Meeting or beating analyst forecast trough loan loss provisioning
at European Banks
Name: Maarten Joost Student number: 10644962
Thesis supervisor: Dr. S.W. Bissessur Date: June 25, 2018
Word count: 12.327
MSc Accountancy & Control, specialization Accountancy Faculty of Economics and Business, University of Amsterdam
Statement of Originality
This document is written by Maarten Joost, 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.
Abstract
This thesis examines the relation between earnings management and the loan loss provisions (LLP) for European banks. More specifically, I examine if the loan loss provision is used for earnings management through meet or beating expectations. The results show first that banks do not use the loan loss provision for meet or beat purposes. Second, when only using the discretionary part of the loan loss provision for earnings management purposes, no significant result is found. This paper makes the following contributions to the current literature. First, the obtained data is from the European Union, whereas current literature mostly investigated the United States and Asia. Second, while other papers use income smoothing as earnings management measure, this paper focuses on meeting or beating a target. And third, this paper contributes to the current literature by investigating the use of the loan loss provision after the financial crisis.
Contents
1 Introduction ... 6
2 Literature review ... 8
2.1 Introduction ... 8
2.1.1 Agency Theory ... 8
2.1.2 Conservatism in financial reporting ... 8
2.2 Earnings management ... 9
2.3 Earnings management at banks ... 10
2.4 Loan Loss Provision ... 12
2.4.1 IAS 39 ... 12
2.4.2 IFRS 9 ... 13
2.5 Meet or Beat ... 14
2.6 Earning management and Loan Loss Provision ... 15
3 Hypothesis Development ... 17
3.1 Hypothesis 1 ... 17
3.2 Hypothesis 2 ... 17
4 Methodology. ... 18
4.1 Total loan loss provision in relation to MBE ... 18
4.2 Discretionary Loan Loss provision in relation to MBE ... 18
5 Data ... 20
5.1 Data Collection ... 20
5.2 Descriptive data ... 20
5.3 Spearman correlation ... 21
6 Results ... 23
6.1 Model 1: Total loan loss provision ... 23
6.2 Model 2: discretionary part of the loan loss provision ... 26
6.3.1 Model 4: Income smoothing ... 28
6.3.2 Model 5: Capital requirements model. ... 29
6.4 Results summary ... 31
7 Paper Summary ... 32
8 Conclusion ... 33
Appendix ... 35
1 Introduction
Banks have played a major role in the credit crisis of 2007 and credit risk has been identified as a starter of this crisis (Gebhardt, 2016). Gebhardt (2016) concluded that the current standard International Accounting Standard 39 did not lead to sufficient timeliness of loss recognition. The losses were incurred too late; therefore banks recognized their losses to late and had too little reserves to compensate for this loss (Hashim, Li, & O’Hanlon, 2016). Hashim et al. (2016) argues that the crisis has been exacerbated because of the delay in recognition of losses. For this reason the IASB and FASB came together in April 2009, to discuss improvements in the standards. (Curcio & Hasan, 2015) But as Hashim explains, the forming of this new standard IFRS 9 turns out to be very difficult. The forming of new rules is an arduous task. The effective date of IFRS 9 is January 2018, so it is arguable that it has taken a very long time. IAS 39 has its limitations, such as the incurred loss model, this model recognizes losses too late. That’s why the FASB and IASB were thinking about a new standard. They were trying to encourage banks to recognize losses earlier, so that problems that caused the crisis can be prevented. That is the reason they introduced the expected loss model instead of the incurred model in IFRS 9. With the change from IAS 39 to IFRS 9 the whole concept of loan loss provision has changed. Firms have to adopt a provision for every loan they issue (PWC, 2017). This provision is based on a lot of different factors that rely on the discretion of managers. The incurred loss model (IAS 39) claims to reduce manager’s discretion and decrease earnings management (Gebhardt, 2016).
Critics claim that the opportunity for earnings management increases under IFRS 9 (Gebhardt, 2016). However, it is not clear that the change from IAS 39 to IFRS 9 increases earnings management. Camfferman argues that there is an earnings management opportunity under IFRS 9 as well as under IAS 39. In both cases it is shown that earnings management is
possible (Camfferman, 2015).
The loan loss provision is built out of two components: an obligated part and a discretionary part. For earnings management an accrual is needed (Dechow & Skinner, 2000), in this case the discretionary part of the loan loss provision. When the criteria for earnings management are met, the following question arises, do managers use this discretion in the loan loss provision to meet or beat their target? Given that it is not clear whether the change to IFRS 9 will help reduce earnings management, this paper will focus on IAS 39 and see if there is a way for managers to use the discretionary part of the loan loss provision for meeting or beating their target. This will provide a baseline for earnings management using
loan loss provisions, against the new IFRS 9 standard that can be evaluated. The research question is as follows.
RQ: Do managers use the loan loss provision as an accrual to meet or beat their targets in the European Union under IAS 39?
This paper makes a contribution to the current literature in multiple ways. Firstly, because the data used will be from the European Union and current literature only investigates the United States and Asia. Secondly, this research focuses on meeting or beating a target, which hasn’t been done before in the perspective of the loan loss provision. And thirdly, this paper contributes to the current literature by investigating the use of the loan loss
provision after the financial crisis.
Results of this paper show that the loan loss provision doesn’t contribute to meeting or beating an analysts forecast. The total loan loss provision and the discretionary part of the loan loss provision both are not significant in the regressions. Possible explanations are the differences in accounting regimes, meaning the US-GAAP in comparison with IFRS. Prior literature examined mostly US-GAAP firms. Their results may not extend to IFRS. The second explanation is the changes made in the banking regulation after the financial crisis. This potentially could have reduced earnings management using the loan loss provision. The remainder of the paper is as follows. The second chapter consists of literature review where the background of the loan loss provision will be sketched, earnings management will be reviewed and the current literature will be discussed. After the literature review the hypothesis and methodology of this paper will be explained. Followed by the collection of data and the descriptive of the data. The results will be discussed after the data and at last a summary and conclusion will be given.
2 Literature review
2.1 Introduction
In this section the overall theory for this paper will be discussed. Problems in the accounting and economy are mostly problems about the principal and the agent. In this paper agency theory and conservatism plays an important role.
2.1.1 Agency Theory
The main theory that underlines financial accounting research in general, and earnings management research in particular is the agency theory. The conflict is about the manager (agent) and the shareholder (principal). For instance, a potential source for the conflict is a possible bonus or benchmark that has been set. The manager tries to comply with the benchmark and will be rewarded for this, the idea of sharing risk (Eisenhardt, 1989). The interest of the manager is thus meeting or beating the target. Sometimes this means for the manager to use accounting discretion that is detrimental to the principal. In this paper the agency problem is about meeting or beating the target and the resource to achieve this is the
loan loss provision.
Demsetz et al. (1997) argues that one of the agency problems at banks relates to the deposit insurance. This deposit assurance leads to excessive risk taking on the part of bank owners. This problem can be mitigated in two ways. First the managers must be risk averse. This has the consequence that managers do not take the excessive risk that stems from moral hazard. Secondly the franchise value can mitigate the agency problem. Demsetz et al. (1997) found in their paper that both these measures decrease the agency problem. More importantly, they found that in smaller banks excessive risk taking plays a bigger role than in
large banks.
Besides Demsetz et al. (1997), Macey and O’Hare (2005) argue that the agency problem is more important at banks. Banks have a social responsibility that should be more important than that of normal firms. Macey and O’Hara argue that the capital requirements are an agency problem, because banks try to achieve these capital requirements by all means, when this is not the desired way.
2.1.2 Conservatism in financial reporting
Besides the agency theory, conservatism plays a significant role in this paper. Watts and Basu give their definition of conservatism: Conservatism is defined as the differential verifiability
required for recognition of profits versus losses. Its extreme form is the traditional conservatism adage: "anticipate no profit, but anticipate all losses." (Watts, 2003), (Basu, 1997)
This paper is about the loan loss provision, which is the result of applying conservatism in financial reporting. This provision is formed because banks are not sure if they will receive the interest payments as expected. As Smiths (n.d.) points out that, conservatism is used by standard setter to require higher verification standards for profit recognition in comparison to the recognition of losses (Smith, n.d.). The loan loss provision is required for banks to reflect uncertainty with regards to losses on future (re)payments of loans, as obligated by the standard setters. Following Smith (n.d.), the loan loss provision can thus be seen as a form of conservatism.
2.2 Earnings management
In the following section the definition of earnings management will be given and the way to
measure earnings management.
According to Healy and Wahlen earnings management is: “Earnings management occurs when managers use judgment in financial reporting and in structuring transactions to alter financial reports to either mislead some stakeholders about the underlying economic performance of the company or to influence contractual outcomes that depend on reported accounting numbers.” (Healy & Wahlen, 1999) Shipper has her own view on earnings management. She claims that earnings management is a form of disclosure management. This means that earnings management is a purposeful intervention in the external accounting process.
The intent of earnings management is gaining a private gain (Schipper, 1989). She argues that there are two main forms of earnings management. The accrual based earnings management and the real earnings management. She describes accrual based earnings as the steering of accounting numbers in the financial reports to the advantage of the company. Real earnings management is for instance timing your investment decision or financing decision to smooth your earnings. Schipper explains that real earnings management is difficult to detect, because companies normally do not disclose information about timing decisions. This makes it hard to judge what the justified timing is. In firms it is difficult to decide what the appropriate amount of research development costs should be. Firms prefer real earnings management to accrual based management (Schipper, 1989).
There are multiple ways of detecting earnings management. One approach is through the use of aggregate accrual models. Dechow, Sloan and Sweeney (1995) describe five aggregate accruals models. In this paper, aggregate models are not employed. Therefore, for a discussion of these models, the reader is referred to Dechow, Sloan and Sweeney (1995). Besides the aggregate accrual methods, there is also the method of discovering earnings through the distribution of earnings after earnings management. The studies of Burgstahler and Dichev and DeGeorge et al. argue for this approach of detecting earnings management. They claim that you must focus on the distribution of earnings after earnings management. In other words, focus on the normal distribution of the earnings. After earnings management there should be a difference in the distribution. Before earnings management the distribution will show a normal distribution. After earnings management the distribution will have an abnormal distribution, a skew to the right (a more positive distribution).
Companies have an incentive to perform above the preconceived targets, because of this incentive earnings just below 0 will not be normally distributed, but will show a sudden increase after 0. The distribution shifts to a more positive distribution. Besides incentives to perform above targets, companies tend to avoid reporting losses. Burgstahler and Dichev and DeGeorge show in their paper that this incentive gives the normal distribution a shift to the right. By comparing the normal distribution before and after earnings, it should be possible to detect earnings. (Burgstahler & Eames, 2006) (Degeorge, Patel, & Zeckhauser, 1999) In this paper the earnings management detection method of Burgstahler and Dichev and DeGeorge et al. will be used. This paper will detect earnings management through the meet or beat of preconceived targets. Following the studies of Burgstahler and Dichev and DeGeorge et al. the method of meet or beat, measures the normal distribution of the earnings after earnings management. Meet or beat measures the difference in the normal distribution around the zero point.
2.3 Earnings management at banks
In the following section the motives of banks for earnings management will be given. According to Agarwal et al. there are multiple reasons why banks use earnings management (Agarwal, Chomsisengphet, Liu, & Ghon Rhee, 2007). They first argue that banks use earnings management to comply with capital requirements. The second use earnings management to smoothen income. Lastly, managers use earnings management for their own purpose in other words; managers use it to gain their own bonus or promotion.
Beginning with the use of earning management to comply with the capital requirements. Shrieves and Dahl argues that Japanese bank used accrual earnings management to comply with the regulatory capital (Shrieves & Dahl, 2003). In their paper they set out that there are multiple ways to comply with these capital requirements. The first possibility is to reduce lending or lower dividends to comply with the capital requirements. By lowering lending the required capital decreases (less reserves are obligated) and by lowering dividend increases the equity of the bank. The second possibility is using discretionary earnings management to comply with the capital requirements. Besides accounting standards a bank needs to comply with the Basel requirements. In this case Tier 1 and Tier 2 capital (Shrieves & Dahl, 2003). Banks can increase the Tier 1 and Tier 2 in multiple ways. First they can increase the loan loss provision as discretionary accrual to increase the Tier 1 and 2 capitals. The loan loss provision falls in three capital categories. All these capital categories influence Tier 1 capital through earnings. When the loan loss provision increases the Tier 2 capital also increases, this is because Tier 2 capital consists of reserves. As mentioned above the sale of securities increased the Tier 2 capital.
Knowing this, banks are capable of using their accruals to steer their, Tier 1 and 2 capitals to the correct ratio. Cummings and Durrani (2016) support this view. They found that banks in Australia use a capital surplus to create a buffer for future capital requirements. Banks in Australia tend to move capital surplus in a provision to foresee future credit losses. To continue with the paper of Ahmed et al. this paper highlights the possibility for banks to use accounting accruals to boost capital requirements. In their paper there is a significant result between the loan loss provision and capital requirements. Besides that they found that the loan loss provision is used for a boost in required capital in firms that are having difficulties to comply to the capital requirements (Ahmed, Takeda, & Thomas, 1999). Furthermore, Agarwal et al. argue that banks use earnings management to smoothen income (Agarwal e.a., 2007). According to the paper of Beidleman it is a possibility that firms use accruals to normalize earnings on the short run (Beidleman, 1973). According to Fonsenca and Gonzalez there is a possibility to smoothen earnings through accounting accruals, provided that there are some additional conditions. When there is not sufficient regulation about disclosure or insufficient supervision the chance of earnings smoothing
increases. (Fonseca & González, 2008)
Besides income smoothing there are some other factors that can influence earnings management at banks. According to Kim and Kross banking regulation also plays a role. Since the change of banking regulation in 1989, earnings management has decreased in the
banking sector that complies with these new regulations (Kim & Kross, 1998). An important factor for the occurrence of earnings management is the regulatory environment. Lastly, Gibbons argues that there are also some incentives for managers to use earnings management. He argues that management uses earnings management to gain an advantage for their own. When there are bonuses involved, managers are incentivized to reach a certain target. When earnings management is available, managers can use this for
their own benefit (Gibbons, 1998).
In this section we have seen the reason banks engage in the use of earnings management. There are three possibilities given. First there are the capital requirements, which can be achieved with the use of earnings management. Second earnings are managed to smooth earnings. And lastly earnings management is used by managers to gain their own bonus. In this paper I will focus on the use of earnings management for the smoothing of earnings.
2.4 Loan Loss Provision
In the following section, I will discuss what the loan loss provision is, its origins and the difference between IFRS 9 and IAS 39
2.4.1 IAS 39
Under IAS 39 the IASB tries to restrict the earnings management possible under IAS 25, the accounting principle concerning accounting for investments. IAS 39 replaced IAS 25 in 1998. So it introduced the incurred loss model (“IFRS - IAS 39 Financial Instruments: Recognition and Measurement - IAS 39 Financial Instruments: Recognition and Measurement”, z.d.). This model looks as follows:
EL = (Carrying Amount – Collateral) x PD x LIP (1)
Where EL is the expected loss, PD is the possibility of default and LIP denotes the loss identification period. Problems of this method are the late recognition of losses (Duh, Hsu, & Alves, 2012). The purpose of this model is to enable banks to use the loan loss provision when losses occur. Under IAS 28 there weren’t any restrictions to form a loan loss provision. The loan loss provision under IAS 39 is the expected loss. IAS 39 is based on the expected loss model and making a provision is not mandatory. In this case the loan write offs are almost similar to the loan loss provision.
2.4.2 IFRS 9
After the financial crisis the IASB wanted to prevent another financial crisis, because of credit risks. The main reason why the IASB founded this provision was to increase the reserves of banks. Banks face the risk that the loans outstanding could not be repaid (Hlawatsch & Ostrowski, 2010). The formula for the loan loss provision under IFRS 9 is as follows:
LLP = EAD · PD · LGD · LIP (2)
Where LLP denotes the loan loss provision, EAD denotes the exposure at default, LGD denotes Loss given default and LIP denotes the time adjustment of the default probability. PD denotes the probability of default (Hlawatsch & Ostrowski, 2010). In this formula the probable credit risk of the current loans is also displayed. The loan loss provision adjusts the loans outstanding with a risk of not being repaid. By forcing banks to use this provision it should prepare banks against sudden defaults of loans and prevents too late recognition of incurred losses. Because there are a lot of estimations in this model, there is a lot of discretion for managers. Because of this discretion it enables them to steer the loan
loss provision.
Differences between IFRS 9 & IAS 39 are according to Gebhardt that impairments will occur earlier in comparison with IAS 39. Besides this, he finds that with IFRS 9 a lot more management estimations are involved than with IAS 39 (Gebhardt, 2016). According to Gornjak there are multiple strengths and weaknesses to IFRS 9. He argues that the major strengths are: standards are more understandable for companies and for investors; accounting is aligned with business models and the addressing of the factors influencing the financial crisis.
Big weaknesses of IFRS 9 according to Gornjak are: more professional judgment and IFRS 9 does not solve the big problems of financial instruments like hedge accounting (Gornjak, 2017). Novotny argues that there is another downside of IFRS 9: it increases volatility of the regulatory capital of banks. This could result in a decrease of integrity of
financial reports (Novotny-Farkas, 2016).
Overall the conclusion is that the biggest change from IAS 39 to IFRS 9 is the change from the incurred loss model to the expected loss model. And with these changes there are some advantages and some disadvantages. Biggest advantage is the prevention of recognition of big losses and the biggest disadvantage is a bigger opportunity for earnings management.
2.5 Meet or Beat
Why do managers prefer a meet or beat? First of all managers prefer a narrow meet or a super small beat, because: Graham finds that firms build credibility on the capital market, secondly firms try to raise their stock price, thirdly they try to increase their external reputation and lastly firms try to convey future growth. Graham argues that with a failure to miss the target, it creates uncertainty for the firm. This is because analysts could have missed some negative points that the firm tried to hide or even the firm tried to hide deeper problems. Further he argues that missing the benchmark means explaining, why the firm missed the benchmark instead of looking to the future with this benchmark (Graham, Harvey, & Rajgopal, 2005) The paper of Veenman and Bissessur argues that managers prefer a narrow or small earnings beating because of two reasons. The first reason is that consequently beating expectations decreases the magnitude of the beating. They put it another way by explaining that the benefits of small earnings surprise outweigh the consequences of a big earnings surprise. Secondly they argue that the advantages of a positive surprise do not outweigh the disadvantages of an understatement. The revision carried out by analyst cost more (Bissessur
& Veenman, 2016)
Besides these papers Burgstahler and Eames investigated in their paper the relation between earnings management and positive pre-managed earnings and negative pre-managed earnings. They found that firms with negative pre-managed earnings increased their earnings by using earnings management, (Burgstahler & Eames, 2006) Following the reasons why firms prefer a meet or beat, I will now argue why to use a meet or beat as earnings management detection. According to Healy and Wahlen there are two advantages of the meet or beat method (Healy & Wahlen, 1999). First advantage is that this model can be used on a larger sample. By using this model it is not necessary to estimate the abnormal accruals. When this model is used, the user looks at the distribution of reported earnings for abnormal discontinuities. After gathering information about this abnormal distribution, the comparison is made with the earnings distribution of firms in a similar industry. The strength of this model is the identification of firms using earnings management, rather than the discretion over earnings. Secondly they argue that it is possible to estimate
pervasiveness of earnings management at these thresholds.
Managers prefer a meet or a small beat, because of the negative consequences that coherent with missing an analyst forecast. Besides Healy & Wahlen argue that it is a good measure for earnings mangement.
2.6 Earning management and Loan Loss Provision
In the current literature there is already a lot written about the loan loss provision in combination with earnings management, for example Dong, Liu and Hu (2012). Their research is about commercial banks in China and the relation between the discretionary loan loss provision and the earnings before tax. They use the model of Kanageretnam (2003) to split the loan loss provision in a discretionary part and a non discretionary part. They investigate 9 domestic banks and see if these banks use the discretionary part of the loan loss provision as accrual for earnings management and for capital management. Their research shows that there is a significant indication that these banks use the loan loss provision to steer their earnings and use it for capital management. They conclude with adding a suggestion about improving the system: more information disclosure from banks so it should be more
difficult to steer their earnings.
A second paper about the loan loss provision in combination with earnings management is Kanagaretman, Lobo and Mathieu ( 2003). Their research is about the relation between the performance of a bank and the possible relation between earnings smoothing through the loan loss provision. They find that when a bank endures good times, the bank saves money in the loan loss provision and when the bank endures less favorable times it borrows money from the loan loss provision, in a second paper of Kanageretman, he finds that this relation holds in an other sample. In this case a relation between income smoothing and signaling through the loan loss provision. The only remarkable finding is that the use of the loan loss provision for income smoothing is not the same in the sample (K.
Kanagaretnam, Lobo, & Yang, z.d.).
A third paper that examines the earnings management consequences of the loan loss provision is Ahmed, Takeda and Thomas (Ahmed, Takeda, & Thomas, 1999). They investigate whether the loan loss provision has influence on capital management and earnings management. They find that the loan loss provision has influence on the capital management, but they don’t have enough support for a relation between the loan loss provision and earnings management. They refer in their paper to Collins, who sketched that there is only a possibility for earnings management when there are non-discretionary earnings (Collins,
Shackelford, & Wahlen, 1995).
The following paper examines if US-banks have used the loan loss provision for smoothing future earnings (Ma, 1988). Ma found similar results as Kanagaretnam in his paper (2003). He founds that commercial banks use the loan loss provision for income
smoothing. In periods of high operating income, banks increase their loan loss provision. In times of lower operating income, banks write off provision from the loan loss provision to
higher their operating income.
The last paper I mention is the paper from Pérez et al, this paper examines if banks in Spain use their loan loss provision to smooth earnings. In their paper they investigate a large ample of Spanish banks. They try to find if banks use their loan loss provision to steer earnings and capital management. What they find is a significant result for the use of loan loss provision to steer the earnings, but a lack of evidence that they use it to steer their capital
management. (Pérez, Salas-Fumás, & Saurina, 2008)
Summarizing the current literature it could be said that there is enough evidence that there is a relation between loan loss provisioning and earnings management. In the current literature earnings management has only been investigated in the form of income smoothing
3 Hypothesis Development
In this section the hypothesis for this paper will be set out. Following the literature described above, there is a relation possible between the loan loss provision and meeting or beating a target. Agarwal et all argues that the loan loss provision is the biggest accrual at banks (Agarwal, Chomsisengphet, Liu, & Ghon Rhee, 2007). Banks can use this accrual to influence their earnings. As Hlawatsch and Ostrowski show that there is a big discretionary part in the formula of the loan loss provision (Hlawatsch & Ostrowski, 2010). Following Healy and Wahlen, there must be managerial influence to facilitate earnings management. In this case, because of the large amount of discretion managers have in the loan loss provision, earnings management is possible (Healy & Wahlen, 1999).
3.1 Hypothesis 1
Following (Ahmed e.a., 1999; Collins e.a., 1995; Dong e.a., 2012; K. Kanagaretnam e.a., 2003; Ma, 1988), they all find a positive relationship between the loan loss provision and earnings management. Ma (1988) found this relationship in the US, Dong (2012) found this for banks in China. All these papers look for the relationship between earnings smoothing and the loan loss provision. I expect that banks use the loan loss provision not only for earnings smoothing, but that there also will be a positive relationship between meeting or beating a target and the use of the loan loss provision.
H1: The loan loss provision has a positive effect on meeting or beating a target for banks in the EU
3.2 Hypothesis 2
Besides H1, I am interested to know if the discretionary part of the loan loss provision also has a significant effect on the loan loss provision as a whole. Multiple factors can influence the loan loss provision, so the following hypothesis will test if the discretionary part of the loan loss provision has any influence on meeting or beating a target.
H2: The discretionary part of the loan loss provision has a positive effect on meeting or beating a target in the EU
4 Methodology.
In the following section the methodology for this thesis will be outlined. In order to do a good regression analysis, the following two models are formed. By combining these models I hope to give a good insight if the loan loss provision has an influence on the meet or beat. This model will be tested by using a logarithm regression
4.1 Total loan loss provision in relation to MBE
To calculate the effect of the total loan loss provision on the meet or beat of the forecast, I will follow the model presented by Doyle, Jennings, and Soliman (2013). In this model they include a variable that controls the possibility for accounting accruals to have influence on the meet or beat (Doyle e.a., 2013). I will replace this variable with the total Loan Loss provision so it will show if it has any influence on the meet or beat of the target. The model looks as follows:
MBEt = α0+ α1 LLPt + α2 Book-to-Markett + α3 Sales growtht + α4 lnSizet + εt (1)
MBE is an indicator variable, which takes the value of 1 if earnings are between 0.00 and 0.01 of total assets, and zero otherwise.
LLP is the total loan loss provision
Book to market is book value of equity divided by market value equity Sales growth is quarterly change of revenue over the prior year
ln Size is the logarithm of the market value at the end of the quarter
ε is the error of the model
In the model of Doyle et al (2013) the model includes ROA and a dummy variable profitability. These variables are left out, because there is a perfect correlation between MBE and the ROA, dummy profitable.
4.2 Discretionary Loan Loss provision in relation to MBE
For my second hypothesis I want to see the effect of the discretionary part of the loan loss provision on meet or beat. First I will discuss the model for the discretionary Loan Loss provision as described in Kanagaretnam (2003). They situate that the loan loss provision exists of a non-discretionary part and discretionary Loan Loss provision. To make sure whether the discretionary part of the Loan Loss provision has influence on the meet or beat,
this discretionary part needs to be filtered out. I will use a similar equation as used in Beatty and Liao
LLPt = α0 + α1 ALWt−1 + α2 NPAt-1 + α3 COt + α4 LOANt + α5ΔLoant + εt (2) LLP is the provision for loan losses
ALW is allowance for bad loans divided by total loans NPA is nonperforming assets divided by lagged total loans, CO is net charge off divided by lagged total loans
LOAN is total loans divided by total assets
ΔLoan is change in total loans divided by lagged total loans ε denotes the discretionary loan loss provision
ALW differs from the LLP, because the ALW is the accumulated LLP displayed on the balance. ALW could be seen as a reserve for bad loans. In contradiction with the loan loss
provision, which is classified as a provision.
With this model (Beatty & Liao, 2014) I will try to estimate the discretionary loan loss provision, to use it in the following model of Doyle, Jennings, and Soliman (2013). The model is the same as for the total loan loss provision, but this variable is replaced by the discretionary loan loss provision and the non-discretionary loan loss provision.
MBEt = α0+ α1 DLLPt + α2 Non-DLLP α3 Book-to-Markett + α4 Sales growtht (3) +α5 lnSizet + εt
MBE is an indicator variable, which takes the value of 1 if earnings are between 0,00 and 0,01 of total assets, and zero otherwise.
DLLP is discretionary loan loss provision see model 2 Non-DLLP is non-discretionary part of the loan loss provision Book to market is book value of equity divided by market value equity Sales growth is quarterly change of revenue over the prior year
ln Size is the logarithm of the market value at the end of the quarter
5 Data
The following section will discuss the data collection, data processing, the data descriptive and the Spearman correlation of the data.
5.1 Data Collection
All the data collected in this paper is retrieved from the Wharton research data service. The database selected is Compustat – Capital IQ from Standard & Poor’s. Variables collected for the two models are the following: at used for the total assets of the company’s. Ceq is the common equity. Dptc is used for the total deposits outstanding. Ebit is the earnings before interest and taxes. Lntal is used for total loans outstanding. Lt is used for total liabilities. Nco represents the net charge off. Ni stands for the net income. Npat is used for the total non-performing assets of the bank. Pll is used for the provision of the loan losses. Tcor is used for the total operating income and prirow is used to identify banks outside the US. The data for the market value of equity are obtained from the Compustat – IQ global daily database. End price of stocks and shares outstanding are retrieved from the security daily option. All this data is processed in STATA. For the variables the following formulas were used. ROA is computed by dividing at by ni. ALQ is computed by dividing dptc by pll. CO is computed by dividing dptc by nco. The NPA is calculated by dividing dptc by npat. LOAN is calculated by dividing at by dptc. Marketvalue is computed by multiplying the stockprice by the shares outstanding. lnmarkt is the natural logarithm of the marketvalue. The btmr is the marketvalue divided by the ceq. And MBE is a dummy variable that is 1 if the ni is between the 0,00 and 0,01. To create the llpd a regression is run to generate the residuals, which are in this case the discretionary part of the loan loss provision. The data is filtered on EU-banks. The filter comes from the variable prirow, which indicates that banks are from outside the US. The banks that are indicated by the prirow are filtered on the EU.
5.2 Descriptive data
My observations are based on 302 bank-years observations. The sample consists of 16 European banks, within the period of 2011 till 2017. Due to the missing data for net-charge off and non-performing assets for a lot of banks, the number of observations decreased
significantly from 302 to 88, as shown in table 1.
In the table below all the descriptive statistics of the variables are shown. None of the outliers of the loan loss provision are deleted, these outliers are all valuable information for
the research. This in comparison with Kanagaretnam et al. (K. (Giri) Kanagaretnam, Lobo, & Mathieu, 2001), where the loan loss provision was selected on value.
5.3 Spearman correlation
In this section the Spearman correlations will be shown. First the Spearman correlation will be given for the model with the total loan loss provision, followed by the model with the non-discretionary part and the non-discretionary part of the loan loss provision.
Table 2 Correlation matrix total loan loss provision MBE 2 3 4 2_Δrevenue 0,111 0,054 3_ln marktvalue 0,250 -0,010 0,000 0,863 4_LLP -0,004 -0,065 0,282 0,945 0,261 0,000 5_ market to book ratio -0,200 0,027 -0,544 -0,104 0,001 0,646 0,000 0,080 Notes: This table provides the Spearman correlation coefficients for the main variables in the LLP-model. P-values are displayed in italics. Significant correlations are indicated in bold. (ρ < 0,10, two-tailed test)
In table 2 the correlations from the first model are shown. These correlations are done to show the correlations on variable level. There is a significant correlation between MBE
Table 1 Descriptive statistics
Variable N Mean Std. Dev. Min Max
LLP 302 5.108.422 4.703.319 -2.106.146 24.458,71 MBE 302 0,781 0,414 0,000 1,000 DLLP 88 3.339.242 3.045.767 6.848.123 13.112,470 Non-DLLP 88 9.055.363 2.695.065 5.467.344 1.589.753 mrktvalue equity (millions) 283 249.000 894.000 1,840 948.000.000 lnmarktvalue 283 2.386.472 2.674.005 1.442.522 2.987.999 Market to book ratio 283 0,084 0,396 0,002 3.961.226 Δ revenue 301 52057.57 36186.81 -6.749.631 132.504 Notes: This table presents descriptive statistics for the meet or beat of the analyst forecast. LLP = the total Loan loss provision. MBE = a dummy variable which is 1 when the bank has a ROA between 0,00 and 0,01 and 0 when the value falls outside this region. DLLP = the residual of the formula to calculate the LLP. Non-DLLP is the non-discretionary part of the loan loss provision. mrktvalue equity = denotes the shares outstanding multiplied by the closing price of the stock (yearly basis). lnmarktvalue is the natural logarithm of the market value of equity. Market to book ratio = the market value of equity divided by the book value of equity. The calculation of the DLLP is shown in appendix 1
and the variables Δrevenue, ln marktvalue and market to book ratio. Further there is a significant correlation between the LLP and ln marktvalue. And there is a significant correlation between the market to book ratio and ln marktvalue plus the LLP. In table 3 the correlation matrix of the discretionary loan loss model is given. This table describes the variable correlations.
Table 3 Correlation matrix discretionary loan loss provision MBE 2 3 4 5 2_Δrevenue 0,111 0,054 3_ln marktvalue 0,250 -0,010 0,000 0,863 4_DLLP -0,077 -0,285 0,349 0,478 0,008 0,001 5_non-DLLP -0,272 -0,180 -0,038 0,295 0,010 0,095 0,734 0,005 6_market to book ratio -0,200 0,027 -0,544 -0,352 -0,210 0,001 0,646 0.000 0,001 0,058 Notes: This table provides the Spearman correlation coefficients for the main variables in the discretionary loan loss provision model. P-values are displayed in italics. Significant correlations are indicated in bold. (ρ < 0,10, two-tailed test)
There is a significant correlation between MBE and the following variables ln marktvalue, nonDLLP, market to book ratio. There is a significant correlation between the Δrevenue and the variables DLLP and non-DLLP. This significant correlation is as described in the current literature (Ahmed e.a., 1999; Collins e.a., 1995; K. (Giri) Kanagaretnam e.a., 2001). These papers found a positive correlation between the loan loss provision and the revenues. Furthermore it is shown in table 3 that there is a significant correlation between DLLP and the variables ln marktvalue, non-DLLP and market to book ratio. The non-DLLP has a significant correlation with the market to book ratio.
6 Results
In this section the results of this paper will be discussed. First the results for the first hypothesis will be discussed and will be followed by the results for the second hypothesis. Two additional tests will be performed.
6.1 Model 1: Total loan loss provision
First I will test hypothesis 1, this hypothesis is as follows: The loan loss provision has a positive effect on meeting or beating a target for banks in the EU. This hypothesis will be tested according to the model of Kanagaretnam (2003). The results for this model can be found in table 4. This is the logarithm regression of the total loan loss provision on MBE.
Table 4
Meet or beat between 2011-2017: effect of the total LLP
Dep. Var. = MBE (i) (ii) (iii)
Independent Var. Coef. z-stat Coef. z-stat Coef. z-stat Intercept β0 -3,258 -2,26 -3,258 -1,06 -2,269 -0,69 Δrevenue β1 0,107 0,17 0,107 0,13 -0,050 -0,07 Lnmarktvalue β2 0,194 3,38 0,194 1,56 0,137 1,02 Markettobookratio β3 -0,295 -1,06 -0,295 0,52 -0,628 -1,38 LLP (total) β4 0,000 -0,13 0,000 -0,06 0,000 0,71 Bank dummies Included Included Year dummies Included No. Observations 283 283 283 Pseudo R2 6,06% 6,06% 10,40% Notes: This table presents logarithm regression estimates to test the effect of the total loan loss provision on the MBE (model 1) The sample in this test includes 302 observations for 16 European banks spanning calendar years 2011-2017 with the available data from Compustat. All variables are defined in Table 1. Reported t-statistics are under (ii) and (iii) clustered for Company and clustered for year under (iii). Significant relations are indicated in bold. (ρ < 0,10)
Table 4 shows 3 logarithm regressions (i) (ii) and (iii). The first regression (i) is a robust logarithm regression of LLP on MBE. The second regression (ii) is a clustered bank logarithm regression of LLP on MBE. And the third regression (iii) is clustered bank and year logarithm regression of LLP on MBE. The additional regressions are done to correct the statistics errors of STATA. The clustered regression of bank dummies, corrects for counting the same bank multiple times in the regression. The same goes for the clustered year dummy, which corrects for regressing of a year multiple times.
In table 1 for regression (i) I see a very low value for LLP. This low z-value does not substantiate the expected results. The predicted relation would be a
significant negative relation, as the current literature has found a positive effect of earnings management on the loan loss provision (Dong e.a., 2012; K. Kanagaretnam e.a., 2003; Ma, 1988; Pérez e.a., 2008). But as regression (ii) show there is not a significant (z-value of -0,06) relationship between the LLP and MBE. Regression (iii) shows another unexpected result for me. The z-value of 0,71 is not in line with the hypothesis. First the relationship should be negative (positive relation for the LLP and MBE), in this case the relation is positive, that states that if the loan loss provision grows, the MBE grows too. The negative relation should hold, because the LLP is a cost. If the LLP grows the ROA should decrease and with it the MBE. Hypothesis 1 is rejected under the circumstances of model 1.
The first explanation could be the time period. As the current literature uses a time period before the crisis (Dong e.a., 2012; K. Kanagaretnam e.a., 2003; Ma, 1988; Pérez e.a., 2008). The time period used in this paper is after the financial crisis and it is possible to argue that due to economic consequences banks revenues were growing and the possibility to get rid of the bad loans. These loans were acquired before the crisis and turned out to be loans, which have a very low possibility of generating revenue. This phenomenon arises at so called zombie banks (Onaran, 2011). Identification of these banks is very difficult, but there could be a possibility that some of these banks used for my regression were zombie banks. In this case it could explain the high write offs, because there was a possibility of writing them off. The economic state offered the possibility to write
off loans and still make a small meet or beat.
Concluding, a possible explanation for the positive (negative for my hypothesis) relation between MBE and LLP could be compensating for the bad loans acquired before the financial crisis. The loan loss provision increases because of the possibility to still meet or beat the analysts forecast.
A second explanation could be the change in banking regulation. The data from current literature is from outside the EU and before the crisis. In the EU the Basel requirements were introduced in 2006 (Catarineu-Rabell, Jackson, & Tsomocos, 2005). These requirements require European banks to hold certain capital requirement. The banks require a certain level of solvency. To reach these requirements banks need to write off bad loans. Writing of loans means under IAS 39, loan loss provisioning. As El Sood has researched, there is a difference in pre
financial crisis loan loss provisioning regime and an after crisis loan loss provisioning regime (El Sood, 2012). After the financial crisis of 2007, banking regulation became more important and especially the rules for loan loss provisioning. He finds in his paper that the banks used the loan loss provision more intensively before the crisis, than after the crisis. Secondly, he mentioned that banks decreased their non-performing assets. To decrease these bad loans the loans are written off, resulting in a loan loss provision. This increased the loan loss provisioning, but the economic growth kept going. So the revenues and ROA were growing, just as the loan loss provision to Basel requirements.
Basel requirements forced banks to write off bad loans resulted in an increased loan loss provision. This in combination with the growing economy could explain the positive relation in regression (iii).
Another possible explanation for the positive relation found in regression (iii) is the difference in accounting regime. Current literature used data from the US or China, these countries use US-GAAP or China-GAAP. In the EU IFRS is mandatory. As Leventis et al. have shown in their research. (Leventis, Dimitropoulos, & Anandarajan, 2011). They found that after the mandatory implementation of IFRS in the EU, European banks used less earnings management through the loan loss provision. This could be an explanation for my findings.
Concluding from this I assume that a possibility for the difference in outcome between the current literature and my results is the difference in accounting regime. Whereas the data from the current literature is obtained from
GAAP based accounting principles.
Cucio and Hasan supports this argument (Curcio & Hasan, 2015). In their paper they try to investigate whether there is a difference between European banks and non-European banks and the use of the discretionary part of the loan loss provision. Their findings are that European banks use the discretionary loan loss provision much less than non-European banks. The difference could be that European banks are more concerned about their capital requirements (Basel) than non-European banks. Non-European banks are more concerned about their earnings. This could also explain the difference with the paper of Pérez et al., which argues that Spanish banks use the loan loss provision to smoothen earnings. This paper has data used from 1985 till 2002 and IFRS became obligatory in 2005
(Pérez e.a., 2008).
6.2 Model 2: discretionary part of the loan loss provision
In the following section the results for model 3 will be discussed. This model includes the discretionary part of the loan loss provision and the non-discretionary part of the loan loss provision. This model seeks to find differences between the total loan loss provision and the discretionary part of the loan loss provision and the effects on the meet or beat.
Table 5 shows 3 logarithm regressions (i) (ii) and (iii). The first regression (i) is a robust logarithm regression of discretionary part on meet or beat. The second regression (ii) is a clustered bank logarithm regression of discretionary part on meat or beat. And the third regression (iii) is a clustered bank and year logarithm regression of discretionary part on meet or beat. The additional regressions are done for the same reason as for model 1 to correct the statistic errors of Stata. The clustered regression of bank dummies corrects for counting the same bank multiple times in the regression. The same applies for the clustered year dummy, which corrects for regressing of a year multiple times. The regression output for the discretionary part of the loan loss provision can be found in appendix 1. The regression is based on model 3 discussed in section 4.2. Starting with regression (i) the discretionary part of the loan loss provision
Table 5
Meet or beat between 2011-2017: effect of the discretionary LLP
Dep. Var. = MBE (I) (ii) (iii)
Independent Var. Coef. z-stat Coef. z-stat Coef. z-stat
Intercept β0 -24,290 -2,570 -24,290 -2,810 -48,746 -3,090 Δrevenue β1 -3,370 -2,240 -3,370 -2,480 -19,937 -5,620 Lnmarktvalue β2 1,267 3,320 1,267 2,960 2,970 4,830 Markettobookratio β3 131,342 2,220 131,342 5,280 320,852 5,670 D LP β4 0,000 -1,270 0,000 -0, 70 0,000 1,060 non-DLLP β5 -0,001 -0,840 -0,001 -0,480 -0,007 -3,540
Bank dummies Included Included
Year dummies Included
No. Observations 82 82 59
Pseudo R2 40,50% 40,50% 72,25%
Notes:
This table presents logarithm regression estimates to test the effect of the discretionary loan loss provision on the MBE (model 2) The sample in this test includes 82 observations for 16 European banks spanning calendar years 2011-2017 with the available data from Compustat. All variables are defined in Table 1. Reported t-statistics are under (ii) and (iii) clustered for Company and clustered for year under (iii). Significant relations are indicated in bold. (ρ < 0,10)
isn’t significant in explaining the meet or beat. The same applies for the non-discretionary part of the loan loss provision. After correction for the bank dummy the DLLP still remains insignificant in regression (ii). Also the non-discretionary part of the loan loss provision remains insignificant in regression (ii). In regression (iii) that is corrected for the dummy year, the discretionary part still remains insignificant. In contradiction to the discretionary part, the non- discretionary part shows a significant effect on the meet or beat in regression (iii). The discretionary part of the loan loss provision doesn’t have an effect on the meet or beat, but the non-discretionary part of the loan loss provision does have an effect on the meet or beat.
After an additional test (the Wald test) to see whether the non- discretionary part is similar to the discretionary part a significant result is shown (Prob > chi2 = 0.0009). This means that the non-discretionary part differs from the discretionary part and represents a different value than the discretionary part. Concluding from the results hypothesis 2 is rejected, because the discretionary part of the loan loss provision does not have an effect on meet or beat. But in contrast to my expectations the results of the non- discretionary part
do have an effect on the meet or beat.
Both results are not in line with the current literature. The current literature argues that there is a relation between earnings management and the loan loss provision (Anandarajan, Hasan, & Lozano-Vivas, 2003; Dong e.a., 2012; K. (Giri) Kanagaretnam e.a., 2001; K. Kanagaretnam e.a., 2003; Ma, 1988). Meaning the discretionary part of the loan loss provision should have an effect on meet or beat. This data shows that European banks do not use the loan loss provision for earnings management.
If it was the case that banks used the loan loss provision for earnings management, there should be a significant effect of the discretionary part of the loan loss provision (resource of earnings management) on the MBE (measure of earnings management)
6.3 Additional tests
To support my assumption that the loan loss provision is not used for earnings management by European banks, I will use my data to perform two additional
tests. One test that will tests the earnings management through income smoothing and one test that will test for earnings management through capital requirements. The purpose of these tests is to compare the results of current literature with my results. If the results of these additional tests are different from those in the current literature it support my view about the use of earnings management in the EU. 6.3.1 Model 4: Income smoothing
The first additional test is a replication of the model of Greenawalt and Sinkey. They designed a model to detect income smoothing (Greenawalt & Sinkey, 1988). The model they used is as follows:
LLP = β0 + β1 EBITt + β2 ΔNon-performing Loans + β3 Loan lossest + (4) β4 ΔLoanst + εt
Where LLP is the total loan loss provision, already calculated in the previous model, EBIT is the earnings before interest and taxes, represented by the variable tcor as used in the previous models. Δ non-performing loans is the difference in non-performing loans compared to last year, npat is used from the previous models and is used to calculate Δ non-performing loans. Loan loss is the loans that are written off. Net charge off is used from the previous models. ΔLoans is the difference in loans outstanding compared with the loan outstanding of previous year, Δloans is used from model 2. This model tries to find the relation between
earnings and the loan loss provision.
According to Greenawalt and Sinkey there should be a negative relation between earnings and the loan loss provision in order to state that these banks don’t use earnings management. This negative relation must exist, because high levels of income should indicate prosperity and reduces the likelihood of loan defaulting. In this dataset a correlation exists between the loan loss provision and earnings, however this correlation is not significant (see table 2). So following the paper of Greenawalt and Sinkey (1988) I would expect a significant negative relation between earnings and the loan loss provision indicating that there is no earnings management
Table 6
Regression of Δrevenue on the total LLP
Dep. Var. = LLP (i)
Independent Var. Coef. z-stat
ntercept β0 -7.127.569 -0,37 Δrevenue β1 -8.860.145*** -9,97 Δnon - performing loans β2 0,135*** 6,48 Total loans β3 1.970.785 0,61 Net Charge off β4 -83.984,74 * -1,70 Year dummies Included Bank dummies Included No. Observations 8 Pseudo R2 46,42% Note :
This table presents an regression estimates to test the effect of the Δrevenue on the total loan loss provision The sample in this test includes 83 observations for 16 European banks spanning calendar years 2011-2017 with the data collected from Compustat. Reported t-statistics are clustered for Company and clustered for year. ***,**,* indicates a significance at the 0.01, 0.05, 0.10 level, respectively for a two tailed test.
The regression of model 4 can be found in table 6. There is a significant result between the Δrevenue and the total loan loss provision. This result is as expected compared with the results of model 1 + 3. The negative relation between Δrevenue and total loan loss provision means that earnings management has not
taken place.
As Greenawalt and Sinkey (1988) point out there must be a positive relation between earnings and the total loan loss provision to state that there is any form of earnings management through the loan loss provision. The aim of this extra regression is to test if other earnings management measures indicate the same result as the regression from model 1 + 3. In this case it does. The income smoothing model gives the same results as the previous described models, this data set shows that there is no earnings management through the loan loss provision. The insignificant relation between the loan loss provision and meet or beat can also be found between the loan loss provision and income of the banks.
6.3.2 Model 5: Capital requirements model.
At last an extra regression is done to support the view that earnings management through the loan loss provision is not done in this data set. Current literature tested this relation before and found that the loan loss provision is used in certain banks to raise capital management just as Ma and Anandarajan et al. found (Anandarajan
e.a., 2003; Ma, 1988; Moyer, 1990). These papers are performed outside the EU and this data set could give another result. Expected from the current literature is that there is a relation between the loan loss provision and capital requirements. My expectations are that the loan loss provision is not used for steering the required capital. Consistent with the rest of the paper, it would be an unexpected result if this model points out that there is a relation between capital requirements and loan loss provision.
This model is based on the model of Anandarajan et al. ( 2003). The model looks as follows:
LLPt = β0 + β1 MCAPt + β2 EBTt + β3 TAt + εt (5) Where LLP denotes the total loan loss provision as used before. MCAP is
the capital requirement ratio of the bank, obtained from Compustat. EBT is the earnings before taxes, as calculated before. TA is the total assets as calculated before. In the paper of Anandarajan et al. the model also tests the change in banking regulation, this test is taken out of consideration.
Table 7
Effect of MCAP on the total LLP
Dep. Var. = LLP (i)
Independent Var. Coef. z-stat
Intercept β0 102,351 0,69 LOAN β1 6506, 69*** 3,70 Total assets β2 -0,002*** -2,45 MCAP β3 -262,085 0,02 Year dummies Included Bank dummies Included No. Observations 2 8 Pseudo R2 41,21% Notes: This table presents an regression estimates to test the effect of the MCAP on the total loan loss provision The sample in this test includes 288 observations for 16 European banks spanning calendar years 2011-2017 with the data collected from Compustat. Reported t-statistics are clustered for Company and clustered for year. ***,**,* indicates a significance at the 0.01, 0.05, 0.10 level, respectively for a two tailed test.
Table 7 shows the results of model 5. The main variable is the MCAP. Anandarajan et al. (2003) states that a significant negative relation between MCAP and LLP, points to earnings management out. In this model a significant negative correlation between MCAP and LLP is found. This points out that there is earnings management used to steer the capital requirements. This result is not in
line with the rest of the paper. The expected result would be that there is not a relationship between capital requirements and the loan loss provision. The results show a match with the current literature (Anandarajan e.a., 2003; Moyer, 1990), which also found a relation between the loan loss provision and earnings management.
This relation could be explained by changing regulations after 2006. As mentioned above Basel requirements were introduced in 2006. The loan loss provision played a different role in the capital requirements. To achieve the requirements banks could have used the loan loss provision to steer the capital requirements. Indicating that the loan loss provision is not used for financial incentives, but for legal incentives in the EU.
6.4 Results summary
The results of the regression are clear. The total loan loss provision isn’t used to achieve a meet or beat. The discretionary part of the loan loss provision doesn’t have a significant effect on a meet or beat. This is in contradicting my own hypothesis and the expectations formed by the current literature. The non-discretionary part of the loan loss provision shows a significant effect on the MBE indicating that the economic situation has an influence on the meet or beat. To support my view that the loan loss provision is not used for earnings management purpose in the EU, two additional tests were performed. The income-smoothing model shows the same result as the MBE. The Loan loss provision is not used for earnings management purposes. The capital requirements model shows a significant result between the capital requirements and loan loss provision, indicating that the loan loss provision is not used for the financial incentives, but for legal incentives.
7 Paper Summary
I started this paper with the change from IAS 39 to IFRS 9 and it implications for earnings management. Critics of the change to IFRS 9 claim that there are more possibilities for earnings management under IFRS 9. This paper is about the earnings management possibilities under IAS 39. This is done by looking at the discretionary part of the loan loss provision. The detection method for earnings management is the one described by Burgstahler and Dichev (Burgstahler & Eames, 2006). They look at the distribution as a normal distribution. This method is called meeting or beating an analyst’s forecast. Managers don’t prefer a negative earnings surprise so they will try to prevent that from happening. In the normal distribution a slight skew to the right will be seen around the 0 point. In this paper 16 banks from the EU were used. Which resulted in 302 observations. The model used for this paper is the meet or beat model. The meet or beat is tested on the effects of the loan loss provision. The results of the first model are that there is no a relation between the total loan loss provision and meet or beat. The third model does not show a significant relation between the meet or beat and the discretionary part of the loan loss provision. The non-discretionary part of the loan loss provision does show a significant effect on the meet or beat. Both hypothesises are therefore rejected. Explanations for this result are the differences between the EU and the US. Current literature gathered data from the US and other US-GAAP countries. The first explanation could be the difference between
US-GAAP and IFRS.
Another explanation is the consequences of the financial crisis. Banks in the EU are obligated to adhere to Basel requirements, which cause other incentives for these banks. Beside this consequence, after he financial crisis revenues started to grow and the need to steer the revenues through the loan loss provision declined. Instead, the growing revenues offered an opportunity to get rid of the bad loans acquired before the financial crisis. To support the view that the loan loss provision is not used for earnings management additional test are preformed.
8 Conclusion
The aim of this paper is to know more about the effects of loan loss provision on meet or beat and seeks to find a relation between earnings management and the loan loss provision in the European Union. This paper is based on a quantitative study and makes use of the Burgstahler and Dichev model to detect earnings management (Burgstahler & Eames, 2006). I expected to find a positive relation between meeting or beating and the loan loss provision. Especially the discretionary part of the loan loss provision I thought would have a significant effect on the meet or beat. The discretionary part of the loan loss provision is the part where managers have their discretion and could use this discretion to manage
their earnings.
My findings suggest different. The loan loss provision does not have an effect on meeting or beating a target. I found an insignificant result for the total loan loss provision and also an insignificant result for the discretionary part of the loan loss provision. These findings are in contrast with the current literature, where a positive relation exists between earnings management and the loan loss provision. Possible explanations of these findings are that there are different accounting regimes between the current literature and the data in this paper. The current literature researches banks that are obligated to US-GAAP and this paper is about European firms, where IFRS is obligated. The difference between these regimes could explain the different results of the findings. Another explanation could be the time frame that is chosen. This paper is about the following period: 2011-2017 and the current literature describes periods from 1989 until 2005. Difference here is the pre-financial crisis period and the after financial crisis period. The financial crisis could have an impact through the growth in earnings that could compensate for earnings management. If we compare the impact of the financial crisis on the regulation for banks, there is a significant change. After the financial crisis European banks are obligated to follow the Basel requirements. These requirements have an effect on the use of the loan loss provision. This change in banking regulation could also be an
explanation for the difference in outcomes.
Besides the test if the loan loss provision has effect on meeting or beating a target, additional regression are performed. Current literature argues that the loan
