regressors	Gini
	GDP per capita < 4,291.63		GDP per capita > 4,291.63
	Estimates	t	Estimates	t
Long-run parameters
L.GDP_PC	0.01298	0.57	0.00471**	1.49
L.GDP_PC ²	–2.64 · 10 ^{− ⁶}	−0.70	–2.47 · 10 ^{− ⁷}*	–1.45
Short-run parameters
D.GDP_PC	0.2486	1.00	0.00200*	0.35
D.GDP_PC ²	–9.66 · 10 ^{− ⁶}	–0.20	–1.3 · 10 ^{− ⁶}*	0.36
Constant	32.53*	1.19	22.30**	1.50
ECT	–0.0053***	–18.04	–0.0418***	–13.26
Observations	480		476

Significant at 10%. **Significant at 5%. ***Significant at 1%.

ECT: error correction term; GDP: gross domestic product.

Thus, as mentioned previously, we re-estimate equation (4), this time differentiating the sample by whether they are above or below the tipping point. The results of these estimates are shown in Table 10. In this table, we can see that the relationship between β₁ and β₂ necessary to discuss Kuznets’ hypothesis in a meaningful way is indeed fulfilled in those countries that have not yet reached the turning point. Meanwhile, for those countries that have surpassed the threshold of US$8683.87, we can see the inverse relationship: it is apparent that β₁ has a negative coefficient and β₂ a positive one, thus describing a U-shaped relationship totally opposite to the one described by Kuznets, which for high levels of income would describe a decreasing curve. These regressions, if we observe them together, also indicate that the Kuznets curve is verified in the first instance, since from a minimum income to US$8683.87, an increase in inequality is observed that mitigates the inflection point, from which the equation gives us an inverse result and thus begins to decrease. However, the mathematical equation tells us that from a certain level onwards, a second turning point will be triggered, at which point inequality will go from decreasing to increasing again. This effect of inequality, which in the very long term takes the pattern of a sinusoid, can be taken as the continuation of Kuznets’ hypothesis, an approach initially proposed by Milanovic (2016). Finally, it is worth noting that the speed of adjustment among countries with an income level below the tipping point is very low indeed. However, when we look at the ECT of the countries that have passed the turning point, we see that it has increased, offering a speed of adjustment of 41%. This leads us to conclude that the speed of adjustment accelerates as the country’s GDP per capita increases.

Table 10.

The role of development in inequality before and after the turning point.

Regressors	Gini
	GDP per capita < US$8683.87		GDP per capita > US$8683.87
	Estimates	t	Estimates	t
Long-run parameters
L.GDP_PC	0.00019**	1.96	–0.0048*	–0.59
L.GDP_PC ²	–1.38· 10 ^{− ⁸}*	–2.19	2.51 · 10 ^{− ⁷}*	0.59
Short-run parameters
D.GDP_PC	0.00106*	1.08	–0.00074**	–0.74
D.GDP_PC ²	–5.8· 10 ^{− ⁸}*	–0.91	3.86 · 10 ^{− ⁷}*	0.73
Constant	44.35*	1.84	25.24	0.61
ECT	–0.061***	–11.29	–0.41***	–7.84
Observations	704		252

Significant at 10%. **Significant at 5%. ***Significant at 1%.

ECT: error correction term; GDP: gross domestic product.

In conclusion, the Kuznets curve can indeed be seen in Africa. However, we find that a minimum income is required to differentiate the countries by their level of income and a second turning point can be found for high levels of income, at which inequality increases again.

The results of the extended model are quite good (see Table 11) – all the coefficients are significant and we have a better fit than in previous models. Again, coefficients are shown that support Kuznets’ hypothesis, as we get a positive β₁ and a negative β₂ and lower in absolute value than the previous models, so we would be facing an inverted U-shape to be graphed. As for the public spending variable, as expected, it has a negative coefficient associated with it, which would imply that the larger the size of the public sector, the smaller the inequality experienced in the country, which is in line with Kuznets (1955) himself and other authors such as Lee (2005) and Dotti (2020) about the fact that the state tends to improve the social situation of a country. Electricity access shows a similar result: it is understood that a country where more of the population have access to electricity is a country where there are more equal opportunities than in countries where fewer people have access to this energy resource (Sarkodie and Adams, 2020). Finally, the dummy variables indicate that the French colonies have a negative coefficient, that is, lower levels of inequality, while on the contrary the British colonies have a positive coefficient, that is, higher levels of inequality. These coefficients are in line with what is shown in the earlier section on income and inequality in Africa, reaffirming that there is a verifiable legacy from the European powers in the distribution of income in Africa today, in this case, of greater inequality in the former British colonies and less inequality in the former French colonies.

Table 11.

The role of development in inequality – expanded model.

Regressors	Gini
Regressors	Estimates	t
GDP_PC	0.00013*	1.12
GDP_PC ²	–5.55 · 10^–9*	–1.18
Public spending	–0.005*	–0.21
Electricity access	–0.0143**	–2.34
French colony	–2.942***	–3.56
British colony	1.265*	1.48
Constant	46.22***	63.22
R²	0.9604

Significant at 10%. **Significant at 5%. ***Significant at 1%.

GDP: gross domestic product.

Conclusions and policy implications

This study has tested Kuznets’ hypothesis in 45 African countries from 1975 to the latest available data. Not many studies have focused on contrasting the non-linear relationship between inequality and the level of development in Africa, which is one of the novelties of this work. In order to deal with the large heterogeneity found among African countries as well as the unit-root processes, estimations have been made based on the PMG technique which takes into account non-stationary and heterogeneous panels.

Using this technique, we have been able to document the validity of the Kuznets hypothesis which was put forward more than half a century ago. Our results show that Kuznets’ patterns can be verified in African economies, which may still be in the upward phase of the inverted U-curve or very close to the turning point. We base this assertion on the high levels of inequality in African economies and the fact that many of them have incomes below US$8683.8 b in constant 2011 terms, which we estimate to be approximately the inflection point at which inequality will start to decline.

Furthermore, we have seen how Kuznets’ hypothesis correctly describes the relationship between inequality and development but requires a minimum level of income to demonstrate the turning point at which inequality decreases. In addition, for economies with relatively high levels of income, a second turning point seems to occur, which in the long term resembles a sinusoidal pattern rather than an inverted U, according to Milanovic’s (2016) contribution.

African economies have a long way to go to achieve sustainable welfare states, as their income levels are insufficient to reach the turning point of the Kuznets curve. However, the policy implications of this study suggest that the way forward is to develop their public sectors and invest in infrastructure that will allow electricity to reach every household, enabling both greater opportunities for the population and a boost to infant industry. According to Jayne et al. (2018), the economic transformation that has been underway in Africa since 2000 must be underpinned by improved governance, which should especially support rural activities, as it is these that will generate multiplier effects through complementary sectors (distribution and related industry). Moreover, these efforts by governments should focus especially on improving productivity and not on expanding the area devoted to agriculture. In this sense, as Haltiwanger and Spletzer (2020) point out, the development of industry will initially increase inequality, as Kuznets also argued, but it is a necessary stage in the process of achieving a modern economy in which the population enjoys a welfare state, and such development is strongly linked to access to electricity. This assertion is based on the fact that access to electricity will increase equality of opportunity for the population, positively impacting the economy as a whole as Marrero and Rodríguez (2019) argue, because a population with access to electricity will be able to develop their skills and engage in more specialised sectors. With all this, African countries will be able to increase their income levels and, according to Kuznets, this will lead to a fall in inequality in the future.

Furthermore, the author is aware that inequality must be explained by numerous variables, as in our models we observe low significance in some cases due to the omission of relevant variables. Nevertheless, it is interesting to note that in this field, it has been shown that the GDP per capita variable explains changes in inequality much better than the HDI variable. Therefore, it would be interesting to explore why the HDI does not capture inequality behaviour as powerfully as GDP per capita does.

And finally, it should be noted that the current situation of the African economies is a direct legacy of their journey with the European powers as argued by Amate-Fortes et al. (2015). These European powers colonized these lands in the past, taking advantage of their resources and preventing the development of national industry, among other things. In this work, it is recorded that the countries that suffer most from these effects are those that were colonised by the United Kingdom, which have higher levels of inequality than any other countries.

Footnotes

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Diego Martínez-Navarro

Author biographies

Diego Martínez-Navarro Graduate in Economics from the University of Almería. PhD in Economics from the University of Almería. He is a professor in the Department of Applied Economics at the University of Almeria. Author of several articles in the field of the Economics of Inequality.

Ignacio Amate-Fortes Degree in Economics from the Complutense University of Madrid. PhD in Economics from the University of Almeria. He is a lecturer in the Department of Applied Economics at the University of Almeria and Vice-Secretary of the Faculty of Economics and Business Studies at the University of Almeria. Author of numerous articles in the fields of Institutional Economics, International Economics, Public Economics, Development Economics and Economics of Inequality.

Almudena Guarnido-Rueda Graduate in Economics and Business Studies from C.E.U. “Luis Vives”. PhD in Economics from the University of Almeria. She is a lecturer in the Department of Applied Economics at the University of Almeria. Author of numerous articles in the fields of Institutional Economics, Regulation, Public Policy Evaluation, Applied Microeconomics and Economics of Inequality.

References

Adams

Klobodu

EKM

(2019) Urbanization, economic structure, political regime, and income inequality. Social Indicators Research 142(3): 971–995.

Ahmad

Muslija

Satrovic

(2021) Does economic prosperity lead to environmental sustainability in developing economies? Environmental Kuznets curve theory. Environmental Science and Pollution Research 28(18): 22588–22601.

Amate-Fortes

Guarnido-Rueda

Molina-Morales

(2015) On the perpetuation of the situation of economic and social underdevelopment in Africa. Journal of African American Studies 19(3): 233–263.

Anand

Kanbur

(1993) The Kuznets process and the inequality—Development relationship. Journal of Development Economics 40(1): 25–52.

Antonakakis

Collins

(2018) A suicidal Kuznets curve? Economics Letters 166: 90–93.

Aziz

Sharif

Raza

, et al. (2020) Revisiting the role of forestry, agriculture, and renewable energy in testing environment Kuznets curve in Pakistan: Evidence from Quantile ARDL approach. Environmental Science and Pollution Research 27(9): 1–14.

Barro

(2008) Inequality and Growth Revisited. Working Paper Series on Regional Economic Integration: 14. Asian Development Bank.

Baymul

Sen

(2019) Kuznets revisited: What do we know about the relationship between structural transformation and inequality? Asian Development Review 36(1): 136–167.

Baymul

Sen

(2020) Was Kuznets right? New evidence on the relationship between structural transformation and inequality. The Journal of Development Studies 56(9): 1643–1662.

10.

Blackburne

III Frank

(2007) Estimation of nonstationary heterogeneous panels. The Stata Journal 7(2): 197–208.

11.

Bukhari

Kalim

Arshed

, et al. (2021) Prevailing poverty in SAARC countries: Can education help? Asia-Pacific Social Science Review 21(1).

12.

Buonanno

Fergusson

Vargas

(2017) The crime Kuznets curve. Journal of Quantitative Criminology 33(4): 753–782.

13.

Caglayan-Akay

Kangalli-Uyar

(2021) Endogeneity and nonlinearity in the environmental Kuznets curve: A control function approach. Panoeconomicus 12.

14.

Cantarero-Prieto

Pascual-Saez

Gonzalez Diego

(2019) Examining an alcohol consumption Kuznets curve for developed countries. Applied Economics Letters 26(17): 1463–1466.

15.

Castelló-Climent (2010) Inequality and growth in advanced economies: an empirical investigation. Journal of Economic Inequality 8(3): 293–321.

16.

Chancel

Cogneau

Gethin

, et al. (2019) How large are African inequalities? Towards distributional national accounts in Africa, 1990-2017. WID.world Working Paper No. 2019/13, HAL.

17.

Cole

Rayner

Bates

(1997) The environmental Kuznets curve: An empirical analysis. Environment and Development Economics 2(4): 401–416.

18.

Dasgupta

Laplante

Wang

, et al. (2002) Confronting the environmental Kuznets curve. Journal of Economic Perspectives 16(1): 147–168.

19.

Deininger

Squire

(1996) A new data set measuring income inequality. The World Bank Economic Review 10: 26.

20.

Deininger

Squire

(1998) New ways of looking at old issues: inequality and growth. Journal of Development Economics 57: 259–287.

21.

Demissew Beyene

Kotosz

(2020) Testing the environmental Kuznets curve hypothesis: An empirical study for East African countries. International Journal of Environmental Studies 77(4): 636–654.

22.

Dogan

Inglesi-Lotz

(2020) The impact of economic structure to the environmental Kuznets curve (EKC) hypothesis: Evidence from European countries. Environmental Science and Pollution Research 27(11): 12717–12724.

23.

Dotti

(2020) Income inequality, size of government, and tax progressivity: A positive theory. European Economic Review 121 103327.

24.

Eberhardt

Teal

(2011) Econometrics for grumblers: A new look at the literature on cross-country growth empirics. Journal of Economic Surveys 25(1): 109–155.

25.

Feenstra

Inklaar

Timmer

(2015) The next generation of the Penn World Table. American Economic Review 105(10): 3150–3182.

26.

Fortin

Lemieux

(2016) Inequality and Changes in Task Prices: Within and between Occupation Effects. In Inequality: Causes and Consequences. Cheltenham: Emerald Group Publishing Limited. pp.195–226.

27.

Ghosh

Mitra

(2021) Tourism and inequality: A relook on the Kuznets curve. Tourism Management 83: 104255.

28.

Grecu

Rotthoff

(2015) Economic growth and obesity: Findings of an obesity Kuznets curve. Applied Economics Letters 22(7): 539–543.

29.

Greenwood

Jovanovic

(1990) Financial development, growth, and the distribution of income. Journal of political Economy 98(5, Part 1): 1076–1107.

30.

Hakkert

de Zeeuw

(1968) Population, development, and human natures: A comment. Environment and Development Economics 7(1): 174–177.

31.

Haltiwanger

Spletzer

(2020) Between firm changes in earnings inequality: The dominant role of industry effects. No. w26786. National Bureau of Economic Research.

32.

Higgins

Williamson

(1999) Explaining inequality the world round: Cohort size, Kuznets curves, and openness. National Bureau of Economic Research Working Paper 7224: 57. http://www.nber.org/papers/w7224 (accessed 9 December 2020).

33.

Ike

Usman

Sarkodie

(2020) Testing the role of oil production in the environmental Kuznets curve of oil producing countries: New insights from Method of Moments Quantile Regression. Science of the Total Environment 711: 135208.

34.

Jayne

Chamberlin

Benfica

(2018) Africa’s unfolding economic transformation. The Journal of Development Studies 54(5): 777–787.

35.

Jovanovic

(2017) When is there a Kuznets curve? Some evidence from the ex” socialist countries. Economic Systems 42(2): 248–268.

36.

Kacprzyk

Kuchta

(2020) Shining a new light on the environmental Kuznets curve for CO₂ emissions. Energy Economics 87: 104704.

37.

Kanbur

(2017) Structural transformation and income distribution: Kuznets and beyond. Working Paper No 266, African Development Bank, Abidjan, Côte d’Ivoire.

38.

Kaulihowa

Adjasi

(2018) FDI and income inequality in Africa. Oxford Development Studies 46(2): 250–265.

39.

Kuznets

(1955) Economic growth and income inequality. The American Economic Review 45(1): 1–28.

40.

Nguyen

, et al. (2020) The Kuznets curve for export diversification and income inequality: Evidence from a global sample. Economic Analysis and Policy 65: 21–39.

41.

Lee

(2005) Income inequality, democracy, and public sector size. American Sociological Review 70(1): 158–181.

42.

León

(2021) The Kuznets curve in Brazil, 1850-2010. Revista de Historia Economica-Journal of Iberian and Latin American Economic History 39(1): 37–61.

43.

Lindert

(1986) Unequal English wealth since 1670. Journal of Political Economy 94: 1127–1162.

44.

Marrero

Rodríguez

(2019) Inequality and growth: The cholesterol hypothesis Working Papers 501, ECINEQ, Society for the Study of Economic Inequality.

45.

Martínez-Navarro

Amate-Fortes

Guarnido-Rueda

(2020) Inequality and development: is the Kuznets curve in effect today? Economia Politica 37(3): 1–33.

46.

Meniago

Asongu

(2018) Revisiting the finance-inequality nexus in a panel of African countries. Research in International Business and Finance 46: 399–419.

47.

Milanovic

(2016) Introducing Kuznets waves: How income inequality waxes and wanes over the very long run. Voxeu, 24 February.

48.

Piketty

(2013) Capital in the Twenty-first Century. Cambridge, MA: The Belknap Press of Harvard University Press.

49.

Pesaran

(2006) Estimation and inference in large heterogeneous panels with a multifactor error structure. Econometrica 74(4): 967–1012.

50.

Pesaran

Shin

Smith

(1999) Pooled mean group estimation of dynamic heterogeneous panels. Journal of the American statistical Association 94(446): 621–634.

51.

Purcel

(2020) New insights into the environmental Kuznets curve hypothesis in developing and transition economies: A literature survey. Environmental Economics and Policy Studies 22(4): 585–631.

52.

Raza

Shah

(2017) Tourism growth and income inequality: Does Kuznets Curve hypothesis exist in top tourist arrival countries. Asia Pacific Journal of Tourism Research 22(8): 874–884.

53.

Sarkodie

(2018) The invisible hand and EKC hypothesis: What are the drivers of environmental degradation and pollution in Africa? Environmental Science and Pollution Research 25(22): 21993–22022.

54.

Sarkodie

Adams

(2020). Electricity access, human development index, governance and income inequality in Sub-Saharan Africa. Energy Reports 6: 455–466.

55.

Sarkodie

Ozturk

(2020) Investigating the environmental Kuznets curve hypothesis in Kenya: a multivariate analysis. Renewable and Sustainable Energy Reviews 117: 109481.

56.

Shahbaz

Loganathan

Tiwari

, et al. (2015) Financial development and income inequality: Is there any financial Kuznets curve in Iran? Social Indicators Research 124(2): 357–382.

57.

Solt

(2020) Measuring income inequality across countries and over time: The standardized world income inequality database. Social Science Quarterly 101(3): 1183–1199.

58.

Tchamyou

(2021) Financial access, governance and the persistence of inequality in Africa: Mechanisms and policy instruments. Journal of Public Affairs 21(2): e2201.

59.

Tenaw

Beyene

(2021) Environmental sustainability and economic development in sub-Saharan Africa: A modified EKC hypothesis. Renewable and Sustainable Energy Reviews 143: 110897.

60.

The Heritage Foundation (2001) United States. [Web Archive] Retrieved from the Library of Congress. https://www.loc.gov/item/lcwaN0002695/

61.

The World Bank, World Development Indicators (2012) GNI per capita, Atlas method [Data file]. http://data.worldbank.org/indicator/NY.GNP.PCAP.CD

62.

Theyson

Heller

(2015) Development and income inequality: A new specification of the Kuznets hypothesis. The Journal of Developing Areas 49(3): 103–118.

63.

Thomas

(2015) Income inequality and economic development in Latin America: A test for the Kuznets inverted-u curve. Indian Journal of Economics and Business 14(1): 143–149.

64.

Tsurumi

Managi

(2010) Decomposition of the environmental Kuznets curve: Scale, technique, and composition effects. Environmental Economics and Policy Studies 11(1–4): 19–36.

65.

United Nations Development Programme (2019) Beyond income, beyond averages, beyond today: Inequalities in human development in the 21st century. Human Development Report 2019. Human Development Report Office, pp.22–25.

66.

Usman

Olanipekun

Iorember

, et al. (2020) Modelling environmental degradation in South Africa: The effects of energy consumption, democracy, and globalization using innovation accounting tests. Environmental Science and Pollution Research 27(8): 8334–8349.

67.

Wittenberg

(2015) Problems with SWIID: The case of South Africa. The Journal of Economic Inequality 13(4): 673–677.