Performance target choice and
writing styles of compensation reports
‘An exploratory study in the UK’
ABSTRACT!
In an attempt to make remuneration policies transparent for shareholders, the 2002 DRR Regulation was introduced. The regulation requires firms to disclose their remuneration policy including a narrative on performance target setting and although it instructs on what information to disclose, it lacks to instruct on how to disclose it. Motivated by this ‘presentation gap’ and evidence from prior studies that found executives tend to behave opportunistically driven by self-interest, this paper examines the relationship between performance target choice and writing styles of compensation reports. After reviewing literature I argue that, on the one hand, organizations might use remuneration reports as a mechanism to signal prospects, credibility and legitimacy and as an opportunity to overcome information asymmetries. On the other hand I argue that organizations might intentionally employ biased writing styles as a way to impress shareholders and camouflage adverse compensation information and concurrent rent extraction. I find evidence for both rationales. Overall, the results suggest that the 2002 DRR regulation did not per se result in more effective communication to investors, but left considerable room for opportunistic behavior and impression management. Additional tests showed that this room is particularly exploited by self-interested top managers that extract rents, by highly powerful CEOs and finally by executives employed in poor governed firms.
Key words: Corporate Governance, DRR Regulation, Performance Target Choice, Writing styles.
Supervisor: Dr. B. Qin
1. INTRODUCTION
Corporate governance in general and executive compensation in particular has attracted a lot of media attention and has been the subject of considerable academic controversy and debate. The public outrage tends to focus on the (rising) level of compensation, observed again in the outrage over executive compensation paid at many financial organizations that were hit hardest by the credit crisis (Conyon and Sadler, 2010). However one cannot infer unadulterated conclusions about the (rising) level of compensation, without looking (in depth) into the specifics of the compensation contract; the design of performance targets. As a reaction to the constant debate and in an attempt to improve (the lack of) transparency in compensation contracts several corporate governance codes have been installed in the UK. The introduction of the DRR regulation in 20021 resulted in the mandatory yearly publication of significant executive compensation disclosure reports. The report has to include a statement of the firm’s remuneration policy including a detailed narrative, for every executive director, on the performance conditions to which any entitlement to options or shares is subject. Although the 2002 DRR regulation dictates the precise content of the remuneration report, the DRR regulation does not, like in the US, contain a ‘plain English’ clause to instruct on the method of disclosure. In the US the ‘plain English’ clause was introduced because “(1) firms could use vague language and format in disclosure to hide adverse information, and (2) the average investors may not understand complex documents and this could result in capital market inefficiency” (Li, 2005: 2). Since a similar clause is missing in the UK, the regulation left room for opportunistic behavior through impression management versus only the provision of useful incremental information. This will be the first study to explore whether performance target choices set for executive compensation in any way influence the way DRRs are crafted.
Previous studies on writing styles and impression management found that executives might use bias in their choice of language and verbal tone as a tool to ‘fool’ investors and ‘manage’ their impressions (e.g. Courtis, 1998; Sydserff and Weetman, 1999; Lang and Lundholm, 2000; Libby, Bloomfield and Nelson, 2002; Bloomfield, 2002; Tetlock, 2007; Sadique, In and Veeraraghavan, 2008; Tan, Wang and Zhou, 2011; Loughran and McDonald, 2011a). The clarity or opaqueness as well as the optimism or pessimism in communication has a big impact on investors’ ability to assess performance target choice. The introduction of mandatory disclosures might therefore not automatically have resulted in improved
1
shareholder transparency. Bettis et al. (2010) found that the nature of performance target provisions, as well as the level of detail provided in the remuneration reports varies considerably across firms. But what varies most substantially is the terminology and tone used to describe the performance target choice and remuneration contract features. Given the widespread take up of performance target provisions and the ongoing interest in reforming remuneration and concurrent disclosure (Bettis et al., 2010), I believe that it would be interesting to find out whether differences in performance target choice are systematically related to differences in writing styles of remuneration reports. More specifically, I ask the following question: Is there a relation between firms’ performance target choice and writing
styles of compensation reports? Does the DRR create valuable incremental information for
shareholder decision-making and stimulate confidence though legitimacy and credibility or is the DRR ‘presentation gap’ indeed exploited and is writing style used for opportunism (i.e. as a form of impression management to camouflage suboptimal target design and rent extraction)? This report attempts to provide the first empirical evidence on this relation.
Since the DRR regulation was operative for firms with fiscal years ending on or after December 31, 2002, I examine the writing styles of British DRRs in the period 2002 to 2009. I consider writing styles in the context of readability and linguistic tone. Using data on 1837 company-year observations from 258 British listed firms (both FTSE 350 and SmallCap), I use regression analysis to examine if performance target choice impacts DRR writing styles. Controlling for several firm- and governance characteristics as well as industry and year fixed effects I find evidence that the remuneration committee can use the writing style of compensation disclosures in two opposing ways. On the one hand the results suggest that disclosure transparency improved (at least partially) under public pressure and the regulatory oversight of the 2002 DRR regulation. The reports are used as a mechanism to signal prospects, credibility and legitimacy and as an opportunity to overcome information asymmetries and rationalize target choice. However, on the other hand the regulation also left considerable room for impression management that is exploited by executives as a way to reach self-interested goals. This room is particularly exploited by self-serving top managers that extract rents, by CEO’s with high managerial power and by executives employed in poor governed firms. Overall I infer from the results that the regulation merely partially succeeded in its goal to make remuneration policies transparent for shareholders and firms do not per se provide effective incremental information.
and structure of compensation and the rate of increase without looking (deeper) into the specifics of the compensation contract such as the design of performance targets. This occurs mainly due to the difficulty of collecting data in this respect; however the findings on executive pay may be misleading without taking into account other important elements in a compensation package. For instance, executive compensation may be high due to the additional risk associated with the difficulty of achieving the target(s). So this paper aims to explore and address the compensation contract more broadly.
Secondly, this paper adds to the nascent stream of research using textual analysis of various types of corporate documents such as environmental disclosures (Cho and Patten, 2010), financial statements (Tan et al., 2011; Loughran and McDonald, 2011a), annual reports or 10-K forms (Lehavy, Lee and Merkly, 2011; Courtis, 1995; Courtis and Hassan, 2002; Li, 2008, You and Zhang, 2009; Loughran and McDonald, 2011b), letters to shareholders (Hooghiemstra, 2010) and earnings announcements (Sadique et al., 2008). However the readability of the compensation report is a topic that “has only recently come to the attention of the investing public” (Laksmana, Tietz and Yang, forthcoming in 2012: 5). After reviewing the literature, I am aware of only one working paper on compensation report writing styles in the UK and say-on-pay votes (Hooghiemstra, Kuang and Qin, 2012) and one study examining readability of remuneration reports in the US (Laksmana et al., forthcoming in 2012). But even for this last paper the authors limit their data on the level of executive compensation without looking (in depth) into the composition or structure of the compensation package.
Thirdly, previous empirical linguistic style studies mostly focus on either readability or tone and examine the association with other variables, including future earnings (Li, 2008; Li, 2010b), stock prices (Lee, 2010; Lawrence, 2011; Davis, Piger and Sedor, 2008; You and Zhang, 2009), analyst behaviour (Lehavy et al., 2011; Kothari, Li and Short, 2011; Kravet and Muslu, 2010), information environment (Kothari et al., 2011; Muslu et al., 2010), cost of capital (Kothari et al., 2011) and shareholder litigation (Nelson and Pritchard, 2007; Rogers, Van Buskirk and Zechman, 2011). However, I am not aware of any prior research on the relation between performance target choice and textual analysis, both readability and tone, of remuneration reports. My study aims to address this void and extend the literature. As Caskey (2009) discusses, while information intermediaries (the remuneration committee) may not be able to control the degree of ambiguity in the underlying information, they can control how they present this information.
Merkl-Davies and Brennan (2007) who state future research needs to incorporate both the possibilities of incremental information and impression management into the research design. These perspectives lead to competing hypotheses; however ignoring them would limit my thinking process.
Lastly, the paper of Li (2010a) discusses some challenges for literature concerning textual analysis of corporate disclosures and proposes some opportunities for future research among which one avenue is corporate governance and the informativeness of textual disclosures. There is a strong link between the particular topic of executive pay and performance target setting and the wider literature on governance, monitoring and entrepreneurship. For instance, performance conditions in equity incentive plans may improve incentive contracting and thereby corporate governance. Bettis et al. (2010) studied stock and option grants with performance-based vesting provisions and state that when these awards are used to improve managerial incentives, their use is associated with ‘stronger’ corporate governance. In the current study, I control for the effect of several corporate governance characteristics and I particularly consider the interacting effect of corporate governance in the context of managerial power and board structure. This study thus answers Li (2010a)’s call for future research.
2. INSTITUTIONAL BACKGROUND
2.1 UK Corporate Governance Codes
Executive compensation and its effect on firm performance has been the constant target for government regulation, capital market’s self-regulation, the accountancy profession, institutional investors’ associations and media criticism (Buck et al., 2003). As a reaction to large scale corporate scandals and the constant scholarly and popular controversy and debate several corporate governance codes have been installed in the UK (e.g. the Cadbury Report, 1992; the Greenbury Report, 1995). They aim to (re)direct firms’ focus on the three pillars of accountability, performance linkage and transparency (Al-Issa, 2009; Conyon and Sadler, 2010). In what follows the paper will discuss the two pillars ! performance linkage and shareholder transparency ! where corporate governance codes have been most influential. 2.2 Pillar I: Performance Linkage
2.2.1 UK Performance-Based Compensation
Appropriate incentives for executives are vital to firm performance (Buck et al., 2003). “Two issues that have received considerable attention in the accounting and economics literatures are the use of stock option grants and choice of performance measures in executive compensation contracts” (Gerakos, Ittner and Larcker, 2007:1). By ‘paying-for-performance’ or ‘optimal contracting’, a proportion of executive compensation is linked to equity and compensation becomes more directly and tightly linked to shareholder wealth (Holmstrom and Milgrom, 1987; Jensen and Murphy, 1990). Historically, traditional stock options, known as Executive Share Options (ESOs) and restricted stock awards based on simple time-vesting provisions, accounted for the majority of performance-based pay. However, both investors and academics have expressed concern that these provisions do not create sufficiently strong financial incentives and have referred to them as ‘pay for pulse’ (Bebchuk and Fried, 2004).
a group of comparator companies so as to prevent making rewards that are not contributable to executives’ efforts. The recommendations of the Greenbury Committee, together with other institutional pressure (e.g. from the Association of British Insurers), “provided considerable impetus for the widespread, voluntary adoption of performance conditions in equity incentive plans for UK executives” (Carter et al., 2009: 272).
Thus in the last two decades, we have witnessed the emergence of more novel executive compensation instruments. Greenbury urged firms to replace traditional stock options (ESOs) with conditional stock options such as Performance-Vested Stock Options (PVSOs) and / or Long-Term Incentive Plans (LTIPs). These instruments provoked particular interest with both investors and academics in their ability to reinforce corporate governance robustness. The purpose of these new instruments is to create stronger incentives for executives to improve stock market and accounting performance, while at the same time incentivising executives to make riskier corporate policy choices (Bettis et al., 2010). Traditionally, the interests of shareholders and managers in terms of risk appetite are not aligned. While investors can relatively easily tolerate risk by diversifying their portfolio, managers generally do not have these options / benefits available to them and might be hurt by risk. Managers therefore usually forego the more risky but potentially more value increasing financial and investment policy projects that can increase firm performance. Shareholders can incentivise risk-averse managers to make riskier choices with the use of performance-vesting provisions. “Performance-vesting provisions can do so by increasing the sensitivity of managerial wealth to firm performance (delta) and to stock price volatility (vega) (Johnson and Tian 2000; Brisley 2006; Kuang and Suijs 2006) beyond the delta and vega provided by the underlying options or stock” (Bettis et al. 2010: 3851). Bettis et al. (2010) find that firms using performance-vesting provisions to reward their executive have significantly stronger operating performance outcomes over the five years following the grant compared to the firms that do not. They conclude that performance-vesting provisions provide significant incentives for executives and create stronger governance.
2.2.2 UK Performance-Vested Equity Contracts
to stock market performance, accounting results, other performance targets or non-financial targets. The two most frequently used targets are targets based on EPS (earnings per share) growth in excess of inflation and / or performance relative to a peer group such as relative TSR (total shareholder return). A peer (or ‘comparator’) group is usually either a self-defined group or a pre-defined market-index group such as FTSE350 (Carter et al., 2009). Vesting can follow different ‘patterns’ such as lump-sum (hurdle) vesting, linear (pro-rata) vesting, step vesting or unlimited vesting (Carter et al., 2009). Most firms use the same performance targets for all executives and typically use a three-year performance period, although some compensation contracts contain provisions that allow re-testing in case targets are not achieved within the set (three-year) performance period (Carter et al., 2009).
Carter et al. (2009) find that although remuneration contracts are relatively stable within firms (from year-to-year), they differ considerably between firms. The authors state that “considerable differences exist in the minimum percentile ranking needed for vesting to begin, the use of performance hurdles versus performance payout ranges, the spread between the performance ranking needed for minimum and maximum vesting, and the extent to which the grants actually vest” (Carter et al., 2009: 270).
2.3 Pillar II: Transparency
2.3.1 The Disclosure Requirements of the 2002 DRR Regulation
Besides the influence on performance linkage the UK corporate governance codes have also been influential in improving shareholder transparency. In 1995 the Greenbury Report included regulatory initiatives aimed at expanding disclosure rules for executive compensation (Al-Issa, 2009) and in 2002 the Directors’ Remuneration Report (DRR)
Regulation highlighted the need for shareholder transparency again. The DRR regulation
needs to contain a narrative on the performance conditions employed for the granting and / or vesting of bonuses, share options and long-term incentive schemes with an explanation as to why those targets were chosen (DRR, 2002; Ferri and Maber, 2009). If the firm employs relative targets, the peer group selection needs to be justified. In addition, if executive directors are entitled to shares or options not subject to performance targets, a detailed narrative has to be given as well to justify why that is the case. Finally, the firm needs to state to what extent the performance targets were actually attained.
The DRR created a ‘voice’ for shareholders by means of their vote to approve the report at the annual meeting. This ‘voice’ is limited because the DRR regulation states that “No entitlement of a person to remuneration is made conditional on the resolution being passed (The Companies Act 2006, Section 439)” (Conyon and Sadler, 2010: 298) and as such the resolution is merely advisory (non-binding). A failure to approve the report doesn’t result in any sanction and has no effect on the underlying remuneration policy. Nevertheless a negative vote still ‘voices’ shareholders lack of confidence (Conyon and Sadler, 2010) in the proposed policies and remuneration committee itself, which could significantly hurt share price and thereby impact firm performance.
2.3.2 The ‘presentation gap’ of the 2002 DRR Regulation
2.3.3 The Communication Effectiveness of the 2002 DRR Regulation
The Department of Trade and Industry (DTI) claimed that the DRR regulation was designed to help enhance transparency in setting directors remuneration, improve accountability to shareholders and provide a more effective link to firm performance. After its introduction in 2002 the DTI asked Deloitte to perform research on the actual effectiveness of the DRR regulation and this resulted in some interesting findings. The study showed that even though nearly all firms comply with the regulation and disclose what the regulation requires, a significant portion of companies merely comply and do not go beyond the basic (i.e. minimum) requirements in order to provide shareholders with a clear insight into their policies and practices. An analysis on communication effectiveness revealed that only 2% of the FTSE 250 firms have an ‘excellent’ rating for communication effectiveness, meaning that firms “fully comply with the regulations, provide a detailed, understandable description of their remuneration policy and use tables and charts to illustrate the details” (Deloitte, 2004: 10). The study reports several remarkable figures concerning communication effectiveness2. Firstly, only 4% of shareholders considered the communication of remuneration strategy, philosophy and practice to be very effective (Deloitte, 2004). Secondly, 63% of shareholders considered the information provided to make an assessment of the appropriateness of the remuneration plan (very) insufficient (Deloitte, 2004). Lastly; 81% of shareholder considered the information provided to explain the relationship between awards actually earned and the performance achieved in the period for which they have been earned (very) insufficient (Deloitte, 2004). After combining the questionnaire responses with subsequent shareholder discussions, Deloitte concluded that shareholders do not want or need more (incremental) information, but they apparently do need better (more informative) communication (Deloitte, 2004).
2
3. LITERATURE REVIEW AND HYPOTHESIS DEVELOPMENT
In an agency setting, the authority relationship between shareholders and executives has always been complicated. The agency theory offers a useful way of understanding the complex authority relationship between principals and agents, who generally have very different interests, risk appetite and access to information. As soon as shareholders decide to hire executives and delegate decision-making authority and responsibility for value creation to them, accountability problems arise. The performance of executives is often difficult to evaluate since it can be measured only after a considerable elapse of time or at a very high cost and executives might exploit this information advantage. No wonder why firms spend considerable time and resources to find ways to overcome or at least reduce agency problems. The two most frequently used instruments are 1) creating a proper set of (monetary) incentives to align the interests of shareholders and executives and 2) improving monitoring effectiveness to reduce information asymmetry. Performance target choice is an example of the former and transparency in corporate disclosures is an example of the latter and these two developments, instruments might be related. It is known that remuneration reports have to disclose the performance target choice, but there are no instructions on the method of disclosure. This raises the question if performance target choices set influence the way DRRs are crafted. In what follows, I develop my hypotheses regarding the relationship between performance target choice and DRR disclosures.
3.1 Performance Target Choice.
However, contrary to expectations based on Locke’s goal setting theory (1968) the study of Merchant and Manzoni (1989) found that, in practice, it is not so straightforward and most budget targets are set at highly achievable levels, (achievable 80 to 90% of the time). Based on their field study they conclude that high achievability creates many advantages for the firm including improved resource consumption and planning, reduced risk of non-attainment (which reduces the need to compensate managers for this risk), improved control, improved corporate reporting (no incentive for harmful earnings management practices) and improved motivation by ensuring a competitive compensation package. Conyon and Murphy (2000) question whether the high achievability of performance targets in practice is the result of their vulnerability to executive manipulation. Finally Chow (1983) and Mowen, Middlemist and Luther (1981) suggest that the impact of goal achievability on firm performance is dependent on the financial incentives associated with the performance, for instance they report a negative relation under a piece-rate incentive plan but they do not find this same negative relation under a hurdle bonus plan.
Locke et al. (1981) also provide several conditions in which goal setting is most likely to increase performance; firstly goals should be specific and sufficiently challenging; secondly managers should be able to reach the goals; thirdly managers have to accept the goals; fourthly regular feedback has to be provided to show progress in relation to the goal and finally (monetary) rewards should be given for goal attainment. However, the optimal amount of challenge in a target is hard to decide upon. The theory regarding the effects of performance target choice on motivation is complex. If you set target difficulty too low, people will not be motivated to produce superior results because they are able to achieve the targets with minimal effort. Vice versa, if you set target difficulty too high, people will be discouraged and loose their commitment and motivation to achieve the target. The optimal amount of challenge should be set at the point which is perceived as the limit of executives’ ability, in the performance target literature known as targets that are ‘challenging but
achievable’ (Merchant and Van der Stede, 2007; Erez and Zidon, 1984).
FIGURE 1
Relationship between Performance Target Difficulty and Motivation / Performance.
(Source: Merchant, K. 1998, P 388)
Besides the abovementioned abundant research on goal setting in general, recently and in line with the trend in ‘best practices compensation packages’, the academic community has shown interest in research on performance-vesting provisions in equity contracts. The results of studies on performance targets with vesting provisions are mixed. Buck et al. (2003) studied the relation between LTIPs and firm performance in the UK and found that the use of such plans results in lower pay-performance sensitivity. The authors explain this lower sensitivity arises in the context of managerial opportunism in the design of LTIPs at the expense of shareholders. In contrast, theoretical studies claim that performance-vesting provisions for option schemes result in higher pay-performance sensitivity (Johnson and Tian, 2000; Kuang and Suijs, 2006). Empirically Bettis et al. (2010) studied stock and option grants with performance-based vesting provisions in the US and found that such provisions are associated with stronger governance and provide significant incentives for executives. The empirical study of Kuang and Qin also found that “the use of PVSO schemes in executive compensation contracts is associated with greater interest alignment and PVSO schemes outperform traditional schemes in providing incentives” (2009: 46). However, these authors also found, consistent with prior studies on the target difficulty-managerial effort interaction, that increased difficulty in vesting targets is associated with lower interest alignment (Kuang and Qin, 2009). Difficult targets are defined as those that are beyond executives’ ability, point A in Figure 1 (Locke and Latham, 2002).
but achievable performance-vesting targets promote firm performance and shareholder value and are designed in shareholders’ best interests.
3.2 The Implications of Performance Target Choice.
3.2.1 DRR Readability.
writing styles that make information look more (or less) favourable. The obfuscation hypothesis asserts that messengers may purposely try to hide negative info behind complex wording, while positive info may be highlighted through easy wording and readily understandable language (Courtis, 1998). Courtis defined obfuscation as “a narrative writing technique that obscures the intended message, or confuses, distracts or perplexes readers, leaving them bewildered or muddled” (Courtis, 2004: 292). The managerial power perspective of agency theory (Bebchuk and Fried, 2003) and the ‘self-serving management’ hypothesis (Benston, 1985) emphasize the significant discretionary power senior executives possess, which they can exploit to place their own private benefits before those of the shareholders. The theory asserts that executives have potency to bargain and influence their own remuneration contracts with the board of directors (e.g. Core, Holthausen and Larcker, 1999; Bebchuk, Cremers and Peyer, 2009; Bushman and Smith, 2001; Bebchuk, Fried and Walker, 2002; Murphy, 2002). Previous studies examining proxy disclosures found evidence that demonstrates executives tend to behave opportunistically driven by self-interest (Murphy, 1996; Yermack, 1998; Baker, 1999; Lewellen, Park and Ro, 1996). Linking this evidence to the knowledge of managerial risk aversion (Benson and Davidson, 2009; Janakiraman, Radhakrishnan and Tsang, 2010), executives might use their power to avoid or reduce the portion of equity-linked compensation and strive for easy performance targets rather than difficult ones (Oyer and Schaefer, 2005). Equity-linked compensation and difficult targets are not only (more) volatile but also much riskier, executives lose for instance unvested equity-linked compensation when they decide to leave the company and would thus not prefer this type of compensation. More importantly (vesting) target difficulty will largely determine the
ex post realization of performance-based equity pay. Vesting targets clearly impose additional
risk on executives and therefore they might try to avoid such pay.Stemming from the agency theory in general and managerial power in particular, the compensation contract is thus clearly influenced by executives, and DRR readability might be a means to reach self-interested goals (avoid uncertainty and extract rents). Taking together all the abovementioned theoretical perspectives and the assumption that executives are risk averse and easy targets are suboptimal for shareholders (Section 3.1), remuneration committees may opportunistically choose low readability (i.e. complex wording) of remuneration reports to ‘camouflage' easy target information from investors. The following hypothesis was developed:
Following the obfuscation reasoning, adverse target information will thus be shielded from investors through low readability and complex wording. While many studies have been supportive of this theory, there is a significant body of research that is in disagreement. The disclosure credibility view asserts that presenting clear, adequate and understandable disclosures might be beneficial for the company. Several studies (Tan et al., 2011; Jennings, 1987; Mercer, 2004) demonstrate that low readability will impair disclosure credibility. If investors do not believe the message disclosed by management, they will be reluctant to trade on or react to the disclosed information both now and in the future. “Compared to a costly information processing argument, a deterioration in investors’ perceptions of disclosure credibility in response to low readability may have more severe longer-term implications for management” (Tan et al., 2011: 5). More importantly, these authors conclude that if executives are aware of the credibility-impairing role of low readability, they may be less motivated to use difficult-to-read disclosures to camouflage information. Complex disclosures and low readability may backfire and have the opposite effect by increasing analysts’ concerns and shareholder pressure. Furthermore empirical studies on linguistic complexity find that information is more persuasive when it is easily readable (MacDonalds-Ross, 1977). A transparent report would thus be the most persuasive to justify performance target choices. If committees are aware of this, then they might want to prevent shareholders getting confused and starting to ask questions. A transparent DRR disclosure is an opportunity to clearly rationalize compensation choices before public pressure and shareholder criticism can arise. Legitimacy theory claims firms improve disclosure readability under public and shareholder pressure and that disclosures are used with the purpose to alter shareholder perceptions of the firms’ legitimacy. A transparent report would be a common interest for executives and the firm in general as it can enhance firm reputation and secure valuable shareholder confidence which in turn can boost share price performance.
means to enhance share price performance. The presentation of challenging performance targets (which are conceptually complex and have significant consequences) might result in reduced readability. The following hypothesis was developed:
Hypothesis 1b: Ceteris paribus, performance target difficulty is inversely related to the readability of Director Remuneration Reports.
3.2.2 DRR Tone
As already mentioned above, successfully communicating information in corporate reports requires adequate and understandable data as well as readily understandable narrative information and there is empirical evidence that shareholder and analyst perceptions are not only influenced by the content of report disclosures, but also by the way the information is presented. Several studies (e.g. Tetlock, 2007; Loughran and McDonald, 2011b) show stock markets react significantly to the sentiments expressed in disclosures and this can be extended to the use of an optimistic/pessimistic tone (Hooghiemstra et al., 2012). Disclosures with a more optimistic tone are associated with increased share price performance (Lang and Lundholm, 2000), increased returns and decreased volatility (Sadique et al., 2008). Davis et al. (2008) also find that employed disclosure tone triggers a significant incremental market response and moreover impacts shareholder credibility and confidence. The use of optimistic tone in DRRs is thus clearly beneficial.
manipulation of qualitative or ‘narrative’ disclosures is less constrained and relatively easy. Accounting principles and independent monitoring are not designed with the purpose to provide credibility to narratives. For instance, the DRR regulation contains no requirements for verification of the report by an independent, external auditor (DRR 2002) and a ‘plain English’ clause is missing. Furthermore, as also already mentioned, previous studies found evidence that shows executives behave opportunistically (Murphy, 1996; Yermack, 1998; Baker, 1999; Lewellen et al., 1996) and optimistic tone decreases volatility (Sadique et al., 2008). The ‘self-serving management’ hypothesis (managerial power theory), linked with the knowledge of managerial risk aversion (Section 3.2.1) can thus also be used to explain biased DRR tone. As explained in the former section under these theoretical perspectives the compensation contract is influenced by executives, and DRR tone might be a means to reach self-interested ends / goals (avoid uncertainty in compensation contracts and extract rents). Considering the importance of compensation issues to executives, they hope that by persuading the remuneration committee to use a positive tone to hide targets that are too easy (adverse information) from investors, they will ‘trigger’ shareholder confidence and get their vote at the annual meeting. Taking together all the abovementioned theoretical perspectives and the assumption that executives are risk averse and easy targets are suboptimal for shareholders (Section 3.1), remuneration committees may opportunistically choose highly (overly) optimistic tone of remuneration reports to ‘camouflage' easy target information from investors. The following hypothesis was developed:
Hypothesis 2a: Ceteris paribus, performance target difficulty is inversely related to optimistic tone of Director Remuneration Reports.
invest in a way to signal; their managerial reputation. The signaling theory predicts that good-type managers, compared to bad-type ones, would make different decisions. Good-type managers would take certain actions that bad-type ones are either unwilling to or cannot afford to take (Jegadeesh, Weinstein and Welsch, 1993; Garfinkel, 1993; Michaely and Shaw, 1994). Good-type managers would for instance, value their reputation as an executive and be reluctant to use impression management because good reputation signals the value of their services to the (executive) labor market. The idea behind signaling theory is especially applicable to target setting because good-type managers are likely to use challenging targets (to signal promising future) which are costly for bad-type managers to mimic as they are less likely to reap the same benefits. More importantly, the signalling theory predicts that in firms with optimal target choice, firms will signal this superiority by greater transparency and probably optimism in their disclosure (Merkl-Davies and Brennan 2007). The idea behind the signaling theory is also that ‘good-governance’ firms use performance-vesting difficulty as a means to align interests, so as to credibly separate themselves from ‘poor-governance’ firms. Furthermore, challenging targets would result in more optimistic tone as they signal committees / executives’ confidence in future firm performance. Performance-vesting difficulty could be used as a vehicle whereby firms with favorable future prospects can signal this and thereby increase shareholder confidence as a means to improve (future) share price performance. Besides, when executives’ interests are aligned by means of a strong set of motivational incentives, this directly benefits them as well.
Besides signalling, more precise tone has other benefits as well and prevents for instance the risk of litigation involved with the use of unusually overly (biased) optimistic tone (Rogers et al., 2011). Less bias in tone results in more precise information for shareholders, and more precise information causes investors to initiate more trades (e.g. Karpoff, 1987; Kim and Verrecchia, 1991) which is beneficial to both executive and the firm. Taking together the signalling theory, the empirical evidence of (un)biased disclosure tone and the assumption that challenging targets are optimal (Section 3.1), remuneration committees may present challenging performance targets with more optimistic tone so as to signal superiority and legitimacy, reduce cost of capital and thereby enhance share price. The following hypothesis was developed:
4. SAMPLE SELECTION AND RESEARCH DESIGN
This section quantifies the extent to which writing styles of compensation reports are sensitive to performance target choice. To test the previously developed hypothesis I will translate these concepts into measures and use regression analysis.
4.1 Sample Selection and Data Sources
The sample for the study consists of 258 UK listed firms (both from FTSE 350 and SmallCap). The examination period for the DRR readability covers the fiscal years 2002 to 2009. I chose this particular period as the new DRR regulation became effective for UK listed firms starting in fiscal years ending on or after December 31, 2002. The initial sample consisted of 3120 firm-year combinations from 390 firms but after excluding the financial and public firms (which are incomparable due to their regulatory and financial structure) and accounting for the missing data the final sample included 1837 firm-year observations from 258 different firms. Table 1 describes the sample selection process.!!
TABLE 1
Sample Selection Process.
Procedure Number of firms
FTSE 350 and 40 small cap 390 firms
(-) minus: firms in financial or public sectors (with SIC in 4900-4999 or 6000-6999)
(111 firms) (-) minus: firms without sufficient financial
information and information on governance
(21firms)
Final sample 258 firms (1840 firm-year observations)
(-) minus firm-years without information of DRR (3 firm-year observations)
Final sample 258 firms (1837 firm-year observations)
Sample composition by fiscal year
2002 198 2003 204 2004 222 2005 230 2006 237 2007 247 2008 253 2009 246
collection manual. Lastly, while there is abundant research on performance-related remuneration, empirical studies on performance target choice are, by contrast, relatively few. One of the main reasons for the paucity of empirical work is that traditional accounting databases provide very little readily available information on the performance targets of the firm. As a consequence, this information was collected manually as well, from the remuneration reports of all companies via a detailed data collection procedure.
4.2 Research Design
4.2.1 Dependent Variables: Writing Style
Table 2 defines all the variables that were included in this study. The dependent variable takes two different measures of writing style for the two regression models. The first variable to express DRR writing style is DRR readability. Consistent with prior literature (Jones and Shoemaker, 1994; Courtis, 1995, 1998; Linsley and Lawrence, 2007; Li, 2008) I will measure DRR readability using a linguistic index called the Gunning Fog Index (Fog_Index). The index expresses ease of readability based on the average sentence length (number of words) and the average word complexity (defined as the number of words with three or more syllables) per report. The index score provides an indication of the number of years of formal education that someone needs in order to easily understand the text after first time reading it. Technical reports typically have an index between 10 and 15; the fog index of this paper is, for instance, 14.34. Texts with a fog score under 12 are considered to have a reasonable readability for a wide audience and professional texts with a fog score over 18 are considered hardly readable. A high (low) index thus indicates a low (high) readability. Despite criticism on the index sometimes being mis-specified when used in the business context “the Fog Index has become a popular measure of financial disclosure readability in recent accounting and finance research” (Loughran and McDonald, 2011a: 24). The popularity is mainly due to its simplicity and objectivity of calculation and in the US “the SEC has even contemplated the use of the Fog Index to help combat poorly written financial documents” (Loughran and McDonald, 2011a: 24).
for analyzing annual reports (10-K’s). Loughran and McDonald (2011b) categorised between different types of words and created a list for positive words. I will use this list to measure optimistic tone and divide the occurrence of positive words by the total number of words (POS_Words).
4.2.2 Independent Variables: Performance Target Choice
The explanatory (independent) variable, performance target choice, cannot be measured directly, so I will use proxy variables to measure target setting and performance vesting target difficulty3. First of all I will create two dummies (Dummy_PVSO and
Dummy_LTIP) to determine if firms use performance-vested equity contracts to grant options and / or shares. The dummy will equal 1 if firms use PVSO / LTIP compensation and 0 otherwise.
Secondly I will create four dummies to determine what kind of target measures firms use. I will distinguish between accounting performance targets (Dummy_EPS_Options and
Dummy_EPS_Shares) and stock performance targets (Dummy_TSR_Options and
Dummy_TSR_Shares). I assume that TSR, compared to EPS, represents more challenging performance-vesting targets for two reasons. Firstly, TSR targets are based on share prices, which are, compared to accounting earnings, less sensitive to executive opportunism and are designed to better align interests of executives and shareholders. Secondly, TSR targets are relative targets and economic theory states that measuring firm performance is more meaningful when it is done in comparison to a comparator group (Holmstrom, 1982). I therefore expect that there might be a relation between Fog_Index / POS_Words on the one hand and these four dummies on the other hand.
Thirdly I will use two variables to determine the performance vesting target difficulty of EPS based targets for PVSOs and LTIPs (EPS_Threshold_Options and
EPS_Threshold_Shares). The variables measure the threshold level of EPS targets, in % over RPI per annum.
Lastly I will create two discrete variables to determine performance vesting target difficulty. Adapted from the ideas of previous studies on performance target difficulty in PVSO design (Sautner and Weber, 2011; Kuang and Qin, 2012), I will construct a
3 Performance targets for the tranche of vesting equity in a particular year can be different from equity granted in
the three years following the grant, but some compensation contracts contain provisions that allow re-testing of the performance targets in later years if the targets were not achieved. Such retesting is only permitted in some contracts (23% of all PVSO observations and 2% of all LTIP observations) and would make the chances of achieving targets easier. All five dimensions are measured on the same scale (3-1) and the higher (lower) the overall aggregated proxy score the more difficult (easy) the performance vesting target. Following my hypothesis I therefore expect that there might be a relation between Fog_Index /
POS_Words on the one hand and Difficulty_Options / Difficulty_Shares on the other hand.
4.2.3 Control Variables
Control variables may also need to be utilized to account for possible influences on DRR writing styles other than performance target choice. Therefore, the regressions will need to control for the three main influences prior found in the literature. First prior studies found mixed evidence on the relation between firm size and disclosure. Ahmed and Courtis, (1999) found firm size to be positively related to disclosure, similarly Hermalin and Weisbach (2012) show that larger firms will tend, ceteris paribus, to have better disclosure regimes. Contrary, Li (2005) found that larger firms tend to have annual reports that are longer and more difficult (complex) to read. In order to limit potential size biases I included both FTSE 350 and SmallCap firms. I will control for firm size with the use of the natural logarithm of market value as a proxy (Ln_MV). Second previous studies found mixed evidence on the relation between firm growth prospects and disclosure regimes. Rogers et al. (2011) found firms with more promising growth prospects to have more favourable information to display. Contrary, Li (2005) suggests that growth firms have more complex and uncertain business models and found that growth prospects are negatively related with annual report readability. I will measure growth opportunities by using firm’s market-to-book ratio as a proxy (MtB). Third firms performing well will obviously have better results to convey which would influence readability and tone. Subramanian et al. (1993) found annual reports of profitable firms are significantly easier to read than those of poor performers. I will control for firm performance using firm’s return on assets as an approximation (ROA).
relation between employing a compensation consultant and executive remuneration contracts (Murphy and Sandino, 2010). Therefore I will control for compensation consultants using a dummy that equals 1 if the firm employs "1 consultant(s) and zero otherwise (Consultant). Second, prior studies found evidence of executives’ potency to bargain and influence their own remuneration contracts with the board of directors (e.g. Core et al., 1999; Bebchuk et al., 2009; Bushman and Smith, 2001; Bebchuk et al., 2002; Murphy, 2002). Consequently I will control for the level of board dependence by measuring the amount of dependent directors divided by total number of directors serving on the board (Board_Dependence). Third, corporate governance literature asserts that greater CEO ownership is associated with stronger interest alignment which might influence DRR readability and tone. Hence I will control for the level of CEO equity holdings by the natural logarithm of 1 plus the ownership of the CEO (CEOHold). Fourth, previous studies found evidence of a direct monitoring role of major shareholders, which might impact target choice and DRR writing style. So I will control for the presence of block holders with a discrete variable that equals 2 if there is at least one block holder with "10% of shares; 1 f there is at least one block holder with "5% but <10% of shares and 0 otherwise (BlockHold). Lastly I will control for the monitoring role and the activity of the remuneration committee with the committee size (RC_Size) and the number of committee meetings (RC_Meet) respectively.
Variable name Description
Dependent Variables
Fog_Index The Fog Index is a measure of readability; it is defined as a linear combination of average sentence length and percentage of complex words (words with ! 3 syllables) whose scale provides an estimate of grade level. Calculation = 0.4 (average # of words per sentence + percent of complex words)
POS_Words The pre-specified positive wordlist is a measure of tone; it is the degree of positive words used (compared to the total number of words) to express optimistic sentiment
Independent Variables
Explanatory Variables
Dummy_PVSO A dummy variable to measure the use of PVSO contracts; it equals 1 if the firm uses PVSOs and 0 otherwise
Dummy_LTIP A dummy variable to measure the use of LTIP contracts; it equals 1 if the firm uses LTIPs and 0 otherwise
Dummy_EPS_Options A dummy variable to measure PVSO target choice; it equals 1 if there are EPS based PVSO vesting targets and 0 otherwise
Dummy_TSR_Options A dummy variable to measure PVSO target choice; it equals 1 if there are TSR based PVSO vesting targets and 0 otherwise
Dummy_EPS_Shares A dummy variable to measure LTIP target choice; it equals 1 if there are EPS based LTIP vesting targets and 0 otherwise
Dummy_TSR_Shares A dummy variable to measure LTIP target choice; it equals 1 if there are TSR based LTIP vesting targets and 0 otherwise
EPS_Threshold_Options A variable to measure the difficulty of EPS based PVSO vesting targets; it is the EPS threshold level
EPS_Threshold_Shares A variable to measure the difficulty of EPS based LTIP vesting targets; it is the EPS threshold level
Difficulty_Options A discrete variable to measure PVSO vesting-target difficulty; it is a multidimensional aggregated score based on five aspects.
Difficulty_Shares A discrete variable to measure LTIP vesting-target difficulty; it is a multidimensional aggregated score based on five aspects.
Control Variables
Ln_MV Ln_MV is a measure of firm size; it is the natural logarithm of market capitalization MtB MtB ratio is a measure of firm growth; it is the market value of the shares / book
value of the shares
ROA ROA is a measure of firm performance: it is the return on assets
Consultant A dummy variable to measure firms' use of compensation consultants; it equals 1 if the firm employs at least one compensation consultant and 0 otherwise.
Board_Dependence Board_Dependence is the number of dependent directors divided by the total number of directors serving on the board
CEOHold CEOHold is the natural logarithm of 1 plus the ownership of the CEO
BlockHold BlockHold is a discrete variable that equals 0 if no shareholders hold at least 5 percent ownership; 1 if at least one shareholder owns 5–10 percent; and 2 if at least one shareholder owns more than 10 percent of the firm shares
RC_Size RC_Size is the number of directors serving on the remuneration committee RC_Meet RC_Meet is the number of remuneration committee meetings during the fiscal year Overpay A dummy variable to measure excess pay; it equals 1 if the difference between actual
and expected compensation is located in the top quartile of the full sample and 0 otherwise
4.2.4 Empirical Models
To test my hypotheses I constructed the following models:
H1: Readability = Constant + # Target_Choice + $1 Ln_MV + $2 MtB + $3 ROA + $4 Consultant + $5 Board_Dependence + $6 CEOHold + $7 BlockHold + $8 RC_Size + $9 RC_Meet + $10 Overpay + % Industry +% Year + &
H2: Tone = Constant + # Target_Choice + $1 Ln_MV + $2 MtB + $3 ROA + $4 Consultant + $5 Board_Dependence + $6 CEOHold + $7 BlockHold + $8 RC_Size + $9 RC_Meet + $10 Overpay + % Industry +% Year + &
Where Readability = Fog_Index and where Tone = POS_words to capture the readability and
tone of the compensation report; where Target_Choice = Dummy_PVSO, Dummy_LTIP, Dummy_EPS_Options, Dummy_TSR_Options, Dummy_EPS_Shares, Dummy_TSR_Shares, Difficulty_Options or Difficulty_Shares, where & = error term; all other control variables are
defined as previously.
Panel A: Summary Statistics
Variable N Mean StDev Min 25% Median 75% Max
Dependent Variables Fog_Index 1769 17.37 1.62 11.33 16.50 17.49 18.37 23.31 POS_Words 1748 0.84 0.28 0.07 0.64 0.82 1.01 1.75 Independent Variables Explanatory Variables Dummy_PVSO 1837 0.33 0.47 0.00 0.00 0.00 1.00 1.00 Dummy_LTIP 1837 0.54 0.50 0.00 0.00 1.00 1.00 1.00 Dummy_EPS_Options 611 0.80 0.40 0.00 1.00 1.00 1.00 1.00 Dummy_TSR_Options 611 0.20 0.40 0.00 0.00 0.00 0.00 1.00 Dummy_EPS_Shares 986 0.52 0.50 0.00 0.00 1.00 1.00 1.00 Dummy_TSR_Shares 988 0.74 0.44 0.00 0.00 1.00 1.00 1.00 EPS_Threshold_Options 454 3.66 1.60 1.00 3.00 3.00 4.00 12.00 EPS_Threshold_Shares 413 3.80 2.24 0.00 3.00 3.00 5.00 16.20 Difficulty_Options 611 6.27 2.31 2.00 5.00 6.00 8.00 13.22 Difficulty_Shares 989 8.24 3.21 2.00 7.00 8.00 9.00 15.00 Control Variables Ln_MV (=size) 1729 6.76 1.67 2.58 5.62 6.61 7.69 11.59 MtB (=growth) 1729 1.13 0.90 0.02 0.54 0.87 1.39 4.51 ROA (=performance) 1729 0.06 0.07 -0.19 0.03 0.06 0.09 0.26 Consultant 1729 0.81 0.40 0.00 1.00 1.00 1.00 1.00 Board_Dependence 1837 0.50 0.14 0.08 0.40 0.50 0.58 0.92 CEOHold 1783 10.31 5.56 0.00 10.57 12.75 13.81 16.15 BlockHold 1783 1.32 0.78 0.00 1.00 2.00 2.00 3.00 RC_Size 1783 3.76 1.05 0.00 3.00 4.00 4.00 7.00 RC_Meet 1636 4.38 1.98 0.00 3.00 4.00 6.00 10.00 Overpay 1729 0.26 0.44 0.00 0.00 0.00 1.00 1.00
Panel B: dummy statistics
Explanatory Variables Freq. % total EPS TSR Both PVSO targets 611 33.3% 79.7% 20.5% 7.9%
5. EMPIRICAL RESULTS AND DISCUSSION 5.1 Descriptive Statistics and Correlations
Table 3 panel A reports the descriptive statistics of the dependent and independent variables. Firstly, with respect to the independent variables; readability and tone, the following insights emerge. The mean and median readability (Fog_Index) are 17.37 and 17.49 meaning that readers need more than 17 years of education to understand the information in an average DRR. Technical reports typically have an index between 10 and 15 and this score thus indicates, in line with previous research, that most DRRs are very difficult to read. The mean and median optimism (POS_Words) are 0.84% and 0.82% meaning that a little less than 1% of all words used in the DRR express optimistic sentiment. This finding is consistent with the findings of Loughran and McDonald in the US (2011b).
TABLE 3
Secondly, with respect to the explanatory variables; performance target choice, the following insights emerge. Table 3 panel B reports the results of the dummy variables for performance-vested equity contracts. It shows most firm-year observations use PVSOs (33.3%) and / or LTIPs (53.8%) in the compensation contracts of their directors with the latter used the most. Of all observations with PVSO targets, 79.7% are EPS based, 20.5% are TSR based and 7.9% use a combination of EPS and TSR targets. Of all observations with LTIP targets, these percentages are 51.6%, 73.6% and 32.2% respectively. Table 3 panel A indicates that the mean and median EPS threshold targets for options are 3.66 and 3.00 and for shares are 3.80 and 3.00. Furthermore it shows that the multidimensional target difficulty score is substantially different for PVSOs and LTIPs. The mean and median score of PVSOs (Difficulty_Options) are 6.27 and 6.00 and the mean and median score of LTIPs (Difficulty_Shares) are 8.24 and 8.00. This difference indicates that, on average, the vesting targets attached to LTIPs are more difficult than the ones attached to PVSOs.
Lastly, with respect to the control variables, the following three insights emerge. Firstly, Table 3 panel A shows most firms (81%) employ at least 1 compensation consultant. This widespread practice seems to indicate that organizations recognise the value of the services offered by these consultants. Second, on average boards are quite dependent; with 50% of all directors in the board being dependent / insiders. Lastly, most firms (84%) have at least 1 large shareholder with "5% of all shares that might provide a direct monitoring role.
Year 0 1 0 1 0 1 0 1 0 1 0 1 2002 count 100 98 142 56 19 79 85 13 24 32 19 37 5.22 7.14 % total 49% 28% 81% 13% 57% 66% 2003 count 102 102 141 63 14 88 86 16 31 32 20 43 5.63 7.32 % total 50% 31% 86% 16% 51% 68% 2004 count 122 100 127 95 17 83 82 18 51 44 27 68 6.19 7.83 % total 45% 43% 83% 18% 46% 72% 2005 count 146 84 109 121 18 66 65 19 67 53 30 90 6.69 8.12 % total 37% 53% 79% 23% 44% 74% 2006 count 165 72 89 148 13 59 56 16 73 75 36 112 7.10 8.29 % total 30% 62% 82% 22% 51% 76% 2007 count 187 60 86 161 12 48 47 13 75 86 39 122 6.93 8.70 % total 24% 65% 80% 22% 53% 76% 2008 count 202 51 77 176 12 39 38 13 75 101 49 127 7.06 8.53 % total 20% 70% 76% 25% 57% 72% 2009 count 202 44 77 169 19 25 27 17 80 87 40 129 6.34 8.50 % total 18% 69% 57% 39% 51% 76% Full sample 1226 611 848 989 124 487 486 125 476 510 260 728 6.27 8.24 % total 33% 54% 80% 20% 52% 74% mean mean PVSO difficulty LTIP diffculty Dummy_ PVSO Dummy_ LTIP Dummy_ Options_ EPS Dummy_ Options_ TSR Dummy_ Shares_ EPS Dummy_ Shares_ TSR TABLE 4
Development Performance Target Choice Over Time (2002-2009).
The medians of the explanatory variables in Table 3 are used as a cut-off point to compose sub samples. In the first four subsamples that are reported in Table 5 (Panels A to D) I compare two groups; a group equal to or smaller than the median target difficulty (for options and shares) and a group larger than the median target difficulty (median
EPS_Threshold_Options 3; Difficulty_Options 6; EPS_Threshold_Shares 3; Difficulty_Shares
8). I use an Independent Samples T Test to compare the mean scores and a Mann-Whitney Test to compare the median scores between the two subgroups4. These tests showed some statistically significant differences between the groups that will be discussed below. Panel A indicates that for more difficult EPS_Threshold_Options targets (>3) the mean and median tone is more optimistic (higher percentage of positive words) than for easier
4 A t-test is a parametric test for the joint equality of means of each variable across the two samples that assumes
EPS_Threshold_Options targets ('3) (p = <0.01). Panel B indicates that for the higher overall Difficulty_Options (>6) group the average tone is more optimistic than for the lower overall Difficulty_Options ('6) group (p = 0.1). Panel C shows that for more difficult EPS_Threshold_Shares targets (>3) the mean and median optimistic tone is higher (use of
positive words is higher) than for easier EPS_Threshold_Shares targets ('3) (p = <0.01). Secondly, this panel indicates that when EPS_Threshold_Shares targets are more difficult, there is more overpay (higher mean and median) than when EPS_Threshold_Shares targets are easier (p = <0.1). Panel D shows that for higher overall Difficulty_Shares (>8) the readability is lower (mean and median fog index is higher) (p = <0.01) than for the group of lower overall Difficulty_Shares ('8). Secondly, this panel indicates that when
Difficulty_Shares is higher, there is less overpay (lower mean and median) than when Difficulty_Shares is lower (p = <0.1).
TABLE 5
Summary statistics of subsamples.
Variable Mean Median Mean Median T-test Mann_
Whitney Panel A: Fog_Index 17.437 17.610 17.276 17.470 0.319 0.194 POS_Words 0.790 0.770 0.879 0.840 0.002 *** 0.001 *** Overpay 0.271 0.000 0.269 0.000 0.961 0.961 Panel B: Fog_Index 17.305 17.535 17.417 17.560 0.394 0.472 POS_Words 0.810 0.770 0.849 0.810 0.100 * 0.136 Overpay 0.278 0.000 0.289 0.000 0.758 0.757 Panel C: Fog_Index 17.396 17.535 17.384 17.600 0.945 0.491 POS_Words 0.810 0.785 0.892 0.880 0.003 *** 0.002 *** Overpay 0.207 0.000 0.280 0.000 0.096 * 0.092 * Panel D: Fog_Index 17.346 17.340 17.707 17.790 0.000 *** 0.000 *** POS_Words 0.883 0.860 0.868 0.850 0.423 0.210 Overpay 0.289 0.000 0.237 0.000 0.073 * 0.075 * Panel E: Fog_Index 17.355 17.440 17.570 17.640 0.007 *** 0.031 ** POS_Words 0.823 0.800 0.872 0.840 0.001 *** 0.008 *** Dummy_PVSO 0.323 0.000 0.370 0.000 0.073 * 0.068 * Dummy_LTIP 0.533 1.000 0.562 1.000 0.283 0.284 Dummy_EPS_Options 0.817 1.000 0.750 1.000 0.071 * 0.071 * Dummy_TSR_Options 0.183 0.000 0.244 0.000 0.100 * 0.100 * Dummy_EPS_Shares 0.540 1.000 0.482 0.000 0.116 0.115 Dummy_TSR_Shares 0.750 1.000 0.719 1.000 0.346 0.337 Difficulty_Options 6.311 6.000 6.215 6.000 0.652 0.952 Difficulty_Shares 8.318 8.000 8.193 8.000 0.603 0.371
Notes:***, **, * Coefficient is statistically significant at the 1%, 5%, and 10% level, respectively (2-tailed).
p-value
EPS_Threshold_Shares ! 3 EPS_Threshold_Shares > 3 EPS_Threshold_Options ! 3 EPS_Threshold_Options > 3
Difficulty_Options ! 6 Difficulty_Options > 6
No overpay ( = 0) Yes overpay (= 1)
In the last subsample that is reported in Table 5 (Panel E) I use the same tests to compare two groups; a group where overpay is absent (overpay=0) and a group with overpay present (overpay=1). Panel E first reports the relation between overpay and writing style and indicates that for the subgroup with overpay (=1), the mean and median readability is lower (p = 0.01) and the mean and median optimistic tone is higher (p = 0.01) than for the subgroup without overpay (=0). Finally Panel E reports the relation between overpay and the explanatory variables for performance target choice. It indicates that when overpay is present, the overall target difficulty score for options (Difficulty_Options) and shares (Difficulty_Shares) is lower but the differences are not significant. The panel shows that the subgroup with overpay (=1) uses (comparing mean and median) more PVSO contracts, less EPS targets and more TSR targets in PVSO contracts (p = 0.1) than the subgroup without overpay. The other differences are not significant. Regression analyses will have to reveal further details on the differences found.
Table 6 reports Pearson correlations among the regression variables from which the following three findings emerge. It shows firstly, that the correlation among the dependent variables (readability and tone) is weak (<0.017) which indicates that they measure different aspects of writing style. Secondly, it shows that some of the explanatory variables are highly correlated; for instance between Dummy_EPS_Options and Dummy_TSR_Options (-0.521). The correlations between Dummy_PVSO, Dummy_EPS_Options, Dummy_TSR_Options and
Difficulty_Options cannot be computed because at least one of the variables is constant. The
same applies to Dummy_LTIP, Dummy_EPS_Shares, Dummy_TSR_Shares and
Difficulty_Shares. This is very likely since the former variable is an aggregation of the later
ones. Finally, it shows that the correlations between the control variables are weak (all <0.477), which implies that they measure different dimensions of firm- and governance characteristics and do not merely substitute one another. The sign of their coefficients also show significant correlations with the writing style measures which indicates they are relevant to include in the regression analyses. In addition to these three findings the Pearson correlations matrix was used to check for multicollineairity with p-value > 0.5 as a cut-off point for large correlation (Cohen, 1988). The correlation coefficient between any pair of variables indicates no severe multicollinearity, except for the one logical high correlation between the two explanatory variables stated above (Dummy_EPS_Options and
Dummy_TSR_Options), but since I will used them in separate regression models I considered
5.2 Main Results
The findings for the two empirical models are summarized in Tables 7 and 8. I first test the models without any explanatory variables (control variables only) to measure the model fit (Model 1). Afterwards I perform the regression in three steps. The first step is to test if the use of any performance-vesting equity contracts itself impacts writing style (Models 2 and 3). The second step is to test if the choice of performance targets, accounting measures and / or stock measures, impacts writing style (Models 4 to 7). The third and final step is to test if the difficulty of the performance-vesting targets impacts writing style (Models 8 and 9). The most remarkable results, per step, will be discussed below.
5.2.1 Performance target choice and DRR Readability.
Table 7 summarizes the empirical results for Hypothesis 1a/1b. All nine regressions show that the F-value is positive and significant at p<0.01 which suggests that the models are valid and have explanatory power. I will discuss my findings on the impact of the explanatory variables first and my findings on the control variables later. Note that a higher score in
Fog_Index indicates lower readability. The first step (Models 2 and 3) shows that the use of performance-vesting equity contracts impacts readability; Dummy_LTIP is positively related to Fog_Index (+0.171**). The second step (Models 4 to 7) shows that the choice of performance measures impacts readability; Dummy_EPS_Options and Dummy_TSR_Shares are positively related to Fog_Index (+0.527*** and +0.372***). The third step (Models 8 and 9) shows that the difficulty of LTIP targets, Difficulty_Shares is positively related to
Fog_Index (+0.048***); indicating that more challenging targets result in lower readability.
The results show two different outcomes; a negative and a positive relation between difficulty and readability, both will be discussed below.
performance is more meaningful when it is done in comparison to a comparator group (Holmstrom, 1982). Lastly, adapted from the ideas of previous studies (Sautner and Weber, 2011; Kuang and Qin, 2012) on performance target difficulty in PVSO design I constructed a multidimensional target difficulty score for PVSOs and LTIPs based on five aspects and I assumed that higher scores represent more challenging targets than lower scores.
