Beyond Aid Volumes: Multidimensional Aid Dependency, Institutional Quality and Economic Growth in Sub-Saharan Africa

Highlights

Please indicate how your work links to systems science via your contributions to sys-tems practice, theory, and/or methodology.

• A multidimensional ODA Dependency Index captures the interconnected fiscal, external-financing and debt dimensions of aid dependence, extending conventional single-ratio measures.
• An integrated empirical framework combines institutional moderation, dynamic panel estimation, cross-sectional-dependence controls and threshold analysis to examine aid–growth relationships as a conditional and potentially nonlinear system.

What are the main findings and/or the implications of the main findings?

• Higher ODA dependency is consistently associated with lower economic growth, while stronger institutional quality independently supports growth but does not significantly moderate the aid-growth relationship.
• Threshold effects are sensitive to index specification, underscoring the need for domestic resource mobilization, productive investment and institutional capacity to reduce structural aid dependence and strengthen sustainable growth.

Abstract

Economic growth and the effectiveness of official development assistance (ODA) have long been a topic of debate, with the current concern being that there are high levels of aid dependency in Sub-Saharan Africa, despite low levels of structural transformation. The study analyses the growth impacts of multidimensional ODA dependency, the moderating effect of institutional quality and the possible thresholds of aid dependency in an unbalanced panel of 48 Sub-Saharan African countries over the period 2000–2025. Two-way fixed effects, System Generalized Method of Moments, Difference Generalized Method of Moments, and Common Correlated Effects Pooled are used to analyze a principal component-based ODA Dependency Index, and least-squares threshold regression is used to investigate multiple thresholds. These results show that there is a strong negative correlation between aid dependency and economic growth, while institutional quality is consistently positive to growth, but does not significantly moderate the aid–growth relationship. There are no stable thresholds for aid dependency as a function of the specification of the index. The findings highlight the need to improve domestic resource mobilization, productive investment and institutional capacity to decrease reliance on structural aid and ensure sustainable economic growth in the long term.

1. Introduction

For more than 60 years, official development assistance (ODA) has been a key source of external development finance for Sub-Saharan Africa (SSA). Despite these relatively large concessional resource inflows, the region is still facing persistent structural constraints, such as low domestic revenue mobilization, growing fiscal deficits, increasing external debt, and donor priority shifts [1]. Given these ongoing difficulties, questions arise about the effectiveness of long-term aid dependence on development, as well as if long-term dependence on external support has led to structural dependency or self-reliant development. According to the Organization for Economic Co-operation and Development’s Development Assistance Committee, ODA is concessional financing provided by official agencies with a grant element of at least 25%, for the economic development and welfare of developing countries and excluding military and commercial transactions [2]. In the last year reported, DAC members disbursed some USD214.5 billion in ODA, which equates to 0.34% of the combined gross national income (GNI) of all DAC members [3]. In view of the size of such transfers, it is crucial to policy and research to understand when and why dependency on aid impacts economic growth.

Although foreign aid has been a subject of empirical research over many years, it is not clear how aid and economic growth are connected. The initial research claimed that aid stimulates economic development, especially in nations where macroeconomic policies and institutions are good [4,5]. Subsequent studies have cast doubt on the above finding, showing that estimates of the relationship between aid and growth are highly dependent on the model specification, the method of estimation and the sample, and that, when accounting for endogeneity, many such studies have found only weak or statistically insignificant relationships between the two [6,7]. More recent reviews suggest that the reasons for these disparate results have to do with the fact that dependency on aid is a multidimensional and potentially nonlinear phenomenon whose impacts are partly shaped by the institutional and structural context, rather than with any uniform average effects of aid across countries [8,9].

There are three important gaps in the current literature. Previous research mainly looks at one indicator of aid dependency, usually net ODA as a percentage of the Gross Domestic Product (GDP) or gross national income (GNI). Though popular, this measure provides only a measure of the volume of inflows of aid and fails to consider longer-term impacts of external dependency, such as the decline in domestic tax collection that follows years of external dependence, and the per capita amount of aid received [10,11]. As a result, countries can have vastly varying degrees of fiscal dependency and economic vulnerability despite having similar levels of aid-to-GNI. Moreover, institutional quality is usually included as an exogenous variable affecting growth instead of as an intermediary variable linking aid dependency to the performance of economic institutions. This does not allow for an assessment of whether stronger institutions enhance the effectiveness of aid, through better resource allocation, accountability and productive investment [12,13]. Likewise, the presence of an aid dependency trap is regularly suggested, but with few formal studies being able to estimate whether the aid–growth relationship changes after an endogenous threshold despite the availability of the econometric framework developed by [14]. To overcome these shortcomings, this study develops a multidimensional ODA Dependency Index based on principal component analysis of four indicators of ODA dependency: net ODA received (% GNI), net ODA per capita, external debt stocks (% GNI), and tax revenue (% GDP). It also constructs an Institutional Quality Index based on the six Worldwide Governance Indicators, explicitly accounts for institutional moderation, and calculates endogenous threshold levels of aid dependence for 48 Sub-Saharan African countries for the period 2000–2025 based on World Development Indicators data.

This study therefore takes a different approach from traditional studies which ask whether foreign assistance can foster economic growth by investigating under which institutional conditions and at what level of dependence on foreign aid ODA affects economic growth in Sub-Saharan Africa. The study’s focus on conditional and nonlinear relationships contributes to a broader understanding of aid effectiveness in developing economies, rather than focusing solely on average effects.

The research presented in this study makes several important contributions to the literature: creating an ODA Dependency Index that is multidimensional and can be replicated by using PCA to offer a more holistic assessment of aid dependence than the traditional single-ratio indicators; directly testing the potential moderating role of institutional quality in addition to being an explanatory variable, as it is explicitly modelled as such; using an endogenous threshold method to detect structural breakpoints in the distribution of aid dependency and presenting novel evidence on the presence of an aid dependency trap in SSA; and offering one of the most comprehensive analyses of the region at hand, covering up to 48 SSA countries between 2000 and 2025, which enhances the robustness, external validity and policy relevance of empirical evidence on aid dependency and economic growth.

2. Literature Review

The relationship between official development assistance (ODA) and economic growth is one of the most debated in the field of development economics. Aid–growth literature spans more than 60 years, ranging from the belief that concessional finance can alleviate the savings and FX constraints that low-income economies face [15,16] to the opposite view that aid, especially in the context of poor governance, stifles institutional development, encourages rent-seeking, and never leads to sustained growth [17,18]. The debate is interwoven with the question of where and how much money should be invested, as Sub-Saharan Africa (SSA) is at the heart of the controversy: it is the region that has received the most aid as a share of gross national income (GNI) over the years, and it is the region where the relationship between aid and growth remains the most difficult to prove empirically. This section summarizes the theoretical underpinnings, the seminal empirical controversies, the institutional-feedback literature, and the new set of panel-data nonlinearities and thresholds that justify the focus of the present study on thresholds of aid dependency between 2000 and 2025.

2.1. Theoretical Framework

The theoretical framework of this research is based on the institutional economics approach that considers institutions as the main factors that cause long-term economic growth and the channels through which foreign aid affects development indicators. In contrast, modern development literature has become increasingly sceptical of the effectiveness of aid as a transfer of resources and increasingly holds that the institutional context in which aid is distributed, managed and productively used is key to its effectiveness. This angle goes beyond the conventional models of resource gaps, to acknowledge how governance systems, capacity to implement policy, and institutional accountability affect the productivity of externally provided development assistance. As a result, the association between aid dependency and economic growth is likely to be conditional, not universal. The intellectual foundations for this argument stem from the institutional framework outlined by [19], who defines institutions as formal and informal rules that reduce uncertainty, influence incentives, and enable economic exchange. This hypothesis was later formalized in the seminal work by [20], which used the variation in settlement mortality in Europe as an instrument for institutional development and showed that the difference in the quality of institutions (defined as the risk of expropriation) accounts for about 75% of the cross-country variation in income per capita. More significantly, they found that after accounting for institutional quality, geography in Africa is no longer a good predictor of income level, moving institutions from being a proximate factor of development to its fundamental factor. This has been further advanced by [21], who found that there are three competing explanations for the persistence of differences in economic performance between countries institutions, geography and trade integration—and suggested that institutional quality is the best explanation of the long-run differences in economic performance.

The institutional approach has marked a major change in the literature on aid effectiveness. Where institutions are the main engines of economic development, foreign aid is unlikely to lead to the same results in different recipient economies. In contrast, the flow of identical aid can result in significantly different development outcomes depending on the quality of the governance, institutional capacity and accountability mechanisms in recipient countries. Thus, institutional quality is not just a control variable, but also a structural moderator which can either complement the domestic development process or consolidate existing inefficiency. This conditional interpretation offers theoretical grounds for modelling the aid–growth relationship as an interaction effect.

The theoretical framework is also strengthened by incorporating complementary viewpoints that help explain how aid dependency can affect economic growth. The Two-Gap Theory [16] assumes that foreign aid alleviates saving-investment and foreign-exchange constraints which hinder the build-up of capital stocks in developing economies, thus providing the basis for including gross capital formation and trade openness as important control variables. Conversely, Dependency Theory [10,22] suggests that heavy reliance on external support can lead to a lack of resort to own resources and a lack of institutional accountability, as well as structural dependence. This view underpins the inclusion of a multidimensional ODA Dependency Index, given that aid reliance carries macroeconomic consequences that cannot be captured by a single indicator. The framework also draws on the concept of Institutional Complementarity theory [23], which states that institutional architecture functions as a web of interdependent institutions rather than independent ones. The moderating effect of institutional quality is therefore not a foregone conclusion and may differ depending on institutional arrangements, so the productivity of aid is expected to differ across different institutional configurations. A positive interaction would suggest that better institutions increase the developmental benefits of aid, while a negative one would imply that institutional improvements are not sufficient to make aid successful in boosting a country’s economic growth. Taken together, these complementary theoretical approaches offer a plausible basis for analyzing the conditional and nonlinear dynamics between aid dependence, institutional quality and economic growth in Sub-Saharan Africa.

2.2. The Reverse Channel: How Aid Dependency Shapes Institutional Outcomes

Another analytically pertinent question is whether long-term aid dependency affects institutions. This channel is key to the “aid-dependency-threshold” concept used in this study, as it suggests that the relationship between aid and growth might not necessarily be one-way, and that an aid-dependent economy may follow a particular governance path distinct from a moderately aid-dependent economy.

Some of the first systematic evidence for this reverse channel was offered in a study by [10] dedicated to Sub-Saharan Africa. They surveyed the extensive literature on the links between aid and governance outcomes and bureaucratic capacity and identified a series of mechanisms through which high aid levels are negatively correlated with governance indicators and bureaucratic capacity, including a decline in government accountability to domestic taxpayers, increased aid reliance and coordination failures, and donors displacing tax-based revenue mobilization, as noted in previous research linking higher aid dependence with lower domestic tax effort. Significantly, ref. [10] did not argue that the amount of aid should be curtailed, but that the form and selectivity of aid allocation could strengthen rather than weaken governance.

Ref. [18] extended this argument to the global level by demonstrating the quantitative impact of foreign aid on democratic institutions using a panel-data set of 108 recipient countries from 1960 to 1999. They found that for the 75th percentile of the aid-to-GDP ratio in their sample, the democracy index declines by about half of a point. Interestingly, the authors found the institutional “curse” effect of aid to be more significant than that of natural resource rents, an intriguing result in the context of the many studies on the resource curse in Africa’s petrostates. This paper is directly relevant to the dependency-threshold framing of the present study, as it explicitly identifies its own threshold of aid intensity beyond which institutional damage becomes detectable—an empirical approach structurally similar to the present study’s panel threshold regression (PTR) approach, though focused on a democracy outcome rather than growth.

2.3. Nonlinearity and Threshold Effects in the Aid–Growth–Institutions Nexus

The most methodologically relevant literature for the present study introduces explicit modelling of nonlinearities in the aid–growth relationship, typically through panel threshold regression [14], smooth transition panel regression (PSTR), and quantile regression techniques. This literature broadly agrees that the aid–growth relationship is not adequately captured by a single linear coefficient, and that assuming otherwise is likely to account for much of the instability found in the Burnside Dollar/Easterly–Levine–Roodman exchange [4,24,25]. In a UNU-WIDER working paper, ref. [26] apply a nonlinear panel threshold framework to a panel of African economies between 1980 and 2007, and do not find evidence of a positive relationship between aid and growth under a linear specification but do observe significant threshold effects when the aid variable is interacted with macroeconomic policy variables, confirming that linear cross-country specifications may mask conditional effects that exist in specific policy or institutional settings. At a specific threshold of 12.74% of GDP, Hansen’s bootstrapped panel threshold regression analysis for the West African Economic and Monetary Union (WAEMU) over 1980–2015 shows that aid positively affects growth by 0.69 percentage points per percentage point of aid, suggesting that aid works best as a complement to domestic investment rather than a substitute for it [27].

Ref. [28] tackled the threshold question from a distributional perspective, demonstrating that the aid–growth linkage is nonlinear not only with respect to the intensity of aid received, but also with respect to the recipient country’s initial development level: countries in the top quantiles of the growth distribution gain more from aid than countries in the bottom quantiles, with the correlation between aid and growth systematically larger in the top quantiles. This has direct implications for interpreting the heterogeneity found across the SSA panel in this study, signalling that thresholds are unlikely to be uniform across the panel.

Newer evidence specifically for SSA has used threshold approaches to examine the direct link between institutional quality and aid outcomes. A nonlinear threshold study of governance quality and poverty reduction across South Asia and SSA using dynamic panel threshold specifications [29] revealed that the effect of governance quality on reducing extreme poverty is only realized above certain thresholds of the governance index (0.20, 0.62, and 0.70, corresponding to poverty lines of $2, $3.20, and $5.50 per day, respectively) a methodological caution of great relevance to the threshold-based growth specifications used here. Similarly, studies on institutional-quality thresholds in the finance–economic nexus in Africa [30] revealed a threshold score of 5.73 on a 10-point scale, below which financial development does not achieve productive economic upgrading, with only a handful of African countries above the threshold, further highlighting the distance of the institutional frontier in SSA from the level needed for complementary macro-financial flows, including aid, to be growth-enhancing.

2.4. Institutional Quality and Growth in Sub-Saharan Africa: Recent Panel Evidence

A reasonable amount of SSA-specific panel-data research has examined institutional quality and growth without accounting for aid, which sets the baseline institutional-growth elasticities that need to be compared with the elasticity of the aid-interaction effect. New findings from [31], using a two-step system GMM estimator for 31 SSA countries between 1991 and 2015 and principal component analysis to decompose institutional indices into separate institutional clusters, suggest that investment-promoting and democratic/regulatory institutions have significant positive impacts on growth, while conflict-preventing institutions become immaterial once other clusters are controlled for cautioning against lumping “institutional quality” into a single homogeneous construct in panel growth models. In contrast, ref. [32], using system-GMM on a panel of 43 SSA countries, found no significant direct causal effect of aggregate institutional quality on growth, highlighting how model dynamics, sample composition, and the degree of institutional-measure aggregation shape empirical results on the institutions–growth nexus in SSA.

Using a panel autoregressive distributed lag (ARDL) model on 15 SSA countries over 2002–2021, ref. [33] found that institutional quality plays a significant role in short-run growth, whereas ODA and the square of ODA are not significant in the short run; however, under mean-group and dynamic fixed-effects estimation, ODA, the square of ODA, and institutional quality are jointly significant short-run determinants of growth, with the squared term suggesting a nonlinear (concave) relationship at high aid intensities, implying diminishing or reversing returns. Ref. [34], disaggregating aid by type for 45 SSA countries between 1980 and 2017, found that the adverse effect of aid volatility on growth is not fully mitigated by institutional quality, challenging the simple “good institutions neutralize bad aid” narrative and indicating that aid volatility, rather than aid volume, is a first-order determinant of the aid–growth relationship in the region. Illustrating the sector-specific nature of aid’s growth and structural-transformation effects, ref. [35] apply Driscoll–Kraay fixed-effects estimators, fixed-effects panel threshold regression, and method-of-moments quantile regression to sectoral aid allocation in 32 SSA countries during 2002–2019 and conclude that aggregated ODA measures may mask significant heterogeneity relevant to threshold estimation. Ref. [36] also finds that physical and digital infrastructure positively condition the aid–growth relationship in SSA, while corruption control determines the scope of aid-financed infrastructure’s contribution to broad-based growth.

2.5. Synthesis and Research Gap

This review gives rise to three general conclusions. First, while the theoretical case for institutions as a structural moderator rather than a simple control variable in the aid–growth relationship is fairly well established, empirical operationalization of moderation in SSA panels has not been consistent: some studies suggest that institutions moderate the impact of aid on growth [31,36], while others find no significant moderating effect once dynamics are modelled [32], and still others show that institutions do not shield countries from the adverse growth impacts of aid volatility specifically [34]. Second, the reverse-causality literature [10,18] shows that aid dependency itself can deteriorate the institutional environment, suggesting that panel models of aid’s impact on growth in SSA should treat institutional factors as both an endogenously determined moderator and a pure exogenous one. Third, and most directly relevant to the present study, the threshold and nonlinear literature [26,27,28] indicates that linear aid–growth relations are likely mis-specified, and that thresholds in institutional quality and aid intensity may exist beyond which the direction and size of the aid–growth relation shift.

What remains unanswered is whether aid-intensity and institutional-quality thresholds, estimated as two separate processes, can be generalized across the entire Sub-Saharan African panel as [27] did for WAEMU economies or whether a unified dynamic panel threshold framework exists. The current research addresses this gap by estimating a dynamic panel threshold model of ODA, institutional quality and economic growth for Sub-Saharan Africa over 2000–2025, explicitly identifying the values of ODA within which its effect becomes negative and/or differs in magnitude, conditional on the institutional-quality regime of the recipient country.

Building directly on the gaps identified above, four hypotheses are advanced and tested including: H1: multidimensional ODA dependency is negatively associated with economic growth; H2: institutional quality is positively associated with growth; H3: institutional quality moderates the aid–growth relationship, such that the marginal effect of aid dependency becomes less negative as institutional quality improves; and H4: the aid–growth relationship is nonlinear, with the marginal effect differing above and below an endogenously estimated threshold. H1 and H2 are supported under fixed effects and CCEP; H3 is not supported under any estimator; H4 is only partially supported, since the threshold location is sensitive to index specification.

3. Data and Methodology

3.1. Research Design

The study employs a quantitative longitudinal design using an unbalanced panel of 48 Sub-Saharan African countries observed annually from 2000 to 2025. The unbalanced structure reflects differences in data availability across countries and years. Secondary data are obtained exclusively from World Bank databases. The World Development Indicators (WDI) provide macroeconomic measures, including GDP, net official development assistance (ODA), external debt, tax revenue, gross capital formation, and trade openness. Institutional quality is measured using the Worldwide Governance Indicators (WGI), covering government effectiveness, control of corruption, regulatory quality, rule of law, voice and accountability, and political stability. To ensure parsimony, these six dimensions are aggregated into a single Institutional Quality Index using principal component analysis (PCA). Dimension-specific regressions interacting each governance indicator separately with ODAI are not estimated in the present analysis because six repeated interaction tests would require multiple-testing correction and substantially increase model complexity. This limitation is explicitly acknowledged, with dimension-specific interaction analysis identified as a priority for future research to determine which governance dimensions drive the aggregate institutional effect.

3.2. Variable Definition and Data Source

Table 1. Variable definition and data source.

Variable	Measure/Construction	Data Source	Remarks
GDP Growth (GDPG) (Dependent Variable)	Annual GDP growth rate (%) based on constant 2015 US$.	World Bank World Development Indicators (WDI)	Measures annual economic growth and serves as the dependent variable.
ODA Dependency Index (ODAI) (Independent Variable)	The ODA Dependency Index (ODAI) is constructed using principal component analysis (PCA) from three standardized indicators: net ODA received (% of GNI), net ODA received per capita (current US$), and external debt stocks (% of GNI). Tax revenue is excluded listwise because of missing data. To assess robustness, a parallel four-indicator index incorporating tax revenue is estimated for the 32-country subsample and directly compared with the three-indicator index covering 41 countries.	World Development Indicators (WDI)	Three-indicator PCA index selected based on superior data coverage (41 countries; 819 estimation observations) and stationarity (CIPS t-bar = −2.57). The first principal component explains 54.9% of total variance (eigenvalue = 1.65).
Institutional Quality Index (IQI) (Moderating Variable)	Composite index derived from six Worldwide Governance Indicators using PCA: government effectiveness, control of corruption, regulatory quality, rule of law, voice and accountability, and political stability and absence of violence.	World Bank Worldwide Governance Indicators (WGI)	Existing governance composite adopted following the methodology of [12] and used without re-estimation. Higher values indicate stronger institutional quality.
Gross Capital Formation (GCF) (Control Variable)	Gross capital formation as a percentage of GDP.	World Development Indicators (WDI)	Proxies domestic investment and productive capital accumulation, consistent with growth theory.
Trade Openness (TRADE) (Control Variable)	Sum of exports and imports of goods and services expressed as a percentage of GDP.	World Development Indicators (WDI)	Captures the degree of integration with the global economy and controls for external sector effects on economic growth.

Source: Authors’ compilation.

presents the description of variables and data sources.

3.3. Model Specification

Four complementary specifications are estimated. The baseline static model is:

GDPGit = α + β1ODAIit + β2IQIit + β3(ODAI × IQI)it + β4GCFit + β5TRADEit + μi + λt + εit

(1)

estimated with two-way (country and year) fixed effects and standard errors clustered by country, where GDPG is GDP growth, ODAI the ODA Dependency Index, IQI the Institutional Quality Index, GCF gross capital formation (% GDP), and TRADE trade openness (% GDP). The dynamic specification adds the lagged dependent variable:

GDPGit = δGDPGit − 1 + β1ODAIit + β2IQIit + β3(ODAI × IQI)it + β4GCFit + β5TRADEit + μi + εit

(2)

Given the baseline dynamic specification, the threshold model allows the effect of the ODA Dependency Index (ODAI) to differ depending on whether ODA dependency is below or above an estimated threshold value (γ). Following [14], the threshold equation can be written as:

GDPG = f(ODAI ≤ γ, ODAI > γ, IQI,GCF,TRADE) + μi + εit

(3)

where GDPGit is the GDP growth rate of country i at time t; GDPGit−1 is the lagged dependent variable; ODAIit is the ODA Dependency Index; γ is the estimated threshold value; I(·) is an indicator function taking the value 1 when the stated condition holds and 0 otherwise; IQIit is the Institutional Quality Index; GCFit denotes gross capital formation; TRADEit represents trade openness; μi captures unobserved country-specific effects; and εit is the idiosyncratic error term.

This specification estimates separate marginal effects of ODA dependency on economic growth below (β1) and above (β2) the estimated threshold γ. The threshold value is identified by conducting a grid search over the central 10th–90th percentiles of the ODAI distribution and selecting the value that minimizes the residual sum of squares, consistent with the least-squares threshold estimator of [14].

3.4. Estimation Technique

To obtain robust inferences, several complementary estimation methods are used. First, two-step system GMM [37,38] handles endogeneity, reverse causality, and dynamic persistence through the inclusion of both level and differenced moment conditions. Second, one-step difference GMM [37] is estimated as an alternative specification, since the smaller number of instruments makes it more stable in small panels, although it is less efficient than system GMM. Estimation is performed in Stata/SE 18, using collapsed instruments and lagged-depth restrictions to avoid instrument proliferation, as proposed by [39]. Where cross-sectional dependence is present, year fixed effects are added, and robustness is checked using the Common Correlated Effects Pooled estimator [40]. Nonlinearities are examined using the least-squares threshold method [14], which identifies the ODA-dependence level that minimizes residual variation across model specifications. Given potential bidirectional causality between aid dependence and institutional quality, the analysis explicitly addresses endogeneity rather than treating institutional quality as purely exogenous. System and difference GMM instrument the Institutional Quality Index and the ODAI × IQI interaction using lagged levels and differences, while country fixed effects absorb time-invariant sources of reverse causality. Nevertheless, internal instruments cannot fully eliminate reverse causality; therefore, the absence of a valid external instrument for institutional quality is acknowledged as a limitation and priority for future research.

4. Empirical Results

4.1. Pre-Estimation and Diagnostic Test Analysis

4.1.1. Descriptive Statistics

Table 2. Descriptive statistics.

Variable	Mean	Std. Dev.	Min	Max
GDP Growth (%)	4.08	5.38	−46.08	63.38
ODA Dependency Index (3-indicator PCA)	0.00	1.32	−1.77	6.31
Institutional Quality Index	−0.69	0.63	−2.50	+2.50
Gross Capital Formation (% GDP)	22.63	9.62	−3.68	76.78
Trade Openness (% GDP)	68.21	34.08	2.00	222.18

Source: Authors’ computation, 2026.

reports descriptive statistics for the core variables over 2000–2025, based on the three-indicator index sample.

Table 2 shows great heterogeneity between economies in Sub-Saharan Africa. The average GDP growth rate is 4.08%, but the wide dispersion of the variable and the presence of extreme values denote high volatility across economies. The ODA Dependency Index is near zero, reflecting PCA standardization, while the wide dispersion indicates high variance in aid dependency between countries. Institutional quality is low on average, with substantial variation indicating poor governance across countries with high heterogeneity. Gross capital formation averages 22.63%, but the dispersion shows unequal investment. Trade openness has the highest variance, reflecting differences in foreign economic exposure.

4.1.2. Pesaran Cross-Sectional-Dependence Test Analysis

Table 3. Pesaran cross-sectional-dependence (CD) test analysis.

Variable/Residuals	CD Statistic	N Countries	Mean Corr. (Signed)	Mean |Corr.|	Interpretation
GDP Growth	21.602	48	0.126	0.238	p < 0.0001—reject independence
ODA Dependency Index	7.731	44	0.053	0.340	p < 0.0001—reject independence
Institutional Quality Index	1.269	48	0.009	0.424	p = 0.204—fails to reject independence
Gross Capital Formation	6.065	45	0.034	0.377	p < 0.0001—reject independence
Trade Openness	8.006	45	0.389	0.053	p < 0.0001—reject independence
Growth-Equation Residuals (entity FE)	14.429	41	-	0.266	p < 0.0001—reject independence

Source: Authors’ computation, 2026.

reports the [41] CD test on every variable in levels, on the growth-equation residuals, and on the full estimating sample underlying the three-indicator ODA Dependency Index.

Four of the five variables, together with the growth-equation residuals, strongly reject cross-sectional independence, indicating that Sub-Saharan African economies are exposed to common shocks beyond those captured by country and year fixed effects. This finding justifies incorporating year dummies in the GMM specification and retaining the Common Correlated Effects Pooled (CCEP) estimator as a robustness check. The Institutional Quality Index is the only variable for which the CD test fails to reject cross-sectional independence, despite high average correlations. This pattern suggests that institutional dynamics vary across countries, with positive and negative institutional interactions offsetting one another and reflecting heterogeneous development trajectories across subregions. Although year fixed effects capture common temporal shocks, they do not fully absorb cross-sectional dependence generated by regional co-movement. Accordingly, CCEP is retained as the primary robustness estimator because it explicitly and directly accounts for unobserved common factors through cross-sectional averages, providing a more rigorous response to the detected dependence than year dummies alone.

4.1.3. CIPS Second-Generation Panel Unit-Root Test Analysis

Table 4. Pesaran CIPS second-generation panel unit-root test.

Variable	CIPS (T-Bar)	Reject Unit Root @5%	Reject Unit Root @10%	Order of Integration
GDP Growth	−3.968	45/47 (96%)	46/47 (98%)	I(0)
ODA Dependency Index (3-indicator)	−2.570	29/44 (66%)	30/44 (68%)	I(0)
Institutional Quality Index	−2.357	24/48 (50%)	25/48 (52%)	I(0)
Gross Capital Formation	−1.924	18/45 (40%)	23/45 (51%)	I(1)
Trade Openness	−1.666	12/45 (27%)	16/45 (36%)	I(1)

Source: Authors’ computation, 2026.

reports the [42] CIPS second-generation panel unit-root test for the same five variables, including the share of countries individually rejecting a unit root at the 5% and 10% levels; full per-country CADF t-statistics are reported in the underlying estimation output.

The [42] CIPS test indicates a mixed order of integration across the variables. GDP growth is stationary at levels (I(0)), while the ODA Dependency Index is also predominantly stationary at levels and is therefore treated as I(0). Institutional quality exhibits weak level stationarity, with the unit-root hypothesis rejected for approximately half of the sampled countries. By contrast, gross capital formation and trade openness are predominantly non-stationary in levels, indicating I(1) processes. This combination of I(0) and I(1) variables is accommodated through the study’s estimator selection rather than left unresolved. Because the dependent variable, GDP growth, is already I(0), the specification does not constitute the conventional I(1)-on-I(1) setting associated with classical spurious regression. Accordingly, the two-way fixed-effects, CCEP, and GMM estimators provide complementary safeguards against instability associated with non-stationary regressors: fixed effects control for unobserved heterogeneity and CCEP accounts for common factors through cross-sectional averages, while GMM relies on differenced transformations that mitigate level non-stationarity. A full-panel error-correction analysis is additionally retained as a robustness extension. This strategy therefore aligns empirical design with the mixed integration properties observed in the data.

4.1.4. Multicollinearity and Model Specification Diagnostics

Prior to empirical estimation of the FE and dynamic panel results, the regressors are checked for multicollinearity and the choice between fixed effects (FE) and random effects (RE) is formally tested using the Hausman specification test [43]. The pairwise correlation matrix of regressors is shown in

Table 5. Pairwise correlation matrix of regressors.

Variable	ODAI	IQI	ODAI × IQI	GCF	Trade Openness
ODA Dependency Index (ODAI)	1.000	−0.220	−0.605	0.136	0.019
Institutional Quality Index (IQI)	−0.220	1.000	0.207	0.085	0.324
ODAI × IQI (interaction)	−0.605	0.207	1.000	0.102	−0.040
Gross Capital Formation (GCF)	0.136	0.085	0.102	1.000	0.337
Trade Openness	0.019	0.324	−0.040	0.337	1.000

Source: Authors’ computation, 2026.

, the variance inflation factors (VIF) in

Table 6. Variance inflation factors (VIFs).

Variable	VIF	Tolerance (1/VIF)
ODA Dependency Index (ODAI)	1.723	0.580
Institutional Quality Index (IQI)	1.201	0.833
ODAI × IQI (interaction)	1.719	0.582
Gross Capital Formation (GCF)	1.234	0.810
Trade Openness	1.286	0.778
Mean VIF	1.433	-

Source: Authors’ computation, 2026.

, and the Hausman test comparison in

Table 7. Hausman specification test: fixed effects versus random effects.

Variable	Fixed-Effects Coef.	Random-Effects Coef.	Difference (FE − RE)
ODA Dependency Index (ODAI)	−1.168	−0.568	−0.600
Institutional Quality Index (IQI)	5.906	0.228	5.677
ODAI × IQI (interaction)	−0.767	−0.365	−0.402
Gross Capital Formation (GCF)	−0.028	0.015	−0.042
Trade Openness	0.108	0.021	0.088

Source: Authors’ computation, 2026. χ²(5) = 54.335, p < 0.001 (two-way model: χ²(5) = 25.639, p < 0.001).

.

The pairwise correlations show low-to-moderate correlations between regressors, with no concern of a high level of multicollinearity. Correlations between ODAI and other variables are relatively weak except for institutional quality (−0.220) and its interaction (−0.605), which are negatively correlated. These patterns include all explanatory variables simultaneously without incurring serious estimation bias in later regressions.

Correlation and VIF diagnostics indicate that multicollinearity is not a serious concern. The strongest pairwise correlation is between ODAI and its interaction term (ODAI × IQI) (−0.605; Table 5), reflecting the mechanical construction of the interaction rather than problematic collinearity. All VIFs remain below the conventional threshold of 5 (mean VIF = 1.433), while the interaction term’s VIF (1.719) is comparable to that of ODAI (1.723). Accordingly, the interaction term’s consistent insignificance across estimators (

Table 8. Two-way fixed-effects results.

Variable	Coefficient	Std. Err.	p-Value
ODA Dependency Index	−1.126	0.343	0.001 ***
Institutional Quality Index	3.386	1.618	0.037 **
ODAI × IQI (interaction)	−0.550	0.481	0.253
Gross Capital Formation	0.076	0.045	0.092 *
Trade Openness	0.050	0.023	0.028 **
Constant R2 = 0.072; F(5.752) = 9.88;	1.235	1.788	0.490

Source: Authors’ computation, 2026. p < 0.001. Notes: ***, **, and * denote statistical significance at the 1%, 5%, and 10% levels, respectively.

and

Table 9. Two-step system GMM and one-step difference GMM.

Variable	System GMM Coef. (p-Value)	Difference GMM Coef. (p-Value)
GDP Growth (t−1)	0.184 (0.002) ***	0.162 (0.028) *
ODA Dependency Index	−0.369 (0.343)	−0.639 (0.351)
Institutional Quality Index	0.458 (0.414)	8.619 (0.0005) ***
ODAI × IQI (interaction)	−0.194 (0.725)	−0.197 (0.863)
Gross Capital Formation	0.079 (0.067)	0.004 (0.964)
Trade Openness	−0.009 (0.430)	0.019 (0.698)
Year 2020 Dummy	−4.828 (0.000) ***	−4.768 (0.000) ***

Source: Authors’ computation, 2026. Note: * p < 0.05 and *** p < 0.001 denote the 5% and 0.1% levels of statistical significance, respectively.

) is interpreted as evidence against an institutional-quality moderating effect rather than as an artefact of multicollinearity.

The Hausman specification test strongly rejects the null hypothesis that the random-effects estimator is consistent (χ²(5) = 54.335, p < 0.001 for the one-way model), corroborated by the two-way (entity and time) fixed-effects model with clustered standard errors, which also rejects the null (χ²(5) = 25.639, p < 0.001). Unobserved country-specific effects are therefore correlated with the regressors, violating the orthogonality assumption of the random-effects model. The two-way fixed-effects model is accordingly used as the baseline estimator, with system GMM, difference GMM, and CCEP estimators used as robustness checks to address endogeneity and cross-sectional dependence.

4.2. Fixed-Effects Results

Table 8 reports the two-way fixed-effects estimates (country and year effects, standard errors clustered by country), on the expanded 41-country sample.

Table 8 shows the results of a two-way fixed-effects regression for 41 Sub-Saharan African countries, accounting for unobserved heterogeneity and common time shocks, with standard errors clustered at the country level. The coefficient on the ODA Dependency Index is negative and highly significant (−1.126; p = 0.001), signifying that greater aid dependence is associated with slower GDP growth, consistent with the possible demotivating effect of aid dependency. Institutional quality has a positive and statistically significant impact (3.386; p = 0.037), implying that better governance, regulatory quality, accountability, and public-administration efficiency are conducive to higher growth. The interaction between aid dependency and institutional quality is negative but insignificant (−0.550; p = 0.253), meaning that within this fixed-effects framework, institutional quality does not significantly alter the marginal effect of aid dependency on growth, even though the negative sign hints at a more complex relationship than a simple linear moderation. Among the controls, gross capital formation is positive and marginally significant (p = 0.092), and trade openness is positive and significant (p = 0.028). The model is jointly significant (F = 9.88, p < 0.001), despite a low R² reflecting substantial unexplained variation in growth.

4.3. Dynamic Panel Results: Two-Step System GMM and One-Step Difference GMM

Table 9 presents the dynamic panel results for two-step system GMM and one-step difference GMM, using collapsed instruments and year dummies.

The lagged dependent variable is positive and statistically significant in both specifications, indicating moderate persistence in economic growth. The ODA Dependency Index remains negative but is not statistically significant under either estimator, indicating a persistent negative relationship whose magnitude is not precisely estimated. Under difference GMM, institutional quality has a positive and significant effect on growth, whereas under system GMM it has no significant effect, suggesting sensitivity to the identification assumptions of the alternative GMM estimators. The interaction between the ODA Dependency Index and institutional quality remains consistently negative but statistically insignificant in both specifications, offering little empirical support for the notion that institutions systematically reduce the growth impact of ODA dependency. Trade openness is insignificant in both specifications, and gross capital formation shows only a weak positive effect under system GMM. The large, significant negative coefficient on the 2020 dummy reflects the severe impact of the COVID-19 pandemic on most Sub-Saharan economies, underscoring the importance of accounting for common time-specific shocks.

4.4. Threshold Results

Because the location of structural breaks proved sensitive to the specification of the ODA Dependency Index, least-squares grid-search threshold estimates are reported in

Table 10. Panel threshold estimates for the aid-growth relationship.

Variable/Statistic	Estimate	Robust SE	t-Statistic	p-Value
Threshold estimate	1.300	-	-	-
95% confidence interval	(−1.200, 1.300)	-	-	-
Bootstrap threshold test: (F_0)	4.690	-	-	0.256
ODAI	−1.045	0.488	−2.14	0.033 **
ODAI	0.038	0.292	0.13	0.897
IQI	5.738	1.091	5.26	<0.001 ***
GCF	−0.041	0.038	−1.09	0.277
TRADE	0.112	0.019	5.97	<0.001 ***
Model statistics
Within (R2)	0.120
Observations (N)	453
Countries	31
Bootstrap replications	500

Source: Authors’ computation, 2026. Notes: ***, ** denote statistical significance at the 1%, 5% levels, respectively.

under both index specifications.

Table 10 Threshold Re-Estimation Summary: Formal Bootstrap Testing under Two Index Specifications (Two-Way Fixed Effects).

The threshold estimates provide limited evidence of nonlinearity in the aid–growth relationship. The estimated threshold occurs at an ODAI value of approximately 1.30, with a 95% confidence interval ranging from −1.20 to 1.30. However, the bootstrap threshold test yields an F-statistic of 4.69 with a p-value of 0.256, indicating that the null hypothesis of no threshold effect cannot be rejected. Thus, although the grid search identifies 1.30 as the best-fitting threshold, the evidence is insufficient to establish a statistically significant regime shift in the aid-growth relationship. Nevertheless, the regime-specific estimates reveal an economically meaningful pattern. Below the estimated threshold, ODA dependence has a negative and statistically significant association with economic growth (β = −1.045, p = 0.033), whereas its effect becomes statistically insignificant above the threshold (β = 0.038, p = 0.897). This suggests that higher aid dependence may be associated with weaker growth when dependence remains below the estimated threshold, but the relationship effectively disappears at higher levels. Institutional quality exerts a positive and highly significant effect on growth (β = 5.738, p < 0.001), while trade openness is also positive and significant (β = 0.112, p < 0.001). Gross capital formation is statistically insignificant.

The findings suggest that aid dependence should be catalytic rather than substitutive, with emphasis on productive investment and domestic resource mobilization. Strong institutional quality and trade openness are associated with higher growth, highlighting the importance of governance, accountability, policy credibility, trade integration, and strategic use of ODA for sustainable development.

4.5. Cross-Sectional-Dependence-Robustness Check: Pooled Common Correlated Effects (CCEP)

The key coefficients are re-estimated after including the cross-sectional averages of the dependent variable and of each regressor as proxies for unobservable common factors in the dynamic specification, using the Pooled Common Correlated Effects (CCEP) estimator [40]. Results are reported in

Table 11. Pooled Common Correlated Effects (CCEP).

Variable	Coefficient	Std. Err.	p-Value
GDP Growth (t−1)	0.127	0.064	0.046 **
ODA Dependency Index	−1.030	0.318	0.001 ***
Institutional Quality Index	2.820	1.374	0.041 **
ODAI × IQI (interaction)	−0.572	0.447	0.201
Gross Capital Formation	0.069	0.042	0.098 *
Trade Openness	0.042	0.020	0.039 **

Source: Authors’ computation, 2026. Note: * p < 0.05, ** p < 0.01, and *** p < 0.001 denote the 5%, 1%, and 0.1% levels of statistical significance, respectively.

(for brevity, the cross-sectional-average terms are omitted). N = 819 country-years, 41 entities, entity fixed effects, and standard errors clustered by country.

The CCEP estimates are consistent with the fixed-effects estimates for all coefficients: the ODA Dependency Index remains negative and statistically significant (−1.030, p = 0.001, versus −1.126), the Institutional Quality Index remains positive and statistically significant (2.820, p = 0.041, versus 3.386), and the interaction term remains negative and statistically insignificant (−0.572, p = 0.201, versus −0.550, p = 0.253). This demonstrates that the significance of the aid-dependency and institutional-quality variables is not merely an artefact of neglecting cross-sectional dependence, and that the lack of a significant interaction term is likewise not an artefact of the baseline specification an estimator explicitly designed to account for CSD produces results consistent with the naive fixed-effects model.

4.6. Cross-Estimator Consistency

Table 12. Cross-estimator consistency of the three headline coefficients.

Estimator	ODA Dependency Index	Institutional Quality Index	ODAI × IQI interaction
Two-way fixed effects (Table 8)	−1.126, p = 0.001 ***	3.386, p = 0.037 **	−0.550, p = 0.253 (n.s.)
System GMM, collapsed + year dummies (Table 9)	−0.369, p = 0.343 (n.s.)	0.458, p = 0.414 (n.s.)	−0.194, p = 0.725 (n.s.)
Difference GMM, collapsed + year dummies (Table 9)	−0.639, p = 0.351 (n.s.)	8.619, p = 0.0005 ***	−0.197, p = 0.863 (n.s.)
CCEP, CSD-robust (Table 11)	−1.030, p = 0.001 ***	2.820, p = 0.041 **	−0.572, p = 0.201 (n.s.)

Source: Authors’ computation, 2026. Note: ** p < 0.01 and *** p < 0.001 denote the 1%, and 0.1% levels of statistical significance, respectively. “n.s” means none significant.

summarizes the three main coefficients; the ODA Dependency Index, the Institutional Quality Index, and their interaction across all four regressions on the extended dataset.

Table 12 demonstrates that the negative association between ODA dependency and growth remains robust under two-way fixed-effects and CCEP estimators but becomes insignificant under dynamic GMM, suggesting that dynamic adjustment and endogeneity partly explain the relationship. Institutional quality consistently exhibits a positive growth effect, although its interaction with ODA dependency remains negative but insignificant across specifications, indicating no significant moderating effect. Nevertheless, the country-level design captures average growth effects and cannot address within-country distributional outcomes or Pareto efficiency. Worthy of note is the imperative to examine who benefits from aid and whether inflows deepen institutional dependence and donor conditionality through institutional and micro-level analyses in future studies.

4.7. Discussion of Findings

This study explored the direct impact of ODA dependency on economic growth, the moderating effect of institutional quality, and the potential nonlinearity of the aid-growth relationship in Sub-Saharan Africa (SSA). The findings consistently show that higher aid dependency is associated with lower economic growth in fixed-effects and cross-sectional-dependence-robust estimations, while institutional quality has an independent positive effect on growth. However, institutional quality’s moderating role in the aid-growth relationship was not significant under any estimator, and threshold estimates were sensitive to index specification, suggesting nonlinearity that should be interpreted cautiously.

The negative relationship between aid dependency and growth is consistent with the concerns raised by Dependency Theory and with empirical evidence from [7,10,18], that long-term dependence on external financing can adversely affect domestic revenue mobilization, policy autonomy, and institutional accountability. The results also align with recent trends in SSA, including the shift in donor funding towards more urgent priorities during the COVID-19 crisis and the Russia-Ukraine war, which highlight the risks of relying heavily on concessional financing and underscore the importance of building fiscal resilience rather than continued reliance on external assistance.

The positive effect of institutional quality reinforces the institutional-economics perspective of [19,20,21], which emphasizes governance, regulatory quality, accountability, and government effectiveness as determinants of long-run economic performance. This finding accords with [31,36], who report positive growth effects of improved institutions in SSA. However, the minimal interaction between aid dependence and institutional quality across estimators suggests that stronger governance does not necessarily translate into sustained growth under aid dependence, consistent with [32,34], who show that structural constraints associated with aid dependence and volatility cannot be fully mitigated through institutional improvement. Methodologically, the two-way fixed-effects estimates are transparent and efficient under strict exogeneity but cannot address reverse causality. Two-step system GMM is suitable for short, unbalanced panels but is vulnerable to instrument proliferation, mitigated here through instrument collapsing, whereas difference GMM avoids the initial condition assumption but may amplify measurement error in persistent series. CCEP accommodates cross-sectional dependence but is less established for dynamic panels. Accordingly, multi-method triangulation distinguishes robust findings from estimator-sensitive results.

The dynamic estimations show that the negative relationship between aid dependence and growth loses statistical significance once endogeneity and growth persistence are accounted for. This does not contradict the fixed-effects results but likely reflects the smaller effective sample size and lower statistical power of dynamic GMM estimation. Importantly, the coefficient remains negative across all specifications, while the significant 2020 dummy captures the sharp regional contraction during the COVID-19 pandemic. The positive and significant persistence coefficient on lagged GDP growth in both GMM specifications (Table 9) confirms that growth is a slow-moving, autocorrelated process within which contemporaneous aid dependence is embedded. Accordingly, static (Table 8) and dynamic (Table 9) specifications are reported jointly rather than relying exclusively on static estimates. Donor allocation also responds to need, including weak growth and fiscal distress, potentially biasing naïve aid–growth estimates toward zero or positive values. Nevertheless, the consistently negative fixed-effects and CCEP coefficients suggest that the documented association is unlikely to be solely driven by need-based aid allocation. A fully specified dynamic model with donor-side instruments remains necessary to disentangle causality and is therefore identified as a priority for future research in Section 5.

The threshold analysis indicates that excessive aid dependency may, in some circumstances, become growth-reducing; however, the estimated breakpoint differs markedly across index specifications. This sensitivity reinforces the caution urged by [14] regarding the use of threshold estimates to guide policy before further validation. Overall, the evidence suggests that sustainable growth in SSA depends less on large increases in aid and more on building domestic investment, institutional capacity, fiscal mobilization, and resilience to external shocks.

5. Conclusions, Policy Recommendations and Limitations

This study examined the relationship between official development assistance dependency (ODA), institutional quality, and economic growth in 48 Sub-Saharan African countries over 2000–2025, using fixed-effects, dynamic GMM, threshold, and cross-sectional-dependence-robust estimators. The evidence is strong and consistent: high levels of aid dependency are associated with lower economic growth, and higher institutional quality is associated with better economic performance. However, there is no clear evidence that institutional quality systematically moderates the growth impact of aid dependency. The threshold analysis indicates that high levels of aid dependency could be harmful to growth, but the estimated breakpoints vary with index specification, so the nonlinear results should be regarded as exploratory rather than conclusive.

These findings suggest that development strategies should prioritize improving domestic resource mobilization, public financial management, productive investment, and trade competitiveness, in order to reduce structural reliance on external assistance. Donors should increasingly link aid to institutional capacity development, fiscal sustainability, and productive sectors capable of generating long-term growth and investment. The threshold estimates warrant confirmation through fully bootstrapped dynamic panel threshold estimation and more comprehensive data sources, and institutional quality should be disaggregated into its individual governance dimensions to better understand how aid can be deployed to support sustainable economic development.

For countries below the threshold in Table 10, where the growth penalty associated with aid is small or statistically indistinguishable from zero, priority should be given to preserving productive uses of concessional finance while progressively strengthening domestic revenue mobilization. Above the estimated threshold, governments should pursue phased reductions in aid-financed budget shares, sequenced against concrete domestic-resource-mobilization milestones to avoid abrupt fiscal adjustment. Donors should complement this transition by shifting toward budget support linked to tax-effort benchmarks and productive-sector programmes rather than balance-of-payments support. However, the insignificant interaction term suggests that generic governance-strengthening measures alone may not neutralize the growth costs of high aid dependency. Moreover, country fixed effects do not fully capture structural heterogeneity between resource-exporting and non-resource economies or conflict-affected and stable states.

The study has several limitations that provide avenues for future research. Tax revenue was excluded from the ODA Dependency Index because of substantial missing observations, despite its theoretical importance. Future studies should employ fully dynamic threshold models and disaggregated institutional-quality indicators, as the present threshold estimates rely on static fixed-effects procedures rather than fully bootstrapped dynamic threshold models. Additionally, measures of international openness and domestic fiscal and monetary policy, including exchange-rate regimes, monetary-policy independence, and broader trade-freedom indices, were excluded because of data-coverage gaps across the 48-country panel for 2000–2025.