South Africa plays a dominant role in the Southern Africa region both on the political and economic front. Economically, South Africa accounts for approximately 50% of the region’s output (African Development Bank [AfDB] 2018b). Since the onset of the democratic dispensation in 1994 following the collapse of apartheid regime, the developmental policies and programmes, from the Reconstruction and Development Programme (RDP) to the current NDP (Vision 2030), have focussed on improving economic growth and alleviation of poverty and inequalities (Sulla & Zikhali 2018). However, achieving such goals has become difficult given the low rate of economic growth over the last 10 years.

The country has recorded the slowest rate of growth since the end of 2007/2008 global financial crisis compared to the decades before the crisis. For instance, between 1994–2000 and 2001–2008, South African economy expanded by an average of 2.9% and 4.2%, respectively. By contrast, the average growth expanded by 1.7% over the period of 2010–2019. The sluggish economic growth over these periods can be attributed to several economic events such as global financial crisis, Eurozone debt crisis, commodity market crash and Brexit. As shown in Figure 1, South Africa has experienced several periods of negative economic growth, particularly during the early 1990s, 2009 and 2020. The economy was in a recession from 1990 to 1992, largely in response to the effect of international sanctions on the apartheid regime (The Presidency 2014; World Bank 2020). The periods 2009 and 2020 coincided with the global financial crisis and COVID-19 pandemic. In a nutshell, South Africa’s economic growth has been consistently below the 6.1% target of the Growth Employment and Redistribution (GEAR) strategy introduced in 1996.

Bonga-Bonga and Guma (2017) argue that savings are crucial for economic growth in South Africa. Unfortunately, South Africa has experienced a decline in the domestic savings over the 1990–2019 period. For instance, the average gross domestic saving (GDS) grew from 11.8% between 1994 and 2000 to 13.9% over the period 2001–2008, before declining to 5.1% during the period 2010–2019. This rate of domestic saving is less than the target rate of 16% – 25% set in the NDP (South African Government 2012). As a result, a study by the National Treasury (2019) conceded that South Africa’s low domestic savings are likely to constrain the growth of the economy. As a proportion of GDP, the GDS accounts for approximately 17.7% and 20.1% of the GDP over the period 1994–2000 and 2001–2008, respectively. From 2010 to 2019, the share of GDS declined to approximately 17.5%.

Since its independence in 1966, Botswana has experienced rapid economic development owing to its diverse mineral wealth, prudent economic management and good governance (World Bank 2022). While Botswana has historically experienced one of the highest economic growth rates in world, the rate has slowed in recent years (International Trade Administration [ITA] 2022). For instance, from 1990 to 1999, the economy expanded by an average of 5.4% compared to average growth of 4.2% over the 2000–2008. After experiencing a negative growth of -14.1% during the global financial crisis in 2009, the economy expanded by 10.1% in 2010. However, the average growth rate of 4.7% over the period of 2010–2019 was lower compared to the period 1990–1999.

The high dependence on diamond export which accounts for over 80% of total export, and public sector-driven model make the economy vulnerable to external shocks (World Bank 2022). In addition, diamond export revenue and custom revenue from SACU account for about two-third of the total revenue, implying a large share of the country’s revenue falls outside its control (ITA 2022). As a result, the negative economic growth experienced over the period of 2012, 2015 and 2020 indicates the exposure of the economy to external shocks. Moreover, these periods were associated with different external economic events (including the Eurozone sovereign debt crisis, commodity market collapse and COVID-19 pandemic) that affected commodity export.

Given the vulnerability of the country to external shocks, policy makers would be under pressure to plan conservatively, and savings is seen as one of the policy measures to mitigate this problem. While Botswana has maintained a higher domestic saving compared to other SACU members over the years, the growth rate of savings has remained relatively stable over the periods of 1990–1999, 2000–2008 and 2010–2019, averaging 16.0%, 9.8% and 16.1%, respectively. As a percentage of the GDP, the average rate of growth increased from 36.7% during the period of 1990–1999 to 39.2% between 2000 and 2008, before declining to 26.4% over the period of 2010–2019 (see Figure 2).

The Kingdom of Lesotho is a small country surrounded by the Republic of South Africa, and is classified as a low-income developing country (World Bank 2020). It is an open economy, with imports accounting for over 90% of GDP, of which around 80% originate from South Africa (Ministry of Finance and Development Planning 2016). The government of Lesotho depends on external sources of revenue, such as receipts from the customs union (SACU), as well as royalties from transfers of water resources to South Africa. Additionally, households rely heavily on remittances from family members working in South African mines, farms and domestic workers (AfDB 2018b). The country’s ability to sustain strong economic growth is closely linked to the economic and political dynamics in South Africa and other members of Southern African Development Community as well as weather conditions (United Nations Conference on Trade and Development [UNCTAD] 2013).

In recent decades, Lesotho has experienced declining economic performance from average growth of 4.4% over the period of 1990–1999 to 3.5% during the period of 2000–2008 and further to 2.4% during the period of 2010–2019. The declining economic performance towards the early 2000s has been attributed to drop in Lesotho Highland Water Project investment, weak growth in agriculture and manufacturing as well as declining migrant workers remittance from South Africa (World Bank 2002). The weak economic performance over the period 2010–2019 has been attributed to weak growth in South Africa, lower global economic prospects and effect of drought (UNCTAD 2013). The fact that Lesotho economic growth path is linked to external environment, the role of domestic savings will be critical in ensuring stable economic growth. Although the proportion of domestic savings as a percentage of GDP was negative over the period 1990–2019, it has progressively improved from -66.3% in 1990 to -16.0% in 2019. The fluctuations experienced in the economic performance and savings over the period of 1990–2021 are presented in Figure 3.

Namibia is a small open economy, sharing borders with Angola, Botswana, South Africa and Zambia (AfDB 2018a). The economy is closely linked to South African economy through several institutional relationships including membership of SACU and Common Monetary Area (CMA) as well as through trade, investment and common monetary policies (International Monetary Fund [IMF] 1997). Since independence in 1990, the major goal of the government is to reduce the legacy of apartheid regime through sustained economic growth, employment creation, reducing inequality and poverty. These socio-economic challenges are addressed through various 5-year national development planning strategies which consist of both medium- and long-term development perspectives (National Planning Commission 2004). The adoption of various National Development Plans (NDP 1–5) together with the implementation of Namibia Vision 2030 have assisted in addressing some of these issues.

However, the economic performance in Namibia has continued to be subject to external shocks and erratic growth since the independence. The real GDP grew at an average rate of 3.5% annually over the period of 1990–1999. This period was marked with the implementation of the First National Development Plan (NDP1). Thereafter, the economy expanded by an average of 4.8% annually during the period of 2000–2008 following the implementation of the NDP2, before contracting to 3.1% during the 2010–2019 period. While Namibia weathered the effect of the global financial crisis compared to most SACU countries, the economy contracted to -1.0% in 2017 and -8.0% in 2020 because of uncertainties around Brexit and COVID-19 pandemic. In terms of domestic savings as a proportion of GDP, the share of saving increased from 16.0% over the period 1990–1999, to 17.8% during the 2001–2008 period, and declined to 6.5% during the period of 2010–2019. Figure 4 presents the trends in the economic growth and domestic savings in Namibia over the period.

The Kingdom of Eswatini formerly known as Swaziland gained its independence in 1968. The country is classified as a lower middle-income country according to the World Bank (2020). It is a small open economy with close economic ties to South Africa, which receives 60% of its exports and from which it receives over 80% of its imports (AfDB 2020). In addition, it is a member of CMA with South Africa, Lesotho and Namibia, and its domestic currency is pegged to the South African rand. Furthermore, the economy is dependent on workers remittances and customs union revenue from SACU (Moody’s Analytics 2023). The overarching development goal of the government is to maintain a sustainable and inclusive economic growth that is resilient to shocks through structural transformation (Central Bank of Eswatini 2020). The major policy framework adopted by the government to address the socio-economic challenges facing the country is the National Development Strategy (NDS, 1997–2022). The implementation of NDS is carried out through the medium-term programmes such as the Economic Social Reform Agenda (ESRA), the Millennium Action Programme (MAP), the Smart Programme for Economic Development (SPEED), and the Poverty Reduction Strategy and Action Plan (PRSP) (Ministry of Economic Planning and Development [MEPD] 2006).

Although there has been a shift in the economic structure of the country from primary to secondary production, the manufacturing sector particularly the sugar and soft drink concentrate are the leading export industries (Moody’s Analytics 2023). However, the rate of economic growth has declined over the years. From 1990 to 1999, the economy expanded at an average of 4.9% per annum but contracted to 3.8% during the period of 2001–2008, and further to 2.7% during the period of 2010–2019. The recent poor economic performance is partly attributed to low agricultural productivity, a slowdown in manufacturing output, declining foreign direct investment inflows and drought (MEPD 2019). The domestic savings rate is very low, averaging about 2.0% of the GDP from 1990 to 1999 before rising to 14.1% during 2000–2008, and falling to 9.5% during 2010–2019. Figure 5 shows the growth and savings trend over the period 1990–2021.

This article utilised a balanced panel data over the period 1990–2021 to investigate the savings-growth nexus among five SACU member countries (South Africa, Botswana, Lesotho, Namibia and Eswatini). The data were obtained from the World Bank Development Indicators, African Development Bank Socioeconomic database and African statistical yearbooks. The key variables of interest are GDP which is used as a proxy for economic growth and GDS as a percentage of GDP. Following previous studies, the control variables included in the analysis are: the gross fixed capital formation (GFCF) to GDP a proxy for investment; the Consumer Price Index (CPI) as a proxy for inflation; and Trade openness (TRADE), which is the ratio of total exports plus imports of goods and services to GDP (Joshi et al. 2019; Mohanty 2018; Nguyen & Nguyen 2017; Rahman & Ferdaus 2021; Sujianto et al. 2020). It is important to highlight that panel data were transformed to natural logarithm to allow the estimates to be analysed as elasticity coefficients.

Table 1 provides both the descriptive statistical properties (Panel A) and correlation analysis (Panel B). The data in Panel A show that the average economic growth in SACU countries is about 3.2%, while the ratio of domestic savings to GDP is approximately 7.6% over the period. This finding shows that the current economic growth is well-below the target rate of 6% GDP growth. The rate of investment in SACU as proxied by the GFCF to GDP, indicates that investment accounts on average about 21.0% of GDP. Looking at the skewness coefficients, it is shown that all the variables are positively skewed except domestic saving. These findings suggest that domestic savings have declined or that there is a high level of dissaving in SACU countries over the period.

The data in Panel B show that there is a positive correlation between domestic savings and economic growth with a correlation efficient of 7.0%. Such a positive correlation implies that domestic savings and GDP move in the same direction, and that an increase in domestic savings is more likely to lead to an increase in economic growth. In addition, it suggests the likelihood that domestic savings and economic growth could share long run relationship. Contrary to the expectation, the correlation between investment and GDP is found to be negative, implying de-anchoring or divergence between investment and economic growth in SACU countries. However, TRADE and CPI have the expected signs and are consistent with economic theory.

Following Verhoef et al. (2014), the econometric model used in this article is specified as: GDPit=α+βiGDSit+βGFCFit+β3TRADEit+β4CPIit+uit[Eqn 1] where α represents constant term while the parameters β₁, β₂, β₃ and β₄ are coefficients of the respective explanatory variables. The subscript i denotes individual SACU member country with i = 1, 2, 3, 4 and 5. The time subscript t is the time dimension in years and u_it indicates the composite error term and is independent and identically distributed normally (U ~ NIID,0,δ²)

To analyse the relationship between savings and economic growth as well as the effect of saving on economic growth, this article applied several complementary panel estimation techniques to Equation 1. Specifically, the article first utilised the combination of panel techniques namely the panel least square (POLS), fixed effect (FE) and random effect (RE), fully modified ordinary least square (FMOLS) and dynamic ordinary least square (DOLS). Apart from the above panel techniques, this article follows recent development in the literature where it is now a common practice to distinguish the short run impacts from the long run impacts (Agu & Omolade 2021; Bolarinwa & Obembe 2017; Gatsi & Appiah 2020; Joshi et al. 2019; Keho 2018; Kok & Munir 2021; Miah & Majumder 2020; Mohanty 2018). For this reason, the article applies the panel ARDL model proposed by Pesaran, Shin and Smith (1999). The panel ARDL model for the long run and the short run relationship between saving and economic growth can be obtained by transforming Equation 1 as: ΔGDPit=ωi+Σi=1pα1iΔGDPit−1  +Σi=1qα2iΔGDSit−1+Σi=1qα3iΔGFCFit−1  +Σi=1qα4iΔTRADEit−1+Σi=1qα5iΔCPIit−1+  +π1GDPit−1+π2GDSit−1+π3GFCFit−1  +π4TRADEit−1+π5CPIit−1+εit[Eqn 2]

In Equation 2, all the variables are allowed to be integrated level I(0), first difference I(1), or the combination of I(0) and I(1), respectively. The short run parameters are α_1i – α_5i while the long run parameters are π₁ – π₅. The constant is ω₁ and Δ represents first difference operator. Finally, p, q are the optimal lag orders which are determined using the Schwarz Information Criterion (SIC). The model is flexible such that time, trends and other fixed regressors may be included (Cameron & Trivedi 2010). The cointegration test for long-run relation is examined by testing the null hypothesis of no cointegration such that π₁ = π₂ = π₃ = π₄ = π₅ = 0 against the alternative that there is cointegration, π₁ ≠ π₂ ≠ π₃ ≠ π₄ ≠ π₅ ≠ 0 However, it is important to point out that the application of the above hypothesis in a panel setting is difficult. For this reason, the Pedroni (2004) cointegration test is mostly used by studies to determine the long-run relationship (Ekanayake & Dissanayake 2022). Therefore, cointegration between savings and economic growth is tested in this article using the Pedroni (2004) cointegration test.

In order to determine the short-run relationship between savings and economic growth, Equation 2 can be reformulated into the error correction model (ECM) representation. However, it is important to note that the estimation of the ECM is dependent on presence of a long-run relationship. The ECM for this article is expressed as: ΔGDPit=α01+Σi=1pα1iΔGDPit−1+Σi=1qα2iΔGDSit−1  +Σi=1qα3iΔGFCFit−1  +Σi=1qα4iΔTRADEit−1  +Σi=1qα5iΔCPIit−1+λECTit+εit[Eqn 3]

The parameter ECT_it is the error correction term. The validity of the error correction term depends on the magnitude and the sign of the coefficient λ, representing the speed of adjustment. Theoretically, the coefficient of the error correction term is expected to be a negative value. Such a negative value indicates how long it takes for shocks to work through the system.

The modified ARDL specification allows for a dynamic structure wherein three different estimators, namely the pool mean group (PMG), mean group (MG) and dynamic fixed effect (DFE), can be estimated by Equation 2 and Equation 3. Importantly, all the three estimators consider the long-run and short-run relationship and the heterogeneity of the dynamic adjustment process. However, they differ in several ways because of the underlying assumptions. For example, DFE estimator assumes that data for each group are pooled but allows the intercepts to differ across groups. The PMG estimator relies on pooling and averaging by allowing the intercept, short-run coefficients, and error variances to differ across group, while the MG estimator relies on averaging of coefficients (Pesaran et al. 1999). In this article, these three estimators are applied, and the results are compared to ensure the robustness of the estimates.

Nevertheless, one major weakness of the above panel techniques is that they do not account for existence of nonlinear relationship between variables. In other words, these models assume that the saving is linearly related to economic growth. To address this weakness, this article applies the dynamic threshold regression model introduced by Hansen (1999) and extended by Caner and Hansen (2004) which controls for endogeneity and serial correlation. The threshold regression can be represented by: GDPit=μi+β1 GDSit I(GDSit≤γ)+β2 GDSit I(GDSit>γ)+εit[Eqn 4]

In Equation 4, the subscript i = 1… …5 indicates SACU member countries, and t = 1, …. T represents time dimension. While μ_i denotes country specific effect and ε_it is the error term. Kremer, Bick and Nautz (2013) proposed different regimes that correspond with the optimal threshold value. Hence, I is an indicator function, showing the regime. GDS_it is gross domestic savings indicating threshold variable and regime-dependent explanatory variable, and the threshold level is given by γ.

The unit root properties of the variables were analysed given the need to investigate the long run relationship between domestic saving and economic growth. For this reason, the panel unit root tests by Im, Pesaran and Shin, (IPS 2003), Levin, Lin and Chu (LLC 2002) were implemented, and the results are reported in Table 2 for both level and first difference. The unit root tests are conducted in the general form of intercept and intercept with trend to ensure robustness of the stationary tests at level and first difference.The LLC shows that all the variables are nonstationary in level but stationary in first difference, implying that the variables are integrated of order one or I(1). Also, the results based on the IPS test deviate slightly from those of LLC with respect to GDP. The IPS results show that GDP is stationary in level, suggesting that it is I(0). Overall, the results of the unit root tests indicate that the variables are either I(0) or I(1), which are consistent for the application of the ARDL model.

In this section, the long run relationship between domestic saving and economic growth in SACU is assessed using the Pedroni cointegration test. The results of this test are provided in Table 3. As reported in the table, the panel v-statistic is a one-sided test with large positive test statistic value rejecting the null hypothesis of no cointegration among the variable. Also, the null hypothesis of no cointegration is rejected by rho, PP and ADF-statistics for both with-dimension and between-dimension. This finding implies existence of long run relationship between domestic saving and economic growth. In addition, this finding suggests that while both variables may diverge over a short-term because of economic shocks, the relationship will converge to their long-term trend. This result is consistent with the previous studies by Bolarinwa and Obembe (2017), Gidigbi and Donga (2020), Mohanty (2018), Sellami et al. (2020), Sere and Tchereni (2020), and Van Wyk and Kapingura (2021).

After determining evidence of cointegrating relationship between domestic savings and economic growth, the next step is to estimate various panel regression models, specifically POLS, FE, RE, FMOLS and DOLS. Table 4 reports the results for each of these regression models. Interestingly, the results from the POLS, RE, FMOLS and DOLS show that domestic saving has positive and statistically significant effect on economic growth. The elasticity coefficient across the models ranges from 0.06 to 0.20. This finding suggests that an increase in domestic savings by 1% would likely contribute between 6% and 20% increase in economic growth of SACU. This also suggests that accumulation of domestic savings is important to stimulate economic growth in SACU countries. In other words, mobilisation of domestic saving should be a priority for SACU countries to realise sustainable economic growth. This finding supports the role of saving in promoting growth as suggested by economic theories such as Harrod-Domar model, Lucas (1988) and Romer (1986). Empirically, this finding is consistent with previous studies that found positive effect of savings on growth (Gidigbi & Donga 2020; Jagadeesh 2015; Kok & Munir 2021; Siaw et al. 2017; Soylu 2019; Van Wyk & Kapingura 2021).

However, it is shown that investment has a negative and statistically significant effect on growth in SACU countries according to the results from FE and FMOLS. This result is inconsistent with the proposition of the economic theory that GDP and investment are positively related. This negative effect could be because the level of investment is well below the required level to promote economic growth or that the available savings are not invested in productive assets. On the other hand, the coefficient for trade openness is positive and significant in all the models except for DOLS regression. This is consistent with economic theory because accumulation of domestic savings encourages importation of intermediate materials and technological transfer which enhances growth. The analysis also revealed that inflation rate has a negative effect on economic growth in the SACU countries. This finding suggests SACU countries should strive for low rate of inflation in order to promote economic growth.

While the results from various panel regression models in Table 4 appear consistent, these models have some weaknesses. In particular, these models are unable to distinguish between the short run effects from the long run effects. To address this weakness, the MG, PMG and DFE models were estimated using the optimal lag length based on SIC (the lag length selection criteria is provided in Table 1-A1 in the Appendix 1).

Table 5 reports the results from the MG, PMG and DFE model. As shown in the table, while the result from MG model shows that domestic saving has a positive long run impact on growth, the results from the PMG and DFE models show insignificant effect. The result from MG model collaborates evidence from POLS, RE, FMOLS and DOLS models. This finding is in line with the previous studies that found evidence of positive long run impact of domestic savings on economic growth (Gatsi & Appiah 2020; Joshi et al. 2019; Miah & Majumder 2020; Ribaj & Mexhuani 2021). Also, the results show that investment as proxied by GFCF has negative long run impact on economic growth, while TRADE promotes economic growth in the long run.

Looking at the short run impact (Panel B), the results from the PMG and DFE models show that domestic savings contributes positively to economic growth. It is shown that a percentage increase in domestic saving would likely lead to approximately 8% to 14% increase in GDP. This finding is consistent with previous studies that found a positive relationship between saving and economic growth in the short-term (Kok & Munir 2021; Sere & Tchereni 2020; Van Wyk & Kapingura 2021). The error correction terms from three models are negative and significant as expected, implying existence of long run relationship between domestic savings and economic growth. This finding supports the evidence from the Pedroni cointegration test result. Furthermore, the error correction term shows that speed of adjustment to equilibrium ranges from 90% to 100% across the three models. That is, it suggests that it takes about 1 year for the system to revert to equilibrium after a given shock. The model analysis using the Hausman test shows that the PMG model provided more robust results than DFE. Subsequently, the stability of the estimated models was performed using the cumulative sum (CUSUM) and cumulative sum of squares (CUSUMQ) tests. The results, as displayed in Figure 1-A1 in Appendix 1, show evidence that the estimated models are stable and consistent.

The dynamic threshold regression is estimated with GDP as the dependent variable while domestic saving is treated as a threshold variable. The estimated model is divided into two regimes, the lower regime and the upper regime (Kremer et al. 2013). The result of the dynamic threshold regression is provided in Table 6. The result shows that domestic saving is a suitable threshold variable as confirmed by the linearity test which is significant at 5% level. In addition, the probability values for the two regimes indicate the rejection of the null hypothesis of linearity which confirms the existence of nonlinear or asymmetric relationships between domestic savings and economic growth. This finding is interesting given that domestic savings stimulate economic growth through investment channel, rather than by itself as suggested by economic theories (such as Domar 1946; Harrod 1939; Romer 1986; Solow 1956).

Table 6 also revealed that domestic saving contributes approximately 6% to economic growth at a lower regime whereas it reduces economic growth by approximately 13% at the upper regime, with estimated threshold value of approximately 16%. This finding implies that for SACU countries to achieve the growth target of 6%, the domestic saving to GDP ratio of less than or equal to 16% will be required. By contrast, saving to GDP greater than 16% is likely to reduce economic growth by about 13%. The above finding is consistent with the idea of Solow (1956) that rising savings influence economic growth until the economy reaches a steady state. In addition, this threshold level might be the result of low rate of saving among SACU countries given the low rate of income coupled with high level of unemployment, poverty and inequality. Consequently, a higher threshold level may lead to crowding out household spending and consumption, which may in turn affect economic growth.