According to the agency theory, while the shareholders (principals or owners) of a firm strive to attain the objective of firm value maximisation, the managers (agents) pursue the objective of their own utility maximisation (Jensen 1968). These self-interests may include extracting perks from the firm’s resources (Morck et al. 1988), empire building (Ongore 2011), suboptimal resource allocation (Khan et al. 2014), and establishing bureaucratic inefficiencies and burdens to further their own agenda (Iwasaki, Ma & Mizobata 2022). As a result of the information asymmetries, shareholders must implement mechanisms such as increased monitoring, debt covenants, and management incentive plans, such as share option schemes, to align the interests of both parties (Fleming, Heaney & McCosker 2005; Komati 2017). The cost of these alternatives, known as agency costs, impacts the firm’s bottom line and its ability to maximise shareholder value.

Managerial ownership is recognised as a mechanism that converges the interests of managers and shareholders, as managers will have an incentive to increase the firm’s value if they own a stake in the company (Dube 2018). An increase in managerial ownership thus reduces both information asymmetry as well as the agency costs that other shareholders need to incur, as it aligns the incentives of all parties (Shleifer & Vishny 1986). However, high levels of managerial shareholding can induce an entrenchment effect, where managerial owners who own a significant portion of a company’s shares aim to cultivate their private interests at the expense of minority shareholders and the other executives in the firm (Naidu et al. 2022; Ongore 2011). These managers may become resistant to external scrutiny, display a lack of accountability and prioritise decisions which ensure their job security and control, even though it may be suboptimal for the other shareholders. The use of share option schemes, which grant managers the exclusive rights to purchase shares of the firm at a pre-determined exercise price, is also proven to cause managers to engage in speculative stock actions that cause short-term volatility in the share price, at the expense of long-term value (Bolton, Scheinkman & Xiong 2006; Naidu et al. 2022).

While studies such as Ongore (2011) and Din et al. (2021) found evidence of a positive relationship between managerial ownership and firm performance for Kenyan and Pakistani firms, respectively, Mohd (2020) found no significant relationship between these two variables in the Malaysian context. The South African evidence is largely negative. Komati (2017) and Dube (2018) utilised a sample of cross-industry data on the JSE, and both studies found evidence of a negative relationship between managerial ownership and firm performance.

While the aforementioned studies represent linear representations of the model, there is a further stream of evidence in the field which tests the relationship using non-linear models. Khan et al. (2014) is one of these studies, and the results produced indicate a U-shaped relationship between managerial ownership and firm performance. This implies that managers can be entrenched at low levels of ownership, and this may only be eliminated when they own enough shares to have their interests aligned with the owners. In contrast, Hu and Zhou (2008) and Din and Javid (2011) found a non-linear inverse U-shaped relationship between managerial ownership and firm performance. This result suggests that managerial shareholding below the threshold improved firm performance; however, once shareholding surpassed the threshold, firm performance was impaired (Naidu et al. 2023). Hu and Zhou (2008) attributed the positive effect at lower levels to interest alignment and the negative effect at higher levels to entrenchment. The threshold value was estimated to be 75% by Hu and Zhou’s (2008) study of the Chinese market, whereas Din and Javid’s (2011) study of listed firms in Pakistan estimated a threshold value of 25%.

Foreign ownership includes both strategic and institutional investors (National Treasury 2017). This study focuses on foreign strategic investors, defined as those holding at least 10% of a firm’s equity (Orlic, Hashi & Hisarciklilar 2018), who are more likely to engage in long-term commitments and play an active operational role to enhance profitability and productivity.

The resource-based theory refers to the competitive advantages that firms obtain from strategic resources that are scarce, inimitable and irreplaceable (Naidu et al. 2022). This theory considers foreign shareholders as important suppliers of scarce resources, such as cutting-edge technologies and advanced managerial capabilities (Swart 2013). This may improve profitability through increased sales and cost reduction because of superior resources (Naidu et al. 2022). In emerging markets, foreign owners have been found to be efficient monitors of the firm (Khanna & Palepu 1999). Foreign investors who risk transferring their funds across borders are also found to be strongly motivated to maximise returns and are likely to demand high corporate governance standards (Iwasaki et al. 2022). However, foreign shareholders are also found to be susceptible to the entrenchment effect under large ownership levels, similar to managerial owners (Phung 2015). Similarly, foreign shareholders can potentially increase information asymmetries, particularly when the target companies are remote and across different time zones (Naidu et al. 2022).

South African studies such as Swart (2013) found that there is no relationship between ownership structure and performance, while Dube (2018) found a positive relationship between foreign ownership and return on assets (ROA), but negative impacts on return on equity (ROE) and Tobin’s q. International studies of foreign ownership, which account for non-linearity, such as Greenaway, Guariglia and Yu (2014) and Phung (2015), found evidence of a non-linear, inverse U-shaped relationship between foreign ownership and firm performance. The only South African study to date which evaluated nonlinearity in the relationship between foreign ownership and performance was Naidu et al. (2022), and the results found evidence of a positive relationship at low levels of shareholding, up to a level of 40.1%. Anything above this threshold was found to cause negative performance impacts.

Institutional ownership is generally large companies which invest on behalf of their clients, such as insurance companies, pension funds and investment funds. These domestic institutional owners are often active in monitoring and enhancing corporate governance strategies and tend to place emphasis on the growth of firms that they are invested in (Iwasaki et al. 2022). According to the agency theory, these investors can therefore reduce information asymmetries and agency costs and thus improve financial performance.

However, these owners can also become entrenched at high ownership levels, in which case they may prioritise their own interest in business negotiations, rather than that of the firm, an issue which is particularly relevant if the institutional investor has existing business relations with the firm (Iwasaki et al. 2022). Further research has shown that as institutional investors vote in terms of their fiduciary responsibilities to their investors, they are more risk-averse and may pressure managers into avoiding projects which are risky in the near-term, but provide opportunities for long-term value creation (Yang 2021).

As such, the literature displays mixed results of a positive association (Komati 2017), a negative association (Dube 2018), a non-linear U-shaped relationship (Wardhana & Tandelilin 2011) and an inverse U-shaped relationship (Daryaei & Fattahi 2020). While a U-shaped relationship (observed in Indonesian firms) indicates that at high levels of ownership, institutional investors attempt to maximise Tobin’s q, the inverse U-shaped relationship, which was observed in Iran, is attributed to the institutional investors developing a strategic relationship with managers to pursue their interests.

According to the resource-based theory, government owners can also provide firms with competitive advantages, but through political resources, easier access to capital and profitable government contracts (Dube 2018; Habtoor, Hassan & Aljaaidi 2019). Political resources include strong connections to political officials, which may promote the firm’s public reputation and legitimacy (Komati 2017). However, unlike other types of owners, government shareholders can be motivated by social and political objectives rather than shareholder wealth, and the agency issue between management and government can be exacerbated, particularly when the state is a majority shareholder (Dube 2018; Iwasaki et al. 2022). From the perspective of agency theory, government shareholders are thus often seen as detrimental to firm performance (Phung 2015).

While studies like Mugobo, Mutize and Aspeling (2016), Al-Matari, Al-Swidi and Fadzil (2013), Le and Buck (2011) observed that government ownership has a positive impact on firm performance, Ongore (2011) found a negative impact in Kenya, and Komati (2017) found no relationship present in South Africa. Evidence of non-linearity in the relationship was also found by Hess, Gunasekarage and Hovey (2010) and Phung (2015).

In family-owned firms, Srivastava and Bhatia (2020) suggest that agency conflicts are reduced in family-owned firms, as families tend to either oversee operations directly or maintain close oversight of management. The resource-based theory also proposes that family firms possess a unique bundle of resources, which includes loyalty, willingness to work long hours and easy development of tacit knowledge (Srivastava & Bhatia 2020; Venter & Farrington 2009). Studies by Rajput and Joshi (2015) and Wang and Shailer (2017) confirm this relationship in their evaluations of the Indian and emerging markets, respectively. However, with family ownership, entrenchment, relational and altruistic aspects surrounding self-control problems, excessive generosity and nepotism can also contribute to agency costs (Srivastava & Bhatia 2020). This hypothesis was confirmed by the study by Al Farooque, Buachoom and Sun (2020), who evaluated the issue in the Thailand market. Studies that accounted for non-linearity with quadratic specifications provided strong evidence of inverse U-shaped relationships. For example, in India, Srivastava and Bhatia (2020) observed that the initial positive effect of family ownership began diminishing at approximately 30% for ROA and ROE and 42% for Tobin’s q. In the United Kingdom (UK), Poutziouris, Savva and Hadjielias (2015) observed an optimal value of 31% of family ownership, claiming that ownership levels above 31% increased family opportunism, which hinders performance.

Based on the evidence, there is no definitive conclusion as to the impact each ownership category has on firm performance, with the results not only country-specific but also subject to the assumptions made regarding endogeneity.

The sample includes all non-financial firms listed on the JSE from 2004 to 2021, covering the implementation of King II, King III and King IV. This period spans key events such as the 2008–2009 financial crisis and the COVID-19 pandemic, both of which may have impacted ownership structures. Notably, the pandemic saw sharp declines in international investment (Strydom 2022) and several firms delisting from the JSE in 2020 and 2021 (Prinsloo & Henderson 2021; Wilson 2021). The study includes both listed and delisted companies but excludes financial firms because of their distinctive financial reporting standards, asset profiles and regulatory conditions compared to other sectors. The final dataset contains 267 firms, resulting in an unbalanced panel of 3246 annual observations. Financial statement data were sourced from Bloomberg, and ownership data from IRESS and Equity Real Time.

Firm performance is quantified using accounting-based measures, such as ROA and ROE, as well as a market-based indicator, namely Tobin’s q. Tobin’s q is regarded as a forward-looking approach, as it reflects expectations about future earnings, growth and market perceptions of the value of the company (Demsetz & Villalonga 2001). The use of both accounting-based and market-based metrics enhances the validity of a study’s results (Dube 2018).

The primary independent variables in the analysis are the five ownership types of focus: managerial, foreign, institutional, government and family ownership. Managerial ownership is measured by the shares held by insiders, including both directors and managers (Din et al. 2021). Foreign ownership is measured by two variables: a dummy variable that accounts for strategic foreign investors who own at least 10% of shares, as per the definition of foreign-owned firms by the International Monetary Fund (IMF) and a continuous variable that measures the percentage of shares owned by all foreigners (institutional and strategic) in the firm. The continuous variable allows for the detection of non-linearity as well as the optimal percentage of foreign shareholding.

Institutional ownership is measured by the percentage of shares held by asset management firms, banks, brokerage houses and insurance companies (Dube 2018). Following Dube (2018), government ownership is measured by the percentage of shares held by the Public Investment Corporation, Government Employees Pension Fund and SOEs such as Transnet, South African Broadcasting Corporation, Eskom, Passenger Rail Agency of South Africa and the Industrial Development Corporation. Lastly, family ownership is measured by both a dummy variable and a continuous variable. The dummy variable captures family trusts holding at least 20% of shares. This follows previous studies in which a company was classified as family-owned if families held more than 20% of the shares (Lodh, Nandy & Chen 2014).

Dummy variables for foreign and family ownership are used in Equation 1, as shareholdings below their minimum values do not align with how these ownership types are conceptually defined. In contrast, managerial, institutional and government ownership are treated as continuous variables, because they are not dependent on predetermined ownership levels.

Various firm-specific control variables are outlined. Firm size is considered because of its potential impact on performance, where larger firms are often more efficient and competitive, yielding better results (Gurbuz & Aybars 2010). Yet, Dube (2018) argued that larger firms may suffer from performance-hampering information asymmetry. Thus, the link between performance and size is unclear. Firm performance may also be impacted by the firm’s age. Komati (2017) posits that older firms have superior performance because of experience, suggesting a positive relationship between age and firm performance. The leverage ratio is included, but its impact on performance is uncertain. The signalling theory suggests positive effects as high-quality firms use debt to showcase value, while the pecking order theory predicts negative impacts because of agency costs (Dube 2018).

The fourth control variable is the dividend payout ratio. Distributing dividends diminishes free cash flow, forcing firms to seek external funding for new investments. This, in turn, elevates the level of external monitoring and improves firm performance (Jiraporn, Kim & Kim 2011). To account for macroeconomic factors affecting firm performance, the South African gross domestic product (GDP) growth rate is included as a control variable. Higher GDP boosts consumer purchasing power, demand and profitability, suggesting a positive relationship. Lastly, industry and year dummy variables are included to account for sector-specific traits as per the Industrial Classification Benchmark, and contemporaneous correlations among firm errors, respectively.

First of all, to examine the linear impact of ownership structure on firm performance, Equation 1 is estimated: Yit=α+βOit+γzit+ei[Eqn 1] where i = 1 … N and t = 1 … 18; Y_it is either ROA, ROE or Tobin’s q; O_it is the group of ownership variables; z_it is the set of control variables and e_it is the random error term. Table 1 outlines the measurement of the variables.

To capture potential non-linear effects of ownership types on firm performance, Equation 1 is modified by adding a quadratic term as shown in Equation 2: Yit=α+βOit+λ(Oit)2+γxit+eit[Eqn 2]

Equation 2 uses the continuous variables of foreign ownership and family ownership as the dummy variables lack the structure required to address non-linearity.

Any significance of the quadratic term λ; indicates the existence of a non-linear relationship. A U-shaped curve is characterised by a negative slope at lower values that becomes positive at higher values, whereas an inverse U-shaped curve initially exhibits a positive slope that turns negative at higher levels. To identify the precise shape of the relationship, the following composite null hypothesis (inverse U-shape) and alternative hypothesis (U-shape) are tested (Equations 3 and 4): H0:α+2λ (0min)≥0 and/or α+2λ (0max)≤0[Eqn 3] H1:α+2λ (Omin)<0 and/or α+2λ (Omax)>0[Eqn 4]

The minimum and maximum ownership values used in the hypotheses are based on the data ranges observed for each of the ownership variables. Managerial, institutional and government ownership have minimum values of 0%, while the minimum values for foreign ownership and family ownership are set at 10% and 20%, respectively, as per the definitions of strategic foreign investors and family-owned firms. The output from the test provides Fieller confidence intervals to identify threshold values at which the null hypothesis can be rejected.

Endogeneity is a frequent concern in ownership and firm performance research, which can stem from dynamic endogeneity, simultaneity and unobserved heterogeneity (Schultz et al. 2010). While dynamic endogeneity occurs when past performance determines the firm’s current ownership structure and control characteristics, simultaneity occurs when ownership levels are determined concurrently with a firm’s performance. Unobserved heterogeneity can arise from firm-specific attributes that may impact the ownership structure, control characteristics and firm performance; however, this is difficult to measure.

If any of the three above-mentioned forms of endogeneity is found in the data, conventional panel models such as the Fixed Effects Model (FEM) may yield biased coefficients (Phung 2015). In such cases, the system GMM model is a reliable alternative, as it delivers unbiased and consistent estimates in the presence of all forms of endogeneity (Schultz et al. 2010). However, in the absence of endogeneity, the FEM specifications deliver more efficient estimates than the GMM (Schultz et al. 2010). This study, therefore, employs three tests of endogeneity, as discussed in the ‘Results and analysis’ section. If endogeneity is detected, system GMM will be used to estimate Equations 1 and 2. If the endogeneity tests indicate that the variables are exogenous, the FEM is instead adopted.

An application for full ethical approval was made to the Human Research Ethics Committee and an ethics waiver was received on 01 July 2025. The ethics clearance number is HRECNMW25/07/04.

Table 2 presents the descriptive statistics for the variables.

Consistent with earlier South African studies (Dube 2018; Komati 2017; Zhang 2016), institutional investors represent the largest shareholder group on the JSE, with an average stake of 37.7%. Foreign ownership ranks second, averaging 18.1% of total shareholding.

Managerial ownership and government ownership constitute an average of 14% and 5.1% of shareholding on the JSE, respectively. Family ownership, in contrast, accounts for a negligible average of only 0.07%, which can likely be attributed to the fact that most family-owned businesses in South Africa are concentrated in unlisted small and medium-sized enterprises rather than on the public exchange (Venter & Farrington 2009). The average ROA and ROE over the sample period are 6% and 12.7%, respectively. These values are lower than Komati’s (2017) findings of 9.48% and 15.27% from 2004 to 2014, thus inferring a decline in performance for JSE-listed firms from an accounting perspective. The average Tobin’s q of 1.186 suggests that over the sample period, JSE shares are mostly overvalued.

Table 3 displays the correlation matrix for the variables. As expected, ROA and ROE are strongly correlated given their similar computations. Both metrics show weak positive correlations with Tobin’s q, indicating that the measures capture distinct aspects of firm performance. All performance indicators exhibit positive correlations with foreign and government ownership, while showing negative correlations with institutional and family ownership. Managerial ownership displays positive correlations with ROA and ROE but negative correlations with Tobin’s q.

Given that the existence of at least one source of endogeneity can produce severely biased and invalid results (Schultz et al. 2010), we conduct three endogeneity tests to determine a suitable estimation approach for this study. As per Wintoki et al. (2012), firm age and the industry and year dummy variables are treated as exogenous.

The first test of dynamic endogeneity is implemented by regressing current performance against lagged performance and control variables and assessing the significance of the performance lags (Wintoki et al. 2012). Significant lagged values indicate that past performance impacts current performance. These results are displayed in Appendix 1A-1. In panel A, when four performance lags are incorporated, the initial lag is significant across all performance metrics, while the second lag demonstrates significance for ROA and Tobin’s q, and the third lag for ROA and ROE. The fourth lag, however, is insignificant for all performance measures. When the first and second lags are omitted (panel B), the third lag is significant for all performance measures, whereas the fourth lag is significant for ROE only. This implies the existence of dynamic endogeneity, as past performance influences current performance across all four indicators. Furthermore, the significance of later lags (years 3 and 4) when the first and second lags are excluded suggests that, although these later lags hold relevant information, the more recent lags (years 1 and 2) already subsume most of this information. Hence, two lags of past performance are considered sufficient to capture dynamic endogeneity and are thus incorporated into Equations 1 and 2.

The second method used is the weak test for exogeneity, which is applied to detect reverse causality (Wintoki et al. 2012). The findings of the weak exogeneity test in Appendix 1A-2 reveal significant reverse causality between the control variables and firm performance. Aside from the GDP growth rate, all control variables are significantly related to past ROA. Similarly, the lag of ROE is significant in explaining most control variables, except for the GDP growth rate and leverage. The lag of Tobin’s q is significantly related to all control variables. This suggests that the control variables retain a certain level of endogeneity with firm performance. Managerial ownership and family ownership are the only ownership variables significantly related to past performance. Specifically, managerial ownership shares positive associations with the past values of ROE and ROA, while demonstrating negative relationships with past Tobin’s q. This result indicates reverse causality, where past performance, from a year ago, influences the current percentage of managerial shareholding, but not vice versa.

Finally, Wooldridge’s (2002) test of strict exogeneity is performed to examine whether past performance affects the future ownership structure. Appendix 1A-3 presents the results of the strict exogeneity test with different subsets of ownership and control variables. In panel A, the future values of managerial ownership (in specifications 1 and 7), foreign ownership (in specification 6) and government ownership (in specification 6) display coefficients that are significantly different from zero, whereas panel B shows that all specifications for the forward values of government ownership are significantly different from zero. Contrastingly, panels C and D do not present any significant future values of ownership variables. Hence, aside from family ownership and institutional ownership, the results of the strict exogeneity test suggest that future realisations of ownership variables are related to current performance and, therefore, cannot be considered strictly exogenous. Moreover, the future values of all control variables are significant in either panel.

Based on the results of the tests of dynamic endogeneity, weak exogeneity and strict exogeneity, all control and ownership variables (except family ownership and institutional ownership) exhibit significance in at least one of the tests, indicating a certain degree of endogeneity with firm performance. Therefore, all variables are treated as endogenous in this study, leading to the adoption of the system GMM for estimation purposes.

Table 4 presents the regression results for Equations 1 and 2, estimated using the system GMM. Two lags of past performance are included to address the dynamic endogeneity.

Based on the Arellano-Bond autocorrelation test, there is evidence of first-order but not second-order serial correlation in the first differenced residuals, thereby satisfying the exogeneity assumption. The Hansen test also reinforces the exogeneity of the variables.

The role of the control variables in explaining firm performance varies across each measure. In Equation 1, firm size has a significant positive effect on ROA and ROE but a negative effect on Tobin’s q. The positive impact suggests that larger firms are more diversified and employ better technology, which may increase profitability (Phung 2015). In contrast, the negative effect can be interpreted as the diseconomies of scale that manifest in the firm’s market valuation (Lawson & Osaremwinda 2019). As firms grow, they may experience inefficiencies (diseconomies of scale) that diminish their value.

Dividend payout positively impacts all performance metrics in Equations 1 and 2, consistent with the view that higher dividend payouts increase the level of external monitoring and thus improve firm performance (Jiraporn et al. 2011). The asset turnover ratio positively influences both ROA and ROE, indicating that more efficient asset utilisation enhances accounting performance.

As per Equation 1, managerial ownership has a significant positive linear effect on ROE but diminishes Tobin’s q. The positive impact can be attributed to the convergence of interest between managers and shareholders, while the negative effect may stem from possible entrenchment and stock liquidity effects, because Tobin’s q is related to concurrent liquidity (Fabisik et al. 2018). Specifically, Fabisik et al. (2018) indicate that managerial shareholders may decrease stock liquidity by leveraging insider information and selling only under conditions that are acceptable to them. Similarly, foreign investors, who own at least 10% of shares, positively influence the ROE of JSE-listed firms. Firms with higher ROE typically possess competitive advantages that yield superior investor returns. Thus, as per the resource-based theory, the increase in ROE could result from the transfer of scarce resources by foreign investors to JSE-listed firms, which is not captured by ROA or Tobin’s q.

In contrast, family ownership exhibits negative effects on Tobin’s q. The negative impacts may arise from entrenchment, where institutional investors enable entrenched managers for personal gain, and family owners make decisions that prioritise family interests over the firm’s financial health (Al Farooque et al. 2020). The insignificant effects of institutional ownership on ROA, ROE and Tobin’s q reinforce King IV’s critique of the lack of involvement by institutional investors (Zhang 2016). Government ownership has no significant impact on firm performance, which may stem from the low level of government shareholding on the JSE because most SOEs, aside from Telkom, are unlisted enterprises.

Equation 2 shows that a non-linear relationship exists between foreign ownership and ROE. Based on the positive linear coefficient of foreign ownership, along with the negative quadratic term, this relationship is characterised as an inverse U-shaped pattern.

The SLM test results in Table 5 verify an inverse U-shaped relationship between foreign ownership and ROE, evidenced by the positive slope at the minimum level (0.661) and the negative slope (−1.341) at the maximum level of foreign ownership. Moreover, the extreme point (0.396) and the Fieller interval (0.310–0.784) fall within the foreign ownership range (0.1–0.997), thus violating the conditions of a U-shape in Equation 3.

The extreme point of 0.396 reflects the optimal level of foreign ownership. Accordingly, as foreign ownership rises, the ROE of JSE-listed non-financial companies initially improves, but once foreign ownership exceeds 39.6%, its impact turns negative, as displayed in Figure 1. An inverse U-shaped relationship between foreign ownership and firm performance coincides with prior studies (Greenaway et al. 2014; Gurbuz & Aybars 2010; Naidu et al. 2022).

From a theoretical standpoint, the improved firm performance stemming from low levels of foreign ownership can be ascribed to increased monitoring and the supply of scarce resources and expertise by foreign investors (Douma, George & Kabir 2006). Kollamparambil and Jogee (2018) claimed that foreign enterprises in South Africa are deemed more technologically advanced than their domestic counterparts because of foreign investors sharing their technical knowledge and introducing cutting-edge technology. Moreover, South African multinational corporations allocate significant funds to training and developing local staff, frequently bringing in international experts to enhance the skills of local employees.

According to the agency theory, the decline in performance at higher levels of foreign ownership can be ascribed to the entrenchment effect on minority shareholders, which hinders firm performance. Additionally, the poor performance can also be a consequence of the liability of foreignness, which is the additional costs a firm operating in a market overseas incurs that a local firm would not incur (Naidu et al. 2022). Local alliances offer benefits to domestic firms that cannot be achieved by majority foreign-owned firms (Gurbuz & Aybars 2010). In South Africa, domestic owners may have a greater understanding of the dynamics of the South African environment, including corporate governance structures, the target market, business practices and the legal and regulatory frameworks. Therefore, when foreign ownership surpasses 39.6% of total shareholding on the JSE, the decline in firm performance may arise from entrenchment and the liability of foreignness. Based on the optimal level of foreign ownership, it is recommended that domestic shareholders comprise at least 61% of the total ownership to retain optimal firm performance and prevent the expropriation of minority local shareholders (Phung 2015).