Amsterdam Business School
The effect of CEO turnover on the analysts’ forecasts and the
influence of board of directors.
Name: Guo Hu
Student number: 10885439
Thesis supervisor: Alexandros Sikalidis Date: 14 August 2016
Word count: 9000
MSc Accountancy & Control, specialization Accountancy Faculty of Economics and Business, University of Amsterdam
Page 2 of 33
Statement of Originality
This document is written by student Guo Hu who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
Page 3 of 33
Abstract
This paper examines how a CEO turnover affects the quality, quantity and dispersion rate of the analyst forecasts. Furthermore, I assess whether and how the quality of the board of directors influences the analysts in the year before, during and after a CEO turnover. In order to investigate this, 297 firms with a CEO turnover in the years between 2009-2014 will be examined and categorized into forced and natural turnover. Also, 3445 unique analyst-firm combinations will be examined. The quality of the board of directors will be split up in three variables: size, expertise and independence of the company. This paper finds significant differences in the quality, quantity and dispersion rate over the three years compared. Suggesting that a CEO turnover has a negative effect the quality of forecasts, the number of forecasts and the dispersion rate between the analysts. Furthermore, it finds that the board of directors has a moderating effect for the quality of forecasts and the dispersion rate. This would imply that that a CEO turnover (both forced and natural) has negative effects on the analysts’ forecasts and that this can be reduced by a stronger board of directors.
Page 4 of 33 Contents 1 Introduction ... 5 2 Literature review ... 7 3 Hypothesis development ... 11 4 Research design ... 14 5 Conclusion ... 27 6 Bibliography ... 29 7 Appendices ... 31
Page 5 of 33
1 Introduction
Financial analysts are important intermediaries in the capital market. Their forecasts have a big impact on investor’s decisions and stock market. They try to remove the information asymmetry between investors and management, by either providing the publicly known information to the right parties or using private information in their (earnings) forecasts. Considering analysts might have access to inside information, the stock market immediately reacts to their forecasts because they contain new information.
Analyst forecasts also affect managers’ actions. Researchers found a significant amount of zero and small positive earnings surprises (abnormal earnings) (Burgstahler & Eames, 2006). This is related to earnings management in the form of real operating decisions and accruals. Also management issues pessimistic management forecasts to dampen analysts’ forecasts (Baik & Jiang, 2006). Baik and Jiang’s reasoning for this are to lower the earnings to achievable levels for their own performance. Which in return is to maximize their own stock rewards (Aboody & Kasznik, 2000).
Another reason to dampen the analyst forecast can have to do with the post-earnings announcement drift. Prior studies have shown that the stock market will continue to drift in the direction of the results released, but only if it is unexpected (Ball & Brown, 1968; Bernard & Thomas, 1989). The projected results are usually based on the market consensus and the analysts’ forecasts. Hence if the company manages to beat the forecast’s predictions, it will result in an increase of stock prices and abnormal results. Moreover, this effect will last much longer due to the post-earnings announcement drift. However, the effect of a negative result will also be prolonged. Thus a small positive earnings surprise can result in an increase in future estimated abnormal returns and managers will try and prevent a negative earnings surprise.
Many situations can affect an analyst’s forecast. For example, corporate governance transparency of a company can positively influence analyst forecasts’ accuracy (Bhat, Hope, & Kang, 2006). Or as said before the (voluntary) disclosures and earnings forecasts by management will be used in analyst’s forecasts too (Baik & Jiang, 2006). Moreover, research has shown that analysts influence each other and calls this herding behavior (Trueman, 1994).
Furthermore, events can change an analyst’s forecast. These events can lead to changes in an analyst’s forecast throughout the fiscal year. These changes are called forecast revisions and provide information about corporate’s earnings (Gleason & Lee, 2003). They acknowledge the importance of the revisions due to the investor’s direct reaction to these revisions. Standard events like quarterly announcements can result in a revision. However, revisions during non-standard events provide even
Page 6 of 33 more information, because it might indicate that the analyst has private information. But unforeseen events can also change the forecasts, like a (forced) CEO turnover. Thus, it is interesting to understand how forecasts react to unforeseen events and in specific environments.
A recent study analyzed the effects of (forced) CEO turnover on the analysts forecast (Choi, Chen, Wright, & Wu, 2014 ). They concluded that the analysts are less accurate and more optimistic about earnings after a forced CEO turnover. However, they only analyzed the forecasts around the financial year-end of the year when the CEO changed. They indicate that there’s a lack of research in the pattern of the earnings forecast over a longer period of time after a CEO change. This paper will respond directly to the research of Choi, Chen, Wright and Wu (2014).
Furthermore, the board composition will be taken into account for the research. This is to find out whether a strong board of directors can counteract the negativity around a (forced) CEO turnover. Assuming that stronger corporate governance will lead to a more accurate forecast (Bhat, Hope, & Kang, 2006), and a forced CEO turnover means less accurate forecasts. It can provide more insight in the behavior of the analyst and what information they use and deem important about a firm. The research question that will be answered is: to what extent does the quality of the board of directors influence the analyst’s forecasts after a forced CEO turnover.
The paper is structured in the following way to answer the research question. In the next section a theoretical background of the subject will be described. Following in the third section will be the hypothesis development. The fourth section will provide the sample selection and statistical data. This paper concludes in the fifth section.
Page 7 of 33
2 Literature review CEO Turnover
CEO’s are in the top management level and are supposed to meet the demands of, for example, the investors, customers and the firm itself. They do this by giving direction to the strategy, by prioritizing certain projects and leading the senior executive team. Thus they hold considerable power in a firm and are responsible for the firm’s results. A change in CEO might result in a change in direction and strategy of the firm and thus creating more information asymmetry between the firm and the stakeholders.
Prior studies have mostly made a distinction between forced CEO turnover and a natural or voluntary CEO turnover. This is because the board of directors usually intervenes in bad times and will force a new CEO to step in, whereas a voluntary CEO turnover has nothing to do with the firm’s performance. This distinction is being made because of the (performance) information it provides to the outsiders and the different effect it has on the market.
This is because the board decides to let go of the CEO if the quality of the CEO falls below their expected quality (Jenter & Kanaan, 2015; Weisbach, 1988; Farrell & Whidbee, 2002). However, they also mention that the board structure also influences the rate of forced CEO turnover. A stronger board of directors will be more critical in the performance of a CEO and thus increasing the likelihood of a CEO turnover during bad performances. In any cases the board hopes to resolve agency conflicts and a better performance by replacing the CEO. This by possibly changing the future direction of the firm and probably the strategy to achieve this.
However, a forced CEO turnover comes with negative publicity. This might even become a bigger problem than the change of CEO itself or the changes in strategy. It might even imply the initial quality of the board of directors, because they have chosen the fired CEO in the first place. For these reasons, the board will try its hardest to prevent negative publicity by announcing that there was already a succession plan in place or that there will be a transitioning period to the new CEO. These plans will temporarily keep the old CEO in place in a more advisory role.
Clayton, Hartzell and Rosenberg (2005) have also shown that the stock prices are a lot more volatile after a forced CEO turnover in comparison with a voluntary turnover. They hypothesize that it is because of the uncertainty regarding the CEO’s ability to run the firm and the CEO’s decisions
Page 8 of 33 about the direction the firm should take. It is known that the lack of information and higher information asymmetry will lead to more volatile and lower stock prices (Healy & Palepu, 2001).
At the same time a voluntary CEO turnover says nothing about the current performance of the CEO and the firm. Thus the board didn’t have a need to change of the CEO. Possible reasons could be sickness, age or a new opportunity. Therefore, there is no reason to believe that the upcoming CEO will (drastically) change the direction and strategy of the firm. However, there is still uncertainty in the market regarding the new CEO. This uncertainty is evident in the volatility of the stock prices (Clayton, Hartzell, & Rosenberg, 2005).
Board of directors
The board of directors is a board selected by the shareholders and has the highest authority in the management of the corporation. They serve the shareholders and have different duties that are important for the organization and the corporate governance. One of their duties is to appoint and monitor the CEO (Weisbach, 1988). Furthermore, they govern the organization by setting up certain policies and they are the bridge between the shareholders and the firm’s performance.
Because of the importance of the board of directors, there have been many studies relating to them. Studies have shown that the board of directors plays an important part in the quality of the corporate governance in a firm (Fama & Jensen, 1983). They argue that outside directors have more incentives to monitor the management, because they are not part of the company. A reason for this is that directors wants to show they are the experts in their field and that they take their job seriously. Further studies have shown that board independence (Klein, 2002), outside directors (Peasnell, Pope, & Young, 2005) and board members with financial or corporate background (Xie, Davidson III, & DaDalt, 2003) leads to less earnings management and thus a more trustworthy information of the firm.
A stronger board of directors leads to a more reliable and stronger corporate governance (Khanchel, 2007; Bhat, Hope, & Kang, 2006). They give two important reasons why corporate governance is important for financial analysts. The first reason is that the financial disclosures of the management are more reliable. The second reason is that the corporate governance can reduce uncertainty surrounding future performance. Multiple studies have shown that investors are willing to pay extra for good corporate governance. Due to the fact that (in theory) a strong corporate governance will have less agency costs, which will lead to a better firm performance. Thus the board
Page 9 of 33 of directors has a say in a CEO turnover and should have to some extent influence on the quality of the analyst forecasts.
Analyst forecast revisions
Analyst forecasts are an important source of information for both investors and the firms. Firms can benefit a lot from the forecasts. If the firm manages to beat the analyst forecast, it will benefit the management and possibly increasing the stock prices. This effect is so beneficial for management, that they might even manipulate analysts (Richardson, Teoh, & Wysocki, 2004; Baginski & Hassell, 1990; Burgstahler & Eames, 2006). However, this does not mean that the forecasts are not useful. Because investors are still using analyst forecasts as an important source of information, it is a way to decrease the information asymmetry between the investor and the firm. The reason for this is because analysts might have non-publicly disclosed information about the firm and incorporated that in its forecast.
Changes in an analyst forecasts are being called forecast revisions, and the revisions itself can also be seen as highly informative. To find the added value of a revision, it is necessary to distinguish them between a low-innovation revision and a high-innovation revision (Gleason & Lee, 2003). A low innovation is simply a revision based on the forecasts of other analysts, by moving towards the consensus forecast. A high innovation revision is a revision that provides new information to the investors, as such this will also decrease the information asymmetry. This is the mostly the case after a (forced) CEO turnover.
To determine the quality of the analyst forecast revision you can look at the analyst’s prior forecasts and use that as the benchmark for that same analyst (Gleason & Lee, 2003). This will better show the quality of the revisions and the possible reasons behind the revisions. Furthermore, the amount of revisions can say a lot about the situation. Because the quantity shows how many times the analyst feels necessary to change their forecast and indirectly saying they got new information.
Analyst forecast dispersion
Forecast dispersion is the differences in the forecasts between analysts. Even though this doesn’t measure the earnings uncertainty (Johnson, 2004), it is still being used to determine the uncertainty of the information environment. Johnson assumes that analysts don’t think about their colleagues and competitors in the field and that they create their own forecast to their own best knowledge. However, whether this is always the case remains a question, because the users of the forecasts will compare the
Page 10 of 33 forecasts. As said before some analysts might have the tendency to put out forecasts towards the consensus. But it is still very usable as a proxy for information asymmetry between the analysts and the firm and the quality of the information environment.
Studies have shown a relationship between analysts’ forecasts dispersion and stock value. Diether, Malloy and Scherbina (2002) uses analyst forecast dispersion as a proxy of differences in opinion about stock and basically uncertain earnings. They have found that a higher forecast dispersion results in lower future stock returns, because the investors pay a premium price due to information asymmetry. Therefore, a negative relationship between dispersion and future returns has been found. However, Johnson (2004) has another reason for this negative relationship, and this is the increase in option value of the firm due to uncertainties. More evidence seems to support Johnson argument that the forecast dispersion is due to uncertainties (Barron, Stanford, & Yu, 2009). However, they do show that an increase in dispersion does mean more information asymmetry.
Page 11 of 33
3 Hypothesis development
Looking at the current theory regarding the effect of an unexpected event like a CEO turnover on the analyst forecast, a similar result can be expected. That is an increase in analyst forecast revisions or the quantity of forecasts. Under the assumption that analysts want to provide quality information and accurate forecasts, they will have to acquire new information. Because of this extensive information gathering process, they have more opportunities and possible reasons to adjust their forecast. Thus assuming that there will be an increase in analyst forecasts revisions after a CEO turnover.
Not only will the quantity increase, the quality of the forecasts will also be affected by the CEO turnover. The turnover itself provides the analysts with information that the leader will change and possibly the performance of the company. But a forced turnover will be usually related to bad performances, which in turn could mean that the current forecasts of analysts are instantly inaccurate. Furthermore, due to the lack of information after a CEO turnover it will become more difficult to provide their users with high quality forecasts. Therefore, we assume that the forecast quality will drop.
Regardless of the type of CEO turnover, an increase of analyst revisions will still occur. This is mainly due to the fact that the analysts do not know about the skills and capabilities of the new CEO. Therefore, information asymmetry will exist and new information will have to be gathered, thus increasing the revisions. However, we do expect that the revisions will be even larger for forced CEO turnover situations. Because analysts also have to consider the fact that the firm is not performing well enough and that the company will have to change their strategy and plans.
An increase in analyst forecast quantity and a decrease in analyst forecast quality after a CEO turnover will be expected. However, we are not sure whether this happens in the year of the CEO turnover or the year after and whether this effect persists in the first case. We formulate the following hypotheses:
H1a: There is a significant change in analyst forecasts quantity after a voluntary CEO turnover took place
H2b: There is a significant change in analyst forecasts quality after a voluntary CEO turnover took place.
H1b: There is a significant change in analyst forecasts quantity after a forced CEO turnover took place H2b: There is a significant change in analyst forecasts quality after a forced CEO turnover took place.
Page 12 of 33 After looking at the individual behavior of the analysts, we want to see the effect the CEO turnover has on the dispersion rate of the analysts. As said before the analysts will have a difficult time to assess the firm’s future performance due to the (possible) changes in the CEO’s ability and the changes in the strategy. It will be interesting to see how the analysts will perceive these changes and their ability to predict the earnings per share based on the information they have and can acquire. Because each analyst might have a different source for their information and might interpret the same information differently.
The prediction is that the dispersion rate will differ between a forced and voluntary turnover. A forced CEO turnover happens when the board thinks the CEO doesn’t perform anymore, therefore a change will result in both the leaderships ability and the strategy the firm will take on. This creates a huge amount of information asymmetry between the analysts and the firm. One analyst might think this will boost the firm’s performance over time because of a fresh CEO. However, another analyst might determine the opposite or another analyst knows the ability of the new CEO and can therefore give a more educated forecast. Due to high number of factors analyst have to take into account the prediction will be that the dispersion rate will be higher when there was a forced CEO turnover.
This won’t be expected during and after a voluntary CEO turnover. Under the assumption that there won’t be any significant changes in the performance of the firms, there would be no need to expect that analysts’ forecasts will suddenly drift apart.
H2a: There is no significant difference in forecast dispersion after a voluntary CEO turnover took place
H2b: There is a significant difference in forecast dispersion after a forced CEO turnover took place.
Next we will look at the quality of the board of directors and their influence on the analysts’ forecasting capabilities. The reason for this is because they play a big role in the supervision and selection of the CEO. But they also play a big part in the quality of the corporate governance of a firm. A stronger corporate governance will mean a more trustworthy firm that works responsibly. In that case third parties can rely more on the performance of the firm and the information it provides to their stakeholders. The influence of the board of the directors may certainly affect the analysts.
A forced CEO turnover happens under the direction of the board of directors. Whether this is a good or bad decision depends on the quality of the board. The board will most certainly try to get
Page 13 of 33 the most capable CEO to run the firm. If the stakeholders can trust the board on this decision a better future will be expected. Analysts will still have to take into account the sudden turn of events and the possible changes in the CEO’s ability and the firm’s strategy. But in this case they can trust on the board’s decision and therefore limiting the panic surrounding a forced CEO turnover.
The same effect can be assumed in the event of a voluntary CEO turnover. The appointment of a CEO is still done by the board of directors, thus maintaining the assumption that the ability to run the firm is more than capable of the new CEO. However, there is no reason to believe that the strategy of the firm will drastically change after a voluntary CEO turnover. Apart from the changes, a stronger board will still lead to a more reliable and comprehensive information output of the firm. Thus, the analysts have even less factors to keep in mind that might set them on the wrong foot. Also in this case a stronger board will lead to a higher quality forecast in comparison with a weak board.
H3: The effect of analyst forecasts revisions, quality and dispersion is being diminished by a stronger board of directors in the year of a CEO turnover and the year after.
Page 14 of 33
4 Research design Sample selection
Our sample consists of data of U.S firms and analysts ranging between the years 2008 and 2015. The sample will be extracted from the WRDS datasets. Most recent data will be used to get the most relevant information and not be influenced by old regulations. Furthermore, no distinction will be made based on industry or size.
The first step is to gather data on whether a company had a CEO turnover in these years through the use of Compustat. However there’s no information on the reasoning behind the CEO turnover. This classification has to be done manually and based on public information. Because this classification is very subjective, we make use of certain keywords to identify the type of turnover. However most companies will announce a turnover, but won’t reveal the reason for it. For this reason we consult public press articles from third parties, seeming they should be more objective (Choi, Chen, Wright, & Wu, 2014 ).
A voluntary turnover could have many different reasons. Most common are retirement due to age, but there is also the possibility of sickness, death, personal reasons or a better business opportunity elsewhere for the CEO. Moreover, the change could also be part of a business plan, or due to personal wishes. In almost all voluntary turnover cases they will announce that the change is based on a planned succession. Meaning they have already prepared for the change.
However this planned succession could also be a mask to hide a forced CEO turnover. Because there is almost no way to prove that this turnover is truly planned beforehand. And the company obviously wants to prevent a negative press release and therefore will place the turnover under this context. This is a big reason why we will look at public articles instead of the company’s announcement.
To determine a forced CEO turnover we will look at certain keywords in a press release of the company or a public article. They keywords that will be used are mostly negative, for example: scandal, bad performance, issue, abrupt, sudden, unexpected. A substantial amount of forced CEO turnovers are regarded as abrupt or sudden for the outsiders.
Based on this list of companies with a CEO change we will gather the analyst forecasts through the use of IBES. By using detail history all the forecasts of a specific firm within a specific data can be found and used. Moreover the actual EPS value is also available in the same fashion.
Page 15 of 33 ISS (RiskMetrics) will be used to gather information about the board of directors. We will gather the data of the companies for the year of the CEO change. Furthermore ISS makes it
possible to gather data of all the board members and their affiliation with the company and whether they are financial experts. This data will be used to determine the quality of the board of directors.
Regression models
This paper investigates whether the CEO turnover and the typing of the CEO turnover has any effect on the forecast quality and the number of forecasts per analyst. Also it investigates whether the CEO turnover has any effect on the dispersion rate per company. Moreover, the board of directors will be taken into account and see if they have any effect. The regression models are shown below:
REVQUALt = α0 + β1*TURNOVERt + β2*OUTSIDEt2 + β3*EXPERTISEt2 + β4*BOARD SIZEt2 + 1
REVQUANt = α0 + β1*TURNOVERt + β2*OUTSIDEt2 + β3*EXPERTISEt2 + β4*BOARD SIZEt2 + 1
DISPERSIONt = α0 + β1*TURNOVERt + β2*OUTSIDEt2 + β3*EXPERTISEt2 + β4*BOARD SIZEt2 + 1
REVQUAL = The quality is based on the accuracy of the analyst forecast. The quality number shows the relative forecasting error by using the average forecast of an analyst’s entire year compared to the actual EPS value of the same year.
REVQUAN = The quantity is the relative number of forecasts of an analyst in a year compared to the total number of forecasts of all observed years. DISPERSION = The forecast dispersion is the standard deviation of the value of all the
forecasts of one company in a year compared to the average forecast of the same year.
FORCED = This is an indicator variable to determine the type of CEO turnover. Value is 1 if it is a forced CEO turnover, otherwise it is a 0 (for a voluntary turnover)
TURNOVER = This is an indicator variable to determine the year in which a CEO turnover (forced and voluntary turnover) took place. Value is 1 if a turnover took place, otherwise it is a 0.
Page 16 of 33 OUTSIDE = The relative number of independent members in the board of directors
of a company in the year in which a CEO turnover (forced and voluntary turnover) took place.
EXPERTISE = The relative number of members who are considered a financial expert in the board of directors of a company in the year in which a CEO turnover (forced and voluntary turnover) took place.
BOARD SIZE = Absolute value of the number of board members in the board of directors in the year in which a CEO turnover (forced and voluntary turnover) took place.
The regression models above are testing whether a CEO turnover has any effect on the models on separate datasets where we split the forced CEO turnovers with the voluntary CEO turnovers. However, we also want to measure whether the type of turnover has any effect on the models, for this we replace the TURNOVER variable with the FORCED variable. Furthermore, we will also examine the changes over the years per analyst. By using an analysts own benchmark we can more clearly see the surprise in the event and the effect (Gleason & Lee, 2003).
Variable measurement
In this study we want to look at the quality and quantity differences per analyst in three time periods1.
To determine which forecasts, belong in which year, we look at the period for which the forecast is based on. Meaning a forecast can be issued a year before the time frame, but will be taken into the database. Because the forecast will be based on the performances of that financial year. Also this makes it possible to see how many forecasts given out for the performance of one book year, making this more comparable with other years and with other analysts.
The quality number is calculated as the relative error compared to the actual value of that year. Meaning a higher QUALITY value shows a higher difference between the forecast and the actual value, which in turn would show that the quality has dropped. A lower QUALITY will mean less error and thus higher quality (accuracy). For the sake of clarity all numbers are shown as the absolute value, making all negative ratios positive.
The dataset is slightly different for the analysis of dispersion, but based on the same data as for the quality and quantity revisions. The dataset has to be summarized per company. Furthermore, the value of dispersion will be transformed into a ratio number based on the standard deviation and
1 Three time period are three years and are coded. T
1 is year before the CEO turnover, T2 is the year in which a CEO
Page 17 of 33 average value of the forecasts. This change will make it possible to compare different companies with each other and not worry about the actual dispersion sizes. Also here all the dispersion ratio’s will be shown has a positive number.
To determine the quality of the board of directors we use three components out of Khanchel’s study (2007). These are the directors size, the financial expertise and outside directors in the board. Considering that the directors size differs per company, we want to make the financial expertise and outside directors of a board more comparable. For this reason, we take the relative number of outsiders and experts.
Descriptive statistics
The sample size consists of 297 unique companies of which 44 companies have undergone a forced CEO turnover and 253 companies have had a voluntary CEO turnover. Under the 297 companies lies 3445 unique analyst-company combinations of whom we have gathered and summarized the forecasts of three years. Of the 3445 analysts, 466 has analyzed a company in which a forced CEO turnover took place and 2979 has analyzed companies with a voluntary CEO turnover. Furthermore, we’ve winsorized the dependent variables quality to a maximum of 2, and the depended variables dispersion to a maximum of 1.
The average quality shows an increase in every year, meaning a higher forecast error and thus a lower quality. Interestingly the minimum of quality is also 0, meaning the analyst has an average forecast value of exactly the actual EPS value. This seems to be entirely coincidental. Also both REVQUAN and DISPERSION seems to show a higher mean value in year 2 (year in which a CEO turnover took place) than year 1 or year 3.
The board of directors shows a huge variety in size, expertise and outside directors. The board size can vary between 4 directors and 32 directors, however the mean shows an average of 10 directors in the board. Also all board of directors in our sample consists of at least 50% independent members, meaning the independence is being sought out to some extent. The same cannot be said about financial experts in the board, showing that some boards have exactly 0 financial experts in their board. However, on average, at least 20% are financial experts, which is a positive indication.
Looking at the total sample size we see a difference of 12% quality between forced and voluntary turnover. This shows that in this sample that the quality is worse for a forced turnover. The quantity percentage of 33,33% is voluntary due to the calculation of this variable, because the three years combined adds up to 100%. And it also shows a higher dispersion value of 8% for forced CEO
Page 18 of 33 turnover, possibly indicating that a forced CEO turnover has a relationship. Also there seems to be very little differences between the board of directors’ size and composition.
Variables N Minimum Maximum Mean Std. Deviation
REVQUAN t1 3445 0,0417 0,8182 0,32555 0,10619 REVQUAN t2 3445 0,0526 0,8 0,353612 0,097415 REVQUAN t3 3445 0,0357 0,7143 0,320838 0,102949 REVQUAN t1 3445 0 1,9971 0,115427 0,229414 REVQUAN t2 3445 0 1,9771 0,125948 0,239612 REVQUAN t3 3445 0 1,9845 0,126393 0,231524 FORCED 3445 0 1 0,14 0,342 BOARD SIZE t2 3445 4 32 10,43 2,781 EXPERTISE t2 3445 0 0,6667 0,213932 0,119843 OUTSIDE t2 3445 0,5 0,9333 0,796761 0,10142 DISPERSION t1 297 0,0014 0,9899 0,075603 0,143205 DISPERSION t2 297 0,0005 0,9596 0,077009 0,143836 DISPERSION t3 297 0,0012 0,8144 0,064952 0,113442 Table 1. Descriptive statistics of entire sample divided per year
Variables Forced CEO turnover Voluntary CEO Turnover
N 1398 8937 QUALITY 0,2212 0,1072 QUANTITY 0,3333 0,3333 DISPERSION 0,14474 0,0630 SIZE 10,8498 10,3662 EXPERTISE 0,2141 0,2139 OUTSIDE 0,7953 0,7970
Page 19 of 33
Correlations between variables
Table 4 summarizes the entire sample used for both datasets. It seems that the quality (QUALITY) variable is significantly correlated with the type of turnover (FORCED). However, the fact that a turnover happened in the second year doesn’t seem to have an interaction. Furthermore, there is a significantly negative correlation between the quality and the directors size (SIZE), director’s
expertise (EXPERTISE) and the outside directors (OUTSIDE). This suggests that that the board of directors are taken into account by analysts. Almost the same significant correlations can be found against the dispersion rate (DISPERSION). The difference is that the director’s expertise is not statistically significant, however it shows the same coefficient as with quality.
The amount of revisions (QUANTITY) seems to have a correlation with the turnover itself and not the type of turnover. Due to the design of the dataset of quantity, a correlation between other variables are not shown. For this reason, the quantity dataset will be split up per year as shown in table 5. These table shows no significant correlation between the type of turnover and the
quantity variables. Moreover, the correlation between the different variables of the board of directors and quantity does not show many significant correlations. Also the variables that are significant correlation shows different kind of effects suggesting it to be random.
The correlation coefficients between the dependent variables and independent variables seems to be in line with our predictions, with the exception of the dependent variable quantity. Showing that the board of directors has a negative correlation with the dependent variables and that the type of turnover and the turnover itself has a positive correlation.
Page 20 of 33 Correlation table
Quality Quantity Dispersion
rate
Turnover Type Turnover Directors
size Directors Expertise Directors outsiders Quality 1 ,083** N/A ,003 ,095** -,215** ,013 -,100** Quantity ,028** 1 N/A ,123** ,002 -,001 ,001 ,001
Dispersion Rate N/A N/A 1 -0,028 ,190** -2,37** 0,36 -,126**
Turnover ,010 ,139** 0,009 1 0,000 0,000 0,000 0,000 Forced ,167** ,000 0,199** ,000 1 -,008 ,011 -,019* Directors size -,125** ,000 -0,125** ,000 ,059** 1 -,243** ,218** Directors Expertise -,021* ,000 -0,32 ,000 ,001 -,183** 1 ,094** Directors outsiders -,055 ** ,000 -,055 ,000 -,006 ,192** ,137** 1
**. Correlation is significant at the 0.01 level (2-tailed). *. Correlation is significant at the 0.05 level (2-tailed).
Table 3. Pearson and Spearman correlation of the entire sample. Whereas the lower data triangle under the 1’s is Pearson’s correlation data and the upper data triangle above the 1’s is Spearman’s’ correlation data. This figure is a summarization of two different datasets, which is the reason there is no data correlation available between dispersion rate and Quality/Quantity.
Correlation table of quality and quantity divided by year
Table 4. Pearson and Spearman correlation of quantity and quality variables divided per year. Whereas the lower data triangle under the 1’s is Pearson’s correlation data and the upper data triangle above the 1’s is Spearman’s’ correlation data.
QUANTIT
Y t1 QUANTITY t2 QUANTITY t3 QUALITY t1 Y t2 QUALIT QUALITY t3 Type turnover Directors Size Directors Expertise Directors Outsiders QUANTITY t1 1 -,464** -,552** ,170** ,033* ,009 ,010 ,036* -,063** ,028 QUANTITY t2 -,491** 1 -,395** -,061** ,085** ,016 ,010 -,010 ,022 -,038* QUANTITY t3 -,566** -,439** 1 -,111** -,098** -,010 -,016 -,030 ,047** ,009 QUALITY t1 ,094** -,032 -,067** 1 ,396** ,317** ,085** -,235** ,003 -,072** QUALITY t2 ,045** ,042* -,086** ,273** 1 ,400** ,113** -,215** ,066** -,064** QUALITY t3 ,015 ,042* -,055** ,267** ,381** 1 ,084** -,197** -,027 -,164** Type turnover ,004 ,011 -,015 ,072** ,291** ,133** 1 -,008 ,011 -,019 Directors Size ,031 ,010 -,041* -,141** -,094** -,141** ,059** 1 -,243** ,218** Directors Expertise -,061 ** ,027 ,037* -,026 -,038* ,004 ,001 -,183** 1 ,094** Directors Outsiders ,024 -,029 ,003 -,085 ** -,011 -,072** -,006 ,192** ,137** 1
**. Correlation is significant at the 0.01 level (2-tailed). *. Correlation is significant at the 0.05 level (2-tailed).
Page 21 of 33
Regression results
This study investigates whether the event of a CEO turnover affects the quality, quantity and dispersion rate of the analyst forecasts. The first step is to examine whether the event of a turnover and the type of turnover has an effect on the dependent variables. Next is to examine the event of a turnover per type of turnover, to see if the effects are different in another situation.
First of all, we find that the type of turnover (FORCED) is significantly positively associated with the quality of the forecasts (QUALITY) (B = 0,12, t-stat = 18,215). This suggests that in the event of a forced CEO turnover the amount of error in the forecast increases, thus that the quality decreases. Furthermore, all forms of a strong board of directors (SIZE, EXPERTISE, OUTSIDE) has a significantly negative relationship with the quality of the forecasts. Further indicating the positive effect the board of directors can have. However, it seems that the event of a turnover is not statistically significant. But if we look at the analysis made on the forced and voluntary turnover separately, we do see a significant effect. Showing a decrease in quality if it is a forced CEO turnover and an increase in quality if it is a voluntary CEO turnover. This effect seems to be much larger (B = 0,122) in the case of a CEO turnover. Meaning the event of a turnover leans to a negative effect on the quality of the forecasts. Also it seems that the financial expertise of a board of directors has a huge impact on the quality (B = -0,462, t-stat = -4,730). Clearly suggesting that financial experts influence the analysts in some way.
Next we find that the number of forecasts is significantly larger in the year the CEO turnover happens than in the other two years (B = 0,03, t-stat = 14,252). However other
relationships could not be found in the dataset’s current form, for the reason that the independent variables does not vary over the forecasts of years and that the dependent variables over the three years combined is equal to 1.
Regarding the dispersion rate, we find that the type of turnover is significant for the dispersion rate (B = 0,086, t-stat = 6,340). This suggests that the analysts’ forecasts have more spread in the event of a forced CEO turnover. In turn this means that there is more information asymmetry between the analysts. Furthermore, the board size seems to have a significant negative relationship with the dispersion rate (B = -0,007, t-stat = -3,661).
The results indicate that the event of a turnover and that a forced CEO turnover has mostly a negative effect for all dependent variables and that the board of directors has to some extent a
Page 22 of 33 moderating effect on this. For the next part we will investigate whether there is a change over the years for the dependent variables.
Additional tests of dependent variables per year
We have already established significant relationships between the dependent vaiables (QUALITY and DISPERSION) and the independent variables (TURNOVER, TYPETURNOVER and board of directors) with the use of a regression model. However, no tests were done for quantity due to the structure of the data. By dividing the values of quantity by year we can examine whether there is a significant change in value over the years. The results are found in table 6.
For all analyst forecast determinations, it shows a significant value at the Mauchly's Test of Sphericity, indicating a violation of sphericity. This means that the variances between the dependent variables are not equal and that a correction needs to be made. For this who look at the sphericity value of Huynh-Feldt which shows a value of larger than 0,750 and that means that the Huynh-Feldt variables needs to be used. This shows that there is a significant difference over the years for all dependent variables (F=25,249; F=50,084 and F=3,690). This means that one mean of the years is significantly different than the others. Suggesting that the event of a turnover that happens in t2 might have a significant influence on the quality, quantity and dispersion rate.
By comparing the years with each other in the pairwise Bonferroni test in table 7, it shows that the variable in year 2 has a significant higher value compared to year 1 and year 3 for quality and quantity. The exception lies with the dispersion rate, which shows only a significant change between year 2 and year 3. Figure 3 in the appendix shows an increase between year 1 and 2 for the
dispersion rate, but it seems that this difference is not considered significant. Also suggesting a relationship with the CEO turnover event. The values over the years are placed in a graph, shown in figure 1, figure 2 and figure 3 in the appendices.
Furthermore, it seems that the type of the turnover (TYPETURNOVER) only shows a significant value for the variables Quality (F=70,605) and Dispersion rate (F=2,512). This indicates that the value of the quality and dispersion rates will significantly change with the type of turnover. This is also shown in the figures in the appendices. The quantity results in the Bonferroni test in combination with the regression Huynh-Feldt value means that the statistically significant difference (increase) in t2 will occur during a CEO turnover, regardless whether this is forced or natural.
Page 23 of 33 Regression coefficients
Quality of forecasts
(QUALITY) Quantity of forecasts (QUANTITY) (DISPERSION) Dispersion rate
Variables Entire
dataset Turnover Forced Voluntary Turnover dataset Entire Turnover Forced Voluntary Turnover dataset Entire Turnover Forced Voluntary turnover
Constant 0,285*** 0,552*** 0,2533*** ,323*** 0,322*** ,323*** 0,177*** ,146 ,211*** (15,168) (6,838) (14,162) (38,377) (13,857) (35,271) (4,512) (0,221) (5,766) Turnover 0,005 0,122*** -,013** 0,030*** 0,035*** 0,030*** 0,003 0,010 0,002 (1,059) (6,119) (-2,936) (14,252) (6,05) (12,937) (0,288) (0,221) (0,190) Type Turnover (18,215) 0,12*** 0,086*** (6,340) Size -0,012*** -,014*** -0,011*** -0,007*** -0,003 -0,009*** (-13,861) (-6,477) (-12,033) (-3,661) (-0,486) (-4,607) Expertise -0,084** -,462*** -0,037** 0,020 0,211 -0,007 (-4,35) (-4,730) (-2,077) (,478) (0,936) (-0,183) Outside -0,50*** -0,158 -0,019 -0,062 -0,026 -0,067 (-2,191) (-1,467) (-,898) (-1,325) (-,122) (-1,573)
P-values are two-sided; ***, **, and * represent p-values <1%, <5%, and <10% respectively.
Table 5. A summary of the regressions results of the dependent variables QUALITY, QUANTITY and DISPERSION. Results based on either the entire dataset and divided by the type of turnover. First value of the variables is the coefficient B and the value underneath is the t-statistics value.
Page 24 of 33
Tests of within-subjects effects
The first value is the partial ETA squared to show to what extent the variables explain the differences, with exception of the sphericity. The number is parentheses are the F-value. P-values are two-sided; ***, **, and * represent p-values <1%, <5%, and <10% respectively. Table 6. Results of the repeated measures test. This is created by dividing the dependent variables quantity, quality and dispersion rate into the values per year. The results show whether there is a significant change between the three years. Moreover, taking into account a possible interaction with the various independent variables.
Pairwise Comparisons (Bonferroni test)
Quantity t1 Quantity t2 Quantity t3
Quantity t1 1 -0,029*** 0,007
Quantity t2 0,029*** 1 0,036***
Quantity t3 -0,007 -0,036*** 1
Quality t1 Quality t2 Quality t3
Quality t1 1 -0,067*** -0,026***
Quality t2 0,067*** 1 0,041***
Quality t3 0,26*** -0,041*** 1
Dispersion t1 Dispersion t2 Dispersion t3
Dispersion t1 1 -0,007 0,022
Dispersion t2 0,007 1 0,030**
Dispersion t3 0,022 -0,030** 1
P-values are two-sided; ***, **, and * represent p-values <1%, <5%, and <10% respectively.
Table 7. Results of the pairwise comparison test per dependent variable. This compares the values of each year with each other showing a statistical change where possible.
Quantity Quality Dispersion Rate
Mauchly's Test of
Sphericity Huynh-Feldt Epsilon 0,991*** P-value 0,000 0,992*** 0 0,975*** 0,005
Dependent variable Huynh-Feldt 0,008*** 0,014*** 0,012**
(25,249) (50,084) (3,690) Dependent variable * typeturnover Huynh-Feldt 0,000 (0,394) 0,020*** (70,605) 0,008* (2,512) Dependent variable * Size Huynh-Feldt 0,006*** (3,922) 0,044*** (30,703) 0,027 (0,474) Dependent * Expertise Huynh-Feldt 0,018*** (1,921) 0,111*** (12,957) 0,026 (0,009) Dependent variable * Outsiders Huynh-Feldt 0,011** (1,442) 0,066*** (9,096) 0,042 (0,743)
Page 25 of 33
Moderator analysis
We’ve tested to see if there is any relationship between the board of director’s variables and the number of quantity, quality and dispersion rate. To test whether these variables have any moderating effect on the relationship between the type of turnover and the dependent variables, the data has been transformed. The board of directors’ variables and the type of turnover will be standardized and multiplied with each other to create three new moderator variables to perform a moderator analysis.
The results of the type of turnover and the board of directors’ size, expertise and outsiders are in the same line as with the regression models. In this table we find significant values for type*expertise (B = -0,16; t-stat = -6,553) and type*outside (B = -0,13; t-stat = -5,834) for the quality of forecasts. Moreover, they are of negative value, while the type of turnover is positive value. Suggesting that if the expertise and outside board members are larger, that the quality of forecasts will increase. But there is no relationship found with the type of turnover and the board size. This could suggest that a higher board size is not always beneficial for the company and the analysts. Looking at the dispersion rate we see all significant values for the type*board of directors variables. These moderating values (B=0,008 B=0,010 B=0,007) all show a lower and significant B-value than the TypeTurnover (B=0,029). It means that when taking into account the board of directors size, expertise and outsiders, the effect on the dispersion rate is lower. Suggesting a moderating effect of the board of directors on the type of turnover for the dispersion rate.
Page 26 of 33
Moderator effects table
Variables Quality of forecasts
(QUALITY)
Dispersion rate (DISPERSION)
Standardized type turnover 0,039*** 0,029***
(17,162) (6,084) Standardized size -0,031*** -0,023*** (-12,509) (-4,371) Standardized expertise -0,06*** -0,002 (-2,682) (-0,517) Standardized outside -0,31*** -0,022*** (-12,509) (-4,407) Type*Size 0 0,008** (-,182) (2,219) Type*Expertise -0,16*** 0,010** (-6,553) (2,052) Type*Outside -0,13*** 0,007* (-5,834) (1,934)
P-values are two-sided; ***, **, and * represent p-values <1%, <5%, and <10% respectively. Table 8. Analysis of the effect of the type of turnover and the various board of directors’ variables separately and combined. First value of the variables is the coefficient B and the value underneath is the t-statistics value.
Page 27 of 33
5 Conclusion
This paper has examined how analysts reacts in an event where information asymmetry increases. In particular in the event of a forced CEO turnover and natural turnover. Prior studies have already examined this effect on the performances of analysts on short term. This paper tries to extent this literature by looking at the effect on quality, quantity and dispersion over a longer time period of three years and looking at individual analysts instead of the consensus forecasts. Furthermore, it tries to see how the analysts make use of a stronger board of directors.
The results show that a forced CEO turnover is negatively associated with the quality and positively associated with the dispersion rate of analyst forecasts. In case of a natural turnover the effect seems to be smaller and minimal. There is no evidence suggesting the same effect on the quantity of the forecasts. However, the event of a turnover does seem to increase the number of forecasts in that year. Furthermore, we find a positive correlation between the accuracy/quality of the forecasts and the board size, independence rate and outsiders rate, suggesting that the quality increases with a stronger board of directors. A negative correlation is non-existent with quantity and limited for the dispersion rate.
These results are supporting the hypotheses that a forced CEO turnover will change and negatively affect the quantity, quality and dispersion rate of the analyst forecasts. A limited effect has been found of the effects in the event of a natural turnover. This confirms multiple studies that a forced CEO turnover will negatively influence the analyst forecast, meaning they have troubles in acquiring accurate information for their forecasts.
Moreover, a moderator effect has been found for the board of director’s size, expertise and outsiders. But this is only in the case of the quality of forecasts and the dispersion rate. This
implicates that a stronger board of directors will limit the negative influence of a forced CEO turnover. This is also partially supporting the hypothesis of the role of the board of directors in the event of a CEO turnover.
The implications of these results are that the information asymmetry increases for analysts due to the forced leave of the CEO and the takeover of a new CEO. Analysts doesn’t have enough information to issue a qualitative accurate forecasts, resulting in more revisions during that year and thus a higher discrepancy between the forecasts. However, a stronger board of directors will mean more transparency limiting the effect of information asymmetry in this event, thus giving the analysts the opportunity to issue better forecasts.
Page 28 of 33 This paper only uses the CEO turnover and the board of directors as a factor that can influence the analyst forecast. Future research may investigate whether this effect is still significant when taking into account the economic environment of the company and or market. Another possibility is taking into account the old CEO’s tenure, expecting that a relatively short tenure might increase the effect, due to the lack of trust in the company. Finally, another extension could be to go even more in depth by looking at each individual forecast instead of the of each year.
Page 29 of 33
6 Bibliography
Aboody, D., & Kasznik, R. (2000). CEO stock option awards and the timing of corporate voluntary disclosures. Journal of Accounting and Economics (29), 73-100.
Baginski, S. P., & Hassell, J. M. (1990). The Market Interpretation of Management Earnings Forecasts as a Predictor of Subsequent Financial Analyst Forecast Revision. The Accounting
Review (65)1, 175-190.
Baik, B., & Jiang, G. (2006). The use of management forecasts to dampen analysts’ expectations.
Journal of Accounting and Public Policy (25), 531-553.
Ball, R., & Brown, P. (1968). An Empirical Evaluation of Accounting Income Numbers. Journal of
Accounting Research (6)2, 159-178.
Barron, O. E., Kim, O., Lim, S. C., & Stevens, D. E. (1998). Using Analysts' Forecasts to Measure Properties of Analysts' Information Environment. The Accounting Review (73)4, 421-433. Barron, O. E., Stanford, M. H., & Yu, Y. (2009). Further Evidence on the Relation between
Analysts’ Forecast Dispersion and Stock Returns. Contemporary Accounting Research (26)2, 329-357.
Bernard, V. L., & Thomas, J. K. (1989). Post-Earnings-Announcement Drift: Delayed Price Response or Risk Premium? Journal of Accounting Research (27), 1-36.
Bhat, G., Hope, O.-K., & Kang, T. (2006). Does corporate governance transparency affect the accuracy of analyst forecasts? Accounting and Finance (46), 715-732.
Burgstahler, D., & Eames, M. (2006). Management of Earnings and Analysts’ Forecasts to Achieve Zero and Small Positive Earnings Surprises. Journal of Business Finance & Accounting, 33(5) &
(6), 633–652.
Choi, K., Chen, X., Wright, S., & Wu, H. (2014 ). Analysts’ Forecasts Following Forced CEO Changes. A Journal of Accounting, Finance and Business Studies (50)2, 146-173.
Clayton, M. J., Hartzell, J. C., & Rosenberg, J. (2005). The Impact of CEO Turnover on Equity Volatility. Journal of Business (78)5, 1779-1808.
Diether, K. B., Malloy, C. J., & Scherbina, A. (2002). Differences of Opinion and the Cross Section of Stock Returns. The Journal of Finance 57(5), 2113-2141.
Fama, E. F., & Jensen, M. C. (1983). Seperation of Ownership and Control. Journal of Law &
Economics (26)2, 301-325.
Farrell, K. A., & Whidbee, D. A. (2002). Monitoring by the financial press and forced CEO turnover. Journal of Banking & Finance (26), 2249-2276.
Page 30 of 33 Gleason, C. A., & Lee, C. M. (2003). Analyst Forecast Revisions and Market Price Discovery. The
Accounting Review (78)1, 193-225.
Healy, P. M., & Palepu, K. G. (2001). Information asymmetry, corporate disclosure, and the capital markets: A review of the empirical disclosure literature. ournal of Accounting and Economics (31), 405-440.
Jenter, D., & Kanaan, F. (2015). CEO Turnover and Relative Performance Evaluation. The Journal of
Finance (70)5, 2155-2183.
Johnson, T. C. (2004). Forecast Dispersion and the Cross Section of Expected Returns. The Journal of
Finance 59(5), 1957-1978.
Kasznik, R., & McNichols, M. F. (2002). Does Meeting Earnings Expectations Matter? Evidence from Analyst Forecast Revisions and Share Prices. Journal of Accounting Research (40)3, 727-759.
Khanchel, I. (2007). Corporate governance: measurement and determinant analysis. Managerial
Auditing Journal (22), 740-760.
Klein, A. (2002). Audit committee, board of director characteristics, and earnings management.
Journal of Accounting and Economics (33), 375-400.
Peasnell, K. V., Pope, P. F., & Young, S. (2005). Board Monitoring and Earnings Management: Do Outside Directors Influence Abnormal Accruals? Journal of Business Finance & Accounting
(32)7&8, 1311-1346.
Richardson, S., Teoh, S., & Wysocki, P. D. (2004). The Walk-down to Beatable Analyst Forecasts: The Role of Equity Issuance and Insider Trading Incentives. Contemporary Accounting Research
(21)4, 885-924.
Sheng, X., & Thevenot, M. (2012). A New Measure of Earnings Forecast Uncertainty. Journal of
Accounting and Economics 53, 21-33.
Trueman, B. (1994). Analyst Forecasts and Herding Behavior. The Review of Financial Studies (7)1, 97-124.
Weisbach, M. S. (1988). Outside Directors and CEO Turnover. Journal of Financial Economic (20), 431-460.
Xie, B., Davidson III, W. N., & DaDalt, P. J. (2003). Earnings management and corporate governance: the role of the board and the audit committee. Journal of Corporate Finance (9), 295-316.
Page 31 of 33
7 Appendices
Page 32 of 33 Figure 2. Showing the average value of Quantity per year per type of turnover.
Page 33 of 33 Figure 1. Showing the average value of dispersion rate per year per type of turnover.
