Despite the proven relationship between road infrastructure, trade openness and economic growth, SADC countries face persistent challenges in leveraging these synergies. The region’s road infrastructure varies widely in quality, with many remote areas struggling to access markets because of poor connectivity (Blyde 2013). This disparity hinders the flow of goods and services, exacerbating economic inequalities and slowing progress towards regional integration.

Trade openness in SADC remains limited, with intraregional trade accounting for a small proportion of overall trade activities. This is despite evidence that enhanced trade openness significantly contributes to economic growth in both developed and developing regions (Udeagha & Ngepah 2020). Additionally, factors such as population growth, inflation and limited government spending on infrastructure complicate the relationship between road quality and economic growth, necessitating a nuanced understanding of how these variables interact.

This study addresses the pressing need for a conceptual framework to explore the relationships between road quality, trade openness and economic growth in the SADC. By identifying key mediating and moderating variables and synthesising insights from empirical and theoretical literature, this research aims to provide actionable recommendations for policymakers to optimise road infrastructure investments and achieve regional economic integration.

The study is guided by the following specific objectives:

This study adopts a panel data regression framework to analyse the relationship between road quality, trade openness and economic growth in the SADC region, focusing on Botswana, Mozambique, South Africa and Zimbabwe over the period of 2010–2020. The other SADC member states were not included because of the unavailability of adequate and consistent data for the variables of interest over the study period. Furthermore, even for the selected countries, the study period could have been extended, but data availability limited the analysis to the years 2010–2020. Panel data regression is chosen for its capacity to handle both cross-sectional and temporal variations, enhancing the robustness of estimates by accounting for unobserved heterogeneity (Wooldridge 2020). The analysis includes fixed effects (FE) and random effects (RE) models, with the Hausman test employed to select the appropriate specification.

Data are sourced from reliable databases, such as the World Development Indicators (WDI) and the Global Economy Database, ensuring comprehensive coverage of variables like road quality, trade openness, economic growth and key control variables – population growth, inflation, government spending, employment and FDI. This dataset allows for an empirical assessment of contemporary trends in infrastructure and economic performance in the SADC region.

The key variables under study are operationalised as follows.

Panel data regression is applied to evaluate two key relationships: (1) the effect of road quality on trade openness and (2) the effect of trade openness on economic growth. Control variables are included in both models to account for confounding factors. Fixed effects (FE) and RE models are estimated, with the Hausman test determining the preferred approach for each analysis.

This approach allows for the examination of within-country and between-country variations, aligning with established methodologies for longitudinal and cross-sectional analysis in economic studies (Baltagi 2021). Diagnostic checks, including tests for multicollinearity and heteroskedasticity, ensure the reliability of findings.

While the models in this study focus on estimating the association between trade openness and economic growth, the potential for reverse causality – where higher economic growth could also lead to increased trade openness is acknowledged. The analysis mitigates this concern to some extent by incorporating multiple macroeconomic control variables and exploiting the panel structure of the data, which captures both cross-country and temporal variation. However, the primary aim is to establish the strength and direction of relationships within the SADC context rather than to claim strict causality. The findings should therefore be interpreted as indicative of associations under the specified model, with causality subject to further investigation in future studies using advanced techniques, such as instrumental variables or dynamic panel estimators.

All modelling and diagnostics were carried out in Stata, version 17 (StataCorp LLC, College Station, Texas, United States [US]). We estimated fixed- and random-effects models with xtreg and used post-estimation diagnostics, including the Breusch-Pagan test for heteroskedasticity and the Information Matrix (IM) decomposition. The analyses are conducted using STATA 17 in order to leverage its advanced capabilities for panel data regression. This ensures precision in handling complex models and the computation of robust statistical inferences.

By applying panel data regression with control variables, this study contributes to the understanding of how road infrastructure quality influences trade openness and economic growth in the SADC region, aligning with its objective to inform strategies for regional economic integration.

Ethical clearance to conduct this study was obtained from the University of the Witwatersrand, Johannesburg, Human Research Ethics Committee (Non-Medical) (No. H23/03/22).

The regression analysis evaluates the relationship between RQ and TO, incorporating key variables, such as PG, INF, GS, EMP and FDI. Fixed-effects (Figure 1) and RE (Figure 2) models are employed, with the Hausman test (Table 1) determining the appropriate specification.

In the fixed-effects model, the within R² value of 0.247 indicates that 24.7% of the variability in TO is explained by the predictors within countries over time. Population growth (PG) emerges as the sole significant variable (p = 0.045), suggesting that higher population growth is associated with decreased trade openness. This aligns with findings by Udeagha and Ngepah (2020), who emphasised the complex interplay of demographic factors and trade dynamics in Africa. Other variables, including RQ, are statistically insignificant, reflecting potential limitations in the model or the need for more granular data.

The random-effects model, with an overall R² of 0.7106, provides stronger explanatory power for cross-country variations. Government spending (GS) significantly predicts TO (p = 0.000), indicating its role in enhancing trade openness, consistent with literature that highlights infrastructure investment as a catalyst for trade (Duernecker et al. 2022). FDI approaches significance (p = 0.093), resonating with Ahmad et al. (2020), who identified FDI as critical in integrating economies into global trade networks. However, RQ remains insignificant, suggesting potential indirect or moderated effects rather than direct impacts on TO.

The Hausman test (χ²(6) = 14.97, p = 0.0205) indicates that the fixed-effects model is preferred, emphasising the importance of within-entity variations. This result underscores that trade openness in the SADC region is influenced by domestic factors, such as population trends and policy implementation. Recent studies, including Esaku (2021), support the view that trade openness benefits from tailored domestic policies that address country-specific barriers. Within the SADC framework, initiatives such as the SADC Protocol on Trade and the Industrialisation Strategy and Roadmap (2015–2063) aim to enhance intra-regional trade by reducing tariff and non-tariff barriers, harmonising customs procedures and promoting cross-border infrastructure development. These policies, when effectively implemented at the national level, can strengthen the link between openness and economic growth by addressing structural constraints that vary across member states.

For the model specification tests, the following results were found (Table 2):

The Breusch–Pagan/Cook–Weisberg test for heteroskedasticity produced χ²(1) = 0.69 with (probability) Prob > χ² = 0.4076. At the 5% level, the null hypothesis of constant variance could not be rejected, indicating no evidence of heteroskedasticity in the regression residuals. Cameron and Trivedi’s IM decomposition confirmed this, with the heteroskedasticity component yielding p = 0.2609.

Cameron and Trivedi’s decomposition also examined skewness (p = 0.5396) and kurtosis (p = 0.3038), with neither being statistically significant. These results indicate that residuals did not exhibit abnormal skewness or kurtosis, supporting the normality assumption (Table 3). The Ramsey RESET test returned F (3, 30) = 1.96, p = 0.1405, suggesting no strong evidence of omitted non-linear terms or functional form misspecification.

Variance inflation factors (VIF) indicated that while some predictors, such as road quality (10.63) and employment (9.70), were on the higher end, the mean VIF was 5.44, which is within an acceptable range. This indicates that multicollinearity was not a severe concern for the overall model Refer to Table 4.

Effect size analysis showed that the model explained a substantial portion of the variance in the dependent variable, with an overall eta-squared of 0.7106. Government spending exhibited the largest partial eta-squared (0.5298), indicating it was a strong predictor, while other variables contributed smaller but meaningful proportions to the explained variance.

The regression analysis examines the effect of TO on EG using a panel dataset from four SADC countries (Botswana, Mozambique, South Africa and Zimbabwe) between 2010 and 2020. Two models were evaluated: FE (Figure 3) and RE (Figure 4), with the Hausman test (Table 5) determining the most suitable specification.

The FE model indicates that TO has a positive but statistically insignificant effect on EG (p = 0.121), while inflation (INF, p = 0.025), employment (EMP, p = 0.011) and population growth (PG, p = 0.033) significantly affect EG. Specifically, higher inflation and population growth negatively influence EG, while employment positively contributes to economic outcomes. These findings align with Udeagha and Ngepah (2020), who argue that demographic and macroeconomic factors play crucial roles in shaping trade-related economic growth.

The RE model corroborates some of the FE findings, showing significant positive effects of TO (p = 0.004) and employment (p = 0.005) on EG. Conversely, inflation (p = 0.001) and population growth (p = 0.020) exert negative impacts, consistent with the literature linking inflation to economic instability and population pressures to strained resources (Esaku 2021; Jalil & Rauf 2020) (Table 5). Government spending (GS) and FDI remain insignificant in both models, suggesting that their roles may be context dependent or moderated by other factors, as discussed by Tosunoğlu (2023).

The Hausman test (χ²(6) = 7.32, p = 0.2927) fails to reject the null hypothesis, indicating no systematic difference between the FE and RE models. Thus, the RE model is preferred for its efficiency. The RE results, particularly the positive relationship between TO and EG, echo findings from global and regional studies (Khan et al. 2020), which highlight trade openness as a driver of economic performance when coupled with conducive policy environments.

The specification tests run for the second model also produced ideal results as follows (Table 6):

The Breusch–Pagan/Cook–Weisberg test for heteroskedasticity returned χ²(1) = 1.09 with Prob > χ² = 0.2956, meaning the null hypothesis of constant variance could not be rejected (Table 7). This indicates no evidence of heteroskedasticity in the residuals.

Cameron and Trivedi’s IM-test confirmed this result, with the heteroskedasticity component producing p = 0.3005. The skewness (p = 0.0722) and kurtosis (p = 0.3349) components were also not statistically significant, suggesting that the residuals followed an approximately normal distribution without severe departures in symmetry or tail behaviour.

The VIF diagnostics showed that all predictors had VIF values below 6, with a mean VIF of 3.53, well within conventional thresholds for concern. This indicates that multicollinearity was not a major issue and that the predictors could be reliably interpreted in the model.

Effect size analysis revealed that the model explained 48.86% of the variation in economic growth (eta-squared = 0.4886). Trade openness (eta-squared = 0.1868), inflation (eta-squared = 0.2161) and employment (eta-squared = 0.1775) emerged as the most influential predictors, indicating that these variables contributed meaningfully to explaining variation in economic growth across the sample (Table 8).