The effect of mandatory disclosure on the informativeness and
opportunistic appearance of alternative earnings measures
Name: Joëlle Veltman Student number: 10794352 Thesis supervisor: Réka Felleg Date: 24-06-2018
Word count: 15,477
MSc Accountancy & Control, Accountancy
1
Statement of Originality
This document is written by student Joëlle Veltman 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.
2 Abstract
I examine whether mandatory disclosure of an alternative earnings measure decreases the opportunistic appearance of this measure and increases the difference in informativeness between GAAP earnings and the alternative earnings measure compared with voluntary disclosure. In my main tests I find evidence that firms in a mandatory disclosure setting are less likely to disclose higher alternative earnings figures than GAAP earnings. However, for other proxies of difference in opportunistic appearance I find no evidence, and I do not find evidence for a different informativeness gap between alternative earnings and GAAP earnings for voluntary disclosure firms compared with mandatory disclosure firms. In my robustness tests I do find that the alternative earnings measure is lower and that the extent to which the alternative earnings measure is more informative than GAAP earnings is larger for firms in a mandatory disclosure setting. Thus, there is some evidence for a higher quality of the
alternative earnings measure for firms that are mandated to disclose them compared to firms that do so voluntarily.
3 Table of contents
1. Introduction ... 4
2. Literature review and hypothesis development ... 6
2.1 Regulation of headline earnings and non-GAAP/non-IFRS earnings ... 6
2.2 Alternative earnings measures ... 7
2.3 Effects of regulation on alternative earnings measures ... 9
2.4 Effects of mandatory disclosure of alternative earnings measures ... 10
2.5 Voluntary versus mandatory disclosure ... 11
2.6 Theory on additional disclosure... 12
2.7 Hypotheses development ... 14
3. Sample Selection and Research Design ... 16
3.1 Sample Selection ... 16
3.2 Research Design... 18
4. Results ... 21
4.1 Descriptive Statistics ... 21
4.2 Effect on the difference in reconciliations ... 24
4.3 Effect on the informativeness of alternative earnings ... 29
4.4 Additional analysis ... 31
4.4.1 Median split in Model (1) ... 31
4.4.2 Scaling variables in Model (2) ... 32
5. Conclusion ... 34
4 1. Introduction
Since 2000, South African companies listed on the Johannesburg Stock Exchange Limited [JSE] are required to disclose non-IFRS earnings, which is referred to as headline earnings (South African Institute of Chartered Accountants [SAICA], 2015). Headline earnings is an earnings measure that excludes separately identifiable re-measurements, and is meant to inform investors better about the operating and trading activity of an entity (SAICA, 2015). More specifically, it is stated that headline earnings will reflect performance in the reporting period and future periods better than IFRS earnings as it will only include the earnings from ‘core operations’ of a firm. For this reason, SAICA decided to meet the large demand for an alternative earnings number by making headline earnings a mandatory figure in the financial statements.
In other voluntary disclosure regimes, there is an ongoing debate on the advantages and disadvantages of alternative earnings measures. Several studies have found that
alternative earnings measures are used opportunistically by managers (e.g. Doyle, Jennings & Soliman, 2013; Walker & Louvari, 2013), but other studies also show that these earnings can be more informative for investors than GAAP earnings in these regimes (e.g. Aubert, 2010; Bhattacharya, 2003). However, the results of studies in voluntary regimes may not be valid in a mandatory regime, as several researchers have found that effects can differ when disclosure is mandatory compared to voluntary (e.g. Butler, Kraft & Weiss, 2007; Venter, Cahan & Emanuel, 2013). Thus, it is important to compare the effect of mandatory versus voluntary disclosure on alternative earnings measures, as the results might be different.
To the best of my knowledge, only the studies by Venter, Emanuel and Cahan (2014) and Venter et al. (2013) have investigated the value relevance of headline earnings compared to mandated earnings and respectively the correct pricing of earning components by investors in South Africa, which is the only country where alternative earnings disclosure is mandatory. However, these studies only investigate the mandatory setting and do not compare this with a voluntary setting. I expect that firms that have relatively low alternative earnings are the firms that do not disclose an alternative earnings measure, and thus, mandatory disclosure will decrease the opportunistic appearance of alternative earnings measures. Furthermore, I assume that the firms that do disclose alternative earnings in a voluntary regime do not disclose every exclusion because some information is unfavourable or too costly to disclose, however they choose to disclose exclusions that are a part of core operations. Therefore, I expect that mandatory disclosure, where firms are provided with a detailed description of which expenses and revenues should be excluded, increases the extent to which the alternative
5
earnings measure is more informative than the GAAP earnings measure. Thus, a comparison is important for standard setters, policy makers, and investors, as differences across the settings will likely require the alternative earnings number to be interpreted differently dependent on the disclosure regime, which in turn may alter their investment and policy decisions. Therefore, this study examines the difference in the informativeness and
opportunistic appearance of non-GAAP earnings in a mandatory disclosure setting compared to voluntary disclosure settings.
My measure for the informativeness of the alternative earnings measure is the extent to which future net operating cash flows are predicted by exclusions, and opportunistic appearance is measured by the number of reconciliations, the absolute amount of
reconciliations, and the probability of the alternative earnings measure being higher than GAAP earnings. I hand-collected this data for the largest listed firms in South Africa, Australia, and the US to be able to compare the mandatory disclosure setting with the voluntary disclosure setting. The US and Australia were chosen as reference countries because they are in the same accounting cluster as specified by Mueller, Gernon and Meek (1994). Moreover, as South Africa and Australia use IFRS for reporting and the US uses US-GAAP, I can control for the choice of reporting standards as well.
In my main analysis I find that firms in a mandatory disclosure setting are less likely to disclose an alternative earnings measure that is higher than GAAP earnings compared to firms in a voluntary disclosure setting. However, for other proxies I find no difference in the opportunistic appearance of alternative earnings measures between a mandatory and a voluntary disclosure setting. Furthermore, the extent to which the alternative earnings measure is more informative than GAAP earnings is not different for firms in a mandatory disclosure setting than for firms in a voluntary disclosure setting. However, in my robustness tests I do find evidence that the alternative earnings measure is relatively lower in a
mandatory disclosure setting compared to voluntary disclosure setting. Moreover, I find that the extent to which the alternative earnings measure is more informative than GAAP earnings is larger in a mandatory disclosure setting. Hence, there is some evidence to infer that the alternative earnings measure is of higher quality for firms that are mandated to disclose them compared to firms that do so voluntarily.
This paper contributes to the current literature in several aspects. It is the first study to compare mandatory and voluntary disclosure of alternative earnings numbers. In this way this study also contributes to the existing literature on the benefits and disadvantages of
6
information and disclosure is always better. Furthermore, the results are important for
investors, as the results imply that investors should bear in mind that the quality of alternative earnings is lower for firms in a voluntary disclosure setting compared to a mandatory
disclosure setting when interpreting firms’ financials. Lastly, the result is relevant for policy makers as this study concludes that making alternative earnings measures mandatory is beneficial for the quality of the measure, which can be taken into consideration by
policymakers when making a cost-benefit analysis in deciding to mandate earnings for more stock exchanges or as an accounting standard.
The paper proceeds as follows. Section 2 provides an overview of prior literature and the hypotheses development. Section 3 introduces sample selection procedures and my research design. Section 4 presents the results and further analysis. Section 5 concludes and discusses the limitations of this paper.
2. Literature review and hypothesis development
2.1 Regulation of headline earnings and non-GAAP/non-IFRS earnings
Headline earnings, which is obligatory to be present in the annual reports of JSE listed companies, is a supplement to the basic earnings prescribed by IAS 33 (SAICA, 2013). This number excludes re-measurements net of tax and non-controlling interest, with some
exceptions as some re-measurements are specifically expected to be included in BOTH headline and IFRS earnings. A re-measurement is defined by the SAICA (2013, p. 6) as: “an amount recognised in profit or loss relating to any change (whether realised or unrealised) in the carrying amount of an asset or liability that arose after the initial recognition of such asset or liability”. The specific rules per IFRS and IAS standards can be found in SAICA (2013, pp. 9-18) publication on headline earnings, but the underlying criterion for excluding
re-statements from headline earnings is that the re-statement should not be related to the core operations and trading of an entity. Furthermore, headline earnings must be presented with a reconciliation between the basic earnings number and headline earnings, showing all excluded re-measurements. Because of this, the difference in reconciliations made to get to headline earnings can be observed easily. It is important to note that there are detailed rules on what is and what is not included in headline earnings and that this is not something that is at the companies’ discretion since 2007 (SAICA, 2013, p. 8). This has explicitly been done to
7
ensure consistency of reporting for investors and other stakeholders to be able to compare the non-IFRS earnings measure across companies.
In the US, disclosing non-US GAAP measures, including non-US GAAP earnings, is a voluntary decision. However, these measures are subject to Regulation G since the Sarbanes-Oxley act of 2002 (US Securities and Exchange Commission [SEC], 2003). This regulation, similar to that of headline earnings in South Africa, has a reconciliation requirement; the requirement to first present a financial measure that is most directly comparable and then present reconsolidations to eventually arrive at the non-GAAP measure. Also, a statement must be provided by the company that explains why this measure provides relevant
information. Furthermore, the regulation prohibits items to be classified as non-recurring/re-measurements when it is highly likely that such a charge or gain reoccurs in the coming two years, or has occurred in the past two years.
In Australia, just as in the US, disclosing non-GAAP earnings is also a voluntary decision. Furthermore, similar to Regulation G in the US, there are regulations imposed by the Australian Securities & Investments Commission [ASIC] on the disclosure of non-GAAP earnings (ASIC, 2011). Australian companies also have to meet the regulatory requirement of disclosing a reconciliation between IFRS and non-GAAP earnings and issue a statement on why the information is useful. Furthermore, the Australian regulation also states that items that have a high likelihood to reoccur are not allowed to be classified as non-recurring and, therefore, cannot be excluded from non-GAAP earnings.
In sum, South Africa, the US, and Australia are regulated with regard to the disclosure of an alternative earnings measure. In all three countries firms can exclude items only if the likelihood of reoccurrence of this item is low, and firms have to provide a reconciliation containing details on the separate exclusions made. However, South Africa is the only country for which it is mandatory to disclose this alternative earnings measure with a detailed
description of items that can be excluded per reporting standard.
2.2 Alternative earnings measures
On the one hand, providing alternative earnings measures is stated to be used at a manager’s discretion to enhance their performance as perceived by investors and other stakeholders. On the other hand, several studies find that alternative earnings measures are more informative than GAAP earnings, implying that the alternative earnings measure is a fair representation of the core performance of a firm instead of an enhanced representation.
8
Several studies have found that alternative earnings measures have been used
opportunistically by managers to improve the performance of their firm in voluntary settings. Moreover, firms that benefit the most from reporting alternative earnings tend to be firms that actually disclose this alternative earnings measure. For instance, in a study by Walker and Louvari (2013) performed in the UK, they find evidence that disclosing alternative earnings per share [EPS] measures is a deliberate choice by managers. More specifically, they find a positive relation between firms voluntarily disclosing an alternative EPS measure and the amount by which this measure exceeds the earnings measure provided under FRS3. Furthermore, they find that firms are more likely to disclose alternative earnings measures when these measures increase reported profits compared to when the measure decrease reported profits. Similarly, firms are also more probable to report an alternative EPS measure when this alternative number is positive whereas the mandatory FRS3 figure is negative. The results of studies by Bradshaw and Sloan (2002) and Aubert (2010) performed in the US and France respectively are similar: they also show that firms are more likely to report street earnings or pro forma earnings, another term for non-GAAP earnings, which exceeds the GAAP earnings number. Aubert (2010) furthermore states that pro forma earnings measures are used to meet or beat analysts’ forecasts. Namely, while only 18% of the firms met or beat analysts’ forecasts on the basis of the audited GAAP earnings number, approximately 89.4% did so based on the non-GAAP earnings number. Bhattacharya et al. (2003) find similar but less drastic results in their study in the US; whereas 39% of the firms beat or meet analysts’ forecasts based on GAAP earnings, 80% of the firms does so based on non-GAAP earnings. This is also in line with the findings by Doyle et al. (2013) in the US, as they find the numbers reported by firms as non-GAAP earnings are more likely to meet or beat analysts’ forecasts. In particular, especially other (unexpected) exclusions, which are exclusions other than non-recurring special items, are found to be the primary means by which these forecasts are met or beat. Altogether, there is evidence from different countries that the alternative earnings
measure is used opportunistically because it is more often than not higher than GAAP earnings and meets analysts’ forecasts.
Although the alternative earnings figure is used to boost the performance of a company, there is also literature that shows that this earnings number is more informative than non-GAAP earnings. For example, Aubert (2010) concludes that pro forma earnings have a higher correlation with abnormal daily equity returns compared to the GAAP earnings measure, and thus, non-GAAP earnings is more useful in assessing the firm’s future
9
(2003), who also find pro forma earnings to be more informative based on short window abnormal returns. Furthermore, although Walker and Louvari (2013) found that firms that have a negative FRS3 earnings number are more likely to report an alternative earnings measure, Leung and Veenman (2017) show that even for loss firms the non-GAAP earnings measure is highly informative. A slightly different interpretation of this matter is provided by Lougee and Marquardt (2004), as they find that firms that have less informative GAAP earnings are more likely to disclose non-GAAP earnings than other firms. This would imply that non-GAAP earnings are not per se more informative, but the firms that choose to report non-GAAP earnings have GAAP earnings with low informativeness.
In conclusion, alternative earnings appear to be used opportunistically by managers as firms are more likely to report alternative earnings when this measure is higher than GAAP earnings, turns a GAAP loss to a profit, or meets or beats analysts’ forecasts. However, because there is evidence that alternative earnings is a more informative measure in predicting future performance, it could also be that alternative earnings is a faithful reflection of good core performance that firms cannot show with GAAP earnings.
2.3 Effects of regulation on alternative earnings measures
Several studies have investigated the effect of regulation on the quality and
opportunistic appearance of alternative earnings measures and overall found mixed evidence for a quality of informativeness increase but one-sided evidence for the decrease in
opportunistic appearance of non-GAAP earnings. Heflin and Hsu (2008), Kolev, Marquardt and McVay (2008), and Entwistle, Feltham and Mbagwu (2006) all have studied the impact of Regulation G on non-GAAP earnings in the US. Heflin and Hsu (2008) concluded that the regulation has decreased the opportunistic use of non-GAAP earnings as they found that the amount of GAAP earnings disclosure, the exclusion magnitude, the probability of non-GAAP earnings meeting or beating analyst forecasts, and the relation between returns and forecasting errors have all decreased. However, not only did firms disclose fewer other-items, which are generally perceived to be used opportunistically, but also fewer special items. Therefore, Heflin and Hsu (2008) concluded that Regulation G also had the negative side-effect of firms not disclosing non-opportunistic non-GAAP earnings because the costs of disclosure have increased with the implementation of the regulation. Similarly to Heflin and Hsu’s (2008) results, Entwistle et al. (2006) have also found that the portion of firms
disclosing non-GAAP earnings and the magnitude of exclusions have decreased. Furthermore, they find that non-GAAP earnings are presented less prevalently and less misleadingly in
10
press releases. Kolev et al. (2008) have researched the effect of Regulation G on the quality of non-GAAP earnings. They found that the quality of non-GAAP earnings has increased as non-GAAP earnings exclusions are more transitory after the implementation of the regulation. Furthermore, the firms that have stopped disclosing non-GAAP earnings after the regulation was implemented had exclusions that were of lesser quality; i.e. were less transitory.
However, when decomposing exclusions in special and other items Kolev et al. (2008) found that the quality of other items has increased while the quality of special items has decreased.
In sum, managers appear to use alternative earnings less opportunistically under regulation, as indicated by the decrease in the amount of non-GAAP disclosure, exclusion magnitude, probability of meeting or beating analyst forecasts, and misleadingness of the presentation of non-GAAP earnings. However, non-opportunistic use also seem to have decreased as firms also excluded fewer items that are not part of core performance. Overall, the informativeness of alternative earnings has improved after regulation implementation.
2.4 Effects of mandatory disclosure of alternative earnings measures
There are two studies that used the unique South African setting for new insights in alternative earnings measures. The study by Venter et al. (2013) investigates whether the non-recurring items that are subtracted from IFRS earnings to arrive at headline earnings are mispriced. Previous research has pointed out that these components are mispriced in the US, but Venter et al. (2013) study this in the South African setting, where disclosing an
alternative earnings measure is mandatory. They find that only the cash flow component of earnings is mispriced, while other components do not show this anomaly. Furthermore, they find that the headline earnings figure is a more reliable predictor for future earnings than headline earnings exclusions. More specifically, the likelihood that the current level of earnings will recur in feature periods of headline earnings was significantly higher for headline earnings than the for headline earnings exclusions.
Another study by Venter et al. (2014) examines the value relevance of the different earnings components, where the value relevance is defined as the investors’ decisions reflected by the share prices as a response to these headline earnings similar to the
informativeness of alternative earnings measures as discussed above. Their findings are that headline earnings are indeed value relevant, and the reconciliations made to arrive at headline earnings starting from IFRS earnings are value irrelevant. This is in line with the findings in other countries where disclosing an alternative earnings measure is not mandatory.
11
Furthermore, Venter et al. (2014) also analyse whether the value relevance is different for firms with positive earnings compared to loss firms. They find that headline earnings is value relevant for both loss firms and firms with positive earnings, where headline earnings from loss firms had a negative effect on the share price and headline earnings from positive profit firms had a positive effect on the share price. Lastly, Venter et al. (2014) also analyse the difference in value relevance between firms with extreme headline earnings exclusions, defined as the lower and upper 20% of headline earning exclusions observations, compared to less extreme headline earnings exclusions, which is the 60% of the observations that is left. They find that headline earnings is a better measure of performance than IFRS earnings for firms with high headline earnings exclusions, but this is not the case for firms with low headline earnings exclusions.
In conclusion, Venter et al. (2013) and Venter et al. (2014) have studied the mandatory disclosure setting in South Africa. Venter et al. (2013) show that only the cash flow
component of non-recurring items is mispriced in this setting, in contrast to studies in the US setting that find that all components are mispriced. Venter et al. (2014) show that the earning components are value relevant for both loss firms and firms with positive earnings and exclusion are value irrelevant, in line with research studying the voluntary disclosure setting.
2.5 Voluntary versus mandatory disclosure
Some empirical studies show different findings in a mandatory setting than in a voluntary setting when looking at the disclosure of an alternative earnings measure, whereas others do not. Venter et al. (2014) find that in South Africa non-GAAP earnings is an
informative measure, which is in line with findings from the US and France (Aubert, 2010; Bhattacharya et al., 2003). In contrast, Venter et al. (2013) find that the earnings components for South African companies are not mispriced, although other research in the US finds that non-recurring items that are excluded from GAAP earnings to arrive at non-GAAP earnings on a voluntary basis are mispriced (Doyle, Lundholm & Soliman, 2003). This shows that although some findings concerning the mandatory disclosure of pro forma earnings may hold for voluntary disclosure of pro forma earnings, others may not.
Empirical research that examines the difference between voluntary and mandatory disclosure for other subjects also shows different results for the two settings. For example, Butler et al. (2007) have examined the effect of financial reporting frequency on timeliness, which is the time it takes for the information to be reflected in the stock price of a firm. They find that firms that voluntarily increased their reporting frequency experienced increased
12
timeliness while firms that were obliged to increase reporting frequency did not experience any change in timeliness (Butler et al., 2007). Another example is a study performed by Polinsky and Shavell (2010), in which they study the effect of mandatory and voluntary disclosure on product risk. More specifically, they find that under voluntary disclosure firms gather more information about product risk, which could lead to higher social welfare. However, this is under the condition that the firms are not liable for any harm their product has caused. Under strict liability there is no difference between voluntary and mandatory disclosure as the product risk has now become irrelevant for consumers because they will be compensated for any harm. Although the previous two studies showed that mandatory disclosure can have negative side-effects, the study by Ferrell (2007) finds that mandatory disclosure can also have positive side-effects. His study examines the effect of the mandatory disclosure requirements that were imposed in 1964 on the over-the-counter market on stock volatility and returns. He found that stock returns were less volatile after the mandatory disclosure requirements, and stocks enjoyed positive abnormal returns as well.
In sum, mandatory disclosure can have a different impact on variables studied under voluntary disclosure in a positive or negative way, implying that the difference in disclosure policy is important to research.
2.6 Theory on additional disclosure
Disclosure theory also points out that firms that voluntarily engage in additional disclosure are the ones that disclose good news, and the firms that do not disclose attempt to withhold bad news. Milgrom (1981) showed this by modelling a game in which an interested party tries to influence a decision maker by selectively providing information on the decision to be made. The result was that, under the assumptions of rational expectations, costless communication, and the ability to detect the withholding of information, the decision maker judges every piece of information that is withheld as unfavourable. As a result, it is optimal for the interested party to fully disclose all information. Grossman (1981) models a similar situation in which sellers can disclose information about the quality of their product, and furthermore provide a warranty that the information given is actually true. He also finds that when sellers engage in less than full disclosure, the consumers will assume the worst possible quality of the information that is not disclosed. As a result, the seller’s best strategy is to fully disclose all information.
The effect of changing from voluntary to mandatory disclosure in theory has different effects depending on how informed investors are and/or how easy it is to understand the
13
information provided by the disclosure. Fishman and Hagerty (2003) model situations in which high quality and low quality sellers disclose information voluntarily and mandatorily in the case of many, an intermediate number, or relatively few informed investors. In the case of many informed investors, and in the case that the information is easy to comprehend, there is an equilibrium for both high and low quality sellers to disclose, in line with the theory by Grossman (1981) and Milgrom (1981). When there is an intermediate number of informed investors there are equilibria both for voluntary and mandatory disclosure in which there is disclosure, and in which there is no disclosure. However, Fishman and Hagerty (2003) argue that the equilibria in which there is no disclosure are implausible in practice. This is because the equilibrium relies on the assumption that in the case of non-disclosure, both types of sellers can charge the same certain maximum price for their product, above which the sellers think the product is overpriced. However, there is also an equilibrium in which the high quality seller can sell the product above this price, therefore the buyers do not have to
consider the product overpriced. Hence, making disclosure mandatory would create additional value for investors as it is optimal to disclose for both types of sellers and no new equilibria will arise from implementing mandatory disclosure. Lastly, when there are few informed investors compared to informed investors, there will be no disclosure as high quality sellers do not benefit from disclosure and low quality sellers can never be better off disclosing their low quality when the sellers can voluntarily choose their disclosure behaviour. Making disclosure mandatory will make informed investors better off, as they are able to use the additional information provided by disclosure effectively. However, there is no gain or loss for uninformed investors and sellers are worse off because the informed sellers will not buy from the low quality seller anymore as they now know the product is overpriced.
However, the assumptions of these disclosure theories are not realistic and are incomplete, which may lead to other results than full disclosure for all firms in the case of informed investors and no disclosure in the case of uninformed investors. Healy and Palepu (2001) sum up six factors provided in theoretical and empirical studies that can alter the degree to which management engages in voluntary disclosure: capital market transactions, corporate control contests, stock compensation, litigation, proprietary costs, and management signalling. From these factors, litigation cost and proprietary costs can serve as a motivation to not fully disclose all information available to management. More specifically, litigation costs arise because unexpected forecast errors may be seen as deliberate management bias for which the company is then sued. Proprietary costs can arise because full disclosure would also mean disclosure on a firm’s competitive advantage, which the firm will consequently
14
lose. This would also mean that not disclosing certain information does not necessarily mean that it is withheld because it is bad news, but may also be too costly to disclose. However, as withholding information does not signal bad news anymore, it provides an opportunity for firms to withhold bad news without the investor being able to identify this. This allows the low and high quality sellers as modelled by Fishman and Hagerty (2003) to behave differently from each other.
In conclusion, Milgrom (1981) and Grossman (1981) use models to show that it is optimal for both high and low quality firms to disclose all information. However, Fishman and Hagerty (2003) argue that this equilibrium relies on unrealistic assumptions and Healy and Palepu (2001) mention several factors that can alter firms’ disclosure decisions such that it is not optimal for good quality firms to fully disclose. As a result, low quality firms can choose not to disclose bad information, without investors being able to identify the firm as low quality.
2.7 Hypotheses development
When considering an investment in a certain firm it is important to be able to interpret the earnings measures correctly. Similarly, it is important for policymakers to know whether making alternative earnings disclosure mandatory harms or benefits the informativeness of alternative earnings figures and whether it affects the opportunistic use of alternative earnings numbers found by several researchers in a voluntary disclosure setting.
Research on the timeliness of disclosure, social welfare, stock volatility, and abnormal stock returns in a mandatory versus voluntary disclosure consistently shows differences between the effects of the two forms of disclosure. Butler et al. (2007) show that mandatory disclosure harmed the timeliness of disclosure and Polinsky & Shavell (2010) show
mandatory disclosure decreases social welfare. However, the effects of mandatory disclosure are not always detrimental. Ferrell (2007) has found that mandatory disclosure requirements decreases stock volatility and lead to positive abnormal stock returns. The research executed on the mandatory disclosure setting in South Africa does not consistently show differences with findings in a voluntary regime. The study by Venter et al. (2013) on the pricing of earnings components in a mandatory setting shows different results compared to similar research in voluntary disclosure settings, whereas the findings of the study by Venter et al. (2014) on value relevance are in line with those of other research in voluntary disclosure settings. The fact that one study on mandatory disclosure of alternative earnings is in line with research on voluntary disclosure and the other one is not shows that mandatory disclosure of
15
alternative earnings can alter the findings of voluntary disclosure research, but does not necessarily have to.
As the empirical evidence is mixed with respect to whether mandatory disclosure changes findings on research on voluntary disclosure, and also whether this effect is positive or negative, I use the theoretical literature on additional disclosure to posit the hypotheses for this research. Due to the factors as summed up by Healy and Palepu (2001), not disclosing certain information does not necessarily mean that the information is unfavourable, in contrast to the beliefs of Milgrom (1981) and Grossman (1981). As a result, low quality firms can choose not to disclose without signalling investors that the firms performed badly. Therefore, I expect that firms with an alternative earnings measure showing bad performance will choose not to disclose this measure. This expectation is also on the research by Heflin and Hsu
(2008), Entwistle et al. (2006) and Kolev et al. (2008) where they compare regulating non-GAAP disclosure to not regulating it, as the expectation is an explanation for their findings. All these studies found a decrease in the amount of firms reporting non-GAAP earnings, because it was not in their best interest anymore to disclose non-GAAP earnings, and in the magnitude of exclusions. However, because several studies show that the alternative earnings figure is an informative measure for future performance, it is also a possibility that the
alternative earnings measure is higher because the firms actually performed better than GAAP earnings show and not because firms choose to exclude and include items in such a way that it increases their alternative earnings measure. In this case, making alternative earnings a
mandatory measure and providing detailed guidelines on what to exclude would not alter the difference between GAAP earnings and the alternative earnings measure. Despite this possibility, I choose to rely on the theoretical literature in forming my expectations.
Therefore, I expect that most firms that do not disclose an alternative earnings measure do so because it is unfavourable for them to do so, and thus, mandatory disclosure of an alternative earnings measure will result in a less positive or more negative difference in reconciliation between GAAP earnings and alternative earnings.
H1: Ceteris paribus, the reconciliation between GAAP earnings and alternative earnings measures is smaller for firms that are obliged to report alternative earnings measures compared to firms that voluntarily disclose alternative earnings measures.
Venter et al. (2014) have found that headline earnings are more informative than GAAP earnings in South-Africa, but not whether mandatory disclosure makes headline earnings
16
more, less or similarly informative as compared to voluntary disclosure. Heflin and Hsu (2008) and Kolev et al. (2008) find that firms reported less non-opportunistic special item exclusions because the disclosure cost increased when Regulation G was implemented, which resulted in a lower quality of special item exclusions. In South Africa’s mandatory disclosure environment with a detailed outline of what exactly should be included in special items and what not, these firms would be mandated to disclose these special items again, resulting in an increase in the quality of alternative earnings. Therefore, I expect that mandatory disclosure of an alternative earnings measure will result in a relatively more informative alternative earnings measure.
H2: Ceteris paribus, the extent to which the alternative earnings measure is more informative than GAAP earnings is larger for firms that are obliged to report alternative earnings
measures compared to firms that voluntarily disclose alternative earnings measures 3. Sample Selection and Research Design
3.1 Sample Selection
The research question is answered by comparing the effects of disclosure of non-GAAP earnings in the mandatory South-African setting with the voluntary setting in the US and Australia. The US and Australia were chosen as reference countries because there are regulations with regard to non-GAAP disclosure in these countries, in contrast to European countries for example, such that this research is not influenced by the difference in exclusions between a regulated and a non-regulated environment as found by Heflin and Hsu (2008), Entwistle et al. (2006) and Kolev et al. (2008). Also, South Africa, Australia and the US are categorized to have similar accounting practices, specified as the British-American
accounting cluster by Mueller et al. (1994). Furthermore, an international study by Hung (2000) on the accounting standards in relation to value relevance of financial statements shows that indexes on legal system, tax-book conformity and antidirector rights are
approximately the same for South Africa and the US and exactly the same for South Africa and Australia. Therefore, although the cultural differences might be significant, the financial reporting systems are comparable. Furthermore, Australia and South Africa both use IFRS Accounting Standards, while the US uses US GAAP, such that the difference in the earnings measures as a result of different accounting standards can be controlled for by selecting three countries.
17
For these countries data is hand collected from the annual reports or earnings
announcements for the largest firms listed in South-Africa, Australia, and the US, similar to the studies by Entwistle, Feltham and Mbagwu (2005) and Black et al. (2017). While this brings about a bias towards large and successful firms, it contributes to investors and policy makers made by this research as the sample consists of firms that have the largest percentage of investments and that are most affected by accounting policies in their reporting. The largest firms are selected by obtaining the composition and market capitalization of the JSE all shares, ASX Top 300 and S&P500 from Datastream and sorting them per stock exchange from largest market capitalization to smallest. Not all American and Australian companies report on non-GAAP earnings, those that do not are excluded from the sample before collecting the data. Shares that are listed both on the JSE and on the ASX or NYSE or NASDAQ are excluded from the sample as well.
The amount of companies listed in the US, Australia and South-Africa approximately adds up to 7500 firms. In order to be able to obtain significant results with a 95% confidence level I have to collect a minimum of 366 observations according to sample size calculation statistics. To be sure, I collected a final sample of 400 firm-year observations. As the
regulation in Australia was implemented in December 2011, the dataset consists of firm-year observations over 5 years from 2012 onwards up until 2016. The sample selection details per country are shown in Table 1. Furthermore, sales is collected from a year before the time period to calculate sales growth, and operating cash flows from the year after the time period to have data on future operating cash flows for the last year. To obtain 400 firm-year
observations over this time period, I collected data from 83 firms. As shown in Table 1 the panel is unbalanced as some US firms did not report on non-GAAP earnings every year, such that more firms were needed to obtain the same amount of firm-years. The annual reports are retrieved from the firms’ individual websites or from the websites of the stock exchanges they are listed on. The variables industry, total assets, total liabilities, sales, total equity, and net operating cash flows are obtained from Compustat Global. The data that was obtained in local currency was converted to US Dollar using the currency translation rate obtained from
Compustat North America. Market value of equity and firms firm age are obtained from Datastream. Other variables such as headline earnings are extracted from the individual annual reports by hand. Also, firm age was collected from the individual firms’ websites for the firms that had no data available on Datastream for firm age.
18 Table 1: Sample Selection Details
Description SA US AUS Total
n largest firms 32 36 39 107
Drop because data is not available (1) (1)
Drop because cross-listed (4) (1) (5)
Drop because no disclosure of alternative
earnings measure (7) (11) (18)
Total n firms 27 29 27 83
Total n firm-years 132 135 133 400
3.2 Research Design
Hypothesis 1 predicts that the reconciliation between GAAP earnings and alternative earnings measures is smaller for firms that are obliged to report alternative earnings measures compared to firms that voluntarily disclose alternative earnings measures. This hypothesis is tested according to regression equation 1, for every company i in period t:
𝐷𝐼𝐹𝐹𝑖𝑡 = ∝𝑖𝑡+ 𝛽1𝑀𝐴𝑁𝐷𝑖𝑡+ 𝛾𝐶𝑜𝑛𝑡𝑟𝑜𝑙𝑠 + 𝜀𝑖𝑡 (1)
The dependent variable in the first hypothesis, the difference in reconciliations (DIFF), is measured by three different proxies that are based on the measures used by Entwistle et al. (2005) and Epping and Wilder (2011). However, in contrast to the study by Entwistle et al. (2005) I study the difference in the alternative earnings measure using three multivariate multiple regressions instead of a comparison of means, as this allows me to control for covariates. The first proxy for DIFF is the absolute difference (ABS) between the GAAP earnings measure and the alternative earnings measure, measured by
𝐴𝑙𝑡𝑒𝑟𝑛𝑎𝑡𝑖𝑣𝑒 𝑒𝑎𝑟𝑛𝑖𝑛𝑔𝑠−𝐺𝐴𝐴𝑃 𝑒𝑎𝑟𝑛𝑖𝑛𝑔𝑠
|𝐺𝐴𝐴𝑃 𝑒𝑎𝑟𝑛𝑖𝑛𝑔𝑠| . The second proxy is the number of reconciliations (NUMB) made to GAAP earnings to arrive at non-GAAP earnings. This excludes reconciliations involving tax and non-controlling interest remeasurements, as these are a result of other reconciliations. The third proxy is the probability of the alternative earnings measure being higher than GAAP earnings (HIGH), where HIGH is a dummy variable that takes on the value of 1 when the alternative earnings measure is higher than GAAP earnings and zero otherwise. The first two dependent variables are tested using an OLS regression and the third one using a logistic model, both using robust standard errors clustered by firm.
19
The independent variable of interest is a dummy variable (MAND) that takes on the value of 1 if disclosure of non-GAAP earnings is mandatory, and zero otherwise.
Several control variables are incorporated in the OLS and logistic regressions. I include the control variables for IFRS accounting standards, size, leverage, sales growth, high-tech industry, and fiscal year following Epping and Wilder (2011), as they also study adjustment characteristics and reconciliation quality between two populations. The dummy variable for IFRS accounting standards (IFRS) is incorporated because previous research shows that US GAAP compared to IFRS accounting standards can lead to different reporting outcomes. For example, Collins, Pasewark, and Riley (2012) have found that firms are more likely to classify leases as operating leases than under IFRS, and they also found some
evidence that there is less dispersion in lease classification under IFRS. The variable IFRS has a value of 1 if the firm applies IFRS accounting standards and 0 otherwise. Total assets (SIZE) is incorporated to control for firm size, as common in disclosure research. Heflin and Hsu (2012) for instance found that larger firms, measured by total assets, have lower
exclusions in magnitude compared to smaller firms. Similar to this research, I define SIZE as the log of total assets at the end of the firm’s fiscal year. Leverage (LEV), measured by the debt-to-equity ratio, is incorporated as Epping and Wilder (2011) show that more leveraged firms tend to make a higher number of adjustments, and also a higher number of income-increasing adjustments. LEV is calculated by dividing total liabilities by shareholders’ equity. Sales growth (GROW) is defined as the change in sales from previous year, and is included because low-growth firms tend to be more aggressive in the magnitude and the number of exclusions made to arrive at non-GAAP earnings (Epping & Wilder, 2011). Furthermore, as Bowen, Davis, and Matsumoto (2005) showed that high-tech industry firms are more
aggressive in their non-GAAP earnings disclosure, I also control for high-tech industries. This is done using a dummy variable (IND) which takes on 1 for the following high-tech industries based on SIC code, and 0 otherwise: Drugs (283); Computer and Office Equipment (357); Electrical Machinery and Equipment, Excluding Computers (360); Electrical Transmissions and Distribution Equipment (361); Electrical Industrial Apparatus (362); Audio, Video Equipment, Audio Receiving (365); Communication Equipment (366); Electronic
Components, Semiconductors (367); Computer Hardware (368); Telephone Communications (481); Computer Programming, Software, Data Processing (737); Research, Development, Testing Services (873). A dummy variable for every fiscal year (YEAR) in the sample is included to control for any external effects or time trends that influence alternative earnings of
20
every firm in the same year. All continuous dependent and control variables are winsorized at the 1% and 99% level to eliminate outliers.
Hypothesis 1 is confirmed if coefficient β1 is negative and significant in Model (1). In
line with previous research, I expect the coefficients for the control variables SIZE and
GROW to be negative and the coefficients for the control variables IND and LEV to be
positive. For IFRS and the coefficients for every YEAR I do not expect a certain outcome for the coefficient, as there is no study on alternative earnings measures that incorporates this variable.
Hypothesis 2 predicts that the extent to which the alternative earnings measure is more informative than GAAP earnings is larger for firms that are obliged to report alternative earnings measures compared to firms that voluntarily disclose alternative earnings measures. This is tested according to regression equation 2, that is based on the study by Bentley et al. (2017), for every company i in period t:
𝐹𝑂𝐶𝑖,𝑡+1 = ∝𝑖𝑡+ 𝛽1𝐴𝐿𝑇𝐸𝐴𝑅𝑁 + 𝛽2𝐸𝑋𝐶𝐿 + 𝛽3𝑀𝐴𝑁𝐷 + 𝛽4𝑀𝐴𝑁𝐷𝑖𝑡∗ 𝐸𝑋𝐶𝐿 +
𝛽5𝐴𝐿𝑇𝐸𝐴𝑅𝑁𝑖𝑡∗ 𝐸𝑋𝐶𝐿 + 𝛾𝐶𝑜𝑛𝑡𝑟𝑜𝑙𝑠 + 𝜀𝑖𝑡 (2)
The dependent variable in the second hypothesis, the informativeness of non-GAAP earnings (FOC), is measured by future operating cash flows as in the research performed by Leung and Veenman (2017) and Doyle et al. (2003), as informing investors about future cash flow is seen as one of the main objectives of earnings measures. Future operating cash flows is defined as the operating cash flows in the next period.
The independent variable mandatory (MAND) is defined in the same way for
hypothesis 2 as it is defined for hypothesis 1. The independent variables alternative earnings number (ALTEARN) and alternative earnings exclusions (EXCL) are incorporated, similar to the studies performed by Bentley et al. (2017), Whipple (2015), Leung and Veenman (2017), Doyle et al. (2003) and Kolev et al. (2008). Alternative earnings exclusions is calculated by subtracting the alternative earnings number from GAAP earnings.
For testing the second hypothesis, this paper includes control variables for IFRS accounting standards, size, sales growth, book-to-market assets ratio, firm age, and fiscal year based on other informativeness studies (Doyle et al., 2003; Leung & Veenman, 2017;
Whipple, 2015; Kolev et al., 2008). The control variables IFRS accounting standards (IFRS), firm size (SIZE), fiscal year (YEAR), and sales growth (GROW) are defined as in the previous
21
regression. Following Bentley et al. (2017) I also incorporate the book-to-market assets (BTM) ratio, defined as the book value of equity divided by the market value of equity. Furthermore, I incorporate firm age (AGE) to control for effects of a firm’s maturation on the amount and use of future cash flows and earnings (Kolev et al., 2008). AGE is defined as the natural logarithm of the amount of years since the company has been founded. Accruals, although used in other research, is intentionally left out as also done in Whipple (2015), because this would mean the accruals are incorporated into the regression twice; as a control variable and as the accrual part of the exclusion. The model is tested with an OLS regression using robust standard errors clustered by firm. All continuous dependent, independent variables are winsorized at the 1% and 99% level to, as before, eliminate outliers.
For confirming hypothesis 2 I look at the interaction term of mandatory and exclusions in Model (2), as the quality of alternative earnings measures is higher when the association between exclusions and future firm performance is lower (Bentley et al., 2017). Thus, if the coefficient β4 is negative and significant hypothesis 2 is confirmed. Furthermore, in line with
the findings of previous studies I expect the coefficient β1 to be positive and β2 to be negative
(Doyle et al., 2003; Bentley et al., 2017; Kolev et al., 2008). I do not expect β3 to be
significant, as I do not expect that mandating disclosure of an alternative earnings measure has a direct effect on future operating cash flows. I do expect β5 to be positive and significant,
as I expect alternative earnings measures to be more informative in a mandatory disclosure setting. For the control variables, I expect SIZE and AGE to be positively related to net operating cash flows, and GROW and BTM to be negatively related. As before, I have no expectations with regard to the variables IFRS and YEAR.
4. Results
4.1 Descriptive Statistics
Shown in Panel A of Table 2, the full sample consists of 400 firm-year observations from 83 firms. The sample of 400 observations consists of 79 observations from the year 2012, 81 observations from the years 2013, 2014, and 2015, and 78 observations from the year 2016. The univariate descriptive statistics in Panel A support the opportunistic
appearance of alternative earnings, as the alternative earnings measure is more than GAAP earnings in 68.8% of the observations. Furthermore, the average of exclusions made from GAAP earnings is -633 million dollars (i.e. the alternative earnings measure is 633 million
22 Table 2: Descriptive statistics
Panel A: Descriptive statistics full sample
Variable N Mean Std. Dev. Minimum Maximum
ALTEARN 400 4379.302 5720.185 -529 24324.5 EXCL 400 -632.650 1636.896 -7218.86 3876.78 ABS 400 0.532 1.664 -0.585 12.315 NUMB 400 4.628 3.036 0 12.5 HIGH 400 0.688 0.464 0 1 FOC 400 5969.326 9653.398 -142 42585.5 MAND 400 0.33 0.471 0 1 SIZE 400 10.403 1.726 6.896 14.559 IFRS 400 0.663 0.473 0 1 LEV 400 3.220 4.431 -4.050 16.309 IND 400 0.218 0.413 0 1 GROW 400 -0.012 0.193 -0.582 0.867 BTM 400 0.581 0.467 -0.034 2.566 AGE 400 4.061 0.885 1.609 5.308
Panel B: Descriptive statistics mandatory versus voluntary disclosure regime
Voluntary Mandatory Diff.
t-test1 Variable N Mean Std. Dev. N Mean St. Dev
ALTEARN 268 6103.898 6261.787 132 877.849 1142.469 9.505*** EXCL 268 -898.135 1877.755 132 -93.636 733.179 -4.745*** ABS 268 0.556 1.589 132 0.485 1.812 0.398 NUMB 268 4.519 3.241 132 4.848 2.567 -1.022 HIGH 268 0.739 0.440 132 0.583 0.495 3.186*** FOC 268 8225.426 11042.77 132 1388.686 1933.103 7.056*** SIZE 268 10.896 1.630 132 9.40 1.466 8.906*** IFRS 268 0.496 0.501 132 1 0 -11.546*** LEV 268 3.167 0.436 132 3.329 4.277 -0.345 IND 268 0.254 0.436 132 0.144 0.352 2.516** GROW 268 -0.015 0.161 132 -0.006 0.245 -0.453 BTM 268 0.553 0.438 132 0.638 0.519 -1.718* AGE 268 4.164 0.890 132 3.851 0.820 3.369***
23 Table 2: Descriptive statistics (continued)
1 Student t-test on the difference of the means. Significance at a level of 10%, 5%, 1% level is indicated by
*,**,*** respectively, based on a two-tailed t-test.
ALTEARN is the alternative earnings measure, in millions of dollars. EXCL are the adjustments made to arrive
from GAAP earnings to the alternative earnings measure, in millions of dollars. ABS is the absolute difference between GAAP earnings and the alternative earnings measure, scaled by the absolute value of GAAP earnings.
NUMB is the number of reconciliations made to arrive from GAAP earnings to the alternative earnings measure. HIGH takes on the value of 1 if the alternative earnings measure is higher than the GAAP earnings number, and
zero otherwise. FOC is net operating cash flows at time t+1, in millions of dollars. MAND takes on the value of 1 is firms are mandated to disclose an alternative earnings measure, and zero otherwise. SIZE is the natural logarithm of total assets, in millions of dollars. IFRS takes on the value of 1 if a firm prepares its annual reports in accordance with IFRS, and zero otherwise. LEV is the debt-to-equity ratio. IND takes on the value of 1 if the firm has a high-tech industry SIC code, as defined in section 3.2, and zero otherwise. GROW is sales growth compared to the previous year. BTM is the book-to-market ratio. AGE is the natural logarithm of the amount of years since the company has been founded.
dollars higher), resulting in an average increase of 53.2% from alternative earnings to GAAP earnings.
Panel B of Table 1 divides the full sample into two sub-samples: firms that voluntarily disclose an alternative earnings measure and firms that are mandated to disclose alternative earnings. In line with my expectations in hypothesis 1, comparing the difference in means with a student t-test shows that the amount of expenses that are excluded is significantly lower in the mandatory disclosure setting compared to the voluntary disclosure setting (p < 0.001, one-sided). According to my results, firms in a mandatory disclosure setting on average exclude 94 million USD worth on expenses, while firms in a voluntary disclosure setting on average exclude 898 million USD worth on expenses. Furthermore, as expected in hypothesis 1, the alternative earnings measure is significantly less often higher than GAAP/ in the mandatory disclosure setting compared with the voluntary disclosure setting (p = 0.001, one-sided). More specifically, under voluntary disclosure firms report an alternative earnings number that is higher than GAAP earnings in 73.9% of the observations, while for mandatory disclosure this happens in 58.3% of the observations. However, for the other proxies ABS and
NUMB I find no significant difference using univariate analysis. The descriptive evidence
also shows that the firms under mandatory disclosure are significantly smaller (p < 0.001, two-sided) which is the result of choosing the US as part of the sample, as this is a large country with large firms compared to South Africa. This finding is also a possible cause for
24
the increase in expense exclusions, therefore multivariate analysis is necessary to be able to make correct inferences.
Table 3 presents the Pearson correlation matrix, providing the correlation coefficients between the variables for further univariate analysis and for indicating
multicollinearity. The correlations show similar results as the descriptive statistics; there is a positive significant relationship between the mandatory disclosure of alternative earnings and the probability that alternative earnings is higher than the GAAP earnings measure, but the other proxies ABS and NUMB do not show this effect. Furthermore, there is a positive significant relationship between exclusions and mandatory disclosure. The only coefficient that exceeds 0.8, which is the threshold for multicollinearity, is the correlation between the dependent variable future net operating cash flows and the alternative earnings measure (ρ = 0.842, p < 0.001, two-sided). As multicollinearity is only a potential issue amongst the independent variables, this correlation only indicates that the alternative earnings measure is a good predictor for future performance, in line with previous research (e.g. Bhattacharya et al, 2003; Venter et al., 2014). The correlation coefficients between IFRS and the alternative earnings measure (ρ = -0.748, p < 0.001, two-sided), and between IFRS and net future operating cash flows (ρ = -0.734, p < 0.001, two-sided) are relatively high as well. The latter is probably caused by the fact that US firms are bigger, and therefore also have higher future operating cash flows, compared to Australian and South African firms that report under IFRS. These variables are checked for multicollinearity by using the variance inflation factor [VIF] test based on the OLS regression equation (2).
The results of the test on OLS regression equation (2) are shown in Table 4. A VIF value above 5 is seen as a sign of multicollinearity. As every VIF value is lower than 3.07, the VIF test does not indicate a sign of multicollinearity. Therefore, I can conclude the coefficient estimates will not respond to small changes in the model or in the data, and the individual predictors are reliable.
4.2 Effect on the difference in reconciliations
Hypothesis 1 expects the difference in reconciliations from GAAP earnings to the alternative earnings measure to be lower in a mandatory disclosure than in a voluntary disclosure setting. The results are shown in Table 5 Column 1 to 3 for the three different proxies for the difference in reconciliations. For the first proxy, the probability of the
alternative earnings measure being higher than GAAP earnings (HIGH), there is statistically significant evidence at a 10% level that in a mandatory disclosure setting there are lower odds
25 Table 3: Pearson correlation matrix
ALT-
EARN EXCL ABS NUMB HIGH FOC MAND SIZE IFRS LEV IND GROW BTM AGE
ALTEARN 1 EXCL -0.406*** 1 ABS -0.069 -0.251*** 1 NUMB -0.112** -0.015 0.054 1 HIGH 0.182*** -0.407*** 0.268*** 0.043 1 FOC 0.842*** -0.360*** -0.039 -0.167*** 0.109** 1 MAND -0.430*** 0.231*** -0.020 0.051 -0.158*** -0.325*** 1 SIZE 0.696*** -0.308*** -0.059 0.069 0.128** 0.578*** -0.408*** 1 IFRS -0.748*** -0.336*** 0.074 0.040 -0.173*** -0.715*** 0.501*** -0.544*** 1 LEV 0.066 0.047 -0.067 0.126** 0.056 -0.008 0.017 0.555*** 0.138*** 1 IND 0.437*** -0.283*** -0.088* -0.107** 0.055 0.343*** -0.125** 0.189*** -0.367*** -0.121** 1 GROW 0.021 0.099** -0.165*** -0.058 -0.067 0.045 0.023 0.012 0.022 0.063 0.053 1 BTM -0.078 -0.129*** 0.258*** 0.069 0.051 -0.002 0.086* 0.109** 0.224*** 0.011 -0.219*** -0.055 1 AGE 0.278*** -0.192*** -0.026 0.082* 0.179*** 0.181*** -0.167*** 0.431*** -0.254*** 0.262*** -0.008 -0.076 -0.078 1 For every variable there are 400 observations from 83 firms the years 2012-2016. Significance at a level of 10%, 5%, 1% level is indicated by *,**,*** respectively, based on a two-tailed t-test.
ALTEARN is the alternative earnings measure, in millions of dollars. EXCL are the adjustments made to arrive from GAAP earnings to the alternative earnings measure, in millions of dollars. ABS is the absolute difference between GAAP earnings and the alternative earnings measure, scaled by the absolute value of GAAP earnings. NUMB is the number of reconciliations made to
arrive from GAAP earnings to the alternative earnings measure. HIGH takes on the value of 1 if the alternative earnings measure is higher than the GAAP earnings number, and zero otherwise.
FOC is net operating cash flows at time t+1, in millions of dollars. MAND takes on the value of 1 is firms are mandated to disclose an alternative earnings measure, and zero otherwise. SIZE is
the natural logarithm of total assets, in millions of dollars. IFRS takes on the value of 1 if a firm prepares its annual reports in accordance with IFRS, and zero otherwise. LEV is the debt-to-equity ratio. IND takes on the value of 1 if the firm has a high-tech industry SIC code, as defined in section 3.2, and zero otherwise. GROW is sales growth compared to the previous year. BTM is the book-to-market ratio. AGE is the natural logarithm of the amount of years since the company has been founded.
Table 4: VIF test regression equation (2)
Variable VIF 1/VIF
ALTEARN 3.33 0.300 EXCL 1.38 0.727 MAND 1.41 0.710 EXCL*MAND 1.16 0.860 IFRS 2.70 0.370 SIZE 2.47 0.405 GROW 1.10 0.910 BTM 1.28 0.779 AGE 1.29 0.776 YEAR 2013 1.66 0.603 YEAR 2014 1.64 0.609 YEAR 2015 1.73 0.578 YEAR 2016 1.62 0.619
This variance inflation factor test has been executed on the OLS regression on future net operating cash flows (regression equation 2). ALTEARN is the alternative earnings measure, in millions of dollars. EXCL are the adjustments made to arrive from GAAP earnings to the alternative earnings measure, in millions of dollars.
MAND takes on the value of 1 is firms are mandated to disclose an alternative earnings measure, and zero
otherwise. EXCL*MAND is an interaction variable multiplying EXCL with MAND. SIZE is the natural logarithm of total assets, in millions of dollars. IFRS takes on the value of 1 if a firm prepares its annual reports in
accordance with IFRS, and zero otherwise. IND takes on the value of 1 if the firm has a high-tech industry SIC code, as defined in section 3.2, and zero otherwise. GROW is sales growth compared to the previous year. BTM is the book-to-market ratio. AGE is the natural logarithm of the amount of years since the company has been founded. YEAR 2013, YEAR 2014, YEAR 2015, YEAR 2016 are dummy variables taking on the value of 1 when the observations is made in fiscal year 2013, 2014, 2015, and 2016 respectively, and zero otherwise.
of the alternative earnings measure being higher than the GAAP earnings measure (β1 =
-0.450, p = 0.080, one-sided). More specifically, the odds of the alternative earnings measure being higher than GAAP earnings for firms in a mandatory disclosure setting is e-0.450 = 0.637
times that of firms in a voluntary disclosure setting. This confirms hypothesis 1 as, when keeping all else equal, firms in a mandatory disclosure setting report higher alternative earnings than GAAP earnings 36.3% fewer times compared to firms in a voluntary disclosure setting. Furthermore, the result is economically significant as 36.3% of all the annual reports of a stock exchange concerns a large amount of firm. However, the other two proxies, the
27 Table 5: Difference in reconciliation results
Dependent variable: DIFF
Variable Expected sign HIGH (1) ABS (2) NUMB (3) MAND - -0.450* (0.321) -0.214 (0.225) 0.452 (0.753) IFRS ? -0.991* (0.586) 0.554** (0.278) 0.012 (1.034) SIZE - -0.143 (0.180) 0.091 (0.088) 0.137 (0.352) IND + 0.078 (0.376) -0.262* (0.163) -0.742 (0.645) LEV + 0.080* (0.056) -0.053** (0.026) 0.049 (0.107) GROW - -0.933* (0.632) -1.255*** (0.372) -0.938 (0.968)
YEAR 0 YES YES YES
CONSTANT + 3.047* (2.083) -0.653 (0.961) 3.056 (3.849) Observations 400 400 400 R2 0.0451 0.060 0.034
1: Pseudo R2 for this logistic regression model.
All regressions are OLS regressions, except the regression on variable HIGH, which is a logistic regression. All regression use robust standard errors clustered by firm. Significance at a level of 10%, 5%, 1% level is indicated by *,**,*** respectively, based on a one-tailed test for expected signs +/- and two-tailed t-test for the sign ?.
ABS is the absolute difference between GAAP earnings and the alternative earnings measure, scaled by the
absolute value of GAAP earnings. NUMB is the number of reconciliations made to arrive from GAAP earnings to the alternative earnings measure. HIGH takes on the value of 1 if the alternative earnings measure is higher than the GAAP earnings number, and zero otherwise. MAND takes on the value of 1 is firms are mandated to disclose an alternative earnings measure, and zero otherwise. SIZE is the natural logarithm of total assets, in millions of dollars. IFRS takes on the value of 1 if a firm prepares its annual reports in accordance with IFRS, and zero otherwise. LEV is the debt-to-equity ratio. IND takes on the value of 1 if the firm has a high-tech industry SIC code, as defined in section 3.2, and zero otherwise. GROW is sales growth compared to the previous year. YEAR consists of four dummy variables with a value of 1 when the observations are made in fiscal year 2013, 2014, 2015, and 2016 respectively, and zero otherwise.
absolute difference between GAAP earnings and alternative earnings (ABS) and the number of reconciliations (NUMB), are not significantly different from zero (β1 = -0.214, p = 0.173; β1 =
0.452, p = 0.275, both one-sided). This means that firms that are mandated to disclose alternative earnings do not make a higher number of exclusions or do not have a higher absolute amount of exclusions. Therefore, hypothesis 1 is confirmed only when the difference
28
in reconciliations is measured by the probability of the alternative earnings measure being higher than GAAP earnings. The control variable GROW is negative and significant for the proxies HIGH and ABS (β6 = -0.933, p = 0.070; β6 = -1.255, p < 0.001, both one-sided), in line
with my expectations and the findings by Epping and Wilder (2011). The variable LEV is positive and significant for the proxy HIGH (β5 = 0.080, p = 0.078, one-sided), however, is
negative in the regression with the other proxy ABS (β5 = -0.053, p = 0.024, one-sided). This
mixed evidence is not in accordance with Epping and Wilder (2011) as they found a positive relationship between the extent to which firms are leveraged and the amount of adjustments and income increasing adjustments, and no significant effect between the leverage of firms and the absolute adjustment magnitude. Furthermore, the coefficient for IND is significantly negative (β4 = -0.262, p = 0.056, one-sided) in contrast to the findings by Bowen et al. (2005).
For IFRS I did not expect a certain effect, and according to the results reporting according to the IFRS standards does not alter the difference in reconciliations in only a positive or a negative way as the results are mixed for the proxies HIGH and ABS (β2 = -0.991, p = 0.090; β2 = 0.554, p = 0.050, both two-sided). All other control variables are insignificant.
The fact that my results are so different from previous research could mean that the results of previous research are not valid anymore due to changes in the reporting behaviour of firms, as noted by Black et al. (2017). Another possibility is due to my research design. For example, the difference in findings between the three proxies can be design-related because a logistic regression is used to test the proxy HIGH, and OLS regressions for the other proxies
ABS and NUMB. Testing the proxies with the same regression method could enhance the
consistency of the findings amongst the difference proxies. Furthermore, this sample incorporates different countries. As a result, possible inter-country differences could be included in the individual predictors. For example, Epping and Wilder (2011) show that US firms are more aggressive in reporting alternative earnings measures compared to foreign firms listed on a US stock exchange. Therefore, there might be differences between Australia, South Africa and the US in the aggressiveness with which they report, that alters my findings. Furthermore, it could be a possibility that the small sample size of this study (n = 400)
compared to others may cause the results to be insignificant. Another possibility, especially for the proxy NUMB (F = 0.79, p = 0.641, two-sided), is that the models are not correctly specified to investigate the research question of this study. For example, it could be that the number of reconciliations (NUMB) is affected by differences in how firms classify the
reconciliations. In this case, similar firms could show differences in reconciliations, distorting the effects any other independent variables could have on the difference in reconciliations.
