The value relevance of non-GAAP performance information as published by a firm’s management

(1)

Amsterdam Business School

The value relevance of non-GAAP performance information as published by a firm’s management.

Final Version Name: Erik Verduin Student number: 5613310 Date: 7 October 2018

First supervisor: Dr. S.W. Bissessur Second supervisor: drs. J.F. Jullens

Executive Master of Finance and Control (‘EMFC’)

Faculty of Economics and Business, University of Amsterdam Number of words (excl. appendices): 12.735

(2)

Statement of originality

This document is written by Erik Verduin, who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document are 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.

Verklaring eigen werk

Hierbij verklaar ik, Erik Verduin, dat ik deze scriptie zelf geschreven heb en dat ik de volledige verantwoordelijkheid op me neem voor de inhoud ervan.

Ik bevestig dat de tekst en het werk dat in deze scriptie gepresenteerd wordt origineel is en dat ik geen gebruik heb gemaakt van andere bronnen dan die welke in de tekst en in de referenties worden genoemd. De Faculteit Economie en Bedrijfskunde is alleen verantwoordelijk voor de begeleiding tot het inleveren van

(3)

Abstract

The value relevance of non-GAAP performance measures is the subject of the research performed in this thesis. Prior research has been performed on this subject. However, this is the first known research to use the non-GAAP earnings per share as published by managers, by using a recent hand collected data set provided by Bentley et al. (2018). Their data set comprise data of North-American firms from 2003 to 2015. Prior researched by among others Albring (2010) and Brown and Sivakumar (2003) use other, not directly prepared by a firm’s management, non-GAAP performance metrics in their research. The final data set used in my research consist of 38.541 firm years.

The research has been performed by using the Ohlson-model for value relevance (Ohlson, 1995). The regression model has been modified based on a research by Balachandran and Mohanram (2011) to control for losses and inter-industry impacts. This is the first known non-GAAP value relevance research to do so. The outcome of this research is that non-GAAP performance information as published by managers is value relevant for investors and often has a stronger relation with the market value of a firm than GAAP

performance information has. My research also shows that the value relevance of non-GAAP performance information is not uniform across industries and is also different for firms reporting a profit or a loss. The contribution of this research on controlling literature is that I have used non-GAAP performance information as published by managers. This non-GAAP performance information is the result of the finance processes within a firm and are meant to provide users of the financial reporting provide with information which add to the mandatory reported GAAP financials. This to mitigate the information asymmetry from the agency theory (Jensen and Meckling, 1976). Prior research by among other Lev and Zarowin (1999), Francis and Schipper (1999) and Belachandran and Mohanram (2011) show that GAAP financials have lost relevance and usefulness. One way manager can adept to this is by publishing non-GAAP earnings, which are closer to the ‘core earnings’ of a firm than GAAP earnings are (Heflin and Hsu, 2008).

Key words: GAAP, Non-GAAP, earnings, usefulness, value relevance, Fama-French, controlling, financial accounting, investors.

(4)

1. Introduction

1.1. Introduction of the subject

The purpose of this thesis is to determine the usefulness of non-GAAP (‘Generally Accepted Accounting Principles’) performance measures by investigating the value relevance of non-GAAP disclosures. There is a distinction between GAAP performance measures, such as accounting revenue and accounting earnings, and non-GAAP performance measures, such as earnings before interest, taxes, depreciation and amortization (EBITDA), adjusted EBITDA, adjusted earnings per share, constant currency earnings and free cash flows (PwC, 2016). This distinction exists because GAAP performance measures are not always considered to be the optimal representation of a firm’s performance and condition. Dichev et al. (2014) explain that Chief Financial Officers (‘CFOs’) believe that high quality earnings are sustainable and repeatable. Their research also explains that GAAP impacts earnings quality as it restricts managers to use reporting discretion to present sustainable and repeatable earnings, cleaned for e.g. one-off items. CFOs interviewed for their research indicate that GAAP financial reporting has turned into a compliance exercise instead of a tool to provide the best information to constituents. Disclosing non-GAAP performance information together with mandatory GAAP reporting can solve this problem.

Black et al. (2017) explain that non-GAAP information is sometimes used to influence stakeholder

perceptions. Although managers primarily use (real) earnings management within GAAP reporting, they also use non-GAAP reporting to be able to influence perceptions and present desired performance levels. The research shows that firms are more likely to disclose non-GAAP information if they not meet analyst

forecasts. Prior researches on this subject, among others by Doyle (2013), showed the same. Doyle explains that the market is aware of this and discounts positive earnings surprises, although Bhattacharya et al. (2003) find that analysts are more skeptical and discount more than investors.

The different purposes and uses of reporting non-GAAP information make it a relevant subject for a research. Therefore this research investigates the value relevance of non-GAAP performance information, especially using the information as it is published by managers.

In recent years, favorable macroeconomic market conditions in combination with low interest rates, led to an increase in mergers and acquisitions (FD, 5 December 2017). In the process around a merger or an acquisition, (potential) investors are looking for the best information to value the (potential) acquisition target. My motivation for this research comes from this search for the best information by investors as I am working in Transaction Services where I advise my clients in the deal process. The outcomes of this research will help me, and other professionals working in Transaction Services, to be able to further enhance the quality of our advises.

(7)

There is a variety of information available for investors to base their investment decisions on. Financial statements as published on a yearly or quarterly basis by (large) firms are an important source for financial information. The information in financial statements is required to adhere to Generally Accepted Accounting Principles (e.g. IFRS for international firms, and US GAAP for US firms). The requirement to adhere to GAAP is for comparability purposes and financial statements are usually audited by an auditor1_{. One of the main}

purposes of publishing financial statements is for investors to base their investments decisions on (Scott, 2014). However, Lev and Zarowin (1999) conclude in their research that the usefulness of accounting earnings, cash flows and book values, including equity, has declined over the twenty years prior to the research. A comparable research by Francis and Schipper (1999) comes to the same conclusion. Therefore, investors have to look for other sources of information which are more useful to base their investment decisions on.

To ultimately value a business, investors usually use the discounted cash flow valuation (hereafter: ‘DCF’). Disadvantage of using the DCF-method to value a business is that it is not easy to use and it is sensitive to a lot of assumptions (Lie and Lie, 2002). As a consequence, investors often use valuation by multiples, such as a price/earnings multiple instead or supplementary to the DCF valuation (Lie and Lie, 2002). The earnings multiples are frequently based on adjusted (non-GAAP) earnings.

Kaplan and Ruback found in their research that the DCF valuation performed at least as well as using the multiple approach but they also find that the multiple approach is useful, especially when combined with the DCF method (1995). Note that by using the DCF-method, non-GAAP performance measures are also commonly used as a starting point for the calculation of a firm’s value (CFA, 2015).

Prior research has been performed on the value relevance of non-GAAP performance measures. For instance, Brown and Sivakumar (2003) conclude in their research that non-GAAP operating income is more value relevant than GAAP operating income is. They have compared the value relevance of GAAP and non-GAAP operating income by assessing (1) the ability of predicting future earnings, (2) the association of earnings levels with stock prices and (3) the correlation of earnings surprises with abnormal stock returns, which they call the ‘information content’. Albring et al. (2010) also conclude that the non-GAAP performance measures they have included in their research have a higher value relevance than traditional GAAP performance measures. However, they have used the S&P core earnings metric as non-GAAP measure, which is a more explicitly defined metric than other non-GAAP performance measures which usually have more classificatory discretion.

This research contributes to existing literature by using a recent, hand collected data set including non-GAAP adjustments published by managers. The researches by Brown and Sivakumar (2003) and Albring (2010) use data up to 2004 and are not using non-GAAP information as published by the people managing a firm. They have used analyst forecast data or the S&P core earnings metric as non-GAAP performance information. Also, this research controls for loss firms and industry impacts and is the first known research on non-GAAP value relevance to do so.

(8)

The specific contribution to controlling literature is that this research investigates if investors value the non-GAAP performance information as used and published by managers. This research shows that investors do value non-GAAP adjustments as published by managers. Executives, controllers and other finance

professionals can use this outcome to improve their external reporting in order to mitigate the agency problem of information asymmetry (Jensen and Meckling, 1976).

1.2. Research question

The objective of this thesis is to examine the value relevance of non-GAAP performance measures. The following research question has been formulated:

Are non-GAAP performance measures value relevant for investors to value a business and do they contribute to GAAP information as included in financial statements?

1.3. Research methodology

In this thesis, a combination of publicly available information and a recent data set provided by Bentley et al. (2018), containing non-GAAP performance information, is used. All the data used relates to public firms in the United States. The Warton Research Data Service (‘WRDS’) is used to collect all data used in this research except for the non-GAAP performance measure data. Within the WRDS database, COMPUSTAT is used to collect the financial information of the firms included in the sample. The final sample comprises data of public companies in the United States and consist of 38,541 firm years in the period 2003 to 2015. The model of Ohlson (1995) will be used to assess the value relevance of non-GAAP performance measures. The Ohlson model is commonly used and widely recognized model to analyze the value relevance of financial information (Barth et al., 2001).

The non-GAAP performance measure in this thesis is the reported non-GAAP earnings per share as hand collected by Bentley et al. (2018) from form 8-K filings in the United States.

Form 8-K filings are a mandatory report companies must file with the SEC in case of material events that the SEC and shareholders should know about.

(9)

1.4. Contribution of the research

The research performed in this thesis contributes to existing literature in several ways. At first, in this research hand collected data is used which is provided by Bentley et al. (2018) as part of their research. They have prepared a publicly available data set comprising non-GAAP earnings of listed companies in the United States. Traditionally, research related to non-GAAP performance measures, such as the research of Brown and Sivakumar (2003), uses analyst forecast data which are provided by so called forecast data providers (hereafter: ‘FDPs’) or other sources such as the S&P core earnings measure (Albring et al., 2010). This thesis is the first known research that investigates the value relevance of the non-GAAP earnings as collect by Bentley et al. (2018) by using the Ohlson-model.

Secondly, this research contributes by using a recent data set, covering the years 2003 to 2015. The earlier mentioned research performed by Brown and Sivakumar (2003) is based on a data set with a sample period that ends in 1997, which is already more than twenty years ago at the moment this thesis was written. The research of Albring et al. (2010) uses a sample period ending in 2004.

Third, information regarding differences in value relevance of financial information between industries is provided. Among others Lev and Zarowin (1999) and Francis and Schipper (1999) already provide information that multiple factors impact value relevance of financial information to be different among industries, for instance driven by the level of ‘high-tech’ a business is. For managers, it is important to understand the effect of industry heterogeneity in their investment decision-making.

The fourth contribution relates to loss making firms. Lueng and Veenman (2018) explain that the relevance of GAAP earnings published by loss making firms is lower than the relevance of earnings of profitable companies. This is due to the fact that this loss in some cases is driven by non-recurring expenses which are not a good predictor for the future performance of firms. Therefore by excluding these items, and

presenting non-GAAP performance information, the relevance of the published information is increasing. This research provides further research on the value relevance of non-GAAP performance information for loss making firms. For investors, it is important to understand the effect of accounting losses in their investment decision-making.

Not only the impact of firms being loss making is included in this research, but the regression model has also been adjusted to control for the impact of industry differences in accordance with the approach of

Balachandran and Mohanram (2011). This research is the first known research investigating the value relevance of non-GAAP performance information to do so.

(10)

Finally, this research also contributes to the field of financial accounting. This research provides additional information for standard setters with respect to non-GAAP disclosures. The outcomes of the research provide information which standard setters, such as the FASB and the IASB, can use for setting standards regarding non-GAAP performance information.

It also contributes by showing the added value of non-GAAP performance information to users of financial statements and therefore the importance for standard setters to include this non-GAAP information.

1.5. Structure of the research

This research will be structured as follows. First, I start with a discussion of prior literature. This literature review is the second section of this thesis, directly following the introduction. In the literature review, prior literature on value relevance, decision usefulness and non-GAAP performance measures is set-out and discussed. In the third section the hypotheses tested in this thesis will be described. In the following forth section, the empirical research will be discussed, describing the research method, data and sample selection and the variables used. Results from the empirical research are presented in the fifth section. This includes the descriptive statistics and correlation and regression analyses. In the sixth section, 3 additional tests are performed to test the robustness of the model used in the empirical research. In the final sixth section the results will be analyzed in order to form an answers to the research question. Furthermore, suggestions for further research and a description of limitations will be set out in the sixth section. The outcome of regressions ran by industry and the outcomes of the robustness tests are included in the appendix.

(11)

2. Review of prior literature

In this section of the thesis, the prior research on value relevance, decision usefulness and non-GAAP performance measures is reviewed and discussed.

2.1. Value relevance and decision usefulness

As already being mentioned in the first section of this thesis, value relevance is the dependent variable in this research. In this paragraph, a further description of value relevance is presented on the basis of prior literature. Furthermore, the model used to determine value relevance is introduced. This model, introduced by Ohlson more than twenty years ago, is widely recognized as model to analyze value relevance (Barth et al., 2001).

Value relevance can be descripted as the relation between the accounting value of a firm, which is the value according to the (audited) financials statements of a firm, and the market value of a firm (Barth et al., 2001). The market value of a firm is in this case measured by the (net) value of stocks (Ohlson, 1995). Barth et al. (2001) state that an accounting amount is value relevant if it has a predicted association with the market values.

Research on value relevance date back to the 1960’s, when Nobel Prize winning authors Modigliani and Miller (1966) investigated the association between accounting values and market values of a firm. This can be seen as the basis for today’s research on value relevance as these two variables are still the core of the value relevance model as provided by Ohlson (1995). The first known paper to introduce the current concept of value relevance is performed by Amir (1993). Amir investigates in his research whether the estimating the present value of postretirement benefits and pensions is value relevant for investors.

Scott (2014) explains that one of the main purposes of financial report is to provide information to investors to base their investment decisions on. This is in line with the purpose of the Financial Accounting Standards Board (hereafter: FASB). In their Statements of Financial Concept No. 8, dated September 2010, they explain the conceptual framework for financial reporting, starting with the objective and usefulness. There the following description is included (FASB, 2010):

The objective of general purpose financial reporting is to provide financial information about the reporting entity that is useful to existing and potential investors, lenders, and other creditors in making decisions about providing resources to the entity. Those decisions involve buying, selling, or holding equity or debt instruments and providing or selling loans and other forms of credit.

(12)

Williams and Ravenscroft (2015) highlight the necessary characteristics of accounting information to make it useful, being representational faithfulness and relevance. Representational faithfulness is describes as complete, neutral and error-free presentation and portrayal of the economic phenomena. The (accounting) information is relevant if it is capable of making a difference in the decisions made by users in their capacity as capital providers. Capital provider can both be equity investors or providers of debt.

Lev and Zarowin (1999) argue that the usefulness of accounting financials is deteriorating as the association between market values and financial information in GAAP financial statements such as earnings, cash flows and book values is weakening. This phenomenon of declining value relevance of GAAP accounting

information was also subject of the research of Balachandran and Mohanram (2011). They have investigated the impact of increasing conservatism on the decline in value relevance of GAAP accounting information. Their conclusion is that conservatism is not the driver of the decline in value relevance of GAAP accounting information.

The research in this thesis will also investigate the association between market values and financial information, but focused on non-GAAP information which was not included in the research of Lev and Zarowin.

This research qualifies as a value relevance study. Holthausen and Watts (2001) describe three different types of value relevance studies:

1. Relative association studies – These studies compare the association between stock market values (or changes in the stock market values) and alternative bottom-line measures. They explain that in these studies usually the R-squared of the regressions is tested for differences between different bottom line accounting numbers. The bottom-line accounting number with the highest R-squared is considered to be the most value relevant.

2. Incremental association studies – In these studies, it is investigated whether the accounting number of interest is helpful in explaining value or returns (over long windows) given other specified variables. The basis of these studies is to test the estimated regression coefficient between the value or returns and accounting number with the specified variables. If the regression coefficient is significantly higher or lower than zero, the accounting numbers is considered value relevant.

3. Marginal information content studies – Investigation on whether a particular accounting number adds to the information set available to investors. These are usually based on event studies where is investigated to what extent release of information (i.e. the accounting number tested) is associated with changes in the firm’s market value.

The research performed in this thesis is classified as a relative association studies (#1) as it tests the association between non-GAAP performance measures and the market value of a firm.

(13)

Important to note is that in their paper, Holthausen and Watts (2001) criticize value relevance research. They have investigated the contribution of value relevance research to standards setting but conclude that this contribution is fairly limited. One of the main reasons for that is that literature is almost solely based on equity valuation. Furthermore they argue that by using many assumptions in order to estimate value

relevance, important factors are ignored.

Barth et al. (2001) have responded to the article of Holthausen and Watts (2001) by countering the

conclusion that the contribution of value relevance literature to standard setting is limited. One of the main arguments, which is the most important argument for the research in this thesis, is that a primary focus of the Financial Accounting Standards Board (hereafter: FASB) is equity investment. Although financial statements are used for many more purposes than only equity investment, the possible contracting uses of financial statements in no way diminish the importance of value relevance research, which also focusses on equity investment (Barth et al., 2001). Value relevance studies contribute because they are designed to assess whether particular accounting amounts reflect information that is used by investors in valuing the equity of a firm (Barth et al., 2001).

2.2. Non-GAAP performance measures

Non-GAAP performance measures have been subject in many researches before. In short, non-GAAP performance measures can be explained as a customized version of GAAP performance measures, usually earnings, which are disclosed by a firm’s management to financial statement users (Bentley et al., 2018). Bentley (2018) further explains that the customization of the earnings usually consist of excluding earnings components which are deemed less representative of the core operations of the firm. The CFA Society of the UK (hereafter: ‘CFA UK’), representing more than 10.000 investments professionals in the United Kingdom, explains that there are three categories of adjustments made to GAAP earnings:

1. To adjust for the effect of one-off, unusual or non-recurring items;

2. To remove volatility associated with economic events outside the control of a firm’s management; 3. To remove the effect of accounting treatment that in the view of management does not reflect the

firm’s performance during the period (CFA, 2015).

Heflin and Hsu (2008) explain that there are at least two motives for firms to disclose non-GAAP performance measures. First, managers can improve the perceived performance by excluding certain cost items which analysts might not exclude in their (forecast) analysis. Secondly, firms choose to disclose non-GAAP performance measures to more effectively communicate permanent (recurring) earnings.

(14)

Dichev et al. (2013) found in their research that CFOs believe high quality earnings are sustainable, meaning that earnings should exclude one-off items. GAAP however restrict management’s possibilities to exclude one-off items from reported earnings.

A research by Graham et al. (2005) shows that managers consider GAAP earnings as the most important earnings metric and that they believe it is important to meet earnings benchmarks e.g. to build credibility with the capital market, maintain or increase stock prices or improve the external reputation of

management. Their research, and the research by Dichev et al. (2013), shows that managers are reluctant to use accruals management since the introduction of the Sarbanes-Oxley act. They now often use real

economic actions to meet earnings benchmarks which sometimes means that long-term value is sacrificed. This is not in the interest of shareholders and investors. Managers also use voluntary disclosures to provide the market with information to be used next to GAAP performance information. Non-GAAP (pro-forma) performance information could be an example of a voluntary disclosure to limit the risks and impacts resulting from not meeting earnings benchmarks.

Disclosing non-GAAP performance information can be seen as an attempt to reduce the information asymmetry between a firms’ management (agents) and the shareholders (principals). This information asymmetry is an important part of the agency theory as provided by Jensen and Meckling (1976) and the contract theory. One of the most difficult questions in relation to the agency and contract theory is how to measure the performance of a firms’ management (Lambert, 2001). The performance of a company is largely driven by actions taken by a firms’ management. If non-GAAP performance information is better reflecting the underlying performance of a company, ‘cleaned’ for gains and losses resulting from events outside the control of management, this contributes to the performance evaluation of management.

Examples of GAAP performance measures are among others provided by PwC (2016). Examples of non-GAAP performance measure they give are earnings before interest, taxes, depreciation and amortization (EBITDA), adjusted EBITDA, adjusted earnings per share, constant currency earnings and free cash flows. Adjusted in this case means that specific items are excluding in calculating the non-GAAP performance measure, for instance because these have a one-off nature and are not expected to recur in the future. Black et al. (2018) found that the importance of non-GAAP performance measures is increasing. This appears to be a continuing increase in importance as Brown and Sivakumar (2003) already drew this conclusion fifteen years before. In their research, Black et al. (2018) found that across all industries the frequency of non-GAAP disclosures has increased significantly. According to the authors, this is undesired by the standard setters. Lueng and Veenman (2018) also mention in their research that standard setters have concerns about the disclosure of non-GAAP information as the subjectivity and possibility to strategically present the

information can be a risk for investors.

Both the Financial Accounting Standards Board (hereafter: FASB) in the United States as the IASB consider non-GAAP performance measures inconsistent with the motives of their standards which are focused on consistent and credible information.

(15)

The chairman of the IASB, Hans Hoogervorst (2015), explains that the problem with non-GAAP performance measures is that the people preparing these alternative performance measures are free to extract and include items. The selection of items to include or exclude is not free of biases.

Contrary to the opinions of the FASB and IASB, Bhattacharya et al. et al. (2003) found that non-GAAP earnings are significantly more informative that GAAP operating earnings and appear to be a more

permanent summary measure. They argue that market participants consider non-GAAP earnings to be closer to the ‘core earnings’ of a firm than GAAP earnings are.

Further evidence is available in existing literature that non-GAAP performance information is relevant to investors. Leung and Veenman (2018) investigated the value relevance on non-GAAP performance

information for loss making firms. They conclude that especially for loss reporting firms non-GAAP earnings are a significantly better predictor for the future performance of a firm than GAAP earnings and therefore more value relevant. The difference in value relevance is even bigger for firms reporting a GAAP loss and a non-GAAP profit in a particular fiscal period.

For these firms, the conclusion is that the GAAP earnings have almost no predictive capacity for the future performance of a firm. This indicates that firm’s management is well capable of adjusting the GAAP earnings by excluding expenses which are causing the firm to report a loss but are not informative and relevant. Firms reporting non-GAAP profits together with a GAAP loss are based on their research not overvalued by

investors.

CFA UK (2015) list in their article a number of items that investors should bear in mind when using alternative (non-GAAP) performance measures as starting point for their valuation of a firm.

First, they explain that there is a potential positive bias. Firm’s quite often remove expenses that in fact are recurring, leading to non-GAAP performance measures to systematically overstate their earnings potential. Black et al. (2017) add to that by showing that managers use non-GAAP performance information to

influence shareholders perceptions, especially when expected results are just missed. Heflin and Hsu (2008) discuss that reporting ‘core earnings’ as determined by a firm’s management could lead to for instance reporting a loss under GAAP principles but a non-GAAP calculated profit as managers sometimes

opportunistically portray their performance. This led to non-GAAP earnings disclosure rules issued by the Securities and Exchange Commission (hereafter: SEC) which require that if a firm discloses non-GAAP earnings in any public communication, it must also disclose the most directly comparable GAAP number, disclose a reconciliation between the non-GAAP number and the most directly comparable GAAP number and it must file a Form 8-k containing an explanation why they believe the non-GAAP performance measure is useful for investors (Heflin & Hsu, 2008). Reporting a reconciliation increases the transparency and make reporting opportunistic non-GAAP performance measures potentially costly as substantial costs are associated with SEC enforcement actions and reputation damage could have significant consequences. The second item to consider according to CFA UK is that using alternative performance measures impacts the balance sheet and related ratios.

(16)

They provide an example on the return on invested capital ratio (hereafter: ‘ROIC’). Alternative

performance measures usually provide an alternative view on a firm’s income statement but they do not adjust the balance sheet accordingly. Therefore measures based on return on assets, such as ROIC, are based on hybrid information and consequently hard to interpret and less useful. If amortization is excluded from to an alternative operating profit and this alternative performance measure is used to calculate ROIC, this measure is likely to be overstated as the GAAP intangibles amount is net of amortization (CFA, 2015). The third and final item CFA UK mentions is that non-GAAP performance measures are often used by investors to identity potential acquisition targets, for instance by using data from data service providers. This could make firm’s using a more aggressive approach on non-GAAP performance measures appear to be a more attractive investment target than firm’s using a more conservative approach. Therefore CFA UK stresses that clarity and balance in using non-GAAP performance facilitates the investment selection process and that in order to bring this clarity and balance to non-GAAP performance measures further guidance and requirement are needed (CFA, 2015).

In prior literature, usually two different sources are used for non-GAAP performance measures. The first source, which is most commonly used (Bentley et al., 2018) are analyst forecast data providers (hereafter: FDPs) such as the International Brokers’ Estimate System (hereafter: I/B/E/S). In this I/B/E/S database, non-GAAP earnings per share information is collected since the early 1980s from earnings announcements (Brown and Sivakumar, 2003). The earlier mentioned research performed by Lueng and Veenman (2018) uses data the I/B/E/S database for their main research.

The second source is to hand collect disclosed non-GAAP information which is very time consuming and often results in smaller sample sizes. Bentley et al. (2018) argue that FDPs might not be the best source to collect non-GAAP performance measures as the non-GAAP information in these databases is based on how analysts forecast firm performance. This does not have to be in line with the non-GAAP performance measures management is using to monitor the firm’s performance. Therefore Bentley et al. (2018) prepared the first large-sample data set of manager disclosed non-GAAP earnings metrics which is publicly available for further research. They have collected the data from quarterly earnings announcements in SEC 8-K filings. After completing the collection of the non-GAAP data, they compared this data in their data set with the I/B/E/S data sets and found that they generally overlap. However, 23 percent of the non-GAAP earnings measures included in the I/B/E/S database are not actually disclosed by managers and 17 percent of the manager non-GAAP earnings measures in the author’s data set were not included in the I/B/E/S data sets.

(17)

2.3. Literature summary

The literature review covered existing literature on value relevance in combination with decision usefulness and non-GAAP performance measures. I have summarized below the main results of the literature review:

- Standard setters, such as the FASB, state that the purpose of financial information is to provide useful information for among others investors;

- Research by Lev and Zarowin (1999) tells us that the usefulness of GAAP financials is deteriorating; - Value relevance research is subject to assumptions and almost solely based on equity valuation,

which could make the outcomes less relevant for standard setting;

- Contrary to that, because it is based on equity valuation this makes value relevance research particularly relevant for investors;

- Non-GAAP performance measures are used present earnings cleaned of e.g. non-recurring items and events outside management’s control to communicate the more permanent earnings or to improve the perceived performance. CFO believe high quality earnings are sustainable, meaning that no one-off items are included in high quality earnings;

- Non-GAAP performance information can help to mitigate the agency problem by reducing information asymmetry;

- A firm’s management is free to include and exclude items from earnings. This could make them less relevant as the decision of which items are included and excluded is not free of biases;

- Despite the shortcomings of non-GAAP performance measures, prior researches performance by among others Brown and Sivakumar (2003), Albring et al. (2010) and Lueng and Veenman (2018) show that non-GAAP performance measures have a higher value relevance than GAAP earnings, especially for firms reporting a GAAP loss.

(18)

3. Hypotheses development

In the previous section, the existing literature related to value relevance, decision usefulness and non-GAAP performance measures has been set out. In this section, the hypotheses tested will be introduced.

The purpose of the research is to investigate whether non-GAAP performance measures are useful for investors to use in their investment decisions. Therefore the value relevance of the non-GAAP earnings per share, as provided by Bentley et al. (2018) in their data set, will be tested using the Ohlson model (1995). Prior literature already tells us that non-GAAP financial information can be relevant for investors. For instance, Brown and Sivakumar (2003) and Albring (2010) already found that non-GAAP performance measures are more value relevant that GAAP net income is. However, these researches are based on other data sets than the data set used in this research. Lueng and Veenman (2018) have focused their research on the value relevance of non-GAAP performance information of firms reporting a GAAP loss. Their conclusion for these specific firms is in line with the broader conclusion of Brown and Sivakumar (2003) and Albring (2010) that non-GAAP performance information is more value relevant than GAAP performance information. Contrary to the findings of Brown and Sivakumar (2003), Albring (2010) and Lueng and Veenman (2018), some institutions have concerns in relation to non-GAAP performance measure. Among others CFA UK (2015), the FASB and the IASB explain that users of non-GAAP information should bear in mind that the preparation of non-GAAP financial information is not free of biases. Managers are relatively free to include and exclude items, although under SEC regulations companies have to include a reconciliation from the non-GAAP financials to GAAP financials since 2003 (Black et al., 2018).

The information above, including the other information from prior researches as set out in the literature review, was used to develop the following main hypothesis for empirical research:

H1: Non-GAAP earnings are a value relevant performance measure.

Prior research also indicates that the value relevance of financial information is different among industries. Francis and Schipper (1999) have investigated the difference between high- and low technology firms’. This is based on the fact that expenses related to research and development are expensed when incurred but not capitalized on the balance sheet as an assent until certain requirements are met. This means that results of those firms are negatively impacted by research and development efforts although these investment

potentially lead to high future profits. However, this potential is not reflected in financial statements. Furthermore, Lueng and Veenman (2018) show that the frequency of non-GAAP reporting varies substantially across industries and that reasons for reporting non-GAAP information are not determined by equal factors. This leads to the second hypothesis tested, which is a sub hypothesis to the main hypothesis tested (H1): H1a: The value relevance of non-GAAP earnings is not uniform across industries.

(19)

Studies by among others Lueng and Veenman (2018), Balachandran and Mohanram (2011) and Holthausen and Watts (2001) state that information provided by firm’s reporting a loss is considered to be less informative than information provided by firm’s reporting a profit. Reasons for this are that losses are considered to be less persistent than profits and that losses are associated with an uncertainty about future earnings (Leung and Veenman, 2018). Holthausen and Watts (2001) explain that profit and loss firms are valued differently following the conservatism theory: “anticipate no profits but anticipate all losses”.

Leung and Veenman (2018) argue that non-GAAP performance information could help investors to better understand the nature and implications of GAAP losses and are therefore a relevant source of information. Their research shows that non-GAAP information provided by loss making firms is significantly more predictive of future firm performance than GAAP information and therefore more informative. Summarizing, GAAP financial information is valued differently across profit and loss firms. Non-GAAP performance information is considered to be relevant for investors when a firm is reporting a loss. This lead to the second sub hypothesis:

(20)

4. Research design

In this section, the research design will be discussed and explained. First, the sample selection process will be set out followed by the detailed introduction of the research model. This will include a description of the variables included in the model for the empirical research and the modifications that are made to the original research model as provided by Ohlson (1995).

4.1. Sample selection

This research qualifies as a quantitative research. The data is collected from public sources. First, the data related to non-GAAP information is provided by Bentley et al. (2018). The data set of Bentley et al. (2018) contains data of United States listed companies for the years 2003 to 2015. The data for these years have been collected on a quarterly basis. The total data set comprise 146,121 observations, including both firms that have published non-GAAP earning per share information and companies that have not published this information. The total number of non-GAAP earnings per share included in their data set is 49,918. The authors are constantly updating the data set by including additional more recent periods. The data set used is the version dated 11 July 2018.

The financial data was collected using the Wharton Research Data Services (hereafter: ‘WRDS’). This database is developed and managed by The Wharton School of the University of Pennsylvania in the United States and consist of a number of sub-databases containing specific financial- and business information. One of these sub-databases, COMPUSTAT, is used to collect all financial information used in this research which was not provided by the data set of Bentley et al. The total set of quarterly data from North American companies consisted of 448,118 observations. After excluding companies which were not included in the data set of Bentley et al., 250,549 observations were left. The companies which were not included in the data set of Bentley et al. were excluded because it is unclear whether they report non-GAAP earnings or not. The non-GAAP earnings per share from the Bentley et al. (2018) data set were combined with the data set derived from COMPUSTAT.

Starting point to create the final data set were these 250,549 quarterly observations. This set was first cleaned for firms which had one or more quarters in a fiscal year missing in the data set. It was needed to have complete information for four quarters to be able to calculate annual GAAP earnings and the annual non-GAAP earnings adjustment. Secondly, data from financial institutions was excluded from the final data set. Lueng and Veenman (2018) explain that the nature of financial institutions’ non-GAAP disclosures is systematically different from other institutions. Therefore data from firms with a SIC code between 6000 and 6999, being the financial institutions related SIC codes, was excluded.

(21)

Next, observations where required information was missing, such as non-GAAP earnings per share, the number of shares outstanding, the share price at the end of the fiscal year or the net book value at the end of the fiscal year were removed. These steps left a data set of 162,500 quarterly observations. Because the analysis is performed on annual numbers, this resulted in 40,625 annual observations.

The data set with annual data was trimmed for outliers. To mitigate the impact of the extreme values of the outliers, the 1% highest and lowest values of variables share price, net book value of equity per share, GAAP earnings per share and non-GAAP adjustment per share were indicated as outlier. In total, 2,084

observations were excluded following being indicated as outlier. Ultimately, a data set with 38,541 (N=38,541) observations was left. Refer to the table below for a summary of the data selection process.

The data set of Bentley et al. only provide non-GAAP earnings per share. This information sometimes was only provided for one, two or three quarters in a fiscal year. To calculate the total non-GAAP earnings, this research assumes no non-GAAP earnings adjustments in the other quarters of the fiscal year. Therefore the total non-GAAP earnings per share is calculated as the sum of the quarters with non-GAAP earnings per share according to the data set of Bentley et al. (2018), complemented with the GAAP earnings per share for quarters where no non-GAAP earnings per share were published as derived from COMPUSTAT.

4.2. Research model - The Ohlson Model (1995)

In this thesis, the value relevance of non-GAAP performance measures is investigated. To investigate this value relevance, the Ohlson model (1995) is used as this is the most commonly used in model in value relevance research (Barth et al., 2001).

The Ohlson model is explained as follows. According to Ohlson (1995), the model is based on what he calls ‘three analytically straightforward assumptions’.

Table 1

Data sample selection

Number of observations

Quarterly data North-American firms 2003-2015 (COMPUSTAT) 448.118

(-) Firms not in data set Bentley et al. (2018) -197.569

(-) Not 4 quarters of data available -10.597

(-) Financial institutions (SIC code 6000-6999) -53.284

(-) Missing data points -24.168

Total quarterly observations 162.500

Total annual observations 40.625

(-) Outliers (highest and lowest 1% per variable) -2.084

(22)

1. The first assumption is that based on neoclassical models of security valuation, being that the market value of a firm is the present value of expected future dividends (PVED). In order to keep the model simple, risk neutrality applies so that the discount factor is equal to the risk-free rate:

𝑃𝑟𝑖𝑐𝑒𝑡= ∑ 𝑅𝑓−𝜏𝐸𝑡[𝑑𝑡+𝜏] ∞

𝑡=1

Where 𝑃𝑡 is the price (market value) of a firm at date 𝑡, 𝑑𝑡 are the dividends paid at date 𝑡, 𝑅𝑓 is the risk-free rate plus one and 𝐸𝑡 is the expected value based on date 𝑡 information.

2. Second assumption is the Clean Surplus Relation (CSR). This assumes all changes in a firm’s equity are driven by the net result, which is the outcome of all information flowing through a firm’s income statement, or by dividend payments to the shareholders:

𝑁𝐵𝑉𝑡−1= 𝑁𝐵𝑉𝑡+ 𝑑𝑡+ 𝑁𝐼𝑡

Where 𝑁𝐵𝑉 reflects the firm’s net book value of equity, 𝑑𝑡 are the dividends paid at date 𝑡 and 𝑁𝐼𝑡 is the net income in period 𝑡.

3. The third assumption is that there is a relation between abnormal earnings in consecutive periods. Ohlson furthermore included a variable for ‘other information’ to satisfy an autoregressive process. This information included in ‘other information’ is not explained in detail and could potentially be any (non-)financial information.

Combining assumption 1 with the model from assumption 2 leads to what today is known as the Ohlson model for value relevance:

(1) 𝑃𝑟𝑖𝑐𝑒𝑡= 𝑎0+ 𝛽1𝑁𝐵𝑉𝑡+ 𝛽2𝑁𝐼𝑡+ 𝜀

Before testing how value relevant non-GAAP performance information is, I first analyze whether firms reporting non-GAAP performance information are valued differently than firms only reporting GAAP performance information. Therefore the following model is used:

(2) 𝑃𝑟𝑖𝑐𝑒𝑡= 𝑎0+ 𝛽1𝑁𝐵𝑉𝑡+ 𝛽2𝑁𝐼𝑡+ 𝑁𝐺𝑡+ 𝛽3𝑁𝐺𝑡∗ 𝑁𝐼𝑡+ 𝜀 Where:

𝑁𝐵𝑉𝑡 = a firms’ net book value of equity in fiscal year 𝑡; 𝑁𝐼𝑡 = a firms’ GAAP net income per share in fiscal year 𝑡;

𝑁𝐺𝑡 = a dummy variable indication whether a firm has reported non-GAAP performance information in a particular fiscal year 𝑡.

(23)

To analyze the value relevance of non-GAAP performance information, the Ohlson model will be adjusted to address two econometric problems as identified by Balachandran and Mohanram (2001) in their value relevance research. First, they explain that the incidence of losses has increased over time. This is likely to lower the goodness of fit of value relevance regressions because the informativeness of losses tends to be lower than the informativeness of profits. To control for losses, the authors have included separate

intercepts for and slopes for loss firms based on research by Core et al. (2003). In my research, I will control for this in a similar way.

Secondly, value relevance regressions as the Ohlson-model, assume the coefficients on earnings and book values to be the same for all industries. Balachandran and Mohanram (2001) explain that if inter-industry heterogeneity has increased over time, the value relevance and meaningfulness would decrease. This

decrease is in that case not driven by a decrease of the value relevance of the accounting information but by the fact that this information is becoming increasingly different between industries. To control for this, I use roughly the same approach as Balachandran and Mohanram (2001) have used. They have used the

classification of SIC codes as provided by Fama and French (1997) to divide the firms in the sample into 48 different industry groups. These 48 industry groups are used in the model to let earnings and book value coefficients to vary between industry groups. In this research, the Fama and French classification is used which divides SIC coded in 12 different industry groups. The same classification is used in many researches, among other in the earlier mentioned research by Lueng and Veenman (2018).

The adjustments as described above lead to the following adjusted value relevance model:

(3) 𝑃𝑟𝑖𝑐𝑒𝑡= ∑ 𝑎𝑗 12 𝑗=1 ∗ 𝐼𝑁𝐷𝑗+ ∑ 𝛽1𝑗 12 𝑗=1 𝐼𝑁𝐷𝑗∗ 𝑁𝐵𝑉𝑡+ ∑ 𝛽11𝑗 12 𝑗=1 𝐼𝑁𝐷𝑗∗ 𝑁𝐵𝑉𝑡∗ 𝐿𝑂𝑆𝑆 + ∑ 𝛽2𝑗 12 𝑗=1 𝐼𝑁𝐷𝑗∗ 𝑁𝐼𝑡+ ∑ 𝛽22𝑗 12 𝑗=1 𝐼𝑁𝐷𝑗 ∗ 𝑁𝐼𝑡∗ 𝐿𝑂𝑆𝑆 + ∑ 𝛽3𝑗 12 𝑗=1 𝐼𝑁𝐷𝑗∗ 𝑁𝐺𝑎𝑑𝑗𝑡+ ∑ 𝛽33𝑗 12 𝑗=1 𝐼𝑁𝐷𝑗∗ 𝑁𝐺𝑎𝑑𝑗𝑡∗ 𝐿𝑂𝑆𝑆 + 𝜀 Where:

𝐼𝑁𝐷𝑗 = dummy variable for the 12 industry groups based on the Fama and French classification of SIC codes; 𝐿𝑂𝑆𝑆 = dummy variable indicating if firms report negative GAAP net income per share;

𝑁𝐵𝑉𝑡 = a firms’ net book value of equity in fiscal year 𝑡; 𝑁𝐼𝑡 = a firms’ GAAP net income per share in fiscal year 𝑡;

𝑁𝐺𝑎𝑑𝑗𝑡 = the reported non-GAAP adjustment to GAAP net income per share by a firm in fiscal year 𝑡; The analyses will be performed using SPSS, a software tool designed by IBM to use for statistical analysis. A linear weighted ordinary least squares regression is ran where the market capitalization of the firms in the data set is used as the weight factor. Easton and Sommers (2003) argue that it is difficult to argue that there

(24)

are better measures of scale than market capitalization when performing market based accounting research. This because of the high influence of large firms, which according to them is called the ‘scale effect’. The scale effect results in coefficient biases and heteroscedasticity in unweighted regression analyses.

4.3. Variables

Table 1

Variable

Description of the variable

Dependent variable

Price This variable reflects a firms’ share price at the end of the fiscal year. (COMPUSTAT _{code: PRCCQ)}

Price +3m This variable is used in the robustness test and reflects a firm’s share price 3 months after the end of the fiscal year (COMPUSTAT code: PRCCQ). The variable is calculated using the share price at the end of the first quarter in the next fiscal year.

Independent variables

NBV

This variable reflects the net book value of equity per share at the end of the fiscal year. This is calculated by dividing the net book value of equity (COMPUSTAT code: CEQQ) by the number of outstanding shares at the end of the last quarter of the fiscal year (COMPUSTAT code: CSHOQ)

NI This variable reflects the GAAP net income per share of the fiscal year. This is calculated by adding up the quarterly GAAP net income (COMPUSTAT code: NIQ) per share (COMPUSTAT code: CSHPRQ)

NGadj

This variable represents the difference between the GAAP net income per share and the non-GAAP net income per share for a fiscal year. The non-GAAP net income per share is calculated as the total of the quarters where firms have reported non-GAAP earnings per share supplemented with the GAAP earnings per share for quarters where no non-GAAP earnings per share were reported. The non-GAAP earnings per share are derived from the data set of Bentley et al. (2018).

IND This variable is a multiple dummy variable for the 12 Fama-French industry classifications that equals 1 if a firm is classified in the industry of the dummy and a 0 otherwise.

LOSS This variable is a dummy variable that equals a 1 if a firm reports a loss for the fiscal _{year and a 0 otherwise.} NG This variable is a dummy variable that equals a 1 if a firm has reported non-GAAP _{earnings per share in a fiscal year and a 0 otherwise.}

Weight Market

capitalization

Market capitalization is used as weight factor in the OLS regression. This variable is calculated as the price at the end of a fiscal quarter (COMPUSTAT code: PRCCQ) multiplied by the number of outstanding shares at the end of a fiscal quarter (COMPUSTAT code: CSHOQ).

(25)

5. Results

This section of the thesis will include the descriptive statistics and the empirical results of the analyses. This starts with the descriptive statistics, followed by the outcome of the conducted analyses. There, the

outcomes will be discussed and interpreted.

5.1. Descriptive statistics

In this section, the descriptive statistics will be presented along with the correlations between the variables. Refer to section 4.1 on this research for a description of the sample selection process. This process resulted in 38,541 observations that were used for the empirical research. Table 3 provides a summary of the statistics of the variables used in the empirical research. The share price ranges from $16 cents to $125.35, with an average of $21.15. The average net book value of equity of the firms included in the final data set is $8.77 per share and the average GAAP net income per share is $0.66 with a minimum of $7.41 negative and a maximum of $8.79. Note that these are the minimums and maximums after trimming the data set by 1% for the highest and lowest values.

From the firms reporting non-GAAP earnings per share, the average adjustment per share in a fiscal year was $14 cents with a lowest value of $8.01 negative and a highest value of $14.25. It is remarkable that the highest and lowest value of the non-GAAP adjustment per share are more extreme that the GAAP net income per share. By taking a closer look at a number of the extreme non-GAAP adjustments, it turned out that these extreme values are caused high impact events such as goodwill impairments. The value of the impairment in some cases is significantly higher than the net result. For instance, Plains Exploration & Production Company, a firm active in petroleum and natural gas exploration, reported a GAAP net loss per share of $6.52 in fiscal year 2008. The non-GAAP earnings per share were a positive amount of $4.81, being $11.33 higher. This non-GAAP adjustment of $11.33 largely relates to an impairment of the oil and gas properties (representing $33.81 of the non-GAAP adjustment per share), partly offset by a gain on the mark-to-market derivative contracts (offsetting the non-GAAP adjustment by $14.50 per share) and some other adjustments (further offsetting the impairment adjustment by $7,98).

The NG dummy has a mean of 0.38, meaning that 38% of the fiscal years included in the sample include a non-GAAP adjustment.

The market capitalization of a firm is used as a weight in the regressions for scaling purposes. Note that this is not a variable in the regression.

(26)

The minimum market capitalization of a firm included in the sample is $0.45 million, the highest market capitalization is $504.24 billion with an average of €3.68 billion and a standard deviation of $15.6 billion. This shows the substantial variation in market capitalization of the firms included in the final sample.

In regression model 3, I have corrected for industry effects and for the impact of reporting a GAAP loss by using dummy variables. Table 4 shows how the observations are distributed among the industries and between reporting a loss or profit.

The observations are not equally distributed among the twelve Fama-French industry classifications. In the final sample, no observations are included in industry class 11 (Money), as this class includes the SIC codes (6000-6999) related to Financial Institutions. As described in paragraph 4.1, all observations from financial institutions are excluded from the final data set as the nature of financial institutions’ non-GAAP disclosures is systematically different from other institutions.

Tabel 3

Variable N Minimum Maximum Mean Std. Deviation

Price 38.541 ,16 125,32 21,15 20,87 NBV 38.541 -6,21 48,06 8,77 8,79 NI 38.541 -7,41 8,79 ,66 1,82 NG (dummy) 38.541 0 1 ,38 ,48 NGadj 38.541 -8,01 14,25 ,14 ,63 Mark et cap. 38.541 ,45 504.239,58 3.683,72 15.612,92 All variables are defined in Table 2

Descriptive Statistics

Table 4

Distribution of observations by industry and profit/loss

Industry Description # observations As a % of total

1 Consumer non-durables 2.242 5,8% 2 Consumer durables 932 2,4% 3 Manufacturing 4.366 11,3% 4 Energy 2.164 5,6% 5 Chemicals 1.052 2,7% 6 Business equipment 9.141 23,7% 7 Telecom 1.230 3,2% 8 Utilities 1.353 3,5% 9 Shops 4.260 11,1% 10 Health 6.338 16,4% 11 Money 0 0% 12 Other 5.463 14,2% 38.541 Reported a profit 25.077 65,1% Reported a loss 13.464 34,9% 38.541 Total observations Total observations

(27)

Table 5 shows the correlation coefficients and indicates the degree of correlation between the variables in the linear regression models. The correlation coefficients in the table are the bivariate Pearson coefficients, which can have a value between minus 1 and 1.

All variables are significantly correlated at a two-tailed 1% significance level. There is no multicollinearity between the variables which could affect the reliability of the models. The correlation between net income per share (NI) and the non-GAAP adjustment per share (NGadj) is particularly interesting as they are

negatively correlated. This means that when net income per share increases, the magnitude of the non-GAAP adjustments decreases.

5.2. Results of the empirical research

The results from empirical research will be presented and discussed in this section. This will be structured based on the hypotheses tested in this thesis. First, the outcomes of testing the value relevance of non-GAAP performance measures will be presented and discussed.

Tabel 5

Price NBV NI NGadj. Loss

(dummy) Pearson Correlation 1 Sig. (2-tailed) N 38.541 Pearson Correlation ,666 ** ₁ Sig. (2-tailed) ,000 N 38.541 38.541 Pearson Correlation ,649 ** ,564** 1 Sig. (2-tailed) ,000 ,000 N 38.541 38.541 38.541 Pearson Correlation ,053 ** ,078** -,283** 1 Sig. (2-tailed) ,000 ,000 ,000 N 38.541 38.541 38.541 38.541 Pearson Correlation -,442 ** -,410** -,669** ,158** 1 Sig. (2-tailed) ,000 ,000 ,000 ,000 N 38.541 38.541 38.541 38.541 38.541 Loss (dummy)

All variables are defined in Table 2. **Correlation is significant at the 0.01 level (2-tailed). Correlations

Price

NBV

NI

(28)

This will be followed by presenting and discussing the outcome of testing the value relevance of non-GAAP performance measures among the different industries classified in accordance with the Fama-French 12-industries classification. Next, differences in value relevance of non-GAAP performance information between loss and profit firms will be presented and discussed. All regressions are scaled by the market capitalization of the firms included in the sample.

5.2.1. Results from preliminary analyses

To examine if firms reporting non-GAAP performance measures are valued differently than firms reporting no non-GAAP performance measures, I fist run the Ohlson-model regression without non-GAAP earnings per share. Refer to paragraph 4.2, model (1). The outcome is included in table 6. Table 6 shows an adjusted R-square of 0,458, meaning that 45.8% of the variation in price is explained by the mode land shows that NBV and NI are both positively significant. This means that both NBV and NI have impact on the value of a company. Therefore, this information can be considered to be value relevant.

Table 8 present the outcomes of the Ohlson-model, including a dummy indication if a firm has reported non-GAAP performance information in a certain year. Refer to paragraph 4.2, model (2). The adjusted R-square is 0,461, which is slightly higher than the adjusted R-square without the non-GAAP variable. Table 9 shows that the non-GAAP dummy, NG (dummy)*NI, is significantly associated with the depended variable. However, the Beta is negative (-0,030). This means that firms reporting non-GAAP performance information are generally valued lower than firms not reporting this information. This could be driven by the fact that non-GAAP adjustments sometimes are used to remove costs which in fact are recurring to report a better performance (CFA UK, 2015).

Nevertheless, the outcomes presented in table 6 to 7 indicate that firms reporting non-GAAP performance information are valued differently compared to firms who do not.

Table 6

Model 1: The Ohlson-model

Coeffic. t-stat Intercept 23,890 130,598 *** NBV ,480 42,621 *** NI 7,043 128,725 *** N 38,541 Adjusted R² ,458

The sample period is 2003 to 2015. The annual regression is run with price per share (Price) as the dependent variable and net book value of equity per share (NBV) and GAAP net income per share (NI) as independent variables. N reflects the number of observations.

(29)

Table 7

Model 1: The Ohlson-model including a non-GAAP adjustment dummy

Coeffic. t-stat Intercept 21,499 82,757 *** NBV ,479 42,486 *** NI 7,303 95,979 *** NG (dummy) 3,808 12,130 *** NG (dummy)*NI -,407 -4,313 *** N 38,541 Adjusted R² ,461

The sample period is 2003 to 2015. The annual regression is run with price per share (Price) as the dependent variable and net book value of equity per share (NBV) and GAAP net income per share (NI) as independent variables. The non-GAAP dummy (NG (dummy)) is 1 if a firm has reported non-GAAP information in a certain fiscal year. N reflects the number of observations.

5.2.2. Results from testing Hypothesis 1

The outcome of the testing the first hypothesis (H1) is presented in table 8. Refer to paragraph 4.2, model (3) for the regression model used to test the hypothesis.

Firstly, the adjusted R-square is 0.911, meaning that 91.1% of the variation in stock price is explained by the variables included in the model. This can be seen as the value relevance of the combination of the variables included in the model. An adjusted R-squared of 91.1% is considered as high, indicating the goodness of fit of the model and implicating that the information included is value relevant.

Secondly we see that for the vast majority of the industries, NBV and NI are positively associated with firm value. Only the NBV in Industry 1 is not positively associated but this relation is fairly weak (–0,072). All NBV and NI relationships with the stock price are significant. This is in line with the expectation as this is a one of the assumptions the Ohlson model (Ohlson, 1995) is based on.

Thirdly, we see that in general reporting NI*LOSS and NGadj*LOSS are negatively associated with firm value. This means that for most industries reporting a loss lowers the relevance of the reported numbers or

adjustments. Refer to the outcomes of testing hypothesis 1b related to differences in value relevance between profit and loss firms for more details.

Finally and most important, we find that for all 11 industries in the sample the non-GAAP adjustment to earnings (per share) is positively associated with the share price of a firm. This association is significant at a 1% level of all industries. This positive relation means that for every 1 dollar adjusted to the GAAP earnings, the share prices moves in the same direction with the value of the coefficient. This is in line with the outcomes of among others the researches by Albring (2010) and Brown and Sivakumar (2003). They have also found that non-GAAP performance information is value relevant. For 5 of the 11 industries (i.e. industries 2, 5, 6, 9 and 12) in the sample the coefficient of the non-GAAP adjustment is visibly higher than the GAAP net income coefficient.

(30)

This means that there is a stronger impact on a firm’s value from non-GAAP performance information than from GAAP performance information. Refer to the outcomes of hypothesis 1a for more details.

Table 8

Outcome of testing the value relevance of non-GAAP performance information

Coeffic. t-stat Industry dummy 1 13,999 17,595 *** ID1*NBV -,072 -2,003 ** ID1*NBV*LOSS ,950 5,175 *** ID1*NI 13,866 57,611 *** ID1*NI*LOSS -12,456 -7,267 *** ID1*NGadj 11,600 16,070 *** ID1*NGadj*LOSS -10,188 -5,427 *** Industry dummy 2 11,745 8,614 *** ID2*NBV ,523 5,214 *** ID2*NBV*LOSS ,896 2,318 ** ID2*NI 7,397 12,736 *** ID2*NI*LOSS -9,871 -6,558 *** ID2*NGadj 13,367 13,212 *** ID2*NGadj*LOSS -15,299 -6,913 *** Industry dummy 3 17,160 25,483 *** ID3*NBV ,358 10,182 *** ID3*NBV*LOSS ,468 4,329 *** ID3*NI 9,724 56,832 *** ID3*NI*LOSS -8,112 -10,140 *** ID3*NGadj 7,832 9,854 *** ID3*NGadj*LOSS -6,557 -6,051 *** Industry dummy 4 22,498 31,563 *** ID4*NBV ,631 21,990 *** ID4*NBV*LOSS ,270 3,896 *** ID4*NI 5,326 37,961 *** ID4*NI*LOSS -2,246 -3,624 *** ID4*NGadj 3,693 6,570 *** ID4*NGadj*LOSS -1,592 -2,151 ** Industry dummy 5 21,316 19,455 *** ID5*NBV ,132 2,484 ** ID5*NBV*LOSS ,795 2,792 *** ID5*NI 10,606 32,621 *** ID5*NI*LOSS -7,847 -3,216 *** ID5*NGadj 11,841 14,710 *** ID5*NGadj*LOSS -12,506 -5,548 *** Industry dummy 6 13,105 37,363 *** ID6*NBV ,752 21,686 *** ID6*NBV*LOSS ,622 6,276 *** ID6*NI 8,641 44,914 *** ID6*NI*LOSS -7,703 -12,607 ***

(31)

ID6*NGadj 11,384 27,784 *** ID6*NGadj*LOSS -10,041 -14,987 *** Industry dummy 7 5,127 6,573 *** ID7*NBV ,540 12,942 *** ID7*NBV*LOSS ,555 3,672 *** ID7*NI 10,924 39,061 *** ID7*NI*LOSS -13,405 -16,439 *** ID7*NGadj 7,220 11,130 *** ID7*NGadj*LOSS -10,517 -5,882 *** Industry dummy 8 14,519 15,602 *** ID8*NBV ,427 8,571 *** ID8*NBV*LOSS ,208 1,277 ID8*NI 7,763 20,066 *** ID8*NI*LOSS -6,364 -4,381 *** ID8*NGadj 5,599 6,749 *** ID8*NGadj*LOSS -3,683 -2,716 *** Industry dummy 9 12,535 20,260 *** ID9*NBV -,202 -5,217 *** ID9*NBV*LOSS 1,614 9,810 *** ID9*NI 14,633 63,741 *** ID9*NI*LOSS -12,034 -10,774 *** ID9*NGadj 14,978 18,558 *** ID9*NGadj*LOSS -10,235 -8,809 *** Industry dummy 10 18,765 36,789 *** ID10*NBV ,510 11,516 *** ID10*NBV*LOSS ,824 7,205 *** ID10*NI 9,091 43,729 *** ID10*NI*LOSS -12,471 -23,486 *** ID10*NGadj 8,454 21,595 *** ID10*NGadj*LOSS -7,668 -12,567 *** Industry dummy 12 18,931 33,320 *** ID12*NBV ,034 1,143 ID12*NBV*LOSS ,868 8,068 *** ID12*NI 11,680 58,349 *** ID12*NI*LOSS -9,826 -13,166 *** ID12*NGadj 12,611 25,216 *** ID12*NGadj*LOSS -12,587 -13,752 *** N 38.541 Adjusted R² ,911

The sample period is 2003 to 2015. The annual regression is run with price per share (Price) as the dependent variable and net book value of equity per share (NBV), GAAP net income per share (NI) and non-GAAP adjustments per share (NGadj) as independent variables. Dummies variables are used to allow separate coefficients for NBV, NI and NGadj for loss firms (LOSS) and separate coefficients based on industry (IDxx) (refer to paragraph 4.2). N reflects the number of observations.