The earliest and largest strand in the literature examines the links between economic globalization and growth. Dreher (2006) introduced the first index of globalization (composed of three categories) and observed consistent positive growth outcomes for members. This evidence was corroborated by research on OECD panels (Chang & Lee, 2010), transition economies (Gurgul & Lach, 2014), and Sub-Saharan Africa (Zahonogo, 2018).

Subsequent granular study confirmed this effect in banking sectors (Ghosh, 2017), the insurance industry (Lee, Lee, & Chiou, 2017), via financial and exchange-rate channels (Gaies, Goutte, & Guesmi, 2019; Gaies, Goutte, & Guesmi, 2020), in terms of the proliferation of entrepreneurship (Coulibaly, Erbao, & Metuge, 2018) and increasingly, via digital infrastructure and ICT-enabled FDI (Latif et al., 2018; Ze et al., 2023; Elfaki & Ahmed, 2024).

These impacts are not uniform: outcomes seem dependent upon the quality of institutional arrangements and baseline income levels (Muye & Muye, 2017; Sandanayaka et al., 2025). Furthermore, literature in the African context highlights that these benefits remain conditional on levels of democratic governance and stock of human capital (Glennie Fosah et al., 2023; Baidoo et al., 2023; Radulović & Kostić, 2024).

The second strand decomposes globalization, finding that political globalization, a composite of treaty membership, embassy proliferation, and participation in international peacekeeping and military activities, is often the largest sub-component but is the least adequately conceptualized in terms of causal growth mechanisms (Ehigiamusoe, 2021; Marques, Fuinhas, & Marques, 2017). Social globalization, tracking cross-border migration, tourism, information flows, and cultural convergence, has been shown to interact with labour markets and demographics (Gaddis & Klasen, 2014; Burhan et al., 2015). Research also suggests poverty-alleviating pathways from social globalization, primarily via migrant remittances, diaspora business ties, and broader financial inclusion (Omar & Inaba, 2020; Polloni-Silva et al., 2021; Neaime & Gaysset, 2018).

A novel group of academic papers relates globalization to problems of poverty and food security. Scholars argue that, in poorer nations with a larger volume of food imports than food exports, volatility of commodity prices generated by trade might be expected to lead to aggregate welfare losses for poor families (Ivanic & Martin, 2008; De Hoyos & Medvedev, 2011; Mary, 2019; Ng & Aksoy, 2008; Chikhuri, 2013; Subramaniam, Masron, & Loganathan, 2024).

However, evidence from research points out that export crop agriculture is potentially able to boost rural incomes (Sanjuán-López & Dawson, 2010; Huang et al., 2007; Sawadogo, 2024; Artuc, Porto, & Rijkers, 2021). The literature on labour markets shows significant adjustment costs that fall disproportionately on vulnerable workers (Porto, 2008; Ball, de Roux, & Hofstetter, 2013) and the literature on crisis episodes shows the vulnerability of impoverished households to contagion from global economic shocks (Nikoloski, 2011; Ivanic, Martin, & Zaman, 2012; Laborde, Martin, & Vos, 2021; Cuesta & Pico, 2020; Jara, Montesdeoca, & Tasseva, 2022).

Literature focusing on the African political economy takes a dimmer view of the globalization thesis. This neo-structuralist lens argues Africa remains positioned on the periphery in core-periphery dynamics, so that increased formal participation in global institutions often hides the thin nature of Africa’s industrializing links (Langan, 2018; Bassey & Ayara, 2017; Ibietan, 2010; Dickson, 2021; Ugumanim, Jacob, & Abunimye, 2017). As Arop & Inyang (2017) and Juma (2016) suggest, regional blocs are a proposed remedy-an intermediate scale from which Africa can assertively negotiate with the global core.

Current global winds complicate this: talk of “de-globalization” in a pandemic-racked world (Altman & Bastian, 2024; Autio, Mudambi, & Yoo, 2021), the acceleration of remote-work-based offshoring (Choudhury, 2020; Sytsma, 2022), the rise of economic nationalism (Colantone & Stanig, 2019) and “great power” bifurcation of trade systems (Farrell & Newman, 2020; Aiyar et al., 2023; Gopinath et al., 2024) have massive and sudden implications for how a small open regional bloc can exist.

The gap in this literature is a long-run, comparative, cross-dimensional empirical evaluation of an African regional economic community through this series of acute, overlapping, global shock events.

The study doesn’t utilize a singular framework but rather two lenses that provide distinct analytic capabilities for this phenomenon. Transformationalism frames the outside; Neo-functionalism provides the engine on the inside. Transformationalism, according to Held et al. (1999) and Scholte (2005), proposes that globalization is the restructuring of the sovereign state amid intense trans-border connectivity. States do not disappear (Hyperglobalists) nor do they endure as historical artifacts (Sceptics).

Instead, they are reimagined across tiers. This creates a baseline expectation for the EAC: integration is an existential necessity-as value chains (World Bank, 2019), global norms, technology, and systemic geopolitical risks emerge, and the small and individual state loses control over domestic destiny unless pooled bargaining is employed. Neo-functionalism defines the inside mechanism: Haas (1958) and Schmitter (1969) describe how initial integration creates “spillovers.” Cooperation in sector A (the Customs Union signed in 2005) generates technical/functional hurdles that lead to integration in sector B (the Common Market and aspirational Monetary Union, then political federation), while loyalties from local interest groups/elites gradually migrate upwards.

Combined, this views the EAC as a transformationalist response to global pressure powered by a neo-functionalist engine. Transformationalism predicts the exogenous force/shock, and neo-functionalism describes the transmission path of response. The framework leads to hypothesis formulation: if a functional spillover takes place, then this research expects a strong correlation across all three dimensions and evidence of convergence where “laggard” members or sub-categories catch up with the pack. Conversely, if transformationalist pressures dominate, the evidence would show dramatic synchronous structural impact from exogenous global shocks, such as in 2008-2009 and 2020.

This paper adopts a quantitative, longitudinal, panel design, consistent with the purpose of tracking multidimensional globalization over a longitudinal period, as well as to identify evidence of convergence or divergence between the countries in focus. This study draws on a balanced panel for the five EAC countries (Burundi, Kenya, Rwanda, Tanzania and Uganda) during the years 2006 to 2023, providing a total of N = 90 observations (five countries × 18 years; n = 18 annual observations per country). These five countries constitute a purposive subgroup of the EAC membership, selected for data availability and cross-country comparability over the full study period; the Community also includes the Democratic Republic of the Congo, South Sudan and Somalia, which are not covered here. Although the subgroup is not a probability sample, this does not dispense with inferential caution: composite indices embody measurement and modelling choices, and coverage of the five countries eliminates neither measurement nor model uncertainty. All findings are accordingly interpreted as applying to the five study countries rather than to the EAC as a whole.

Data for four variables were extracted from The Global Economy.com, which republishes the revised KOF Globalisation Index produced by the KOF Swiss Economic Institute (0-100 scale): Globalization Index (GI) overall; Economic (EGI), Political (PGI) and Social (SGI) Globalizations.

The GI is higher when the country is more integrated (The Global Economy, 2023). This KOF-type composite instrument is considered standard and robust in such studies (see Dreher, 2006). The study period (2006-2023) is useful for theoretical leverage as it traverses milestones like the consolidation of the EAC Customs Union, EAC Common Market Protocol, the global financial crisis of 2008-2009, the COVID-19 pandemic and, recently, trends toward geo-economic fragmentation.

The research uses existing secondary measures based on observed flows and restrictions in the KOF style. EGI is based on data for trade, FDI and portfolio investment and (de jure) capital-account openness. PGI is an aggregation of the number of embassies in a country, the number of international organizations in which a country is a member, and the number of international treaties signed; SGI aggregates flows of personal contacts (migration, tourism, international telephone traffic), information flows, and cultural proximity.

As the instrument has a large body of peer-reviewed applications (see references in Dreher, 2006 and the references reviewed in Section 2 above), the measures have satisfactory construct validity for the study purpose.

The reliability of the indices is based on the robust nature of their properties and the documented and accessible guide to construct them such that any researcher can precisely replicate (Gygli, Haelg, Potrafke, & Sturm, 2019). Reproducibility statement. The four series analysed are the overall KOF Globalisation Index and its economic, political and social sub-indices in their revised form (Gygli et al., 2019), obtained from TheGlobalEconomy.com, which republishes the index produced by the KOF Swiss Economic Institute. Under the revised KOF methodology, the economic, social, and political dimensions are aggregated into the overall index with equal weights of one-third (Gygli et al., 2019); this weighting scheme is corroborated internally by the decomposition coefficients of approximately 0.33 reported in Section 4.2, which recover the construction weights of the composite. The raw series are publicly and persistently available from the KOF Swiss Economic Institute (https://kof.ethz.ch/en/forecasts-and-indicators/indicators/kof-globalisation-index.html) and from TheGlobalEconomy.com (https://www.theglobaleconomy.com/); the extracted spreadsheet and the Python analysis code used to produce all statistics and figures are available from the corresponding author upon reasonable request, allowing full replication of the analyses.

The study results are divided into four main parts corresponding to these four objectives. Regarding (i) description of dynamics, calculating the arithmetic mean, standard deviation, and compound annual growth rate (CAGR) by country and by dimension is adopted in this section:

For Objective (ii), convergence, sigma-convergence is assessed from the cross-sectional standard deviation in each year:

and beta-convergence regresses average annual log growth on the initial level:

where g i = [ ln ( X i , 2023 ) − ln ( X i , 2006 ) ] / 17 ; β < 0 implies catch-up. For Objective (iii), co-movement, Pearson correlations and panel regressions estimate:

estimation proceeded by 3 methods: pooled OLS; fixed effects (FE) using country and year dummies, with cluster-robust standard errors (CRSE) at the country level; and a first-difference (FD) “growth” specification, in an attempt to address any spurious correlation deriving from joint non-stationarity (common trends) (Granger, 1969; Pesaran, 2007; Pesaran & Yamagata, 2008; Breitung & Das, 2005; Love & Zicchino, 2006). It must be kept in mind that GI is a weighted summation of its inputs and, as such, (4) ought technically to be considered purely a consistency decomposition of the GI. Coefficients are expected to be roughly equal to the weights used to construct the index and should not be interpreted as causal estimates. Future work could pursue causal estimation using instruments or system-GMM (Roodman, 2009; Bun & Kiviet, 2006).

Future work could also potentially use time-series causal tools like variants of Granger-causality (Ren et al., 2024; Yuan & Qin, 2012), though they are not used here as T = 18 is likely too short for valid time-series causal inference. Concerning Obj. (iv), results are interpreted within the overarching conceptual framework derived in Sec. 2.4. Python 3.12 (pandas 2.x, statsmodels, matplotlib) was used for analysis, with plots exported at 500 dpi.

This study relies exclusively on secondary, aggregated, and publicly available macroeconomic data and does not contain personal information, such that the use of human beings or primary data collection is not applicable in this study; hence, informed consent or prior ethics committee approval for this research is unnecessary. However, the study still follows ethically relevant standards, including honesty, clear methodology, and full data availability (cf. Sect. 7).

The descriptive statistics of the total panel are contained in Table 1 and reveal that the most globalized country is Kenya, with an average Globalization Index (GI) equal to 55.27, followed by Uganda with 52.62, Tanzania with 50.23, Rwanda with 48.10, and finally Burundi with 38.82. In general, it seems that the political aspect is the most globalized (five-country mean, 67.82), whilst the social one is the least globalized dimension (five-country mean, 37.96) and the economic one is an intermediate dimension (five-country mean, 41.22). Finally, Kenya has the highest average with the lowest dispersion, whilst Rwanda has the highest variability, which could imply an accelerating structural change.

Table 1.Descriptive statistics of globalization indices by country, 2006-2023 (total N = 90 country-year observations; n = 18 observations per country).

Source: TheGlobalEconomy.com; author’s computations. GI = overall; EGI = economic; PGI = political; SGI = social globalization index (0 - 100). Each country contributes n = 18 annual observations (2006-2023); the total panel comprises N = 90 country-year observations.

The five series show a general trend of rising to mid-decade peaks: Kenya at 57.04 (2014), Tanzania at 52.52 (2014), Burundi at 41.39 (2015), Rwanda at 52.39, and Uganda at 54.23 (2017), and then declining around 2020, coinciding with the COVID-19 pandemic, and partially recovering thereafter (see Figure 1).

Table 2 quantifies the dynamics. Rwanda is the clear leader in overall growth (ΔGI = +12.98 points; CAGR 1.76%), driven by an extraordinary economic opening (ΔEGI = +20.29; CAGR 3.11%). Burundi also converges from a low base (+9.34 points). By contrast, Tanzania’s economic globalization contracted sharply (ΔEGI = −10.80; CAGR −1.61%), leaving its overall index flat (−0.36 points), while Kenya and Uganda advanced only marginally overall despite strong social gains (SGI +9.38 and +7.16, respectively).

Table 2.Globalization levels and dynamics by country and dimension, 2006-2023.

Source: TheGlobalEconomy.com; author’s computations. CAGR computed per Equation (1).

The identical information, grouped by dimension, is shown in Figure 2 and Figure 3. The community-wide picture seems quite different. Political globalization is high and stable; social globalization is low but rising everywhere; economic globalization is variable and may explain the bulk of cross-country variation in outcomes.

The 2023 set reveals a regional bloc having relatively higher levels of political openness, combined with relatively lower levels of economic and social openness. Broadly speaking, the same qualitative hierarchy for the three dimensions seems to hold for the bloc’s members in that political openness appears deeper than both economic and social openness. This point signifies the convergence in the three dimensions’ development levels across the various bloc countries, although the level of intensity may have differed, as shown in Figure 4.

Table 3 reports pooled Pearson correlations. Overall globalization correlates strongly with each component PGI (r = 0.883), SGI (r = 0.890), EGI (r = 0.757), while the inter-component correlations are moderate to strong (0.437 - 0.708). Figure 5 visualizes the matrix. Because these correlations are computed over pooled country-years, they conflate persistent between-country differences with within-country co-movement over time and must be read as associations, not as evidence of temporal co-movement. Temporal co-movement is assessed separately through the first-difference specification reported in Table 4, which differences out persistent country characteristics; the coefficients of that specification likewise recover the construction weights of the composite rather than behavioural elasticities.

Table 3. Pearson correlation matrix of globalization indices (pooled panel, N = 90).

Source: Author’s computations. All correlations are significant at p < 0.001.

The composite index appears to be almost perfectly balanced, with a weighting for each component of about one-third. Specifically, the pooled OLS coefficients for EGI, PGI and SGI are 0.330, 0.336 and 0.336 respectively (R² = 0.9997). This pattern holds up when allowing for country and year fixed effects and when transforming to first differences. Also consistent with treating the composite as a monolithic aggregate, all of the intercept terms are statistically insignificant (See Table 4). These near-equal coefficients are mechanical: because the GI is constructed as an equal-weighted average of its three dimension indices, each dimension receiving a weight of one third (Gygli et al., 2019), regression coefficients of approximately one third simply recover the construction weights of the composite. The decomposition is thus an accounting consistency check and can establish neither spillovers among the dimensions nor equal substantive contributions of the dimensions to globalization outcomes.

In the growth models, the effects on growth in the globalization index of a one-point increase in the economic, political or social indices are 0.32, 0.34, and 0.34 index points, respectively (all p < 0.001, R² = 0.9831). These first-difference coefficients are likewise recovered construction weights rather than estimated behavioural effects. Co-movement of the three dimensions in first differences is consistent with a coupled-spillover interpretation but does not constitute evidence for it; the accounting identity underlying Equation (4) and the high multicollinearity among components preclude disentangling separate dimensional effects within this design.

Table 4.Panel estimates of the composite-index decomposition (dependent variable: GI).

Source: Author’s computations. t/z statistics in parentheses; Note: ***p < 0.001; (n.s.) = not significant. Fixed-effects model uses cluster-robust standard errors by country. First differences: N = 85. Coefficients are interpreted as consistency weights of the composite, not causal effects (Section 3.3).

The process of sigma-convergence (visible in Figure 6) seems material. Between 2006 and 2023, cross-country standard deviation in GI fell by around 40% (from 8.53 to 5.16), while standard deviation in political globalization was cut by nearly 50% (from 14.83 to 8.90). Also, in Table 5/Figure 6, average growth from 2006 to 2023 is negatively and significantly correlated with the starting value (β = −0.084; t = −3.33; p = 0.045; R² = 0.787).

The least globalized members in 2006, Rwanda (followed by Burundi), grew fastest (+1.76% and +1.53% per year, respectively); Kenya, Tanzania, and Uganda, relatively globalized at the start, stagnated or grew more slowly.

Taken together, and subject to the small-sample caveats reported below, these diagnostics are consistent with convergence within the subgroup, although they cannot rule out alternative explanations such as mean reversion or common shocks. They are reported alongside the cautionary results, the statistically insignificant intercepts in Table 4 and the small-sample warning in Table 5, in the interest of transparency rather than suppression.

Table 5.Beta-convergence regression (dependent variable: average annual growth of GI, 2006-2023, %; Equation (3)).

Source: Author’s computations. R² = 0.787; n = 5 (one observation per member state); intercept and slope jointly significant at the 5% level. Coefficients per Equation (3).

Small-sample robustness. The beta-convergence regression is estimated on only five country-level observations, and the panel significance tests reported in Table 4 rely on cluster-robust standard errors with only G = 5 country clusters; with so few clusters, cluster-robust inference is unreliable and may overstate precision (Cameron, Gelbach, & Miller, 2008). As a sensitivity check, Equation (3) was re-estimated omitting each country in turn (recomputed from the endpoint values reported in Table 2). The beta coefficient remains negative in all five leave-one-out runs, with point estimates ranging from −0.070 (omitting Rwanda) to −0.116 (omitting Burundi); the baseline estimate (β = −0.084) is therefore not driven by any single member state. However, none of the leave-one-out estimates is statistically significant at the 5% level (p-values between 0.075 and 0.127), and only four of the five are significant at the 10% level, confirming that formal significance is fragile at n = 5. Given the five-country cross-section, neither bootstrap refinements nor cluster-robust corrections can restore formal inference; the convergence diagnostics are accordingly read as suggestive rather than definitive.