The effects of reconciliation quality on the investment decision of non-professional investors

(1)

The effects of reconciliation quality on the investment

decision of non-professional investors

Name: J.B. (Jelmer) van Ommen Student number: 11080663

Thesis supervisor: dr. J.J.F. (Jeroen) van Raak Date: 26 June 2017

Word count: 22689, including appendices

MSc Accountancy & Control, specialization Accountancy & Control Faculty of Economics and Business, University of Amsterdam

(2)

Statement of Originality

This document is written by student J.B. (Jelmer) van Ommen 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.

(3)

Abstract

Prior research (Elliott, 2006; Dilla, Janvrin & Jeffrey, 2013) find contradictory evidence to whether reconciliation between non-GAAP and GAAP mitigates the upwards effect of the investment decision caused by non-GAAP measures for non-professional investors. This study is a response on the call of Dilla, Janvrin & Jeffrey (2014) to examines the impact of reconciliation quality on the investment decision of non-professional investors. In this experiment, participants make multiple investment judgments, based on a manipulated press release. My participants are economic-students and economic graduates. This study examines the impact of the reconciliation quality following the decision-making model of Maines & McDaniel (2000). (1) As predicted, I find no significant differences in the information acquisition. Contradictory to my expectation, (2) no significant differences in performance evaluation and (3) no differences in the weighting of the performance evaluation are observed. (4) I find some evidence of an increase in investment valuation if middle-quality reconciliation is provided. Moreover, I find limited evidence of an decrease in investment valuation if high-quality reconciliation is provided. In conclusion, this study reveals that the quality of the reconciliation between non-GAAP and GAAP is a moderating factor when the mitigating effect occur, observed by Elliott (2006).

Keywords: non-GAAP; non-professional investors; reconciliation quality, mitigating effect Data Availability: Data used in this study is self-collected and only available for the first and the second reader for grading my thesis.

(4)

1 Introduction

This study examines the influence of reconciliation quality of non-Generally Accepted Accounting Principles (GAAP) and GAAP measures on the investment decision of non-professional investors. Regulators and lawmakers are concerned that non-GAAP measures might be misleading, especially, for non-professional investors (Burns, 2001), and therefore require reconciliation between non-GAAP and the comparable GAAP measures (Young, 2014). A reconciliation between non-GAAP and GAAP provides additional information by indicating the effects of the excluded account names on key numbers and rations (Nelson & Tayler, 2007), however, the representation format and the quality differ between firms and over time (Hitz, 2010, Zhang & Zheng, 2011, Aubert & Grudnitski, 2014). Prior research indicates that, especially, non-professional investors are influenced by the presence of non-GAAP measures (Frederickson & Miller, 2004; Elliott, 2006). Moreover, non-professional investors hold around one-third of all outstanding shares worldwide (Cohen, Holder-Webb, Nath & Wood., 2011), and thus, they have a substantial effect on the capital markets.

Chen, Krishnan and Pevzner (2012) define non-GAAP measures as a type of voluntary earnings disclosures in which earnings releases show financial numbers, even though certain income statement components are excluded. In addition, Andersson and Hellman (2007) define non-GAAP measures as financial disclosures that represent a situation if certain situations occur or never occur. The basic principle is that GAAP earnings are modified when certain unusual events never happened. Note, that earnings-before metrics are not to be qualified as non-GAAP earnings measures (Andersson & Hellman, 2007). Non-GAAP earnings are also known as pro forma earnings, street earnings or adjusted earnings (Young, 2014). In this study, non-GAAP measures are defined as voluntary disclosures in which excluded certain income statements component, to represent unusual events never occur.

As an example of the use of non-GAAP reporting, LinkedIn reported a GAAP net loss of 166 million and a non-GAAP net income of 373 million over 2015. The GAAP net income have been adjusted for stock-based compensation, non-cash interest expense related to convertible senior notes, amortization of intangible assets, accretion of redeemable non-controlling interest, fair value adjustment on other derivative and income tax effects and adjustments. They include a reconciliation table between non-GAAP and GAAP and an explanation of these measures (LinkedIn Corporation, 2016).

There is contradictory evidence about the effects of reconciliation between non-GAAP and GAAP on the investment valuation of non-professional investors. Elliott (2006) indicates

(7)

measures, however, she finds that the effect is mitigated by the presence of simultaneously reconciliation between non-GAAP and GAAP. In contrast, Dilla, Janvrin and Jeffrey (2013) find that graphical displays of non-GAAP measures influences the non-professional investors judgement, even when reconciliation between non-GAAP and GAAP is present. Dilla, Janvrin and Jeffrey (2014) find the occurrence of this effect depends on that the level of knowledge. Since both, Elliott (2006) and Dilla et al. (2013) use MBA-students as proxy for non-professional investors. The level of investor sophistication might not explain this contradictory evidence. Both studies use different reconciliation representations, and the results of archival studies suggest that quality of the reconciliation might be a moderating factor (Zhang & Zheng, 2011; Aubert & Grudnitski, 2014). This research answers the call of Dilla et al. (2014) for an experimental study to examine the effects of quality of the reconciliation representation on the investment decision of non-professional investors. Research also indicates that, especially in Europe, there are differences in the quality of the reconciliation (Hitz, 2010; Aubert & Grudnitski, 2014), which makes it useful for examining.

This study contributes to the literature by examining the effect of the quality of the reconciliation on the investment decision of non-professional investors. Archival studies examine valuation differences between the different reconciliation quality but do not provide evidence on whether non-professional investors use or are affected by reconciliation quality in their investment decisions. Experimental studies give the possibility to examine how investors use different types of information and how this information affects their investment valuation (Frederickson & Miller, 2004). From a societal point of view, examining the effects of the reconciliation quality is also interesting for regulators and lawmakers because provides evidence that reconciliation quality effects the investment decision. Moreover, non-professional investors have substantial effects on the capital market, thus, regulation should prevent this investors group also.

This study follows the decision-making framework of Maines & McDaniel (2000) to examine how the quality of the reconciliation between GAAP and non-GAAP affects the investment valuation. This model assumes that the decision making is a three-stage process: acquiring, evaluating and weighting. Individuals evaluate and weight the gains and losses which can be affected by formulations and the expectation of the individual (Kahneman & Tversky, 1979). These expectations can be irrational or influenced by uninformative information, but can be mitigated by simultaneously explanation of the information (Nelson & Tayler, 2007). The reaction is also affected by the presentation of information, because this should fit the task

(8)

(Vessey, 1991; Vessey, 1994; Vessey, Galletta, 1991). Based on these behavioural theories, I hypothesize that the only high-quality reconciliation mitigates the upwards effect of the investment decision caused by non-GAAP measures.

To answer this research question, a case-based experiment with four experimental conditions is set-up. I use economic and master students and economic and business-graduates as a proxy for non-professional investors. The task of the experiment was to make multiple investment judgements based on a press release with the earnings announcement. The press release follows the standardized pattern and is manipulated for the presence of non-GAAP measures and the reconciliation quality. One group receives a press release with only GAAP measures and is used as a control group. The other groups receive a press release including non-GAAP measures and a reconciliation of low, middle or high quality.

Consistent with my expectation, I find no significant result that the reconciliation quality affects the information acquisition. Contradictory to my expectations, I find no significant results indicating that the performance evaluation is affected by non-GAAP measure or the quality of the reconciliation. Moreover, I find no significant differences in the weighting of the performance evaluation between the quality levels. In addition, the study shows weak evidence that non-GAAP measures increases the final investment valuation. I some evidence find of an increase in investment valuation when middle-quality reconciliation is present. Furthermore, I find weak evidence that high-quality reconciliation mitigates the upwards the effect caused by non-GAAP measures. In conclusion, the findings reveal that the quality of the reconciliation is a moderating factor when the mitigating effect, as assessed by Elliott (2006), of reconciliation occur for non-professional investors. These findings contribute to the prior literature by revealing that the quality of the reconciliation moderates the effects of reconciliation between non-GAAP and GAAP on the investment valuation. This study contributes to the behavioural theory by indicating that the representation of financial measures effects the final investment decision of non-professional investors.

The rest of the paper is organized as follows: in the second chapter, I review the prior literature about non-GAAP measures and develop my hypotheses. Chapter 3 describes the experimental set-up. Chapter 4 describes the analyse of the data and presents the results. The final chapter concludes the results describes the implications of the study.

(9)

In this section I review the prior literature to provide a background on the use of non-GAAP measures. Furthermore, I provide a background about the different reconciliation representations quality. I review the prior research about the impact of non-GAAP measures on the investment decision of non-professional investors. I introduce the decision making of Maines and McDaniel (2000) Lastly, I develop my hypothesis based on this model and behavioural theory. This section starts with explaining the regulatory requirements about the use of non-GAAP measures.

2.1 The regulatory landscape

The regulatory landscape for non-GAAP measures varies per country and reporting channel (Young, 2014). US-GAAP regulations prevent firms to present non-GAAP measures as part of the audited financial statements, but non-GAAP measures outside these financial statements are allowed (Hitz, 2010). In contrast, International Financial Reporting Standards (IFRS), which is used in the Europe Union, has lower requirements about the presentation of certain line items and subtotal in the income statement, such as non-GAAP measures (Hitz, 2010). For example, IAS 33 allows management to report non-GAAP earnings per share on the face of the income statement or in the notes. Non-GAAP measures must then be accomplished with a reconciliation between non-GAAP and comparable GAAP financial measure reported in the income statement.

In the USA, Regulation G covers press releases, conference calls, investors presentations and other communicating channels. This regulation requires that non-GAAP measures includes a comparable GAAP measure and reconciliation between non-GAAP and GAAP. Other regulation, such as Item 10(e) and Item 12, also require reconciliation between non-GAAP and comparable GAAP measure. The European Union has a low degree of regulation reporting outside the financial statements and only restrict firms to report misleading financial measures (Hitz, 2010, Aubert, 2009). Because of the low degree of regulation, the European securities regulators have, published voluntary recommendations about non-GAAP measures, with comparable requirements as the USA regulation (Young, 2014).

Young (2014) argues that the inconsistency about non-GAAP regulation between the important reporting standards is insufficient, because non-GAAP measures can be informative for every investor, but current regulations signals that non-GAAP measures are unreliable and less trustworthy. Furthermore, the current regulation limits the informativeness of the

(10)

non-GAAP measures, because firms are only required to explain why they excluded certain income line items, but not why some of the line items are included (Young, 2014).

2.2 The use of non-GAAP measures

GAAP – for example US GAAP or IFRS - ensures reliable and consistent information across firms and time with respect to the financial reporting outputs, however, these standard regulations also limits reporting practices. Firms are heterogeneous, and therefore, non-GAAP measures can be seen as a market response for more firm-specific financial information (Young, 2014). In the past two decades, there has been an increase in the use of non-GAAP information in firms’ communication. Moreover, financial analyst and media write more about the non-GAAP information (Walker, 2016). Bhattacharya, Black, Christensen and Mergenthaler (2004) reveal that in the period of 1998-2000 the use of non-GAAP measures increased with 417 percent, even after SEC intervention, the use of non-GAAP measures increases. In 2005 44% of the S&P100 firms reports non-GAAP measures and increased to 60% of the S&P100 firms in 2010 (Webber, Nichols, Street & Cereola, 2013). Similar research reveals the same trends for European firms (Isidro & Marques, 2015).

Consistent with the theory of Grossman (1981) and Verrecchia (1983), and stating that firm managers only disclose beneficial measures, Bhattacharya, Black, Christensen and Mergenthaler (2003) reveal that the most non-GAAP measures overstates the GAAP measures. Some firms report a GAAP loss a non-GAAP profit. Secondly, firms place more emphasis on the non-GAAP measures than the comparable GAAP measures. Furthermore, only 10% of the firms made identical adjustment every period (Bhattacharya et al., 2004). The inconsistent definition of their non-GAAP measures decreases the comparability and might even decreases the usefulness of non-GAAP measures.

Prior research indicates two motives why firms reports non-GAAP measures (Heflin & Hsu, 2008). The first objective is to mitigate information asymmetry and agency conflicts between the management and the stakeholders by providing more firm-specific information (Choi, 2015). Research indicates that non-GAAP measures can be a response on the market demand for firm-specific measures (Bowen, Davis & Matsumoto, 2005; Levi, 2008; Curtis, McVay & Whipple, 2014). For example, Curtis et al. (2014) find that non-GAAP measures increase the amount and precision of earnings information. Research indicates that non-GAAP measures can be more informative than the comparable GAAP measure and thus narrow information asymmetry (Curtis et al, 2014; Heflin, Hsu & Jin, 2015; Leung & Veenman, 2016;

(11)

better indicators for future cash flows than GAAP measures for GAAP-loss firms.

The second motive to disclose non-GAAP measures is to improve the performance perceptions by excluding recurring expenses or including one-time benefits. The SEC Chief Accountant said in a speech that non-GAAP measures are ‘’Bad stuff’’ (SEC, 2000). Heflin & Hsu (2008) define this type as opportunistic use of non-GAAP measures. A practical example, the SEC concludes that the third-quarter 1999 earnings release of Trump Hotels & Casino was misleading for investors because they include special gains and exclude special losses (SEC, 2002). Research indicates that firms not only exclude non-recurring items, but also recurring items (Bradshaw & Sloan, 2002; Doyle, Lundholm & Soliman, 2003; Brow, Christensen, Elliott & Mergenthaler, 2012a). Moreover, research also indicates that non-GAAP measures are more influenced by the reporting strategy than by the demand of the financial markets (Bradshaw & Sloan, 2002). Additional Brown, Christensen and Elliott (2012) find when firm managers excluding recurring line items they are more likely to use less transparent reconciliation representations.

The mixed evidence suggests that these measures are used to be informative about future cash flows, but can be used to mislead investors through suggesting higher future cash flows. Black and Christensen (2009) argue that non-GAAP earnings are more informative if they exclude one-time expenses and non-recurring expenses, but they are misleading if recurring expenses are excluded. Researchers (Elliott, 2006; Young; 2014) argue that reconciliation between non-GAAP and GAAP helps investors to determine the report motives of the firm managers and increase the usefulness of the non-GAAP measures. Research indicates that the representation of the reconciliation varies between firms and over time (Bhattacharya et al., 2003; Hitz, 2010; Zhang & Zheng, 2011; Aubert & Grudnitiski, 2004). Zhang & Zheng (2011) find less differences after the SEC intervention, but after the intervention of the European securities regulators there is still variety in reconciliation representation (Hitz, 2010; Aubert & Grudnitiski, 2014).

The variety in reconciliation representation represents a variety in quality of the reconciliation between non-GAAP and GAAP. Reconciliation quality is the degree to which a non-GAAP measure fully articulates the difference between the non-GAAP measure and comparable GAAP measure (Aubert & Grudnitski, 2014). The highest reconciliation quality can be received if the degree of transparency of the relationship between GAAP and non-GAAP is the highest. Prior research (Bhattacharya et al, 2003; Hitz, 2010; Zhang & Zheng, 2011; Aubert

(12)

& Grudnitiski) have different reconciliation quality classifications. Aubert & Grudnitski (2014) classify three different reconciliation qualities: no disclosures, disclosures of adjusted account names and the last is full reconciliation between GAAP and non-GAAP measures. Zhang and Zheng (2011) classify five different reconciliation qualities: no disclosures, disclosures of adjusted account names, disclosure of adjusted account names including magnitudes, non-GAAP income statement and the last, table of reconciliation between GAAP and non-GAAP measures. However, the direction of the reconciliation quality is similar for all studies. In all studies, the lowest level of reconciliation are firms without disclosures of their non-GAAP measures and the highest level is full reconciliation.

In this study reconciliation quality is classified in accordance of Hitz (2010), because their classification is more sophisticated than Aubert & Grudnitiski’s (2014), but less extensive than Zhang and Zheng (2011). Hitz (2010) defines four different levels of reconciliation between GAAP and non-GAAP representations (table 1). The lowest level of reconciliation quality are the firms without disclosure about their non-GAAP measures. The second level are firms with disclosures about the adjusted account names without magnitudes. The third level are firms with the adjusted account names with the magnitudes This quality of reconciliation representation is quite similar to the reconciliation representation of Dilla et al. (2013). The highest level are firms with full reconciliation. This quality of reconciliation representation is quite similar to the reconciliation representation of Elliott (2006).

Table 1 - The classification of Hitz (2010) of reconciliation quality – This table presents the classification of the quality of reconciliation information. In the first column, the quality level is given. The second column is the group name of this experiment presented. The lowest level is excluded, because prior experimental research (Frederick & Miller, 2004; Elliott, 2006) provides some explanations about their non-GAAP measures, equal with the second level of Hitz (2010), even when the condition is without reconciliation. The last column presents the disclosures.

Quality level Hitz

Research

group Reconciliation between GAAP and non-GAAP 0 excluded No disclosures of adjusted account names or magnitudes 1 Non-GAAP/

Low Disclosures of adjusted account names without magnitudes 2 Non-GAAP/

Middle Disclosures of adjusted account names with magnitudes 3 Non-GAAP/

High Disclosures of complete reconciliation. 2.3 The effects of non-GAAP on the investment decision

The archival study of Johnson & Schwartz (2005) reveals that firms with non-GAAP measures in their press releases are higher prized than firms with other reporting strategies, but they argue that this relates to the underlying characteristics of the firms and less to the presence of

(13)

non-non-GAAP measures and affects the investment decision (Bhattacharya et al., 2003; Bhattacharya, Black, Christensen & Mergenthaler, 2007), because non-professional investors find non-GAAP measure more useful than professional investors (James & Michello, 2009). Frederickson and Miller (2004) examine this finding in an experimental setting, and they find that non-professional investors who received non-GAAP measures value the shares prices significantly higher than non-professional investors who only received GAAP earnings. In contrast, there was no significant increase for professional investors. They find that non-professional investors use less sophisticated models, which are more affected by non-GAAP measures than professional investors. These results suggest that, mostly, non-professional investors are influenced by the presence of the non-GAAP measures.

Furthermore, Elliott (2006) finds that the emphasis on non-GAAP measures influences the investment decision of non-professional investors. More emphasize on non-GAAP measures increase the investment valuation. Furthermore, Dilla et al. (2013) find that graphical displays of non-GAAP measures increase the investment valuation of non-professional investors. The results of the experiments of Frederickson and Miller (2004) and Elliott (2006) are verified in an archival setting by Allee, Bhattacharya, Black and Christensen (2007). They find consistent evidence that GAAP measures in press release affects the trading behaviour of non-professional investors. Secondly, they find that non-non-professional investors are affected by the emphasis place on these non-GAAP measures. The findings of Reimsbach (2014) indicate that non-professional investors evaluate firms with non-GAAP measures higher than firms that provide no non-GAAP measures, earnings before measures or both meausres. This research suggest that non-professional investors are influenced by the presence of non-GAAP measures.

Elliott (2006) finds that the presence of reconciliation between non-GAAP and GAAP mitigates this upwards effect in investment valuation. Contradictory, Dilla et al. (2013) find no mitigating effects if reconciliation between non-GAAP and GAAP is present. Allee et al. (2007) find little archival evidence that reconciliation between non-GAAP and GAAP mitigates this upwards effect. Dilla et al. (2014) examine the effects of reconciliation between non-GAAP and GAAP measures and find that level of sophistication is dependent or this effect occurs. They find that this only occurs for high sophisticated non-professional. Furthermore, archival studies suggest that the reconciliation quality is dependent on occurrence of this mitigating effect (Zhang & Zheng, 2011; Aubert & Grudnitiski, 2014).

(14)

In conclusion, prior research indicates that non-professional investors are more influenced by the presence of non-GAAP financial measures than professional investors. These studies find an increase in investment valuation if non-GAAP financial measures are present (Frederickson & Miller, 2004; Elliott, 2006; Dilla et al, 2013; Reimsbach, 2014). Research is contradictory about the mitigating effects of reconciliation between non-GAAP and GAAP (Elliott, 2006; Dilla et al., 2013). Research indicates that the level of sophistication moderates the effects of the reconciliation between non-GAAP and GAAP. Moreover, archival studies suggest that the reconciliation quality also moderates the effects of reconciliation between non-GAAP and GAAP (Zhang & Zheng, 2011; Aubert & Grudnitiski, 2014).

2.4 Framework of decision-making

A common assumption in the field of economics is that people behave purely rational. The expected utility theory states that individuals make decision under risk by comparing their expected utility values multiplied by the probabilities (Mongin, 1997). Behaviour studies, however, indicates that individuals are risk averse, and want to avoid risk even when this is irrational (Kahneman, Tversky & Michael, 1984). The prospect theory of Kahneman and Tversky (1979) assumes that individual’s analyse the potential outcomes and evaluate and weight their options and choose the option with the highest value. Individuals perceive the outcome as gains or losses and not rational as wealth or welfare. The gains and losses can be affected by the formulation and the expectations of the decision maker. The findings of the effect of the presence of non-GAAP measures (Frederikson & Miller, 2004, Elliott, 2006, Dilla et al., 2013; Reimsbach, 2014) fits the prospect theory, because non-GAAP measure are other formulation of performance measures than the comparable GAAP measure.

Behavioural studies indicate that making judgements and decisions is a three-phase process: information acquisition, information evaluation and information weighting (Maines & McDaniel, 2000). During the information acquisition phase individuals read information and determine whether this information relevant. Afterwards, the individual is able to remember the information. This model (figure 1) assumes that the process only starts after the information is acquired and during the information evaluation assess, the characteristics of the information and integrates the observations in a performance evaluation. During information weighting the individual weight reliability of the information for the final decision. Bouwman (1984) finds that although professional and non-professional investors have the same phases, there are differences between them in each phase. Non-professional tries to link findings together to explain each

(15)

and they use standardized patterns to make a decision.

Consistent with prior research (Frederickson & Miller, 2004; Elliott, 2006; Reimsbach, 2014), I use the model of Maines and McDaniel (2000) for examining the impact of reconciliation quality on the investment decision of non-professional investors.

Figure 1– Framework for the effects of non-GAAP financial measures on the non-professional investors decision making (Adapted from ‘’Effects of Comprehensive-Income Characteristics on Nonprofessional Investors’ Judgment: The Role of Financial-Statement Presentation Format’’ by L.A. Maines & L.S. McDaniel, 2000, The accounting review, 75(2), p. 184. Copyright 2000 by JSTOR.

This model visualized the decision-making process of non-professional investors. This figure shows that the investment valuation (IV) is a function of the information evaluation phase and the information weighting phase. The model shows that if the information is not acquired, there won’t be an investment valuation.

2.5 Hypothesis development

In this paragraph, I develop my hypothesis for every phase in the decision-making model (Maines & McDaniel, 2000): (H1) the information acquisition, (H2) information evaluation, (H3) information weighting and (H4) the final investment evaluation.

2.5.1 H1: information acquisition

During the information acquisition the individual acquires the information and is able to remember the information. The non-professional investors focus, primarily, on the balance sheet, the income statement and the cash flow statement. They discard the notes of the financial statements entirely (Cascino et al., 2014). Prior research indicates that non-professional investors acquire information and solve their problems in an unstructured way (Chi, Feltovich & Glaser, 1981) and focus on the overview of information. Individuals mitigates a potential information

Is information acquired? Information evaluation E(Ij) Information weighting βj Investment evaluation (IV) IV = α + ∑βj * E(Ij) + ε No information evaluation E(Ij) = 0 No information weighting βj = 0

(16)

overload (Miller, 1956) by an absolute stopping rule (Pennington & Kelton, 2016). Therefore, non-professional investors have not a complete overview about the trends. Bhattacharya,

Hackethal, Kaesler, Loos and Meyer (2012) reveal that non-professional investors do not acquire all useful information through their information acquisition strategy.

The cognitive fit theory assumes that only efficient and effective judgments can be made if the information representation support the decision-making strategy (Dilla, Janvrin & Raschke, 2010). Individuals choose an information acquisition strategy that minimize the cost of effort and errors and maximize the accuracy (Vessey, 1994). Humans have limited information processing capacity they must change sometimes from information strategy (Vessey, 1991; Vessey, 1994). This because individuals try to mitigate information overload (Miller, 1956).

Prior research suggest that non-professional investors do not acquire all useful

investment-related information. In my understanding this research (Vessey, 1991; Vessey, 1994) does not suggest that non-professional investors are unable to acquire some specific types of information. Moreover, the studies of Frederickson and Miller (2004), Elliott (2006) and Reimsbach (2014) indicate no significant differences in the acquisition of the information between different types of information or formats. This suggest that reconciliation quality does not have an impact during the information acquisition:

H1: The acquisition of performance measures does not differ significantly between press releases.

2.5.2 H2: information evaluation

In the information evaluation phase, the individuals take the number of distinct characteristics of an information into account (Elliott et al, 2007) and integrates the information (Bouwman, 1984). During this phase, the individual processes the information for understand the meaning of the metrics and analyse the given trends (Speier, 2006). During these phase people evaluate the performance of the firm and formulate the performance evaluation (Frederickson & Miller, 2004).

Non-professional investors have less understanding about the meaning of non-GAAP measures and the relationship between non-GAAP measures and their comparable GAAP measure (Elliott, 2006). Still non-GAAP measures can affect the performance evaluation. The anchoring theory assumes that individuals can be influenced by unrelated or uninformative

(17)

Kleinmuntz and Linsmeier (2001) argue that when individuals makes investment decision, the individuals are biased through prior expectations and they adjust in the general direction of the news revealed in the financial statement. Non-GAAP measure are primarily more positive than the comparable GAAP measures (Bhattacharya et al., 2003), thus, non-GAAP measures can bias the performance evaluation (Anderson & Hellman, 2007), even when investors argue that these metrics are uninformative (Chapman & Johnson, 1999: Chapman & Johnson 2000). Furthermore, Dietrich et al. (2001) find that one side positive management disclosers have an upwards effect on the market prices. Frederickson and Miller (2004), Elliott (2006), Dilla et al. (2013) and Reimsbach (2014) find an upwards effect on the performance valuation of non-professional investors. In all these studies, the non-GAAP measure(s) exceed their comparable GAAP measure. The following hypothesis summarize my prediction. These results suggest an increase in performance evaluation when non-GAAP measures are presented in the press release:

Middle-quality reconciliation provides adjusted account names and the magnitudes about the non-GAAP measures. Magat, Payne and Brucato (1986) find that making information available is not always sufficient, but the same available information must be presented in way that the information is easy to acquire. According to the cognitive fit theory of Vessey (1991), the information representation must fit the task. The cognitive fit theory distinguishes two types of information representation: spatial and symbolic. Spatial representation emphasizes trends and relationships in data and symbolic representation emphasize direct data values (Vessey & Galleta, 1991). For non-professional investors making investment judgments is a complex task (Elliott, 2006), because evaluating detailed-orientated information is difficult (Elliott, Hodge and Jackson, 2008). Research of Beach & Mitchell (1978), Einhorn and Hogarth (1981) and Payne and Miller (1982) suggest that for complex judgments, the judgment strategies influence the representation fit. Individuals shift from a symbolic to a spatial strategy (Einhorn & Hogarth, 1981), even when this results in decrease in accuracy (Russo, Dosher & Shiffrin, 1983; Speier, 2006). To reduce the cognitive effort the information representation should be more spatial than symbolic. Middle-quality reconciliation can be categorized as spatial, thus won’t fit the task. This results that individual makes insufficient adjustments resulting in more biased performance evaluation

H2A: Press release with low-quality reconciliation will have higher performance evaluation than press releases with only GAAP

(18)

(Kruger, 1999; Kelton & Pennington, 2012). Furthermore, the results of Hogan, Krishnamoorthy and Maroney (2017) indicates that the nature of details of reconciliations increases the performance evaluation. These findings suggest upwards effect on the performance evaluation because of the less accuracy in the decision-making process and an increase in detailed information about the non-GAAP measures:

H2B: Press release with middle-quality reconciliation have higher performance evaluation than press releases with low-quality reconciliation.

High-quality reconciliation provides a complete reconciliation in a table. This representation is more spatial than symbolic, and thus, decreases the cognitive effort (Vessey, 1991) resulting in higher accuracy in the performance evaluation (Vessey, 1991; 1994). Young (2014) argue that providing more insight in the relation between the non-GAAP and the GAAP mitigates the upwards effect in the performance evaluation. The high-quality reconciliation provides the most insight in non-GAAP measures. However, high-quality reconciliation represents a greater detail about the relation between the non-GAAP measure and comparable GAAP measure and provides insight in the trends of these measures. Hogan et al. (2017) indicates an increase in performance evaluation because of the greater nature of details in the reconciliation. These theories are contradictory and thus I think that there will be no differences in the performance evaluation:

H2C: Press release with high-quality reconciliation have an equal performance evaluation than press releases with middle-quality reconciliation.

(19)

Figure 2 – Summary of hypothesis 2 – This figure presents the predicted trend of the performance evaluation.

2.5.3 H3: information weighting

According to the decision-making model of Maines and McDaniel (2000) individuals weight information before they make an investment decision. Jennings (1987) find that the weighting of information is dependent on the perceived creditability of the information. The creditability is affected by the content of the disclosures. Positive news is perceived less creditable than negative news (Mercer, 2004). Furthermore, investors overweight private management information and underweight public signals (Daniel, Hirshleifer and Subrahmanyam, 1998). Non-GAAP measures are private management disclosures, but the content is mostly more positive than the GAAP measures (Bhattacharya et al., 2004). This suggest that professional investors find the non-GAAP measures less creditable, but, they weight non-non-GAAP measures more than the non-GAAP measures. This expectation is confirmed by Reimsbach (2014), who find an increase in the weighting of the performance evaluation:

H3A: Non-professional investors place more weight on the performance evaluation if the press release contains low-quality reconciliation than press release with only GAAP measures

Hutton, Miller and Skinner (2003) indicates that the presences of management explanations increase the perceived creditability of voluntary disclosures. Reconciliations

(20)

between non-GAAP and GAAP can be a way of signalling the creditability of the voluntary disclosures. The results of Marques (2006) and Black, Black, Christensen and Heninger (2012) indicates that because of the presence of reconciliation after the SEC intervention investors rely more on non-GAAP measures. This suggest that non-GAAP measures are more creditable with middle-quality reconciliation, which results in an increase in weighting of the performance evaluation:

H3B: Non-professional investors place more weight on the performance evaluation if the press release contains middle-quality reconciliation than press release with low-quality reconciliation.

Furthermore, the results of Hogan et al. (2017) indicates that summarized and isolated information increases the weighting non-professionals have on this information. Middle quality reconciliation is more summarized and isolated compared to high quality reconciliation. This suggest an decrease in information weighting of the performance evaluation. I predict that the reconciliation quality 3 have a mitigating effect on the weighting of performance evaluation.

H3C: Non-professional investors place less weight on the performance evaluation if the press release contains high-quality reconciliation than press release with middle-quality reconciliation.

(21)

Figure 3 - Summary of hypothesis 3 – This figure presents the predicted trend of the weighting performance evaluation.

2.5.4 H4: investment valuation

According to the model of Maines and McDaniel (2000) the investment valuation is a product of the information evaluation and information weighting. Non-professional investors are biased by the presence of non-GAAP measures (Anderson & Hellman, 2007), even when non-professional investors find these measures uninformative (Chapman & Johnson, 2000). Therefore, I expect that the presence of non-GAAP measures increases the performance evaluation. Moreover, I expect an increase in the weighting of the performance evaluation because investors overweight private management disclosures (Daniel et al., 1998). Prior research (Frederickson & Miller, 2004; Elliott, 2006; Dilla et al. 2013; Reimsbach, 2014) find an increase in the investment valuation if non-GAAP measures are present.

H4A: Press releases with low-quality reconciliation does have a higher investment valuation than press releases with only GAAP measures.

Middle quality reconciliation includes the magnitudes of the non-GAAP measures. I expect an increase in the performance evaluation. Research indicates that providing more details about the measures increase the performance evaluation (Hogan et al., 2017). Further, I expect that the cognitive effort of integrating this information is high which results in less accurate

(22)

decisions (Vessey, 1991; Vessey & Galleta, 1991), resulting in a high-performance evaluation. Moreover, I expected an increase in the weighting of the performance evaluation of the increase in creditability of these measures. Results of prior studies (Hutton et al., 2003; Marques, 2006; Black et al., 2012) indicate that manager’s explanation of their private disclosures increases the creditability. An increase in performance evaluation and increase in weighting results in an increase investment valuation.

H4B: Press releases with middle-quality reconciliation does have higher investment valuation than press releases with low-quality reconciliation.

The representation of high quality reconciliation reduces the cognitive effort resulting in a higher accuracy of the judgements and thus a decrease in the performance evaluation (Vessey, 1991; Vessey & Galletta, 1991). Providing more details increases the performance evaluation (Hogan et al., 2017). Therefore, I expected nor an increase or decrease in performance evaluation. Further, I expected a decrease in weighting of the performance evaluation. Providing more insight in the relationship between the non-GAAP measures and GAAP measures decreases the weighting of the performance valuation (Hogan et al., 2017). This result in a decrease in the performance evaluation. Moreover, I expect that high-quality reconciliation mitigates the upwards effect cause by non-GAAP measures:

H4C: Press releases with high-quality reconciliation has lower investment valuation than press releases with middle-quality reconciliation.

H4D: Press release with high-quality reconciliation does not differ in investment valuation press release with only GAAP.

(23)

(24)

3 Research method

To answer the research question, I set-up an experiment to test the hypotheses. The design of my research is based on prior research, such as, Frederickson and Miller (2004), Elliott (2006) and Dilla et al. (2013). In short, participants will get a case about a hypothetical firm. Participants are asked to make multiple investment-related judgements and answers multiple series of questions about their judgements about the earnings announcements.

3.1 Participants and task

Non-professional investors are heterogonous, which makes it difficult to identify a single group that is representative for all non-professional investors. Prior research uses MBA-student as a proxy for non-professional investors (Frederickson & Miller, 2004; Elliott, 2006, Dilla et el, 2013, Reimsbach, 2014). Moreover, Elliott, Hodge, Kennedy and Pronk (2007) predict and find that using MBA-students as a proxy for non-professional investors is a valid methodology choice. I have limited access to MBA-students, thus, adjusted my proxy group to economic and business-students and economic and business-graduates. My participants are required from the different universities in Amsterdam, a big-four accounting firm and Prolific.ac. The participants from Prolific.ac (n=89) received 0,90 GBP for their response1_.

In total 157 people finished the experiment between 24th_{of April and the 24}th_{of May. 16}

people are excluded because they had an education outside field of economics and business. 23 people are excluded because they do not have the required educational level. Resulting in 118 subjects in the study (table 2, panel A). The deceptive statistics of the participants are in table 2 (panel B). The participants have an average age of 27 and a mean of 24. Of my participants 71,2% is male and 28,8 is female. 44,9% of the participants have a HBO-bachelor degree, 29,7 have a WO-bachelor degree, 21,2 % have a WO-master degree, and last 4,2% have a post-master degree. Last I ask my participants to indicate how many years of investment experience they have. 39,8% have no investment experience, 21,2% have less than one year experience, 28,8% have between one and three years of investment experiences and 10,2% have more than 3 years’ investment experience. To mitigate the risk that my results are biased by unequal deviation, I

1_{Because of the experimental set-up, the distribution channels are not reported. Therefore, I assume that everyone}

(25)

two-are not significantly unequally divided over the groups2_.

Table 2 – Overview of the sample selection and the descriptive statistics of the participants – Panel A reports the sample selection. The first column presents the participants characteristics for the total sample selection. The second column reports the characteristics of participants who correctly answer the manipulation check.

Panel A: Sample selection

n %

Completed the experiment 157 100%

Excluded:

None-economic and business participants -16 -10,2% No HBO-bachelor degree or higher -22 -14,0%

A PhD-degree or higher -1 -0,7%

Total respondents 118 75,1%

Panel B: Descriptive statistics of the participants

Total sample After the manipulation checkA (n=118) (n=87) Gender Male 84 71,2% 60 69,0% Female 34 28,8% 27 31,0% Education

HBO bachelor degree 53 44,9% 37 42,5%

WO bachelor degree 35 29,7% 26 29,9%

WO master degree 25 21,2% 23 26,4%

Post-master degree 5 4,2% 1 1,1%

Investment experience

No investment experience 47 39,8% 32 36,8%

Less than one year 25 21,2% 20 23,0%

Between one and three years 34 28,8% 24 27,6%

More than one year 12 10,2% 11 12,6%

Participants type

Unpaid participants 29 24,6% 25 28.7%

Paid participants 89 75,4% 62 71,3%

A_{– Maines and McDaniel (2000) assume that investors have to acquire the information before this can have an impact in the} investment valuation. I exclude the participants who fail the manipulation check if I test the information evaluation (H2), the information weighting (H3) and the investment valuation (H4).

All participants receives a press release about a hypothetical company called Titian Technology Inc. – hereafter Titian - which is based on a game-developing company that trades on the NASDAQ. The press release contains background information about the firm and financial information about the fourth-quarter 2016 and the final year earnings announcements.

2_{A Pearson Chi-squared test is conducted if all groups have estimation above 5, otherwise the Fischer’s Exact test is}

conducted (Field, 2014). A Pearson Chi-squared test (χ2_{=2,538; p=0,468) indicates that gender is not unequally}

divided over the conditions. The Fischer’s Exact test (χ2_{=10,484; p=0,270) indicates that the education levels are} not unequally divided, and the Fischer’s Exact test (χ2_{=4,244; p=0,911) indicates that the investment experience is} equal over the groups. These results indicate that none of the participants characteristics are not unequally divided over the experimental conditions.

(26)

The participants are randomly assigned over the experimental conditions. The provided press release is manipulated per experimental condition. Participants then are asked to make multiple investment-related judgments and answer question about the perception of the relevance, reliability and the usefulness of the information.

3.2 Experimental design and press release

The experiment contains four different conditions. The first condition is the control group, and is added to examine whether there is a mitigating effect through the presence of the reconciliation (GAAP). For the other conditions, I include the non-GAAP measures and the manipulation the quality of the reconciliation. I exclude the first reconciliation quality of Hitz (2010), because prior experimental studies always provide limited background about the non-GAAP measures comparable the low-quality reconciliation in this study. Furthermore, providing no disclosures is limited used in practice (Aubert & Grudnitski, 2014).

The press release contains two GAAP measures – a GAAP revenue and non-GAAP income measure. I modified one of the non-non-GAAP measures, because this mitigates the risk that differences in judgement can be attributed to differences in information availability (Elliott, 2006). The non-GAAP revenue measure is adjusted for the change in deferred revenue. The non-GAAP income measures is adjusted for provision for income tax, other income expenses, interest income, depreciation and amortization, impairment of intangible assets and stock-based expenses. The non-GAAP income measure includes the adjustment of the change in deferred revenue.

Table 3 – experimental conditions – This table presents the four experimental conditions of this study. The second column presents the used names in this experiment. The third column presents the information in the press release which is manipulated. Group

number Research group Received information

1 GAAP GAAP measures.

2 Non-GAAP/Low GAAP measures, non-GAAP measures and low-quality reconciliation.

3 Non-GAAP/Middle GAAP measures, non-GAAP measures and middle-quality reconciliation.

4 Non-GAAP/High GAAP measures, non-GAAP measures and high-quality reconciliation.

Consistent with prior research (Frederickson & Miller, 2004; Elliott, 2006, Dilla et al., 2013; Dilla et al, 2014; Reimsbach, 2014), the press release follows the typical and standardized pattern and consistent out of three sections (appendix 12). The press releases begins with a headline that stated ‘’A net loss of 108,2 million over the year 2016’’, for the control group. For

(27)

income of €68.8 million over 2016’’. Consistent with prior research (Frederickson & Miller, 2004: Elliott, 2006: Reimsbach, 2014), I emphasize the use of non-GAAP measure through the headlines. Consistent with empirical evidence (Bhattacharya et al., 2004) the non-GAAP measures exceeded the comparable GAAP measures and consistent with Dilla et al. (2013), Titian has a GAAP loss and a non-GAAP profit.

The second section of the press release contains a narrative that stated current and comparative amounts of the quarterly and annual earnings of the reporting year including non-GAAP measure dependent on the condition (appendix 9). The narratives also include forward looking information and information about the share repurchase. The narratives include the reconciliation between GAAP and non-GAAP, which varies with the condition (appendix 10). The first condition receives the lowest level of reconciliation that contains the adjusted account names without magnitudes (non-GAAP/low). The second condition contains the adjusted account names with the magnitudes (GAAP/middle). The third condition contains a non-GAAP income statement (non-non-GAAP/high). The third section contains the financial statements of the firm (appendix 12), including the GAAP income statement, GAAP balance sheet and the GAAP cash-flow statement.

3.3 Independent and dependent variables

The experimental design contains two manipulation (independent) variables: presence of non-GAAP and the reconciliation quality. These independent variables vary in the press release per experimental conditions. After the press releases the participants receive a questionnaire (appendix 11) that is similar to prior research (Elliott, 2006; Reimsbach, 2014). Because of the experimental set-sup the participants could not go back to the press release after they receives the questionnaire, consistently with Elliott (2006).

In the first section of the survey, the participants have to make multiple investment-related judgments. Participants are asked to indicate on a 1-to-7 Likert-scale, to judge about the earnings performance of the reporting year [Performance], and indicate how they judge about the earnings potential for the next two years [Potential]. These questions are proxies for the performance evaluation of the participants (Frederickson & Miller, 2004; Reimsbach, 2014; Hogan et al., 2017).

I ask four investment related question as indicators for the investment decision. I ask the participants to indicate the attractiveness [Attractiveness] and the overall favorableness of the

(28)

Titian [Favorableness] (Reimsbach, 2014) on a 7-point Likert-scale. The third indicators is how they value the Titian’s shares [stock] (Reimsbach, 2014). My participants could response on a 11-point Likert-scale between 20 EUR and 30 EUR, with steps of 1 EUR. Last, I ask my participants to indicate how much they would invest in the firm [investment] (Elliott, 2006; Hogan et al., 2017), they could response on a 11-point Likert-scale between 0 and 5.000 EUR with steps of 500 EUR. For both questions, I provide a benchmark of 25 EUR per share. Reimsbach (2014) has a benchmark of 2,50 USD and Elliott (2006) of 2 USD. My benchmark is 25 EUR, because low fair value may signal low business performance.

In the second section of the questionnaire, I ask the participants to indicate their perception of the non-GAAP measures. I ask how relevant [Relevant] and reliable [Reliable] the participants believe the non-GAAP measures are (Frederickson & Miller, Reimsbach, 2014). They could response on a 7-point Likert-scale between not at all relevant/reliable and very relevant/reliable. Last, I ask the participants to indicate how useful the press release is for indicating the managerial motives for this information [useful]. For these question, they could response on a 7-point Likert-scale. These questions might provide additional insight in how reconciliation quality the investment decision. The last set of questions, contains questions which are used to collect the characteristics of the participants.

3.4 Manipulation check

To check if my manipulation variables are sufficient, I ask the participants to indicate which information components the press release contains. The participants could choose out of GAAP measures, non-GAAP measure, financial highlights, forward looking information and a table of reconciliation (Frederickson & Miller, 2004; Elliott, 2006). I provide no brief explanation about GAAP and non-GAAP measures. However, the study of Reimsbach (2014) provides brief explanation about GAAP and non-GAAP measures. Participants gave feedback (personal communication, 24 April 2017) and said that they don’t understand the meaning of non-GAAP and GAAP measures. I adjust my criteria to an easier level, because of these the feedback and Reimsbach (2014) and to increase the group sizes. However, this can have an effect on my results, and therefore act as a limitation for this study.

Participants of the GAAP condition answer the question correctly if they indicate that their press release won’t include non-GAAP measures. Participants of the other conditions correctly answer the question if they indicate that that their press release contains non-GAAP measures. For these conditions, I ask the participants to indicate which measure exceeds, the

(29)

total, 26,3% failed these manipulation checks. A Pearson Chi-squared test indicates that the educational level (χ2_{=11,332; p=0,008) and if the participants get paid (χ}2_{=3,091; p=0,093)}

affects whether they correctly answer the manipulation check questions. The educational level might be dependent whether they correctly answer the manipulation checks. Furthermore, the participants who get paid the manipulation checks are used to determine if they reive their compensation. In common practice, some of the paid-participants may not be that cautious during the experiment.

Before the manipulation check, I had group sizes between the 28 and 31, however, because of the model of Maines and McDaniel (2000) I exclude the participants who fail the manipulation check. Thus, after the manipulation check I have group sizes between 18 and 25. Prior research (Maines & McDaniel, 2000; Frederickson & Miller, 2004, Dilla et al., 2013; Reimsbach, 2014) have similar group sizes, however, these small group sizes can affect my results (Field, 2014), thus this act as an limitation for this study. Because of the exclusion, again, I check whether I find significant unequal participants characteristics between groups. I find no significant results indicating that the participants characteristics are unequally divided over the groups3_{. This result reduce the risk that results are unexpected affected the characteristics of the}

participants

3_{A Pearson Chi-squared test (χ}2_{=2,788; p=0,438) indicates that gender is not unequally divided over the conditions.}

The Fischer’s Exact test (χ2_{=9,111; p=0,400) indicates that the education levels are not unequally divided, and the} Fischer’s Exact test (χ2_{=5,320; p=0,828) indicates that the investment experience is equally divided. A Pearson} (χ2_{=0,693; p=0,873) indicates that the paid-participants are equally divided over the groups. These results indicate}

(30)

4 Results

In this section I examine the data through testing the hypotheses. This study follows the research method of Maines & McDaniel (2000), because we examine the impact of reconciliation quality according to their decision-making model. To test the first hypothesis, I conduct a Chi-squared test. I test the second hypothesis using non-parametric test and analyse of variance. Thirdly, I conduct multiple regression analysis to test the weighting of the performance evaluation. Lastly, I perform non-parametric tests and multiple analyse of variance (ANOVA) to test the fourth hypothesis.

4.1 Testing H1 comparison of the information acquisition

Hypothesis 1 predicts that non-professional investors will acquire the investment related information regardless of quality of the reconciliation. Consistent with the approach of Maines & McDaniel (2000) and prior research (Frederickson & Miller, 2004; Reimsbach, 2014), I assume that participants who succeed the manipulation checks correctly acquire the performance measures. Table 3 (Panel A) presents the number of participants and the number of participants who acquire the information correctly separated per group.

Table 4 – Information acquisition – Panel A presents the number of participants per group and the number of participants who acquire the non-GAAP information. Panel B presents the results of the Pearson Chi-squared test.

Panel A: Descriptive statistics

GAAP Non-GAAP /low Non-GAAP /middle Non-GAAP /high Total

Number of participants 28 30 29 31 118 Number of participants acquiring the informationA 18 (64,3%) 21 (70,0%) 25 (86,2%) 23 (74,2%) 87 (73,7%) Panel B: Test of independence between information acquisition and experimental condition Pearson Chi-squared 3,839

Degrees of freedom 3

P-valueB _0,279

* indicates a significant level of 0,10; ** indicates a significant level of 0,05; *** indicates a significant level of 0,01

A_{– I ask the participants to indicate whether they receive GAAP measures, non-GAAP measures, financial highlights, forward} looking information and table of reconciliation. Participants correctly answer this check if they indicate whether they receive non-GAAP measures or not. Secondly, I ask the participants of the conditions with non-GAAP information to indicate what the highest measures was. If one of these questions are correct, then I assume that the participant acquired the information. B _{– The P-value is two-tailed}

A Pearson Chi-squared test (χ2_{=3,839; p-value=0,279) indicates that information}

acquisition is not unequally divided over the groups. This result is consistent with prior research (Maines & McDaniel, 2000; Frederickson & Miller, 2004; Elliott, 2006; Reimsbach, 2014). This

(31)

non-professional investors, as predicted by H1. However, non-significant results won’t indicate that there are no differences between in the information acquisition between the groups (Field, 2014).

All participants who failed to acquire the performance measures (n=31) are removed from this study. Maines and McDaniel (2000) assume that information must be acquired before individuals evaluate and weight information. Only if information is acquired, then this can have an effect on the final investment valuation, this approach is consistent with prior research (Maines and McDaniel, 2000; Frederickson & Miller, 2004; Reimsbach, 2014).

4.2 Testing H2: information evaluation

Hypothesis 2 predicts that the GAAP conditions have the lowest performance evaluation and that non-GAAP/middle and non-GAAP/high have the highest performance evaluation. To measure the information evaluation, I combine the performance of current year [Performance] and the earnings potential of the coming two years [Potential]. A Cronbach’s Alpha score of 0,742 indicates that both measures correlate highly and therefore, consistent with prior research (Elliott, 2006; Reimsbach, 2014; Hogan et al. 2017), these measures can be combined. Performance Evaluation Reconciliation Quality Level [PERQ] is the proxy for performance evaluation.

[𝑃𝐸𝑅𝑄𝐿] = ([𝑃𝑒𝑟𝑓𝑜𝑟𝑚𝑎𝑛𝑐𝑒] + [𝑃𝑜𝑡𝑒𝑛𝑡𝑖𝑎𝑙]

2 )

(1)

There are no outliers in the data, as assessed by inspection of a boxplot. For comparison of the mean, any data points that are more than 1.5 box-lengths from the edge of the boxplot are classified as an outlier. The Levene’s test (p=0,491) indicates homogeneity of variance between the groups. I transform my data because the Shapiro-Wilk test (p<0,05) indicates that the data is not normally distributed. I transform the data through the square root of the highest value plus one minus the actual value (Laerd Statistics, 2015b). These results indicate no violation of assumptions of the ANOVA. Regardless of the ANOVA results, I conduct a Fisher’s Least Significant Differences (LSD) for specific testing of the hypotheses. A post-hoc LSD test is primarily a set of individual t-test (Field, 2014). Hsu (in Laerd Statistics, 2015b) argues that a

(32)

post-hoc LSD test can be conducted regardless of the ANOVA results. Furthermore, transforming data increases Type I errors, compared to the consequences of analysing the untransformed data (Flied, 2014), to mitigate this risk, I perform non-parametric tests with the untransformed data. For the Jonckheere-Terpstra test I reorder the groups in (1) GAAP, (2) non-GAAP/high, (3) non-GAAP/low and (4) non-GAAP/middle, because this test only test for linear ordering.

Table 5 – Comparison of the information evaluation: performance evaluation – Panel A reports the descriptive statistics of the performance evaluation. Panel B presents the results of the analyse of variance (ANOVA). Panel C reports the outcome of the non-parametric tests.

Panel A: Descriptive statistics (mean, [standard deviation])

n GAAP N Non-GAAP /low n Non-GAAP /Middle n Non-GAAP /low

PERQL 18 3,861

[1,391] 21 3,929 [1,076] 25 4,040 [1,127] 23 4,196 [1,346] Panel B: Test of comparison of the mean (ANOVA)B

Sum of

squares Df Mean square F-statistics P-value

C

Between groups 0,125 3 0,042 0,377 0,770

Within groups 9,203 83 0,111

Total 9,328 86

Panel C: Test of comparison of the mean (non-parametric test) P-valueC

Independent sample Kruskal-Wallis test 0,874 Independent sample Jonckheere-Terpstra test 0,437 Panel D: Post-hoc LSD test

GAAP Non-GAAP

/low Non-GAAP /low Non-GAAP /low

GAAP 0,962 0,701 0,364

Non-GAAP/low 0,962 0,726 0,360

Non-GAAP/middle 0,701 0,726 0,563

Non-GAAP/high 0,365 0,370 0,563

* indicates a significant level of 0,10; ** indicates a significant level of 0,05; *** indicates a significant level of 0,01

A _{– I asked the participants to indicate their judgement about the current year earnings performance [Performance] on a} 7-point Likert-scale between very weak (1) and very strong (7). Secondly, I ask to indicate their earnings potential for the two coming years [Potential] on a 7-point Likert-scale between very weak (1) and very strong (7). A Cronbach’s Alpha score of 0,742 indicates that these measures have high internally consistently.

B_{– Shapiro-Wilk test (p<0,05) indicates that GAAP condition is not normally distributed. After the transformation of the} data the Shapiro-Wilk test (p>0,05) indicates that all conditions are normally distributed. The Levene’s test (p=0,491) indicates homogeneity of variance between the groups.

C _{– P-values are two-tailed}

The descriptive statistics are in table 5 (panel A) and it increases from GAAP (mean=3,861; SD=1,391) to non-GAAP/low (mean=3,929; SD=1,076), to non-GAAP/middle (mean=4,040; SD=1,127), to non-GAAP/high (mean=4,196; SD=1,346). The results of the one-way ANOVA (F=0,377; p=0,770) indicate that the differences between the groups are not significant. The Kruskal-Wallis test (p=0,874) also indicates no significant differences between

The effects of reconciliation quality on the investment decision of non-professional investors