Culture and accounting : cultural determined uncertainty avoidance and earnings quality

(1)

Culture and Accounting

Cultural determined uncertainty avoidance and earnings quality

Name: Ines Goetz

Student number: 11692871

Thesis supervisor: Alexandros Sikalidis Date: June 24, 2018

Word count: 14,133

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

(2)

Statement of Originality

This document is written by student Ines Goetz 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

In this paper I examine whether culturally determined risk aversion of a CFO leads a reporting style involving less earnings management and better earnings quality. I measure culturally determined risk aversion by the Uncertainty Avoidance Index values of Hofstede (1980), respective to the CFOs’ nationality. Earnings management is proxied for by abnormal discretionary accruals computed according to the modified Jones model used in Kim et al (2012), and three different measures of real earnings management, namely abnormal cashflows from operations, abnormal discretionary expenses and abnormal costs of production. I find that the CFOs’ uncertainty avoidance is positively correlated with the negative magnitude of discretionary accruals and abnormal discretionary expenses. This relationship is found to be more pronounced for firms with CFOs that score low on Hofstede’s power distance index scale.

(4)

Contents

Introduction ... 5

Institutional Background, Theory & Literature Review ... 6

Hypothesis development ... 11

Research method and design ... 13

Results and Discussion ... 19

Conclusion ... 32

References ... 34

Appendice A: Libby box ... 38

Appendice B: Regression Model of Kim, Park and Wier (2012) ... 39

Appendice C: Cultural dimensions’ index scores for all countries covered by Hofstede(2004) ... 41

Appendice D: Correlation table ... 43

Appendice E: Further interaction regression models ... 44

Appendice F: Further regression output tables ... 45

(5)

Introduction

In the past, the combination of accounting and culture has been a rarely investigated field. However, nowadays in course of globalization that increasingly affects all parts of our life and the business world on a day-to-day basis, it is obvious that investigation on the effect of culture in the accounting profession might lead us to interesting and useful insights.

In my thesis I want to answer the question whether culture has an impact on accounting choices. Expressed more specifically, I am investigation in the question whether cultural determined risk aversion embedded in the nationality of top management will lead to less earnings management and hence better earnings quality.

I believe finding an answer to this question is important for several reasons: First, the progressing process of globalization leads every one of us to face the impact of different cultures in an increasing sphere of our life, especially doing business; Second, evidence on this empirical question can be interesting and helpful not only for boards and regulators, but also for auditors and investors that want to get an additional clue about the risk-aversion implemented in a CFO’s reporting style; Third, an answer to my research question could possibly provide employers with knowledge that can help them choosing the right person for a job in top management, according to the needs of the firm.

To find an answer to this question, I will measure risk aversion tied to nationality according to the uncertainty avoidance index of the cultural dimensions framework by Hofstede (1980) and examine this score in relation to accounting conservatism implemented in the respective firms’ financial statements, measured by an abnormal discretionary accruals measure as well as three different proxies for real earnings management, namely abnormal cash flows from operations, abnormal discretionary expenditures and abnormal production costs.

(6)

Institutional Background, Theory & Literature Review

Organizational Behavioral Theory

The basis for my research provides the upper echelons theory developed by Hambrick and Mason (1984). It is founded on several previous research [Cyert and March, 1963; March and Simon, 1958] opposing the inertial organization theory by Hall (1977) and contrarily providing evidence that an organization’s behavior is largely dependent on it’s top management’s personal characteristics and the top management’s perceptions of a situation and environment other than only the objective economic aspects of a firm and its environment. Following Hambrick and Mason’s paper published in 1984, there have been various studies drawing on their upper echelons theory with supporting findings: Bertrand and Schoar (2003) were one of the first to conduct research about differences in managements’ styles. They remark that a great part of heterogeneity in corporate practices cannot be explained by firm-, industry-, or market-level characteristics leading them to investigate on the influence of individual managers’ characteristics. In their sample they find that the age of managers is positively correlated with them being more conservative. Moreover, holding an MBA degree reveals to be a proxy for more aggressive strategies. Following their footsteps, Dyreng et al (2010) study the effects of individual managers on tax avoidance of firms. Though they cannot identify specific characteristics being responsible for the differences between managers, but they find their Hypothesis supported that the implementation of different individual managers can change a firm’s level of tax avoidance. Ge et al (2011) found evidence that every CFO has his own distinctive style that impacts upon a firm’s reporting practice, however they cannot show a distinctive effect of the variables gender, age and educational background on reporting styles. Some studies however, can identify style-determining characteristics. Bamber et al (2010) for example finds that personal backgrounds can reveal evidence. They find that managers from finance, accounting and legal career tracks, as well as managers born before the second world war and managers with experience in the military show a stronger conservative bias. Parts of this group are also especially favorable towards more detailed and extensive voluntary disclosure. Francis et al (2015) found that the implementation of female CFOs alters accounting conservatism respective to the implementation of a male CFO.

(7)

Cultural determined characteristics

Culture on itself seems to be a very vague and hardly definable characteristic but it is evident that different cultures can lead to different basic characteristics in people. According to Kluckhohn(1951), the probably most widely accepted definition of culture is one offered by Herskovits (1948). It defines culture as being learned, derived from the biological, environmental, psychological and historical components of human existence, being structured, dividend into different aspects, and at the same time being dynamic and variable. Further it describes culture as incorporating regularities that make it analyzable by science, and as being a means the individual adjusts to and uses when expressing himself.

Anthropologic research did during the 20th century come to the conclusion that humans in all kinds of societies experience problems very similar if not the same in their nature during their life. Different cultures do not impact upon the existence of certain issues, but they may lead to different visible outcomes by determining different ways of how to deal with these problems [Kluckhohn and Morgan, 1952; Hofstede, 1980]. These different ways of dealing with issues have to be captured in every culture’s general framework consisting of universal categories of culture [Kluckhohn, 1952]. These universal categories of culture do contain different specific content in every culture, hence they incorporate different cultural characteristics and can be used for relative cultural analysis.

One of the most influential contemporary researchers to distinct different characteristics inherent to a collection of national cultures is Geert Hofstede. He follows Kluckhohn’s definition of culture although using the more restricted definition of culture as “the collective programming of the mind which distinguishes the members of one human group from another”1

in his work. As described in Hofstede (1983), he conducted a large-scale cross-cultural empirical study that led him to develop four dimensions that capture systematically differing characteristics between people of 40 national cultures in 1968 as well as in 1972. One distinctive point in the nature of Hofstede’s data is that it was collected among marketing- and sales employees of one single multinational, who mostly had in common their belonging to the middle-class and young professional age. Hence the only comprehensive major difference in the background of the employees taking part in the study was their nationality. Differences in the participants values due to company culture, profession or social class are prevented in this

(8)

way. Later more countries were added to the study, so that there are index scores available for 74 countries available now [Hofstede, 2004].

The four dimensions developed by Hofstede are power distance, individualism, masculinity and uncertainty avoidance. Power distance has to do with inequality and hierarchy. Societies with low power distance usually have flatter organizational hierarchies and give less importance to social status. Individualism describes the relationship between the individual and the collectivity. Some societies place a heavier emphasis on the individual and individual freedom while others stress collectivism. The cultural dimension masculinity points towards the promotion of values in a culture that are typically perceived as feminine, nurturing values or as masculine, more assertive values. The uncertainty avoidance index measure describes the degree to which people in one culture are averse to uncertainty. The higher the score of one culture on his uncertainty avoidance index, the more do people in this culture dislike to take risk and bear uncertainties. People from cultures with a low uncertainty avoidance index do in general have a better ability to tolerate uncertainty, feel less threatened by the uncertainty that is unavoidable in life and feel less need to compensate for this uncertainty by for example implementing strict rules creating a kind of felt certainty and security.

In this paper I am particularly focusing on Hofstede’s uncertainty avoidance dimension, also combined with the power distance dimension. These two dimensions combined are most frequently put into context with organizational theory in Hofstede (1980), and also with accounting. Hofstede (1980) points out the fact that the same accounting principles can be interpreted very differently by members of different cultures, a finding uncovered in a research case with IBM.

Especially the uncertainty avoidance index shows a closely tied connection to accounting matters. In Hofstede (2004) it is mentioned that accounting itself can be an expression of an uncertainty-avoiding ritual for societies. Hofstede (2004) further cites Gambling (1977) and states that accounting is often more used as a tool to comforting support than to depict facts. Cleverly (1973) compares accountants to the function of the priests in a religion, who serve a vital role in reducing the perceived uncertainty existing in an organization.

Nowadays, the Hofstede framework has great importance all over the field of cultural studies and is frequently considered by multinationals taking cross-border strategic decisions.

Research about culture in accounting and its influence on accounting has been relatively rare in the accounting research literature. However, there have been a couple of accounting

(9)

researchers drawing upon Hofstede (1980). Gray (1988) developed a theory about the influence of culture measured by Hofstede’s dimensions on different national accounting systems. Against the background of today’s global industry context, recently the interest in this topic has grown. A still unpublished working paper by Marra (2017) studies the impact of a CFO’s country of origin, not necessarily consistent with nationality, on accounting quality. His findings confirm that there are large differences existing within the EU, and foreign CFOs seem to be able to improve accounting quality more often than local CFOs. He further makes a distinction between the CFOs being originally from an outsider- or insider-economy. CFOs who’s country of origin is an outsider economy show to have a more beneficial impact on financial reporting quality when appointed as a CFO in an insider economy than vice versa.

Earnings quality

In the accounting literature, earnings quality is an especially popular topic. Earnings quality is high if the financial statements give an adequate picture of the company and its operating performance. Penman and Zhang (2002) define the quality of earnings as “reported earnings, before extraordinary items that are readily identified on the income statement, is of good quality if it is a good indicator of future earnings”. Earnings quality is closely related to and correlates negatively with earnings management. Earnings management is defined by Healy and Wahlen (1999) as “managers use judgement 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”. According to their literature review, two of the main incentives for earnings management are the capital market and contractual incentives. Earnings management is frequently used to positively influence the stock price in the short run or to prevent negative stock market reactions. On the other hand side, managers might use earnings management for personal gain to alter their compensation that is often partly tied to certain accounting measures as a means of aligning managers’ incentives with the company owners.

The earnings quality can be determined by a lot of different factors. There seem to be indicators as to whether a firm will be inclined to engage in earnings management and to what extent. Leuz et al (2003) find an association between the investor protection environment of a country and earnings management levels in the country’s firms. The identity of a country’s economy as outsider economy, developed stock markets, dispersed firm ownership, strong legal investor

(10)

protection and generally strong law enforcement are found to contribute to earnings quality. Further, earnings quality can be impaired by weaknesses in the internal control system [Doyle et al, 2007; Dechow et al, 1996], a dominance of executive directors in the supervisory board [Dechow et al, 1996] and private ownership [Burgstahler et al, 2006].

Kim et al (2017) examine a cross-cultural data set comprising 38 different nations. Their findings support their Hypothesis that managers from countries with languages that use present tense to describe future actions and events engage less in earnings management. The explanation used is that the negative consequences of earnings management may feel more immediate to these managers because their language implies less distance to the future by not grammatically separating the future with the present.

One paper that involves the study of the relationship of culturally determined uncertainty avoidance, measured by Hofstede’s uncertainty avoidance index is the paper by Han et al (2008). They do find an average negative relationship between uncertainty avoidance and earnings management. Guan and Pourjalali (2010) do also find a negative association between Hofstede’s uncertainty avoidance index and earnings management as well as positive relationships between the other cultural dimensions individualism, power distance and masculinity and the extent of earnings management. However, their analysis was only conducted on firm level, not on the level of CFO nationality. Further, the previous’ sample involved firms all around the world, including a large proportion of US firms, while the latter geared their firm nationality sample towards the set of countries included in Hofstede (1980). Neither of them took into account different accounting standards, by that firms might be constrained.

In this paper, I am looking for evidence in a purely European sample, which increases comparability in terms of accounting standards applicable, and on the level of individual CFO nationality.

(11)

Hypothesis development

The upper echelons theory suggests that it is actually the people in an organization and their mindset that determines the behavior of the organization, rather than only objective economic factors. One essential characteristic with a great impact on a person’s values and perceptions is culture. Thus in my study I suppose that the CFO of a firm, who has considerably the most influence over a firm’s financial reporting because of being responsible for it, does considerably influence a firm’s accounting and reporting style.

Accounting can be viewed as a means to reduce uncertainty [Hofstede, 2004]. This can be interpreted in two different ways. First, accounting can help to decrease uncertainty within the firm by generating a record about past economic events serving as a basis to make assumptions about the future and facilitate decision making. Secondly, accounting can mitigate uncertainty in the sense that it provides information about the performance of the firm and its managers to outside investors and creditors. By dint of the financial statements, potential capital providers can evaluate their risk in investing in or lending to a firm and present shareholders and lenders can investigate on the viability of their investment. Thus it seems likely that a cultural mindset with a low score on uncertainty avoidance would feel less need do decrease perceived uncertainty by sticking to a particularly high quality reporting style that does give the most accurate and faithful view of the company’s performance. Such a mindset, though not necessarily wanting to falsify any information should be still more lenient towards a little earnings management if otherwise seemingly more troublesome or bothering procedures or events can be avoided like this.

Moreover, there is risk involved in earnings management [Han et al, 2008]. Once the management of earnings becomes public, extensive negative stock market reactions can be visible, and although lowering cost of capital initially, in the long run earnings management will lead to higher cost of capital [Dechow et al, 1996]. Even worse, regulators and politics can impose punishments on the firm, if it is believed to manage earnings.

There is some literature about the influence on overconfident managers on a firm. Overconfidence can be assumed to correlate to a low uncertainty avoidance index defined by Hofstede (1980), because overconfidence implies the contrary of anxiety. Further, there are findings that overconfident managers are more willing to take risks in terms of their own compensation contracts and are more eager to pursue risky projects [Gervais et al, 2011], are more prepared to issue earnings forecasts and do more favorably assess their own abilities to

(12)

contribute to the firm value [Libby and Rennekamp, 2011]. More optimistic CFOs tend to overinvest [Campbell et al, 2011], and overconfident managers do also lead to less conservative accounting methods within a firm [Ahmed and Duellman, 2012]. Schrand and Zechman (2012) find that the optimistic bias that overconfident managers have in reporting can easily lead firms into a vicious cycle, slipping into committing accounting fraud because they might have to engage in stronger and stronger earnings management to conceal their excess optimism in the statements of previous periods.

The risk associated with engaging in earnings management and the lower risk affinity provide a reason why I expect CFOs with a culturally determined higher uncertainty avoidance score, to be less inclined to engage in earnings management. They do also have less a problem of unintentionally incorporating an optimistic bias which might force them to drift into earnings management in later periods.

Summing up the above, my research Hypothesis is the following:

H1: Firms with CFOs with a nationality with a high score on uncertainty avoidance do less frequently engage in earnings management.

Further, linking organizational theory to his cultural dimension indexes, Hofstede (1980) makes a distinction between firms combining low uncertainty avoidance with high power distance and firms incorporating low power distance and high uncertainty avoidance. Accordingly, the former are more strongly relying on power of the upper management to regulate organizational functioning, while the latter insure proper operation of the entity by mainly relying on formal rules.

If my first Hypothesis is supported by the data, CFOs from countries with low uncertainty avoidance are prone to higher proxies of earnings management. A more power distant management style should then lead to employees being more inclined to follow instructions by their boss, in this case the CFO, even if they might not agree with him. A low power distance implies flatter hierarchies in organizations. A CFO from a country with a lower power distance is thus likely to have a more anti-authoritative management style, leaving more discretion in tackling the work tasks to the employees themselves.

Thus I secondly hypothesize:

H2: In firms with CFOs with a nationality with a high score on power distance, the impact of the CFO’s culturally determined uncertainty avoidance on earnings management is more pronounced.

(13)

Research method and design

In this paper, I aim to investigate upon the effect that the culturally determined uncertainty avoidance of a CFO has on earnings quality of a firm. Culturally determined uncertainty avoidance is proxied for using the Uncertainty Avoidance Index Score developed by Hofstede and measured for 70 countries respectively. Earnings quality is proxied for by earnings management, which is an inverse function of earnings quality. For visualization of my research concept, please see the Libby box in Appendix A.

Data Collection

I will focus my analysis on European firms, specifically the firms comprised by the EuroSTOXX600 Index. Since all countries part of the European Union are obligated to adopt IFRS beginning from the start of the year 2005, this is a way to control for potentially different regulatory settings that could influence a firm’s accounting choices and in particular managerial discretion. I get the financial data for the six years in the sample period from 2012 through 2017 for the firms through the WRDS databases, primarily using Compustat global, supplementing this data by the data items inventories, market capitalization and the market-to-book ratio obtained from Datastream. Further, information about governance is acquired using the former KLD-databank, now called MSCI.

My independent variable is CFO risk aversion (CFO_UAI), measured as the respective score to the CFO’s nationality on Hofstede’s uncertainty avoidance index scale. For the CFO nationalities, I could obtain 328 observations from BureauVanDijk, which I accessed through the WRDS database. These do however contain repetitions and observations that do not fall within my sample period. The rest of the CFO nationalities were hand collected by consulting annual reports and websites of the EUROSTOXX600 companies. To the extent that annual reports and websites of the sample companies were not sufficient to obtain all the data needed, I further consulted websites and annual reports of other former or present employers of the respective CFOs and of companies where they might hold board positions, as well as online press releases and other sources which I could find through the search engine Google. If the nationality was not stated explicitly in any of the sources consulted, assumptions about the nationality were made taking into account biographical data such as place of birth, place of secondary and higher education and main country of residence. Wherever the information of the biographical data was considered ambiguous or insufficient, the CFO nationality was treated as a missing value. Altogether, I collected 2735 CFO nationality observations. After

(14)

merging the data with the data obtained from Compustat and dropping observations with missing values, as well as firm year observations with CFOs with double or triple nationalities, there are 970 firm year observations available to perform my statistical regression analysis. The biggest national groups of CFOs in my sample are CFOs from Germany, the United Kingdom and France. Since especially Germany and the United Kingdom are often used as examples for the difference between high- and low-UAI countries by Hofstede (2004), this can be considered as a good fundament of my analysis.

CFONationality Frequency Percentage of total observations (N=22) New Zealand 4 0.41 Brazil 5 0.52 Czech Republic 5 0.52 India 8 0.82 Argentina 10 1.03 South Africa 10 1.03 Belgium 12 1.24 Denmark 13 1.34 Netherlands 15 1.55 Ireland 16 1.65 Austria 19 1.96 Spain 19 1.96 Australia 23 2.37 Norway 26 2.68 United States 26 2.68 Italy 39 4.02 Finland 45 4.64 Switzerland 70 7.22 Sweden 80 8.25 France 160 16.49 United Kingdom 164 16.91 Germany 201 20.72 Total 970 100

Table1: CFO Nationality Frequency distribution

In the second step I assign a risk averseness value and the corresponding power distance index, index for individualism and the index for masculinity to each nationality value, using the level on the uncertainty avoidance index and the power distance index and the two other cultural dimensions for every country developed in the cultural framework by Hofstede (1980). The

(15)

distribution of the UAI-scores of the sample is depicted in Graph1. The decrease in nationality indicators when CFO nationality is proxied for by the UAI index score from 22 to 16 is due to the fact that multiple countries score equally on the UAI index scale. France, Spain and Argentina are all assigned a UAI index of 86. The UAI score of New Zealand and South Africa is both respectively equal to 49, and Ireland and the United Kingdom both score 35. Switzerland and Belgium are separated according to language in Hofstede’s framework. For my analysis I took the average of the separate values as single index values for the respective country.

Data Treatment

I will estimate the relation between culture and earnings quality using the regression model of Kim , Park & Wier (2012), making small modifications to it to suit my specific research topic. In particular I replaced their independent variable describing social corporate responsibility by my independent variable which is culturally determined uncertainty avoidance, named UAI_CFO, containing uncertainty avoidance index score by Hofstede that is ascribed to the respective firm’s CFO’s nationality. Furthermore, I eliminated the variable associated with the listing of the company in an American company popularity index since this variable does not apply to my EuroSTOXX600 sample. Furthermore, I do not take into account equity offerings

0 50 100 150 200 250 23 29 35 40 46 49 50 51 53 59 63 65 70 74 75 76 86 95 N u m b er o f firm y ear o b se rv at ion s UAI Score

(16)

and firm age. Due to almost the whole sample engaging a Big4 firm as an auditor, I restrain from involving this control variable. Because my research focuses on CFO characteristics and culture, I add control variables to control for other cultural determined characteristics than uncertainty avoidance, namely power distance (PDI), cultural individualism (IDV) and the masculinity of cultural values (MAS) as well as the CFO’s gender (CFO_M), a variable that takes the value 1 if the CFO is male and 0 if the CFO is female. Besides that, I do not only concentrate on absolute discretionary accruals (ABS_DA) but I do also examine the relationship with the first model on the relative value of discretionary accruals (DA), because this might give us further insight on the direction of earnings management. These are both different discretionary accruals proxies (DA_PROXY). I end up with the following two models:

(1) DA _PROXY= + UAI_CFO + COMBINED_RAM + SIZE

+ MB + ADJ_ROA + LEV RD_INT

GOVERNANCE + PDI + IDV + MAS

+ CFO_M +

(2) RAM_PROXY = + UAI_CFO + ABS_DA + SIZE + MB

+ ADJ_ROA + LEV RD_INT

GOVERNANCE + PDI + IDV + MAS

+ CFO_M

SIZE is a variable controlling for firm size by market capitalization, MB is the market-to-book ratio, ADJ_ROA the industry adjusted ROA, LEV indicates the firm’s leverage measured by long term debt, and RD_INT stands for R&D intensity. The GOVERNANCE variable captures the governance mechanisms of the firm. I compute it adding up the governance strength dummy variables “Corruption and Political Instability” and “Financial System Instability” and deducting the dummy variables for governance weaknesses “Governance Structures”, “Controversial Investments”, “Business Ethics” and “Governance other concerns”. Like this, I obtain a governance score that can theoretically take values between -4 and +2, where positive values indicate overall strong governance, negative values indicate overall weak governance and a score equal to zero is governance neutral. In my actual sample all governance scores lie between -1 and +1.

In these two models, earnings quality is measured by (1) discretionary accruals, and by (2) real activities manipulation. Earnings management is an inverse function of earnings quality, it

(17)

obscures a firm’s true underlying economics in the financial statements. Discretionary accruals are the probably most widely used proxy for earnings management. However, real activities manipulation is a form of earnings management that is also gaining increasing attention in the financial accounting literature. Cohen et al (2008) found that many firms use real activities management as a supplement to or a substitute for accruals management, because it is harder for auditors to detect and nearly impossible for regulators to persecute. On the basis of these previous research findings, association of my independent variable with both of these types of earnings management should be examined in order to find evidence concerning my Hypothesis. COMBINED_RAM is a variable capturing real earnings management. Following Kim, Park and Wier (2012), I compute it as a function of abnormal cash flows from operation (AB_CFO), abnormal production costs (AB_PROD) and abnormal discretionary expenses (AB_EXP), as COMBINED_RAM= AB_CFO - AB_PROD + AB_EXP. Different to Kim, Park and Wier (2012) I do due to a lack of available data not define discretionary expenses (EXP) as a sum of R&D (RD), Advertising (AD) and SG&A (SGA) expenses but as EXP=RD+SGA.

I do again follow the paper of Kim, Park and Wier (2012) and estimate abnormal discretionary accruals according to the modified Jones Model. The Jones Model after Jones (1991) and modifications of it have been widely used in the accounting literature to calculate abnormal discretionary accruals [Kothari et al (2005), DeFond and Subramanyam (1998)]

To control for variations due to industry specific business models I perform the estimation regressions for abnormal discretionary accruals and real earnings management indicators in loops by groups of firm year observations formed using the Global Industry Classification Sectors. Sectors for which not at least ten firm year observations are available are excluded from the analysis.

My second Hypothesis examines the impact of the CFO’s nationality’s inherent power distance on the relationship between the CFO’s culturally determined uncertainty avoidance and earnings quality. Because higher power distance generally leads to higher centralization of power and thus more impact of the CFO personally on actions of the firm, I predict that higher PDI scores lead to a stronger association between the UAI assigned to the CFO and the corresponding firm year observation, and earnings management. To test whether these predicted conditional differences are supported by my data, I use a standard interaction model. First, I have to split my data into two groups that can be compared. Therefore, I create the dummy variable high_PDI that takes the value of 1 for every firm year observation with a PDI

(18)

value higher than the median, here p50=35, and 0 for every firm year observation with a PDI value lower than the median. Observations that are equal to the median of the PDI value distribution are dropped for this analysis. Besides, the interaction variable UAIPDI=UAI_CFO*high_PDI is created, to capture the excess effect of high-PDI observations over low-PDI firm year observations on the coefficient for UAI_CFO. The models for the analysis of the second Hypothesis are as follows:

(3) DA _PROXY = + UAI_CFO + high_PDI + UAIPDI + COMBINED_RAM + SIZE + MB

+ ADJ_ROA + LEV RD_INT

GOVERNANCE IDV + MAS

+ CFO_M +

(4) RAM_PROXY = + UAI_CFO + high_PDI + UAIPDI

+ ABS_DA + SIZE + MB + ADJ_ROA

+ LEV RD_INT GOVERNANCE + IDV

(19)

Results and Discussion

After merging all data and dropping observations with lacking data items, there are 970 firm year observations available to conduct data analyses. My sample comprises 229 firms out of nine different Global Industry Classification Sectors with a number of firm year observations ranging between one and five. The most strongly represented Sectors in my sample are Industrials and Materials. Male CFOs are prevailing in my sample and make up 91% of all observations. Sector N Percentage Energy 54 5.57 Materials 174 17.94 Industrials 230 23.71 Consumer discretionary 115 11.86 Consumer staples 88 9.07 Healthcare 124 12.78 Information Technology 73 7.53 Telecommunication 52 5.36 Utilities 60 6.19 Total 970 100

Table 2: Sample distribution after sector

If firms do engage in real earnings management regarding the cash flows from operations, we would normally expect them to overstate their cash flows from operations and manipulate their sales recording, leading to a positive value of abnormal cash flows from operations. In the descriptive statistics in Table 3 however, the mean abnormal cash flows from operations is negative and close to zero, suggesting that on average, firms in my sample do not engage in operational cash flow manipulation. Similarly, a positive abnormal expenditure mean indicates that firms do on average not cut discretionary expenses in order to manage earnings. An abnormal production mean of 1.099 does however point to a part of my sample firms being heavily involved in overproduction. The standard deviation here of 17.9 is very big though, indicating a great degree of dispersion in this matter among my sample firms. Looking at the descriptive statistics of my accruals management proxy it is showed that on average, firms do engage in managing accruals. A view at the numbers for relative discretionary accruals as compared to absolute discretionary accruals reveals that in my sample, negative accruals management is prevailing.

(20)

Mean

Standard

Deviation Median min max

1st Quartile 3rd Quartile ABS_DA 0.04853 0.02128 0.0444 0.0127 0.1249 0.0337 0.0578 DA -0.04834 0.02205 -0.0444 -0.1396 0.0903 -0.0578 -0.0337 AB_CFO -0.00329 0.04967 -0.0085 -0.1803 0.2773 -0.0347 0.0218 AB_PROD 1.09867 17.90871 -0.4144 -59.4716 133.6970 -6.4724 5.5270 AB_EXP 0.17636 0.08807 0.1691 0.0104 0.4999 0.1308 0.2286 COMBINED_RAM -0.76958 16.58871 0.5253 -75.8404 42.3794 -5.4031 6.5752 lag_SIZE 16.44964 1.30181 16.3784 13.8778 19.3514 15.4269 17.4194 lag_MB 2.83735 2.06368 2.3450 0.4600 11.9500 1.4100 3.5200 lag_ADJ_ROA 0.04978 0.04572 0.0473 -0.0973 0.1824 0.0228 0.0748 lag_LEV 0.18400 0.11050 0.1759 0.0002 0.5064 0.1068 0.2468 RD_INT 0.03924 0.05447 0.0198 0.0001 0.2673 0.0040 0.0474

Table 3: Descriptive statistics for selected variables

Notes: This table presents the descriptive statistics for the main inputs in my regression models. The minimum (maximum) represents the minimum (maximum) after winsorizing at the 1st_{percentile (99}th_percentile).

The model fit of my models is rather low. The independent variables in my first regression model can only explain between 4.94 and 6.17 percent of the change in the dependent variable for DA and ABS_DA respectively. Model 2 does also have a low model fit to estimate the dependent variables COMBINED_RAM (R2=0.0411) and AB_PROD (R2=0.0419). For estimating abnormal cashflows and abnormal expenditures however, the second model is a very good model fit. It can explain 45.16% and 38.77% in the change of the dependent variable here respectively.

Examining the results of my regression using regression model 1, I find that the uncertainty avoidance score after nationality is significantly positively correlated on a 95% confidence level with the absolute level of accruals management (see Table 4). This means that the more uncertainty avoidant a CFO is, the higher is the level of accruals management. This finding is

(21)

contradictory of my first Hypothesis, that CFOs with a high level of uncertainty avoidance would less frequently engage in earnings management.

Coefficient Std. Error t P>t [95% Conf. Interval] UAI_CFO 0.0001631** 0.000064 2.55 0.011 0.000038 0.000289 COMBINED_RAM 0.000022 0.000041 0.54 0.590 -0.000058 0.000103 lag_SIZE -0.000376 0.000565 -0.67 0.506 -0.001485 0.000732 lag_MB -0.0012246*** 0.000398 -3.08 0.002 -0.002005 -0.000445 lag_ADJ_ROA 0.007774 0.017894 0.43 0.664 -0.027341 0.042889 lag_LEV 0.0324939*** 0.006494 5.00 0.000 0.019749 0.045239 RD_INT -0.007416 0.012810 -0.58 0.563 -0.032554 0.017723 Govscore -0.0043502** 0.001692 -2.57 0.010 -0.007671 -0.001029 PDI -0.0001249* 0.000072 -1.74 0.083 -0.000266 0.000016 IDV 0.0001422* 0.000084 1.69 0.092 -0.000023 0.000308 MAS -0.000060 0.000038 -1.59 0.112 -0.000134 0.000014 CFO_M 0.0040988* 0.002426 1.69 0.091 -0.000661 0.008859 Interval 0.037140 0.013327 2.79 0.005 0.010985 0.063294

Table 4: Absolute discretionary accruals

Notes: This table presents the output of a regression using Model 1, on the value of absolute discretionary accruals, for my sample that after merging all relevant data and dropping all missing values consists of 970 firm year observations (N=970). R2 _=0.0617

* indicates significance at the 10% level; ** indicates significance at the 5% level; *** indicates significance at the 1% level

Coefficient Std.Err. t P>t [95% Conf. Interval] UAI_CFO -0.0001555** 0.000067 -2.33 0.020 -0.000286 -0.000025 COMBINED_RAM -0.000026 0.000043 -0.60 0.552 -0.000109 0.000058 lag_SIZE 0.000118 0.000589 0.20 0.842 -0.001038 0.001274 lag_MB 0.0012602*** 0.000415 3.04 0.002 0.000447 0.002074 lag_ADJ_ROA -0.006630 0.018659 -0.36 0.722 -0.043246 0.029987 lag_LEV -0.0268576*** 0.006772 -3.97 0.000 -0.040147 -0.013568 RD_INT 0.006080 0.013357 0.46 0.649 -0.020133 0.032293 Govscore 0.004566** 0.001765 2.59 0.010 0.001103 0.008029 PDI 0.000106 0.000075 1.41 0.160 -0.000042 0.000253 IDV -0.000161* 0.000088 -1.83 0.067 -0.000334 0.000011 MAS 0.000058 0.000039 1.48 0.140 -0.000019 0.000136 CFO_M -0.0041857* 0.002529 -1.65 0.098 -0.009149 0.000778 Interval -0.031936 0.013897 -2.30 0.022 -0.059209 -0.004664

Table 5: Relative discretionary accruals

Notes: This table presents the output of a regression using Model 1, on the value of relative discretionary accruals, for my sample that after merging all relevant data and dropping all missing values consists of 970 firm year observations (N=970). R2 _=0.0494

(22)

Taking into account the relative magnitude of earnings management (see Table 5), the story becomes clearer. Using relative discretionary accruals instead of absolute discretionary accruals as the dependent variable of my regression, I have a negative coefficient for the CFO’s UAI, that is also significant at the 95% confidence level. This unfolds that there is not only more accruals management present in firms with CFOs that are more uncertainty averse, but that the accruals management also gets more negative with the level of CFO uncertainty aversion rising. Though these findings are from some point of view contradictory to my first Hypothesis, but however, they are not illogical either. Ahmed and Duellman (2011) find that overconfident managers are less conservative in their reporting style. This does in return mean that risk averse, alias rather underconfident managers, will have a more conservative reporting style. Accounting conservatism though, does not mean that earnings management is absent. Contrarily, accounting conservatism is associated with the earlier recognition of losses, compared to gains, which can lead to a kind of negative reporting bias. CFOs that score higher on Hofstede’s uncertainty avoidance scale might be less optimistic or just more cautious reporting gains. This caution might lead them to recognize more negative accruals, exhibiting a stronger propensity to engage in negative accruals management. My findings are further consistent with the finding of DeFond and Subramanyam (1998) who find that firms exposed to greater litigation risk are more likely to engage in income-decreasing earnings management. Though I did not consider the actual risk exposure of the firms in my sample, but higher uncertainty avoidance levels with CFOs might lead to perceived higher risk exposure at the same level of actual risk, leading these CFOs to take more use of income-decreasing accruals management.

The output of regression model 2 shows that a higher level of CFO uncertainty avoidance is associated with lower abnormal cashflows from operations, lower abnormal discretionary expenses and lower abnormal levels of production, though only the negative relationship between the value of abnormal discretionary expenses and the CFO’s UAI is significant on a 95% confidence level (see Table 7). Nevertheless, observing the single pair of variables correlations doing a correlation analysis (see Appendice D), I can show a negative correlation between CFO_UAI and AB_CFO as well as AB_PROD that are significant on the 95% and on the 99% confidence level respectively.

(23)

Coefficient Std. Error t P>|t| [95% Conf. Interval] UAI_CFO -0.000182 0.000111 -1.64 0.101 -0.000399 0.000035 ABS_DA 0.8170339*** 0.055654 14.68 0.000 0.707817 0.926251 lag_SIZE 0.0038043*** 0.000971 3.92 0.000 0.001898 0.005711 lag_MB 0.0040259*** 0.000687 5.86 0.000 0.002677 0.005375 lag_ADJ_ROA 0.4835536*** 0.030700 15.75 0.000 0.423306 0.543801 lag_LEV -0.002237 0.011227 -0.20 0.842 -0.024269 0.019796 RD_INT 0.066039*** 0.022131 2.98 0.003 0.022609 0.109469 GovProxy 0.0021906* 0.001222 1.79 0.073 -0.000207 0.004588 PDI 0.0004403*** 0.000124 3.54 0.000 0.000197 0.000684 IDV 0.000057 0.000145 0.39 0.696 -0.000228 0.000342 MAS -0.000065 0.000066 -0.99 0.323 -0.000194 0.000064 CFO_M -0.0103792** 0.004156 -2.50 0.013 -0.018536 -0.002222 Interval -0.1461802*** 0.023037 -6.35 0.000 -0.191389 -0.100972

Table 6: Abnormal Cash flows from operations

Notes: This table presents the output of a regression using Model 2, on the value of abnormal cash flows from operations, for my sample that after merging all relevant data and dropping all missing values consists of 970 firm year observations (N=970). R2 _=0.4516

Coefficient Std. Err. t P>t [95% Conf. Interval] UAI_CFO -0.0004513** 0.000212 2.13 0.034 -0.000868 -0.000035 ABS_DA -0.6683517*** 0.106809 -6.26 0.000 -0.877959 -0.458745 lag_SIZE -0.0250685*** 0.001864 -13.45 0.000 -0.028727 -0.021410 lag_MB 0.0075838*** 0.001319 5.75 0.000 0.004995 0.010172 lag_ADJ_ROA 0.1999612*** 0.058919 3.39 0.001 0.084336 0.315586 lag_LEV -0.1751181*** 0.021547 -8.13 0.000 -0.217403 -0.132834 RD_INT 0.3820731*** 0.042472 9.00 0.000 0.298723 0.465423 GovProxy -0.0096059*** 0.002345 -4.10 0.000 -0.014208 -0.005004 PDI 0.000309 0.000238 1.30 0.196 -0.000159 0.000777 IDV -0.000100 0.000279 -0.36 0.720 -0.000647 0.000447 MAS -0.0002678** 0.000126 -2.13 0.034 -0.000515 -0.000021 CFO_M 0.001728 0.007977 0.22 0.829 -0.013927 0.017382 Interval 0.654620 0.044211 14.81 0.000 0.567857 0.741383

Table 7: Abnormal discretionary expenditures

Notes: This table presents the output of a regression using Model 2, on the value of abnormal discretionary expenditures, for my sample that after merging all relevant data and dropping all missing values consists of 970 firm year observations (N=970). R2 _=0.3877

(24)

Altogether, the results indicate that the more risk averse the CFO is, the more does a firm cut discretionary expenditures. This is a very common form of real earnings management. Moreover, abnormal cash flows of operations and cost of production do decrease with an increase in the CFO’s UAI. I follow that the firms in my sample with highly risk averse CFOs do more strongly engage in real earnings management associated with discretionary expenditures, they are however less likely, or at least not more likely than other firms to engage in real earnings management by overstating cash flows from operations or to overproduce to reduce cost of goods sold. The decrease in discretionary expenditures associated with higher levels of uncertainty avoidance does confirm Gervais et al (2011), who found that overconfident managers are more likely to pursue risky projects. In my study, discretionary expenditures are defined as the sum of sales, general and administrative costs and research and development expenses. Especially research and development projects have a great inherent risk and do often consume great amounts of capital without any security about the value of the outcome. CFOs with a higher level of uncertainty avoidance might shy away

Coefficient Std. Err. t P>t [95% Conf. Interval] UAI_CFO -0.037392 0.050681 -0.74 0.461 -0.136851 0.062067 ABS_DA -12.350220 25.497130 -0.48 0.628 -62.386950 37.686510 lag_SIZE 0.289652 0.445035 0.65 0.515 -0.583705 1.163009 lag_MB 0.099767 0.314885 0.32 0.751 -0.518177 0.717711 lag_ADJ_ROA 45.93211*** 14.064920 3.27 0.001 18.330460 73.533760 lag_LEV -4.738093 5.143588 -0.92 0.357 -14.832110 5.355921 RD_INT 6.266111 10.138880 0.62 0.537 -13.630890 26.163110 GovProxy 0.893464 0.559791 1.60 0.111 -0.205096 1.992023 PDI -0.066927 0.056912 -1.18 0.240 -0.178614 0.044761 IDV -0.080513 0.066501 -1.21 0.226 -0.211017 0.049992 MAS -0.048639 0.030060 -1.62 0.106 -0.107630 0.010352 CFO_M -0.245327 1.904231 -0.13 0.898 -3.982278 3.491624 Interval 7.027934 10.554020 0.67 0.506 -13.683760 27.739620

Table 8: Abnormal production cost

Notes: This table presents the output of a regression using Model 2, on abnormal production cost, for my sample that after merging all relevant data and dropping all missing values consists of 970 firm year observations (N=970). R2 _=0.0419

(25)

Coefficient Std. Err. t P>t [95% Conf. Interval] UAI_CFO 0.036705 0.050652 0.72 0.469 -0.062696 0.136106 ABS_DA 12.448240 25.482330 0.49 0.625 -37.559460 62.455950 lag_SIZE -0.312415 0.444777 -0.70 0.483 -1.185265 0.560435 lag_MB -0.088052 0.314702 -0.28 0.780 -0.705637 0.529534 lag_ADJ_ROA -45.23738*** 14.056760 -3.22 0.001 -72.823020 -17.651740 lag_LEV 4.540431 5.140604 0.88 0.377 -5.547727 14.628590 RD_INT -5.847961 10.133000 -0.58 0.564 -25.733420 14.037500 GovProxy -0.899990 0.559466 -1.61 0.108 -1.997912 0.197932 PDI 0.067564 0.056879 1.19 0.235 -0.044059 0.179186 IDV 0.080201 0.066462 1.21 0.228 -0.050228 0.210630 MAS 0.048226 0.030043 1.61 0.109 -0.010731 0.107183 CFO_M 0.231802 1.903126 0.12 0.903 -3.502981 3.966584 Interval -6.449923 10.547900 -0.61 0.541 -27.149600 14.249750

Table 9: Real earnings management combined

Notes: This table presents the output of a regression using Model 2, on real earnings management combined, using COMBINED_RAM calculated as AB_CFO - AB_PROD + AB_EXP as dependent variable, for my sample that after merging all relevant data and dropping all missing values consists of 970 firm year observations (N=970). R2 _=0.0411

from the risk of incurring major costs without the security of payoff. In addition, in the course of sales, general and administrative costs, many costs are incurred for future periods. An example is that a firm might purchase office supplies once a year to get sales volume discounts. Highly risk averse CFOs though might prefer to purchase the same goods in smaller volumes more often throughout the year to avoid incurring costs for the acquisition of goods that might not be consumed or needed later due to changes in circumstances.

On basis of my findings associated with negative accruals management and negative discretionary expenses with firms with CFOs with higher UAI scores, I reject my first Hypothesis that predicted such firms to engage less frequently in earnings management. Contrarily I found, that these firms exhibit a greater likeliness to involve in negative accruals management and real earnings management concerning discretionary expenditures.

(26)

Coefficient Std. Err. t P>t [95% Conf. Interval] UAI_CFO 0.0004419*** 0.000119 3.71 0.000 0.000208 0.000676 high_PDI 0.0192822** 0.008891 2.17 0.031 0.001820 0.036744 UAIPDI -0.0003907*** 0.000143 2.74 0.006 -0.000671 -0.000111 COMBINED_RAM 0.000073 0.000049 1.48 0.141 -0.000024 0.000170 lag_SIZE -0.000588 0.000810 0.73 0.468 -0.002178 0.001003 lag_MB -0.001034* 0.000607 1.70 0.089 -0.002227 0.000159 lag_ADJ_ROA 0.020873 0.024148 0.86 0.388 -0.026554 0.068299 lag_LEV 0.041783*** 0.008654 4.83 0.000 0.024786 0.058780 RD_INT 0.000504 0.016336 0.03 0.975 -0.031580 0.032588 GovProxy -0.0026804*** 0.000888 3.02 0.003 -0.004424 -0.000937 IDV 0.000030 0.000106 0.28 0.776 -0.000178 0.000238 MAS -0.0001337** 0.000059 2.27 0.023 -0.000249 -0.000018 CFO_M 0.003223 0.002916 1.11 0.269 -0.002503 0.008949 Interval 0.035545 0.019353 1.84 0.067 -0.002463 0.073553

Table 10: Uncertainty avoidance – power distance interaction model (ABS_DA)

Notes: This table presents the output of a regression using Model 3 with absolute discretionary accruals as dependent variable. In this regression variable UAIPDI that is computed as a product of the uncertainty avoidance index of the CFO (UAI_CFO) and the proxy variable high_PDI that is 1 for firms with a CFO who is assigned a high power distance index, and 0 for firms with a CFO assigned a low power distance index. The threshold for high and low values is the median (p50=35). Because of eliminating all firm year observations with a PDI value equal to the median, the sample size for this analysis decreased from 970 firm year observations to only 605 (N=605). R2 _=0.0999

The interaction coefficient in model 3, that is the coefficient for the interaction variable UAIPDI that captures the difference between my two sets of firms with high and low PDI values, reveals that firms with high PDI values of the CFO do indeed exhibit lower levels of accruals earnings management than firms with CFOs that score higher on Hofstede’s PDI scale (see Table 10). The difference is significant on a 99% confidence level. Comparing the regressions for the CFO_UAI on relative discretionary accruals for high- and low-level PDI firms respectively (see Table 11 and 12), the relationship between the UAI and the value of discretionary accruals is both negative, but the coefficient for uncertainty avoidance is only significant, on a 95% confidence level, for firms with CFOs from countries with lower power distance cultures. For firms with CFOs scoring high on power distance, the coefficient for CFO_UAI is also slightly negative, but it is too close to zero to be significant, indicating that among this group of firms, the uncertainty avoidance of the CFO has a minor impact on the extent of accruals management.

(27)

Coefficient Std. Err. t P>t [95% Conf. Interval] UAI_CFO -0.000065 0.000089 -0.73 0.468 -0.000240 0.000110 COMBINED_RAM -0.000066 0.000072 -0.91 0.361 -0.000208 0.000076 lag_SIZE 0.000693 0.001078 0.64 0.521 -0.001427 0.002813 lag_MB 0.000246 0.000833 0.30 0.768 -0.001392 0.001884 lag_ADJ_ROA 0.028473 0.034009 0.84 0.403 -0.038390 0.095337 lag_LEV -0.0377*** 0.011166 -3.38 0.001 -0.059652 -0.015748 RD_INT -0.017525 0.022385 -0.78 0.434 -0.061536 0.026485 GovProxy 0.0026974** 0.001253 2.15 0.032 0.000234 0.005161 IDV -0.000014 0.000118 -0.12 0.907 -0.000245 0.000217 MAS 0.000033 0.000099 0.34 0.735 -0.000161 0.000227 CFO_M -0.003913 0.003878 -1.01 0.314 -0.011536 0.003711 Interval -0.051341 0.024246 -2.12 0.035 -0.099010 -0.003672

Table 11: Relative discretionary accruals – high PDI firms

Notes: This table presents the output of a regression using Model 1 for the group of firms with CFOs that score high on Hofstede’s power distance index scale. The threshold for high and low values is the median (p50=35). Relative discretionary accruals (DA) are used as dependent variable. Because this analysis examines only a subgroup of all sample firms and all firm year observations with a PDI value equal to the median were eliminated, the sample size for this analysis decreased from 970 firm year observations to only 402 (N=402). R2 _=0.0676

Coefficient Std.Err. t P>t [95% Conf. Interval] UAI_CFO -0.000456** 0.000203 -2.24 0.026 -0.000857 -0.000055 COMBINED_RAM -0.000102 0.000067 -1.51 0.133 -0.000235 0.000031 lag_SIZE 0.000171 0.001505 0.11 0.910 -0.002797 0.003139 lag_MB 0.0020095** 0.000939 2.14 0.034 0.000157 0.003862 lag_ADJ_ROA -0.0935772** 0.036321 -2.58 0.011 -0.165220 -0.021935 lag_LEV -0.0480956*** 0.016278 -2.95 0.004 -0.080204 -0.015987 RD_INT 0.008398 0.026128 0.32 0.748 -0.043137 0.059934 GovProxy 0.0025781* 0.001454 1.77 0.078 -0.000290 0.005447 IDV 0.000148 0.000574 0.26 0.797 -0.000985 0.001280 MAS 0.0001889** 0.000076 2.49 0.014 0.000039 0.000339 CFO_M -0.001798 0.004993 -0.36 0.719 -0.011646 0.008049 Interval -0.040327 0.051466 -0.78 0.434 -0.141841 0.061188

Table 12: Relative discretionary accruals – high PDI firms

Notes: This table presents the output of a regression using Model 1 for the group of firms with CFOs that score low on Hofstede’s power distance index scale. The threshold for high and low values is the median (p50=35). Relative discretionary accruals (DA) are used as dependent variable. Because this analysis examines only a subgroup of all sample firms and all firm year observations with a PDI value equal to the median were eliminated, the sample size for this analysis decreased from 970 firm year observations to only 203 (N=203). R2 _=0.2071

(28)

The difference between high-level PDI firms and low-level PDI firms with regards to abnormal discretionary expenditure is statistically not significant (see Table 13). Comparing the regression analysis for the high-level PDI firm set and for the low-level PDI firm set shows that abnormal discretionary expenses have a stronger negative relationship with CFO_UAI among the firms of the latter set (see Appendice F, Tables 3 and 4).

To evaluate whether these findings support or contradict my second Hypothesis, it is needed to further clarify the Hypothesis’ content. I predicted that “In firms with CFOs with a nationality with a high score on power distance, the impact of the CFO’s culturally determined

Coef. Std. Err. t P>t [95% Conf. Interval] UAI_CFO -0.0006423** 0.000280 -2.30 0.022 -0.001192 -0.000093 PDI 0.0008768** 0.000418 2.10 0.036 0.000057 0.001697 UAIPDI -0.000148 0.000251 -0.59 0.555 -0.000641 0.000344 ABS_DA -0.656976*** 0.120774 -5.44 0.000 -0.894162 -0.419790 lag_SIZE -0.0289554*** 0.002472 11.72 0.000 -0.033809 -0.024101 lag_MB 0.0095754*** 0.001738 5.51 0.000 0.006163 0.012988 lag_ADJ_ROA 0.096640 0.069473 1.39 0.165 -0.039797 0.233077 lag_LEV -0.1765001*** 0.027057 -6.52 0.000 -0.229637 -0.123363 RD_INT 0.3376467*** 0.042512 7.94 0.000 0.254158 0.421135 GovProxy -0.0054334** 0.002722 -2.00 0.046 -0.010779 -0.000088 IDV 0.000320 0.000339 0.94 0.346 -0.000346 0.000986 MAS -0.00051*** 0.000169 -3.02 0.003 -0.000842 -0.000179 CFO_M 0.008993 0.009175 0.98 0.327 -0.009026 0.027011 Interval 0.678267 0.062628 10.83 0.000 0.555273 0.801262

Table 13: Uncertainty avoidance – power distance interaction model (AB_EXP)

Notes: This table presents the output of a regression using Model 3 with abnormal discretionary expenditures as dependent variable. In this regression variable UAIPDI that is computed as a product of the uncertainty avoidance index of the CFO (UAI_CFO) and the proxy variable high_PDI that is 1 for firms with a CFO who is assigned a high power distance index, and 0 for firms with a CFO assigned a low power distance index. The threshold for high and low values is the median (p50=35). Because of eliminating all firm year observations with a PDI value equal to the median, the sample size for this analysis decreased from 970 firm year observations to only 620 (N=620). R2 _=0.4583

uncertainty avoidance on earnings management is more pronounced.” I have found, that the impact the CFO’s culturally determined uncertainty avoidance on earnings management is quite different from what I expected. Instead of decreasing the absolute level of earnings management, higher uncertainty avoidance scores of CFOs were found to lead to stronger negative accruals management and greater negative discretionary expenses. Viewing my

(29)

second Hypothesis against this background, findings that support this Hypothesis would exhibit significantly more negative coefficients for CFO_UAI in the regression with DA_PROXY=DA and RAM_PROXY=AB_EXP for the firm group with high_PDI=1 than for the firm group with high_PDI=0. What I found is the exact contrary. I found that the impact of the CFO’s culturally

Coef. Std. Err. t P>t [95% Conf. Interval] UAI_CFO -0.0004313*** 0.000122 -3.55 0.000 -0.000670 -0.000193 high_PDI -0.0185855** 0.009078 -2.05 0.041 -0.036415 -0.000756 UAIPDI 0.0003762** 0.000146 2.58 0.010 0.000090 0.000662 COMBINED_RAM -0.000077 0.000050 -1.54 0.125 -0.000176 0.000022 lag_SIZE 0.000718 0.000827 0.87 0.386 -0.000906 0.002342 lag_MB 0.0010599* 0.000620 1.71 0.088 -0.000158 0.002278 lag_ADJ_ROA -0.022857 0.024656 -0.93 0.354 -0.071281 0.025567 lag_LEV -0.0414149*** 0.008836 -4.69 0.000 -0.058769 -0.024061 RD_INT -0.001041 0.016680 -0.06 0.950 -0.033799 0.031717 GovProxy 0.0026624*** 0.000907 2.94 0.003 0.000882 0.004443 IDV -0.000030 0.000108 -0.28 0.782 -0.000242 0.000183 MAS 0.0001338** 0.000060 2.23 0.026 0.000016 0.000252 CFO_M -0.003186 0.002977 -1.07 0.285 -0.009033 0.002660 Interval -0.038172 0.019760 -1.93 0.054 -0.076979 0.000635

Table 14: Uncertainty avoidance – power distance interaction model (DA)

Notes: This table presents the output of a regression using Model 3 with relative discretionary accruals as dependent variable. In this regression variable UAIPDI that is computed as a product of the uncertainty avoidance index of the CFO (UAI_CFO) and the proxy variable high_PDI that is 1 for firms with a CFO who is assigned a high power distance index, and 0 for firms with a CFO assigned a low power distance index. The threshold for high and low values is the median (p50=35). Because of eliminating all firm year observations with a PDI value equal to the median, the sample size for this analysis decreased from 970 firm year observations to only 605 (N=605). R2 _=0.0959

determined uncertainty avoidance on earnings management is actually more pronounced for firms with CFOs from countries with lower power distance cultures. Concerning the impact of a CFO’s culturally determined uncertainty avoidance on discretionary accruals management I thus reject my second Hypothesis on a 99% confidence level. The coefficient on the interaction variable for the regression on discretionary expenses is not significant. Though it is pointing in the opposite direction of my Hypothesis, but it does also not allow me to reject my Hypothesis on a confidence level.

Examining the reason for a stronger correlation between the UAI score of the CFO for firms with CFOs from low power-distance cultures than for high power-distance cultures, I examine the overall correlation of the UAI score of my observations with their respective PDI scores

(30)

(Table 15). Comparing the correlation between UAI and PDI in my sample to the overall correlation existing between those two indexes among all the nationalities covered by Hofstede I find that that PDI and UAI are positively correlated in general with 10% significance (Table 16), whereas this positive correlation is still a lot stronger in my sample where it is significant on a 99% confidence level.

UAI PDI IDV MAS

UAI 1 PDI 0.7042*** 1 0 IDV -0.5064*** -0.1787*** 1 0 0 MAS 0.1709*** -0.0217 0.1790*** 1 0 0.5005 0

Table 15: Correlation of Hofstede’s culture dimensions in my sample

UAI PDI IDV MAS

UAI 1 PDI 0.2220* 1 0.0573 IDV -0.1818 -0.5983*** 1 0.121 0 MAS -0.071 0.0993 0.118 1 0.5476 0.3998 0.3168

Table 16: Overall correlation of Hofstede’s culture dimensions

This leads me to examine whether the stronger correlation for the CFO_UAI with negative accruals management for low-PDI CFO firms is driven by a stronger CFO_UAI – negative accruals management correlation for firms with CFO’s with culturally determined lower levels of uncertainty avoidance. To investigate on this issue, I create the proxy variable UAI_Proxy that equals 1 for firm-year observations with UAI_CFO levels higher than the median (UAI_CFO=63), and 0 for firm-year observations with UAI_CFO levels lower than the median. Firm-year observations with UAI_CFO levels equal to the median are dropped. Subsequently I run a regression for each of the two UAI_Proxy groups separately but I find that there is only a minor difference between the two groups. Both show a coefficient significant at the 95% confidence level for CFO_UAI on abnormal discretionary accruals (See Appendice G). Thus, I follow that the stronger correlation for CFO_UAI with income-decreasing accruals

(31)

management for low-PDI firms is not driven by a correspondent phenomenon existing in high- and low-UAI firm groups.

A possible explanation for the insignificant correlation for uncertainty avoidance of the CFO and accruals management within my high-PDI sample is that, assuming that the CFO does on average have the same nationality as the firm, in firms with high power distance, the CFO does also not have the ultimate decision power over financial reporting but has to comply with more specific guidance of superiors. Another possibility is that the propensity of managers with high power distance values to rely more on formal rules than on their own experience or feelings found by Hofstede (2004) is responsible for this result. This would explain that CFOs with lower power distance value do rely more on themselves and their experience when making decisions, thus letting their personal risk aversion influence upon their work. At the same time, in CFOs with a higher power distance value, the personal degree of uncertainty avoidance would not significantly impact upon their decisions and hence their reporting style.

(32)

Conclusion

Investigating on the association between a CFO’s culturally determined values and firm’s earnings quality I found in coherence with Han et al (2008), Kim et al (2017) and Guan and Pourjalali (2010), that cultural characteristics of a CFO do indeed have an effect on earnings management. Specifically, I found that higher uncertainty avoidance levels, measured by Hofstede’s Uncertainty Avoidance Index, lead firms to exhibit a stronger propensity to negative discretionary accruals management and higher negative levels of abnormal discretionary expenses. I suggest that the former finding is due to less overconfidence in CFOs with stronger risk avoidance that leads to more accounting conservatism, that is to early recognition of losses while gains are delayed. Ultimately this kind of treatment introduces a kind of negative bias to financial reporting. Cutting discretionary expenditures is also a behavior that might be caused by the culturally determined uncertainty aversion of CFOs. Risk averse managers might shy away from incurring costs for projects that might not be successfully bringing rewards for the firm, e.g. Research and development projects, or they might prefer to buy supplies in small quantities rather than in large quantities for consumption over a longer time to avoid spending funds on supplies that might due to a possible change in circumstances not be necessary any more in future periods. Further, the association between a CFO’s uncertainty aversion and negative discretionary accruals as well as negative abnormal expenditures is more pronounced for firms with CFOs from cultures that have low power distance. An explanation for this might be that managers with higher power distance values do according to Hofstede (2004) rely more on formal rules than on their own experiences when making decisions. Hence, for managers with low power distance values, the impact of personal characteristics on their decision making and their reporting style might be more pronounced.

These findings can be helpful for firms that face CFO changes and assist in the decision whom to appoint as a CFO, as well as for auditors during their work. However, my findings have to be viewed with caution, due to the following limitations existent within my study:

1. The cultural dimensions of Hofstede can only be descriptive on a societal and a group level. It is not possible to conclude on the values of individuals on the basis of Hofstede’s cultural dimensions’ index levels and individuals, such as the CFOs in my study, might not hold consistent values in some or even large areas with the aggregate of their national cultures. 2. Culture can be affected by languages. Hofstede (1980) for example found that in Belgium and Switzerland, countries including several different language areas, people from