The effect of mandatory clawback adoption on earnings
quality
Name: Karin Hernández Fick Student number: 10466509
Thesis supervisor: Mr M. Schabus MSc Date: 26/06/2017
Word count: 11650
MSc Accountancy & Control, specialization Accountancy Faculty of Economics and Business, University of Amsterdam
Statement of Originality
This document is written by student Karin Hernández Fick 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.
Abstract
As of 2014, listed companies in the Netherlands have to adopt mandatory clawback provisions. The mandatory clawback provision authorizes the supervisory board to restore or adjust compensation when the compensation is not found to be within the scope of ‘reasonableness and fairness,’ the supervisory board also has to test the compensation of executives when there is a ‘change of control’ i.e. after a public offering. Voluntary clawback provisions are found to improve the perceived earnings quality and financial reporting quality. However, it is uncertain whether the improvements are due to the clawback or due to signaling by management. This research tests whether the earnings quality improvement perseveres when clawback provision adoption is mandatory. To test this, panel data of listed Dutch firms was used to perform a data-analysis. Using the proxies earnings persistence and meet/beat behavior by management, earnings quality was tested. The results show that there is no significant difference in earnings quality after adoption of mandatory clawback provisions. Thus the new law did not result in the enhancement of earnings quality.
Key words: clawback provisions, earnings quality, clawback law.
Preface
The research in front of you has been done as the last phase of my Master Accountancy & Control at the University of Amsterdam. The subject of clawback provisions caught my interest during a lesson of one of our first courses: Management Control. As I’ve always have had a strong sense of fairness, it seemed unfair to me that certain managers would enrich themselves, while harming the organization they work for, without having to repay their compensation they obtained when the damage that was done came out. Looking into the literature, I found an opinion article written by a Dutch accountant. He stated that mandatory clawback provisions would not have the wished effect because the new law was very similar to the previous arrangement. My curiosity was triggered, and so my research question arose.
Most importantly I would like to thank my thesis supervisor Mr M. Schabus, for all the help, advice and clear feedback I got during the thesis process. Given my statistical backlog, he helped me enormously explaining the steps towards quantitative research. I would also like to thank my supervisor at PwC Amsterdam, J. Kooter, who helped me by providing feedback and was always available for all my questions. I would also like to thank my study group friends, which made long days at the library bearable. I feel we’ve gotten to know each other very well this year and hope our friendship lasts, past the study era. At last I would like to thank my family, friends, and boyfriend, who always understood my lack of availability the last period.
Contents
Abstract ... 3 Preface ... 4 Contents ... 5 Introduction ... 6 1 Literature ... 8 2 2.1 Clawback Provisions in the US ... 82.1.1 Agency theory, the arising problem with performance contingent compensation, and the origin of clawback provisions ... 8
2.1.2 Development of Clawback Provisions in the United States ... 9
2.2 Clawback Legislation in the Netherlands ... 12
2.2.1 Clawback in the Corporate Code of Conduct ... 12
2.2.2 Adoption of clawback law in the Netherlands ... 12
2.3 Earnings Quality ... 13
2.3.1 Earnings Persistence ... 14
2.3.2 Meet or Beat behavior ... 15
2.4 Hypothesis development ... 16 Research methodology ... 18 3 3.1 Sample selection ... 18 3.2 Empirical models ... 19 3.3 Control variables ... 21 Results ... 22 4 4.1 Descriptive statistics ... 22 4.2 Multivariate Analysis ... 24 4.2.1 Earnings persistence ... 24 4.2.2 Meet/Beat behavior ... 26 4.3 Sensitivity Analysis ... 28
4.3.1 Robust regression Earnings Persistence ... 28
4.3.2 Sensitivity Analysis Meet/beat behavior ... 30
Summary and Conclusion ... 32
5 References ... 34
6 Appendix ... 38
Introduction
1
Clawback provisions are one of the tools of management control, to prevent executives of firms to act in self-serving behavior. Clawback provisions allow that compensation that is unearned or excessive to be recouped by the company. The clawback provision removes the incentive of executives to manage earnings to obtain higher bonus compensation, as their compensation will be recouped, i.e. ‘clawed back,' when it appears to be wrongfully obtained. In the United States, clawback provisions have gained popularity since the adoption of the Sarbanes-Oxley Act (SOX), as it enforced the SEC to recoup wrongfully obtained compensation and impose the use of a clawback provision in the case of prior restatements. Voluntary clawback provisions are increasing in popularity as well, even though the cause of this increase of popularity is not well defined, as it can be used as a tool for a firm to control their management, as to signal for strength in management control to the public (Chan et al., 2012).
Within the European Union, there is significant support for clawback provisions, as it is encouraged by the European Commission to include clawback provisions in all the member states’ corporate governance code. Germany, the United Kingdom, and Denmark are a few examples of countries that have included clawback provisions in their corporate governance code. The Netherlands initially included clawback provisions as well within their corporate governance code, which all listed firms were expected to apply or explain why they didn’t apply it. The Dutch legislator decided that including clawback provisions in the corporate governance code was not sufficient, and they decided to implement it within the common law. As we see now, Dutch listed firms are obliged to adopt clawback provisions since 2014.
While there has been some research about the effect of voluntary clawback provisions on accounting matters (Chan et al. 2011, 2015; Dehaan et al. 2013, Iskandar-Datta and Jia, 2013), research about mandatory clawback provisions, is rare (Denis, 2012). Probably because in the United States only financial institutions have mandatory clawback provisions, and the US is usually used as a sample in clawback provision research. So far no prior research has investigated what the effect of nation-wide mandatory clawback adoption would be. Therefore my research question is as follows: Does the law of mandatory clawback
enhance firms’ earnings quality?
The study of Chan et al. (2012) investigates the relationship between voluntary clawback adoption and earnings quality. They found that the incidence of accounting
restatements declines after firms initiate such provisions. Also, the research shows that investors and auditors view such provisions as associated with increased accounting quality and lower audit risk. Specifically, they find that the earnings response coefficient (ERC) of firms increases after the adoption of clawback provisions. Denis (2012) discusses issues surrounding government regulation of clawback provisions in particular and corporate behavior more generally. The reason there might be a different result between the voluntary and mandatory clawback provision is that the results of the voluntary clawback provision research could be due to the signaling effect (Chan et al., 2012). This effect is removed when clawback provisions become mandatory. After the adoption of the law, all publicly held firms will have the same clawback conditions, and so the signal effect will be neutralized. The fact that this effect would be offset makes research on whether earnings quality is positively related to mandatory clawback provision very interesting. It will show whether the results found in the voluntary sample persists in the mandatory sample.
The United States is planning to include mandatory clawback adoption as well in the upcoming Dodd-Frank Act 954. The Dodd-Frank Act 954 will require that all listed firms adopt clawback provisions, which will be triggered by a restatement, regardless of fault. Legislators of the United States can profit from the results of this research, as they can foresee the impact of mandatory clawback adoption by looking at the Dutch sample.
Literature
2
2.1 Clawback Provisions in the US
Within this section the origin of clawback provisions will be discussed, there will be looked at the agency problem, which clawback try to diminish. Public scrutiny will be discussed as well. In various newspaper articles, there is a trend of an increasing demand of the public towards organizations to increase the transparency of firms and punish misconduct of managers or CEOs.
The expansion of the use of clawback provisions will also be discussed. The Sarbanes-Oxley Act (SOX) has enacted a law where the Securities and Exchange Commission (SEC) can recover bonus or other incentives earned when there has been misconduct and firms have had to restate their financial statements. Voluntary clawback provisions are increasing in popularity as well, even though the cause of this increase of popularity is not well defined, as it can be used as a tool of management control or as a signal of openness to the public (Chan et al., 2012). It is interesting at looking at the results of the mandatory clawback provisions on earnings quality.
2.1.1 Agency theory, the arising problem with performance contingent compensation, and the origin of clawback provisions
Agency theory has a few assumptions made on human, organizational and information level (Eisenhardt, 1989). On the human level, we assume that the parties have self-interest, bounded rationality and are risk-averse. This means that we all act with the available information in our own best interest without taking too many risks unless we are compensated for it. On the organizational level the assumptions are that 1) there is a partial goal conflict between the parties, 2) efficiency is the effectiveness criterion and 3) there is information asymmetry between the agent and the principal. This means that there is a conflict of interest where the principal and agent do not possess the same information, but they both want to act as efficient as possible. On the information level, the assumption is that you can buy information.
Due to the assumptions made, contracting problems arise. Contracting problems are defined as agency problems (moral hazard and adverse selection) and risk sharing problems. These agency problems can lead for example to:
- Waste of the entities sources - Under- or overinvestment
The conflicts created by the separation of management and ownership are repaired by contracts that try to harmonize the interests of both managers and owners. These contracts give managers efficient incentives to maximize shareholder value, as well as acting in their own best interest. This is referred as the ‘optimal contracting approach.' One of these incentives is contingent compensation. This type of compensation is contingent upon the achievement of certain goals or targets. Incentives-based executive compensation can however also be a part of agency problems when it does not achieve to harmonize the interests of the executives and the owners of the entity. This approach is the ‘managerial power approach,' which highlights rent-seeking behavior by management (Bebchuk and Fried, 2003). Information asymmetry allows executives to influence the targets set and the height of the compensation received if the targets are met.
Executive compensation packages usually exist of a fixed salary, short-term and long-term contingent compensation, and a pension package. To get a fixed salary and pension package is mandatory in the Netherlands (art. 7:610 BW), the contingent compensation is to align the interests of the executives with the owners, in this case, the shareholders of the publicly held firms.
The contingent compensation received depends on the performance of the company, but also on the company characteristics, like industry, size, complexity, etc. Also, the culture and geographical location of the firm is important, as Jansen et al. (2009) find that the incentive (contingent) compensation in the United States is a bigger portion of the total compensation compared to the compensation in the Netherlands.
Prendergast (1999) also mention that executives demand a bigger compensation when they are exposed to increased risk.
2.1.2 Development of Clawback Provisions in the United States
Jensen and Meckling (1992) argue that public scrutiny can discipline an organization in the following way: an organization will try to avoid a negative public opinion, as these could lead to populist political interventions. These interventions could put further constraints on executive pay.
Since the 1990’s executive compensation became higher than ever before, this caused a public outrage about the amount received by executives of large enterprises. Public opinion, when channeled with media attention, can act disciplinary, (Dyck et al., 2006) as it did in this period. The United States Congress passed a law making executive’s compensation more transparent. Executives (CEO’s and CFO’s) had to report how much compensation they received and what the composition of the compensation was, which would be overseen by the SEC. The increase in transparency had two consequences; first, the shareholders and stakeholders, as well as the public, had better insight within the compensation packages received by executives, which made the information asymmetry smaller and empowered public scrutiny. This effect was what the SEC was hoping to achieve. Second, executives became aware of the compensation of their fellow executives, which gave them bargaining power when the compensation packages were defined (if the compensation were not up to the level of their colleagues they would look for other jobs). So the transparency resulted in even higher executive pay than before (Fried and Shilon, 2011).
After the fraud committed by Enron and WorldCom and other cases, there was a massive public outrage, because the compensation the executives earned while committing fraud, was not returned to the company and its shareholders. These cases led to the passing of the SOX (explained below) rather fast, because there was little arguing about these practices being unacceptable.
There have been regulatory responses to respond to the corporate and accounting scandals at Enron, WorldCom and other entities, making executives take responsibility when a misstatement occurs. First, the Sarbanes-Oxley Act of 2002 (SOX) came to light, and later the Dodd-Frank Act was created in the United States, both contain a clawback section. In Europe, most countries adopted clawback provisions within their corporate governance code. These regulatory measures have led to a widespread adoption of (voluntary) clawback provisions.
The Sarbanes-Oxley Act, which passed in 2002 through the US Congress, included a variety of measures that were intended to regain the trust of the investors in capital markets, after the several scandals that happened at the end of the 1990’s and beginning of the 2000’s. It mandates tighter internal controls and increases disclosure requirements as well as enhances penalties to financial misconduct. It also contains a clawback provision in section 304. This provision applies to publicly traded companies and requires that CEO’s and CFO’s bonuses that were obtained as a result of misconduct, “with any financial reporting
requirement under the securities laws” (US Congress, 2002), to be ‘recouped’ or ‘clawed back’ by the SEC. The provision applies to any bonus received, or other incentive-based or equity-based compensation, or the profits received from selling these securities, within 12 months from the issuance of the financial statements.
Even though the SEC can act based on this regulation, the actual recoupments enforced by the SEC fall far behind the amount of financial restatements in the US (Fried and Shilon, 2011). This shortcoming is partial to blame for the fact that ‘misconduct’ and ‘an intentional act to mislead’ are hard to prove, as well as the limited resources of the SEC to enforce the law (Babenko et al., 2012; Chen et al., 2012).
In the United States, where most of the research on clawback provisions has been held, there is no legislation or corporate governance code encouraging companies to adopt a clawback provision. Because databases from the United States are more extensive that those in Europe, most research that has been performed has focused on voluntary clawback adopters.
The voluntary adoption of clawback provisions after the implementation of SOX section 304 has exponentially grown. In the year 2000 less than 1% of the S&P 1500 companies had adopted a clawback provision, while in 2011 almost 50% of SP 1500 and 70% of SP 500 adopted a clawback provision (Babenko et al., 2012). This increase has multiple reasons according to previous research, which will be discussed beneath.
Chan et al. (2012) discuss the benefits the voluntary adoption of clawback provisions can have on an organization. They find that firms with clawback provisions show increases in earnings quality and reduce audit risks compared to the time before implementation, showing a significant decrease in accounting restatements. They argue that this implies that incentives for earnings manipulation have decreased. Chan et al. (2013) find that clawback provisions adoption has a positive effect on the bank loan contracts provided by banks, decreasing interest rates, increasing loan maturity and decreasing loan collateral. The decline in interest rate indicates that banks perceive the adoption of a voluntary clawback as an efficient manner to improve the quality of financial reporting. Dehaan et al. (2013) find that the voluntary adoption of clawback provisions has an improvement on financial reporting quality as well as the investors’ and analysts’ perceptions about financial reporting quality. These findings contributes to the question whether firms adopt clawback provision to improve the firm’s financial statements, decreasing perverse incentives, or they adopt clawback provisions to
signal investors and analysts about the reporting quality of the firm, without the provision having any real effects on reporting quality.
2.2 Clawback Legislation in the Netherlands
Since the 1st of January 2014 Publicly held firms, cooperatives and private limited liability companies that qualify as banks, mutual insurance companies qualifying as an insurance company, and financial institutions have to abide to the law on clawback. It is interesting to look at the reasoning and process of the implementation of the law, asking the following questions: How much does the current situation differ from the situation before? What is the expected effect and or result of the new legislation? These questions will be discussed in this section.
2.2.1 Clawback in the Corporate Code of Conduct
In the Netherlands, before the clawback law, there was already a regulation in place about clawback provisions in the Dutch corporate code of conduct. The corporate code of conduct t applies to all publicly traded entities in the Netherlands. The code of conduct has principles, defined as widely supported general views about good corporate governance. They provide ‘best practices,' which can be seen as the elaboration of the general principles (MCCGC, 2008).
It is, however, possible for the firms to which the code applies, to deviate from it. As long as shareholders have judged about the circumstances that caused the deviation from the code. The company has to mention in their annual report about the way they applied the principles and best practices. If they did not apply one, they have to explain this.
Only the shareholders have to oversee whether the board and the supervisory board have complied with the code and can call them to account.
2.2.2 Adoption of clawback law in the Netherlands
As mentioned in the paragraph before, only the shareholders of the company oversaw the clawback arrangement before. The supervisory board could restore or adjust the compensation, but if there was a possibility to deviate from this, by explaining the situation.
The Dutch government hopes to achieve with this ruling to facilitate that the supervisory board can oversee that compensation obtained by upper management of companies, so it can oversee that the compensation contributes to the long-term interest of the
firm (Tweede Kamer, 2010). The law is intended to clarify that the supervisory board of a public limited company has the authority to recoup (clawback) or adjust the amount of the compensation, regardless of the efforts of individual boards of commissioners to come to contractual agreements, and independently of the 'comply or explain' principle.
The law also contains a “change of control” arrangement, in which the supervisory board is obligated to test whether paid bonuses due to a public offering are unacceptable to the standards of ‘reasonableness and fairness’ (Tweede Kamer, 2010).
Reasonableness and fairness is a commonly used concept in the Dutch law. This concept is used as a supplement of the law. If the law has a significantly unfair or unreasonable outcome which was not intended or foreseen before implementation, there can be appealed to the concept of ‘reasonableness and fairness’ to mitigate the outcome. Reasonable and fair are very subjective concepts, of which the outcome is not fixed. Every case should be judged individually, as stated by the Dutch law (art 3:12 BW)1:
“At determining what the principle of ‘reasonableness and fairness’ demands in a specific situation, one has to take into account the generally accepted legal principles, the fundamental conceptions of law in the Netherlands and the relevant social and personal interests which are involved in the given situation.”
In the case of reasonable and fair bonuses paid before a public offering, there will be looked at the explanation of the bonuses and the underlying financial numbers, to see whether they support the paid bonus.
2.3 Earnings Quality
Earnings Quality is a widely used concept in accounting research, as it is important to determine the usefulness of accounting information. The IFRS conceptual framework mentions two characteristics that information should have, which will help to determine what high-quality information on earnings is, i.e., earnings quality. First, they mention that information should be relevant. Relevance means that the information is capable of contributing to decision-making and is repeatable and sustainable. Second, they mention faithful representation, which is achieved when information is complete, neutral and free from material error (IASplus, 2016). The meaning of these characteristics is dependent on the
context of the decision model and the specific decision-maker (Dechow et al., 2010). The reported earnings should represent the underlying economics of the firm and at the same time be persistent and predictable. If these characteristics are met, investors will be able to value a firm with a solid base.
If the requirements of the users of earnings information are met, we can acknowledge the earnings quality to be high. There are however different reasons why the expectations would not be met. For example, following of poor quality accounting standards, will result in low-quality earnings information (Lo, 2007). The same goes when there are market-wide problems due to the economic crisis or environmental disasters. These problems could harm the persistence of the earnings shown by a firm. In this paper there will be a focus on the human factor behind earnings quality, specifically intentional acts to distort the financial quality of earnings.
Dechow et al. (2010) name three categories in which earnings quality proxies can be classified: Properties of earnings, investor responsiveness to earnings, and external indicators of earnings misstatements. This paper will focus on the earnings quality proxies in the category properties of earnings. Properties of earnings describe how earnings are built, and what can influence them. The reason that the other categories of proxies are not researched in this paper is that the other categories are difficult to investigate with the current data set, as financial restatements in the Netherlands are rare, which is described by Dechow et al. (2010) as the external indicator of earnings misstatements. Also, the clawback law will be of influence to all firms in the Netherlands, so I do not expect that the Earnings Responsiveness Coefficient (ERC), as described in Dechow et al. (2010) as the investor responsiveness category, will be effective to measure the earnings quality. Even though prior research did focus on these proxies (Chan et al., 2012; Dehaan et al., 2013)
2.3.1 Earnings Persistence
Earnings persistence is a largely used way to investigate earnings quality. It especially focuses on the usefulness of earnings to equity investors for valuation (Dechow et al., 2010). The assumption to use this proxy is that more persistent earnings will provide better inputs for equity valuation models, and this way be more helpful in the decision making process of an equity investor.
When an earnings stream is more persistent, current earnings will be more useful to measure future performance given annuity valuations will have fewer errors, which leads to a
higher quality prediction of future earnings. Research of the consequences of earnings persistence has focused on the equity market consequences. There are consequences highlighted in Dechow et al. (2010). Most important, more persistent earnings will yield a higher equity market valuation; therefore it will yield positive equity market returns and stronger stock price response, this is supported by research done by Kormendi and Lipe (1987), Collins and Kothari (1989), and Easton and Zmijewki (1989).
Earnings persistence is not relevant to the valuation of high-growth firms, as these firms usually have negative cash flows (due to big investments made) and small (negative) earnings (because of the expense of the investments made that cannot be capitalized). There can be expected that these firms will make positive future cash flows and higher earnings in the future (Penman and Sougiannis, 1998). For this reason, another valuation technique is required for this type of firms, like income; as a result, earnings persistence does not reflect earnings quality for these firms.
2.3.2 Meet or Beat behavior
The study of Degeorge et al. (1999) describes why earnings management is used to meet or beat thresholds. Earnings can explain the return on equity better than other predictors, like cash flows and dividends. Because of this explanatory quality, the rewards set for executives of firms is also focused on the earnings achieved during their tenure. Executives have discretion on the earnings they present on financial statements, given there is room for flexibility in the accounting treatment within accounting standards, like in matters of sales recognition, inventory, pensions, and so on. Also, managers have opportunities to performing real earnings management for example by boosting sales through cutting prices, or not investing in an opportunity with positive Net Present Value (NPV) to not lower earnings. Through earnings management, current earnings will be stimulated, but it may be at the cost of the future earnings of the firm beyond the tenure of the executive. The short-term rewards explain that executives have the incentives and the opportunity to engage in earnings management.
Executives have a special focus on thresholds, as stakeholders, other than investors, are also concerned about the firm’s viability and profitability, like banks, customers and employees, also focus on them. Degeorge et al. (1999) name three reasons why these stakeholders focus on thresholds. First, they mention that the line between meeting and failing the norm is crucial for humans. We have the tendency to divide the world into
categories (Glass and Holyoak, 1989), where in this case meeting or beating the norm would be seen as the right category versus failing the norm, which would be the wrong category. Second, they mention that stakeholders need to choose from various risky alternatives, this leads to the need of a reference point to compare them to make the best decision (Khaneman and Tversky, 1979). The benchmark is crucial to them, and the utility function is S-shaped around the benchmark, convex beneath the zero-point and concave above it. The S-shaped utility function implies that a small shortcoming from the zero-point would weigh more than a small trespass of the zero-point to a stakeholder in their utility decision. The third reason is the transaction costs involved in assessing a firm. To reduce the transaction costs, stakeholders use ‘rules of thumb’ as the real calculation would be too costly. For example, a bank giving out a loan will use these rules of thumb, involving looking at thresholds (like the threshold of negative or positive earnings), to decide whether to lend money to the firm or not. Other stakeholders will see the decision made by the bank as a signal, and will take this into account in their investment/ delivery/ working, etc. decision; as for them, the real calculation is too costly as well (Vashishtha, 2014).
Using earnings management to meet or beat a threshold would lead to a false representation of the economic truth behind the financial statements, harming earnings quality.
2.4 Hypothesis development
In this section, the hypotheses of the research are provided. There will be two hypotheses, of which the second one is divided into two sub-hypotheses. As mentioned before, earnings quality is driven by two factors: relevance and faithful representation of the information. These hypotheses will help to support or reject the research question.
Given the expectations of the law and the effect of voluntary clawback adoption shown in the studies of Chan et al. (2012, 2013, 2015), I expect that earnings quality will be higher in the period after the adoption mandatory clawback provisions than the period before. To be able to measure whether the earnings quality is enhanced, there will be proxies of earnings quality used in each both hypotheses; the proxies are described in the paper of Dechow et al. (2010).
The first proxy used is earnings persistence. Earnings persistence is a proxy that helps within the decision usefulness of accounting information. So is the assumption that more
persistence firms yield better inputs to equity valuation models, which would enhance the relevance of the information on earnings (Dechow et al., 2010).
H1: Earnings persistence is after mandatory clawback adoption higher than before mandatory clawback adoption.
The second proxy for earnings quality is meet/beat behavior towards benchmarks. The just meeting or beating of a benchmark is a sign of earnings management (Degeorge et al., 1999; Kent and Routledge, 2017), and thus a decrease in this behavior would enhance the faithful representation of the earnings quality. I will test two benchmarks and see if meet/beat behavior is performed on them. The first sub-hypothesis benchmark is the positive earnings change, described as the unwillingness of management to show a decrease in earnings compared to the prior period.
H2a: Meet/Beat behavior performed on the positive earnings change benchmark is lower after the adoption of mandatory clawback provisions than before
The second sub-hypothesis benchmark is positive earnings, described as the resistance of managers to show negative earnings, as these have a significant impact on stakeholders (Vashishtha, 2014), and so they engage in earnings management to avoid it.
H2a: Meet/Beat behavior performed on the positive earnings benchmark is lower after the adoption of mandatory clawback provisions than before.
Research methodology
3
3.1 Sample selection
This research will be performed as an empirical archival research using data from publicly held companies listed on the Euronext Amsterdam Stock Exchange (AEX, AMX, and AScX). The research is focused on the three years before the adoption of the law on clawback provisions and three years after the adoption, which results in the period of 2011-2016. Data is also retrieved from the years 2009 and 2010 to create lagged variables and changes in lagged variables. The data is retrieved from Osiris, a database containing financial information from publicly held firms.
The initial sample of Euronext Amsterdam stock exchange existed of 111 firms. The companies in the financial and real estate industry (NAICS 2012 code 52 and 53) were excluded because they have to obey additional legislation, which may cause a bias within the sample. To test the first Hypothesis, I will need subsamples, of the period before mandatory clawback provision adoption and the period after the mandatory adoption, for this reason, there should be at least two firm years of each period per firm. Hence information of the variables should be available at least from the years 2012-2015.
Also, three firms were manually excluded from the sample. Heineken Holding was excluded because Heineken NV was already in the sample, and including the holding would let some variables to be based on the same results. ING group NV and Van Lanschot NV were excluded, as they are financial institutions. Finally, 44 observations were excluded due to missing control variables in the sample. The total sample contains 61 unique companies and 322 observation years. All continuous variables have been winsorized at the first and 99th percentile, so extreme observations are excluded; this is done because extreme observations can have an enormous impact on the regressions, which produces an unwanted bias.
TABLE 1
Data collection and manipulation
Manipulation of data firms observations
AEX, AMX, AScX 111 666
- firms from financial or real estate industry (NAICS 2012 code 52 &53)
88 528
- firms without enough
firm-years 64 384
- manually removed firms (ING group NV, Heineken Holding, van Lanschot NV)
61 366
- obs. missing control variables 61 44
Total
61 322
3.2 Empirical models
Hypothesis 1 predicts that the period following the mandatory adoption of clawback provisions, show more persistent earnings compared with the period before the mandatory adoption. To test the first Hypothesis the most commonly used empirical model to regress earnings persistence will be used (Dechow et al., 2010). The model is estimated as follows:
𝐸𝐴𝑅𝑁𝐼𝑁𝐺𝑆!!!= 𝛽! + 𝛽!𝐸𝐴𝑅𝑁𝐼𝑁𝐺𝑆!+ 𝛽!𝑆𝐼𝑍𝐸!!!+ 𝛽! 𝐺𝑅𝑂𝑊𝑇𝐻!!!+ 𝛽!𝐵𝑇𝑀!!! + 𝛽!𝐿𝐸𝑉𝐸𝑅𝐴𝐺𝐸!!!+ 𝛽!𝐵𝐵𝑃!!!+ 𝜀!!! 1
EARNINGS 𝑡+1 is the dependent variable reflecting future earnings (expected earnings in the
next period); it is defined by net incomet+1 scaled by lagged total assets (assetst). EARNINGS𝑡
is the independent variable reflecting current earnings, which is defined by net incomet scaled
by lagged total assets (assetst-1). The control variables will be discussed in section 3.3.
Following prior research (Dechow et al., 2010), I will use a baseline regression model where earnings persistence will be calculated looking at the relationship between current earnings and future earnings. The regression will be performed in subsamples, one from the period before mandatory clawback adoption and one from the period after mandatory
clawback adoption. The influence of the mandatory clawback adoption will be exposed by the differences between β1 shown in the different samples, where a higher β1 would represent
a more persistent earnings stream.
Hypothesis 2 predicts that mandatory clawback adoption will lower the incentive to meet or beat thresholds by managers. To test the second Hypothesis, a logit regression will be performed to examine the relation between Meet/Beat behavior and mandatory adoption of clawback. The model is estimated in two ways, testing meet/beat behavior the zero earnings threshold and for the zero earnings change threshold:
𝑀𝐵𝑍𝐸𝑅𝑂! = 𝛽!+ 𝛽!𝐶𝐵!+ 𝛽!𝑆𝐼𝑍𝐸!+ 𝛽!𝐺𝑅𝑂𝑊𝑇𝐻!+ 𝛽!𝐵𝑇𝑀!+ 𝛽!𝐿𝐸𝑉𝐸𝑅𝐴𝐺𝐸!
+ 𝛽!𝐵𝐵𝑃!+ 𝜀! 2
𝑀𝐵𝐶𝐻𝐴𝑁𝐺𝐸! = 𝛽!+ 𝛽!𝐶𝐵!+ 𝛽!𝑆𝐼𝑍𝐸!+ 𝛽!𝐺𝑅𝑂𝑊𝑇𝐻!+ 𝛽!𝐵𝑇𝑀!+ 𝛽!𝐿𝐸𝑉𝐸𝑅𝐴𝐺𝐸! + 𝛽!𝐵𝐵𝑃!+ 𝜀! (3)
Following prior research (Kent and Routledge, 2017; Degeorge et al., 1999) the dependent variable of the third model is MBZEROt. MBZEROt is a dummy variable that is one if
earnings per share (EPS), defined as net income divided by total shares outstanding, is equal or slightly bigger than zero (0≤ EPS<0,05) and zero otherwise. CB is another dummy variable, which indicates the period before and after the mandatory clawback adoption, with 0 before adoption (years 2011-2013) and one after mandatory adoption (years 2014-2016).
MBCHANGEt reflects the incentive to manage earnings towards a zero earnings change. It is
a dummy variable that is one if the change in EPS within two periods is equal or slightly bigger than zero (0≤ change in EPS<0,05). Given the observations with the scope of prior research (MBZERO (MBCHANGE) =1 if 0 ≤ (change in) EPS<0,02) gives a minimal amount of MBZERO and MBCHANGE observations, I broadened the scope2. In the
robustness test there will be a test for the smaller scope (0≤(change in EPS<0,02) (Kent and Routledge, 2017; Degeorge et al., 1999).
3.3 Control variables
Multiple factors can influence the dependent variables in the three models I will test. To be able to control for these effects, they should be adopted in the models. The control variable
SIZE is included because the size of a firm is correlated to the information available in
security markets, more information leads to better decision-making by investors (Easley & O’Hara, 1987). Following previous research, SIZEis measured as the natural logarithm of the total assets of a firm. The variable GROWTH is included because the earnings quality of high growth firms cannot be measured by earnings persistence, as they are expected to have higher earnings each consecutive year (Penman and Sougiannis, 1998). GROWTH is calculated by the difference in sales compared to last year sales divided by last year sales (!"#$!!!!"#$!!!!
!"#$!!!! ). The control variable LEVERAGE is added due to its relation to future earnings (Nissim and Penman, 2003), as an increase of leverage can decrease earnings in the first instance, but is likely to cause higher future earnings and decrease in future liabilities.
LEVERAGE is calculated by dividing total assets by total liabilities. Also, the control variable
Book-to-Market ratio growth is added (BTM). This control variable is known to be able to indicate risk-taking behavior and the use of investment opportunities by management. The market value shows how investors perceive the ability of the assets to create future cash flows. BTM is measured as the difference between the book-to-market ratio of this year compared to last year book-to-market ratio (which is (total assets – total liabilities) divided by market capitalization). At last the control variable GDP is added. This variable shows the economic growth in the Netherlands per year and is measured as the growth in percentage compared to last year (CBS, 2017). The economic growth can be of influence on the dependent variables. Magnan and Markarian (2011) mark the period of 2007 till 2010 one of the most severe economic and financial crisis in history. This period was followed by a slow recovery (2011-2013) and thus shows smaller growth than the period afterward, as the economy was recovering faster at that time (2014-now).
Results
4
In this section, the results of the statistical analyses performed to test the hypotheses will be discussed. First, the descriptive statistics of all the variables will be shown and discussed. After that, the results of the following tests will be discussed: H1 will be tested comparing the correlation between current and future earnings in the period before mandatory clawback adoption with the period afterward. H2a and H2b will be tested by performing a logit regression to calculate the relation between meet/beat behavior towards the zero earnings (change) benchmark and the adoption of mandatory clawback provision. At last the sensitivity analysis will be discussed.
4.1 Descriptive statistics
Table 2 provides an overview of the descriptive statistics of the variables used in both hypotheses. These statistics have been divided into two panels, where Panel A shows the statistics in the period before mandatory clawback adoption and Panel B the period after. This is done to provide a clear view of the data used in the tests concerning earnings persistence, as the sample is also divided into the two periods.
TABLE 2 Descriptive statistics
Panel A shows statistics of the sample before mandatory adoption of clawback provisions.
Panel A
N Mean St.Dev Min. 25% Median 75% Max
MBCHANGE 169 0.071 0.258 0 0 0 0 1 MBZERO 169 0.030 0.170 0 0 0 0 1 EARNINGSt+1 169 0.006 0.162 -0.976 -0.015 0.033 0.075 0.527 EARNINGSt 169 0.017 0.174 -0.976 -0.010 0.045 0.083 0.685 BTM 169 -0.007 0.369 -1.843 -0.121 -0.009 0.126 1.356 GROWTH 169 0.070 0.391 -0.824 -0.048 0.012 0.100 2.096 SIZE 169 12.934 2.361 5.793 11.188 12.816 14.697 17.398 LEVERAGE 169 1.980 1.199 0.399 1.488 1.784 2.171 14.387 GDP 169 0.001 0.012 -0.011 -0.011 -0.002 0.017 0.017
In Panel A we see that the mean of the MBCHANGE dummy is 0.071, this implies that 7,1% of the observations in the subsample has an EPS (earnings per share) value that increased between the 0 and 0,05 compared to the EPS in the previous period. The mean of the
MBZERO dummy is 0.030; this shows that 3.0% of the observations in the subsample have
an EPS between the 0 and 0,05.
Panel B
N Mean St.Dev Min. 25% Median 75% Max
MBCHANGE 163 0.061 0.241 0 0 0 0 1 MBZERO 163 0.031 0.173 0 0 0 0 1 EARNINGSt+1 163 0.010 0.258 -2.667 0.007 0.046 0.077 0.527 EARNINGSt 163 0.009 0.248 -2.250 0.004 0.040 0.076 0.685 BTM 163 -0.057 0.615 -2.560 -0.191 -0.017 0.119 1.865 GROWTH 163 0.026 0.253 -0.824 -0.048 0.021 0.080 2.096 SIZE 163 13.121 2.375 6.719 11.261 13.072 14.860 17.417 LEVERAGE 163 2.645 4.103 0.399 1.391 1.847 2.322 42.084 GDP 163 0.019 0.003 0.014 0.014 0.020 0.022 0.022
Panel B shows statistics of the sample after mandatory adoption of clawback provisions.
In Panel B we see that the mean of the MBCHANGE dummy is 0.061, this implies that 6.1% of the observations in the subsample has an EPS (earnings per share) value that increased between the 0 and 0,05 compared to the EPS in the previous period. Compared to Panel A we see a small decrease, though I cannot infer from here whether the differences are statistically significant. The mean of the MBZERO dummy is 0.031; this shows that 3.1% of the observations in the sample has a small positive EPS result. This result is a slight increase compared with Panel A, implying there is a very little difference in the number of observations with a small positive EPS results between the two samples. However, I cannot infer from here whether the differences are statistically significant.
Table 3 shows the Pearson's correlation between all the variables for the sample containing both the period before as the period after mandatory clawback adoption (2011-2016). By combining the subsamples, the correlation between the dummy variable CB (clawback) is visible for all the variables within the regressions. The table shows the correlation, that has three stars (***) at a significance level of 1% (p<0.01), two stars if the significance level of 5% (p<0.05) and one star at a significance level of 10% (p<0.10).
TABLE 3
Pearson’s correlations between dependent and independent variables
Table 3 shows that none of the explanatory variables are significantly correlated with the CB dummy variable. We can see that variables MBCHANGE and MBZERO have a very small (respectively -0.02 and 0.00), yet not significant correlation with CB. However, there is a significant (5% level) little correlation between CB and LEVERAGE. There is also a correlation between CB on GDP (0.70 with 5% significance level). However this can be explained by external factors; the period before mandatory clawback adoption (2011-2013) can be seen as the period directly following the financial crisis (Magnan and Markarian, 2011) which is marked by slow recovery, resulting in slower growth, compared to the period after the clawback adoption (2014-2016). The economy was in this period significantly recovered, resulting in higher GDP growth.
4.2 Multivariate Analysis
4.2.1 Earnings persistence
Table 4 is divided in Panel A and Panel B, where Panel A shows the Ordinary Least Squares (OLS) regression of the subsample representing the period before mandatory clawback adoption, and Panel B showing the OLS regression of the subsample representing the period after mandatory clawback adoption. The sample is divided to create a simple way to observe
Variable 1 2 3 4 5 6 7 8 9 10 1. CB 1.00 2. EARNINGSt+1 0.01 1.00 3. EARNINGSt -0.02 0.44*** 1.00 4. BTM 0.04 0.22*** 0.29*** 1.00 5. GROWTH -0.07 0.02 0.01 -0.03 1.00 6. SIZE -0.05 -0.17** 0.02 0.14** -0.03 1.00 7. LEVERAGE 0.11** -0.38** -0.01 -0.11* -0.12** 0.18*** 1.00 8. GDP 0.70** 0.00 0.04 0.04 0.01 0.13** 0.09 1.00 9. MBCHANGE -0.02 0.00 0.00 -0.13** -0.01 -0.02 -0.04 -0.01 1.00 10. MBZERO 0.00 0.00 -0.04 -0.07 0.01 -0.02 0.00 -0.02 0.17*** 1.00
the difference between earnings persistence for the pre- and post- mandatory clawback adoption group. The Hypothesis implies that the coefficient of EARNINGSt in Panel A should
be lower than in Panel B as the explanatory power of current earnings on future earnings should be increased by mandatory clawback adoption, making earnings more persistent.
Panel A shows that the coefficient of EARNINGSt on EARNINGSt+1 is 0.4794 and
highly significant (1% level). Which implies that EARNINGSt+1 is moderately correlated with
current earnings.
Panel B shows the results of the analysis on the earnings persistence model after mandatory clawback adoption. The coefficient of EARNINGSt on EARNINGSt+1 is 0.3659
and highly significant (1% level) as well. This result intuitively contradicts the prediction of the hypotheses, showing a decrease in the correlation between the period ex-ante mandatory adoption and the period ex-post. I conduct a test for statistical differences in coefficients to examine whether the differences in magnitudes are significant or negligible.
TABLE 4 OLS regression
Panel A Panel B
Variable EARNINGSt+1 Variable EARNINGSt+1
Coef. t P>t Coef. t P>t EARNINGSt 0,4794 8,1300 0,0000 0,3659 5,4600 0,0000 SIZE 0,0149 3,4300 0,0010 0,0010 0,1300 0,8930 GROWTH -0,0178 -0,7100 0,4780 0,0307 0,4800 0,6340 BTM -0,0306 -1,0300 0,3030 -0,0548 -2,0300 0,0440 LEVERAGE 0,0319 3,8700 0,0000 -0,0301 -7,4200 0,0000 BBP -0,7327 -0,7900 0,4290 1,2067 0,2600 0,7970 _cons -0,2562 -4,1400 0,0000 0,0472 0,3700 0,7090 N 169 N 163 F(6, 156) 20,35 F(6, 156) 18,85 Prob > F 0 Prob > F 0 R2 0,4297 R2 0,4203 Adj R2 0,4086 Adj R2 0,398
Panel A shows the OLS regression before adoption of the mandatory clawback provision. Panel B shows the OLS regression after adoption of the mandatory clawback provision.
The results of the tests show a small decrease in the coefficient of EARNINGSt on
further tests are needed to see if the difference between the coefficients is significant as well. To be able to test this, first I had to make estimation results of both regressions using the
estimate and store command in stata. Afterward, I used the command suest, which combines
the estimation results into one parameter vector and simultaneous variance matrix of the sandwich type. At last I carried out a difference of mean t-test, to find whether the coefficients found in both samples were significantly different. The results showed a chi2 (1) of 0.45 with a Prob > chi2 = 0.5035, which can be interpreted as not significant (UCLA: Statistical Consulting Group, n.d.). As the coefficients are not significantly different from each other, I have to conclude that my first Hypothesis is not supported, meaning that earnings persistence was not statistically enhanced after adoption of the mandatory clawback provisions.
4.2.2 Meet/Beat behavior
Previous research (Kent and Routledge, 2017; Degeorge et al., 1999) shows three different ways to test whether upper management performs meet/beat behavior. Two out of the three ways are tested in this research. Table 5 shows the result of meet/beat behavior towards the zero earnings change benchmark and Table 6 shows the result of meet/beat behavior towards a zero earnings benchmark. Both ways are tested performing a logit regression. Logit regressions are used when there is a dichotomous dependent variable, which applies to this research as both MBCHANGE and MBZERO are defined as a dummy variable, hence binary. Both hypotheses 2a and 2b would expect a negative correlation between CB and the dependent variable, as we expect that mandatory clawback adoption will reduce the incentive of performing earnings management to meet or beat benchmarks.
The results of Hypothesis 2a are shown in Table 5. The results show that mandatory clawback adoption (CB) is slightly negatively correlated (-0.1494) with MBCHANGE. However, this result is not significant, as the P>z value is 0.812. These values show that hypotheses 2a is not supported and the adoption of mandatory clawback provisions did not cause a significant decrease in earnings management to meet or just beat the positive earnings change threshold.
TABLE 5
Logistic Regression zero earnings change benchmark
Variable MBCHANGE Coef. Std. Err. z P>z CB -0,1494 0,6276 -0,2400 0,8120 LEVERAGE -0,2843 0,2873 -0,9900 0,3220 SIZE -0,2395 0,1006 -2,3800 0,0170 GROWTH -0,1380 0,5931 -0,2300 0,8160 BTM 0,1798 0,4248 0,4200 0,6720 GDP 5,0149 25,6833 0,2000 0,8450 _cons 0,9504 1,3997 0,6800 0,4970
Number of obs 332 Prob > chi2 0,2818
LR chi2(6) 7,44 Pseudo R2 0,046
The test of Hypothesis 2b is shown in Table 6. The results indicate that mandatory clawback adoption (CB) is positively correlated (0.516) with MBZERO, which contradicts the expectation of Hypothesis 2b. However, this result is not significant (P>z= 0.617). These values show that hypotheses 2b is not supported. To conclude, the adoption of mandatory clawback provisions did not cause a significant decrease in earnings management to meet or just beat the positive earnings threshold either, so both sub-hypotheses are not supported.
TABLE 6
Logistic regression zero benchmark
Variable MBZERO Coef. Std. Err. z P>z CB 0,516 1,031 0,500 0,617 LEVERAGE -0,016 0,124 -0,130 0,896 SIZE -0,169 0,144 -1,180 0,239 GROWTH 0,187 0,786 0,240 8,120 BTM 0,037 0,606 0,060 0,951 GDP -23,069 40,573 -0,570 0,570 _cons -1,359 1,854 -0,730 0,464
Number of obs 332 Prob > chi2 0,2818
Due to the small sample (332) observations, it is hard to determine that the value observed is significant. Only 25 observations where marked as MBCHANGE=1 (0 ≤ change in EPS<0,05), and only ten observations were marked as MBZERO=1 (0 ≤ EPS<0,05). The small sample is explained by the amount of publicly held firms and limited available observation years. Because this is a rather new law, I could only research three years after the implementation.
4.3 Sensitivity Analysis
In the previous section, the relation between mandatory clawback adoption and earnings persistence is discussed, as well as the relationship between mandatory clawback adoption and meet/beat behavior. In this section, a robust regression will be performed following the method of Li (1985) for the earnings persistence analysis. Leone et al. (2014) recommend the use of robust regressions, to mitigate the effect of extreme data values. For hypotheses 2a and 2b, a sensitivity test will be performed, by adjusting the scope of the variables MBCHANGE and MBZERO to the scope used by previous research.
4.3.1 Robust regression Earnings Persistence
The robust regression is the one as explained by Li (1985), which limits the influence of outliers with a disproportionate impact on the result of an OLS analysis. Leone et al. (2014) find that robust regressions are more effective in mitigating the influence of influential observations than winsorization or truncation. In Table 2 Panel A, we observe a substantial negative outlier (minimum is -2.667 while the median and maximum values are 0.33 and 0.534 respectively) for variable EARNINGSt+1, and in Table 2 Panel B, both EARNINGSt+1
and EARNINGSt have large negative minimum values compared to their median and
maximum. These disproportionate outliers will be treated differently within this regression. Table 7 shows the results of the robustness regressions performed. Panel A indicates that the correlation coefficient of EARNINGSt on EARNINGSt+1 before the adoption of
mandatory clawback provisions was 0.3986 with a high significance level (1%), while Panel B shows a coefficient of 0.4513 (highly significant; 1% level). These values show that the correlation coefficient increased 0.0527 points. The increase would be consistent with Hypothesis 1, although the significance of the difference between both coefficients of
EARNINGSt in Panel A and B still needs to be tested. Observing the 95% confidence interval
difference would not be significant, yet to be sure the difference of mean t-test should be performed. Unfortunately, there is no possibility to run a suest procedure and perform the
difference of mean t-test on robust regressions. Hence I cannot make definitive conclusions
from this robustness test.
TABLE 7
Robust regression Earnings Persistence Panel A Variable EARNINGSt+1 Coef. Std. Err. t P>t [95% Conf. Interval] EARNINGSt 0,3986 0,0298 13,3600 0,0000 0,3397 0,4576 SIZE 0,0023 0,0022 1,0600 0,2890 -0,0020 0,0066 GROWTH -0,0067 0,0136 -0,4900 0,6220 -0,0337 0,0202 BTM -0,0021 0,0148 -0,1400 0,8870 -0,0313 0,0270 LEVERAGE 0,0349 0,0041 8,4800 0,0000 0,0267 0,0430 BBP -0,1232 0,4635 -0,2700 0,7910 -1,0385 0,7922 _cons -0,0712 0,0309 -2,3000 0,0220 -0,1322 -0,0102 N 169 Prob > F 0 F(6, 161) 20,35
Panel A shows the robust regression before adoption of the mandatory clawback provision.
Panel B Variable EARNINGSt+1 Coef. Std. Err. t P>t [95% Conf. Interval] EARNINGSt 0,4513 0,0171 26,3900 0,0000 0,4175 0,4851 SIZE 0,0023 0,0017 1,3400 0,1840 -0,0011 0,0058 GROWTH 0,0221 0,0161 1,3800 0,1710 -0,0096 0,0539 BTM 0,0096 0,0066 1,4600 0,1470 -0,0034 0,0227 LEVERAGE 0,0042 0,0017 2,4300 0,0160 0,0008 0,0077 BBP -1,2107 1,1308 -1,0700 0,2860 -3,4446 1,0232 _cons 0,0102 0,0307 0,3300 0,7390 -0,0503 0,0708 N 161 Prob > F 0 F(6, 161) 141,01
4.3.2 Sensitivity Analysis Meet/beat behavior
Within prior research, the scope of small positive earnings or small earnings change has been mostly defined as an EPS (change) between the 0 and the 0.02 (Degeorge et al, 1999; Kent and Routledge, 2017, Dehaan et al., 2013). Within this scope, only four MBCHANGE and four MBZERO observations would have made, which is approximately 1,2% of the observations. Hence I’ve chosen to broaden the scope in the multivariate analysis to an EPS (change) between the 0 and the 0.05 and show the results of the more narrow scope to perform the sensitivity analysis. For the analysis both dependent variables will be tested with a logit regression.
Table 8 shows the results of the sensitivity analysis of Hypothesis 2a: the effect of mandatory clawback adoption on the incentive to meet or beat the positive earnings change benchmark. This analysis shows that mandatory clawback adoption (CB) is largely negatively correlated (-1.0077) with MBCHANGE, which would support Hypothesis 2a. However, this result is not significant, as the P>z value is 0.21. The fact that the results are not significant in the sensitivity analysis shows a consistent finding that does not support Hypothesis 2a.
TABLE 8
Sensitivity analysis zero earnings change benchmark
Variable MBCHANGE Coef. Std. Err. z P>z [95% Conf. Interval] CB -1.0077 0.8047 -1.2500 0.2100 -2.5849 0.5695 LEVERAGE -0.3514 0.4249 -0.8300 0.4080 -1.1841 0.4813 SIZE -0.3010 0.1389 -2.1700 0.0300 -0.5733 -0.0288 GROWTH -0.0286 0.6735 -0.0400 0.9660 -1.3487 1.2915 BTM 0.0819 0.5775 0.1400 0.8870 -1.0500 1.2137 GDP 25.8341 31.6394 0.8200 0.4140 -36.1780 87.8463 _cons 1.2841 1.8992 0.6800 0.4990 -2.4383 5.0065 Number of
obs 332 Prob > chi2 0.2818
Table 9 shows the results of the sensitivity analysis of Hypothesis 2b: the effect of mandatory clawback adoption on the incentive to meet or beat the positive earnings benchmark. The results show that mandatory clawback adoption (CB) is largely negatively correlated (-1.6507) with MBZERO, which intuitively would support Hypothesis 2b. However this result is not significant, as the P>z value is 0.347. The fact that the results are not significant in the sensitivity analysis shows a consistent finding that does not support Hypothesis 2b.
TABLE 9
Sensitivity analysis zero earnings benchmark
Variable MBZERO
Coef. Std. Err. z P>z Conf. Interval] [95%
CB -1.6507 1.7549 -0.9400 0.3470 -5.0903 1.7890 LEVERAGE 0.0843 0.1632 0.5200 0.6060 -0.2357 0.4043 SIZE -0.1000 0.2133 -0.4700 0.6390 -0.5180 0.3181 GROWTH 0.3631 1.0231 0.3500 0.7230 -1.6421 2.3683 BTM -1.2092 0.8774 -1.3800 0.1680 -2.9289 0.5105 GDP 7.8622 53.0993 0.1500 0.8820 -96.2106 111.9350 _cons -3.0934 2.8059 -1.1000 0.2700 -8.5928 2.4060 Number of
obs 332 Prob > chi2 0.7104
LR chi2(6) 3.75 Pseudo R2 0.0866
To conclude, even though the coefficients found in the sensitivity analysis are in line with hypotheses 2a and 2b, these are not found to be significant. This means that no significant difference between the period before and after mandatory clawback adoption is found. Hence Hypothesis 2a and 2b are not supported in this sample either. This shows a consistent outcome of the findings after sensitivity analysis.
Summary and Conclusion
5
This thesis tries to answer the question: Does the adoption of the law on clawback affect earnings quality? As such, the purpose of this research is to see whether the effect of clawback provisions when the adoption was voluntary persists when the adoption op the provision is mandatory, especially on earnings quality. Chan et al. (2012) focused on the effect of the adoption of clawback provisions on earnings quality using financial report restatements and the Earnings Responsiveness Coefficient as proxies. Dehaan et al., (2013) focused on the effect of the adoption of clawback provisions on financial reporting quality and used meet/beat behavior as one of the proxies. Earnings quality is described as information on earnings that is both relevant and has a faithful representation. To prove both requirements of earnings quality, I chose to use the proxy earnings persistence, which tests the usefulness of earnings information, and the proxy meet/beat behavior by upper management, which examines the faithful representation of earnings quality. Both tests are found to be proxies of earnings quality in the paper by Dechow et al. (2010).
The sample of this research contains publicly listed Dutch firms, who are expected to adopt a clawback provision by enforcement of the Dutch law, since 2014. Before the ignition of the law on clawback, clawback provisions were already encouraged in the Netherlands. Clawback provisions were incorporated in the Dutch corporate governance code, which has an apply or explain principle. The Dutch legislators found the existence of the clawback provision in the corporate governance code insufficient and decided to adopt it within the common law. The effect of the implementation of this law is interesting; hence the Unites States are planning to execute the Dodd-Frank act soon. Which will force all US listed firms to adopt a clawback provision, hence will affect a lot of companies.
To test earnings persistence the most commonly used model in research is tested; where next period earnings are described to be dependent on current earnings and control variables. The sample was divided into two periods to analyze the results: ex-ante and ex-post the mandatory adoption of clawback provisions. Both samples were subject to an OLS regression. The difference in correlation coefficient of next period earnings and current earnings was tested on whether they were significantly different, with the expectation of a significant increase in coefficient between the two periods. The main results of this analysis showed a decrease in the coefficient between the two periods, which contradicts the expectations. These coefficients were not found to be significantly different. Hence the
conclusion that mandatory clawback would lead to more persistent earnings and thus more relevant earnings information is rejected.
To test meet/beat behavior performed by upper management two thresholds were tested. First the positive earnings change benchmark is tested, described as the unwillingness of management to show a decrease in earnings compared to the prior period. To mitigate this, they perform earnings management to meet or just beat the benchmark. Second, the positive earnings benchmark is tested, where managers are resistant to show negative earnings, as these have a significant impact on stakeholders, and so they engage in earnings management to avoid it (Degeorge et al., 1999). Both tests are performed using a model that observes the relation between meet/beat behavior as the dependent variable and mandatory clawback adoption as the independent variable supported by control variable. Because both dependent variables are dichotomous, the logit regression is performed in both tests. The expectations are that both tests would show a significant negative coefficient between meet/beat behavior and the adoption of mandatory clawback provisions. The main results of these analyses do not show significant coefficients supporting the expectations. Concluding that the results do not find support for the adoption of the law on clawback reducing meet/beat behavior by upper management, and thus would lead to a more faithful representation.
To conclude this research finds no evidence of the adoption of the law on clawback affecting earnings quality. This could be explained by the fact that in the period before adoption, firms already adopted clawback provisions through the Dutch corporate governance code. This implies that the new legislation does not have the intended effect of improving earnings quality.
The biggest shortcoming of this research would be the sample size. Sixty-six individual firms were identified and six observation years were analyzed. The tests of meet/beat behavior contained respectively 25 and ten absolute values that showed meet/beat behavior. With this small amount of observations, it is harder to find significant values than, for example, a sample with ten times as much observations, as usually seen in academic research.
My recommendation for further research would be to look at the effect of the adoption of the law on clawback on other accounting matters. Such as earnings management, using other proxies as the relation between earnings management through accruals or real earnings management, like research by Chan et al. (2015) on voluntary clawback provisions.
