National Culture and Earnings Management: A Cross-country Analysis

National Culture and Earnings

Management: A Cross-country Analysis

Master thesis

University of Groningen

Faculty of Economics and Business

MSc Accountancy

Abstract

The purpose of this study is to measure the impact of national cultures on subsequent use of accrual based (AEM) and real earnings management (REM), with the aim to identify factors that can potentially influence the amount of earnings management (EM). This study focuses on Hofstede’s (2010) two added dimensions recently: Indulgence vs. Restraint (IVR) and Long-term orientation (LTO), as it is a relatively unexplored field. The impact of culture on EM is analyzed for a sample of firms from 38 countries from 2000-2016 using Ordinary Least Squares (OLS). The results show that cultures distinguished by long-term orientation are associated with less earnings management when compared to both AEM and REM. The cultural dimension of indulgence is also associated with less REM, while it renders positive results with AEM. These results hold, even after controlling for other cultural variables, investor protection and the substitution effect. The results also indicate that EM is affected by investor protection and that there is a substitution effect between AEM and REM. The consideration of cultural characteristics, investor protection and the substitution effect is important to regulators, auditors and investors as results indicate that common accounting methods do not necessary result in mutual accounting practices.

January 22, 2018

Syvart Visser

S3237206

Supervisor: Wang, S.

Second assessor: Mukherjee, S.

T

ABLE OF CONTENTS

1. INTRODUCTION 3

2. THEORETICAL BACKGROUND AND HYPOTHESIS DEVELOPMENT 4 2.1ACCOUNTING EARNINGS AND EARNINGS MANAGEMENT 4

2.2EARNINGS MANAGEMENT METHODS 5

2.3EARNINGS MANAGEMENT AND CULTURE 6

2.4HYPOTHESIS DEVELOPMENT 10

2.4.1LONG TERM ORIENTATION VERSUS SHORT TERM NORMATIVE ORIENTATION (LTO) 10

2.4.2INDULGENCE VERSUS RESTRAINT (IVR) 11

3. METHODOLOGY AND SAMPLE SELECTION 12

3.1DEPENDENT VARIABLES 12

3.1.1ACCRUAL BASED EARNINGS MANAGEMENT 12

3.1.2REAL EARNINGS MANAGEMENT 12

3.2INDEPENDENT VARIABLES 13 3.3CONTROL VARIABLES 14 3.4SAMPLE SELECTION 15 4. RESULTS 17 4.1DESCRIPTIVE STATISTICS 17 4.2CORRELATIONS 18 4.3MULTIPLE REGRESSIONS 18

4.4LONG-TERM ORIENTATION 20

4.5INDULGENCE VS.RESTRAINT 20

4.6ROBUSTNESS CHECK AND ADDITIONAL TESTS 22

5. CONCLUSIONS 29

6. REFERENCES 30

1.

I

NTRODUCTION

Earnings management (EM) has been the subject of extensive accounting research and since the paper of Jones (1991) the empirical archival research on EM is vast (Walker, 2013). Early literature shows evidence that managers use EM (Healy, 1985; Healy & Wahlen, 1991). Nabar & Boonlert-U-Thai (2007) state that national culture with respect to earnings management has been researched limitedly. This seems rather remarkable, especially because Gray (1988) already points out that national culture could influence the accounting systems and values of countries. This, in turn, means that culture can also influence EM. Apart from this possible relation, the accounting scandals at the beginning of this century spiked the interest of standard setters and academic research with respect to EM as well (Walker, 2013). For example, Cohen et al. (2008) and Zang (2012), show evidence that real earnings management (REM) is to substitute accrual based earnings management (AEM). The apparent trade-off between AEM and REM should be of importance of auditors, regulators and investors, as REM may be harmful in the future (Roychowdhury, 2006).

In addition, the harmonization of accounting principles and the globalization of economies increased the interest in EM. Pacheco Paredes and Wheatley (2017) subsequently state that it is necessary to have a better understanding of the possible impact of a national culture on managers’ decisions. This research therefore focuses on the impact of national cultures on subsequent use of accrual based (AEM) and real earnings management (REM), thereby aiming to identify factors that can influence the amount of real earnings management. The purpose of this paper is to extend prior literature regarding the interrelationships between national culture and accounting practices, especially earnings management, by looking at both AEM and REM.

The research question addressed in this paper is the following: “To what extent does national culture influence use of accrual based and real earnings management?”

Consistent with prior studies, this paper uses the cultural variables developed by Hofstede (1980, 2001). The four variables used in most studies are: Power Distance Index (PDI), Individualism vs. Collectivism (IDV), Masculinity vs. Femininity (MAS) and Uncertainty Avoidance Index (UAI). In addition to prior literature, it is possible to measure the cultural differences using two more recent cultural variables of Hofstede. Since 2010, Hofstede extended the number of countries for the fifth dimension to 93, and added a sixth dimension to the model. These two new variables are: Long Term Orientation vs. Short Term Normative Orientation (LTO) and Indulgence vs. Restraint (IVR) and are still a relatively unexplored field (Quiroz et al., 2016). This study contributes to the existing literature, by including the less studied fifth and sixth cultural dimensions, in addition to Hofstede’s most used cultural dimensions. In the literature reviewed there is none existing literature regarding all six cultural values of Hofstede with respect to accrual based and real earnings management in such wide set of

This paper does this by, first providing a literature review regarding agency theory, earnings management and the relationship with culture. Then, a hypothesis is developed, which is followed by the model construction and sample selection. This paper then continues with the results. Finally the conclusions are presented, and suggestions for further research are offered.

2.

T

HEORETICAL BACKGROUND AND HYPOTHESIS

DEVELOPMENT

In order to develop hypotheses for this research, literature on earnings and earnings management is examined. First, the importance and relevance of accountings earnings and in particular earnings management is discussed. Second, a review of the relevant literature with respect to national culture and earnings management is provided. Based on the literature review, the hypotheses are presented.

2.1

A

CCOUNTING EARNINGS AND EARNINGS MANAGEMENT

Accounting earnings are relevant to the users of accounting information as they contain information that can be used in pricing debts, equities, and investment decisions (Ball and Brown 1968; Kothari 2001; Francis et al. 2004). Managers who are aware of the importance of accounting earnings can be motivated or pressured into altering earnings to ensure a particular earnings result (Gray et al., 2015). The process of intentional manipulating accounting information, and use of discrete choices to arrive at a particular earnings result is known as earnings management (Han, 2010). Earnings management is made possible by the flexibility of different accounting standards and accounting methods (Ajit et al., 2013). Earnings management can be both income increasing and income decreasing. For example, Burgstahler and Dichev (1997), and Brown and Caylor (2005) report that earnings management is used to increase the level of earnings to a particular height. Strong and Meyer (1987), and Kirschenheiter and Melumad (2002), on the other hand, argue that earnings management is used to lower or smooth earnings rather than favoring higher future earnings.

Earnings management is a complex phenomenon where a single theory can only provide a factional explanation. A dominant theoretical perspective with respect to earnings management is the agency theory (Marston and Shrives, 1996; Mishra and Malhotra, 2016). This covers the conflicts that occur in the relationship between the management (agents) and the owners (principals) of an organization. It also includes how managers can be incentivized to ignore the owners’ interests and act in their own interest instead (Fama and Jensen, 1983). Considering the latter from a moral hazard perspective, managers can be incentivized because certain managers’ compensations are tied to the firms’ performance (Walker, 2013). Managers can also indulge in earnings management to report a better, but fictional, performance to the owners. Other motivations to manage earnings are: to meat market or analysts’ expectations, debt covenants, and other factors (Fields et al. 2001). From an information perspective, managers can also use earnings management to discretely exchange private

information, which can be of interest of all parties (Walker, 2013). Nonetheless, the use of earnings management is not risk free and can lead to significant negative stock market reactions, political responses, and allegations of earnings management or perceived earnings manipulation (Leuz et al. 2003; Dechow et al. 1995). This is also why a better understanding of earnings management is important not only to to shareholders, but to creditors, standard setters and regulators in global capital markets as well (Healy and Wahlen 1999; Leuz et al. 2003; Defond 2010).

2.2

E

ARNINGS MANAGEMENT METHODS

According to Healy and Wahlen (1999), earnings management can be classified into two categories: accruals based and real earnings management. Accrual based earnings management (AEM) refers to the manipulation of earnings via accruals. Accrual based earnings management occurs when the manager's various estimations, judgments and accounting reporting choices are included in the process when financial statements are prepared (Enomoto et al, 2015). Accrual based earnings management is divided into discretionary accruals and non-discretionary accruals. The former refers to the accruals over which managers have considerable discretion to manage reported net income (Scott, 2012). Examples are: stock valuation assumptions, bad debt provisions and accounting depreciation rates (Healy & Wahlen, 1999).

“Real earnings management (REM) is defined as management actions that deviate from normal business practices, undertaken with the primary objective of meeting certain earnings thresholds” (Roychowdhury, 2006) With REM, managers take actions that change the timing and/or structure of: operating, investing, or financial decisions (Enomoto et al, 2015). Roychowdhury (2006) defines three methods of real earnings management:

1) Sales manipulation by offering limited price discounts or a lenient credit term. Earnings management through sales manipulation may lead to more sales, but can also lead to lower cash flows from operations and higher production costs than usual.

2) Changing discretionary expenses. When firms reduce their advertising, research, development, selling, and general and administrative expenses, they are able to reduce expenses and therefore increase their earnings.

3) Engage in earnings management by producing more than strictly necessary. This overproduction will lead to a reduction of the fixed costs per product, and thereby increase earnings. The reduction in total costs per unit indicates lower cost of goods sold (COGS) and thus higher net income (Roychowdhury, 2006).

Graham et al. (2005) provide results that shift the focus from AEM to REM. From the executives they include in their survey, 80 percent is willing to use real earnings management to meet earnings benchmarks. This might be the case because, accrual based earnings management is more visible, due to the scrutiny of auditors and regulators, than real earnings management (Francis et al., 2004). Since real earnings management can be masked as normal activity, it is harder for stakeholders to detect (Kothari et al., 2001). However, actions taken in the current period to increase earnings may have a negative effect on cash flows in future periods (Roychowdhury, 2006).

2.3

E

ARNINGS MANAGEMENT AND CULTURE

Despite the growing volume of research concerning earnings management, the impact of national culture on earnings management is still important to explore as Mishra & Malhotra (2016) state that the accounting scandals at the beginning of this century are not the result of a few accounting misstatements, but rather seem to be a culture of widespread earnings management within firms. Pacheco Paredes and Wheatley (2017) confirm it is for these reasons that knowledge on this topic needs to improve to foster a more profound understanding of the possible impact of national culture on a manager’s decisions, and the subsequent differences in accounting practices.

2.3.1

H

OFSTEDE

–

G

RAY FRAMEWORK

Hofstede (1980) defines culture as “collective programming of the mind that distinguishes the members of one group or category of people from another”. Cultural norms and beliefs are influential forces affecting people’s perceptions and behaviors (Markus and Kitayama, 1991), and are expected to influence management behavior as well (Cieslewicz, 2014). Given that societal values such as culture can affect managers’ and shareholders’ perception (e.g., Chui, Lloyd, & Kwok, 2002; Fidrmuc & Jacob, 2010, Hooghiemstra, 2015), it can influence factors in explaining several accounting phenomena (e.g. reporting practices, earnings management). Since earnings management occurs when managers use personal judgment in financial reporting and in structuring transactions to alter financial reports (Healy and Whalen, 1999), their personal norms and beliefs based on cultural differences could affect the magnitude of earnings management. Gray (1988) is one of the first researchers who state that national culture could influence the accounting systems and values of countries. His theory is often used for the effects of national culture on structural differences and accounting values across countries (Han, 2010). He argues that an accounting system can be influenced by Hofstede’s societal values: first by the country’s institutions, such as its capital markets or legal system, and secondly, through accounting values, such as professionalism, uniformity, conservatism or secrecy.

Figure 1: Gray's (1988) model with Doupnik and Tsakumis's (2004) extension. Source: Gray (1988: 7). Items under SV, AV, and AS have been added to the original.

Conservatism versus Optimism as mentioned by Gray, refers to a preference for a more cautious approach, as opposed to a more risk-taking approach considering future uncertainty. Uniformity versus Flexibility indicates the standardized, uniform and consistent accounting practices in contrast to flexibility of individual companies in accounting choices. Professionalism versus Statutory Control can be seen as a preference of individual professional judgment and professional self-regulation opposite to compliance with rigid legal requirements and strict controls. Lastly, Secrecy versus Transparency shows the relation between confidential and restricted information only to those who are closely involved with its management, as opposed to a more transparent, open, and publicly accountable disclosure of information. Gray (1988) formulates four hypotheses about the relationship between his accounting values and Hofstede’s (1980) cultural dimensions:

H1 - The higher a country ranks in individualism, and the lower it ranks in uncertainty avoidance and power distance, the more likely it is for the country to rank highly in professionalism.

H2 - The higher a country ranks in uncertainty avoidance and power distance, and the lower it ranks in individualism, the more likely it is for the country to rank highly in uniformity.

H3 - The higher a country ranks in uncertainty avoidance, and the lower it ranks in individualism and masculinity, the more likely it is for a country to rank highly in conservatism.

H4 - The higher a country ranks in uncertainty avoidance and power distance, and the lower it ranks in individualism and masculinity, the more likely it is for the country to rank highly in secrecy.

Even though Gray (1988) hypothesizes the relationships between the dimensions, he does not empirically test whether there is a relationship between societal and accounting values. Salter and Niswander (1995) test and confirm the relationship and significance between the hypotheses of Gray. In addition, Baydoun and Willet (1995) and Chanchani and MacGregor (1999) show the direction of the relationships between Gray’s (1998) accounting values and Hofstede’s (1980) cultural values. Doupnik and Tsakumis (2004) critically review research covering Gray's theory of cultural relevance and his extended the model as is shown in Figure 1. The relationship between Hofstede’s societal values (SV) and Gray’s Accounting values (AV) are statistically significant (Doupnik & Tsakumis, 2004). Doupnik (2008) pioneers by using Hofstede’s (1980) cultural dimensions to hypothesize and document a direct relation between culture and earnings management in a sample of thirty-one countries. Cultural dimensions of individualism and uncertainty avoidance are found to be significant explanatory factors of earnings discretion and earnings smoothing across the sample of thirty-one countries (Doupnik, 2008). Later on, Hofstede (2010) develops two additional cultural dimensions: Long-term orientation and Indulgence versus Restraint. Borker (2013) adds the two additional indices by Hofstede (2010) and modifies the notations between strongly and average interrelated relationships. The relationship between Gray’s (1988) accounting values and Hofstede’s (2010) cultural dimensions are shown in Table 1 with significantly strong direct relationships indicated with a (+ +) sign and strong inverse relationships with a (- -) sign.

Table 1: Borkers (2013) adapted summary of the relationships between Gray's accounting values and Hofstede’s cultural dimensions

Individualism (IDV) Uncertainty avoidance (UAI) Power distance (PDI) Masculinity (MAS) Long-Term Orientation (LTO) Indulgence (IVR) Conservatism - ++ + - + - Uniformity -- ++ + ? + - Professionalism ++ -- - ? - + Secrecy -- ++ ++ - + -

From Hofstede’s (1980) initial four cultural dimensions, Gray (1988) states that two first dimensions are fully linked to all four accounting values. This also seems to have implications for the use of earnings management, since later research confirms that the cultural dimensions individualism and uncertainty avoidance has explanatory power for earnings management (e.g. Hope, 2003; Han, 2010).

2.3.2

E

ARNINGS MANAGEMENT AND CULTURE

Leuz et al. (2003) investigate explanatory factors of the international variation in earnings management across a broad cross-section of countries. They hypothesize and prove that the quality of the legal system and the level of investor protection are significant in explaining the variation in cross-national differences in earnings management. However, they do not analyze the effect of cross-national culture on earnings management. Where Leuz et al. (2003) find investor protection to be responsible for 39% of the variation in aggregate earnings management, Doupnik (2008) finds that culture explains 49% of the variation in aggregate earnings management across the sample of thirty-one countries. However, both studies show that culture is a better explanatory factor with respect to earnings smoothing, than investor protection is with respect to earnings discretion. Guan et al. (2005) examine the effect of national culture on discretionary accruals in five Asia-Pacific countries. They find that uncertainty avoidance and long-term orientation have a significant negative effect on earnings management, while individualism has a significantly positive effect. In line with Guan et al. (2005), Nabar and Boonlert-U-Thai (2007) document that earnings management in countries with high uncertainty avoidance cultures is elevated. They examine the magnitude of earnings management in a sample of 30 countries. However, the effect of power distance, masculinity and individualism on earnings management does not show a significant result. Based on a sample of 31 countries, Doupnik (2008) investigates the relationship between cultural values, earnings smoothing and earnings discretion. Contrary to the evidence shown by Guan et al. (2005), Nabar and Boonlert-U-Thai (2007) state that uncertainty avoidance is positively associated with earnings management, while individualism has a negative effect on the same variable.

Han et al. (2010) analyzes the interacting effects between culture and formal institutional structure to explain differences in earnings management. Based on a sample from 1992 to 2003 in 32 countries, with 96,409 firm-year observations, the cultural dimensions of individualism, masculinity and uncertainty avoidance, as well as the strength of investor protection, have significant effects on the variation in cross-national differences in earnings management. Individualism and masculinity are positively influences, while uncertainty avoidance has a negative effect related to discretionary accruals. More importantly, Han et al. (2010) provide evidence that the degree of investor protection interacts with both uncertainty avoidance and individualism. For example, in countries with strong investor protection, individualistic managers with high uncertainty avoidance are more likely to manage earnings. Based on a sample of 27 countries Guan and Pourjalalia (2010) document the same evidence as Han et al. (2010). Earnings management is negatively related to the dimension of uncertainty avoidance, while individualism and masculinity have significant, positive effects on earnings management. In addition to the evidence by Han et al. (2010), Guan and Pourjalalia (2010) also document a positive relation cultural dimension power distance and earnings management.

More recently, Gray et al. (2015) investigate the extent to which the association between culture and earnings management found by Han et al. (2010) persists under single General Accepted Accounting Principles (GAAP) environment. International Financial Reporting Standards (IFRS) is a principle-based set of high quality reporting standards that limits alternative accounting treatments and it is for the restricting reasons that Barth et al. (2008) documented that the adoption of IFRS could restrict earnings management. Based on model of Gray (1988) they analyze the impact the mandatory adoption of IFRS on the association between national culture and earnings management in 14 European countries from the year 2000 – 2010. Their results show that the adoption of IFRS does not necessarily lead to accounting harmonization and that national culture is still a significant explanatory factor for earnings management in the post IFRS adoption period. Individualism has a significant and positive effect on earnings management during the entire period. However, uncertainty avoidance only has a significant negative effect on earnings management in the post-IFRS adoption period.

Culture is a fundamental informal institution that remains stable over time (North 1990, 2005) and it is expected to observe a persistent effect of culture on earnings management. What should be pointed out here, however, is that the majority of studies does not include all cultural dimensions, but base the included dimensions on their research topic. Current and future research might benefit from relatively new dimensions introduced by Hofstede in 2010. Gray (2015) proposes that the cultural dimension Indulgence vs. Restraint could also be a focus of future study with respect to earnings management.

2.4

H

YPOTHESIS DEVELOPMENT

Like others, this paper uses Gray’s (1988) model to determine the relationship of Hofstede’s (1980) cultural dimensions with respect to earnings management (e.g. Han, 2010; Gray et al., 2015). It discusses and hypothesizes the relationship of each dimension with respect to earnings management in the following section.

2.4.1

L

ONG

T

ERM

O

RIENTATION VERSUS

S

HORT

T

ERM

N

ORMATIVE

O

RIENTATION

(LTO)

Long Term Orientation (LTO) relates to what extent certain societies maintain, remember, and honor their own past whilst dealing with new challenges in the present and future. The dimension LTO in business contexts is normative (short term) versus pragmatic (long term) (Hofstede, 2001). In short-term oriented businesses, recent results are of great importance, and risk-taking is generally accepted (Hofstede, 2001). Opportunistic earnings management might be expected to achieve short-term goals, as the management reward systems are also short-term orientated.

Long-term orientation is positively associated with conservatism, uniformity and secrecy, and negatively associated with professionalism. Conservatism indicates a cautious approach to deal with the uncertainty of future events” (Gray, 1988). When having a long-term orientation on the uncertainty

of future events, less earnings management can be expected. In addition, less professionalism and more uniformity indicates that managers and accountants might be less flexible in self-governance and measurement. This limits the possibilities for managers to have certain discretion to manage reported net income. Finally, a positive relationship with secrecy implies a more confidential and restricted disclosure of information about the business. Therefore, the dimension Long Term Orientation is expected to be negatively associated with earnings management.

H1a: There will be a negative relationship between the Long Term Orientation dimension of national

culture and the magnitude of accrual based earnings management.

H1b: There will be a negative relationship between the Long Term Orientation dimension of national

culture and the magnitude of real earnings management.

2.4.2

I

NDULGENCE VERSUS

R

ESTRAINT

(IVR)

In 2010, a sixth dimension was added to the model: Indulgence versus Restraint (IVR). This dimension is based on the research of the Bulgarian sociologist Minkov and the extensive World Value Survey, and is known as the “happiness research”. Indulgence is defined as “the extent to which society allows natural desires, like having fun and enjoy life, as opposed to being suppressed with the need for regulations or strict social norms” (Hofstede, 2010). Due to the recent introduction of this dimension, there is little research with respect to accounting practices, accounting values and earnings management in particular.

Indulgence societies are more optimistic and feel healthier and happier, even though objectively this does not have to be the case. These societies are more focused on the gratification of human desires and individual happiness with a focus on personal freedom, leisure, and control. In restraint cultures maintaining order is given more importance than the desire for freedom of speech and fulfillment of personal needs. Following Gray’s model, Indulgence is weakly negatively correlated with long-term orientation. Since high long-term orientation is associated with less earnings management, it would be assumed that there is more earnings management in cultures with high indulgence. Professionalism is the only accounting value positively associated with indulgence, as it is negatively associated with conservatism, uniformity, and secrecy. High IVR can be associated with professionalism, freedom of choice, and the related values of flexibility, and optimism. In business context this would imply more freedom for personal judgments in reporting. The negative associations with uniformity suggest that managers and accountants can also show more flexibility in choosing accounting choices for individual companies. The negative relationship with conservatism might lead to a preference for a more risk taking approach in case of future uncertainty. Based on the relationships between Gray’s accounting values (AV) and the social value (SV) Indulgence, a positive relationship with a high level of IVR and earnings management is expected.

H2a: There will be a positive relationship between the Indulgence dimension of national culture and

the magnitude of accrual based earnings management.

When Indulgence is high, managers and accountants might be inclined to report the most positive numbers allowed. However, this is only possible when earnings management through accruals, as real earnings management activities are transactions of reality. If real earnings management and accrual based earnings management are substitutes, as shown by Zang (2012), then high levels of Indulgence might be associated with less real earnings management.

H2b There will be a negative relationship between the Indulgence dimension of national culture and

the magnitude of real earnings management

3.

M

ETHODOLOGY AND SAMPLE SELECTION

3.1

D

EPENDENT VARIABLES

3.1.1

A

CCRUAL BASED EARNINGS MANAGEMENT

Prior research has established the link between accruals management and culture (Leuz et al., 2003; Nabar and Boonlert-U-Thai, 2007; Han et al., 2010). It is therefore important to measure earnings management through accruals. Following prior research, this paper uses the Modified Jones model to obtain empirical measures of earnings discretion (DACC). To identify the discretionary component of accruals for given country-year observations, the following ordinary least squares (OLS) is used:

Modified Jones model: !"##!

!!!! = α! ! !!!!+ α! (∆!"#!!∆!"#!) !!!! + α! !!"! !!!!+ ε!

where TACC_!, is the total accruals in year t, ∆REV_! is the change of revenue in year t, ∆REC_! is the change of receivables in year t, and PPE_!is the level of gross property, plant and equipment in year t. Since the error term of his regression exhibits heteroskedasticity, each variable is deflated in the model by the lagged book value of total assets (A_!!!), which is in line with prior discussed literature. The alphas (α_!) are the parameters to be estimated. The residuals from the regressions (ε_!) are used as a proxy for discretionary accruals.

3.1.2

R

EAL EARNINGS MANAGEMENT

Three proxies of Roychowdhury (2006) are used to measure REM: abnormal cash flow from operations, abnormal discretionary expenses, and abnormal production costs. It is necessary to generate the normal levels of cash flow from operations (CFO), discretionary expenses (DISX), and production costs (PROD) in order to measure the three variables. The measures of abnormal cash from operations (abnCFO), discretionary expenses (abnDISX) and production costs (abnPROD) are the residuals from the following models.

Two composite measures (REMALL) are also created to control for measurement errors. To obtain the residuals, the following cross-sectional regression for each real management activity are run.

Operating cash flows: !F!_! !""#$"_!!!= α! 1 !""E!"_!!!+ α! !"#$!_! !""#$"_!!!+ α! ∆!"#$!_! !""#$"_!!!+!!

Here, CFO is the cash flow from operations; ASSETS refers to the total assets; SALES are the net sales; and ΔSALES represents the change in sales from time t−1 to t.

Discretionary expenses: !"#$! !""#$"!!!= α! ! !""!!"!!!+ α! !!"#$!!! !""#$"!!!+ !!

Here, DISX is the discretionary expense measured as the sum of advertising expense (R&D) and selling, general, and administrative expenses, and the other variables are as defined above.

Production costs: !"#$! !""#$"!!!= α! ! !""!!"!!!+ α! !"#$!! !""#$"!!!+ α! ∆!"#$!! !""#$"!!!+ α! ∆!"#$!!!! !""#$"!!!+ !! Here, PROD is the sum of the cost of goods sold and change in inventory.

Cohen and Zarowin (2010) argue that PROD and EXP are likely to have opposite effects on the operating cash flow, while sales discount and other operating activities can also affect the operating cash flow. Since CFO has reflected the effect of PROD and EXP, it is counted twice when aggregating the three together. In the first aggregated measurement of real earnings management (REMALL) the results of abnCFO are excluded, and combined with the effects of abnDISX and abnPROD. The second aggregated measurement combines the results of abnCFO and abnDISX, while excluding abnPROD.

3.2

I

NDEPENDENT VARIABLES

In order to examine the association between dependent variables DACC and REM, and culture this paper follows others: e.g. Doupnik, 2008; Han et al., 2010; Gray et al., 2015. They use four of Hofstede’s cultural dimensions employed as proxies for the cultural variables. The dimensions regarding country's cultural values are obtained from Hofstede (1980) and are assumed to be fixed over time (Gray & Vint, 1995). Most research covers the four original cultural dimensions: Power Distance Index (PDI); Individualism vs. Collectivism (IDV); Masculinity vs. Femininity (MAS) and Uncertainty Avoidance Index (UAI). However, Hofstede introduces a fifth cultural dimension in 1998: Confucian dynamics, now known as Long-term orientation. Since 2010, Hofstede (2010) has extended the number of countries for the fifth dimension (LTO) to 93 and has added the sixth dimension

3.3

C

ONTROL VARIABLES

Following Han et al. (2010), this paper also includes control variables considered to be associated with earnings management. One of these is the natural logarithm of annual sales (SIZE), because large firms tend to exercise less earnings management when they are subject to continuous stock market monitoring. The natural logarithm of book-to-market ratio (GROWTH) is controlled for, because it is one of the major risk factors identified in prior finance studies (e.g. Fama & French, 1983). Risky firms possess greater incentives to exercise earnings management in reported earnings, owing to the high variability of earnings. The leverage ratio (LEV) and a dummy variable of stock issuance (ISSUE) are included, because firms that are likely to raise capital often have incentives to manage earnings opportunistically. A dummy variable for loss firms (LOSS) is included because loss firms tend to use more accruals to increase earnings numbers in the current year, or to make cookie-jar reserves to increase earnings in the future.

During the sample period (2000-2016), a large number of countries permit or mandate adherence to IFRS. As a consequence, the tests performed in this paper could be biased by these changes in the reporting and regulatory environments. To control for this possibility, a 0(1) indicator variable equal to 1 is included; if the company’s financial statements were presented in conformity with IFRS and 0 otherwise. Compustat determines the accounting standard designation based on company disclosure of the employed standards. It is therefore possible that some observations may be in accordance with IFRS, but are miscoded as domestic GAAP, because the firm makes no specific disclosure of the employed standards. Furthermore, the fact that national GAAP can vary across countries and across sectors may impact the measures of EM and bias the presented results.

The global financial crisis of 2008-2009 may have influenced managerial behavior in patterns similar to REM (e.g. reducing discretionary expenses and offering sales discounts). Because of this, a control for the financial crisis (CRISIS) is added, by including a 0(1) indicator variable that is equal to 1 if the firm-year is 2008 or 2009, and 0 in other firm-years. Finally, a industry dummy (SIC2: two-digit SIC Code) is included, in order to control for cross-sectional differences in the level of earnings management and regulatory environment.

!"#/!"## = !_!+ !_!!"# + !_!!"# + !_!!"#$ + !_!!"#$%& + !_!!"# + !_!!""#$ + !_!!"## + !_!!"#$ + !_!!"#$!" + !_!"

In this formula, REM is the abnormal: operating cash flows, discretionary expense, production costs, or the combination between two; LTO and IVR the country-specific cultural scores from Hofstede (2010); and the other variables are defined above.

3.4

S

AMPLE SELECTION

The data are obtained from Compustat North America and from Compustat Global. This paper selects firm-years that satisfy the following criteria:

(1) Non-financial firm; (elimination of banks and financial institutions, SIC codes between 6000 and 6500 and regulated industries, SIC codes between 4400 and 5000)

(2) all the necessary financial statement variables in the regression model are available; (3) book value of equity is positive;

(4) country-level variables are available; and

(5) each country-year combination has at least 15 observations to ensure a reasonable sample size for the measurement of earnings management

To mitigate the effects of outliers, the variables of REM are winsorized: abnCFO, abnDISX and abnPROD, and DACC, SIZE, GROWTH, and LEV at the 5th and 95th percentiles of the pooled distribution. Other variables are categorical in nature and do not exhibit extreme observations.

157,421 firm-year observations could be obtained that have the necessary data for the abnCFO model. 150,390 firm-year observations meet the data requirements for the computation of abnDISX. 139,431 observations can be used for the computation of abnPROD and REMALL. Finally 148,874 firm-year observations are used for REMALL1. The sample includes data from firms in 38 countries, covering a period from 2000 to 2016. The largest sample observations by country are from the USA and Japan (63,882 and 14,073 observations, respectively). The sample includes 109,530 non-US firm-years (59,4 percent of the sample). The fewest country- level observations are from Portugal, Argentina and Austria (182, 239, and 291 firm years, respectively).

Table 2 presents the number of firm-year observation per country as well as Hofstede’s scores for IVR range from a high of 91 for the USA, to a low of 14 in Pakistan and Indonesia. Malaysia has the highest PD score (104), while Austria has the lowest (11). Greece has the highest UA score (112) and the lowest is Singapore’s (8). Sweden has the lowest MAS score (5), while Japan has the highest (95). Hofstede’s score for LTO score from a low of 20 for Argentina, to a high of 100 for Korea. Pakistan has the lowest IVR score (0), while Mexico has the highest (97). The average of the ENF scores is 9.32 and the range from a low of 4.06 for the Philippines to a high of 9.99 for Switzerland. The median of SIZE, GROWTH and LEV values average respectively 6.115, 0.438, and 0,461.

Table 2: Country level descriptive statistics

Country N PDI IDV MAS UAI LTO IVR INVPR SIZE GROWTH LEV

Argentina 239 49 46 56 86 20 62 5.64 6.845 .38 .55 Australia 5,466 38 90 61 51 21 71 9.30 2.114 .39 .29 Austria 291 11 55 79 70 60 63 9.47 6.020 .27 .56 Belgium 398 65 75 54 94 82 57 9.48 5.536 .32 .57 Brazil 1,040 69 38 49 76 44 59 6.46 7.381 .06 .58 Canada 13,764 39 80 52 48 36 68 9.58 4.346 .42 .42 Switzerland 822 34 68 70 58 74 66 9.99 6.128 .63 .48 Chile 607 63 23 28 86 31 68 6.77 10.110 .80 .48 Germany 2,427 35 67 66 65 83 40 9.37 4.833 .54 .54 Denmark 479 18 74 16 23 35 70 9.80 6.455 .54 .49 Spain 501 57 51 42 86 48 44 7.87 6.438 .43 .64 Finland 536 33 63 26 59 38 57 9.76 5.674 .50 .57 France 2,596 68 71 43 86 63 48 8.97 4.995 .39 .57 UK 4,75 35 89 66 35 51 69 9.40 4.218 .56 .46 Greece 785 60 35 57 112 45 50 6.84 3.927 -.71 .61 Hongkong 658 68 25 57 29 61 17 8.77 7.483 .06 .36 Indonesia 1,514 78 14 46 48 62 38 4.38 12.220 .79 .49 India 10,608 77 48 56 40 51 26 6.12 7.784 .05 .54 Ireland 504 28 70 68 35 24 65 8.74 7.274 .81 .52 Italy 992 50 76 70 75 61 30 7.95 5.334 .28 .64 Japan 14,073 54 46 95 92 88 42 9.37 10.500 -.14 .49 South Korea 5,566 60 18 39 85 100 29 7.04 12.110 .03 .48 Mexico 391 81 30 69 82 24 97 5.99 9.310 .55 .48 Malaysia 3,794 104 26 50 36 41 57 7.71 5.146 -.08 .36 Netherlands 628 38 80 14 53 67 68 9.87 6.658 .72 .57 Norway 687 31 69 8 50 35 55 9.92 6.454 .68 .57 New Zealand 504 22 79 58 49 33 75 9.80 4.846 .50 .44 Pakistan 1,248 55 14 50 70 50 0 4.30 8.520 -.08 .56 Peru 332 64 16 42 87 25 46 4.83 5.992 -.23 .43 Philippines 667 94 32 64 44 27 42 4.08 7.881 .58 .43 Portugal 182 63 27 31 104 28 33 7.81 6.026 -.03 .71 Singapore 2,587 74 20 48 8 72 46 8.99 4.729 -.14 .42 Sweden 1,816 31 71 5 29 53 78 9.80 5.757 .86 .49 Taiwan 2,158 58 17 45 69 93 49 8.08 7.814 .51 .42 Turkey 1,007 66 37 45 85 46 49 5.46 5.626 .37 .49 Thailand 7,898 64 20 34 64 32 45 5.93 7.755 .31 .40 USA 63,882 40 91 62 46 26 68 9.52 5.250 .73 .46 South Africa 1,024 49 65 63 49 34 63 6.70 7.763 .37 .47 Total 157,421 Average 9.32 6.115 .438 .461

4.

R

ESULTS

This chapter includes the results of the conducted research. First, the descriptive statistics and correlations are shown. Then, the results of the multiple regression analysis are discussed. Finally, the robustness check and additional test are discussed. The hypotheses are tested by the use of STATA.

4.1

D

ESCRIPTIVE STATISTICS

Table 3 presents descriptive statistics for the measures of REM, the Hofstede (2010) variables, and control variables. For abnCFO, REMALL and REMALL1, the mean (median) values are positive at the level of 0.078 (0.092), 0.177 (0.176) and 0.169 (0.118). For abnDISX, abnPROD and DACC the mean (median) values are negative with -0.250 (-0.215), -0.068 (-0.67) and -0.045 (-0.011). The control variables show that the overall mean of firm size is 8.33, the overall mean of the market to book (GROWTH) ratio is 0.485, the overall mean of leverage (LEV) is 0.460, the overall mean of loss (LOSS) is 0.197, the overall mean of issue (ISSUE) is 0.502, the overall mean of IFRS is 0.319 and the overall mean of CRISIS is 0.081.

Table 3: Descriptive statistics

N mean sd 25% 50% 75% Dependent variables abnCFO 148.874 .078 .285 -.037 .092 .253 abnDISX 150,390 -.250 .378 -.567 -.215 .000 abnPROD 139,431 -.068 .247 -.225 -.067 .098 DACC 136,308 -.045 .263 -.111 -.011 .059 REMALL 139,431 .177 .473 -.084 .176 .522 REMALL1 148,874 .169 .410 -.075 .118 .403 Independent variables IVR 157,421 57.0 16.0 45 68 68 LTO 157,421 43.6 23.5 26 32 51 Control variables SIZE 157,421 6.20 3.10 4.00 6.12 8.33 GROWTH 157,421 .485 .905 -.162 .438 1.08 LEV 157,234 .460 .225 .284 .461 .625 LOSS 157,421 .197 .398 0 0 0 ISSUE 157,421 .502 .500 0 1 1 IFRS 157,421 .319 .466 0 0 1 CRISIS 157,421 .081 .274 0 0 0

4.2

C

ORRELATIONS

Table 4 presents the firm-specific Pearson correlations among the selected variables. The first correlation shows that LTO is positively related to abnCFO and abnDISX. DACC and both measurements of REMALL indicate that long-term orientation is negatively related to both forms of earnings management. These findings are consistent with hypothesis 1a and 1b. IVR is positively related to abnCFO, abnDISX and both measurements of REMALL. This indicates high level of indulgence, which is negatively related to real earnings management. The fact that IVR is negatively related to DACC indicates that high level of indulgence is associated with more abnormal accruals. IDV and MAS are positively related to DACC and both measurements of REMALL. PDI is also positively related to both measures of REMALL, whilst negatively related DACC. UAI is negatively related to both measurements of REMALL, while positively related to DACC. This indicates that individualism and masculinity and are positively related to both forms of earnings management, while PDI (UAI) is negatively (positive) related to abnormal accruals. The opposite relation applies to real earnings management.

INVPR, LEV, and ISSUE (SIZE GROWTH and LOSS) are positively (negative) related to abnormal accruals. INVPR, LEV, and LOSS (GROWTH and ISSUE) are positively (negative) related to both measures of real earnings management. The correlation between SIZE and real earnings management is less clear. However, the correlations should be interpreted with caution, as they do not control for differences in other firms or country characteristics.

4.3

M

ULTIPLE REGRESSIONS

4.3.1

M

ODEL ESTIMATION

The normal levels of cash flow from operations, production costs, and discretionary expenditures are estimated cross-sectional by country and year. The residuals from these estimations represent the abnormal levels of cash flows, discretionary expenditures, and production costs for the sample firms. When actual levels are lower than expected, the residuals are lower. Thus, lower (higher) values of abnormal abnCFO, abnDISX, (abnPROD, REMALL, REMALL1 and DACC) indicate more earnings management. The coefficients and t-statistics from the regression analyses are presented in table 5. Most of the results are significant at the level of 1% as denoted with *** (P<0.01).

Ta b le 4 : P ea rs on c or re la ti on s ab n ab n ab n RE M RE M IN V GR O CF O DI S X PR O D DAC C AL L AL L 1 IV R LTO PD I ID V UAI MA S PR SI Z E WT H LEV LO S S IS S U E IF R S ab nC F O 1. 00 ab nD IS X -0. 25 1. 00 ab nP R O D -0. 34 -0. 14 1. 00 DAC C -0. 01 0. 02 -0. 07 1. 00 RE M A L L 0. 02 -0. 85 0. 64 -0. 05 1. 00 RE M A L L 1 -0. 47 -0. 73 0. 37 -0. 01 0. 76 1. 00 IV R 0. 18 0. 07 -0. 21 0. 24 -0. 16 -0. 19 1. 00 LTO 0. 03 0. 20 0. 09 -0. 09 -0. 11 -0. 20 0. 02 1. 00 PD I -0. 31 0. 15 0. 25 -0. 03 0. 02 0. 07 -0. 04 -0. 02 1. 00 ID V 0. 08 -0. 28 -0. 02 0. 05 0. 20 0. 20 0. 01 0. 01 0. 00 1. 00 UAI -0. 11 0. 25 0. 09 0. 06 -0. 14 -0. 15 0. 01 -0. 00 0. 01 -0. 02 1. 00 MA S 0. 08 -0. 11 -0. 07 0. 02 0. 05 0. 05 -0. 01 0. 03 0. 00 -0. 01 0. 07 1. 00 IN V P R 0. 13 -0. 15 -0. 02 -0. 08 0. 11 0. 05 -0. 15 0. 58 -0. 22 0. 34 -0. 17 0. 39 1. 00 SI Z E 0. 05 0. 01 0. 14 0. 01 0. 07 -0. 04 -0. 23 0. 06 0. 32 -0. 19 0. 43 0. 08 -0. 07 1. 00 GR OW T H 0. 13 0. 03 -0. 26 0. 06 -0. 17 -0. 12 0. 09 -0. 04 -0. 23 0. 09 -0. 16 -0. 07 0. 01 -0. 16 1. 00 LEV 0. 02 -0. 11 0. 15 -0. 07 0. 16 0. 09 -0. 09 0. 05 0. 03 0. 08 0. 07 0. 01 0. 08 0. 28 0. 10 1. 00 LO S S -0. 08 -0. 07 -0. 08 0. 07 0. 01 0. 12 0. 14 -0. 08 -0. 30 0. 13 -0. 17 0. 04 0. 12 -0. 41 0. 09 -0. 02 1. 00 IS S U E -0. 36 0. 32 0. 27 -0. 10 -0. 11 -0. 04 -0. 19 0. 14 0. 35 -0. 15 0. 20 -0. 01 -0. 09 0. 10 -0. 11 -0. 02 -0. 17 1. 00 IF R S -0. 22 0. 34 0. 18 -0. 23 -0. 17 -0. 16 0. 02 0. 15 0. 18 -0. 24 0. 05 -0. 44 -0. 28 -0. 03 -0. 11 -0. 02 -0. 26 0. 32 1. 00 CRI S IS 0. 05 -0. 08 -0. 03 0. 02 0. 04 0. 04 -0. 00 0. 00 -0. 05 0. 05 -0. 01 0. 13 0. 11 0. 01 -0. 07 0. 02 0. 08 -0. 06 -0. 19 No te s: ab nCF O is th e a bn or m al ca sh fr om o pe ra tio ns ; ab nD IS X is th e a bn or m al d is cr etio na ry e xp en se ; ab nPR O D is th e a bn or m al pr od uc tio n e xp en se ; D A C C is th e d isc re tio na ry a cc ru als ; I V R , LTO PD I, I D V , M A S an d U A I ar e th e or th og ona liz ed ra nk tra ns for m ati ons of the c ount ry -sp eci fi c cu ltu ral sco res fr om H ofst ed e (1 98 0) ; I N V P R O is th e ran k t ran sfo rm at io ns of th e av er ag e of th e fi ve va ria bl es fr om L a P or ta e t a l. ( 1998) ; S IZ E is the na tur al loga rit hm of m ar ke t va lue of e qui ty; GR OW T H is na tu ra l lo ga rit hm of th e b oo k t o ma rk et va lu e r ati o; L E V is th e s um of to ta l li ab ili tie s d by tot al as se ts; L O S S is a n indi ca tor va ria bl e e qua l to 1 if ea rni ngs in ye ar t a re ne ga tive a nd 0 ot he rw ise ; I S S U E is a n ind ic ato r v ar ia ble e qu al to 1 if the tot al iss ua nc e of e qui ty and de bt is la rge r t 5% of the ye ar e nd tot al as se ts and 0 ot he rw ise ; I F R S an in dicat or v ar iab le eq ual to 1 if acco un tin g st an dar ds in ye ar t a re IF R S an d 0 o th er w ise ; C R IS IS is an in dicat or v ar iab le eq ual to 1 if th e y ea 2008 or 2009 an d 0 o th er w ise .

4.4

L

ONG

-

TERM ORIENTATION

The results of the regression analysis concerning accrual based earnings management show that LTO has a negative significant effect on the model of DACC. This finding suggests that managers in cultures with high long-term orientation make less use of accrual-based earnings management. Therefore, hypothesis 1a is supported.

Additionally, the results of the regression analysis concerning real earnings management shows that LTO is significant and positively related with all individual models of REM. These findings suggest that long-term orientation is associated with less earnings management through sales discounts and discretionary expenses, while more through overproduction. Finally, LTO has a negative significant effect on both models of REMALL. This indicates that managers in cultures distinguished by long-term orientation employ less real earnings management. Overall, the results are consistent with hypothesis 1b, which states that there will be a positive relationship between the Long Term Orientation dimension of national culture and the magnitude of real earnings management.

4.5

I

NDULGENCE VS

.

R

ESTRAINT

The results of the regression analysis concerning accrual based earnings management shows that IVR has a positive significant effect on the model of DACC. This finding suggests that managers in cultures with indulgence make more use of accrual-based earnings management. Therefore, hypothesis 2a is supported.

Furthermore, the results of the regression analysis concerning real earnings management show that IVR is positive and significantly related to the models of abnCFO and abnDISX. In addition, the coefficient of IVR is significantly negative when related to the model of abnPROD. These results indicate that managers employ less earnings management through sales discounts, discretionary expense reduction and overproduction. The results are confirmed, as the coefficient of IVR is significantly negative for both aggregated measures of real earnings management.

Although a positive relation between IVR and REM was expected based on the accounting values of Gray, extended by Borker (2013), the results are consistent with hypothesis 2b and indicate a negative relation between the Indulgence vs. Restraint dimension of national culture and the magnitude of real earnings management. This indicates a trade-off between accrual based and real earnings management. Managers in high indulgence societies may less be inclined to undermine the wealth of the stakeholders with potential harmful earnings management. On top of this, managers do not need status objects to achieve status as they prefer basic and natural human drives related to enjoying life and having fun (Hofstede, 2010). There might be less motivation to meet or beat targets or other financial goals through the use of real earnings management, because bonuses, status objects and other rewards are less important in indulgence societies.

SIZE is positively associated with DACC, whilst negatively associated with REM. The results indicate that bigger firms engage more in accruals management, while less in real earnings management. The DACC result seems inconsistent with the expectations, since bigger firms tend to have less opportunity to manage earnings due to more monitoring. LOSS is not significant for the model of DACC, while it is positive for REM. These firms facing losses seem to manage earnings through sales discounts and overproduction, which is consistent with Gray et al. (2015). ISSUE is positively related to DACC. However, ISSUE is negatively associated with REM, which indicates that firms issuing debt or equity tend to make more use of DACC, as opposed to the more potential harmful REM. CRISIS is only significant for DACC and shows that, consistent with Cimini (2014), financial crisis is associated with less earnings management. Consistent with Pacheco Paredes and Wheatley (2017), IFRS is positively associated with earnings management through sales discounts and overproduction, while it is negatively associated with earnings management through discretionary expense reduction.

Table 5: Regression results - all countries

VARIABLES abnCFO abnDISX abnPROD REMALL DACC REMALL1

LTO 0.0156*** 0.0573*** 0.0162*** -0.0402*** -0.0151*** -0.0721*** (0.000770) (0.000915) (0.000627) (0.00107) (0.000752) (0.00121) IVR 0.0402*** 0.0292*** -0.0390*** -0.0696*** 0.0669*** -0.0695*** (0.00108) (0.000981) (0.000701) (0.00130) (0.000934) (0.00156) SIZE 0.0112*** 0.00101*** -0.00297*** -0.00417*** 0.00732*** -0.0121*** (0.000269) (0.000336) (0.000228) (0.000439) (0.000302) (0.000376) GROWTH 0.0334*** 0.0247*** -0.0566*** -0.0778*** 0.00964*** -0.0593*** (0.000880) (0.00105) (0.000750) (0.00144) (0.000882) (0.00127) LEV 0.00606* -0.209*** 0.137*** 0.335*** -0.111*** 0.207*** (0.00343) (0.00427) (0.00295) (0.00571) (0.00358) (0.00501) LOSS -0.0962*** 0.00673*** 0.00919*** 0.0159*** -0.00223 0.0917*** (0.00185) (0.00245) (0.00166) (0.00337) (0.00176) (0.00271) ISSUE -0.155*** 0.152*** 0.0687*** -0.0771*** 0.0109*** -0.000742 (0.00154) (0.00190) (0.00130) (0.00255) (0.00157) (0.00213) IFRS -0.0770*** 0.188*** 0.0351*** -0.151*** -0.133*** -0.109*** (0.00169) (0.00197) (0.00136) (0.00254) (0.00181) (0.00239) CRISIS 0.0288*** -0.0270*** -0.0255*** 0.000992 -0.0220*** -0.00213 (0.00209) (0.00285) (0.00196) (0.00393) (0.00196) (0.00325)

Industry Yes Yes Yes Yes Yes Yes

Observations 148,606 150,104 139,163 139,163 136,049 148,606

4.6

R

OBUSTNESS CHECK AND ADDITIONAL TESTS

To test whether the results are sensitive to the research design, a sensitivity analyses is performed. First, to test the validity of the results, an additional regression is performed where the non-significant control variables are removed. The result of this test (untabulated) shows no different conclusions. This indicates that the regression models are robust and the results are not influenced by non-significant control variables. In addition, in the next model, the original and Hofstede’s (1980) most used cultural variables will be added to the model, as they may have an impact on the magnitude of earnings management. By adding a model with the inclusion of a measurement of investor protection, it is controlled whether higher levels of investor protection moderate the influence of culture. Finally, it is checked whether there is a trade-off between earnings management through accruals and the management of earnings with the use of real activities, which would be in line with Zang (2012).

4.6.1

O

THER CULTURAL VARIABLES

Other studies have shown that Hofstede’s (1980) other cultural values have an impact on the use of both accrual based and real earnings management (e.g. Han et al, 2010; Gray et al. 2015; Pacheco Paredes and Wheatley, 2017). Using the following model, it is tested whether the other cultural variables influence the relationship of IVR and LTO on the magnitude of earnings management:

!"#/!"## = !_!+ !_!!"# + !_!!"# + !_!!"# + !_!!"# + !_!!"# + !_!!"# + !_!!"#$

+ !_!!"#$%& + !_!!"# + !_!"!""#$ + !_!!!"## + !_!"!"#$ + !_!"!"#$#$ + !_!" Here, REM is the aggregated measurement of discretionary expense and production costs, or operating cash flows and discretionary expense; IVR, LTO, PDI, IDV, UAI, and MAS the country-specific cultural scores from Hofstede (1980); other variables are mentioned before. The additional regression analysis shows that the relation between IVR and LTO on both forms of EM is not affected by adding the other cultural values. However, with the inclusion of the other cultural dimensions SIZE becomes also positively associated with both, while LOSS is now negatively associated with just one measurement of REM. CRISIS becomes negatively associated with all sorts of earnings management.

4.6.2

P

OWER

D

ISTANCE

I

NDEX

(PDI)

Power distance is characterized by acceptance and expectation of power distribution between people (i.e. high levels of authority vested in top managers). Following Gray’s (1988) model, PDI is strongly related with secrecy, which indicates the restricted disclosure of accounting information. Accounting systems are mostly used to benefit those in power in the organization (Hofstede, 2001). Gray (1988) and Dechow et al. (1995) find evidence suggesting that in high power distance societies, when top managers are powerful and information sharing is low, accrual based earnings management is higher. The results presented in this paper indicate the same, but they are not significant. The real earnings management models provide mixed results, as abnCFO and abnPROD are associated with more REM, but abnDISX is associated with less REM.

4.6.3

I

NDIVIDUALISM VERSUS

C

OLLECTIVISM

(IDV)

Hofstede (1980) suggests that in high individualistic societies, managers are more likely to act in accordance with their own interest, instead of looking after the organization like a family defending their interests. Following Gray’s (1988) model, in individualist countries, managers and accountants show more flexibility in self-governance (professionalism) and measurement (flexible or non-uniform). In contrast to the conservative, cautious approach of measurement to deal with the uncertainty of future events, individual managers are more likely to take risk and report to be more optimistic (Gray, 1988). Even though the relation for DACC is not significant, the results indicate that managers in highly individualistic societies are more likely to manage real earnings. This result is consistent with Han et al.’s (2010) finding regarding accruals management that in individualist countries managers use more earnings management.

4.6.4

U

NCERTAINTY AVOIDANCE INDEX

(UAI)

Uncertainty avoidance is the extent to which a society is able to deal with the fact that the future is unknown, and whether they experience stress and feel uncomfortable when faced with ambiguity and an unknown future (Hofstede, 2001). Guan et al. (2006) based on the extensions of Gray’s (1988) model, document that uncertainty avoidance is positively related to the conservative extension of Gray’s (1988) model. This would imply that the use of accrual based earnings management, an aggressive accounting technique, is limited. Han et al. (2010) confirm the reasoning that rigid regulations and control in high uncertainty societies have a negative relationship on abnormal accruals. The results in this paper indicate that, consistent with Pacheco Paredes and Wheatley (2017), as real earnings management might be destructive for future firm performance, conservative managers could be unwilling to engage in REM. Instead, managers might prefer less damaging accruals based earnings management to the more damaging real earnings management.

4.6.5

M

ASCULINITY VERSUS

F

EMININITY

(MAS)

This dimension indicates the societal preference for competitiveness, as opposed to consensus-orientation (Hofstede, 2010). In highly masculine societies, managers find corporate performance important as it gives them social recognition (Zhang et al. 2015). Masculinity is negatively associated with conservatism and secrecy in Gray’s (1988) model. Masculine managers are more likely to focus on the achievement of financial and personal successes to increase their ego, wealth, and recognition. The results imply that the desire to demonstrate achievements and success leads to more real earnings management. These results are consistent with the Gray et al.’s (2015) regarding accrual based earnings management. Consistent with Zang (2012), the results in this paper indicate that AEM and REM might be substitutes. It might be that in high masculine societies there is more external monitoring (transparency and publicly accountable disclosure of information), which, in turn, reduces the opportunities to engage in abnormal accruals management.

Table 6: Regression results - additional cultural variables

VARIABLES REMALL DACC REMALL1

IVR -0.0487*** 0.0661*** -0.0628*** (0.00134) (0.000950) (0.00161) LTO -0.0536*** -0.0135*** -0.0779*** (0.00105) (0.000729) (0.00118) PDI -0.0464*** 0.000896 0.0133*** (0.00143) (0.00106) (0.00107) IDV 0.106*** 0.00121 0.0748*** (0.00110) (0.000768) (0.000907) UAI -0.104*** 0.0139*** -0.0655*** (0.00121) (0.000873) (0.000983) MAS 0.000372 -0.0264*** 0.00779*** (0.00124) (0.000865) (0.000973) SIZE 0.0194*** 0.00596*** 0.000896** (0.000480) (0.000356) (0.000422) GROWTH -0.0987*** 0.00866*** -0.0639*** (0.00146) (0.000922) (0.00129) LEV 0.235*** -0.108*** 0.148*** (0.00565) (0.00365) (0.00500) LOSS -0.00887*** -0.00234 0.0968*** (0.00335) (0.00176) (0.00272) ISSUE -0.0154*** 0.0135*** 0.0152*** (0.00264) (0.00171) (0.00227) IFRS -0.104*** -0.159*** -0.0690*** (0.00296) (0.00229) (0.00269) CRISIS -0.0161*** -0.0163*** -0.0101*** (0.00367) (0.00195) (0.00308)

Industry Yes Yes Yes

Observations 139,163 136,049 148,606

R-squared 0.374 0.164 0.354

4.6.6

I

NVESTOR PROTECTION

Because investor protection is moderated by culture (Leuz et al., 2003), and because Han et al. (2010) show it is associated with accruals management, this paper controls for the interactions of INVPRO. Investor protection is measured as the mean score across five legal variables:(1) the efficiency of the judicial systems, (2) an assessment of the rule of law, (3) the corruption index, (4) the risk of expropriation and (5) the risk of contract repudiation. Graham et al. (2005) show that in certain situations, managers prefer the use of real earnings management instead of accrual based earnings management. One of the differences between the two forms of earnings management is that accrual based earnings management is more visible due to the scrutiny of auditors and regulators (Francis et al., 2004). Real earnings management activities are more difficult to detect, as they can be masked as normal business decisions. In countries with high investor protection it is more likely that managers reduce the amount of accrual based earnings management and have greater incentives to engage in real earnings management.

4.6.7

S

UBSTITUTION EFFECT

Cohen et al., (2008) and Cohen & Zarowin (2010) find that accrual based earnings management and real earnings management can substitute each other. Managers make a trade-off between accrual based and real earnings management techniques based on their relative costs (Zang, 2012). In countries with limited accounting flexibility or high investor protection firms have less opportunity to engage in accrual based earnings management. For example, Cohen & Zarowin (2010) and Zang (2012) hypothesize that after the passage of SOX, firms use more real earnings management, since accrual based earnings management is more constrained. To measure the substitution effect, this paper includes abnormal accruals (DACC) in the models of real earnings management, and controls for real earnings management in the models of discretionary accruals. The added control variables are adjusted for prior year performance in order to measure the substitution effect.

The following model(s) are used to control for both the effects of investor protection and substitution between accrual based and real earnings management. As two proxies of real earnings management are used, it is necessary to run separate regressions to measure the effects of each individually.

!"#/!"## = !_!+ !_!!"# + !_!!"# + !_!!"# + !_!!"# + !_!!"# + !_!!"# + !_!!"#$% + !_!!"##/!"# + !_!!!"#$ + !_!"!"#$%& + !_!"!"# + !_!"!"## + !_!"!""#$ + !_!"!"#$ + !_!"!"#$ + !_!"!"#$#$ + !_!"

REM is the aggregated measurement of discretionary expense and production costs, or operating cash flows and discretionary expense; IVR, LTO, PDI, IDV, UAI, and MAS the country-specific cultural scores from Hofstede (1980); INVPR the average of the five legal variables from La Porta et al. (1998); DACC/REM the lagged variables to control for the substitution effect; and the other variables

Table 7: Regression results - Investor protection and substitution effects

VARIABLES REMALL REMALL1 DACC DACC

IVR -0.0417*** -0.0560*** 0.0605*** 0.0605*** (0.00152) (0.00186) (0.000987) (0.000989) LTO -0.0616*** -0.0902*** -0.00246** -0.00368*** (0.00160) (0.00156) (0.00109) (0.00109) PDI -0.0431*** 0.0178*** -0.00200* -0.00152 (0.00165) (0.00130) (0.00112) (0.00109) IDV 0.101*** 0.0666*** 0.00960*** 0.00989*** (0.00138) (0.00115) (0.000923) (0.000907) UAI -0.103*** -0.0600*** 0.00602*** 0.00814*** (0.00135) (0.00112) (0.000977) (0.000913) MAS -0.00275* -0.000902 -0.0187*** -0.0190*** (0.00145) (0.00123) (0.000988) (0.000955) INVPR 0.0436*** 0.0643*** -0.0727*** -0.0658*** (0.00568) (0.00547) (0.00416) (0.00407) DACC -0.0215*** (0.00595) -0.0366*** (0.00483) REMALL -0.0332*** (0.00225) REMALL1

-0.0316***

(0.00224) SIZE 0.0196*** 0.00159*** 0.00731*** 0.00641*** (0.000515) (0.000455) (0.000382) (0.000363) GROWTH -0.0998*** -0.0626*** 0.000991 0.00391*** (0.00156) (0.00140) (0.00104) (0.000967) LEV 0.230*** 0.126*** -0.0822*** -0.0901*** (0.00606) (0.00539) (0.00388) (0.00371) LOSS -0.0121*** 0.103*** 0.00527*** 0.00314* (0.00355) (0.00300) (0.00184) (0.00178) ISSUE -0.0198*** 0.0115*** 0.00250 0.0101*** (0.00283) (0.00245) (0.00179) (0.00172) IFRS -0.0919*** -0.0607*** -0.166*** -0.164*** (0.00321) (0.00293) (0.00239) (0.00233) CRISIS -0.0151*** -0.00972*** -0.0138*** -0.0139*** (0.00386) (0.00322) (0.00198) (0.00194)

Industry Yes Yes Yes Yes

Observations 126,507 126,567 117,345 126,702

R-squared 0.350 0.329 0.196 0.181