ABSTRACT
1) INTRODUCTION
In 1996, the first Dutch corporate governance code was released by a committee chaired by Jaap Peters. The report this committee produced, known as the Peters Report, was the first attempt in the Netherlands, and one of the first in the world, to improve the transparency of firms by means of a code of conduct of good corporate governance. As a result of the hausse of corporate governance related scandals1 that emerged in the beginning of the 21st century, the Dutch government decided to update the Peters report.
For this purpose, a new committee chaired by Morris Tabaksblat was installed in 2003. In December 2003, this committee published their report on good corporate governance (the ‘Tabaksblat Code’)2. Starting January 2004, all listed Dutch firms had to ‘comply or explain’ with this code. The goal of the Code was to promote good codes of conduct on several corporate governance related issues such as the composition and remuneration of the supervisory board, shareholder rights and disclosure policies. As these goals prescribe increased transparency of firms and increased shareholder rights, shareholders were positive about this initiative to improve corporate governance.
Firms bear the direct costs of complying with a new corporate governance code since this comprises reviewing remuneration policy, supervisory board composition, reporting standards and other internal processes. For this reason, they were less positive about the introduction of the new code and were often reluctant to comply. Besides, with the Peters Report already being published, there already was a Dutch corporate governance code, they argued.
Regarding the introduction of corporate governance codes in general, several scholars have researched the first question. Roughly two approaches exist. The first approach examines the value of firms’ shares before and after the introduction of a new corporate governance code (event studies). Another approach examines the effect of compliance with corporate governance codes on specific firm characteristics such as firm value or performance (primarily cross‐sectional studies). Neither of these approaches show consensus. For corporate governance regimes on a comply or explain basis, as employed by most European Union member countries, both mixed results (f.i. Dedman [2002] and Doble [1997]) and altogether non existing relations are found (De Jong, Dejong, Mertens and Wasley [2005], Alves and Mendes [2004] and Weir and Laing [2000]). Similar studies on the United States Sarbanes Oxley Act, centered on a rules‐based system, show even less consensus. While some authors find positive value effects (f.i. Li, Pincus and Rego [2006] and Rezaee and Jain [2005]), others find exactly the opposite (f.i. Zhang [2007]). Whereas studies examining firm value or performance in light of governance related events do not show consensus, another stream of research is more univocal. This recently evolving stream of research studies the effect of corporate governance within firms on a more direct characteristic by using firms’ cost of equity capital as the dependent variable. This body of research generally finds a negative relationship. I.e., the influence of corporate governance related attributes leads to a reduction in the cost of equity capital. This literature often focuses on one aspect of firms’ corporate governance such as information quality (f.i. Easley and O'hara [2004], O'hara [2003] and Leuz and Verrecchia [2000]) or decision rights (Garmaise and Liu [2005]). Ashbaugh‐Skaife, Collins and Lafond [2004] find this relationship for multiple aspects: information quality, ownership structure, shareholder rights and board structure. This body of research focuses on corporate governance factors within firms. However, no research has yet been done on the influence of the introduction of new corporate governance codes on the cost of equity capital for a sample of firms.
laws. This thesis also adds to the literature on the Tabaksblat Code in the Netherlands, a topic that has not yet been given substantial attention in the academic literature.
The second goal of this thesis is to examine the influence of the Tabaksblat Code vis‐à‐vis the Peters Report on the cost of equity capital of Dutch firms. It does so by examining whether the Peters report and the Tabaksblat Code have had different effects in terms of significance and magnitude on firms’ cost of equity capital. In this way it sheds light on the question whether the market valued the impact of the Peters Report and the Tabaksblat Code differently.
2) LITERATURE REVIEW
“A corporation is a nexus of contracts between principals and agents” (Berle and Means [1932]). This distinction between those who own (principals) and those who control (agents) is the basis of agency theory. Entrepreneurs or managers have a lack of sufficient own funds and therefore need funds from investors to pursue their intended projects. Investors are willing to provide these funds since these managers have the skills and the time to provide a sufficient return on their investment. The question, however, arises how investors can assure that they indeed get something in return (Shleifer and Vishny [1997]).
Agency theory is part of the contractual view of the firm as developed by Coase [1937] and Jensen and Meckling [1976]. Contract theory provides the solution to the problem stipulated above: design a complete contract. That is, a contract that states exactly what the manager is obliged to do with the invested funds (and what not!) in all possible states of the world. Obviously, such a contract is not feasible. Therefore, the contract also specifies who has control in cases not included in the contract (residual control rights).
For investors to keep most control they could specify that all residual control rights are to be located to them. However, this would require them to be constantly involved in the project while this was the initial reason for investors to contract out their funds. Therefore, in general investors let go substantial residual control rights (Shleifer and Vishny [1997]).
The distinction between investors that own a corporation but managers controlling it results in information asymmetry between these two agents. This is the source of agency risk for investors. A moral hazard problem arises if managers have incentives to use investors’ funds for their own private benefit. This can take many forms including shirking, empire building or excessive use of perks (Tirole [2006]) 3. Next to the moral hazard problem, an adverse selection problem results from information asymmetry. The economic value of a corporation is partially dependent on the quality of its management. Investors face agency risk by having imperfect information on the quality of management and the economic value of the firm. Rational investors need to be compensated for
3The excessive use of perks can add up to substantial forgone gains to investors. Burrough and
bearing these two sources of agency risk. They do so by demanding a risk premium, which raises firms’ cost of capital. Managers, knowing this, aim to construct a set of governance practices which provide credible protection for investors at the lowest possible cost (Hart [1995]). This is where the role of corporate governance comes in. The goal of corporate governance codes is to moderate agency risk using codes of conduct for a variety of matters. These include improving the (financial) disclosure of firms, giving more rights to (minority) shareholders, increasing the possibility of monitoring management and limiting the opportunistic behavior of management. All these measures should improve the transparency of firms and lower agency risks. This way, corporate governance can be expected to effectively lower firms’ cost of equity capital since less risk has to be compensated. Following, shares can be issued closer to their true value, instead of at a discount that reflects the agency risk (Ashbaugh‐Skaife et al. [2004]).
Since lowering firms’ cost of capital enables managers to pursue more positive net present value projects, one can expect that the value of firms increases following corporate governance improving related events. McKinsey research indicates that institutional investors attribute as much as 20 to 40 percent of a firm’s value to its governance (Monks [2002]). This has been the subject of an increasingly large body of empirical literature. Within this literature, a logical distinction can be made between research focusing on countries using a rules‐based system and countries using a comply‐or‐explain system. That is, distinguishing between the United States and most other countries that have released corporate governance codes.
Zhang [2007] examined the economic consequences for US firms surrounding news events that gave information about the release of the Sarbanes‐Oxley Act. Examples of these events are the passing of the Sarbanes’ bill by the Senate Banking Committee or the passing by the House of the bill strengthening criminal penalties for non‐compliance4. Using non‐US traded firms as a benchmark, she found that the cumulative abnormal returns on the stocks of US firms were significantly negative after these events. She therefore concludes that complying with these governance provisions imposes a net cost. Li et al. [2006] conduct a similar analysis. They also examine stock price reactions for US firms surrounding comparable events as in Zhang [2007]. They, however, report positive abnormal returns after these events. Rezaee and Jain [2005] support these results. They also find a positive market reaction to events increasing the likelihood of the passage of the Sarbanes Oxley
Act. Interestingly, they note that the positive value effect is stronger for firms that had good corporate governance in place before the Sarbanes Oxley Act. Their finding that on average the Act has a wealth‐increasing effect supports the notion that the Act has a net benefit.
These studies apparently do not arrive at consensus. An explanation (Chhaochharia and Grinstein [2005] and Zhang [2007]) is that each research used a different set of news events. Furthermore, they associated these events differently with respect to these events increasing or decreasing the likelihood of passage of the Sarbanes Oxley Act.
Chhaochharia and Grinstein [2005] try to overcome this problem by using a large event window to capture information overspill. Furthermore, they control for market‐wide shocks that have no relation with the Sarbanes Oxley Act by comparing two portfolios of firms. One portfolio that is less compliant with the Act and one matching portfolio that is more compliant with the act. In addition, they differentiate with respect to firm size. Their results indicate that firms complying less with the act earn positive abnormal returns compared to firms that comply more. However, small firms complying less earn negative abnormal returns. This suggests that some Sarbanes Oxley provisions are damaging to small firms. In general, no consensus exists as to whether the introduction of the Sarbanes Oxley Act has had positive welfare implications overall.
Several scholars have studied the economic effect on firms of comply‐or‐explain type corporate governance codes. The majority of them conclude that no wealth effect exists. A short overview follows.
Weir and Laing [2000] examine the effect of the in 1992 released UK Cadbury Report on the performance of firms in the UK. Using a sample of 200 firms, they conclude that full compliance is not associated with better firm performance than partial or no compliance.
An example of an article on the value effect of the predecessor of the Tabaksblat Code, the Peters report, is De Jong et al. [2005]. They examine the corporate governance characteristics of Dutch firms and their value before and after the release of the Peters Report. They find that the Peters Report has not influenced firms’ corporate governance characteristics. Following this result, no value effect was found either.
A similarly strong conclusion is given by Nowak, Rott and Mahr [2006] for the German case. Using a dataset of 317 listed German firms they conclude that for the long run, high levels of compliance with the German corporate governance code or improved corporate governance are not associated with higher corporate value than low compliance or a reduction in code compliance. Finally, an analysis of welfare effects of announcements of Spanish firms regarding compliance with the Spanish ‘Olivencia Code’ is conducted by Fernández‐Rodríguez, Gómez‐Ansón and Curevo‐García [2004]. They find no general value effects for firms announcing compliance with the code. However, they do find a positive value effect if the announcement is accompanied by the compliance implying a major restructuring of the board of directors. All in all, while some positive relationships are identified, no sound evidence is found of corporate governance codes of the comply‐or‐explain type having a positive effect on firm performance or value. An overview of these empirical results is given in Appendix 1.
Another stream of literature, and the topic of this thesis, examines the effect of corporate governance codes on firms’ cost of equity capital. First, a summary of the related theory is presented. Then, an overview of the empirical literature will be given.
The goal of corporate governance codes is to moderate agency risk using codes of conduct for several matters. Examples of these matters are improving the (financial) disclosure of firms, giving more rights to (minority) shareholders, increasing the possibility of monitoring management and limiting the opportunistic behavior of management. Each of these measures in theory should improve the transparency of firms and decrease the sources of agency risk described above. By reducing agency costs, firms’ cost of equity capital should decrease.
Next to decreasing agency costs, corporate governance codes aim to improve the quality of information disclosure. Clear links exist between the quality of information disclosure and firms’ cost of equity capital. Increased information disclosure reduces the information asymmetry between managers and investors, lowering the cost of equity capital (Diamond and Verrecchia [1991]). Information asymmetry results in investors demanding a higher return in three ways. First, it is the source of information risk for investors, which has to be compensated5. Second, it increases estimation risk. That is, the risk relative to the estimation of the distribution of returns and thereby the value of the asset (Clarkson, Guedes and Thompson [1996]). Finally, as information asymmetry increases the uncertainty and risk of a stock, more research is needed when deciding whether to buy the stock, resulting in shares being less liquid. This results in increased transaction costs. The transaction costs of relatively illiquid shares in turn are reflected in the cost of capital (Leuz and Verrecchia [2000], Diamond and Verrecchia [1991]). Corporate governance codes aim at improving information disclosure by firms which therefore has a positive (decreasing) effect on firms’ cost of equity capital.
Turning to empirics, relatively little research has yet been done on the relation between corporate governance attributes and cost of equity capital. An overview of the exceptions follows.
Cheng, Collins and He Huang [2006] research the influence shareholder rights regimes and the financial transparency of firms have on the cost of equity capital for a sample of 348 S&P‐500 firms. They use the governance index developed by Gompers, Ishii and Metrick [2003] as a proxy for shareholder rights and the financial transparency and disclosure rankings developed by Standard
5Information risk can be interpreted in two ways. One, stressing the information asymmetry
and Poor’s as a proxy for financial transparency6. They find that firms with stronger shareholder rights regimes and higher levels of financial transparency enjoy significantly lower costs of equity capital.
Ashbaugh‐Skaife et al. [2004] examine four specific corporate governance attributes. Specifically, they examine financial information quality, ownership structure, shareholder rights and board structure. They find support for the hypothesis that good corporate governance lessens agency risk, reducing cost of equity capital. Jointly, the corporate governance attributes considered explain approximately 8% of the cross‐sectional variation in firms’ cost of equity capital. They also incorporate these four attributes to construct a composite governance score for the firms in their sample. Following the renowned Fama and French [1993] three‐factor model, they use the risk factors beta, size and market‐to‐book ratio and add this governance score in a rank regression analysis. They find that firms with better corporate governance have on average a cost of equity capital that is 88 base points lower than firms with weaker corporate governance. Furthermore, when adding the corporate governance factor to the Fama and French [1993] three‐factor model they find this factor to be highly significant and positive. This way, the authors make an implicit plea for an additional risk factor next to beta, size and the market‐to‐book ratio to determine firms’ cost of equity capital.
One would therefore expect that the same reduction in cost of equity capital can be observed for firms around the adoption of a new corporate governance regime. Unfortunately, to the author of this thesis only one example of a comparable empirical study of this kind is known: a study by Leuz and Verrecchia [2000]. This study examines German firms that have switched from a German to an international reporting regime and predicts that as the international reporting regime entails increased transparency and disclosure (vis‐à‐vis the German reporting regime), this switch should have a negative effect on the information asymmetry component of the cost of capital. Using three proxies for the change in cost of equity capital and event study methodology, the authors show that the proxies for the cost of equity capital indeed behave in the predicted direction. I.e., they find evidence of the cost of equity capital decreasing as the information asymmetry decreases as a result of the firms switching to an international reporting regime. To the best of the author’s knowledge, the paper by Leuz and Verrecchia [2000] is the only paper using event study methodology to determine the change in cost of equity capital for a sample of firms.
3) HYPOTHESES
The first objective of this thesis is to test the economic impact of the introduction of the Tabaksblat Code. Section two of this thesis provided a theoretical basis for the relationship between corporate governance codes and cost of equity capital. In short, corporate governance codes can be expected to lower the risks that stem from asymmetric information and the principal‐agent problem, the agency risk. This lower risk enables firms to issue shares closer to their true value – at a lower discount – since there is less risk that investors need to be compensated for. Section two also showed that empirics often confirm this relationship. Regarding the case of the Tabaksblat Code, previous empirical research (Akkermans et al. [2007]) has shown that compliance with the Code in the Netherlands is high, especially for large firms. For a sample of large Dutch firms, this leads to the following hypothesis: Hypothesis 1 H10: The introduction of the Tabaksblat Code has not led to significant changes in the cost of equity capital of large Dutch listed firms.
H1a: The introduction of the Tabaksblat Code has led to a significant decrease in the cost of
equity capital of large Dutch listed firms.
value. True, a lower cost of equity capital enables managers to pursue more positive net present value projects. Engaging in more positive net present value projects, in turn, should theoretically lead to an increase in firm value. However, changes in firm value can have numerous other causes ranging from industrial trends to macro economical changes. Therefore, this thesis employs cost of equity capital as a more direct measure, following directly from agency theory which provides a solid explanation as to how improved corporate governance should influence firms’ cost of equity capital. Next to the choice of dependent variable in the research by De Jong et al. [2005] and this thesis, large differences exist between the Peters Report and the Tabaksblat Code that give reason to think that the Tabaksblat Code has had more economic consequences. Research by the Dutch financial newspaper ‘Het Financieele Dagblad’7 (Zevenbergen [2005]) has shown that one and a half year after the recommendations of the Peters Report as little as fourteen Dutch firms had carried out changes concerning their corporate governance. This could be caused by the fact that the Peters Report was not in any way embedded in Dutch law; all recommendations remained mere recommendations and firms were free to choose whether to comply with these recommendations or not. This is in contrast with the Tabaksblat Code. Although firms are still free whether to follow the principles proposed by the Code, non‐compliance has to be explained in firms’ annual reports. This comply‐or‐explain principle is embedded in Dutch law. Failure to comply or explain can be punished by law. Compliance with the code is checked on an individual basis by the Dutch ‘Autoriteit Financiële Markten’8 and for the Dutch market as a whole by the ‘Monitoring Committee’.
Even more, several Dutch politicians, including the minister of finance, have announced that measures will be undertaken in the case the Monitoring Committee should find evidence of Dutch firms not complying on a large scale (Zalm [2004] and De Vries [2003]). This would mean the Tabaksblat Code would no longer be used purely on a comply‐or‐explain basis but that some provisions could be embedded directly into Dutch law. This can be considered a credible threat for managers to, at least partly, comply with the Tabaksblat Code.
Another important difference between the Peters Report and the Tabaksblat Code involves the amount of attention their respective releases have gotten. Around 1996, the time of release of the Peters Report, corporate governance had gotten relatively little attention from the media,
shareholders and the government. By the time the Tabaksblat Committee was formed, however, the Netherlands as well as the rest of the world had recently experienced a hausse of corporate governance related scandals. Firms therefore knew that they could expect a lot more resistance from shareholders and other stakeholders if they reacted to the Tabaksblat Code the same way they reacted to the Peters Report. Therefore, the impact of the Tabaksblat Code as a result of better compliance can be expected to be higher.
4) METHODOLOGY
For both hypotheses, this section describes the sample selection, the relevant data and the methodology used to test the hypotheses. First, this section will explain the different ways to measure the cost of equity capital, the measures used for this thesis and how to control for other determinant of these measures.
4.1) MEASURING THE COST OF EQUITY CAPITAL
Theoretically, a firm’s cost of equity capital is the discount rate used to discount a firm’s expected future cash flows available to equity holders to determine its current share price. An obvious difficulty in this study is measuring firms’ cost of equity capital since it cannot be observed or measured directly. Fortunately, several methods exist to calculate the cost of equity capital indirectly or to approximate it using proxies. Several common methods will be described, followed by a motivation for the choice of the method used by this thesis.
4.2.1) Average realized returns
One method to measure the cost of equity capital indirectly is by using average realized returns. The problem with using average realized returns, however, is that it is heavily influenced by noise. To illustrate this, Botosan [1997] states that research using average realized returns has had difficulty finding a significant relationship between returns and beta, the most commonly accepted measure of risk. Other evidence by Lakonishok [1993] shows that when using average realized returns as a proxy for the cost of equity capital, more than 70 years of data is needed to prove the significance of the risk actor beta. For this research, the average realized return henceforth is not likely to be a useful proxy for the cost of equity capital.
4.2.2) CAPM
as a function of the expected risk‐free rate
E
(
R
f)
, the firm’s betaβ
and the expected market‐risk premiumE
(
R
m−
R
f)
in the following manner:)
(
.
)
(
)
(
R
SE
R
fE
R
mR
fE
=
+
β
−
(1) Although the CAPM is widely recognized, an implication of the model is that the only variation in a firm’s cost of capital is driven by a firm’s beta. Since the CAPM focuses entirely on beta and leaves out any role for corporate governance factors, the model for this thesis is not useful in modeling firms’ cost of equity capital. 4.2.3) Accounting based valuation formula Another method to calculate a firm’s cost of equity capital is by using an accounting based valuation formula as developed by Feltham and Ohlson [1995] and Ohlson [1995]. This method is based on the dividend discount model, a well‐known model for security valuation:∑
∞ = + −+
=
1 1]
[
.
)
1
(
t t t t tr
E
d
P
(2) The dividend discount model states that the current price of a firm’s stockP
t is equal to the sum of the expected future dividendsE
t[
d
1+t]
, discounted at the firm’s cost of equity capital r. Using clean surplus accounting, this formula can be rewritten as a function of a firm’s expected future accounting earnings, expected future book value of equity and expected future net dividends9. Examples of papers using this technique are Hail [2002] and Botosan [1997]. Papers using this method generally use the required inputs from the Value Line database to solve for the cost of equity capital. The paper by Ashbaugh‐Skaife et al. [2004] also uses the Value Line database but extracts the cost of equity capital directly as the annualized expected return over a three to five year period. Unfortunately, the Value Line database is not available to the author of this thesis, which makes the use of these methods not possible.
9The complete derivation is not within the scope of this thesis. See Hail [2002] or Botosan [1997] for
4.2.4) Bidask spread, trading volume and share price volatility
The article by Leuz and Verrecchia [2000], as mentioned in chapter two of this thesis, puts forward three proxies for the cost of equity capital offered by the economics, finance and accounting literature. These proxies particularly address the information asymmetry component of the cost of capital. These are the bid‐ask spread, trading volume and share price volatility. Each one will be discussed.
Bidask spread
Trading volume
Another proxy for adverse selection and the cost of equity capital is the trading volume of firms’ shares. The volume of trade in a firm’s shares reflects the liquidity of the shares. That is, it reflects the willingness of investors to buy or sell these shares. Adverse selection leads to more risk when transacting in shares. Therefore, adverse selection leads to shares being less traded and therefore less liquid. The volume of trade in a firm’s shares therefore is an inverse proxy for information asymmetry. Unfortunately, trading volume can be influenced by numerous other causes as well. One can think of, for instance, liquidity shocks (the sudden need to buy or sell a number of shares) or portfolio rebalancing. The article by Easley [1996], however, supports the choice for this proxy by showing that as trading volumes increase, information based trading decreases. This notion is confirmed by Grammig, Schiereck and Theissen [2001] for the German market. Trading volume will therefore be used as a proxy for the cost of equity capital for the sample of firms in this thesis. However, as a result of other determinants affecting the volume of trade in firms’ stocks as well, this proxy is considered less reliable than the bid‐ask spread.
Share price volatility
A general disadvantage of the proxies for the cost of equity capital put forward by Leuz and Verrecchia [2000] is that they do not measure the cost of equity capital directly nor in absolute terms. Instead, they focus on the part due to information asymmetry. For this thesis, however, this is considered sufficient since it is less relevant to calculate the cost of equity capital in absolute terms. We are interested in the possibility of the Tabaksblat Code decreasing agency risk represented by information asymmetry. We are therefore merely interested in the change in the cost of equity capital. Furthermore, given the fact that the proxies bid‐ask spread and trading volume have proven themselves in prior research, that the data necessary to use these proxies is available to the author of this thesis and that these proxies are intuitively appealing, they will be used in the rest of this thesis to proxy for the change in cost of equity capital.
Controlling for other determinants of bidask spread and volume of trade
expected to fall by 15%. Again, the value of the bid‐ask proxy, controlled for changes in trading volume, in this case would be increased by 15% to offset the influence of the increase in trading volume. For the trading volume proxy, the factors firm size, index inclusion and institutional ownership have been found to be influential (Leuz and Verrecchia [2000]). Since no major changes are expected in these variables as a result of the introduction of a new corporate governance code, no controls for trading volume are used. Assuming that the bid‐ask spread and trading volume are appropriate proxies for the change in the information asymmetry component of the cost of equity capital, this thesis puts forward that the introduction of a new corporate governance code leads to a smaller aggregate bid‐ask spread and higher trading volume for a sample of firms.
4.2) SAMPLE SELECTION
The Tabaksblat Code is applicable to all Dutch listed firms. Previous research in both the UK (Dedman [2000]; Mallin and Ow‐Yong [1998]; Conyon and Mallin [1997]) and Germany (Werder, Talaulicar and Kolat [2005]) as well as Eastern Europe (Berglöf and Pajuste [2005]) has shown that compliance with corporate governance codes increases with company size. Research by Akkermans
et al. [2007] has shown that this also is the case for the Tabaksblat Code in the Netherlands: Dutch
firms that are part of either the large‐cap or mid‐cap index are found to comply better with the Tabaksblat Code provisions than smaller firms10. Therefore, this thesis selects only the firms that make up the large‐cap and mid‐cap index. Both indices consist of 25 firms, resulting in a total sample of 50 firms. In doing so, this thesis follows the sample selection methodology of Leuz and Verrecchia [2000], who only selected the largest German firms composed out of the German DAX‐100 index11.
10The Amsterdam Exchange Index (AEX‐Index) and the Amsterdam Midcap Index (AMX‐Index)
respectively.
11The “Deutsche Aktien Xchange” 100 (DAX‐100) is an index consisting of the 100 major German
The composition of both the large‐cap and mid‐cap index changes over time as firms are being taken over, merged or delisted. Historical data on the composition of both indices is provided by Euronext and the website http://www.behr.nl12. To research the comparison between the Peters Report and the Tabaksblat Code (hypothesis 2), the same uniform sample has to be used. For this purpose, the composition of the indices at time of the introduction of the Tabaksblat Code (i.e. January 2004) is used to the select the firms in the sample to test both the impact of the Peters Report as well as the Tabaksblat Code.
Restrictions in the availability in the data further reduce this sample. While the original sample consists of 50 firms, data for both time periods is available for a smaller number of firms, reducing the sample to this number. A total of 12 firms included in one of the two indices in January 2004 did not have a listing on the Amsterdam Exchange at time of the introduction of the Peters Report (June 1997), reducing the sample to 38 firms. For all of these 38 firms, trading volume information is available. However, complete data on bid and ask prices is lacking for 15 firms, reducing the number of firms for this proxy to 23. This number of observations still is large enough to generate significant results13. The exact sample composition is provided in Appendix 3. 4.3) DATA Following the previous section, for the shares of the firms in the sample, data is needed on trading volume and bid‐ask spreads over time. To control for the effect of stock price movements, stock price information is also needed. This time‐series data is gathered from the DataStream database. Daily figures are extracted of the following data types:
VO ‐ Turnover by volume: the number of shares traded for a stock on a particular day, expressed in
thousands.
12http://www.behr.nl is a website providing independent investor information particularly for
private investors.
13The maximum number of observations in the event study by Leuz and Verrecchia [2000], for
PA ‐ Price, ask: the asking price quoted at close of market, expressed in Euros. PB ‐ Price, bid: the bid price offered at close of market, expressed in Euros. P ‐ Price, closing: the official closing price, expressed in Euros. The difference between the PA and PB data types constitutes the bid‐ask spread14. 4.4) EVENT STUDY DESIGN The hypotheses ask the question whether the cost of equity capital of Dutch listed firms has changed (decreased) as a result of (in case of hypothesis 1) the introduction of the Tabaksblat Code. In other words, it asks whether a specific event has had an influence on a financial characteristic (the cost of equity capital) of a group of firms. A type of methodology to research the impact of an event on financial characteristics of firms is the event study. Event studies have been used extensively in the field of finance and economics in general. Examples of papers using event study methodology go back to as early as 1933 (Dolley [1933]). However, it is widely acknowledged (Binder [1998]) that the seminal article on event study methodology article by Fama, Fisher, Jensen and Roll [1969] led to a true revolution in methodology in finance, accounting and economics.
the efficient market hypothesis, event studies assume that the market only reacts to the unexpected release of new, valuable information. Anticipating January 1st 2004, firms can be expected to have started incorporating changes some time before this date. The question therefore arises at what date investors expected firms to have switched to complying with the Tabaksblat Code. This is the event date.
To determine the actual event date, it is useful to look at the developments surrounding the release of information concerning the Tabaksblat Code and the respective impact on the expectations of investors concerning corporate governance at Dutch firms. A short chronological overview follows15.
Event 1 December 18th, 2002: First news emerges concerning new corporate governance code The Dutch financial newspaper ‘Het Financieele Dagblad’ published an article speculating that a committee, chaired by Morris Tabaksblat, was to be formed in the nearby future to establish a Dutch code of conduct for corporate governance. NB: the contents of the article were at that time based entirely on rumor and speculation.
Event 2 March 10th, 2003: The Tabaksblat Committee was formed
On March 10th, 2003, news was released that a committee, chaired by Morris Tabaksblat, was formed to write a new corporate governance code. The contents of the code, however, weren’t known yet at this date.
Event 3 July 1st, 2003: The Committee released a concept of the Tabaksblat Code
The Tabaksblat Committee released a concept version of the Tabaksblat Code. All interested parties were invited to submit comments in writing before September 5th, 2003. Since the content of the concept code was a good indicator of the definitive code, firms can be expected to have started incorporating changes as a result of the release of the concept code.
15Based on a Lexis‐Nexis search in the Dutch financial Newspaper ‘Het Financieele Dagblad’ (the
Event 4 December 9th, 2003: The Tabaksblat Committee released the definitive Code
The Tabaksblat Committee released the definitive version of the Tabaksblat Code. The definitive Code differed at some points from the concept Code, but the changes made weren’t drastic. Dutch firms knew at this point exactly with what provisions to comply or explain and they can be expected to have acted accordingly, since the date at which the Code became legally effective was within four weeks.
Event 5 January 1st, 2004: The Tabaksblat Code became effective
At January 1st, 2004, al Dutch listed firms had to comply or explain with all provisions of the Tabaksblat Code. They were legally obliged to do so.
Figure 1: Timeline hypothesis 1 2003 (Jan) 2004 (Jan) 2005 (Jan)
Apr 1 2003 – June 30 2003 July 1 2003 – Sept 30 2003
Sept 9 2003 – Dec 8 2003 Dec 9 2003 – March 8 2004 Oct 1 2003 – Dec 31 2003 Jan 1 2004 – 31 March 2004
Event 3 Event 4 Event 5 4.4.2) Hypothesis 1 – Calculating abnormal values and statistical testing The values of the data in the estimation window provide the benchmark to which the values in the event window are compared. For each data point in the event window, the value is compared with the average value in the estimation window and the value of the difference between these two is labeled abnormal. This implies that without the event having taken place, the data would have stayed at the same average value. Henceforth is the name of the model to benchmark the expected values vis‐à‐vis the actual values the constant mean return model (Mackinlay [1997]). The constant mean return model seems like a simple way to forecast normal values in the event period. The model, however, has proven itself to be robust, even compared to more sophisticated models (Brown and Warner [1980]; [1985]).
As an example, the volume of trade of stocks is used. Mathematically, let
AV
i,t be the abnormal volume for firm i at time t so that: t i t i t i V V AV , ^ , , = − (3) WhereV
i,t is the observed volume for firm i at time t and Vi,t ^ is the expected value based on the constant mean return model (the average value in the estimation window for firm i). The size and the sign of the abnormal volume reflect the (positive or negative) change in the volume of stock trade compared to the expected volume of trade. To enable statistical inference, the computed abnormal values are cumulated over time for the firms in the sample.different than zero (Strong [1992]). For this purpose, robustness tests are performed. The statistical test used in this thesis is the t‐test, which is well specified for this purpose (Strong [1992] and Brown and Warner [1985]).
In case the null‐hypothesis cannot be rejected, the event in question did not have a significant impact on the aggregated volume of trade of the firms in the sample. In other words: the event did not change the perception of investors to whether the firms in the sample indeed incorporated the provisions of the Tabaksblat Code and information asymmetry was reduced. The opposite is also true: if the null‐hypothesis can be rejected, investor behavior provide evidence that the firms in the sample actually incorporated the Tabaksblat Code provisions.
4.4.3) Hypothesis 2 – Specifying the time frame
Now that the methodology to test hypothesis 1 is explained, the same structure can be used to describe the methodology for hypothesis two. Recall that hypothesis 2 stated that the Tabaksblat Code has had a bigger impact on the cost of equity capital than the Peters report. An implicit assumption for this statement is that both the release of the Peters Report as well as the Tabaksblat Code have had significant impact on the cost of equity capital of Dutch firms. For the Tabaksblat Code, this assumption is tested in hypothesis 1. For the Peters Report, this first has to be examined using the same event study methodology as used for hypothesis 1. Then, the impact of the two respective corporate governance codes can be compared by looking at the size and the significance of the test results.
“…only (…) the release of the monitoring report and the related corporate governance information that was collected about the companies is associated with a significant stock price reaction.” (De Jong et al. [2005], P. 497)
Indeed, the article by De Jong et al. examined the influence on firm value, not cost of equity capital. We assume, however, considering the results of De Jong et al. [2005] that this event for the Peters Report is potentially significant for our purpose of measuring changes in cost of equity capital as well.
As was the case with the Tabaksblat Code, both a preliminary and a final version of the Peters Report were released. The respective dates of these events were October 28th 1996 and June 25th 1997. Although these events were not found to be significant in the study by De Jong et al. [2005], these events can be considered to encompass considerable information and therefore have the potential to be found to be significant in this thesis. Furthermore, these events closely resemble two of the three events used to test hypothesis 1 and are therefore valuable for the purpose of comparing the respective impacts of the different codes. Therefore, they are included in the statistical testing. In the rest of this thesis, for hypothesis 2, October 28th 1996 will be labeled event 1, June 25th 1997 will be labeled event 2 and December 3rd 1998 will be labeled event 3.
The event window and estimation window
Following hypothesis 1, the event window for hypothesis 2 is exactly three months. Similarly, the estimation window is also three months of length, ranging up to the date of interest, as summarized in the figure below. Figure 2: Timeline hypothesis 2 1996 (Jan) 1999 (Jan) 2000 (Jan) 1997 (Jan) 1998 (Jan)
July 28 1996 – Oct 27 1996 Oct 28 1996 – Jan 27 1997
March 25 1997 – June 24 1997 June 25 1997 – Sept 24 1997
Sept 3 1998 – Dec 2 1998 Dec 3 1998 – March 2 1999
Event 1 Event 2 Event 3
4.4.4) Hypothesis 2 – Calculating abnormal values and statistical testing
Using the same t‐test procedure as for hypothesis 1, the significance of the impact of the Peters Report on the cost of equity capital of the firms in the sample can be tested. Once completed, the results of the event study for hypothesis 1 and hypothesis 2 can be compared. The table below summarizes the implications. In case neither of the event studies shows significant results, hypothesis 20 cannot be rejected. If only one of the event studies is shown to be significant,
hypothesis 20 can be rejected or not rejected, depending on the outcome. In the case of both event studies for hypothesis 1 and hypothesis 2 being significant, we can look at the value of the abnormal values. Further testing, however, in this case is needed to prove statistically which one of the two events has had more impact. Table 1: When to reject H20: scenarios
Impact Peters Report on COEC Impact Tabaksblat Code on COEC Hypothesis H20
Not proven Not proven Ambiguous
Not proven Proven Rejected
Proven Not proven Not rejected
Proven Proven Further testing needed
5) RESULTS
In this section the results of the conducted event studies are described. First, the results for hypothesis 1 are described. Then, the results for hypothesis 2 are described after which a comparison between the two is made. 5.1) HYPOTHESIS 1 In Table 2 below the impact of the three respective events concerning the Tabaksblat Code on the two proxies for the change in cost of equity capital is displayed. Note that Table 2 displays the results for the full sample, i.e. all 50 firms that make up the Dutch large‐cap and mid‐cap indices at time of introduction of the Tabaksblat Code. Regarding the volume proxy, the results of the tests are ambiguous. Two of the three events result in a positive change in trading volume as predicted. These changes in trading volume, however, are not significant at common significance levels. However, event number 5 with a p‐value of 0.105 is nearly significant at a 90% confidence level. In contrast to the expectations, event number 3 (the release of the concept code) even shows a significant (p<0.05) decline in trading volume, giving reason to believe that the introduction of the Tabaksblat Code has decreased the liquidity of the shares of large Dutch firms. These ambiguous results confirm the belief stipulated in chapter 4 that trading volume is a less reliable proxy for changes in cost of equity capital.
In contrast, the results for the bid‐ask spread are unambiguous. All three events show a decline in the bid‐ask spread, indicated by a negative abnormal value (as predicted). This is true for both the uncontrolled bid‐ask spread as well as the controlled bid‐ask spread16, corrected for changes in trading volume and share price movements. This result is significant at p<0.05 for event number 3 and 4 in the case of the uncontrolled proxy and significant at p<0.01 in case of the bid‐ask proxy corrected for changes in trading volume and share price. Looking at the change in share price and trading volume of the sample in the event period can explain the difference in significance between
16The controlled value of the bid‐ask spread is corrected for changes in both share price and trading
the uncontrolled and controlled value of the bid‐ask spread proxy. For instance, in case of event 3, the sample experienced an average share price increase of 16% and a decrease in trading volume of 13% over the event period. In case of event 4, the sample experienced a share price increase of 7% and an increase in trading volume of 16%. This confirms the importance of controlling for other determinants of bid‐ask spreads. While the uncontrolled proxy indicates event 3 and event 4 as most significant events, in reality event 5 has had the most significant impact on the bid‐ask spread of the sample of firms. Besides from the significance, the impact also is highest in case of event number 5 with an abnormal value of ‐0.014 versus ‐0.005 for event 3 and 4.
The results of the bid‐ask spread proxy indicate that the introduction of the Tabaksblat Code indeed has led to less information asymmetry and consequently a lower cost of equity capital for large Dutch firms. Regarding this result, the date at which the code became effective (January 1st 2004) has had the most significant influence. The ambiguity of the results for the trading volume proxy do not alter this conclusion, as this proxy is considered a less direct and less reliable proxy for the change in cost of equity capital.
Table 2: Results for event studies Tabaksblat Code, full sample
Volume proxy Event (n=50) Abnormal value T‐stat P‐value (sig.)
Event 3 (concept code) ‐319.435 ‐2.069 0.044**
Event 4 (definitive code) 381.840 1.215 0.231
Event 5 (date effective) 611.855 1.652 0.105
Bid‐ask uncontrolled Event (n=50) Abnormal value T‐stat P‐value (sig.)
Event 3 (concept code) ‐0.007 ‐2.566 0.014**
Event 4 (definitive code) ‐0.003 ‐2.055 0.046**
Event 5 (date effective) ‐0.002 ‐1.126 0.266
Bid‐ask controlled Event (n=50) Abnormal value T‐stat P‐value (sig.)
Event 3 (concept code) ‐0.005 ‐0.991 0.327
Event 4 (definitive code) ‐0.005 ‐1.195 0.238
Event 5 (date effective) ‐0.014 ‐3.381 0.002***
Asterisks indicate that the means of the two groups are significantly different using a two‐tailed t‐ test with *p<0.1; **p<0.05 and ***p<0.01
data is available for hypothesis 2 as well. This is done to enable comparison between the impact and significance of the Peters Report and the Tabaksblat Code respectively17.
Both proxies show comparable results as in the case of the full sample. Concerning the trading volume proxy, event number 4 and 5 again show an increase in volume while event number 3 shows a decline in trading volume. Again, event number 5 is nearly significant at a 90% confidence level with a p‐value of 0.126. As is the case for the full sample, event number 3 is significant at p<0.05 but behaves in the opposite direction than predicted. The ambiguity of the results for the trading volume again can be attributed to the fact that this is a less reliable proxy for the change in cost of equity capital.
As is the case for the full sample, all abnormal values of the bid‐ask spread proxy are negative, as predicted. In this case, however, none of the uncontrolled bid‐ask proxy values are significant. Comparable to the previous results, event number 5 is found to be significant (again at p<0.01) in case the bid‐ask proxy is controlled for changes in trading volume and share price movements. The impact also is highest for event 5 with an abnormal value of ‐0.013. The difference in significance between uncontrolled and controlled values again can be explained by changes in trading volume and share price the sample experienced in the event period.
All in all, while the trading volume proxy provides ambiguous evidence on the introduction of the Tabaksblat Code having led to lower costs of equity capital, the bid‐ask spread proxy does support this notion. Since this proxy is considered an explicit measure of cost of equity capital (Leuz and Verrecchia [2000]) and that this proxy for all three events significantly behaves in the predicted direction, including one time at high significance for the (most reliable) controlled value , one can conclude that the introduction of the Tabaksblat Code indeed has led to lower cost of equity capital for large Dutch firms. Event number 5 (the date at which the Code became effective) is the most noteworthy event in terms of magnitude and significance.
17Of the firms in the sample, Numico has experienced substantial changes in its organizational
Table 3: Results for event studies Tabaksblat Code, partial sample
Volume proxy Event (n=38) Abnormal value T‐stat P‐value (sig.)
Event 3 (concept code) ‐426.889 ‐2.304 0.027**
Event 4 (definitive code) 460.882 1.187 0.243
Event 5 (date effective) 715.057 1.564 0.126
Bid‐ask uncontrolled Event (n=23) Abnormal value T‐stat P‐value (sig.)
Event 3 (concept code) ‐0.003 ‐1.280 0.212
Event 4 (definitive code) ‐0.002 ‐1.129 0.269
Event 5 (date effective) ‐0.004 ‐1.416 0.169
Bid‐ask controlled Event (n=23) Abnormal value T‐stat P‐value (sig.)
Event 3 (concept code) ‐0.006 ‐0.844 0.406
Event 4 (definitive code) ‐0.004 ‐1.468 0.155
Event 5 (date effective) ‐0.013 ‐3.377 0.002***
Asterisks indicate that the means of the two groups are significantly different using a two‐tailed t‐ test with *p<0.1; **p<0.05 and ***p<0.01
5.2) HYPOTHESIS 2
Turning to the second hypothesis, a comparison between the impact and significance of the Peters Report and the Tabaksblat Code is to be made. Now that the impact of the Tabaksblat Code on the cost of equity capital is described above, the results for the Peters Report are described. Then, a comparison is made.
Concerning trading volume, the results show that event number 1 and 2 have a positive impact on trading volume while event number 3 shows a moderate decline in trading volume. This last result, however, is insignificant while the significance of event number 1 is high at p<0.01.
The results for the bid‐ask spread do not show consensus either. In contrast to the expectations, except for the uncontrolled bid‐ask proxy in event 1, only positive (increasing) effects of the Peters Report on the bid‐ask spread are found. In case of the uncontrolled proxy, event number two is found to have a significant positive effect on the bid‐ask spread with significance p<0.01. In case of the bid‐ask proxy controlled for trading volume and price movements, both event number 1 and 2 are found to have a significant positive influence at p<0.05 and p<0.01 respectively. Again, the difference in significance between uncontrolled and controlled values can be explained by changes in trading volume and share price the sample experienced in the event period.
for further research. At least, the results above do not give reason to believe that the Peters Report has had a significant influence on the cost of equity capital of large Dutch firms.
Table 4: Results for event studies Peters Report
Volume proxy Event (n=38) Abnormal value T‐stat P‐value (sig.)
Event 1 (concept report) 354.694 3.335 0.002***
Event 2 (definitive report) 113.040 1.384 0.175
Event 3 (monitoring report) ‐70.004 ‐0.833 0.410
Bid‐ask uncontrolled Event (n=23) Abnormal value T‐stat P‐value (sig.)
Event 1 (concept report) ‐0.040 ‐1.190 0.246
Event 2 (definitive report) 0.361 11.156 0.001***
Event 3 (monitoring report) 0.012 0.790 0.437
Bid‐ask controlled Event (n=23) Abnormal value T‐stat P‐value (sig.)
Event 1 (concept report) 0.006 2.572 0.017**
Event 2 (definitive report) 0.271 8.267 0.001***
Event 3 (monitoring report) 0.019 0.997 0.328
Asterisks indicate that the means of the two groups are significantly different using a two‐tailed t‐ test with *p<0.1; **p<0.05 and ***p<0.01
6) CONCLUSION Inspired by the introduction of a new corporate governance code in the Netherlands, the Tabaksblat Code, this thesis analyzes the impact of the Code on the cost of equity capital of large Dutch firms. Also, it compares the impact of the Tabaksblat Code with the impact of its predecessor, the Peters Report. Two hypotheses are stated. Hypothesis 1 states that the introduction of the Tabaksblat Code has led to a lower cost of equity capital of large Dutch firms. Hypothesis 2 states that the Tabaksblat Code has had a more significant influence on the cost of equity capital than the Peters Report.
These hypotheses are tested using a sample of large Dutch firms that are part of the large‐cap or mid‐cap index at time of introduction of the Tabaksblat Code. Event study methodology is used to determine the magnitude and significance of several events related to the introduction of the respective codes. Because of difficulties in measuring the cost of equity capital directly, this thesis uses the bid‐ask spread and trading volume to proxy for changes in cost of equity capital. In the case of the bid‐ask spread, controls are used for the influence of changes in share price and trading volume.
indeed benefit from a lower cost of equity capital as computed by an accounting‐based valuation formula. This is also consistent with the results of this thesis.
Also, the importance of using controls for other determinants of the bid‐ask spread is shown. This is consistent with prior research by Leuz and Verrecchia [2000] and Callahan et al. [1997]. The date at which the Tabaksblat Code became effective, January 1st 2004, this way is identified as the most important event in terms of magnitude and significance.
This thesis does not provide complete insight into the question whether compliance with the Tabaksblat Code by large Dutch firms was real or merely formal. However, the results of this thesis do provide evidence that at least to some extent firms incorporated changes that decreased the information asymmetry between the respective firms and their investors.
Concerning the second hypothesis, this thesis has shown that the Tabaksblat Code has had a more significant influence on large Dutch firms’ cost of equity capital than the Peters Report had. It this way confirms the relevance of introducing a new corporate governance code after the Peters Report. Several limitations to this study can be noted. First, it remains difficult to measure changes in cost of equity capital. While the bid‐ask spread is shown to be a robust proxy for changes in cost of equity capital when controlled for other influences, it only measures a change in the proxy value and does not provide an indication of the absolute decrease in cost of equity capital. Furthermore, the second proxy used in this thesis, trading volume, is found to be an unreliable proxy. Using additional ways to measure cost of equity capital that require different data sources could therefore improve the methodology. This is a topic of further research. Second, no clear guidelines exist for the length of the estimation‐ and event windows. While the choice of this thesis to follow Leuz and Verrecchia [2000] produces significant results, an extension of this study is to test different lengths of these respective windows. Again, this is a topic of further research.