The study examines whether executive remuneration remained sensitive to company performance among JSE-listed firms during the COVID-19 pandemic. The literature first sets out the theoretical foundations that motivate performance-linked pay, then considers the main determinants of remuneration beyond performance. It next reviews evidence on how researchers measure performance and on how pay and performance move together in both stable times and in crisis. Finally, it situates the question in the South African governance and macroeconomic context and reflects on ethical and strategic behaviours that can distort incentives.

The focus area of the study was companies listed on the JSE, South Africa. The JSE setting was used because listed firms are required to disclose audited, standardised financial statements and remuneration information, allowing comparable measurement of executive pay and performance across firms and sectors (Kirsten & Du Toit 2018). Firms from all JSE sectors were included to capture remuneration dynamics across cyclical and non-cyclical industries and across firms of varying size and complexity, which are established determinants of executive remuneration (Ciscel & Carroll 1980; Duffhues & Kabir 2008).

Company financial and market data were obtained from the IRESS database. Executive remuneration values were gathered from total directors’ emoluments disclosed in audited annual financial statements and reconciled with IRESS data where available to ensure completeness and accuracy (Kirsten & Du Toit 2018; Muthu & Wesson 2023). The study’s population comprised all firms listed on the JSE main board between 2015 and 2024, totalling 234 companies. A purposive census-style sampling approach was applied by retaining all firms subject to data availability across the period. Firm-year observations were included when directors’ emoluments and the accounting and market variables required to compute the study measures were available for that year. Observations were excluded where remuneration disclosures were missing or where key financial or market variables were incomplete. Firms were removed only when missing data were extensive across the period. After applying these criteria, 12 firms were excluded as a result of insufficient data, leaving 222 firms. The final dataset contains 2010 firm-year observations in an unbalanced panel because some firms do not have data for every year.

The study time frame covers 10 financial years and was divided into three phases: A pre-pandemic period 2015 to 2019, the pandemic period 2020 to 2021 and a post-pandemic recovery period 2022 to 2024. This periodisation allows direct comparison of pay-performance sensitivity under normal conditions and during the COVID-19 shock, consistent with pandemic-era pay-performance studies that isolate COVID effects through period dummies and interactions (Jarby 2022; Tusiime 2024). Monetary values were recorded in South African rands. Where firms reported in foreign currencies, values were converted into rands using average annual exchange rates for balance-sheet items and year-end exchange rates for income-statement items to ensure comparability across firms and years (Shaw 2011).

Executive remuneration is the dependent variable. Executive pay (ExecPay) was measured as total directors’ emoluments disclosed in firms’ annual financial statements. This measure captures fixed salary, bonuses, benefits and other disclosed pay components, providing a comprehensive proxy for executive compensation (Kirsten & Du Toit 2018). Executive pay was transformed using the natural logarithm to normalise the distribution and reduce the influence of extreme observations, consistent with remuneration modelling practice (Cieślak 2018; Gao & Li 2015; Jarby 2022; Tusiime 2024).

Company performance was measured using both accounting-based and market-based indicators. Accounting performance was captured by return on assets (ROA), defined as net income divided by total assets, and return on equity (ROE), defined as net income divided by shareholders’ equity. Return on assets reflects overall profitability and is widely used in pay-performance research, while ROE reflects profitability relative to shareholder capital (Obeidat & Darkal 2018; Shaw 2011). Market performance was captured by Tobin’s Q (TQ), which was computed using the IRESS market capitalisation and book value of assets as a proxy for replacement cost, reflecting investor valuation and growth expectations (Fu, Singhal & Parkash 2016). Using multiple performance measures is appropriate because no single indicator captures all dimensions of firm performance relevant to executive compensation (Bussin 2015).

Two control variables were included to account for firm characteristics that influence remuneration. Firm size (FSize) was proxied by the natural logarithm of total assets, capturing organisational scale and complexity, which are associated with higher executive pay (Ciscel & Carroll 1980). Firm leverage (FLev) was defined as total liabilities divided by total assets, capturing financial risk and capital-structure constraints relevant to remuneration policy (Duffhues & Kabir 2008; Tusiime 2024).

To isolate the COVID-19 shock, a binary indicator (COVIDDummy) was coded 1 for the pandemic years 2020–2021 and 0 for all other years. Interaction terms were constructed between the COVIDDummy, and each performance measure to test whether pay-performance sensitivity differed during the pandemic relative to the surrounding periods (Jarby 2022). Definitions and operationalisation of all variables are summarised in Table 1.

While multiple performance measures are used to capture different dimensions of firm outcomes, each proxy is subject to inherent limitations. Accounting-based measures such as ROA and ROE are derived from financial statement data and primarily reflect historical aspects of firm performance, which may limit their ability to capture future growth expectations and market perceptions (Obeidat & Darkal 2018). In contrast, market-based measures such as Tobin’s Q, obtained from the IRESS database, are sensitive to market volatility and investor sentiment, particularly during periods of economic disruption such as the COVID-19 pandemic (Bussin 2015). Consequently, Tobin’s Q may capture broader market conditions and forward-looking investor expectations rather than purely firm-specific performance (Fu et al. 2016). To address these limitations, the study employs both accounting- and market-based indicators, consistent with prior literature (Kirsten & Du Toit 2018; Nkwadi & Matemane 2022), to provide a more comprehensive and balanced assessment of firm performance and its relationship with executive remuneration (Bussin 2015).

All data were compiled into a firm-year panel and screened for missing observations and consistency before estimation. Continuous variables were winsorised at the 1st and 99th percentiles prior to log transformation to reduce the influence of extreme outliers in line with previous pay-performance studies (Jarby 2022). The analysis began with descriptive statistics to summarise the distributions of remuneration, performance and control variables, followed by a Pearson correlation matrix to examine preliminary associations among variables.

Ordinary least squares (OLS) panel regression with year and industry fixed-effects is adopted as the primary estimation technique because it provides transparent and interpretable baseline estimates of pay-performance sensitivity across the full panel, consistent with prior executive compensation studies (Bussin 2015, Gerhart & Milkovich 1990). This approach is appropriate given the study’s focus on estimating average relationships across a large panel while controlling for observable time and industry effects. However, OLS may not fully account for unobserved firm-specific heterogeneity and potential endogeneity arising from the simultaneous determination of executive pay and firm performance, as remuneration and performance may be jointly determined (Ntim et al. 2015). Accordingly, firm fixed-effects and dynamic generalised method of moments estimations are subsequently employed as robustness checks to control for unobserved heterogeneity, reverse causality and persistence in remuneration dynamics (Arellano & Bond 1991).

Pay-performance sensitivity was examined using panel OLS regression incorporating year and industry fixed-effects, which control for economy-wide shocks and structural sectoral differences in remuneration practices, respectively. The moderating role of the COVID-19 period was tested by introducing a binary pandemic indicator and its interactions with each performance measure into the same regression framework.

Baseline pay-performance sensitivity is estimated as follows (Equation 1):

where i indexes firms, t indexes years, γ_t are year fixed-effects and δ_S are industry fixed-effects.

To test whether pay-performance sensitivity differed during the COVID-19 period, the baseline model is extended to include the pandemic dummy variable and interaction terms:

COVID-19 interaction model (Equation 2):

Robustness checks were conducted in two steps. Initially, a firm fixed-effects model was estimated to account for unobserved time-invariant firm heterogeneity that could affect remuneration, while still controlling for year effects. A dynamic panel generalised method of moments estimator, specifically the difference generalised method of moments (GMM) approach following Arellano and Bond (1991), was employed to address potential endogeneity, simultaneity and persistence in executive pay and performance. Lagged instruments from t-2 onward were used to reduce reverse causality and limit instrument proliferation. Model validity was assessed using the Arellano-Bond AR(1) and AR(2) tests for serial correlation, together with the Sargan and Hansen tests of over-identifying restrictions to confirm instrument validity and overall specification consistency (Arellano & Bond 1991).

Ethical clearance to conduct this study was obtained from the University of Pretoria Faculty of Economic and Management Sciences Research Ethics Committee, University of Pretoria (No. EMS146/25). All variables were derived from publicly available audited secondary sources (IRESS and company annual reports), and no human participants were involved. Consequently, issues of informed consent, confidentiality and anonymity did not arise. The results of the empirical analyses are presented and discussed in the next section.

The descriptive results (Table 2) provide an overview of directors’ emoluments and firm-level characteristics. Director emoluments ranged widely, from approximately R1.9 million to R367 million, reflecting differences in company size and sector. The mean total emolument was R49.1 million, higher than the median value of R28.3 million, suggesting a right-skewed distribution. Firms financed nearly half their assets through debt, and leverage rose slightly during the pandemic years.

Average ROA and ROE were positive at 6.9% and 8.7%, respectively, although both measures displayed large variability. Tobin’s Q averaged 1.13, indicating modest growth expectations. A comparison of non-COVID and COVID means revealed that average directors’ emoluments increased slightly during the pandemic, while all three performance measures declined marginally. The results show that executive remuneration did not move in line with performance across the two periods, indicating a potential disconnect between pay and outcomes during the COVID-19 pandemic.

Table 3 presents the Pearson correlation coefficients and VIF values for the study variables. All coefficients fall well below the multicollinearity threshold of 0.8, indicating that the independent variables are suitable for regression analysis (Gujarati & Porter 2009). In addition, variance inflation factor values are well below the threshold of 10 used to diagnose multicollinearity, indicating that collinearity is unlikely to materially affect the estimation of the regression coefficients (Gujarati & Porter 2009). The strongest bivariate association is observed between directors’ emoluments and firm size (r = 0.70), suggesting that larger firms tend to pay higher executive remuneration irrespective of profitability.

The regression results in Table 4 present five model specifications examining the determinants of directors’ emoluments. Model 1 provides the baseline pay-performance regression, while Models 2 to 4 include the COVID-19 dummy and its interaction with the three performance measures (ROA, ROE and Tobin’s Q). Model 5 is the full specification, incorporating all variables and controlling for firm size, leverage and fixed-effects for industry and year.

The models show stable explanatory power, with R-squared values of about 0.59 and F-statistics significant at the 1% level. The coefficients are consistent in sign and magnitude across models, confirming robustness in the relationships identified.

Among the performance measures, Tobin’s Q shows the strongest positive and significant association with directors’ emoluments. In the full model, a one-unit increase in Tobin’s Q is associated with an estimated 3.1% increase in compensation. This result suggests that boards reward executives in firms with stronger market valuations and higher future growth expectations. Return on assets is also positive and significant, but with a smaller magnitude, indicating that firms with better asset efficiency tend to pay slightly higher remuneration. The stability of this coefficient across specifications implies that operational performance remains a modest but reliable driver of pay. Return on equity is insignificant across all models, indicating that shareholder-based profitability does not play a central role in determining executive pay.

The coefficient on the COVID-19 dummy is positive and significant at the 1% level, showing that directors’ pay was higher during the pandemic years after controlling for performance and firm characteristics. The magnitude of this increase, about 11%, implies that firms may have provided additional incentives to retain or compensate executives under heightened uncertainty. However, the interaction terms between COVID-19 and the three performance indicators are insignificant, suggesting that the relationship between pay and performance did not materially change during the pandemic. Pay levels increased, but sensitivity to performance remained weak.

Among the control variables, firm size has the most pronounced effect on directors’ emoluments. A 10% increase in firm size corresponds to an estimated 3.7% increase in pay, highlighting that larger, more complex firms tend to reward executives more generously. Leverage is also positive and significant, indicating that directors in more highly geared firms receive higher compensation, possibly as a premium for managing financial risk.

Overall, the regression analysis shows a weak but positive link between executive pay and firm performance. Market-based and accounting-based measures have limited explanatory power, while structural characteristics such as firm size and leverage are stronger predictors. Although executive pay rose during the COVID-19 period, this increase was not accompanied by greater sensitivity to performance, pointing to persistent misalignment between remuneration and firm outcomes.

All models control for industry and year fixed-effects. Dependent variable: Natural logarithm of directors’ emoluments.

Results confirm that firm size remains the most consistent determinant of directors’ pay across all panels. Pay-performance sensitivity is weak across industries and time periods, with marginal strengthening in high-performing firms and weakening during the COVID-19 period.

The dependent variable is ln(directors’ emoluments). The fixed-effects model controls for unobserved, time-invariant firm heterogeneity. The dynamic panel GMM model addresses potential endogeneity and persistence by instrumenting explanatory variables with their lagged values, including the lagged dependent variable.

Diagnostic tests for the GMM specification are as follows: Arellano-Bond AR(1): z = −3.50 (p = 0.000); AR(2): z = 2.10 (p = 0.036). Sargan χ² = 33.46 (p = 0.03); Hansen χ² = 24.59 (p = 0.22). While the AR(1) result is as expected, the significant AR(2) statistic indicates potential second-order serial correlation and raises concerns regarding instrument validity. The Hansen test fails to reject the null of valid instruments, whereas the Sargan test rejects it, suggesting mixed evidence. The GMM results are therefore presented alongside the fixed-effects estimates as a complementary specification.

To assess the stability and reliability of the baseline findings, a series of robustness checks was conducted using alternative sample partitions and estimation techniques. These tests address potential concerns related to industry heterogeneity, period-specific effects, performance-level asymmetries, unobserved firm characteristics and endogeneity in the pay-performance relationship.