The theoretical foundation of this paper lies in agency theory. Multiple factors influence the relationship between performance-based remuneration and the overall performance of a company. The agency relationship is deemed the most important and influential of these factors. An agency relationship is defined as a contractual agreement under which one or more parties, known as the principals, engage another party or parties, known as the agents, to perform a specific service on their behalf (Boshkoska 2015; De Wet 2013; Jensen & Meckling 1976; Kuo et al. 2013). Managers, as the agents of a company, would not act on behalf of the company if their interests are not fully aligned with those of the company (Dyl 1988). Jensen and Meckling (1976) claimed that managers are unlikely to maximise the wealth of a firm, even if the manager is the only shareholder. Bryant and Davis (2012), however, argue that the agency problem can be controlled by incentivising managers, as agents, to act in line with the interests of the company. There are various methods by which a company’s shareholders can control the operations of the managers. One of these methods is the use of performance-based remuneration (Boshkoska 2015). Companies thus incur agency costs in the form of, for example, incentive schemes to motivate managers to run the company in the best interest of the shareholders and to maximise shareholder wealth in a sustainable manner (Coughlan & Schmidt 1985; De Wet 2013). Remuneration policies should aim towards long-term value creation in a company (IoDSA 2016).

A variety of studies focus on executive compensation and how this can minimise the agency problem (Bebchuk & Fried 2003). Reward systems are designed to bridge the agency problem between shareholders and managers, but to bridge this gap, reward systems must be related to the performance of managers and the strategy of the company, and take into account the growth and nature of the company. Reward systems should not only motivate employees, but also enhance the overall corporate values of the company (Kuo et al. 2013). Gill et al. (2012) and Shin (2013) suggest that companies with bonus-based pay structures should consider that worker dishonesty might be an unintended consequence.

Studies such as those by Edmans and Gabaix (2009), Tosi and Gomez-Mejia (1994) and Yanadori and Milkovich (2002) narrowed the focus from executive compensation to that of the CEO of a company. The role of the CEO is pivotal in managing the resources of a company, especially in the context of a continuously changing external economic environment, to ensure value creation for shareholders. There is increased scrutiny levelled at the remuneration committees of companies, especially those that have increased CEO compensation, even when they face disappointing financial results (Bussin 2015). Gomez-Mejia, Tosi and Hinkin (1987) used CEO remuneration as a proxy for executive compensation and concluded that the payment of CEOs is more responsive to performance in companies that have dominant shareholders and are owner controlled. The allocation of company profits is generally based on the condition of achieving a certain level of return on equity employed or a specified dividend, or even both. The aim of a bonus fund is primarily to regulate the use of company funds, as well as safeguard the shareholders’ interests (Merrett & White 1968).

The main determinant for executive remuneration is often firm size, based on the revenue figure of the company (Baumol 1959; Conyon 1997; Jensen & Murphy 2010; Murphy 1985). Executives may thus pursue a goal of sales maximisation, rather than wealth maximisation. Coughlan and Schmidt (1985) opine that the practice to base the bonuses of CEOs on sales growth is not always the most optimal compensation policy. In a case were an executive is near retirement and expects a bonus-based on sales or sales growth, this executive might approve excessive expenditures for advertising campaigns or other increases in the sales force to ensure increased sales revenues, even though this might exceed the profit-maximising levels.

A study by Anderson and Bizjak (2003) found that CEOs who are members of their own compensation committees receive less overall compensation and have a large ownership share. According to the study, these compensation committees were doing more to link executive pay to the performance of the company than those of the control sample. To support this, Sigler (2011) found that independent executive directors, and the consultants hired to advise the board, have relatively little motivation to safeguard the interests of the company’s shareholders.

Another method to link executive compensation to the overall performance of the company is using long-term incentive plans (LTIPs). Pass (2003) found that most LTIPs were undemanding and rewarded average instead of outstanding performance. Buck et al. (2003) concluded that LTIPs are generally associated with reductions in the sensitivity of the relationship between executives’ total rewards and shareholder returns and thus questioned their effectiveness. Walker (2010) also concluded that there is an inadequate relationship between executive compensation and long-term company performance.

Various arguments state that executive remuneration policies played a role in the onset of the global economic crisis. According to Bebchuk and Spamann (2010), Fahlenbrach and Stulz (2011) and Bussin (2015), the global economic crisis was a result of dramatic disintegration in the market capitalisation of USA banks. Rajan (2005) suggests that the significant risks experienced by the global financial sector was, in part, a result of the compensation practices used within this sector. He argued that, even though executives in the financial sector were engaging in business practices that were eroding performance and threatening sustainability of the companies and, by implication, the entire financial system, these executives were still receiving substantial incentive pay-outs. This argument was based on the outlook that a weak performance–pay relationship exists, which allows risky business practices.

A perfect compensation plan – one which links performance to payment – seems impossible to construct. However, Hovy (2005) concluded that the most feasible mechanism to align shareholder wealth creation with executive performance is global performance-based remuneration plans. To achieve this, companies can follow the guidelines provided by Van Zyl Smit and Nel (2010), who claim that remuneration policies should embrace three basic principles, namely appropriateness, fairness and effectiveness, and should be tightly aligned with the long-term value and soundness of the company.

According to Coughlan and Schmidt (1985) an optimal remuneration package should not be based solely on share options or other direct relationships between share performance and pay. A more efficient remuneration package is one which combines share-related components with components which are not linked to shareholder wealth, such as a salary or a bonus. At the discretion of the executive directors, a portion of the bonus can even be linked to share price performance as a signal that the share price is regularly being monitored (Coughlan & Schmidt 1985). A mix of the various components of pay into a single compensation package for an executive director allows for the deficiencies of one component to be counterbalanced by another component’s strength. Including a cash bonus in the package motivates executive directors to focus on the short-term goals of the company, which counters the limitations of restricted share, namely only rewarding achievement of long-term goals (Sigler 2011).

According to the economic theories of compensation, the performance of a company should only influence an executive director remuneration to the point where it serves as a substitute for unobservable managerial efforts and productivity. Even though most of the theories suggest a correlation between executive compensation and company performance, most of them do not mention the measure of performance. Since the shareholders of the company are the principles within the agency theory, it would seem appropriate to define and measure performance in terms of shareholder return instead of accounting profits (Murphy 1985).

Measurement tools can be used to indicate whether the company achieved or surpassed their goals and objectives. Performance measures can also be used to evaluate, control and improve the production processes of the company (Ghalayini & Noble 1996). Performance measurement is thus a process which quantifies the efficiency and effectiveness of actions. The action leads to performance, while the measurement is the quantification process. The level of performance a business attains, is a function of how efficient and effective it undertook all of its actions (Neely, Gregory & Platts 2005). The two approaches most commonly used to empirically evaluate the financial performance of companies are accounting-based measures and market-based measures (Cochran & Wood 1984; Spicer 1978). Market-based measures are those that use corporate share returns to indicate the performance of a company (Kuo et al. 2013), while accounting-based measures use performance measures from the financial statements, such as profit.

According to Bourne et al. (2000), performance management has a dual purpose. The first is the success of the chosen strategy’s implementation as a result of the measures derived from the chosen strategy. The second is that the information and feedback received from the measures should be used to test the strategy’s validity and challenge the assumptions made. The main purpose of performance management can thus be seen as an examination of past performance to stimulate future actions (Neely et al. 2005). It is important to note that, to gain the full potential of performance measures, consistency within the implementation of these measures is needed. As Mintzberg (1978) emphasised, strategies can be realised only through the consistency of actions.

There are many empirical studies that focused on the relationship between company performance and executive director remuneration. However, many failed to produce conclusive results and consensus (Bruce, Buck & Main 2005; Bussin & Blair 2015). Some of these papers include those by Cosh (1975), Hirschey and Pappas (1981), Lewellen and Huntsman (1970), Masson (1971) and Meeks and Whittington (1975). All of these papers made use of accounting information to measure the profitability of companies. As a result, the authors encountered measurement errors which led to biased empirical results (Coughlan & Schmidt 1985). Table 1 below provides a summary of some of the previous research that investigated the relationship between company performance and executive director remuneration. The summary is categorised per year and shows the countries in which the studies were conducted.

The performance of both the company and executive directors are important, and both of these may be impacted by the remuneration packages chosen by the company. Although many studies have been performed over a period of 45 years, there is still no conclusive evidence as to the relationship between executive director remuneration and financial performance as many studies show conflicting results. This study attempted to confirm conclusions derived from previous studies conducted in South African and developing countries with more recent data. The following section addresses the research design and methods used in this empirical study.

The focus area of the study was companies listed on the Johannesburg Stock Exchange (JSE), South Africa. To investigate the relationship between performance-based payments and the overall performance of companies, the consumer goods and services industry was used. This industry was chosen because of the companies’ prominent sales figures, which can be used as a proxy for company size in relation to total executive director remuneration. In addition, using one industry eliminates the need to make provision for industry-specific differences between companies.

Company financial data was obtained from INET BFA, South Africa’s leading provider of financial data feeds and analysis tools. Out of the 59 companies listed on the JSE from the consumer goods and services industry, 17 could not be used because of incomplete data. A sample of 42 listed South African companies from the consumer goods and services industry were used. The total population of South African consumer goods and services industry companies is relatively small, which makes it impossible to employ proper random sampling (see for example Vize et al. 2009). The sample companies’ data for 2006 to 2015 were obtained and investigated. The result is 420 individual observations.

The independent variables used in the analysis were return on equity (ROE), return on assets (ROA), earnings per share (EPS), sales figure/turnover (TO) and price per share (PPS). ROE and ROA are presented as percentages, EPS and PPS as a rand value per share and TO as a total rand value. The dependent variable was the total executive director remuneration (REM) as a rand value, defined as executive directors’ emoluments plus executive directors’ share-based payments, both used in this study as disclosed on the Statement of Profit and Loss and Other Comprehensive Income. This figure for total director remuneration thus includes the executive directors’ basic salary and bonus, as well as other forms of remuneration, such as share-based payments and options.

The statistical model set out to establish which of these independent variables show the most significant relationship to and perhaps represent a driving force behind director remuneration. The independent variables were chosen for the ease with which investors or other interested parties can obtain or calculate them, should these measures show a direct relationship with total director remuneration. The first control variable included in the model was the debt/equity ratio (DE) as a proxy for size. Previous research has shown a direct positive relationship between leverage and firm size (Booth et al. 2001; Fama & French 2002; Rajan & Zingales 1995; Titman & Wessels 1988). Even though the market capitalisation of companies is often included as a control variable for company size, market capitalisation correlates too closely with the price per share (PPS), which is considered to be an important variable not previously considered in the study by De Wet (2013). The second control variable was the JSE All Share Index (JSE), which is included to compensate for market-related changes over the period under consideration.

The variables were tested for stationarity, using the Augmented Dickey-Fuller (ADF) test. The results of the stationarity test indicated that the variables were non-stationary. Non-stationary data have a variance and a mean that changes over time, which will affect the results obtained from any statistical analyses that are performed on the data (Nason 2006). Non-stationary data can be converted by using a return that is calculated by the logarithm of a data point divided by the previous data point (Dahlhaus 1997). A drawback to the method of making data stationary is that one data point or sample member is lost in the process. If the sample had 45 points for instance, the regression analysis will run with only 44 points. The following equation is used to do the conversion from non-stationary to stationary data.

Calculating the natural logarithm and return for each variable has the benefit that variables, measured differently, become comparable, for example turnover as a total rand value and ROE as a percentage.

Descriptive statistics for the adjusted variables are presented in Table 2.

The skewness and excess kurtosis figures in Table 2 indicate that the data are not normally distributed. This is confirmed by significant Jarque–Bera test statistics for all series, despite the data being transformed by calculating the return of the natural logarithm of each variable. Departures from normality in financial time series are widely recognised in the literature and the results of the descriptive statistics are an indication of this (see Xiao and Aydemir 2007). Furthermore, all series are negatively skewed, except for changes in executive director remuneration, which is positively skewed.

A lack of normality calls for the use of a non-parametric test or bootstrapping. The use of non-parametric tests was deemed inappropriate for this study, therefore bootstrapping at a 95% confidence interval was applied. Bootstrapping also reduces the effect of a small sample size by adding robustness to the statistical tests (Field 2013). In addition, the data showed signs of heteroscedasticity, which also calls for the use of a regression model. The data show no significant sign of endogeneity, therefore the authors decided to ensure that the issue of heteroscedasticity is sufficiently addressed.

Table 3 presents the Pearson correlation results for the dataset making use of bootstrapping at a 95% confidence interval.

The results of the Pearson correlation (Table 3) indicated a clear correlation between total executive director remuneration (REM) and share price (PPS). Correlations with other financial performance measures were mainly negative and insignificant. A regression model would provide more insight into the relationship between executive director remuneration and share price.

An initial regression analysis on the data indicated by means of collinearity diagnostics, as well as the tolerance and variance inflation factor (VIF), that there is significant multicollinearity between ROA and ROE. For that reason, ROA was removed from the list of independent variables before commencing with the final regression model.

Conducting an analysis on time series data can include delayed effects because of the nature of financial data. The actions of executive directors do not necessarily have an immediate effect on the performance of the company and performance-based remuneration only occurs after said performance has been achieved. The Durbin–Watson test is used to observe the effect of autocorrelation in a dataset; for this study it was calculated at 0.946. This is a clear indication of serial correlation between the errors in the regression model. According to Ghalayini and Noble (1996), the use of time series data can necessitate the use of a lagging metrics effect. This means that linear regression cannot be applied on the data as is and that a lag needs to be introduced for the dependent variable. A lag of AR (1) was thus introduced to the dependent variable in the regression model to eliminate the effect of autocorrelation.

The variables for this study are all of a financial nature and therefore tend to be endogenous. Consequently, the use of only ordinary least squares regression could result in biased findings and erratic results as ordinary least squares regression ignores time-invariant unobserved individual effects and endogeneity (Flannery & Hankins 2013; Maddala & Lahiri 2009). Generalised least squares regression estimates unknown parameters in a linear regression model that can be applied when the variances of the observations are unequal or when there is a certain level of correlation between variables (Rana & Al Amin 2015). The introduction of a lagged effect in the dependent variable significantly reduced the endogeneity of the dataset.

Linear regression was applied to the adjusted dataset, with bootstrapping at a 95% confidence interval (CI) to correct for non-normality and to compensate for the small sample size by adding robustness to the analysis.

The regression model can be expressed as follows:

The results of the regression analysis are presented in the next section.