The Efficient Market Hypothesis formalised by Fama (1970) states that a financial market is efficient when prices reflect all available information. Specifically, three forms of efficiency were defined – strong, semi-strong and weak. In a market that displays strong-form efficiency, there are no information advantages to be obtained from insider trades, as all information is already reflected in the price of the share. However, these circumstances are viewed as unrealistic in practice. With regard to the JSE’s efficiency, researchers have concluded that it is not strong-form efficient (Thompson & Ward, 1995) and that there is room for prediction of share prices based on informational advantages (Nyakurukwa & Seetharam, 2023).

Market efficiency is relevant to this study, as directors’ dealings are publicly disclosed to the market in accordance with the JSE listing requirements. This study considers the market response to directors’ dealings in the context of the information environment and evaluates the market reaction in relation to information-specific factors to determine the firm-specific contexts in which directors’ trades are likely to confer significant informational advantages. Consistent with the efficiency theories discussed in Ryan (2005) and Nyakurukwa and Seetharam (2023), the present study is predicated on the notion that the richer the information environment, the lower the informational advantage to the market from directors’ dealings.

In the context of incorporating information into prices, the development of the Capital Asset Pricing Model or CAPM (Lintner, 1965; Sharpe, 1964), is closely intertwined with theories of market efficiency. The CAPM is an asset pricing model that specifies the relationship between risk and return. It follows, therefore, that the CAPM is used in an event study such as this one. Significant deviations from a CAPM-predicted return following an event can reasonably be inferred to be a result of the information content of that event.

The research surrounding directors’ dealings often draws on signalling theory and information asymmetry, which suggests that directors who have access to private information about a company signal this information to the market through their trading activities (John & Lang, 1991). Directors’ dealings are viewed as informative signals, and multiple studies investigate the profitability of directors’ dealings as signals by calculating the abnormal returns earned by directors on their trades. Many of these studies (Moodley et al., 2016; Neill et al., 2008; Ryan, 2005) have found significant associations between directors’ trades and abnormal returns on company shares, indicating the potential for informational advantages, even amid the relatively strong information environment and trading laws in South Africa (Barth et al., 2017; Ojah et al., 2020; Natesan, 2020; Thaver & Ward, 2011).

In the context of the JSE, Thaver and Ward (2011) examined whether the volume of shares traded could be an indicator of potential insider trading. They found that 34 out of 142 price-sensitive SENS announcements were associated with significant abnormal trading volumes prior to the release of the announcement (emphasis added). It is therefore unsurprising that the investment community shows strong interest in trading by insiders, driven by the belief that insiders possess valuable information about a company’s future prospects (Moodley et al., 2016; Nanda & Barai, 2020), so much so that these trades could be used in developing trading strategies for outside investors (Mordant & Muller, 2003).

The population for this research has been defined as all on-market directors’ dealings for companies on the JSE, as reported in the SENS announcements between 01 January 2022 and 31 December 2023. The recommendation of Ryan (2005), a two-calendar-year period is used to ensure that sufficient historical data can be collected for the estimation period, given the differing financial year ends. The period begins on 01 January 2022, as 2022 is the first year in which the South African economy displayed pre-pandemic levels of operating with minimal business and economic disruption as a result of the COVID-19 pandemic. According to Statistics South Africa (2022), the South African economy took almost two years to recover from the worst of the COVID-19 pandemic and in the first quarter of 2022, the South African gross domestic product rebounded to its pre-pandemic levels. Additionally, the 2-year period minimised the risk of having two directors’ trades that had different relationships with each of the indicators of the information environment caused by the trades occurring at different phases of the economic cycle. This approach helps ensure that the observed relationships of the market reaction to the directors’ dealings and their relationship with the information environment are not confounded by changes in the economic cycle.

The sample for this study includes directors’ dealings from 90 companies, comprising 30 small companies, 30 medium companies, and 30 large companies on the JSE. This size categorisation is important despite the use of ‘size’ as an indicator of the information environment. By having sub-samples of size, we reduced the possibility of the effects of trades in large companies masking the effects of trades in smaller companies.

The JSE’s Large Cap (J205), Mid Cap (J201) and Small Cap Index (J202) classifications were followed to create the sub-samples of size. These classifications are derived by the JSE according to a ranking system based on-market capitalisation (JSE, 2025). Each company’s classification is included in the EquityRT database. For the purposes of this research, the classification data were drawn directly from EquityRT. The 90 companies were selected from the pool of companies that reported the on-market directors’ dealings. Within each size category (small, medium and large), 30 companies were selected based on the presence of directors’ dealings during the event window.

After adjusting for confounding events, the data collection indicates 247 SENS announcements relating to dealings from large companies, 173 announcements from medium companies and 162 announcements from small companies.

An event study method was used to assess the market reaction to directors’ dealings, following the guidance in MacKinlay (1997) and applications thereof in similar event studies (Nanda & Barai, 2020; Nkongho & Makina, 2020).

The abnormal return was calculated as Equation 1:

The actual return for day t was expressed as a percentage and computed as Equation 2:

MacKinlay (1997) suggests two models that can be used to calculate the expected return: the mean return model and the market model. As in Ryan (2005), the present study uses the market return model. A market model was chosen, as it incorporates the entire market’s performance and avoids the situation in which abnormal returns are conflated with general market movements. Specifically, the CAPM, developed by Sharpe (1964) and Lintner (1965), was used to estimate expected returns (ER_i,t) as Equation 3:

where:

The abnormal volume (AV_i,t) was calculated as Equation 4:

where RV_i,t is equal to the actual trading volume of firm i on day t and EV_i,t is equal to the expected trading volume of firm i on day t.

The expected volume (EV_i,t) was determined as follows:

EV_i,t = the average volume traded over the estimation window from t-90 to t-6.

The stock price and trading volume reactions to directors’ dealings were measured by using the CAAR and the CAAV, respectively, over the event window.

The CAAR was calculated as Equation 5:

where N is equal to the number of observations within the sample, T is the number of days post the event date, and AR_i,t is the abnormal return for firm i on day t.

The CAAR were calculated for the anticipation window, the event date, and five days post the event date.

The CAAV was calculated as Equation 6:

where N equals the number of observations in the sample, T is the number of days post the event date, and AV_i,t is equal to the abnormal volume of firm i on day t.

The CAAV were calculated for the anticipation window, the event date, and five days post the event date.

The timeline of the entire event period consists of three periods, namely, the estimation window, the anticipation window and the post-event window, as shown in Figure 1 (not drawn to scale).

Given that this study investigates the stock price and trading volume reaction to directors’ dealings, a shorter estimation window of 90 days is used, which is similar to Nanda and Barai (2020). The anticipation window begins 5 days before the event date because that is the maximum amount of time allowed between the transaction date and the announcement of the transaction (JSE, 2019).

The SENS announcement is t₀ (event date) because before that, the trade is not known to the market. The short event window of 5 days is optimal for this study because it allows the focus to remain on the initial market reaction to the event without the influence of other events in subsequent trading days. Additionally, it has been empirically demonstrated by Ryngaert and Netter (1990) that a shorter event window will usually capture the significant effect of an event.

In line with Thalassinos et al. (2012) and Nanda and Barai (2020), a regression analysis was employed to determine the relationship between the stock price and trading volume reaction to directors’ dealings and the company’s information environment.

The indicators for the company information environment are defined as follows: Firm size is defined as the natural log of market capitalisation and is termed ‘SIZE’. The bid-ask spread is defined as the natural log of the median of the difference between daily closing bid and ask prices divided by their midpoint and is termed ‘BID-ASK’ (Barth et al., 2017). The number of analysts is defined as the number of analysts producing recommendations for our sample of firms and was termed ‘ANA’.

The following models (Equation 7 and Equation 8) were used:

where CAAR and CAAV are the natural log of the CAAR and volume, respectively for event j in company i at time t. SIZE, BID-ASK and ANA are the indicators for the company information environment for company i at time t.

The chosen model was derived from the model used by Ryan (2005), which incorporated firm size, number of analysts and absolute values of cumulative abnormal returns as dependent variables, while the price and trading volume movements driven by directors’ trades served as independent variables.

Confounding events were identified by comparing the actual returns of a firm to the index return of the industry that the firm operates in, and where the returns of the firm deviate significantly from the returns of the industry index is indicative of a possible confounding event. A significant deviation is defined as a return that is greater than three standard deviations (s.d.) from the mean index return, as 99.7% of observations lie between the mean and three s.d. (Mishra et al., 2019). Events that were identified by using the above method were checked for confounding events by investigating whether there were any SENS announcements released related to company actions that would influence the stock characteristics other than the event, such as earnings announcements, dividend declarations or corporate action announcements, such as mergers or divestitures.

The trading restrictions imposed by the JSE listing rules in paragraphs 3.63–3.74 help address the issue of confounding events by prohibiting directors’ trading during closed periods (see section 2). Given that directors will be unable to trade during these periods, there is a reduced likelihood of abnormal returns and volumes being calculated, which have been influenced by the events that take place during closed periods, such as earnings announcements.

In addition, the assumptions of regression were tested. This study tests for (1) an approximately linear relationship between variables, (2) multicollinearity, (3) homoscedasticity and (4) normality (Osborne & Waters, 2002). Linearity was tested by using scatter plots of the residuals; multicollinearity was tested by using the variance inflation factor (VIF) statistic to ensure that the degree of multicollinearity was sufficiently low to perform the regressions (Shrestha, 2020). The Breusch–Pagan test for heteroscedasticity was used to confirm the absence of heteroscedasticity (Klein et al., 2016). Data were tested for normality by using the Kolmogorov–Smirnov test and the normal Q-Q plot to verify that the data followed a normal distribution (Osborne & Waters, 2002; Yazici & Yolacan, 2007).

Ethical clearance to conduct this study was obtained from the University of the Witwatersrand Human Research Ethics Committee (Non-Medical) (No. WSOA-2024-07-22W).

Table 1 presents the final sample of SENS announcements and the number of dealings after adjusting for confounding events.

A single SENS announcement can announce more than one director’s dealing, which explains why the number of dealings exceeds the number of SENS announcements.

Table 2, Table 3 and Table 4 highlight a breakdown of all the directors’ trades recorded in the SENS announcements that were collected. There were 582 SENS announcements in the sample for this study, which recorded 1202 directors’ dealings. In the tables, all transaction values are provided in terms of millions of rand.

In general, the data indicate that the value and size of transactions are positively related to company size. This relationship is shown in the overall average value of transactions in large companies, R13.08 million, being significantly higher than the average value in small and medium companies of R2.76m and R2.81m, respectively. This outcome could be attributed to larger companies having a higher share price and greater liquidity of stocks compared to those of small and medium companies, meaning that each transaction in a large company would have a higher total value, contributing to the higher average.

The remuneration policies of companies which provide share-based incentives to directors further encourage more frequent transactions by directors. As executive directors, including the Chief Executive Officer (CEO) and Chief Financial Officer (CFO), are involved in the daily operations, their remuneration packages often include share-based incentives to encourage better performance by directors, which benefits the stock price as well as their remuneration packages (Baker et al., 1988). These incentives are common in JSE-listed companies and motivate directors to engage in transactions to maximise personal gains, adjust risk exposure or convert them to cash for liquidity purposes, such as to settle taxes related to the receipt of these stock awards. The increased participation in share-based remuneration schemes and the related transactions may explain why executive directors, including the CEO and CFO, have more frequent transactions than non-executive directors do.

More specifically, the chairman of small and medium companies has the least number of transactions, with 12 and 6 transactions, respectively, yet they show the highest average transaction value. A similar result is present in large companies in which the chairman has the fourth-highest average transaction value despite having the least number of transactions (17). This behaviour is possibly explained by the role and influence of the chairman’s position. As a high-ranking director, the chairman typically prioritises longer-term strategic gains over more frequent market activity, resulting in less frequent and higher-value trades (Del Brio et al., 2018). Additionally, the reason for trades by the chairman is often the rebalancing of their investment portfolio, which generally occurs annually. The rebalancing of an investment portfolio typically requires less frequent but larger trades to maintain an intended asset allocation, reduce transaction costs and enhance long-term returns, which explains the less frequent but higher-value transactions by the chairman (Brennan et al., 1997).

An analysis of the purchases and sales by directors produces several interesting insights relating to the trends of directors’ dealings. It is apparent that the number of sales greatly outweighs the number of purchases, evidenced by the consistently higher average number of sales in small, medium and large companies, respectively. Given the increasing amount of equity-based remuneration received by directors, it is expected that some or all the stock received will be converted into cash, depending on the director’s liquidity needs, resulting in a greater number of sale transactions.

The purchase transactions are dominated by CEOs in small companies. In contrast, large companies exhibit a different dynamic with independent non-executive directors trading more frequently than CEOs, who are only ranked fourth in terms of frequency of purchases. This shift may be attributed to the more complex governance structures and diversified management teams in large companies.

The greater number of executive director sale transactions is explained by the size and management structure of medium and large companies. As these companies are larger, they have more individuals in directorship positions and subsequently more individuals with equity-based compensation. Consequently, the higher number of sales transactions reflects the greater number of directors present in medium and large companies.

The dominance of CEOs is not surprising given that CEOs are usually better compensated than other directors, allowing them to engage in both purchase and sale transactions of greater value and size (Wang et al., 2012).

The CAARs and CAAVs, which were calculated from the event study, were tested for a significant statistical difference.

The results in Table 5 show that sale transactions generate higher abnormal returns than purchase transactions do. The sale transactions showed positive abnormal returns across the anticipation window, event day and post-event window; however, none were statistically significant. Conversely, the purchase transactions showed statistically significant negative returns for the anticipation window and the event day, as shown in Table 6. This result suggests that the market has a significant negative reaction before and on the day of a purchase transaction. These results are partially consistent with Boubacar and Morris (2011), who present results finding significant negative CAARs following purchase transactions.

However, the results of this study show that significant negative returns are present before and on the day of the announcement of the purchase reaction. We posit that our findings are possibly a result of incentive plans, which result in directors acquiring shares (which is seen as a purchase by the market). Given that these stocks are awarded to directors as part of their compensation, this ‘purchase’ does not provide as strong a signal compared to when a director conducts a purchase on the open market. An open market purchase is typically associated with a positive signal, as the director is risking their own money in this stock, which attracts market attention. Therefore, the negative results presented in this study may also be a result of negative information leakage or market speculation, which resulted in a negative market reaction before the event.

The results in Table 6 show that sale transactions present statistically significant abnormal volumes during the anticipation window and the event day. The significant abnormal volumes present during the anticipation window and the event day of sale transactions suggest that the market anticipated the sale transactions before their announcement. This outcome is possibly a result of the market anticipating certain sale transactions following certain events, such as the sale of shares following the vesting of shares in share incentive schemes.

In Table 7, the presence of these significant returns during the estimation and event window, respectively, suggests that the market views the announcement of directors’ dealings as new, relevant information and that this information was possibly anticipated prior to the event. This effect further highlights that there are possibly higher levels of information leakage in smaller companies, which allows the market to anticipate the event in small companies.

The post-event window showed negative average abnormal returns for small and medium companies and positive average abnormal returns for large companies. The large p-values indicate a lack of significance, which suggests that there is no continued reaction to the director’s dealings after the event date. These findings are consistent with those of Mordant and Muller (2003), who found that abnormal returns from directors’ trades were caused by market factors.

In Table 8, the estimation window shows that the average abnormal volumes were positive across small, medium and large companies, but none of these abnormal volumes were statistically significant. This finding indicates that there was no market anticipation or anticipatory trading prior to the event. Similarly, the abnormal volumes of small, medium and large companies were positive for the event day, but none were statistically significant. In their event study, Thaver and Ward (2011) found that 34 out of 142 price-sensitive announcements were preceded by significant trading volumes. However, they concluded that although trading volume could be an indicator of insider trading, it should be viewed in conjunction with other indicators. Our findings support this conclusion.

The post-event window shows statistically significant positive abnormal volumes for medium and large companies, while small companies exhibit positive abnormal volumes that are not statistically significant. This finding may be a consequence of institutional investors dominating trading in medium and large companies, as they tend to react quickly to new information, which creates spikes in trading volumes that are significant (Edelen et al., 2016).

In this section, the results of the regression analysis are presented by size category (small, medium and large firms). Findings regarding each indicator of the information environment are discussed for each size category.