1
Trade-off between real and accrual earnings
management in public and private companies
Name: Bas Schoot
Student number: 10195440 Master thesis
Date: 04-06-2015
MSc Accountancy & Control, specialization Accountancy Faculty of Economics and Business, University of Amsterdam First supervisor: Dr. W. Janssen
2 Statement of Originality
This document is written by student Bas Schoot who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.
The faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
3 Abstract
This paper examines whether there is a difference in trade-off decisions between public and private companies regarding the earnings management
techniques. The study is conducted in the United Kingdom, where the accounting is based on a firm’s legal form instead of his listing status. The discretionary accruals based on the modified Jones model are used as a proxy for accrual based earnings management, while the residuals of the Roychowdhury (2006) models are used as proxies for real activity manipulation. The descriptive statistical tests provide
preliminary evidence that public firms engage more in accrual based earnings management relative to accrual based earnings management and private
companies. This can be interpreted as evidence that there is a difference in trade-off decisions between public and private companies. In addition, the main findings provide evidence that public firms opt more for accrual based earnings management relative to private companies. This can be explained through the sample
characteristics and to the fact that in this sample public firms are less constrained and more incentivized to engage in accrual based earnings management than expected.
4
Table of content
1 Introduction ... 5 1.1 Background ... 5 1.2 Research question ... 5 1.3 Motivation ... 6 1.4 Research methodology ... 6 1.5 Contribution ... 7 2 Literature review ... 9 2.1 Earnings management ... 92.1.1 Real earnings management ... 9
2.1.2 Accrual-based earnings management ... 10
2.1.3 Trade-off between real and accrual earnings management ... 11
2.2 The UK setting ... 12
2.3 Hypothesis: Earnings management incentives private vs. public firms ... 13
3 Research methodology ... 14
3.1 Sample selection ... 14
3.2 Measurement of accrual earnings management ... 17
3.3 Measurement for real earnings management ... 18
3.4 Empirical model ... 19
4 Descriptive statistics ... 21
4.1 Summary statistics balance sheet and income statement... 21
4.2 Summary statistics coefficients of earnings management models ... 22
4.3 Descriptive statistics control variables ... 25
5 Results ... 27
6 Additional analysis ... Fout! Bladwijzer niet gedefinieerd. 7 Conclusion ... 33
8 References ... 36
5 1 Introduction
1.1 Background
Many of the international accounting studies have focused on the existence of
earnings management practices in public firms in the U.S. (e.g., Wongsunwai, 2013). The practices consist of two different strategies, real and accrual based earnings management. Following Roychowdhurry (2006), I define real earnings management as a purposeful action to alter reported earnings in a particular direction. Real earnings management sacrifices firm’s cash flow and might hurt firm value in the long run (2006). Such long term costs are driven by temporary price discounts that lower margins on future sales, reductions of investments in research and
developments, and/or overinvestment in unneeded inventory (Chi, Lisic & Pevzner, 2011). Unlike real earnings management, accrual earnings management is achieved by changing the accounting methods or estimates used when presenting a specific transaction in the financial statement. This technique bias reported earnings in a particular direction without affecting the cash flow.
Prior literature have shown evidence that firms make choices between the two earnings management strategies (Zang, 2012). Zang provides evidence that
choices regarding the strategies depends on the relative costliness. He finds that, when accrual based earnings management is constrained due to limited accounting flexibility and higher level of scrutiny, firms use real earnings management to a greater extent (2012). This suggests that managers shift from accrual to real earnings management after increased risk of litigation.
1.2 Research question
Most of the prior studies have focused on public firms in the U.S. (Wongsunwai, 2013; Cohen, Dey & Lys; Cohen & Zarowin, 2011). So there is a lack of research that focusses on private firms outside and within the U.S. To fill the gap in the literature this study examines the trade-off between real and accrual earnings management in private firms. To this end the following research question is formulated:
‘’ Is there a difference in trade-off in earnings management strategies between public
6 1.3 Motivation
The focus of this research question is important for three reasons. First, as mentioned by Zang (2012), examining both earnings management techniques in isolation cannot explain the overall effect of earnings management activities. When managers use both strategies as substitutes of each other, then examining only one type at a time cannot lead to complete conclusions. In addition, the UK setting is important, because according to Burgstahler, Hail and Leuz (2006) the accounting regulation in UK is based on a firm’s legal form, instead of his listing status. So, private companies face largely the same accounting standards as public companies, but the market for financial reporting differs between both. And finally the existence of earnings management in private companies is important, because they constitute the majority of the EU economy and the EU market for audit services (Tendeloo & Vanstraelen, 2008).
1.4 Research methodology
The sample of this study consist of private and public firms in the UK. The study focuses on the period 1999-2004. By setting this sample period, the results will not be influenced by the implementation of IFRS in 2005. The financial data for the private firms are obtained from the Amadeus database and the data for the public firms are derived from Osiris. After deducting all improper data, the databases yield a final sample of 1260 observations.
The discretionary accruals are used as a proxy for accrual-based earnings management. The primary model to measure discretionary accruals is based on the modified Jones model. Furthermore, the proxies of Roychowdhury (2006) are used to measure real earnings management. The metrics are the abnormal levels of cash flow from operations and production costs. The discretionary expense model is not used due to the low data availability for private firms. Subsequent studies of Cohen, Dey and Lys (2008) and Cohen and Zarowin (2010) provide evidence that these proxies capture real activity manipulation.
7 1.5 Contribution
The objective of this study is to examine whether there are differences in the trade-off decisions between public and private companies. I hypothesize that private firms engage more in accrual-based earnings management relative to public firms. And, that public firms engage more in real earnings management relative to private firms.
This research contributes in several different ways. First it extends prior literature that examines the trade-off between real and accrual earnings management in public firms (e.g. Zang, 2012; Cohen & Zarowin, 2010). According to Cohen and Zaraowin, the comparison between public and private firms shed a light on the influence of market forces on the reporting behavior of firms. This has important implications for society. They rely on the quality of the financial statements, so they should be aware that managers can substitute to real earnings management techniques.These techniques are especially important to investors, because research of Gunny (2005) shows that real earnings management has a significantly negative impact on future firm performance.
The main findings of this article indicate that there are significant differences in the degree of earnings management techniques between public and private
companies. The regression, correlation test and mean test of difference all provide evidence that public firms engage more in accrual based earnings management relative to private companies. This can be considered as evidence that there is a difference in trade-off decisions between both types of companies. In addition, the tests do not yield evidence for a significant difference in the degree of real activity manipulation between public and private companies. In addition, the descriptive statistics for each type of company separately provide further evidence that the private firms engage significantly less in accrual based earnings management
relative to real activity manipulation. On the other hand, the publicly listed companies manipulate the earnings more extensively through the accruals relative to real
activity manipulation.
The findings are inconsistent with the hypothesis and can be explained
through the sample characteristics and to the fact that in this sample public firms are less constrained and more incentivized to engage in accrual based earnings
8
Second, this study examines the existence of earnings management in a different setting than prior literature. Many studies have focused on public firms in the U.S., while the European setting is neglected (e.g. Cohen, Dey & Lys, 2008). And finally, this study contributes by analyzing the trade-off between both earnings management techniques in private firms, instead of examining them in isolation (e.g Burgstahler, Hail & Leuz, 2006; Coppens & Peek, 2005).
The remainder of the paper is organized as follows. Section 2 reviews the relevant literature, provide the theoretical background and formulate the hypotheses. Section 3 describes the research methodology, sample selection procedure and the regression model. After that, section 4 provides the descriptive statistics with
preliminary evidence from the correlation and mean comparison test. Section 5 documents further indirect evidence in the difference in trade-off decision. Hereafter, direct evidence is reported in the additional analysis section 6 and finally section 7 concludes this paper with a discussion of the main results, limitations and
9 2 Literature review
2.1 Earnings management
In the literature earnings management is defined in several different ways. Two commonly used definitions are the following:
‘’Earnings management occurs when managers use judgment in financial reporting
and in structuring transactions to alter financial reports to either mislead some stakeholders about the underlying economic performance of the company or to influence contractual outcomes that depend on reported accounting numbers’’
(Healy & Wahlen, 1999, p. 368).
‘’Earnings management is a purposeful intervention in the external financial reporting process, with the intent of obtaining some private gain’’ (Schipper, 1989,p. 92).
Healy and Wahlen (1999) state that managers can use judgment to make the
financial statements more informative, by providing inside information to the users. In this way earnings management decreases the information asymmetry. But the
managers can also use their discretion for their own interest. In this case, managers use earnings management to mislead stakeholders about the underlying
performance of the firm.
The research that is carried out in earnings management has confirmed that managers manage earnings in a variety of ways, ranging from accounting estimates and methods (so-called accrual earnings management) to manipulation of real transactions (so-called real earnings management) (Zang, 2012;Roychowdhury, 2006;Prencipe, Markarian & Pozza, 2008). Both earnings management techniques and the incentives to engage in both strategies are discussed in-depth in the
following sub-sections.
2.1.1 Real earnings management
Real earnings management is a purposeful action to alter reported earnings in a particular direction, by changing the timing or structuring of a transaction
(Roychowdhury, 2006). Real earnings management occurs in different ways, like for example accelerating sales from next fiscal years into the current year by cutting prices, delaying investments in research and developments and by overinvesting in
10
unneeded inventory (Chi, Lisic and Pevzner, 2011). This all is done in effort to boost current period earnings.
A great deal of research in the US have examined the motivation behind real earnings management. Bushee (1998) and Baber et al (1991) find evidence for firms reducing R&D expenditures to meet or beat earnings benchmarks. This is in
accordance with findings of Burgstahler and Dichev (1997), who provide some limited evidence on managers manipulating real activities to meet the zero earnings benchmark. The existence of real earnings management is further supported by the survey of Graham et al (2005). They report that 78 percent of the surveyed CFO’s reduce R&D, advertising and maintenance expenditures to meet earnings targets. The limitations of the studies that examine real earnings management, is that the sample only comprises firms in the US. This study extends prior studies by focusing on the existence of real earnings management in the UK, instead of the US.
The consequences of real earnings management is the subject of some recent research. Zhao et al (2012) find evidence that in general real activity
manipulation is associated with lower firm’s performance in subsequent years. This is consistent with the research of Roychowdhury (2006) and Ewert and Wagenhofer (2005). They state that real earnings management is costly and directly reduces firm value, because the manager is willing to sacrifice future cash flow for current
performance.
Unlike real activity manipulation, accrual based earnings management bias reported earnings in a particular direction without affecting the firm’s cash flow. This technique is covered in the following subsection.
2.1.2 Accrual-based earnings management
Unlike real earnings management, which alters the execution of real transactions, accrual earnings management is achieved by changing the accounting methods or estimates used when presenting a specific transaction in the financial statement (Zang, 2012). In this manner, earnings are manipulated by using the flexibility allowed by regulation and accounting standards regarding the recording and measuring of transactions. Prencipe, Markarian & Pozza (2008) mention for example the provision for uncollectible accounts. The standard requires firms to record a loss for the amount of receivables that the manager think will be
11
uncollectable in the future. The managers can use this flexibility to estimate a low amount of uncollectable receivables, leading to higher earnings (2008). An important characteristic of accrual-based earnings management is that over the long term accruals will total zero. This means that a period with higher-than-normal accruals must be offset against a period with lower-than-normal accruals (Roosenboom, Goot & Mertens, 2003). So managers can bias earnings, but this have no effect on the cash flow of companies.
In the last years, a lot of research gives attention to the incentives behind accrual earnings management. Burgstahler and Dichev (1997) and Coppens and Peek (2005) find that companies engage in accrual earnings management to avoid small losses and to a lesser amount to avoid earnings decreases. Another motive to engage in accrual earnings management is tax avoidance. The study of Sercu, Bauwhede & Willekens (2002) and Ball and Shivakumar (2005) find evidence for firms manipulating earnings to manage the taxes. Tax incentives induce firms to smooth earnings, since avoiding high earnings reduce taxes (Coppens & Peek, 2005). In addition, Coppens and Peek state that tax incentives have stronger influence on the degree of earnings management in countries where accounting practice is strongly aligned with tax practice. Finally, Bergstresser and Phillipon (2006) present evidence that accrual based earnings management happens more in firms with higher level of stock-based incentives. Warfield and Cheng (2005) had already conducted a research in the association between equity incentives and earnings management. They find consistent with Bergstresser and Phillipon, that managers with high equity incentives are more likely to report earnings that meet or beat analyst’s forecasts.
2.1.3 Trade-off between real and accrual earnings management
Previous subsections described the earnings management practices in isolation. This section elaborate more on prior literature, which examines both strategies. First, Zang (2012) investigates how firms trade-off between both earnings management practices, using a large sample of public firms over 1987-2008. The results show that accrual-based earnings management is constrained by the presence of high quality auditors, heightened scrutiny by investors and legislators and firm’s accounting inflexibility. He also finds support for the hypothesis that managers trade-off the two
12
approaches according to their relative costliness (2012). This is consistent with the research of Cohen and Zarowin (2010), who state that firms engage in both earnings management practices during a seasoned equity offering (SEO). Most importantly, they find for SEO firms a positive correlation between the use of real earnings management and the cost of accrual-based earnings management (2010). This implies that managers substitute according to the relative cost of both actions.
Cohen, Dey and Lys (2008) also examined the trade-off. They document that real earnings management increased significantly after the passage of SOX. This suggests that firms substitute accrual for real earnings management after the heightened scrutiny. The authors state that managers substitute to real earnings management, because it is harder to detect.
2.2 The UK setting
This section describes the differences between the UK and the US setting and clarifies the choice for European firms in the sample. First of all, it is hard to study the existence of earnings management practices of private firms in the US, because the lack of publicly available data. The UK provides an unique setting for this
research, because the Compancy act of 1967 required all companies, public and private, to file their annual reports annually to the register. So contrary to the US, in the UK the audited financial statements of private companies are readily available (Coppens & Peek, 2005).
Apart from the data availability, the UK setting provides another factor that contribute to the opportunity for examining the existence of earnings management in public and private companies. First according to Burgstahler, Hail and Leuz (2006) and Ball and Shivakumar (2005), accounting regulation within the UK is based on a firm’s legal form, instead of his listing status. This means that the regulatory regime for financial statement preparation is the same for public firms and all but the
smallest privately held companies, but the market for financial reporting differs
between both. An important legislation that has contributed to the uniformity between the public and private companies, is the Fourth Directive. It requires the firms to prepare audited financial accounts according to a common set of accounting rules and to deliver the financial statements to the corporate registers (2006)
13
Summarizing, the UK setting is unique as it provides a common set of accounting standards for public and private companies, while both facing different institutional incentives and market forces. These differences lead to different motives for public and private companies for their choice between real and accrual earnings management. This is discussed in the next section.
2.3 Hypotheses: Earnings management incentives private vs. public firms
According to Burgstahler, Hail and Leuz (2006), private firms firms face different demands for accounting information than publicly listed companies. They have concentrated ownership structures, since the shares are in the hands of firm management. This makes them less dependent on external financing. They can also effectively communicate accounting information to stakeholders through private channels, thereby reducing the demand for financial reporting quality (Burgstahler, Hail & Leuz, 2006). These factors lead to lower incentives for private firms to provide informative earnings to outside parties. Hence, it is of lesser concern that
manipulating earnings may make earnings less informative to stakeholders
(Burgstahler, Hail & Leuz, 2006). Ball and Shivakumar (2005) already provided some evidence for this statement in their research to private and public companies in the UK. They show that the earnings of private companies are indeed of lower quality than those of public firms, despite being prepared under the same accounting standards. With quality they mean the information usefulness of the earnings to outside parties, like investors or bankers.
The difference in accounting environment, low demand for earnings quality, and the findings of Ball and Shivakumar (2005) leads to the following hypothesis:
H1: Other things being equal, private firms engage more in accrual-based
earnings management relative to public firms.
On the contrary, public firms face a different accounting environment. The importance of external financing in public equity markets creates the demand for accounting information that is useful in evaluating and monitoring of firms. Equity investors do not have access to private information and because of this information asymmetry, they rely heavily on the financial statements. If the quality of this
14
information is low, then the investors are reluctant to supply capital to the firm (2006).
So, public firms have relatively stronger incentives to provide informative earnings to the investors. This is in accordance with the study of Peek, Cuijpers and Buijink (2010), who state that the role of accounting information in communication and contracting with outside parties is more important for public than for private firms. While these arguments suggest that market forces give incentives for public firms to report earnings that reflect true economic performance, there could be trade-offs and countervailing effects. For instance, managerial compensation contracts and analyst’s expectation can give rise to earnings management practices in public firms to manipulate firm’s performance. Whether the manager choses for real or accrual earnings management depends on the level of constraints of each action. Thus, when discretion is more constrained for one earnings management technique, the manager will use the other more extensively (Zang, 2012). The level of constraint is determined by the firm’s accounting environment. In the accounting environment of public firms the financial statements are more scrutinized by investors, analysts and regulatory authorities relative to private firms (Tendeloo & Vanstraelen, 2008). Thus, accrual based earnings management is more constrained for public firms relative to private firms, because it is more likely to be detected due to the heightened scrutiny. Consequently, managers switch to real earnings management because it is harder to detect. This reasoning leads to the following hypothesis:
H2: Other things being equal, public firms engage more in real earnings
management relative private firms.
3 Research methodology
3.1 Sample selection
In this study, I investigate the trade-off between real and accrual earnings
management between the years 1999-2004 in the United Kingdom. The sample period is chosen, because in 2005 IFRS was implemented. The implementation of IFRS has an impact on the degree of earnings management and on the accounting standards for public firms (Jeanjean & Stolowy, 2008; Callao & Jarne, 2010). By setting this sample period, my results will not be influenced by the implementation of
15
IFRS. The sample of private firms consist of firms that are large enough, according to the criteria of the Fourth Directive. This requires that privately held firms should meet at least two of the following criteria: (1) total assets greater than £ 2,5 million, (2) sales greater than € 5 million and (3) number of employees greater than 50. Consistent with previous research I exclude banks, insurance companies and other financial holdings (NAICS codes between 5200 and 5330) and institutions in public administration (NAICS codes above 9200) (Coppens & Peek, 2005; Tendeloo & Vanstraelen, 2008).
Accounting information for private companies are obtained from the Amadeus database. The advantage of the Amadeus database is that it includes standardized financial statement data for private firms in Europe. This makes it possible to
investigate a group of firms that is under-represented in the academic research (Burgstahler, Hail & Leuz, 2006). Data for public companies are derived from the Osiris database, because on the Amadeus database there is no data available for public firms in the period prior to 2002. For the uniformity of the analysis, largely the same data items that were derived from the Amadeus database are deduct from Osiris. The Osiris database is chosen, because it contains financial information on over 80.000 public companies over the world, including the UK.
The initial sample for large enough private firms consists of 28.487
observations. I exclude companies who are not a private liability company and for which the lagged total assets and the lagged change in revenue (∆Revt-1) are not
available. Next, I verify whether there are companies left for which the data related to certain balance sheet items (e.g. tangible fixed assets), P/L items (e.g. cost of goods sold) and changes in balance sheet items (e.g. Δ Sales) are not available. After that, I exclude the firms who have NAICS codes between 5200-5330, codes above 9000 and firms that do not have a NAICS code. And finally, I verify whether there are companies in the remaining sample where the change in data is based on double observations in a year or where they do not have three continuous years (e.g.
2000,2002,2003). Multiple observations in the same year lead to improper change in data, because then for example year 2002 is compared against 2002. On the other hand, with inconsequential years it could happen that the change in data is obtained by deducting the years 2003 and 2001. After deducting all the improper data I
16
The steps that are taken in the private sample, are also conducted to obtain the final sample of public companies. This lead to 363 firm-year observations The steps that are taken to acquire the final samples are illustrated in table 1.
Table 1. Sample selection
Private sample Public sample
Initial sample 28.487 Initial sample 37.624
Deduct all other types of companies (2.732) Deduct companies outside the UK (36.799) Deduct companies without lagged
assets and lagged change in revenue
(24.383) Deduct companies with multiple
observations in the same year (91) Deduct companies with missing
balance sheet data:
Deduct companies without lagged assets and lagged change in revenue
(293)
Missing tangible assets (31) Deduct companies with missing balance sheet data:
Missing Inventory (12) Missed Inventory (27)
Missing creditors (16) Missed debtors (8)
Deduct companies with missing P/L
data: Other current assets (1)
Missing Cost of goods sold (222) Other current liability (5) Missing Depreciation (7) Deduct companies with missing P/L
data:
Missing Net income (1) Cost of goods sold (26)
Deduct companies with missing Δ
Sales data (54) Depreciation (2)
Deduct companies with missing data
for the GROWTH variable (7)
Deduct companies with missing change data:
Deduct companies with missing NAICS codes or codes between 5200-5330 or above 9000
(99)
MissingΔ Other current assets + Δ Creditors + Δ Other current
liabilities
(4) Delete companies with two
observations in one year or that do not have 3 consecutive years
(26)
Deduct companies with missing NAICS codes or codes between 5200-5330 or above 9000
(5)
Final sample 897 Final sample 363
Total sample = 1260 firm-year observations
17
There are more private firm observations (71%) in the total sample relative to public firm observation (29%). This rate is largely in line with studies of Ball and
Shivakumar (2005) and Burgstahler, Hail and Leuz (2006), where the percentages were even higher (respectively 96% and 95,2% private firms in total sample).
3.2 Measurement of accrual earnings management
Following prior literature, I use discretionary accruals to proxy for accrual based earnings management (Cohen & Zarowin, 2010; Zang, 2012). The discretionary accruals are the portion of total accruals unexplained by normal operating activities. The primary model to measure discretionary accruals is based on the modified Jones model (DeFond & Jiambalvo, 1994):
𝑇𝐴𝑖,𝑡 𝐴𝑠𝑠𝑒𝑡𝑠𝑖,𝑡−1 = 𝑘1 1 𝐴𝑠𝑠𝑒𝑡𝑠𝑖,𝑡−1+ 𝑘2 ∆𝑆𝐴𝐿𝐸𝑆𝑖,𝑡 𝐴𝑠𝑠𝑒𝑡𝑠𝑖,𝑡−1+ 𝑘3 𝑃𝑃𝐸𝑖,𝑡 𝐴𝑠𝑠𝑒𝑡𝑠𝑖,𝑡−1+ 𝜀𝑖,𝑡 (1)
Where TA represents the total accruals. Due to the low data availability for earnings before extraordinary items and discontinued operation, I use the metric of Ball and Shivakumar (2005) which measure the total accruals as follows:
𝑇𝐴𝑖,𝑡 = ∆𝐼𝑛𝑣𝑒𝑛𝑡𝑜𝑟𝑦 + ∆ 𝐷𝑒𝑏𝑡𝑜𝑟𝑠 + ∆ 𝑂𝑡ℎ𝑒𝑟 𝐶𝑢𝑟𝑟𝑒𝑛𝑡 𝐴𝑠𝑠𝑒𝑡𝑠 − ∆ 𝐶𝑟𝑒𝑑𝑖𝑡𝑜𝑟𝑠 −
∆ 𝑂𝑡ℎ𝑒𝑟 𝐶𝑢𝑟𝑟𝑒𝑛𝑡 𝐿𝑖𝑎𝑏𝑖𝑙𝑖𝑡𝑖𝑒𝑠 − 𝐷𝑒𝑝𝑟𝑒𝑐𝑖𝑎𝑡𝑖𝑜𝑛
Furthermore the Assets in equation 1 represents lagged total assets, ∆SALES is the change in sales and PPE is the gross value of property, plant and equipment. The
Assets, ∆SALES and PPE variables are included to control for changes in accruals
that are caused by changes in the firm’s economic conditions (DeFond & Jiambalvo, 1994).
The model is separately estimated for every industry (two-digit NAICS-code). The model is not estimated for each year, due to the low sample size per year and industry combination. Furthermore, there are also no important institutional or regulatory changes during 1999 and 2004 in the UK which could have a significant impact on the degree of earnings management. So for this reason, I see the years 1999-2004 as a homogeneous sample period and do not control for year effects. The observations per industry are presented in table 2. The estimated residuals (EM) following from equation 1, captures the level of discretionary accruals and are used as proxy for accrual based earnings management.
18 3.3 Measurement for real earnings management
To measure real earnings management, I rely on prior studies to develop some proxies for real activity manipulation (Cohen, Dey & Lys, 2008; Zang, 2012). Following from Roychowdhury (2006), I consider two proxies to detect the level of real earnings management: the abnormal levels of cash flow from operations (CFO) and production costs. Cohen, Dey and Lys (2008) and Cohen and Zarowin (2010) provide evidence that these proxies capture real activity manipulation.
First to measure the normal level of CFO, I use Dechow et al’s (1998) model which is a linear function of sales and change in sales. To estimate this model, I use the following regression for each industry, consistent with Cohen and Zarowin
(2010): 𝐶𝐹𝑂𝑖,𝑡 𝐴𝑠𝑠𝑒𝑡𝑠𝑖,𝑡−1 = 𝑘1 1 𝐴𝑠𝑠𝑠𝑒𝑡𝑠𝑖,𝑡−1+ 𝑘2 𝑆𝐴𝐿𝐸𝑆𝑖,𝑡 𝐴𝑠𝑠𝑒𝑡𝑠𝑖,𝑡−1+ 𝑘3 ∆𝑆𝐴𝐿𝐸𝑆𝑖,𝑡 𝐴𝑠𝑠𝑒𝑡𝑠𝑖,𝑡−1+ 𝜀𝑖,𝑡 (2)
Because U.S. style cash flow statements are not available for the sample in the UK, I use the balance sheet approach of Leuz, Nanda and Wysocki (2003) to calculate the operational cash flow:
𝐶𝐹𝑂𝑖,𝑡 = 𝑂𝑝𝑒𝑟𝑎𝑡𝑖𝑛𝑔 𝑖𝑛𝑐𝑜𝑚𝑒 − 𝑇𝐴𝑖,𝑡
Where TAt is measured using the metric of Ball and Shivakumar (2005) mentioned earlier in the measurement of accrual earnings management paragraph.
I then calculate abnormal CFO as the residual of equation 2 (REM1).
Second, I follow Roychowdhury (2006), to estimate the normal level of production costs using the following regression model for each industry:
𝑃𝑟𝑜𝑑𝐶𝑜𝑠𝑡𝑖,𝑡 𝐴𝑠𝑠𝑒𝑡𝑠𝑖,𝑡−1 = 𝑘1 1 𝐴𝑠𝑠𝑒𝑡𝑠𝑖,𝑡−1+ 𝑘2 𝑅𝑒𝑣𝑖,𝑡 𝐴𝑠𝑠𝑒𝑡𝑠𝑖,𝑡−1+ 𝑘3 ∆𝑅𝑒𝑣𝑖,𝑡 𝐴𝑠𝑠𝑒𝑡𝑠𝑖,𝑡+ 𝑘4 ∆𝑅𝑒𝑣𝑖,𝑡−1 𝐴𝑠𝑠𝑒𝑡𝑠𝑖,𝑡−1+ 𝜀𝑖,𝑡 (3)
ProdCost is the firm’s production costs, calculated as the sum of cost of goods
sold and changes in inventory. Further, the Assets stands for lagged total assets,
Rev for the firm’s sales and ∆Rev for the changes in sales. The residuals from
equation three represent the abnormal level of production costs (REM2). High residuals indicate that companies have a larger amount of inventory overproduction. This lead to an increase in earnings, which is caused by a reduction in the cost of goods sold (Roychowdhury, 2006).
Consistent with Zang (2012);Cohen, Dey and Lys (2008) and Burgstahler, Hail and Leuz (2006), I scale all the accounting items by lagged total assets to enforce comparability across the sample firms. Furthermore, I report the aggregated
19
measure (REMsum), which is the sum of the two individual measures as well as both single real earnings management proxies (REM1 and REM2). The reason behind this is that both proxies have different implications for the earnings of a firm. So using the aggregated measure alone may lead to a dilution of the results.
(Cohen, Dey & Lyss, 2008; Zang, 2012, Burgstahler, Hail & Leuz, 2006).
3.4 Empirical model
The following regression model is used to examine the degree of accrual based earnings management for public and private firms:
𝐸𝑀 = 𝛽0+ 𝛽1𝑃𝑢𝑏𝑙𝑖𝑐 + 𝛽2𝑆𝐼𝑍𝐸 + 𝛽3𝐿𝐸𝑉 + 𝛽4𝐺𝑅𝑂𝑊𝑇𝐻 + 𝛽5𝑅𝑂𝐴 + 𝛽6𝐼𝑁𝐷 + 𝜀 (4)
Appendix 1A presents a description of the variables and how they are measured. For the dependent variable in equation (5) I use the proxy of accrual based earnings management (EM) measure derived from the residuals of equation (1). In this study there is one independent variable of interest. This is the dummy variable Public taking the value of one when the company is publicly listed and zero when it is a private firm. Due to the incentives and the heightened scrutiny for public firms and the accounting environment and lower demand of earnings quality for private firms, I expect a negative β1. A negative β1 implies that private firms engage more in accrual based earnings management relative to public firms.
Based on prior research, I include some firm characteristics as control variables which might affect abnormal accruals and which are also likely do differ across public and private firms (Burgstahler, Hail & Leuz, 2006). First, I include the natural logarithm of the book value of total assets at the end of the fiscal year (SIZE)
to proxy for the size of a company. According to Tendeloo and Vanstraelen (2008) size affects earnings management, because of the agency problems and
government scrutiny which increases with the size and profitability of a company. Second, I include a leverage variable (LEV) measured by the ratio of total liabilities to total assets,which can impact earnings management in two ways. Highly leveraged firms may engage in upward earnings management to avoid debt
covenant violation (DeFond & Jiambalvo, 1994), but high leverage may also induce downward earnings management in distressed firms for contractual renegotiation reasons (Tendeloo & Vanstraelen, 2008).
20
Third, to control for variation in performance between public and private firms, I use the variables (GROWTH) and (ROA). GROWTHis defined as the annual
percentage change in sales and ROAis the yearly return on assets measured by bottom-line net income divided by lagged total assets.
Finally, I include industry dummies (IND) to control for industry effects on earnings management. The classification and the amount of observation for the private and public sample per industry group is illustrated in table 2.
Next, to examine the degree of real earnings management in public and private firms, I use the following regression model:
𝑅𝐸𝑀 = 𝛽0+ 𝛽1𝑃𝑢𝑏𝑙𝑖𝑐𝑣𝑠𝑝𝑟𝑖𝑣𝑎𝑡𝑒 + 𝛽2𝑆𝐼𝑍𝐸 + 𝛽3𝐿𝐸𝑉 + 𝛽4𝐺𝑅𝑂𝑊𝑇𝐻 + 𝛽5𝑅𝑂𝐴 + 𝛽6𝐼𝑁𝐷 + 𝜀
(5)
The dependent variable consists of the aggregate of real earnings management (REMsum) proxies and the individual measures (REM1 and REM2) of
Roychowdhury (2006). The independent variable of interest isPublic, indicating a
one when the company is publicly listed and zero otherwise.
Following the hypothesis 2, I expect a positive β1 which result in higher level of real earnings management for public firms relative to private companies.
Furthermore the control variables are the same as mentioned earlier and are presented in Appendix 1A.
Table 2. Observation per industry.
Beta No. Industry group NAICS code
Observations in private sample Observations in public sample 1. Agriculture + forest + mining + construction + utilities 111-115, 221, 236-238 158 94 2. Manufacturing 311-315 258 135 3. Transportation + wholesale trade + retail trade 441-445, 481-485 215 67 4. Information 511, 512, 515, 517, 518, 26 37 5. Services 541,551,561,562, 611, 621-624, 711-713, 721, 722, 811-814 240 30
21 4 Descriptive statistics
4.1 Summary statistics balance sheet and income statement
This chapters provides comprehensive descriptive statistics, to provide a further understanding of the differences between public and private firms. In table 3, I describe the most important balance sheet and income statement items. For ease of presentation I divided it in two panels. Panel A represents the amounts for the
publicly listed firms and panel B for the private firms.
The most obvious difference between both types of firms, is size. Private firms in the sample have mean total assets of £ 93 million, compared to £ 15 million for public firms. Also the amount of sales is larger for private firms (£ 60 million) relative to public firms (£ 11,5 million). We should keep these numbers in mind during the analysis, because they are inconsistent with Ball and Shivakumar (2005). They conducted also a research in the UK, but found opposite descriptive statistics. In their case the public firms had significantly more total assets and sales relative to the private firms.
Table 3. Summary statistics important balance sheet and income statement items. Panel A. Public listed firms
Panel B. Private firms
TAperc stands for the tangible assets/total assets and newCOGS for the costs of goods sold.
On the other hand, the results for net income are consistent with their
findings. The net income for private firms amount to £1,2 million and for public firms £0,54 million. This is also the case for the sample of Ball and Shivakumar, where the net income measures for private firms are higher than for publicly listed companies.
Netincome 363 538421.1 2654036 -1.62e+07 1.82e+07 newCOGS 363 7119185 2.27e+07 18800 2.12e+08 Sales 363 1.15e+07 3.01e+07 504000 2.65e+08 TAperc 363 67.67482 17.15884 16.8652 98.35885 Totalassets 363 1.51e+07 3.14e+07 2556000 1.95e+08 Variable Obs Mean Std. Dev. Min Max
Netincome 897 1233270 3.27e+07 -1.43e+08 9.22e+08 newCOGS 897 4.82e+07 1.95e+08 46068 3.78e+09 Sales 897 6.01e+07 2.42e+08 949000 4.63e+09 TAperc 897 28.08213 24.49069 0 98.29227 Totalassets 897 9.37e+07 1.70e+09 2513568 5.06e+10 Variable Obs Mean Std. Dev. Min Max
22
They argue that this is consistent with the notion in the literature that private firms are more profitable.
What can be concluded from this analysis is that there are significant
differences between private and public firm’s balance sheet and income statement items. Therefore, it is important to control for size in the regression model.
4.2 Summary statistics coefficients of earnings management models
In this paragraph the coefficients of the modified Jones model and the
Roychowdhury (2006) models are compared to the outcomes of prior literature and their expectations. I estimate these models using the entire sample. Table 4 panel A presents the estimates of the coefficients with their expected signs according to Roychowdhury (2006) and Siregar and Utama (2008).
Except from the coefficient 1/At-1, all the coefficients in the Jones model are consistent with the expectations of Siregar and Utama (2008) and with the outcomes of Roychowdhury (2006). The adjusted R-squared of the Jones model in this
research is 0,0852. This is in line with the mean adjusted R-squared over the years 1995-2002 (0,0895), measured in the paper of Siregar and Utama (2008).
The coefficients of the real earnings management models are all consistent with the outcomes of Roychowdhury, except for the ΔSales/ At-1 coefficient under the CFO-model. But this is in accordance with the expectations of Roychowdhury, because according to his assumptions this should be negative. His assumption states that any dependence of accruals on changes in sales, so positive coefficient for ΔSales /At-1 in the Jones model, should be offset by a reverse dependence of CFO on change in sales, so negative coefficient in CFO-model (Roychowdhury, 2006). The R-squared of CFO model (0,010) is quite low and not in line with Roychowdhury (0,45). This may be caused due to the low sample size and/or the use of the balance sheet approach for the calculation of the CFO. On the other hand, the R-squared of the Production costs model (0,9335) is very high and in accordance with findings of Roychowdhury (0,89).
So in essence, the coefficients are consistent with prior literature and with their expectations and the R-squared of the models is sufficient for a proper research.
23
Panel B of table 4 reports the summary statistics of the residuals of the Jones and real earnings management models for the entire sample. The residuals
represents proxies for real and accrual based earnings management. Higher value of the proxies EM and REM2 indicate more earnings management through accrual-based earnings management and overproduction. I multiply the residuals of REM1 times -1 to indicate that a higher amount represents higher degree of real earnings management through the use of price discounts and lenient credit terms. The
REMsum represents the total degree of real activities manipulation and is the sum of REM1 and REM2.
Panel B is further divided in part 1 and 2 present respectively the summary statistics of the residuals for private and public firms. What is interesting in these statistics is the difference in mean of EM between public and private firms. With the use of a two sample t-test, I find that the mean is significantly (t-value= 4,7216, significant at the 1% level) higher for public firms relative to private firms. This indicates that public firms engage more in accrual based earnings management relative to private firms. This is inconsistent with hypothesis 1, stating that private firms engage more in accrual earnings management relative to public firms.
On the other hand, the mean of REMsum is higher for private firms relative to public firms. This might indicate that private firms engage more in real earnings management relative to public firms. However the difference is not significant according to the t-test, so further analysis is necessary.
The summary statistics of the residuals are also used in a mean comparison test between the means of the earnings management proxies and the regression line itself. This means that the residuals are tested whether they are significantly different from zero. This is calculated for both types of companies separately and illustrated in table 4 panel C. First, following the hypotheses I expect for private companies that there is a significantly positive difference between mean of accrual based proxy and zero and a negative significant difference for the real earnings management proxies and the regression line. This implies that private firms engage more in accrual based earnings management relative to real activity manipulation. The findings of the t-test provide opposite descriptive statistics and present a negative significant difference (t-value= -2.1878, significant at the 1 percent level). This indicates that private firms does not manage earnings through the accruals which is inconsistent with the hypotheses.
24
For the real earnings management proxies, the test does not provide significant findings.
On the other hand, for public firms I expect a positive significant difference between real earnings management proxies and zero and for accrual based earnings management a negative significant difference. But again the mean t-test provide opposite statistics. The difference between the accrual based proxy and zero is significantly positive (t-value=10,150, significant at the 1 percent level). This
indicate that public firms engage significantly in accrual based earnings
management. The difference between the proxy for real activity manipulation through price discounts and lenient credit terms is significantly negative from zero (t-value=-2,2537, significant at the 1 percent level). This implies that public firms do not use this technique extensively to manipulate the earnings.
So in essence from the descriptive statistics from panel C, it is clear that private firms engage significantly less in accrual based earnings management relative to real activity manipulation while publicly listed companies engage significantly more in accrual based earnings management relative to real activity manipulation. This is inconsistent with the expectations and with the prior literature.
Table 4. Panel A Model parameters a)
Expected signs Jones Model Expected signs REM model (CFO/At-1) Expected signs REM model (ProdCost/ At-1) 1/ At-1 - 160.465 *** - -44.868,69 - -576.130,3 *** Sales/ At-1 + 0,0059827 + 0,880724 *** ΔSales/ At-1 + 0,081665 *** + -0,0393808 *** + 0,0047139 ΔSalest-1/ At-1 - -0,024665 * PPE/ At-1 - -0,002028 R-squared 0,0874 0,0100 0,9335 Adj R-squared 0,0852 0,0076 0,9332
Table 4. Panel B Summary statistics of real and accrual-based earnings management
b)
Variable N Mean Median Std. Dev. 25 % 75 %
EM 1260 3,07ₑ-11 0,0059176 0,1632 -0,0704 0,0588
REM1 (CFO) 1260 -1,96ₑ-11 -0,0128 0,2026 -0,0704 0,0554
REM2
(ProdCosts) 1260 -4,69ₑ-11 0,0334 0,3315 -0,1199 0,1716
25
Table 4. Panel B.1 Private firms
Variable N Mean Median Std. Dev. 25% 75%
EM 897 -0,0137 -0,0298 0,1875 -0,1016 0,0582
REM1 (CFO) 897 0,0045 -0,0143 0,2324 -0,1030 0,0744
REM2
(ProdCosts) 897 0,0019 0,0513 0,3734 -0,1334 0,2040
REMsum 897 0,006399 0,0389 0,5014 -0,1977 0,2663
Table 4. Panel B.2 Public firms
Variable N Mean Median Std. Dev. 25% 75%
EM 363 0,0338 0,0317 0,0635 0,0094 0,05905 REM1 (CFO) 363 -0,0111 -0,0108 0,0938 -0,0518 0,0269 REM2 (ProdCosts) 363 -0,0047 0,0021 0,1924 -0,0859 0,1221 REMsum 363 -0,015813 -0,0023 0,2333 -0,1430 0,1371
Table 4. Panel C Mean test for earnings management proxies
Private firms Public firms
Variable Mean Regression line (residuals=0) t-stat Mean Regression line (residuals=0) t-stat EM -0,0137 0 -2,1878*** 0,0338 0 10,150*** REM1 (CFO) 0,0045 0 0,5786 -0,0111 0 -2,2537*** REM2 (ProdCosts) 0,0019 0 0,1530 -0,0047 0 -0,4669 REMsum 0,006399 0 0,3822 -0,015813 0 -1,2915
a) *, *** indicates statistical significance at the 10 percent and the 1 percent level, respectively. This table reports the estimated parameters of the following regressions:
TAi,t Assetsi,t−1 = k1 1 Assetsi,t−1 + k2 ∆SALESi,t Assetsi,t−1 + k3 PPEi,t Assetsi,t−1 + εi,t CFOi,t Assetsi,t−1 = k1 1 Asssetsi,t−1 + k2 SALESi,t Assetsi,t−1 + k3 ∆SALESi,t Assetsi,t−1 + εi,t ProdCosti,t Assetsi,t−1 = k1 1 Assetsi,t−1 + k2 Revi,t Assetsi,t−1 + k3 ∆Revi,t Assetsi,t + k4 ∆Revi,t−1 Assetsi,t−1 + εi,t
b) EM= proxy for accrual based earnings management derived from Jones model. REM1= proxy for real earnings management derived from Roychowdhury’s CFO model. REM2= proxy for real earnings management derived from Roychowdhury’s Production costs model. REMsum is the sum of both individual proxies.
4.3 Descriptive statistics control variables
This chapter provides more information about the control variables of the regression models (equation 5 and 6) and how they are correlated with each other and with the dependent variables. First, table 5 provides summary statistics of the control
26
in means between public and private companies are calculated with the t-test (Mean-comparison test).
The results indicate that private firms in this sample are significantly larger in size and more leveraged than public firms. On the other hand, public companies have significantly higher return on assets compared to private firms. Finally, there is no significant difference in their growth patterns. In essence, it is clear that there are significant differences and that I have to control for this in the regression.
Once more, it becomes clear from the statistics of the control variables that there are differences between this sample and the sample that Burgstahler, Hail and Leuz use in their research in the United kingdom in 2006. They show opposite descriptive statistics in the differences between public and private companies their means and medians of the control variables.
Table 5. Summary statistics control variables.
Private firms Public firms Difference
Variables Mean Median Mean Median Mean Test of
difference LEV 0,7684886 0,7396797 0,632133 0,6160152 0,1363555 5.9364***
GROWTH 0,1111731 0,0516375 0,076503 0,0378076 0,0346701 0.9720
ROA 0,0190292 0,0269545 0,039757 0,0432843 -0,020728 -2.6283***
SIZE 16,44045 16,20711 15,86224 15,63369 0,57821 8.6373***
*** Indicates statistical significance at the 1 percent level (Two-tailed). SIZE= Natural logarithm of book value total assets. LEV= Total liabilities divided by total assets. GROWTH= Annual percentage change in sales. ROA= return on assets calculated by bottom-line net income/total assets.
The pairwise correlations between the independent and the dependent
variables of the main regression models are presented in table 6. First of all, there is a significant positive correlation between the residuals of both real earnings
management models, REM1 and REM2 (0,3256). This indicates that companies uses both real earnings management techniques at the same time (Zang, 2012). The
REMsum is the sum of the REM1 and the REM2, due to this there is a high
correlation between these residuals (0,7038 and 0,9008 respectively). Furthermore, the EM and the proxies for real earnings management are also significantly
correlated. This indicates that companies are using both earnings management techniques, real activities manipulation and accrual based earnings management as complementary methods. The correlations between the residuals of the models are consistent with the research of Zang (2012).
27
Furthermore, GROWTH and ROA are significantly correlated with accrual earnings management (0,1631 and 0,1438). This may indicate that high performance companies who report growth patterns engage more in accrual based earnings management relative to stigmatized and low performing companies. On the other hand, LEV is negatively correlated with EM (-0,1390). This suggests that highly leveraged firms are more monitored by bankers which constrain the ability to engage in accrual based earnings management. Due to this high leveraged firms may
substitute to real activity manipulation, which is consistent with the positive
significant correlation between LEV and all the real earnings management proxies. The ROA is highly correlated with the REM proxies (-0,4949 , -0,3149 and -0,4638).
Consistent with the results of the t-test for differences from table 5, LEV has a negative significant correlation (-0,1651) and ROA positive significant correlation (0,0739) with the dummy variable Public. This indicates that public companies are less leveraged and show a higher return on assets relative to private firms.
Finally, the independent variable of interest, Public, is significantly correlated with the proxy of accrual based earnings management (0,1320). This suggests that public firms engage significantly more in accrual based earnings management than private firms. This is consistent with the finding derived from the difference of mean test of table 4 panel B and inconsistent with hypothesis 1. Public is not significantly correlated with the real earnings management proxies.
Table 6. Correlation between the dependent and independent variables of the main regression models.
EM REM1 REM2 REMsum Public~e LEV GROWTH ROA logsize
EM 1 REM1 0,6916*** 1 REM2 0,0582** 0,3256*** 1 REMsum 0,3613*** 0,7038*** 0,9008*** 1 Publicvspr~e 0,1320*** -0,0349 -0,0091 -0,0228 1 LEV -0,1390*** 0,1779*** 0,0968*** 0,1544*** -0,1651*** 1 GROWTH 0,1631*** 0,1405*** 0,0383 0,0933*** -0,0274 0,0066 1 ROA 0,1438*** -0,4949*** -0,3149*** -0,4638*** 0,0739*** -0,4147*** 0,1167*** 1 SIZE 0,0890*** 0,0313 0,0227 0,0315 -0,2366*** -0,0015 0,0514* 0,0086 1 *, **, *** Indicates statistical significance at the 10 percent, 5 percent and 1 percent level, respectively.
EM= proxy for accrual based earnings management derived from Jones model. REM1= proxy for real earnings management derived from Roychowdhury’s CFO model. REM2= proxy for real earnings management derived from Roychowdhury’s Production costs model. REMsum is the sum of both individual proxies. SIZE= Natural logarithm of book value total assets. LEV= Total liabilities divided by total assets. GROWTH= Annual percentage change in sales. ROA= return on assets calculated by bottom-line net income/total assets.
28 5 Results
The descriptive statistics provide some preliminary evidence that public firms engage more in accrual based earnings management relative to private firms. This is
inconsistent with hypothesis 1, which states that private firms engage more in accrual based earnings management. For hypothesis 2 there are no significant findings acquired so far. Therefore, this section discusses the results from the main regression models to test both hypothesis.
Table 7 and 8 presents the results from estimating the regression models (equation 4 + 5). Table 7 reports the results when the accrual based earnings management proxy (EM) is used and table 8 when the proxies for real earnings management are used. The results related to accrual earnings management are discussed first.
Consistent with the preliminary findings from the descriptive statistics, I find a significant positive 𝛽1 (0,0483) for the dummy variable. This indicates that public firms engage more in accrual based earnings management relative to private firms. This is inconsistent with hypothesis 1, which states that private firms engage more in accrual based earnings management relative to public firms. One possible
explanation for the opposite result is the difference in characteristics of the private and public firms in this sample. As can be seen in table 3 and 5, the private firms are significantly larger than the public firms in the sample. They have larger total assets, sales, COGS and other balance sheet items (not tabulated e.g. Accounts
receivable). This is inconsistent with prior research (Ball & Shivakumar, 2005; Burgstahler, Hail & Leuz, 2006), where the public firms are significantly larger. So the characteristics of the private firms in this sample differs significantly from samples of related studies and show more similarities with publicly listed firms. In addition, as stated by Tendeloo & Vanstraelen (2008), the size of a company negatively affects the degree of accrual based earnings management, because governmental scrutiny increases with the amount of sales or the size of a company. Due to this accrual based earnings management is more constrained for larger firms. In this case, the private firms are the larger ones and are more constrained to
engage in accrual based earnings management relative to the ‘small’ public firms in this sample.
29
Furthermore, consistent with the correlation results with EM in table 6,the
GROWTH, SIZE and ROA variables are significantly positive. LEV is significantly
negative which is consistent with findings of Vander Bauwhede & Willekens (2004) . According to them highly leveraged firms are more monitored by bankers which makes it harder to engage in accrual based earnings management. As can be seen in table 5, private firms are more leveraged than public firms which is inconsistent with the sample characteristics of Burgstahler, Hail and Leuz (2006). This provides another possible explanation for the positive 𝛽1, because private firms are more constrained to commit accrual based earnings management due to the high degree of monitoring by banks. On the other hand, public firms are less monitored by bankers and more able to engage in accrual based earnings management. This notion is also supported by the significantly positive correlation between LEV and the real earnings management proxies (see table 6). This suggests that more leveraged firms substitute to real earnings management, because accrual based is constrained. In addition, I control for industry effects in the analysis, but all the industry dummies are insignificant and for ease of understanding not presented in table 7.
Finally, Bergstresser and Philippon (2006) provide evidence that incentivized managers are more likely to engage in accrual based earnings management. With incentivized managers they mean CEO’s whose compensation is more sensitive to the share price of the company. They further state that late in the 1990s
management’s compensation became more exposed to changes in the share price of the company. This was caused through the increase of grants in options and stocks. Companies did this to better align the managers’ incentives with those of the shareholders. So managers’ compensation is more dependent on the stock price, which lead to incentives to use their discretion to affect the earnings and increase the stock price (2006). This is especially the case for public companies, because there is dispersion between the owners of the company and the CEO. Thus, managers of public companies are more incentivized to manipulate the earnings than the managers of private companies, because their compensation is more sensitive to the share prices. This will lead to an increase in accrual based earnings management.
So to conclude from the first regression, public firms engage more in accrual based earnings management, because they are less constrained than private firms. This is inconsistent with hypothesis 1. The inconsistency can be explained due to the
30
highly incentivized managers of public companies relative to private firms and because, of the characteristics of private firms in this sample. They are significantly larger and more leveraged relative to public firms. This lead to more scrutiny and monitoring, which impede the possibility to engage in accrual based earnings management.
Table 7. Coefficients of accrual based earnings management .
Accrual based earnings management (EM)
Variables Coefficient t-stat
Publicvsprivate 0,0483525*** 4,57 LEV -0,0358504*** -2,71 GROWTH 0,0433767*** 5,54 ROA 0,1059266*** 2,71 SIZE 0,0166471*** 4,02 Adjusted R-squared 0,0722
**,*** indicates statistical significance at the 5 percent and 1 percent, respectively (Two-tailed). The accrual based earnings management regression is based on the following model:
𝐸𝑀 = 𝛽0+ 𝛽1𝑃𝑢𝑏𝑙𝑖𝑐 + 𝛽2𝑆𝐼𝑍𝐸 + 𝛽3𝐿𝐸𝑉 + 𝛽4𝐺𝑅𝑂𝑊𝑇𝐻 + 𝛽5𝑅𝑂𝐴 + 𝛽6𝐼𝑁𝐷 + 𝜀. EM= proxy for accrual based earnings
management derived from Jones model. Public= one if the company is publicly listed and zero otherwise. SIZE= Natural logarithm of book value total assets. LEV= Total liabilities divided by total assets. GROWTH= Annual percentage change in sales. ROA= return on assets calculated by bottom-line net income/total assets.
Table 8 reports the regression results based on the proxies for real earnings management. It is divided in two panels, panel A for the individual real earnings management proxies (REM1 and REM2) and panel B for the sum of the individual proxies (REMsum).
The variable of interest, Public, is insignificant for all three proxies. This is inconsistent with hypothesis 2, which states that public firms engage more in real earnings management relative to private firms. An explanation for the insignificance is based on the findings of the regression on accrual based earnings management. These findings suggest that public firms engage more in accrual based earnings management than private companies. This may indicate that they are less
constrained to engage in accrual earnings than expected. This influences the trade-off to real earnings management, because according to Zang (2012) firms substitute to real earnings management when accrual based earnings management is more costly (e.g. constrained). I expected that it was more costly for public firms relative to private companies to engage in accrual earnings management due to the heightened
31
scrutiny. However in this sample I find that public firms are less constrained, so there is no need to substitute to real activity manipulation. In addition, the research of Bergstresser and Philippon (2006) indicate that managers of public companies are more incentivized than those of private firms, because their compensation is more tight to the stock price of the company. So CEO’s of public companies have more incentives to manipulate the earnings relative to those of private companies and at the same time they are less constrained to accrual earnings management than expected. In essence, the findings of regression 1 affect the trade-off decision and have a negative impact on the degree of real activity manipulation by public firms. This may eventually lead to insignificant differences between public and private firms regarding real earnings management.
The GROWTH variable is significantly positive and consistent with the correlation results, which also show significant positive correlation with the real earnings management and accrual based proxies. This indicates that firms report growth patterns engage in both earnings management techniques. ROA and LEV are significantly negative for the real earnings management proxies.
Finally, what is interesting to see is that the industry dummies (IND1, IND2 and IND3) are significantly positive for REM2 and REMsum. This suggests that companies in the agriculture, manufacturing and transportation sector engage significantly in real earnings management. This evidence could be used as
preliminary findings for follow-up research in the effect of industry types on the trade-off between real and accrual based earnings management. IND4 is insignificant and
IND5 is omitted because of collinearity.
So in essence, for the variable of interest (Public) I find insignificant results. This can be explained by the findings of regression 1. They show that public firms are not constrained to accrual based earnings management, as was expected and at the same time they are more incentivized to manipulate the earnings. This will
influence the trade-off decision of public firms and will have a negative impact on the degree of real activity manipulation. So public firms will not engage more in real earnings management relative to private firms and this may lead to insignificant differences between both types of companies.
32
Table 8 Panel A Coefficients of real earnings management (Individual proxies).
REM1 REM2
Variables Coefficients t-stat Coefficients t-stat
Publicvsprivate 0,0042829 0,37 -0,0197151 -0,98 LEV -0,024363* -1,69 -0,0388359 -1,54 GROWTH 0,0715972*** 8,38 0,0404597*** 2,71 ROA -0,8601996*** -20,20 -0,922323*** -12,38 SIZE 0,0051699 1,15 0,0085758 1,09 IND1 0,0144394 0,94 0,279793*** 10,46 ÌND2 0,0065913 0,47 0,1431775*** 5,90 IND3 0,0244693* 1,67 0,079816*** 3,11 IND4 -0,0040443 -0,16 0,0334332 0,77 Adjusted R-squared 0,2842 0,1821
Table 8 Panel B Coefficients of real earnings management (Sum). REMsum
Variables Coefficients t-stat
Publicvsprivate -0,0154322 -0,61 LEV -0,0631989*** -2,01 GROWTH 0,1120569*** 6,02 ROA -1,782523*** -19,22 SIZE 0,0137457 1,40 IND1 0,2942324*** 8,83 IND2 0,1497689*** 4,96 IND3 0,1042853*** 3,26 IND4 0,0293888 0,54 Adjusted R-squared 0,2847
*,*** indicates statistical significance at the 10 percent and 1 percent, respectively (Two-tailed). The real earnings management regression is based on the following model:
𝑅𝐸𝑀 = 𝛽0+ 𝛽1𝑃𝑢𝑏𝑙𝑖𝑐 + 𝛽2𝑆𝐼𝑍𝐸 + 𝛽3𝐿𝐸𝑉 + 𝛽4𝐺𝑅𝑂𝑊𝑇𝐻 + 𝛽5𝑅𝑂𝐴 + 𝛽6𝐼𝑁𝐷 + 𝜀. REM= proxy for real earnings management
derived from the residuals of the Roychowdhury models (REM1, REM, REMsum). Public= one if the company is publicly listed and zero otherwise. SIZE= Natural logarithm of book value total assets. LEV= Total liabilities divided by total assets.
GROWTH= Annual percentage change in sales. ROA= return on assets calculated by bottom-line net income/total assets. IND are dummy variables see for classification table 2.
33 6 Conclusion
This study provides evidence regarding the trade-off decision between accrual based earnings management and real activity manipulation of public and private companies in the UK. The research is conducted for the years 1999-2004, because in that time frame the results are not influenced by the implementation of IFRS in 2005. The real activity manipulation techniques I examine, are overproducing of inventory and providing price discounts and lenient credit terms. The models of Roychowdhury (2006) are conducted to estimate the abnormal level of CFO and production costs, which are used as proxies for real earnings management. The proxy for accrual based earnings management is based on the modified Jones model.
This research contributes to prior literature by examining the trade-off decision outside the US and for both public and private companies. The UK is chosen,
because there the data availability is the highest and most importantly, the accounting regulation is based on a firm’s legal form, instead of his listing status. This implies that private companies face largely the same accounting standards as public firms. Although, both types of companies face the same accounting standards, I still hypothesize that there are differences in trade-off decisions. First, I expect that private firms engage more in accrual based earnings management relative to public firms. The reason behind this is that private firms are less dependent on external financing and have lower incentives to provide informative earnings to outside parties. On the other hand, real earnings management is harder to detect. This is important for public firms, because their financial statements are more scrutinized by investors and analysts. This lead to the other hypothesis, public firms engage more in real activity manipulation relative to private companies.
The descriptive statistics for both types of companies separately already provided some evidence that private firms engage significantly less in accrual based earnings management relative to real activity manipulation. On the other hand, the publicly listed companies manipulate the earnings more extensively through the accruals relative to real activity manipulation.
Consistent with the descriptive statistics, the main findings of this article indicate that there are significant differences in the degree of earnings management
