Examining the link between tax revenue mobilization efforts and capital flight in African countries

ABSTRACT

This article examines the impact of tax revenue mobilization on capital flight in 30 African countries, focusing on the role of natural resources. Over the period 1998–2018, econometric analysis based on dynamic generalized method of moments suggests that tax revenue mobilization reduces capital flight in Africa. However, when countries have more natural or oil resources, the negative impact of tax revenues on capital flight weakens. Therefore, despite the importance of the benefits associated with natural resource wealth (in particular, oil wealth), the latter compromises the impact of tax revenue mobilization on stemming capital flight. Finally, greater responsibility in the management of natural resources and more transparent reporting by companies operating in this sector are needed.

KEYWORDS:

• Capital flight; residual method
• natural resources
• tax revenues
• GMM

JEL Classification:

• C23
• E6
• F32

1. Introduction

Public resource mobilization is an important and indispensable strategy for the realization of public services, investment policies, poverty reduction, and economic growth (Culpeper and Aniket 2010; Ezu and Oranefo 2023). However, the acceleration of public revenue mobilization occurs concurrently with a surge in capital flight (Ndikumana, Boyce, and Ndiaye 2014; Orji et al. 2020). Despite the commendable tax collection performance of African economies, a shortage of sufficient savings impedes their ability to finance investments independently, compelling a reliance on diminishing official development assistance in the face of recurring crises.

The literature points out that the discovery of natural resources within an economy attracts the attention of various interest groups, both within the national economy and beyond (Tornell and Lane 1998). Consequently, empirical evidence indicates that numerous resource-rich African nations grapple with elevated levels of poverty, lagging behind their resource-poor counterparts in the pursuit of developmental objectives such as access to critical social services such as health, education, water, and sanitation (Sachs and Warner 1997; Stijns 2005).

This situation raises profound concerns regarding the anticipated benefits derived from natural resources. Historically, in the 1960s and the 1970s, resource-rich African nations exhibited considerably higher per capita incomes than their resource-poor counterparts did. However, over the subsequent decades, many resource-rich countries have witnessed underperformance, with the expected benefits of dissipating natural resource wealth. In Africa, there is negligible disparity in income levels between resource-rich and resource-poor nations.

A vast body of economic literature suggests that the economic and social development of resource-rich countries does not surpass that of countries that are less endowed with natural resources (Ross 2015). In contrast, natural resource wealth is associated with heightened corruption (Akinlo 2023), diminished democratic governance (Andersen and Aslaksen 2013), and a higher likelihood of violent conflict (Collier and Hoeffler 1998). In numerous African nations, the natural resources sector serves as the primary conduit for exports, enabling dictators, corrupt leaders, and unscrupulous entities to exploit tax evasion, transfer pricing, and concealed corporate ownership for profit maximization, whereas millions of Africans suffer deprivation in terms of nutrition, health, and education.

In recent decades, a notable trend of substantial private capital outflows from 30 African countries to Western financial centers has been observed, as illustrated in Figure 1 in the appendix. Ndikumana and Boyce (2018) documented that between 1970 and 2018, capital flows from developing countries reached a staggering $2,010.8 billion, equivalent to 93.8% of GDP. This phenomenon is accentuated in countries with abundant natural resources, wherein a significant proportion of total exports and government revenues stems from these resources, thereby elevating susceptibility to capital flight due to corruption associated with concentrated power. Such circumstances not only reflect weak institutions at large but also signify inadequate governance within the natural resources sector. Exemplifying this trend is the case in Nigeria, where the discovery of substantial oil reserves in 1965 paradoxically resulted in the deterioration of economic conditions. GDP per capita increased from $1,113 in 1965 to $1,084 in 2000.

Figure 1. Real capital flight of the 30 African countries over the period 1998–2018 according to the World Bank residual method. (% of GDP).

Source: Author based on data from Ndikumana and Boyce (2018)

The ratio of average real capital flight to gross domestic product (GDP) in Africa stood at 5.6% between 1998 and 2018. Notably, resource-poor countries exhibit a higher ratio (7.3%) than their resource-rich counterparts (4.4%). However, despite the lower ratio, resource-rich countries witnessed a higher volume of real capital flight ($1,063.16 billion) compared to resource-poor countries ($82.64 billion). Among resource-rich countries, the rate of non-resource tax revenue in non-oil-rich countries stood at 17.6%, in contrast to 10.3% in oil-rich countries, between 1998 and 2018. This discrepancy can be attributed to the lower risk of capital flight in countries with diversified economies and lower dependency on natural resources.

The focal point of concern in many African nations lies within the natural resources sector, which necessitates governments to maximize revenue collection from substantial profits while minimizing capital outflows. The diversification of economies has become imperative to mitigate the over-reliance on natural resources for economic growth. Effective measures to prevent capital flight from the resource sector contribute not only to the development of a robust national financial system but also offer prospects for financing non-resource sectors, yielding positive implications for public revenues at large.

An overarching argument supporting tax revenue mobilization lies in the susceptibility of African economies to global economic crises, which may diminish official development assistance flows, potentially prompting citizens to seek safer havens for their wealth abroad. However, empirical studies indicate that developing countries are more vulnerable to global financial shocks resulting from financial liberalization, leading to heightened capital flight (Pradan and Hiremath 2017). Addressing this research gap, this study seeks to provide novel insights by investigating the impact of tax revenue mobilization on capital flight in selected African countries, with a specific focus on the role of natural resources. In addition, based on the ensuing discussion, two empirical questions arise. Firstly, what is the impact of tax revenue mobilization on capital flight? In other words, what explains capital flight, and to what extent does tax revenue mobilization affect capital flight? Secondly, does natural resource endowment mitigate the impact of tax revenue mobilization on capital flight? Given that resource-rich countries tend to be both zones of armed conflict and political instability, and therefore important sources of capital flight. To achieve the main objective of the study, two hypotheses are proposed: improved tax revenue mobilization significantly determines capital flight in African countries, while the availability of natural resources weakens the negative impact of tax revenue mobilization on capital flight in African countries.

In light of the above, it is somewhat surprising that existing studies on tax revenue mobilization have not yet explored its impact on capital flight through the role of natural resources. This, then, is the primary motivation for this study, and three main contributions flow from it. First, the study adds to the existing literature on the determinants of capital flight, as it complements previous studies, such as Salisu and Isah (2021) and Akinlo, (2023) by incorporating tax revenue mobilization in the presence of natural resources and expanding the data set to a more recent one provided by Ndikumana and Boyce (2018). To our knowledge, this is the first study along these lines. Second, the study extends the trajectory of existing studies and uses a recent econometric methodology on thirty (30) African countries, thus providing a guiding framework for policymakers in the debate and formulation of policies regarding tax revenue mobilization in Africa. Third, we provide empirical evidence of the policy disparity in terms of combating capital flight between resource-rich and resource-poor countries in Africa.

In light of the above, it is somewhat surprising that existing studies on tax revenue mobilization have not yet explored its impact on capital flight through the role of natural resources. This is the primary motivation for this study, and the three main contributions come from it. First, the study adds to the existing literature on the determinants of capital flight, as it complements previous studies, such as Salisu and Isah (2021) and Akinlo, (2023) by incorporating tax revenue mobilization in the presence of natural resources and expanding the data set to a more recent one provided by Ndikumana and Boyce (2018). To the best of our knowledge, this is the first study to be conducted along these lines. Second, this study extends the trajectory of existing studies and uses a recent econometric methodology on thirty (30) African countries, thus providing a guiding framework for policymakers in the debate and formulation of policies regarding tax revenue mobilization in Africa. Third, we provide empirical evidence of policy disparity in terms of combating capital flight between resource-rich and resource-poor countries in Africa.

In this context, the impact of tax revenue mobilization on capital flight over the period 1998–2018 is analyzed using a panel extension of Arellano and Bover (1995) and the Blundell and Bond (1998) method by Roodman (2009a, 2009b) in 30 African countries. An estimation technique can solve econometric problems such as endogeneity, simultaneity, autocorrelation, and heterogeneity in our data.

The total natural resource rents and four rent-related natural resource components were used in a disaggregated manner (coal, natural gas, oil, and forest resources) to facilitate the identification of the channel through which tax revenues influence capital flight. Our results show that non-resource tax revenues are negatively related to capital flight and that tax revenue mobilization efforts are therefore associated with a reduction in capital flight. Similarly, we find a negative relationship between total rents from natural resources and capital flight. However, when total resources or oil rents are non-zero, the impact of non-resource tax revenues on capital flight weakens for countries with more resources or oil rents.

The rest of the paper is structured accordingly. Section 2 presents a review of the relevant literature on tax revenues, natural resources, and capital flight. Section 3 presents the methodology and data used. Next, we present the analysis and discussion of the empirical results in section 4. Section 5 presents the conclusion and policy implications arising from the analysis.

2. Empirical literature on capital flight, tax revenues, and natural resources

Capital flight is determined by a set of variables. Since the pioneering works of Cuddington (1986), Dooley (1988), and Pastor (1990), various studies have identified the structural factors that determine capital flight (Raheem 2015). These empirical studies have explored the factors that are likely to affect capital flight, both within a developing country and across country groups.

Investment diversion theory states that two factors cause capital flight, namely macroeconomic and political uncertainties in developing countries and the existence of better investment opportunities and a stable political and economic environment in developed countries (Ajayi 1995). Better investment opportunities are due to high returns on foreign investments, a variety of financial instruments in which to invest, political and economic stability, a favorable tax system (i.e. lower taxes or tax exemption), and the hiding of accounts in tax-haven countries. The favorable tax climate argument seems to suggest that countries would be better off in terms of capital flows if they had oriented their tax policy towards a system of lower taxes or even tax exemption.

However, Muchai and Muchai (2016) warned that reducing taxes and offering tax incentives to attract or retain capital cause market distortions and tax favoritism, which in turn leads to further capital losses. What is more aggravating is that on the eve of an increase in tax breaks, international investors repatriate their funds to regions with favorable tax regimes. In effect, they avoid paying high taxes and, thus, contribute to a country’s capital flight. For example, Schineller (1997) finds that low taxes contribute to low capital flight for a sample of 18 developing countries over the period 1978–1988, that low taxes contribute to low capital flight. Loungani and Mauro (2000) find that high capital flight from Russia, Central Europe, the Baltic States, and Latin America is significantly associated with high taxation. Hermes and Lensink (2002) also find a positive relationship between capital flight and uncertainty about government tax policy for a sample of 84 developing countries. According to Onwioduokit (2001), most residents anticipate higher taxes and shift their investment abroad. In the event of a large budget deficit, domestic investors have an incentive to move their capital abroad to avoid the risk of future taxation. This is because a large budget deficit would increase future taxes. This is complemented by the work of several researchers (Loungani and Mauro 2000; Schineller 1997) who found that budget deficit uncertainty has a positive effect on capital flight.

Furthermore, the relationship between natural resources and capital flight has received particular attention in recent years. First, corruption in natural resource management is a developmental problem rather than an isolated one (Kolstad and Soreide 2009). High revenues from natural resources lead to rent-seeking and embezzlement by corrupt political and administrative elites and decision-makers, who typically deposit embezzled funds in foreign accounts. Second, the natural resources sector is generally managed by multinationals and local companies, whose technological and financial processes are highly complex. As a result, they tend not to declare actual sales figures to authorities and take advantage of opportunities for false invoicing (Ndikumana, Boyce, and Ndiaye 2014). At the same time, they are bribing decision-makers to award lucrative contracts and practice tax evasion (Mpenya, Metseyem, and Epo 2016). The misinvoicing of oil, gas, and mineral exports is an important channel for the relationship between capital flight and natural resources. For example, using a sample of 30 African countries over the period 1970–2015, Ndikumana and Sarr (2019) examined the triangular relationship between foreign direct investment, capital flight, and natural resource rents.

Third, the complexity of the technological and financial processes involved in exploiting natural resources creates an imbalance in terms of expertise and technical capacity between governments of resource-rich developing countries and multinational companies. This creates opportunities for under-invoicing, smuggling, and other forms of unrecorded outflows of resources from countries. Econometric evidence of the link between capital flight and natural resources is mixed. While studies assessing capital flight from developing countries show that resource-rich countries top the list of countries with the highest capital flight (Ndikumana and Boyce 2010; Ndikumana and Sarr 2019), robust econometric evidence of the direct impact of natural resources on capital flight is relatively scarce.

3. Econometric methodology and model specification

The estimation of capital flight, the methodology to be adopted, the data, and the econometric model used are presented in the following three subsections respectively.

3.1. Estimating capital flight

Despite the substantial literature in recent years, there is no common measure of capital flight (Hermes, Lensink, and Murinde 2002; Ndikumana, Boyce, and Ndiaye 2014). The existing literature has proposed four main methods (the residual method; the Dooley method¹; the trade misinvoicing method; and the hot money method) for estimating capital flight. The residual method remains the most widely used measurement method in the capital flight literature because it relies on commonly available data and is simple to implement (Ndikumana and Boyce 2010). Therefore, indirect methods are used to calculate capital flight (Schneider 2003): (1) $\mathrm{FK}{R}_{\mathrm{rit}}=\mathrm{CDEBTADJ}+\mathrm{FDI}+\mathrm{PI}+\mathrm{OI}\text{–}\left(\mathrm{CC}+\mathrm{RES}\right)+\mathrm{MISINV}$(1) where CDEBTADJ is the change in outstanding external debt adjusted for exchange rate fluctuations, debt forgiveness, and the change in interest arrears; FDI is net foreign direct investment, PI is portfolio investment, OI is other investment, CC is the current account deficit, RES is net additions to foreign reserves, and MISINV is net trade misinvoicing.

In addition, the Morgan Guaranty (1986) measure can be written as follows: (2) $M{G}_{\mathrm{it}}=\mathrm{CDEBTADJ}+\mathrm{FDI}+\mathrm{PI}+\mathrm{OI}\text{–}\left(\mathrm{CC}+\mathrm{RES}\right)+\mathrm{MISINV}-\mathrm{AB}{D}_{\mathrm{it}}$(2) Where MG equals the private claim measure for capital flight (Morgan Guaranty's 1986 measure); ABD equals the foreign short-term assets of the banking system.

3.2. Model specification for empirical estimation

Based on the above statement, we specify a mathematical model as follows: (3) $\begin{array}{rl}\mathrm{FK}{R}_{\mathrm{it}-1}=& f(\mathrm{FK}{R}_{\mathrm{it}-1},\mathrm{TN}{R}_{\mathrm{it},}\mathrm{REN}{T}_{\mathrm{it},}\mathrm{TN}{R}_{\mathrm{it}}\ast \mathrm{REN}{T}_{\mathrm{it}},\mathrm{\Delta GD}{P}_{\mathrm{it}},\mathrm{IN}{F}_{\mathrm{it}},\\ & \mathrm{DEB}{T}_{\mathrm{it},}\mathrm{OPE}{N}_{\mathrm{it}},C{C}_{\mathrm{it}})\end{array}$(3) Where i = 1,2 … 30 is the individual (country) dimension and t = 1998 … 2018, the time dimension. The main objective of this study is to explore the impact of tax revenue mobilization on capital flight through the natural resource channel. The empirical model used draws on the work of authors such as Ndikumana and Boyce (2010).

The dependent variable FKR denotes the volume of capital flight as a percentage of GDP where α is a constant, it is explained by a set of variables, FKR_t–1 represents past capital flight, TNR represents total non-resource tax revenue as a percentage of GDP (our variable of interest) and natural resources are measured by the rent they generate. Thus, RENT is total natural resource rents as a percentage of GDP. We used control variables such as ΔGDP which represents the real GDP growth rate for country i at period t, INF represents the inflation rate measured by the annual change in the consumer price index in country i at period t, DEBT represents the external debt in country i at period t. OPEN is the degree of openness of country i at period t, and CC is the institutional variable of the quality of institutions measured by the control of corruption, which takes values ranging from –2.5 (lowest corruption) to 2.5 (highest corruption) in country i at period t. Among other things, we have decomposed² total resource rents into oil (OIL), gas (GAS), coal (COAL), and forest (FOREST) rents, as different types of natural resources can have a different impact on capital flight (Ndikumana and Boyce 2010; Ndikumana and Sarr 2019). The control variables are included to avoid omitted variable bias.³

The equation to be estimated takes the form of a simple linear regression as follows: (4) $\begin{array}{rl}\mathrm{FK}{R}_{\mathrm{it}}& ={\alpha }_0+{\delta }_1\mathrm{FK}{R}_{\mathrm{it}-1}+{\delta }_2\mathrm{TN}{R}_{\mathrm{it}}\\ & +{\delta }_3\mathrm{REN}{T}_{\mathrm{it}}+{\delta }_4\mathrm{TN}{R}_t\ast \mathrm{REN}{T}_{\mathrm{it}}\\ & +{\delta }_5\mathrm{\Delta GD}{P}_{\mathrm{it}}+{\delta }_6\mathrm{IN}{F}_{\mathrm{it}}+{\delta }_7\mathrm{DEB}{T}_{\mathrm{it}}\\ & +{\delta }_8\mathrm{OPE}{N}_{\mathrm{it}}+{\delta }_{9}C{C}_{\mathrm{it}+}{\vartheta }_i+{ϵ}_{\mathrm{it}}\end{array}$(4) where ϑ is an individual time-invariant fixed effect and ϵ is a set of unobservable factors.

One of the objectives of this study is to test whether tax revenue mobilization can impact capital flight through natural resources. We therefore include the interaction of tax revenues with natural resource rents, as shown in equation (4). Then, differentiating equation (4) concerning tax revenue (TNR_it), we obtain the following: (5) ${\displaystyle \frac{\left(\mathrm{FK}{R}_{\mathrm{it}}\right)}{\left(\mathrm{TN}{R}_{\mathrm{it}}\right)}={\delta }_2+{\delta }_4\mathrm{REN}{T}_{\mathrm{it}}}$(5) δ₂ and δ₄ capture the extent to which the natural resource in tax revenue mobilization countries enhances the impact on capital flight. The introduction of the interaction term means that the impact of tax revenue mobilization on capital flight should be treated as a marginal effect in such a specification (Asongu and Nwachukwu 2016).

3.3. Econometric method

The empirical literature proposes several approaches for estimating a dynamic panel data model with assumed endogeneity problems. To efficiently estimate the dynamic model formulated above, we use the Generalized Method of Moments (GMM) estimation approach (Arellano and Bond, 1995). To overcome the problems of endogeneity, simultaneity, autocorrelation, and heterogeneity of our data. We adopted the endogeneity-resistant GMM estimator, which is an extension of the method of Arellano and Bover (1995) and Blundell and Bond (1998) and is available as xtabond2 in Stata. Blundell and Bond (1998) have shown that the GMM estimator produces dramatic efficiency gains over GMM in the first base difference (Baltagi 2013). Some of the advantages of the GMM estimation approach over other methods and its suitability for our sample are briefly explained here. Firstly, the method is suitable for dynamic or persistent panels. Secondly, GMM eliminates bias due to endogeneity (or reverse causality) by controlling for simultaneity (using an instrumentation process) and unobserved heterogeneity (using time-invariant omitted variables). This objective is partially achieved through the use of lagged explanatory variables as internal instruments. The estimation technique also allows the inclusion of external instruments. Thirdly, the technique is adapted to the ‘small T, large N’ context, taking into account Nickell's (1981) bias and applying the ‘Windmeifer finite sample correction’ (Windmeijer 2005). Fourth, the approach eliminates country-fixed effects by differentiating internal instruments to make them exogenous to fixed effects (Akobeng 2016, 215), but does not eliminate country differences. It controls dependence between countries, and limits instrument proliferation and over-identification (Baltagi 2013; Love and Zicchino 2006; Roodman 2009b). The estimator allows the researcher to control for time-invariant country-specific effects and the endogeneity of foreign aid (Alvi and Senbeta 2012, 955). Fifth, the two-stage GMM approach was adopted in our specification because of its ability to control for heteroscedasticity, instead of the one-stage approach that is compatible with homoscedasticity. We also adopted orthogonal forward deviations, instead of differentiation, to minimize data loss (Roodman 2009b).

4. Data sources and variable definitions

4.1. Data sources

The analysis based on balanced panel methods spans 20 years (1998–2018) and covers 30 African countries. The scope is determined by data availability, in particular the institutional variable of corruption control. Data on capital flight (real, in millions, constant value in 2018 dollars) were extracted from the Capital Flight Series calculated by Ndikumana and Boyce (2018) of the Political Economy Research Institute, University of Massachusetts, Amherst, and available free of charge at https://www.peri.umass.edu/capital-fightfrom-africa. Data to be known on total natural resource rents (oil rents, natural gas rents, coal rents, forestry rents) as well as inflation rate, and GDP growth were extracted from the World Bank's World Development Indicators (WDI 2021). The variable of interest, non-resource tax revenues as a percentage of GDP (NRT), was taken from the International Centre for Tax and Development's (ICTD) UNU-WIDER Government Revenue Dataset 2020. Corruption monitoring is also taken from the World Bank's Worldwide Governance Indicators (WGI). The STATA 15 statistical analysis tool was used for the estimation procedures. Table A1 lists the countries used, and Table A2 gives a full description of the data set (see Appendix).

4.2. Descriptive analysis

Descriptive statistics for the 30 African countries over the sample period are presented in Table 1. The average ratio of non-resource tax revenues to GDP is around 14.16% in countries over the 1998–2018 period. This average share is below the minimum of 25% recommended by the World Bank and the International Monetary Fund (IMF). This suggests that tax revenue mobilization performance in Africa is still very weak. In 2018, the country with the lowest tax mobilization effort was Sudan at 7.407%, while Seychelles was the country with the highest effort at 37.67%.

Table 1. Summary statistics of variables.

Variables	Obs	Mean	Std. dev.	Min.	Max.
FKR	623	5.70	16.171	−53.259	180.992
MG	624	−1.4199	38.900	−144.87	557.41
TNR	622	14.169	7.284	0.573	38.657
RENT	630	12.367	11.147	0.066	58.650
ΔGDP	630	4.488	4.292	−20.599	26.417
INF	630	11.641	33.269	−8.238	513.907
OPEN	630	63.774	36.587	16.141	225.023
DEBT	609	22.639	1.191	19.715	25.900
COAL	630	0.178	0.702	0	8.047
FOREST	630	5.449	6.372	0	40.409
GAS	628	0.307	0.779	0	4.399
OIL	630	5.321	10.957	0	55.458
CC	570	−0.619	0.5792	−1.723	1.217

Source: Authors’ calculations (2022).

However, over the whole period, the minimum non-resource tax revenue is 0.573% for the Democratic Republic of Congo (a resource-rich country) and the maximum is 38.657% for the Seychelles. On average, capital flight is 5.701% of GDP. In real terms, this represents 1145.8 billion US dollars. Its maximum value is 180.992% of the GDP for Sierra Leone, while its minimum value is a net capital inflow of around 53.259% of the GDP for Malawi. Rents from natural resources represent an average of 12.36% of GDP. Their minimum share is that of the Seychelles (0.065%) and their maximum share is that of the Republic of Congo (58.65% of GDP).

4.3. Correlation analysis

The correlation matrix in Table 2 (appendix) presents revealing results on the relationship between tax revenue (the dependent variable) and each of the explanatory variables, as well as between the explanatory variables to mitigate potential problems of multicollinearity. We find that capital flight is positively correlated with all the variables in Table 2 (but significantly so with tax revenue, economic growth rate, and degree of openness). On the other hand, there is a possibility of multicollinearity between natural resource rents and oil rents, which is because the measures give very similar estimates.

Table 2. Correlation matrix.

	FKR	MG	TNR	RENT	ΔGDP	INF	OPEN	DEBT	COAL	FOREST	GAS	OIL	CC
FKR	1.00
MG	0.5459*	1.00
TNR	0.048	−0.1403*	1.00
RENT	0.087*	0.0652	−0.445*	1.00
ΔGDP	0.107*	0.0262	−0.111*	0.012	1.00
INF	−0.042	−0.0157	−0.219*	0.143*	−0.103*	1.00
OPEN	0.191*	−0.0277	0.445*	0.138*	−0.116*	−0.067	1.00
DEBT	−0.104*	−0.0551	0.187*	−0.082*	−0.046	0.039	0.087*	1.00
COAL	−0.002	0.0293	0.313*	−0.092*	−0.154*	−0.047	0.074	0.255*	1.00
FOREST	0.063	0.2056*	−0.304*	0.296*	0.051	0.102*	−0.353*	−0.396*	−0.088*	1.00
GAS	−0.008	−0.1783*	−0.015	0.200*	−0.01	−0.055	0.084*	0.214*	0.079*	−0.187*	1.00
OIL	0.069	−0.0394	−0.304*	0.810*	−0.017	0.103*	0.303*	0.137*	−0.122*	−0.260*	0.273*	1.00
CC	0.008	−0.2126*	0.628*	−0.455*	0.033	−0.153*	0.267*	−0.114*	0.131*	−0.255*	−0.074	−0.323*	1.00

Note: *** Significance at 1% level, ** Significance at 5% level, * Significance at 10% level.

5. Results and discussion

5.1. Main result

In light of the aforementioned criteria for the GMM estimator, the majority of the models used in this study are valid because they meet three conditions. The Arellano and Bond difference autocorrelation test (AR 2), which assumes that the model does not suffer from autocorrelation in the residuals, is not rejected. Second, the over-identifying restriction (OIR) tests based on Sargan and Hansen with null hypotheses of instrument validity or absence of correlation with error are also not rejected. Finally, the homogeneity test for instruments anchored on Hansen's difference test (DHT) was used to examine the validity of Hansen's test results. From this point of view, it is fundamental to mention that to ensure model persistence, the lagged values of the dependent variable must be significant and satisfy the convergence criterion. The convergence criterion is based on the argument that the absolute value of the lagged estimated capital flight indicators must lie in the interval between zero and one. Detailed intuitions describing this criterion are well-advanced in the existing literature using the GMM estimator (Asongu 2013).

Table 3 presents the impact of tax revenues on capital flight using the World Bank methodology while examining the role of natural resources. Our estimation results indicate that an increase in non-resource tax revenue is associated with a decrease in capital flight. The regression results suggest that a 1% increase in non-resource tax revenues is associated with a 0.4022% to 3.378% decrease in capital flight in the same year. Tax revenues hurt capital flight, even though non-resource tax revenues are low in Africa, and the effect is not statistically significant in the forest and natural gas models. In addition, the total rent from natural resources or oil resources is associated with a reduction in capital flight. The impact of the ratio of total natural resource rent to GDP is negative and statistically significant at the 5% level in all the regressions.

Contrary to intuition, a proportional increase in the disaggregated components of total natural resource rents, such as oil, forest, and natural gas, reduces capital flight by 1.633%, 0.636%, and 16.98%, respectively. This result does not corroborate the findings of some previous work (Kwaramba, Mahonye, and Mandishara 2016; Ndikumana and Sarr 2019). On the other hand, it is also interesting to note that the regressions that include coal-related rent are statistically significant and positive. However, the lagged value of the dependent variable is, on the contrary, positively related to capital flight.

Overall, the main results in Table 3 reveal that when total resource/oil rents are non-zero, the impact of non-resource tax revenues on capital flight weakens (becomes less negative and sometimes positive) for countries with more resources or oil rents. It should be noted that the positive interaction term is significant only for oil resources. This means that as African countries’ non-resource tax revenues increase, capital flight intensifies as countries exploit or dispose of more natural resources. The result for total resource rents is similar to the results for oil, gas, and forestry rents. In the coal rent regressions, the results are counterintuitive: the coefficient of the interaction term is negative and significant. Furthermore, these results suggest that the instantaneous impact of resource rents on capital flight depends on the type of natural resources exploited.

Table 3. GMM estimate of the tax revenue mobilization system on capital flight (FKR) using the World Bank method.

Variable	Dependent variable: FKR
	Model 1 RENT	Model 2 OIL	Model 3 FOREST	Model 4 GAS	Model 5 COAL
L.FKR	0.245***	0.192***	0.283***	0.209***	0.229***
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
TNR	−3.378*	−4.023***	−0.869	−1.217	3.044***
	(0.032)	(0.007)	(0.609)	(0.482)	(0.001)
RENT	−0.994**
	(0.018)
OIL		−1.634***
		(0.001)
FOREST			0.636
			(0.380)
GAS				−16.984**
				(0.016)
COAL					28.489***
					(0.000)
TNR*RENT	0.065	0.136***	0.023	1.129***	−0.801***
	(0.127)	(0.003)	(0.647)	(0.000)	(0.000)
ΔGDP	1.091***	0.860**	0.772*	1.449***	0.849*
	(0.002)	(0.019)	(0.021)	(0.000)	(0.038)
INF	−0.075***	−0.033*	−0.044	−0.045**	−0.09***
	(0.001)	(0.052)	(0.125)	(0.004)	(0.000)
OPEN	−0.174**	−0.048	−0.385***	−0.262***	−0.304***
	(0.016)	(0.547)	(0.000)	(0.000)	(0.007)
DEBT	−2.486	0.097	−0.002	0.594	−5.363***
	(0.131)	(0.956)	(0.999)	(0.773)	(0.000)
CC	−1.968	0.723	0.630	3.299*	−8.404***
	(0.583)	(0.700)	(0.763)	(0.085)	(0.002)
_cons	98.033***	39.793	27.300	17.916	107.904***
	(0.002)	(0.321)	(0.388)	(0.681)	(0.002)
AR(1)	0.000	0.002	0.001	0.001	0.002
AR(2)	0.823	0.999	0.867	0.994	0.966
Sargan OIR	0.042	0.017	0.143	0.549	0.309
Hansen OIR	0.435	0.222	0.294	0.407	0.290
Dif(null, H = exogenous)	0.706	0.652	0.582	0.293	0.721
1. (b) IV (years, eq(dif))	0.338	0.375	0.315	0.648	0.466
Dif(null, H = exogenous)	0.559	0.157	0.328	0.167	0.171
Fisher	3871.5***	320.58	595.70	337.65	1999.53***
Instruments	29	29	29	29	29
Country	29	29	29	29	29
Observation	492	492	492	490	492

Note: Numbers in parentheses () are standard errors (SEs). *** Significance at 1% level, ** Significance at 5% level, * Significance at 10% level. (a) estimated parameters and F-statistics and (b) failure to reject the null hypotheses of: (i) the absence of autocorrelation in the AR (2) tests and; (ii) the validity of the instruments in the Sargan OIR test. Source: Authors’ calculations (2022).

We consider all levels of significance to draw the following conclusions. First, in most specifications, capital flight increases systematically with economic growth. Developing countries are known to have high growth volatility (Sheng 2010), which has important implications for various macroeconomic factors (Lin and Kim 2014).

However, the model results also suggest that the coefficient of inflation is negative and statistically significant at the 5% level. Second, the coefficient of the trade openness variable is, in most specifications, negative and statistically significant, suggesting that capital flight decreases with trade openness. In addition, Asongu and Odhiambo (2020) pointed out that there is no confirmed expected effect of exporting, and trade openness in general, on capital flight because its impact depends on whether it is limited to a few sectors of the economy or generalized. Regarding external debt, our results are generally consistent with the literature, with capital flight increasing with external debt; however, in most regressions, the coefficients are not statistically significant. Finally, the coefficient of corruption control is negative and statistically significant at the 10% significance level in Model 5 in Table 3, indicating that these highly corrupt African economies tend to experience more capital flight. This result is consistent with Le and Rishi’s (2006) empirical findings.

The impacts of non-resource tax revenues on capital flight using the Morgan Guaranty method through the role of natural resource rents are presented in Table 4. These results were almost identical to those of previous studies. Thus, in all four regressions, the estimated coefficients of lagged capital flight were positive and statistically significant. Similarly, the empirical results in Table 4 reveal negative and statistically significant impacts of total natural resource rents on capital flight. Thus, the oil rent component reduces capital flight, such that a percentage increase in oil and natural gas rents reduces capital flight by 1.478% and 19.347%, respectively, with the opposite result (an increase) for coal rent of 25.781%. Non-resource tax revenues show a direct and statistically significant relationship with capital flight, implying that an increase in non-resource tax revenues leads to a decrease in capital flight.

Table 4. GMM estimation of the tax revenue mobilization system on capital flight (MG) using the Morgan Guaranty method.

Variable	Dependent variable: MG
	Model 1 RENT	Model 2 OIL	Model 3 FOREST	Model 4 GAS	Model 5 COAL
L.MG	0.315***	0.209***	0.281***	0.246***	0.284***
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
TNR	−6.169**	−4.473***	−1.869	−1.739	2.411*
	(0.011)	(0.003)	(0.401)	(0.371)	(0.048)
RENT	−1.721***
	(0.003)
OIL		−1.478***
		(0.003)
FOREST			−0.365
			(0.696)
GAS				−19.347**
				(0.011)
COAL					25.781***
					(0.001)
TNR*RENT	0.113*	0.095*	1.983***	1.058***	−0.713***
	(0.035)	(0.046)	(0.000)	(0.001)	(0.000)
ΔGDP	0.831**	1.068***	1.983***	1.832***	1.429***
	(0.012)	(0.000)	(0.000)	(0.000)	(0.002)
INF	−0.091***	−0.042***	−0.068*	– 0.038**	−0.083***
	(0.000)	(0.008)	(0.039)	(0.017)	(0.000)
OPEN	0.133	0.008	−0.602***	−0.154	−0.235
	(0.279)	(0.944)	(0.002)	(0.117)	(0.135)
DEBT	1.007	3.837	6.813**	1.938	0.728
	(0.741)	(0.184)	(0.014)	(0.207)	(0.696)
CC	−4.902	−6.784***	1.747	−6.098*	−9.463***
	(0.179)	(0.007)	(0.629)	(0.056)	(0.001)
_cons	23.345	−54.692	−116.338*	−29.163	−34.794
	(0.663)	(0.366)	(0.044)	(0.431)	(0.436)
AR(1)	0.002	0.005	0.003	0.005	0.010
AR(2)	0.673	0.911	0.563	0.901	0.787
Sargan OIR	0.063	0.012	0.062	0.254	0.055
Hansen OIR	0.353	0.385	0.329	0.333	0.311
H excluding group	0.339	0.255	0.487	0.648	0.141
Dif(null, H = exogenous)	0.393	0.554	0.237	0.158	0.647
1. (b) IV (years, eq(dif)) H	0.329	0.620	0.383	0.828	0.459
Dif(null, H = exogenous)	0.415	0.165	0.298	0.050	0.200
Fisher	356.32	1606.57	430.49	1853.87	10360.99
Instruments	29	29	29	29	29
Country	29	29	29	29	29
Observation	492	492	492	490	492

Note: Numbers in parentheses () are standard errors (SEs). *** Significance at 1% level, ** Significance at 5% level, * Significance at 10% level. (a) estimated parameters and F-statistics and (b) failure to reject the null hypotheses of (i) the absence of autocorrelation in the AR(2) tests and; Source: Authors’ calculations (2022).

As the results in Table 4 show, the interaction between non-resource tax revenues and rents from natural/petroleum resources, except coal, is positive and statistically significant. The economic intuition that can be drawn from this result suggests that, as African countries’ non-resource tax revenues increase, capital flight intensifies as countries exploit or dispose of more natural resources. This result is in line with existing studies (Kwaramba, Mahonye, and Mandishara 2016; Mpenya, Metseyem, and Epo 2016; Ndikumana and Boyce 2010; Ndikumana and Sarr 2019) and is attributed to numerous mechanisms such as rent-seeking, corruption, tax evasion, and trade misinvoicing. African countries are endowed with large and varied natural resources and have profited abundantly from them. However, some of these rents appear to be illegally transferred abroad in the form of capital flight, although the sectors are strictly state rather than market-controlled.

6. Conclusion and policy implications

This study examines the impact of tax revenue mobilization on capital flight via the conditioning role of natural resources in 30 African countries over the period 1998–2018.

To facilitate the identification of the channel through which tax revenues influence capital flight, total natural resource rents, and four rent-related natural resource components were used in a disaggregated manner (coal, natural gas, oil, and forest resources). The empirical analysis uses a two-stage method of generalized moments. We find that non-resource tax revenues are negatively related to capital flight and that tax revenue mobilization efforts are therefore associated with a reduction in capital flight. Similarly, we find a negative relationship between total rents from natural resources and capital flight. This conclusion does not apply to coal rents. However, the conditional impact between tax revenue and natural resources, on the one hand, and its oil component on the other, presents positive and statistically significant coefficients. This suggests that when total resource/oil rents are non-zero, the impact of non-resource tax revenues on capital flight weakens for countries with more resources or oil rents.

The results of this study suggest that efforts to improve the quality of institutions and increase domestic investment returns alone will not enable governments to successfully address the problem of capital flight. More targeted strategies are required to curb tax-induced capital flight. These include greater accountability of African governments in the management of natural resources and greater transparency in the reporting of production, sales, profits, and tax payments by companies operating in the natural resources sector. These results also imply that the creation of a democratic environment is linked to stronger governance institutions, reduced corruption, and a better domestic investment climate. Finally, the promotion of good governance at all levels of the socio-economic sphere is necessary to improve tax citizenship and voluntary tax compliance, which are levers for maximizing tax revenues and fiscal consolidation in developing countries.

Similar to similar studies of cross-national panel data, caution should be exercised. Problems associated with heterogeneity, multicollinearity, and endogeneity, particularly the heterogeneity of individual countries, may not have been fully considered. Future research could explore the possibilities of analyzing individual countries, disaggregating countries and examining the direction of causality between tax revenue and capital flight.

Notes

1 The residual method (World Bank 1985; Morgan Guaranty 1986); the Dooley method (Dooley 1986); the commercial misinvoicing method (Bhagwati 1964); and the hot money method (Cuddington 1986).

2 We were unable to include resources from minerals due to data limitations. We have included forestry resources because they are considered an integral part of World Bank data. However, the IMF also found that, on average, effective taxation in the mining sector was significantly lower than in the oil and gas sectors.

3 Depending on the time horizon of the study, Asongu and Nwachukwu (2016) caution against using more than five control variables as this would lead to biases in the estimated coefficients due to the proliferation of instruments.

Appendix

Table A1. List of sampled countries in sub-Saharan Africa.

	Oil rich	Non oil riche	Poor ressource
Country	Cameroon	Botswana	Burkina Faso
	Gabon	Congo, Dem. Rep.	Burundi
	Ghana	Mauritania	Malawi
	Nigeria	Mozambique	Rwanda
	Congo, Rep.	Tanzania	Kenya
	Cote d'Ivoire	South Africa	Uganda
	Angola	Morocco	Zimbabwe
	Sudan	Tunisia	Madagascar
	Algeria	Zambia	Seychelles
	Egypt		Sierra Leone
	Ethiopia

Note: Resource-rich countries are those whose average annual rents from natural resources (excluding forests) exceed 5% of GDP.

Table A2. Variable definitions.

Variables	Signs	Définition	Sources
Economic growth	ΔGDP	GDP growth rate (annual %)	WDI (World Bank), 2021
Inflation	INF	Annual change in the consumer price index (CPI)	WDI (World Bank), 2021
Capital flight from Morgan Guaranty	MG	Where MG is equal to the measure of a private claim for capital flight.	Author's calculation
Capital flight	FKR	The ratio of capital flight to GDP	Ndikumana and Boyce (2018)
Tax revenues	TR	Total non-resource revenues, including social contributions.	Government Revenue Dataset 2020
Natural resource rents	RENT	Total natural resource rents as a percentage of GDP	WDI (World Bank), 2021
Forest rents	FOREST	Forestry rents as a percentage of GDP	WDI (World Bank), 2021
Oil rents	OIL	Oil rents as a percentage of GDP	WDI (World Bank), 2021
Gas rents	GAS	Gas rents as a percentage of GDP	WDI (World Bank), 2021
Coal rents	COAL	Coal rents as a percentage of GDP
Foreign debt	DEBT	Total external debt is measured as total outstanding external debt as a percentage of GDP.	WDI (World Bank), 2021
Corruption control	CC	Perceptions of the degree to which public power is exercised for private ends, including minor and major forms of corruption, as well as the ‘capture’ of the state by elites and private interests	WDI (World Bank), 2021

Note: WDI is a connotation for the World Bank Development Indicator data in the World Bank database.