Do managers engage in earnings management when a firm is near a broad credit rating change: A European perspective
Sebastiaan Pel S2248697 Antillenstraat 1-47, Groningen 06-45580483 s.pel.1@student.rug.nl Supervisor: S. Wang Word count: 11269
2 Abstract
This thesis focuses to what extent managers of firms in France, Germany the United Kingdom, Italy, Luxembourg and the Netherlands engage in accrual and real earnings management techniques when they are near an upgrade (downgrade) of a broad credit rating category (for example: AA- or AA+). My proxy for accruals earnings management are discretionary accruals which are calculated with the Jones Model (Jones, 1991). The proxies for real earnings management are based on sales manipulation, overproduction and cutting discretionary expenses (Roychowdhurry, 2006; Cohen & Zarowin, 2010). I find that European firms engage in both accrual and real earnings management when they are near a broad credit rating change (upgrade and downgrade). I also find that accruals earnings management is used more in code law countries and real earnings management is used more in common law countries.
Keywords: earnings management, real activities, accrual, credit rating agencies, broad credit rating
3 Table of contents 1. Introduction 4 academic contribution 6 2. Theoretical framework 7 Credit ratings 7
Why are credit ratings important to firms? 8
Credit ratings in Europe 9
Earnings management 10
3. Hypotheses development 13
4. Research methodology 15
Sample 15
Accrual earnings management 16
Real earnings management 17
Control variables 19
Model 20
Sample description 22
Correlation matrix 24
5. Results 24
6. Conclusion and discussion 29
7. References 32
4 1.Introduction
Credit ratings are important to investors because they help them in making investment decisions by showing the creditworthiness of a firm. They do this by minimizing information asymmetry between issuers and borrowers. For issuers the credit ratings are equally
important. A high credit ratings will lead to a higher capital gain when issuing bonds (Ali & Zhang, 2008; Petit, 2001).
A broad credit rating refers to a rating without a plus, middle or minus indicator. If a firm has a rating of either AA+, AA or AA- , it indicates that the firm is in the broad credit rating of AA. Figure 1 shows an illustration of a broad credit rating.
Rating Broad credit rating category
AA+
AA AA
AA -
figure 1. An example of a broad credit rating.
Firms are generally more concerned when they are near a change to another broad credit rating compared to when they are in the middle of a broad credit rating (Ali & Zhang 2008). This is because being upgraded (downgraded) into another broad credit category has significant consequences for a firm (Kisgen, 2006). First of all, it can lead to changes in bond coupon rate. Another reason is that a downgrade might lead to a loss of contracts. Third it will increase (decrease) the cost of raising capital for a firm.
The credit ratings are getting assigned by credit rating agencies. The three biggest credit rating agencies are Standard’s and Poor, Moody’s (both US based firms) and Fitch (based in France). A lower rating exhibits a higher chance that a company will be unable to repay its debt. This leads to a higher risk for potential investors or borrowers. Moody’s bases ratings on profitability, capital leverage, asset quality, cash flow positions, industry related risks and several other factors (Moody’s rating methodologies, 2018). Moody uses the following ranges to categorize firms: “Aaa, Aa, A, Baa, Ba, B, Caa, Ca, C”. In which Aaa is the highest credit rating. These categories are followed by a number (1,2 or 3), denoting whether a company will is near an upgrade (downgrade) or holds a middle position within the broad credit rating category. S&P uses the same system, however they use + and – to instead
5
of the numerical system of Moody’s. Their categories are presented as following:”Aaa, Aa, A, Bbb, Bb, B, Ccc, Cc, C”. (Hull, Predescu & White, 2004). Within the remainder of this
research I will use the credit rating system of Standard’s and Poor, because the firms in my dataset have been rated with their system. Since being upgraded (downgraded) can have severe consequences for a firm, managers are generally concerned with having a good credit rating. To avoid being downgraded or to have a higher chance of getting upgraded, managers might engage in earnings management to attain (prevent) this upgrade (downgrade). Earnings management can be defines as the the managerial discretion that manager exhibit over
accounting choices, reporting choices and real economic decisions to influence the earnings of a firm (Walker, 2013). This leads to the following research question:
Do European based firms engage in earnings management when their current credit rating contains a plus or a minus to attain (prevent) an upgrade (downgrade) to another broad credit rating?
Prior research by Ali & Zhang (2008) found that managers in the United States engage in earnings management when they are near a broad credit rating change. They found that firms with a credit rating that contains a plus sign and firms with a credit rating that contains a minus sign are more likely to inflate earnings than firms who are in the middle of a broad credit rating. They accredit this to the fact that firms that get upgraded (downgraded) to another broad credit rating gain (lose) more in comparison to firms that are in the middle. Besides this, they conclude that credit rating agencies are not sufficiently capable of assessing whether a company’s current period’s earning have been inflated. In a different study, Caton, Chiyachantana, Chua, and Goh (2011) tested if firms that use aggressive accrual earnings management techniques are more often downgraded when their earnings management
techniques are detected by credit agencies. Contrary to what would be the expected outcome, these firms are actually downgraded less than their counterparts who didn’t participate in earnings management. A possible explanation for this is that companies have better operating performance when their cost of capital is lower. This could mean that engaging in earnings management actually has a positive effect on your firm. Jung et al (2011) found that
companies within a broad credit rating in general have managers that are more inclined to use accrual earnings management to boost earnings. This is consistent with Ali & Zhang (2008) their research. In yet another study by Alissa, Bonsall, Koharki & Penn Jr. (2013) it was examined if firms that are above or below their expected credit ratings use real and accrual
6
earnings management to deviate towards their expected category. They found that firms indeed use these techniques.
Academic contribution
This study adds to the literature in several ways. First of all it expands the existing literature about earnings management. Using earnings management to attain (prevent) upgrades
(downgrades) between broad credit ratings has not been studied extensively yet. Ali & Zhang (2008) have studied it with a population of American firms. This study will add to their research by looking at European firms.
Secondly, this study adds to the research about European bond market. Prior research has shown that credit ratings are highly important to managers (Graham & Harvey, 2001; Alissa et al, 2013; Ali & Zhang, 2008; Kisgen, 2006). However, without a big capital market a credit rating might be unnecessary for a firm. The European capital market used to be dispersed because it was forced to mostly exists domestically because of the different currencies being used in Europe (Pagano & von Thadden, 2004). After the formation of the European monetary union the capital market within continental Europe has grown rapidly. Even though the market is still significantly smaller in comparison to the US market there has been a shift from bank lending towards public debt (European Commission Expert group on Corporate Bonds, 2017; Blomkvist, Friman & Korkeamäki, 2018). Based on this, there is sufficient reason to suspect credit ratings are of growing importance for European firms. This validates my research as earnings management is linked to attaining an expected credit rating. It also validates the European setting of my thesis as the use of public debt is of growing importance on the continent. I will expand on the European bond market more in section 2 (theoretical framework). Lastly, this study will add to the literature about the differences between common and code law countries. Several studies have been conducted to assess the differences between earnings management in code and common law countries.
The remainder of this thesis is structured as following. In section two I will discuss the related theory about credit ratings, credit rating agencies and earnings management. In the third section I will develop my hypotheses. In the fourth section I will discuss the sample and the research methodology. The fifth section will contain my results. Lastly, in the sixth section I will conclude and discuss the results that I have found.
7 2. Theoretical framework
Credit ratings
A credit rating is a future opinion regarding the creditworthiness of an obligor (firm) with respects to a specific (class of) financial obligation(s) or a specific financial program. Credit ratings are assigned by credit rating agencies. The biggest credit rating agencies are Standard’s and Poor, Moody’s and Fitch. The three big credit rating agencies hold 95% of the market. In the EU there are 15 certified and registered credit rating agencies (Petit, 2001). A credit rating of AAA can be defined as an extremely strong capacity of the obligor to meet its financial commitments. A rating of C indicates that there is a high chance that the financial commitments of the firm will not be met. The broad credit ratings in between AAA and D can be seen as a downwards slope that indicates the capability of a firm to meet its financial commitments. Lastly, a rating of D indicates that a payment has not been made on the due date. It can also be an indicator that the firm has gone bankrupt. (S&P Global Ratings Definitions, 2018). Appendix 1 shows the qualification for each credit rating according to Standard’s and Poor. Dividing the firms into categories gives information to the capital market about how a firm performs. It gives information about the riskiness too, which can help in making decisions. A bond with a higher risk usually has a higher yield. Bonds with a rating above BB are called investment-grade bonds. A bond with a rating lower than BB is called a speculative bond. Junk bonds also fall into this category. From an investor’s point of view, these type of bonds are high risk and high yield bonds. The high risk is among other reasons associated with the assumption that the issuer might not be able to repay the debt. From the point of view of the issuer, these types of rating mean that their cost of capital will increase. Certain investment firm like insurance companies, but also banks, are often not allowed to invest in these types of bonds due to regulatory reasons. Pension funds are also not allowed to invest in speculative bonds. Firms that have a rating of BB or lower therefore have a limited pool of investors, which leads to a higher interest rate for the issuer (Ali & Zhang, 2008).
Credit rating agencies depend on an “issuer pays” business model. The bank that is issuing the debt for a firm or government pays a credit rating agency to rate their
creditworthiness. Investors are dependable on credit ratings to aid in their investments. It is one of their main sources of information. Therefore, it is highly advised that firms (banks) hire a credit rating agency to rate their creditworthiness. In the past, banks would sometimes
8
rate the bonds themselves. However, this is not sustainable because of the high costs associated with the rating process. Because of the mitigating role that credit rating agencies play in the financial markets, they are sometimes referred to as “the gatekeepers of the financial market” (Petit, 2011). However, the issuer pays business model also creates a high-risk situation. Credit rating agencies might be deterred from rating the issuers
creditworthiness with a poor credit rating because they are dependent on the issuer to pay for their rating. This leads to a serious moral hazard problem in which the principle is being paid by the agent (Mählmann, 2009).
The past has shown that these are real risks. In the events leading up to the financial crisis, the credit rating agencies have subsequently graded junk bonds as investment-grade bonds. This has caused credit rating agencies to being subjected more to stringent regulations. In the EU credit rating agencies are now subjected to certain laws instructed by the European security markets agency (ESMA). These regulations aim to limit the risk of conflict of interest, improve ratings quality, improve methodologies and foster transparency (Petit, 2001). Another way to combat this problem is to have an investor paid business model. Following the growth of an investor paid credit rating agency S&P has improved the information quality of their credit ratings (Xia, 2013).
An initial bond offering (IBO) refers to the situation in which a firm is offering bonds to the market for the first time. The firm is assessed by credit rating agencies and assigned a certain rating, which is a thorough process that takes a significant amount of time. The firm is assigned a rating based on its competitive position, financial characteristics, operational results and future prospects. After this, the firm is most often not assessed in depth again, unless there is a triggering effect (Caton et all, 2011). It is therefore important that a firm receives a good initial rating. Graham & Harvey (2001) confirm that managers view a good credit rating as important. They also state that small firms have a lower credit rating in comparison to bigger firms. Firms make capital structure decisions based on financing, investment and hedging opportunities to reach their target credit rating (Hovakimian, Kayhan & Titman, 2009). But credit ratings are not only important to the firm. They are also
important towards other stakeholders. With the publication of specialized information regarding the creditworthiness of a firm, the information asymmetry that exists will get minimized.
9
The paragraph above gives a broad explanation of what credit ratings mean and what credit rating agencies are. In this paragraph, I will show why credit ratings are important for firms. First of all , research by Graham & Harvey (2001) has shown that credit ratings are highly important to managers. According to their survey of 392 CFO’s, it is the second most important debt related factor. However, this does not explain why credit ratings are important. As I stated before, investors rely on credit rating agencies to share information with them. It is an important factor to decide whether they want to invest in this firm. Secondly, due to
regulations, most banks are not allowed to buy debt from firms that have a credit rating below BB. A change in a credit rating thus directly affects the cost of future borrowings
(Jung,Soderstrom & Yang, 2003). A high credit rating is crucial for the successful sale of obligations (White, 2010). Besides this, other parties might also engage in contract with a firm when they have a certain credit rating. The loss of these contracts incur costs for the firm as well. A low credit rating can also negatively affect customer and employee relationships (Kisgen, 2006). Credit rating agencies reduces information cost, expands the pool of potential lenders and promotes liquidity in the market (Petit, 2011). When a company goes below BB, the firm will incur serious consequences when trying to acquire new capital. A rating of BB- is therefore an important threshold. When a firm goes below that, the pool of lenders
decreases significantly. Similarly, a rating of B+ is as important as going up towards BB will increase the pool of potential lenders significantly. This has serious consequences for the cost of raising capital. Besides this, a change of a broad credit ratings leads to a change of coupon rate. This problem is less apparent when a firm switches from a middle position to a plus or minus within the same broad credit rating. The firm might also be forced to repurchase bonds when they are put in a lower broad credit rating. Another reason for a firm to keep an
investment-grade rating is to keep access to the commercial paper market. Losing this access will incur significant costs for a firm (Ali & Zhang, 2008). To reach the firms expected credit rating, managers are willing to make capital structure decisions based on achieving this rating (Graham & Harvey, 2001).
Credit ratings in Europe
This study focuses on European firms. Therefore, we need to take into account the differences between the US and European corporate bond market. The size of the corporate bond market within the EU is three times smaller than the corporate bond market within the USA (10% of GDP in 2017 compared 31). This could indicate that credit rating agencies are of limited importance in the EU. There are more differences between the US and the EU.
10
Europe is a diverse continent with different law systems in place. Some countries have more developed capital markets than others. The countries in continental Europe adhere a code law system, while the United Kingdom adheres a common law system (Maaijoor & Verstraelen, 2006). There are multiple differences between code law and common law countries. First of all, common law countries have more developed capital markets, while code law countries have less active capital markets. Secondly, common law countries have a diverse pool of investors and high investor protection. This is lower in code law countries. Lastly, common law countries have accounting practices that are influenced by the private sector. This is in contrast with code law countries, where accounting practices are mostly influenced by the government (Ball, 2000). Bank financing is traditionally the most used form of financing in Europe, with the public debt market being less frequently used (Blomkvist et al, 2018). Before the introduction of the euro, the European bond markets used to be mostly domestic, which made the market significantly smaller than the US market. Trading identical financial claims in other currency zones used to be imperfect substitutes and traded at different prices. After the introduction of the European monetary union (EMU), the bond market in Europe changed. With the introduction of a new currency in continental Europe financial market the same size as the US market has opened up (Pagano & von Thadden, 2004). The bond market has since grown rapidly and gained popularity among large numbers of European firms. However, as stated, the European market differs significantly from the US market with its heterogeneous cultures, different law systems and different market developments (Blomkvist et al, 2018). Although, even within Europe, a distinction exists. Continental Europe is generally more reliant on bank financing than public debt financing. This is in contract with the United Kingdom, where the public debt market is more utilized (Pattani, Vera & Wackett, 2011). Pattani further states that having a credit rating from an agency, dramatically enhances the chances of issuing bonds to the public market. This shows that in the United Kingdom, the bond market is intertwined with credit rating agencies.
Earnings management
As I stated in the introduction, earnings management can be defined as the managerial discretion that manager exhibit over accounting choices, reporting choices and real economic decisions to influence the earnings of a firm (Walker, 2013). Healy & Wahlen (1999) define it as changing the firms economic performance by insiders to mislead stakeholders or to
influence contractual outcomes. There are two forms of earnings management. Accrual earnings management and real earnings management.
11
Accrual earnings management (AEM) is done by changing the accruals within the financial statement of a firm. This method has several advantages. First of all, it can be done opportunistically by managers. When a manager prepares the balance sheet of the firm in a certain period, he can use his personal estimation and judgement with some of the entries that have to be done. A second advantage arises from this, because of the timing of the financial statements, accrual earning management makes it possible to use this technique even after the end of the financial year (Enomoto, Kimura & Yamaguchi, 2005). However, accrual earnings management also has disadvantages. First of all, this form of earnings management is more easily detectable by auditors and regulator in comparison to real earnings management. This makes it a riskier form of earnings management. Besides this, accrual earnings management can also lead to reversal in future periods, which in turn means that it will have a negative effect on future quarters or years (Enomoto et al, 2005).
Real earnings management is the second form of earnings management. Managers can manage earnings through altering the timing and scale of operating decisions. These decisions are different from the normal business practices and have as a primary objective to mislead stakeholders on the actual economic performance of the firm (Ge & Kim, 2014). It is near impossible to measure real earnings management, as the actual steps the managers take to manage earnings might be hard to trace. Therefore, researchers often use proxies to find real earnings management. Three proxies that are often used in prior research are sales
manipulation, overproduction and cutting discretionary expenses (Ge & Kimg, 2014; Roychowdhurry, 2006; Alissa et al, 2013). Sales manipulation refers to managers trying to boost sales by using certain techniques. Managers might give discounts or agree to more lenient credit terms. When a firm is giving out discounts on the regular price, it will decrease the cashflows for the current period. Additionally, more lenient credit terms might lead to an increase in uncollectable accounts (Ge & Kim, 2014). Overproduction is a second way in which managers can manage earnings. By producing more units, the overhead costs can be smoothed over more units. This leads to a lower cost per unit and in turn to higher earnings. The products can be added to the inventory, which is added to the balance sheet (Ge & Kim, 2014; Cohen & Zarowin, 2010). However, it also means that the firm might have a lot of unusable inventory of finished goods that they are unable to sell. The firm is likely to incur losses in future periods because of this technique. The firm will also see an increase in inventory costs as they need to store the additional finished goods (Roychowdhurry, 2006). Lastly, managers can cut discretionary expenses. Generally, managers will cut on advertising
12
costs, research and development costs and sales, general and administrative costs. Cutting discretionary costs will increase the cash flows of the firm. Investors use future cashflow as a main indicator for their investment. This is because a firm that has a high cash flow is more likely to pay out bond yields and repay the initial debt. However, cutting advertising might lead to the firm incurring a loss of sales in future periods. Similarly, cutting research and development costs can have a competitive disadvantage for the firm (Ge & Kim, 2014; Alissa et al, 2013; Cohen & Zarowin, 2010; Roychowdhurry, 2016).
Real earnings management has an advantage over accrual earnings management, as it is harder to be detected by auditors, regulators and credit rating agencies. The disadvantage is that the future consequences are often way more costly for a firm compares to accrual
earnings management. Before the Sarbanes-Oxley Act managers would more often engage in accruals earnings management. After SOX managers have been engaging in real earnings management more often. This implies that after the scandals leading up to this new regulation, managers want to avoid being detected and shifted from accrual earnings management to real earnings management. (Cohen et al, 2008). Real earnings management has some advantages for managers. First of all, capital markets overvalue the firm in the periods that earnings management occurred (Cohen & Zarowin, 2010; Ge & Kim, 2014). Second of all, firms that engage in real earnings management also have overpriced bonds when they beat analyst forecasts through this (Ge & Kim, 2014).
One of the most important factors for managers is to keep earnings volatility at a minimum. They prefer to have a smooth earning pattern to look more consistent towards credit rating agencies, as found through a survey by Graham & Harvey (2001). This is also one of the factors in the profitability section that determines the overall credit ratings
(Moody’s, 2018). In another study, almost half of the sampled managers say that they expect that smoother earnings (thus less volatile earnings) ensure that the company stays in its current rating group or even pushes it up a notch (Jung et al, 2013; Graham, Harvey &
Rajgopal, 2005). Therefore, we could say that it might be tempting for managers to engage in earnings smoothing when they can enhance their credit ratings with it, as smoother earnings may lead to a higher credit rating. Earnings management can have severe consequences for cutting costs or investment decisions. Moreover, it has a direct effect on the cash flows of a firm.
13
Managers are willing to meet certain goals, even when it might be detrimental for the value of the firm in the future (Roychowdhurry, 2006; Graham et al, 2005). Thus, managers could also be enticed to manipulate earnings to get a better credit rating. Jung et al (2013) note that there are also reasons why it might not always be optimal to engage in earnings managers. Firstly, it is time consuming and diverts the attention of the manager from improving firm value. Besides this, it can also lead to reputational damage for management and the firm can get demoted if the earnings management is detected.
3. Hypothesis development
A change in credit rating can have a significant impact on a firms cost of capital. This could lead to managers who are close to a credit rating, to make changes in their normal firm operation to achieve their desires goals. Graham & Harvey (2001) found in their survey that managers are willing to change capital structures to reach their goals. They also found that managers are willing to engage in earnings management to reach their goals. If managers are willing to manage earnings to reach goals, they might also be willing to manage earnings to get upgraded or to prevent being downgraded to other broad credit rating categories. This leads to the following hypothesis.
H1. When a firm is near a broad credit rating upgrade (downgrade), the amount of discretional earnings management firms use is higher.
Graham and Harvey (2001) also found that 80% of managers are willing to engage in real earnings management by cutting discretionary expenses. Prior research has found that after the SOX – act managers have been engaging more in real earnings management compared to accrual earnings management (Ge & Kim, 2014). Besides that, CFO’s are very concerned with earnings vitality. Which means that they prefer a smooth earnings path. Both credit rating agencies and investors prefer a firm with a smooth earnings path. Credit rating agencies take earnings smoothness into account when rating a firm. Therefore, I expected managers to engage in real earnings management more when they are near a broad credit rating. This leads to the second hypothesis.
H2. When a firm is near a broad credit rating upgrade (downgrade), the amount of real earnings management firms use is higher.
14
Because of Europe’s heterogeneous cultures, different law systems and different market developments differences can arise in which way firms in individual countries manage earnings. In continental Europe, all of the countries have adopted a code law system.
However, the UK but also Ireland have a common law system in place. Prior research indicated that managers in code law countries have the tendency to engage in earnings management more than their counterparts in common law countries (Maijoor & Verstraelen, 2006; Leuz, 2013). This is because legal protection is lower in code law countries. Besides this, code law countries have on average less developed capital markets. However, as I have pointed out in the theory section, there is a rapid growth in the European bond market. Appendix 2 shows that France, a code law country, currently has the biggest bond market within Europe. This could lead to more earnings management within code law countries as credit ratings will gain importance. Furthermore, due to lower investor protection and lower legal protection this effect might increase. This leads to the third hypothesis.
H3. Firms performing in a code law country engage on average more in earnings management technique than firms in common law countries.
One theory that is commonly used regarding earnings management is the agency theory. This theory states there is an information asymmetry between two parties; the agent and the principal. In this study, the credit agency is the principal and the manager is the agent. Walker (2010) describes two problems that could arise from this. The first one is a moral hazard problem in which the credit agency is unable to observe the choices the manager has made to get to his results. There is a second moral hazard problem that I described earlier, this arises because the principle and the agent in this research are not fully independent from each other. The principle is relying on the agent for his payment, which is a conflict of interest and might lead to the principle being less stringent with its judgement. This will negatively affect other stakeholders of the firm. A second problem is adverse selection, in which the manager has more private information about the firm than the credit agency. The second problem is apparent, as credit agencies only use publicly known information. However, managers still engage in earnings management if they expect the benefits to be higher than the possible agency costs that might arise when the credit agency notices their earnings management (Ali & Zhang, 2008).
15 4. Research methodology
Sample
I retrieved my initial dataset from my supervisor professor Shuo Wang. The dataset contained credit ratings from European, Middle-Eastern and African countries. Furthermore, it contained information regarding credit watches. I removed all the data that I didn’t require for my research. From the initial dataset I kept the credit ratings, the company name and the year in which the rating was assigned. Besides this, I removed every firm entry that did not contain a Gvkey. This research is focused on the European market and therefore, I also removed the Middle-Eastern, African and most of the European countries. According to a distribution (appendix 2) made by Bloomberg in 2017 and used in a report of the European Comission regarding the European bond market, 80% of the current outstanding bonds are accounted for by six countries (European Commission Expert group on Corporate Bonds, 2017). As this study focuses on credit ratings, which are related to bond markets and the cost of capital I decided to focus on these six countries. Countries with underdeveloped bond markets would give little insight in the relation between credit ratings and earnings management. These six countries are Germany, France, the United Kingdom, Italy,
Netherlands and Luxembourg. Germany, the UK and the Netherlands account for most of the bonds issued. Furthermore, I removed companies within the utilities industry (SIC codes between 4400 and 5000) and bank and financial institutions (SIC codes between 6000 and 6500) (Roychowdhurry, 2006). Roychowdhurry (2006) states that firms within these industries are subjected to different accounting practices. Hill et al (2018) confirm this and state that financial firms are not comparable with firms in other sectors. This is in line with Kisgen (2006), who also removed financial firms from his sample, as firms within the
financial sector have different rating criteria. Financial firms are prone to different accounting requirements and can’t freely choose the accounting standard they want to use. Firms that did not have sufficient data to calculate either total accruals or discretionary accruals were also removed from the dataset, this is in line with other research in earnings management according to Kothari, Leone & Wasley (2005). Variables relating to the income statement, balance sheet and cash flow statement were retrieved from compustat. I combined the datasets based on the Gvkey and the fiscal year. The total number of observations is 1982, of which 1841 have a credit rating assigned. The final dataset contained 652 observations with
16
variables relating to discretional accruals and 537 observation regarding real earnings management. The data ranges from 2009 to 2018.
Accrual earnings management
In this thesis I will calculate discretionary accruals with the Jones Model (Jones, 1991). His model estimates discretionary accruals by comparing the total amount of accruals of a firm with expected accruals. The total amount of accruals is defined by income before extraordinary items minus cash flow from operational activities. The expected accruals are estimated through the model. The difference between the total accruals and the expected accruals are the discretionary accruals. Deviations between the total accruals and the expected accruals gives insight in whether a manager has or hasn’t used accrual earnings management. The model is illustrated in formula 1.
Total accrualsit Assetsit − 1 = β0 + β1 1 Assetsit − 1+ β2 ΔRevit −ΔRecit Assetsit − 1 + β3 PPEit Assetsit − 1+εit (1)
In this formula, every firm is denoted as j with the year being denoted as t. The formula estimates discretionary accruals by calculating the missing term for firm jt. Total accruals is estimated by subtracting income become extraordinary items with cash flow from operating activities and is scaled by lagged total assets of firm j in year t-1 (Ali & Zhang,
2008; Yu;2008). ΔRevjt is the change of revenue for firm j in comparison to the previous year
(t-1). ΔRecjt is the change in accounts receivable for firm j compared to the previous year (t-1).
We subtract those from each other and scale by lagged total assets. Lastly, PPE is the gross property, plant and equipment for firm j in year t. We also scale PPE by lagged total assets. Lagged total assets are the assets of firm j in year t-1. The first term (β0) is the estimate of the discretionary accruals. Calculating an absolute value for discretionary accruals is necessary because it is an effective tool in capturing income increasing (decreasing) effects that can occur due to earnings management (Becker, DeFond, Jiambalvo & Subramanyam, 1998). Prior research has shown that firms mostly take action when they are downgraded ex ante (Kisgen, 2009;Alissa et al, 2013). In order to capture accruals earnings management, we need to look at discretionary accruals in the next year. When a firm receives a credit rating in a given year, they will take action on that in the next fiscal year to try and achieve their
17
desired rating. When a firm receives an AA- rating in a certain year, the manager will act on this rating the next year. Because of this, we need to look at the discretionary accruals in year t+1. This is the main variable I will use for testing hypothesis 1. I calculated accruals earnings management based on 2-digit SIC codes with a minimum of 10 firms per year. The result of the regressions of formula 1 are my discretionary accruals.
Real earnings management
To calculate real earnings management among European firms, I will rely on the model used by Roychowdhurry (2006) which is also used in other research by Alissa et al (2013) and Cohen & Zarowin (2010). Real earnings management can be calculated through proxies related to abnormal cashflow, abnormal production and abnormal discretionary expenses. I have calculated the proxies exactly as in Cohen & Zarowin (2010). The first metric, which is abnormal cashflow, relates to managers boosting short term sales while hurting future sales and cashflows (Alissa et al, 2013). I used the following formula to calculate this metric:
CFOAit Assetsit − 1 = β1 1 Assetsit − 1+ β2 Salesit Assetsit − 1+ β3 ΔSalesit Assetsit − 1+εit (2)
In this formula CFOA is the cash flow of operating activities in year t for firm j. Sales are the sales of the firm in year t for firm j and ΔSales is the change in sales compared to year t-1 in year t for firm j. Assets are the lagged total assets. Abnormal cashflow is the actual cash flow minus the cashflow we estimated through the model.
The second metric is abnormal production. This is the willingness from managers to overproduce rather than producing for the expected demand (Alissa, 2013). I calculated this as following. This is calculated as the sum of cost of goods sold and the change in inventory. First of all I calculate COGS. This is a linear function of contemporaneous sales (Cohen & Zarowin, 2010). COGSit Assetsit − 1 = β1 1 Assetsit − 1+ β2 Salesit Assetsit − 1+εit (3)
Secondly, I calculate the growth in inventory as a linear function of the contemporaneous and lagged change in sales, which is in line with prior research from Cohen & Zarowin (2010)
18 ΔINVit Assetsit − 1 = β1 1 Assetsit − 1+ β2 ΔSalesit Assetsit − 1+ β3 ΔSalesit − 1 Assetsit − 1+εit (4)
Cohen & Zarowin (2010) then combine both functions to calculate abnormal production cost in the following formula.
PRODit Assetsit − 1 = β1 1 Assetsit − 1+ β2 Salesit Assetsit − 1+ β3 ΔSalesit Assetsit − 1+ β4 ΔSalesit − 1 Assetsit − 1εit (5)
In formula 3 COGS are the cost of goods sold and Sales is the number of sales. In formula 4,
Δinventory refers to the change in inventory. Formula 5 combines the two previous formulas to create a measure for abnormal production.
The third metric to capture real earnings management are abnormal discretionary expenses. What has to be noted is that my university did not have access to advertising expenses. Therefore, I only used research and development costs and selling, general and administrative costs (which can sometimes contain advertising costs). I modeled the formula for abnormal discretionary expanses as a function of lagged sales instead of a function of currect sales. This is to prevent receiving lower residuals from our regressions when firms manages sales upwards to increase reported earnings (Cohen & Zarowin, 2010). Research and development costs were not available for most of the firms within my dataset. I calculated discretionary expenses by adding selling, general and administrative costs and research and development cost. When research and development costs were not available for a firm in a certain year, I calculated used selling, general and administrative costs as discretionary costs. Formula 6 shows how I calculated abnormal discretionary costs.
DisXit Assetsit − 1= β1 1 Assetsit − 1+ β2 Salesit − 1 Assetsit − 1εit (6)
DisX is the amount of discretionary expenses for firm j in year t. Sales is the same metric as I used above. As with the other formulas, everything gets scaled by lagged total assets. I calculated real earnings management based on 2-digit SIC codes with a minimum of 10 firms per year. With the calculated data, I will create two variables that are indicative for real earnings management. Consistent with prior research, I will create a variable RM1. I multiply abnormal discretionary expenses with minus 1 and add this new measure to abnormal
19
production costs (Zang, 2006; Cohen & Zarowin, 2010). As a second measurefor real
earnings managemen which I label RM2, I will multiply abnormal cash flow from operation and abnormal discretionary expenses with minus 1 and add them to each other. This is also in line with Cohen & Zarowin (2010) and Zang (2006). By multiplying the amounts with minus 1 we can easier determine if a firm is engaging in real earnings management. A higher value of RM1 and RM2 indicates that a firm is engaging in sales manipulation and is cutting discretionary expenses.
Control variables
The control variables that I will use in this thesis have been used in prior earnings management research. Declining liquidity or solvency are indicators that a firm might not be performing optimally. It could be an indicator that a firm is heading to bankruptcy.
Companies that are in financial distress are more likely to engage in earnings management (Alissa et al, 2013). To control for the fact that a firm might engage in earnings management because of a looming bankruptcy I control this with two variables. The Altman Z-score (ALTZ) is a formula which is used to predict bankruptcy. The formula consists of five parts that are added to each other. My dataset has limitations with regards to the price of
outstanding shares. Because of this, I could not calculate the market value of equity for most firms. Therefore, I have used the book value of equity to construct the Altman Z-score in my thesis. According to prior research by Range, Njeru & Waititu (2018) this is a sufficient indicator of bankruptcy in firms. The Altman Z-score is calculated with this formula: Z = 1.2X1 + 1.4X2 + 3.3X3 + 0.6X4 + 1.0X5. In this formula X1 is calculated by dividing
working capital by total assets, X2 is calculated by dividing retained earnings by total assets,
X3 is calculated by dividing earnings before interest and taxed by total assets, X4 is calculated
by dividing the book value of equity by total liabilities and lastly X5 is calculated by dividing
sales by total assets.
Firms that are near a broad credit rating are more likely to decrease their leverage to achieve their desired credit rating (Kisgen, 2006). I have calculated financial leverage (LEV) as the total liabilities divided by the total assets. Furthermore, I control for litigation risks within the industry. Conservatism in accounting practices has a bigger occurrence in settings where litigation is high (Ali & Zhang, 2008). Therefore, earnings management might occur less in high litigation industries. I constructed a dummy variable for litigation (LIT). The dummy has the value 0 when the firm is not in a high litigation industry and the dummy takes
20
the value 1 when it is based within a high litigation industry. In line with prior research done by Ali & Zhang (2008) I labeled the following SIC codes as high litigation firms, 2833-2836, 3570-3577, 3600-3674, 5200-5961 and 7370. I also control for firm size. In line with Jung et al (2013), we define size (SIZE) as the natural log of total assets. A firm with a higher size is associated with lower risk. This is because of the assumption that bigger firms are scrutinized by investors when they engage in earnings management and this might deter them from doing so (Katman & Farooque, 2014). However, smaller firms might have little incentive to engage in earnings management when they can’t achieve an investment grade rating. Lastly, I control for the audit firm that is assigned to a firm. Prior research has found being audited by a Big 4 company, puts a constraint on earnings management and deters firms from engaging in it. Big audit firms have an incentive to deliver high quality audits. When a Big 4 audit firm doesn’t report earnings management when the firm did engage in it, it would be considered an audit failure. If the audit failure gets discovered, it can lead to reputational damage for the audit firm. Besides this, it can also lead to litigation expenses through fines (van Tendeloo &
Verstraelen, 2008). We constructed a dummy variable to determine if a company is audited by a Big 4 audit firm (AUDIT) or not. When a firm is audited by a Big 4 firm the dummy has the value of 1. When the firm is audited by another accounting firm the dummy is a 0. Lastly, I control for the return on assets, which is calculated net income divided by total assets. Model
I will be using the following model to test if being near an upgrade (downgrade) to another broad credit rating leads to an increase in accrual earnings management. DACC is the dependent variable in this model, which is an absolute value. In line with research of Ali & Zhang (2008) I will run two different regressions. In the first regression I will run both credit ratings with a plus or minus in the rating.
𝐷𝐴𝐶𝐶 = β0 + β1CR_ALL + β2ROA + β3LEV + β4LIT + β5SIZE + β6AUDIT + β7ALTZ + ε
Model 1. Regression model for hypothesis 1.
β0 is the constant in this model. CR_ALL is a dummy variable which indicates if a firm has a plus or minus sign in its credit rating. ROA is the return of assets, LEV is the leverage of the firm, LIT is a dummy variable which indicates if the firm is within a high litigation industry. SIZE is the size of the firm, AUDIT is a dummy which indicates if the firm is audited by a Big 4 company and ALTZ is the Altman Z-score of the firm calculated with book value of equity. ε is the error term.
This model will show if managers that are near a broad a broad credit rating change
21
that are in the middle. After this, I will run the following two model which will give insight in the behavior of managers when their credit rating has either a plus of a minus sign.
𝐷𝐴𝐶𝐶 = β0 + β1CR_MIN + β2CR_PLUS + β3LEV + β4LIT + β5SIZE + β6AUDIT
+ β7ALTZ + β8ROA + ε
Model 2. Regression model for hypothesis 1.
β0 is the constant in this model. CR_MIN is a dummy variable which indicates if a firm has minus sign in its credit rating. CR_PLUS is a dummy variable which indicates if a firm has plus sign in its credit rating ROA is the return of assets, LEV is the leverage of the firm, LIT is a dummy variable which indicates if the firm is within a high litigation industry. SIZE is the size of the firm, AUDIT is a dummy which indicates if the firm is audited by a Big 4 company and ALTZ is the Altman Z-score of the firm calculated with book value of equity. ε is the error term.
For my second hypothesis I will run a model which measures the effect of real earnings management on being near a broad credit rating change. I will run this model 5 times with different indicators of REM. I will run the model with the real earnings management variables RM1 and RM2 as dependent variables.
𝑅𝑀1 = β0 + β1CR_ALL + β2ROA + β3LEV + β4LIT + β5SIZE + β6AUDIT + β7ALTZ + ε
Model 3. Regression model for hypothesis 2
𝑅𝑀1 = β0 + β1CR_MIN + β2CR_PLUS + β3LEV + β4LIT + β5SIZE + β6AUDIT
+ β7ALTZ + β8ROA + ε
Model 4. Regression model for hypothesis 2
𝑅𝑀2 = β0 + β1CR_ALL + β2ROA + β3LEV + β4LIT + β5SIZE + β6AUDIT + β7ALTZ + ε
Model 5. Regression model for hypothesis 2
𝑅𝑀2 = β0 + β1CR_MIN + β2CR_PLUS + β3LEV + β4LIT + β5SIZE + β6AUDIT
+ β7ALTZ + β8ROA + ε
Model 6. Regression model for hypothesis 2
RM and RM2 are measures that I constructed with the real earnings management proxies ACF, AP and ADX. RM1 one is calculated by multiplying abnormal discretionary
22
expenses with minus 1 and adding this to abnormal production costs. RM2 is calculated by multiplying abnormal cash flow and abnormal discretionary expenses with minus 1 and adding the result. ACF is a measure of abnormal cash flow, AP is a measure of abnormal production and ADX is a measure of abnormal discretionary expenses. Similarly, as with the discretionary accruals I estimated the proxies for the year t+1 in order to estimate earnings management in the year after the firm achieved its credit rating.
Sample description.
Table 1 contains the distribution of credit ratings among the firms within the dataset that are have a variable for earnings management. In total 482 firms have a credit rating. For all of these firms I have calculated discretionary and real earnings management proxies. The data contains 280 investment-grade credit ratings and 202 speculative credit ratings. Table 2 shows the distribution of firm per country. Luxembourg and Italy are underrepresented in the sample. Appendix 2 shows that the bond market is both countries is smaller compared to the rest of the sample. Therefore, we can accept this distribution as it reflects the actual situation on the European bond market.
AAA AA+ AA AA- A+ A
Number of Firms 0 3 8 1 24 32 A- BBB+ BBB BBB- BB+ BB Number of Firms 33 68 71 40 31 27 BB- B+ B B- CCC+ <CCC+ Number of Firms 52 48 24 11 8 1
23 COUNTRY FREQUENCY FRANCE 152 GERMANY 136 UNITED KINGDOM ITALY THE NETHERLANDS LUXEMBOURG 205 31 89 39 Table2. Distribution of firms per country
Mean SD Min Max 25% 75%
DACC ACF AP ADX REM1 REM2 -.0045 .0019 -.0082 .0637 -.0807 -.0656 .081 . 0894 .2099 .1968 .3282 .2267 -.2495 -.3338 -.7112 -.8798 -1.3151 -.8531 .3432 . 3119 .6258 .6039 1.5056 1.2135 -.0379 -0351 -.0967 -.0097 -.2265 -.1709 .0229 .0448 .0815 .1519 .0776 .0147 CR_MIN .2863 .4525 0 1 0 1 CR_PLUS .3796 .4858 0 1 0 1 CR_ALL .6639 .4729 0 1 0 1 ROA .0437 .0100 -.7941 1.1479 .0387 .0629 LEV .6722 .1877 .1541 1.6359 .5535 .7639 ALTZ 1.4957 .8857 -2.241 6.11 .9497 1.9439 LIT .1135 .3174 0 1 0 0 SIZE 3.9519 .6666 2.2454 5.661 3.4498 4.4084 AUDIT .8696 .3370 0 1 1 1 MTB .8441 1.455 -.0264 8.371 .1352 .8597
Table 3. Descriptive statistics variables.
In table 3 I summarized the descriptive statistics of the data. I winsorized the data at the 2% and 98% to remove outliers from the data. My discretionary accruals have a mean of .0044997. In prior research by Alissa et al (2013) the discretionary accruals have a mean of
-24
0,001. Which means my proxy of accrual management is in line with this. The variables relating to a broad credit rating, namely CR_MIN, CR_PLUS and CR_ALL are dichotomous variables and therefore have a minimum of 0 and a maximum of 1. The mean of CR_MIN is 0.2863. This indicates that roughly 29% of the data has a credit rating with a minus in it. The mean of CR_PLUS is 0,3796 which indicates that around 38% of the data has a credit rating with a plus sign in it. Credit ratings with a plus sign are slightly overrepresented in the sample. It also indicates that slightly more than half of the dataset has either a plus or minus sign on its credit rating.
Correlation
According to Katmon and Farooque (2014) a Pearson r correlation has to be
performed on all dependent and independent variables. The results are in appendix 3. When the correlation between two variables is higher than 80%, it indicates that there is multi collinearity between those variables (Katmon & Farooque, 2014; Hair et al, 2006). However, according to the table is appendix 3 there is no correlation higher than 80% between the variables in the sample. We can conclude that the results of our data are free of multi collinearity. There is a correlation of above 80% between most of the real earnings
management variables. However, they will always be used interchanged from each other. The correlation exists because RM1 and RM2 are constructed from the three original proxies.
5. Results
In this section I will explain the results of my thesis. First of all, I run I will run model 1 and 2. I estimated discretionary accruals with the Jones Model (Jones, 1991) and I regressed for each 2-digit SIC code. I will run an ordinary least square regression on the models and expect a positive coefficient on all three independent variables. Hypothesis 1 states that a firm that is near a change in broad credit rating is more likely to engage in discretional earnings management to inflate earnings. Therefore, the coefficient of the regression for CR_ALL, CR_PLUS and CR_MINUS should be positive. It is expected that a firm that is near a broad credit rating inflates earnings with a higher amount of accruals. Ali & Zhang (2008) state that a firm with a greater incentive to inflate earnings, is expected to have a higher amount of discretionary accruals. The results of model 1 and 2 are shown in table 1.
25
DACC Coef. t-value Coef. t-value
CR_ALL .0148*** 1.92 CR_PLUS .0142*** 1.66 CR_MIN .0198** 2.12 ROA .1695** 2.15 .1711** 2.17 SIZE -.0063 -1.12 -.0059** -1.05 LIT -.0161 -1.31 -.0153** -1.25 AUDITOR .0362* 3.30 .0358* 2.36 ALTZ -.0109 -1.56 -.0104 -1.50 LEV -.0316 -1.27 -.0305 -1.23 CONST .0125 0.36 .0086 0.24 n=481 r²=.0258 n=481 r²=.0265
Table 1. This table show the results for hypothesis 1. The model is estimated on a OLS regression with a panel dataset. The discretionary accruals are calculated with the Jones Model 1991 and are regressed for each 2-digit SIC code. ***, ** and * indicate significance levels of respectively 0.10, 0.05 and 0.01. The results show that being near a broad credit rating is an indicator for discretional earnings management. The dependent variable DACC are the discretionary accruals. The variables CR_ALL indicates that a credit rating has either a + or – sign attached to it. The CR_MIN variable indicates that a firm has a minus sign in its credit rating. For the variable CR_PLUS there is a + sign in the credit rating. ROA is income/total assets in year t. SIZE is the natural logarithm of the total assets in year t. LIT is a dummy variable to show if a company is in a high litigation industry. AUDIT is a dummy variable to show if a firm is audited by a big 4 accounting firm. LEV is the total of liabilities / total assets. ALTZ is a calculation of the Altman Z-score by using book value of equity.
Table 1 shows that the coefficient for CR_ALL, CR_PLUS and CR_MIN are all three positive. This indicates that firms which are near a broad credit rating change have a higher amount of discretionary accruals. Besides this, all three variables are significant on a 10% level. Firms with a minus sign in their broad credit rating are significant on a 5% level. CR_ALL has a coefficient of 0.0148 value =1.92), CR_PLUS has a coefficient of 0.0142 (t-value=1.66) and CR_MIN has a coefficient of 0.0198 (t-value = 2.12). Therefore, we can conclude that firms near a broad credit rating change manage earnings upwards through accrual earnings management.
For hypothesis 2 I will determine if being near a broad credit rating change is an indicator for real earnings management. I will run the same OLS regression as I did for hypothesis 1. According to Cohen & Zarowin (2010) the constructed real earnings
26
management proxies indicate that a higher value shows that a firm engages in earnings management. Hypothesis 2 states that firms that are near a broad credit rating change will have a higher measure of real earnings management. Therefore, we expect a positive coefficient. I will run two multivariate OLS regressions. First, I will run models 3 and 4 to assess if firms with both a plus or a minus sign has an effect on real earnings management. I will also assess if firms with either a plus or a minus sign in its credit rating are more likely to engage in real earnings management. Models 3 and 4 will indicate if a manager near a broad credit rating change will overproduce and cut discretionary expenses. The results are shown in table 2.
RM1 Coef. t-value Coef. t-value
CR_ALL .0999* 3.79 CR_PLUS .0968* 3.34 CR_MINUS .0972* 3.03 LEV -.3524* -4.10 -.3531* -4.10 SIZE -.1036* -5.41 -.1034* -5.38 ALTZ -.0738* -3.01 -.0735* -2.99 ROA .2019 0.76 .2062 0.77 AUDITOR .0177 0.46 .0169 0.43 LIT -.1443* -3.45 -.1444* -3.44 CONST .6332* 5.19 .6348* 5.19 n=383 r²=.1694 n=383 r²=.1635
Table 2. This table show the results for hypothesis 2. The model is estimated on a OLS regression with a panel dataset. The real earnings management proxies are estimated as in Cohen & Zarowin (2010) and are regressed for each 2-digit SIC code. RM1 is an indicator of abnormal production and abnormal discretionary expenses. ***, ** and * indicate significance levels of respectively 0.10, 0.05 and 0.01. The results show that being near a broad credit rating is an indicator for discretional earnings management. The dependent variable RM1 is a proxy for real earnings management. The variables CR_ALL indicates that a credit rating has either a + or – sign attached to it. The CR_MIN variable indicates that a firm has a minus sign in its credit rating. For the variable CR_PLUS there is a + sign in the credit rating. ROA is income/total assets in year t. SIZE is the natural logarithm of the total assets in year t. LIT is a dummy variable to show if a company is in a high litigation industry. AUDIT is a dummy variable to show if a firm is audited by a big 4 accounting firm. LEV is the total of liabilities / total assets. ALTZ is a calculation of the Altman Z-score by using book value of equity.
In table 3 I will show the results of the second proxy I used to calculate real earnings management. I will run a regression on models 4 and 5. This proxy will show if a firm will
27
inflate sales and cut discretionary expenses when the firm is near a broad credit rating change. The expected coefficient for this regression is also positive.
RM2 Coef. t-value Coef. t-value
CR_ALL .0665* 3.63 .0662* 3.26 .0623** 2.81 LEV -.1791* -3.03 -.1806** -3.04 SIZE -.0925* -6.91 -.0927* -6.88 ALTZ -.0662* -4.01 -.0663* -4 ROA .0101 0.05 .0129 0.07 AUDITOR .0932* 3.57 .0930* 3.56 LIT -.0612** -2.10 -.0617** -2.11 CONST .4060* 4.91 .4091* 4.91 n=481 r²=.1725. n=481 r²=.1835.
Table 3. This table show the results for hypothesis 2. The model is estimated on a OLS regression with a panel dataset. The real earnings management proxies are estimated as in Cohen & Zarowin (2010) and are regressed for each 2-digit SIC code. RM2 is an indicator of abnormal cashflow and abnormal discretionary expenses. ***, ** and * indicate significance levels of respectively 0.10, 0.05 and 0.01. The results show that being near a broad credit rating is an indicator for discretional earnings management. The dependent variable RM1 is a proxy for real earnings management. The variables CR_ALL indicates that a credit rating has either a + or – sign attached to it. The CR_MIN variable indicates that a firm has a minus sign in its credit rating. For the variable CR_PLUS there is a + sign in the credit rating. ROA is income/total assets in year t. SIZE is the natural logarithm of the total assets in year t. LIT is a dummy variable to show if a company is in a high litigation industry. AUDIT is a dummy variable to show if a firm is audited by a big 4 accounting firm. LEV is the total of liabilities / total assets. ALTZ is a calculation of the Altman Z-score by using book value of equity.
The results for the regressions of models 3,4,5 and 6 indicate that firms near a broad credit rating change engage in real earnings management. Models 3 and 4 predicted real earnings management on overproducing and cutting discretionary accruals. For all three categories of broad credit ratings the coefficient is positive and significant. CR_ALL has a coefficient of 0,0999 and a t-value of 3,79, CR_PLUS has a coefficient of 0,0968 (t-value 3,34) and CR_MIN has a coefficient of 0,0972 (t-value 3,03). The regression results for model 5 and 6 indicate that firms near a broad credit rating also engage in real earnings management through sales manipulation. CR_ALL has a coefficient of 0,0665 and a t-value of 3,63, CR_PLUS has a coefficient of 0,0662 (t-value 3,26) and CR_MIN has a coefficient of 0,0623 (t-value 3,81). Based on the results of the regressions we can conclude that firms near a broad credit rating
28
change engage in real earnings management through overproduction, sales manipulation and cutting discretionary expenses.
To test hypothesis 3 I regressed models 1,2,3,4,5 and 6 again but separated them with a country dummy. The results showed me in which countries managers engage accrual earnings management and real earnings management when they are near a broad credit rating change. In table 4 I have added the results that are significant on a 1,5 or 10% significance level. Due to the limitation of the dataset, which had limited firms in certain countries I only used France, Germany and the United Kingdom for this regression.
WACC Germany Coef. Germany t-value France Coef. France t-value UK Coef. UK t-value CR_ALL .0097 .93 .0234*** 1.77 .0160 .82 CR_PLUS CR_MIN .0061 .0191 .56 1.37 .0444* .0278*** 3.04 1.84 .0085 .0213 .34 .95 n=111 r²=.1316. n=115 r²=.1958 n=124 r²=.0805 RM1 CR_ALL .0370 1.66 -.0100 -.28 .2362* 4.04 CR_PLUS .0388*** 1.66 -.0226 -.56 .2413* 3.24 CR_MIN .0315 1.03 -.0122 -.30 .2322* 3.38 n=95 r²= .4425 n=91 r²= .3536 n=89 r²=.4030 RM2 CR_ALL .0456*** 1.77 .0243 .99 .1319* 3.12 CR_PLUS .0359 1.33 .0321 1.16 .1376** 2.56 CR_MIN .0711** 2.06 .0052 0.18 .1279* 2.65 n=111 r²=.1659 n=115 r²=.1988 n=124 r²=.2542
Table 4. This table show the results for hypothesis 3. The model is estimated on a OLS regression with a panel dataset. The discretionary accruals are calculated with the Jones Model 1991 and are regressed for each 2-digit SIC code. The real earnings management proxies are estimated as in Cohen & Zarowin (2010) and are regressed for each 2-digit SIC code. The models are regressed per country (Germany, France, the United Kingdom). DACC is an indicator of discretionary accruals. RM1 is an indicator of abnormal production and abnormal discretionary expenses. RM2 is an indicator of abnormal cashflow and abnormal discretionary expenses. ***, ** and * indicate significance levels of respectively 0.10, 0.05 and 0.01. The results show that being near a broad credit rating is an indicator for discretional earnings management. The dependent variable RM1 is a proxy for real earnings management. The variables CR_ALL indicates that a credit rating has either a + or – sign attached to it. The CR_MIN variable indicates that a firm has a minus sign in its credit rating. For the variable CR_PLUS there is a + sign in the credit rating. ROA is income/total assets in year t. SIZE is the natural logarithm of the total assets in year t. LIT is a dummy variable to show if a company is in a high litigation industry. AUDIT is a dummy variable to show if a firm is audited by a big 4 accounting firm. LEV is the total of liabilities / total assets. ALTZ is a calculation of the Altman Z-score by using book value of equity.
29
Table 4 contains the regression results per country (Germany, France and the United
Kingdom). Regarding discretionary accruals we can see in table 4 that only in France there is a significance between accruals earnings management and being near a broad credit rating category. There coefficient on all three categories is positive and therefore we can conclude that managers in French firms engage in accrual earnings management. Secondly, firms from the UK have a significance on all forms of real earnings management. The coefficients are all positive (0,2361 ; 0,2413 ; 0,2322 ; 0,1319 ; 0,1376 ; 0,1279) and are significant. This leads to the conclusion that firms near a broad credit rating, regardless of plus of minus indicator, engage in sales manipulation, overproduction and cutting discretionary expenses when they are near a change. Lastly, in Germany we can see that there is no accruals management to get upgraded (prevent getting downgraded) when firms are near a broad credit rating change. However, German firms do engage in real earnings management when they are near a change. Based on these results I can’t conclude that earnings management happens more in code low countries.
Robustness test
I conducted a robustness test assess if a firm that is near a change from an investment grade bond to a speculative bond or vice versa is an indicator for earnings management as in Ge & Kim (2014). I created a dummy to take the value 1 when a firm has either a BBB- or a BB+ rating. I found that firms within this category only engage is overproduction and cutting discretionary expenses. Firms within this category might not engage in accrual earnings management because it is too easily detectable. Firms in this category might not engage in sales manipulation because they do not have the financial leverage to cut prices. However, the results of this test could also be because of the limited sample size.
6. Conclusions and discussion
Limitations
First of all, I want to address to limitation of the study. The dataset that I acquired ranged from 2009 till 2018. Therefore, it only gives information regarding this period. We can’t assess a change in earnings management behavior that is caused by for example the financial crisis. We also can’t assess the more stringent regulations regarding credit rating agencies in Europe that was introduced in 2009. Besides this the data regarding market value was very minimal. I could only compute a market to book ratio for around 60 firms. Therefore I omitted the market to book control variable from my regressions, even though it is often used in earnings management research.
30
Secondly, my study takes only Germany, France, the United Kingdom, Italy, Luxembourg and the Netherlands into perspective. I chose this because 80% of the
outstanding bonds in Europe are from these countries. However, two things need to be stated. First of all, these are all western European countries with a good economy. Therefore, this study only gives information about western Europe and does not consider earnings
management in Eastern Europe. Besides that, even though the European bond market is mostly situated in these countries does not mean that the bond markets are nonexistent in other European countries. For example, Spain and Portugal have relatively big bond markets too, however, on average their bond markets are smaller than the countries within my sample.
Lastly, the sample that I did use still have relatively few observations from both Luxembourg and Italy. Therefore, the conclusions I have drawn in this thesis might be accredited more to the countries with more observations.
Conclusions and discussion
Following the research of Ali & Zhang (2008) and by using real earnings management estimations of Roychowdhurry (2006) and Cohen & Zarowin (2010) I have tried answering the question if European based firms engage in earnings management when their current credit rating contains a plus or a minus to attain (prevent) an upgrade (downgrade) to another broad credit rating?
First of all, I tested it firms near a broad credit rating change engage in more accrual earnings management. Based on the results I presented in section 5 I can accept my
hypothesis. Based on the theory regarding credit ratings the results make sense. There are several reasons why being upgraded (downgraded) has consequences for a firm. First of all, the threshold between investment-grade bonds and speculative bonds is important to firms. When a firm gets downgraded to a speculative bond, they lose a lot of potential investors. Similarly, they gain a lot of potential investors when they get upgrade to an investment-grade bond. Besides this, a change in broad credit rating has serious implication for the cost of capital. A lower rating indicates a higher risk for investors. As credit ratings are a big component in taking away information asymmetry a lower rating will lead to higher interest payments for the issuer.
Secondly, I tested if firms near a broad credit rating change engage in real earnings management. After the SOX- act firms have in general been moving to this form of earnings management. This is because, for an outsider, it is harder to detect. A bondholder might think that a firm is overproducing to meet demands for sales in the next year or that cutting
31
discretionary costs is a smart decision to meet benchmarks. Real earnings management can be viewed positively by some bondholders because of this (Ge & Kim, 2014). I expected that firms near a broad credit rating change show higher forms of sales manipulation,
overproduction and cutting of discretionary expenses. Based on the results I can accept this hypothesis.
Lastly, I tested if there is a difference between code and common law countries. I hypothesized that code law countries on average engage more in earnings management. Based on the results I can’t accept this hypothesis. In France, accruals management is more
significant, while in the UK real earnings management is more significant. Common law countries are influenced more by the private sector (Ball, 2000). There is more investor protection and there are more regulations. Based on this, the usage of less detectable earnings management techniques is understandable. In France, a code law country, discretionary accruals are more used. This could be because of a less developed capital market in which there are not a lot of private investors. According to Ge & Kim (2014) real earnings
management is mostly used to manipulate private bondholders which use bond prices as an estimate for future performance. In a market with more institutionalized investors engaging in accrual earnings management might be more advisable.
