Single Supervisory Mechanism: How did the commercial banks
respond? An accounting perspective.
Name: Thijs Hermans Student number: 10373578
Thesis supervisor: Pouyan Ghazizadeh Date: June 24, 2018
Word count: 11.949
MSc Accountancy & Control, specialization Accountancy Faculty of Economics and Business, University of Amsterdam
Statement of Originality
This document is written by student Thijs Hermans 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
The financial crisis of 2007 led to a discussion on how this could have been prevented and what impact it had on the banking sector. It became clear that a banking union had to be created, which started with the introduction of the Single Supervisory Mechanism (SSM). A single (more superior) supervisor was appointed, which could influence how conservative banks account for their loan loss provisions due to awe for this new supervisor. Since supervisors prefer conservative accounting, according to Leventis, Dimitropoulos, and Owusu-Ansah (2013) and Turner (2010), I predict that banks that fall under the SSM are more conservative in their loan loss provisions accounting. However, I have found mixed results. The empirical results from this thesis indicate that banks are positively sensitive to a change in nonperforming loans in the current or next year while under the SSM in their accounting for loan loss provisions. This indicates more accounting conservatism. However, there is a contradicting sensitivity of banks when there are more net charge-offs recognised in the current year under the SSM in their loan loss provisioning, suggesting that banks account less conservative. My results might provide insights for supervisors, who prefer conservative accounting, on how a change in supervision might influence the accounting of banks falling under that supervision.
Keywords: Single Supervisory Mechanism, accounting conservatism, loan loss provisions, centralised supervision.
Contents
1 Introduction 4
2 The introduction of the SSM 7
2.1 Run-up to the introduction of the banking union and the SSM in the Eurozone 7
2.2 The banking union 8
2.3 Supervision with the SSM in the Eurozone 9
3 Accounting conservatism and supervision 11
3.1 Accounting conservatism in general 11
3.2 Accounting conservatism for banks 12
3.3 Effects of transition to central supervision 13
3.4 Effects of supervision on accounting conservatism 14
4 Sample selection and descriptive statistics 16
4.1 Sample selection 16
4.2 Descriptive statistics 17
5 Empirical tests and results 23
5.1 Testing and results 23
5.2 Implications for results 29
6 Conclusion 33
References 35
Appendices 39
1 Introduction
It took a long time before there was a consensus on the many effects that the financial crisis of 2007 had on the banking sector in Europe (Véron, 2012). Eventually, the importance of a creation of a banking union became clear for the European policymakers, to have a broader vision for crisis management and resolution. On their website, the ECB describes this banking union as an important step towards a genuine Economic and Monetary Union. The purpose of this banking union is to be more transparent by being more consistent with applying rules for supervision. This way, the rules are unified for all the banks of the Eurozone, which might result in a safer situation, since this enables the ECB to intervene more quickly and adequate when banks face problems. The union consists of five pillars, of whom two pillars are in place and fully operational at the moment of writing: the Single Resolution Mechanism (SRM) and the Single Supervisory Mechanism (SSM). The latter was introduced in November 2014, with the objectives to ensure safety in the European banking system, to increase financial stability and integration and to ensure consistent supervision (ECB, 2018a). With this Single Supervisory Mechanism, the ECB gained the authority to take on the supervision role, which was at a national level before. The introduction of the ECB as the supervisor resulted in the same rules and requirements for all the commercial banks in the Eurozone, which meant that some banks had to cope with changes in how they were supervised.
Rochet (2005) described two goals of the supervision on commercial banks. The first goal is to limit the costs and frequency of failing banks in case of bankruptcy. The second goal is to protect the financial stability in the banking sector by limiting the frequency and costs of those failures. Bushman and Williams (2012) argue that the way banks used to account for loan losses is in direct conflict with those goals. Banks took big risks with their loan loss accounting and therefore reinforce pro-cyclical effects of bank capital regulation. They mention that the incurred loss model should be changed to allow bankers to incorporate forward-looking judgments into loan loss provisions to be more conservative. Barth and Landsman (2010) agree with this and add that the now introduced under IFRS 9 expected loss model could have mitigated the effects of the delayed and asymmetric recognition of loan losses. This mitigation is done by recognising loan loss provisions to reflect all expected future changes, that is increases and decreases, in the future cash flows of its loans. It allows banks to be more conservative in their accounting, which is according to Leventis,
Dimitropoulos and Owusu-Ansah (2013) beneficial to financial statement users like supervisors by, amongst others, embedding managerial opportunism.
This thesis wants to analyse the effect the introduction of the SSM, and with that, the change to a central supervisor for most banks had on the loan loss accounting of banks in the Eurozone. Did the SSM affect the way how conservative banks approach their loan loss accounting, now that they are falling all under the same focused supervisory regime? Also, did this result in an improvement in the situation described by Bushman and Williams (2012)?
I predict that banks that fall under the SSM will increase their timeliness and size of loan loss provision recognition since banks that fall under the SSM will increase the awe for their now more superior supervisor, the ECB, which prefers higher accounting conservatism (Leventis et al., 2013). The analysed sample contains 183 different banks in the years 2012-2015, of which 78 different banks that fall under the SSM from a certain point, with a total of 510 analysed bank years. For my regression analysis, I use an adjusted version of the model of Nichols, Wahlen, and Wieland (2009), based on the analysis of the effect that changes in nonperforming loans have on loan loss provisions. They state that an increase in
nonperforming loans, which are borrowings that have not been repaid according to the repayment schedule for at least 90 days, should be an indicator for management to increase provisions. Contradicting my prediction, I have found mixed results in my tests: banks are positively sensitive to a change in nonperforming loans in the current or next year while under the SSM in their accounting for loan loss provisions. This indicates more accounting conservatism since Turner (2010) mentioned that central supervisors prefer conservatism through loan loss provisioning, and therefore confirms my prediction. However, I find that banks are negatively sensitive to current year net charge-offs under the SSM in their loan loss provisioning. I try to explain this by suggesting that banks under the SSM that realised more loan losses through its net charge-offs had to get rid of the relating nonperforming loans and therefore could write-off the loan loss provisions. Another explanation might be that the bank successfully expects to adjust their lending choices as a response to the realised losses of the current year and therefore could lower its provisions.
This study contributes to the literature in several ways. First, it provides an insight into the effects of the implementation of the SSM, several years after the introduction of it. Secondly, it contributes to the research on conservatism and loan loss accounting in the banking sector and factors influencing these. Lastly, the results of this study might interest
supervisors and regulators on the effect of central supervision on the accounting of banks, especially on the accounting conservatism which they seem to prefer.
In the next section, I set the stage by describing how the supervision before the SSM was in the Eurozone for banks, how the introduction of the SSM happened and what changes the SSM brought in the Eurozone. The third section provides a brief discussion of existing literature about conservatism in accounting and of how central supervision may affect this. This discussion results in the presentation of my hypothesis. In the fourth section, I will describe my sample selection and present the descriptive statistics. In the fifth section, I show my empirical research, the main findings and interpretations of them and in the last section, I provide some conclusions and insights for future research.
2 The introduction of the SSM
In this section, I present a description of the development of supervision in the Eurozone. The first section describes the period before the introduction of the banking union and the SSM and the arguments for why they were introduced. The next sections elaborate more deeply on the banking union and the SSM.
2.1 Run-up to the introduction of the banking union and the SSM in the Eurozone Chang (2015) and Beetsma and Eijffinger (2009) illustrated in their articles the development of the financial supervision role of the European Central Bank in the euro area. When the euro was introduced in 1999, monetary policymaking to ensure financial stability became a European level competence. However, financial supervision remained with national
authorities, since there was only a Banking Supervision Committee created at European level. This committee was extended the next years into so-called ‘Level 3 Committees’, comprising national supervisors and thus strengthening the cooperation. In 2007, the ECB and European Commission (EC) suggested proposals, improving the Level 3 Committees in three ways. First, they would receive a stronger legal basis. Secondly, their accountability increased by having them to define their objectives and report on an annual basis to the EC and the European Parliament. Lastly, they now could take decisions by qualified majority. Together, these improvements would lead to increased international collaboration on crisis prevention, crisis management, and crisis resolution.
It became clear that the regulatory competition between the national supervisors, which were and are still not always the national central banks (ECB, n.d.-a), attributed to the failure to detect and prevent the financial crisis of 2007. This competition led to a call from the EC, among others, for a single European supervisory authority (Beetsma and Eijffinger, 2009). In response, the European Banking Authority (EBA) as part of the European System of Financial Supervision (ESFS) was created in 2011, which was an improvement to the previous system by implementing the European Systemic Risk Board and the continued use of national supervisors. Chang (2015) however still asks why was chosen for a rather toothless system instead of creating a single supervisor at the time.
The call for the banking union and the SSM came later, resulting from the sovereign debt crisis that escalated in the spring of 2012 (Chang, 2015). Due to the problems in Greece and Spain, euro area leaders frequently met to discuss direct bank recapitalisation, banking supervision, structural reforms and increased solidarity. When Germany surprisingly agreed
with the direct recapitalisation, it did demand the creation of a single European supervisory mechanism for the banking sector. In June 2012, the ECB was eventually chosen by a meeting of the European Council as the Single Supervisory Mechanism. They were now in charge of bank supervision, to break the vicious circle between banks and sovereigns. 2.2 The banking union
Tröger (2014) mentions that political leaders argue that the establishment of a European banking union would lead to a welcome contribution because this would ensure impartial and uniform implementation of supervision for all the banks in the Eurozone. This banking union, which would provide common safety nets like a reliable deposit guarantee scheme and govern banks’ operations and risk-taking behaviour, could help in preventing or reducing costs of bank failures and other calamities.
According to Véron and Wolff (2013), the marking point of the start of this banking union was the previously mentioned meeting of the European Council in June 2012, in which the creation of the SSM was decided. To build the European banking union further, the European Council had defined an approach that includes four successive steps. The first step was integrated supervision, which included the adoption of the regulation and reforming the EBA to the new situation advent of the SSM. Also, the Capital Requirements Regulation (CRR) had to adapt so that the SSM could implement a harmonised supervisory rulebook based on the Basel III accord, instead of every independent national rulebook that was applicable before. The second and third step of building the European banking union revolved around creating a coordinated framework for bank resolution, leading to a Single Resolution Mechanism (SRM). The goal of these steps was, according to European Council Conclusion of December 2012, organising a harmonised mechanism to “safeguard financial stability and [...] protecting taxpayers in the context of banking crises.” The fourth step is according to Véron and Wolff (2013) a combination of steps like the integration of deposit guarantee schemes, which will be needed in the future but has no priority.
Two of those steps, or pillars, as the ECB calls them, are already fully operational (ECB, 2018b) and the deposit guarantee systems are still under discussion (Wymeersch, 2014). These deposit guarantee systems should alleviate the first consequences of a failing bank, by providing savers’ deposits as reimbursement of their savings at the failing bank. This would increase the confidence of savers in their bank since they would receive a uniform amount of their savings if the bank bankrupts.
Because this thesis focuses on the introduction of the SSM, the following section will provide a zoomed-in scope on this mechanism.
2.3 Supervision with the SSM in the Eurozone
The introduction of the SSM is the direct outcome of a process that was launched by the European Commission, which wrote a proposal in 2012 under the pressure of the financial crisis (Wymeersch, 2014). The ultimate goal of implementing a single supervisory system was to restore the public confidence in the banking sector. This confidence in the banking sector was hurt during the financial crisis that caused huge losses for the banks, but in the process also hurt the state and its taxpayers as spillover effect, since them being the ultimate guarantor. One of the objectives of the bank union is to break this link between the banking sector and the state, by keeping banks under stricter control and getting rid of the idea that at the end of the day, the state and the taxpayers have to step in.
Navaretti, Calzolari, and Pozzolo (2015) say that the SSM had the task to create an institutional harmonisation and central supervision. This task was realised by defining and implementing a coherent harmonised regulatory and legal framework for banks, based on the CRR/CRD IV packages of EU laws1. Also, harmonizing and cooperating the activities of
national supervisory authorities with a single rule-book for the whole sector was achieved. The ECB assessed from then on the balance sheets and activities of 122 significant banks that fell under its supervision, which makes up for 82% (ECB, 2018a) of the bank assets in the euro area (25 trillion in assets).
According to the website of the ECB, to fall under the SSM and thus the direct supervision of the ECB, a bank has to qualify as significant (ECB, n.d.-b). To define
significance, the ECB developed four criteria. If a bank fulfils one or more criteria, it will be recognised as significant and therefore will fall under the SSM. The criteria are as follows: a) Size: the total value of the bank’s assets exceeds €30 billion; b) Economic importance: the bank is important for the specific country of the EU economy as a whole; c) Cross-border activities: if the total value of a bank’s assets exceeds €5 billion and the ratio of its cross-border assets/liabilities in more than one other participating Member State to its total assets/liabilities is above 20% and d) Direct public financial assistance: the bank has requested or received funding from the European Stability Mechanism or the European
1 Please refer to the papers of the ECB (2015) and the Basel Committee (2017) for more information on the
Financial Stability Facility. Instead of testing these requirements myself to find out which Eurozone banks are significant for the ECB, I received lists of significant banks after requesting them from the ECB (see Appendix A). These were used in creating a part of the data described in section 4.1.
3 Accounting conservatism and supervision
The following sections introduce a brief discussion on accounting conservatism in general and for banks. Next, I describe how the change to central supervision can impact banks in general and how it can affect accounting conservatism for banks. After this discussion, I present this thesis’ hypotheses that are developed on the introduced theory in these sections. 3.1 Accounting conservatism in general
A lot of academic research describes accounting conservatism as resulting in earnings reflecting ‘bad news’ more quickly than ‘good news’ (Basu, 1997; Kanagaretnam, Lim & Lobo, 2013). This results in asymmetric timeliness of recognising earnings and losses in the reported income. Basu (1997) describes conservatism with several indicating variables, like the writing down of assets to reflect the obsolescence or impairments, but not revaluating them upwards when there are indicators of a value increase in the assets. Other indicators of conservatism are lower-of-cost or market price accounting for inventories, or the immediate recognition of losses when cost estimates result in a future expected loss on long-term contracts but not if they are expected to increase and thus lead to expected gains.
Leventis et al. (2013) concluded, based on prior literature, that financial statement users benefit from accounting conservatism, since this embeds managerial opportunism, mitigates agency problems associated with managerial investment choices and makes debt agreements more efficient in the presence of asymmetric information. They also mention that evidence in auditing literature suggests that accounting conservatism can lessen litigation risk since lawsuits against auditors are more likely to be related to the overstatement of earnings or net assets.
In the introduction of their research, Beaver and Ryan (p.269, 2005) distinguish two types of conservatism: unconditional and conditional. Conservatism can be unconditional when aspects of the accounting process determined that assets of liabilities yield expected unrecorded goodwill. Conditional conservatism is based on expected events, meaning that book values are written down when unfavourable events are expected, but not written up when favourable circumstances are expected to occur. From now on, when referred to conservatism, I refer to conditional conservatism, since this thesis focuses on this type of conservatism. Specifically, this thesis focuses on conditional conservatism in banks. Therefore, the next section will elaborate more on bank conservatism.
3.2 Accounting conservatism for banks
With the new IFRS 9 standard, introduced in 2014 by IASB, the accounting for loans by banks was changed (Novotny-Farkas, 2016). The incurred loss approach under IAS 39 (Financial Instruments: Recognition and Measurement) was replaced with an expected credit loss (ECL) model. This allowed banks from then on to recognise expected credit losses on their loans earlier. The Basel Committee published in 2015 guidelines on Banking
Supervision in which it referred to this ECL model and the expectations related to this model by supervisors. Novotny-Farkas noticed that these guidelines reflect the Committee’s
preference for conservatism in loan loss accounting. He described this by the example in which the guidelines emphasise that the expected credit loss for the first 12 months should rarely be nil. This means that banks from then on always had to recognise a loan loss provision when a new loan contract was signed.
In their paper, Kanagaretnam, Lobo, and Mathieu (2003) examined other reasons for managers to use their discretion over loan loss provisions, usually the largest accrual for banks, to smoothen the reported income. They found that managers tend to save earnings in relatively good years to compensate for future bad years. This is why the use of loan loss provisioning to smooth income has received a lot of attention from regulators.
Nichols et al. (2009) describe a bank’s loan loss accounting as the reflection of its credit risk management. Loan loss provisions are accrued expenses that reflect the judgment of the manager on the estimation of the changes in future losses from credit risk in the loan portfolio. They state that banks with more conditional conservative accounting recognise provisions for those losses more timely and these are larger than banks with less conditional conservative accounting. Kanagaretnam et al. (2013) found that bank financial reporting incentives and risk-taking are likely to be affected by several factors, such as a bank’s ownership structure, bank monitoring, and bank regulation. They also agree with Nichols et al. (2009) by noticing that loan loss accounting has a material effect on bank earnings and balance strength and requires a certain degree of estimations. These arguments make a bank’s loan loss provisioning a good measure for the conservativeness of a bank’s management in its accounting. The academic literature states that accounting conservatism is especially
important for a bank due to its ambiguousness, complexity, significant information
asymmetry and uniqueness in their contracting (Furfine, 2001; Levine, 2004). Supervisors on banks prefer that banks apply more accounting conservatism by setting aside more loan loss provisions in time of economic growth, to mitigate critical claims when a bank has solvency
issues (Watts, 2003; Turner, 2010). Therefore, it can be concluded that loan loss provisions are a strong measure for accounting conservatism of banks, as central supervisors prefer this according to Turner (2010).
3.3 Effects of transition to central supervision
Navaretti, Calzolari, and Pozzolo (2015) describe in their article that there are cons and pros with a central supervision system as opposed to a local supervision system. Supervising banks exists of at least three main activities: gathering information, processing information and act (or don’t) consequently. The cons and pros directly relate to these main activities.
The first con described by Navaretti et al. (2015) is the difficulty of gathering of information in a centralised system. Although it seems better having not to rely on
information acquired by biased national authorities, the centralised gathering of information on itself is a difficult task when keeping timeliness and the quality of the information in mind. Besides, the transmission of information is also problematic when one deals with “soft,” not quantitative information. Stein (2002) also mentions that dealing with “soft” information, that is information that cannot be directly verified by anyone other than the producing agent, is most efficient in decentralised systems and not in centralised ones. He mentions that in a centralised, hierarchical organisation a lower level manager will perform less research since he will not be able to communicate his findings credibly to his superiors to convince them to allocate capital to him. This effect could also translate to the transition from a (decentralised) national supervisor to the (centralised) ECB acting as the supervisor.
Another mentioned shortcoming by Navaretti et al. (2015) of a centralised supervision, is the fact that centralisation reduces flexibility in designing policies and in tailoring standards to individual countries’ banks and economies under its jurisdiction (and therefore the most optimal policy). Even when the regulatory needs differ across the countries, the central supervisor would prefer to take the perspective of the whole system, rather than on country-level. Lastly, there might be an effect of the safety-nets mentioned by Tröger (2014) on the risk appetite of banks. Studies like Calomiris (1999) and Hovakimian, Kane, and Laeven (2003) found that banks can increase their risk appetite when safety-nets like deposit guarantee systems are introduced since these lower the downward risk for banks.
The main benefit, according to Dell’Ariccia (2015), of centralising regulation is that it internalises any externalities that may exist due to the integration of the financial systems. These externalities are entailed because banks compete internationally. A centralised agency
will apply tighter standards than an independent national one since it ignores independent national incentives to make supervision standards laxer.2 Also, in case of decentralised
supervision, local supervisors might not focus or care about potential negative spillover effects of their supervising actions on foreign banking systems. They simply care more about the effects on domestic banks. A centralised agency does care about these potential negative effects and will act on it since it maintains a higher level of perspective. Navaretti et al. (2015) support this internalisation of externalities theory by mentioning that when a troubled bank has larger international liabilities, this may strengthen the commitment of a national regulator not to bail out the bank, since a portion the expensed money could flow outside of the country. This externality would not exist for a centralised supervisor.
Another pro for the transition from a national supervisor, which is often a national central bank (ECB, n.d.-a), to a more focused central supervisor like the ECB, is mentioned by Barth, Dopico, Nolle and Wilcox (2002). They found that banks being supervised by a national central bank often has more nonperforming loans than banks that are being supervised by a more narrowly focused bank supervisor. So the change from a national central bank as supervisor to the ECB might strengthen the monitoring and control of banks.
Finally, Ongena, Popov, and Udell (2011) found that when home-country regulation for banks is lax, and the home-country supervision is inefficient, domestic banks seek more risk abroad by lending to more opaque or observably riskier companies and lends less in its home country. This suggests that the effectiveness of supervision in a country affects a bank’s lending choices and willingness to take risks.
3.4 Effects of supervision on accounting conservatism
The introduction of the SSM can have a positive effect on accounting conservatism of banks in several ways. First, because of the change from a(n) (inferior) national supervisor to the ECB as the more narrowly focused central supervisor, supervised banks can develop more awe for the new (more superior) supervisor. Because of this increase in awe, banks possibly account more conservative, which can be seen in an increase in loan-loss provisions. Norden and Stoian (2014) might support this theory by finding the opposite, that banks create lower loan loss provisions (LLPs) when regulatory capital requirements have decreased. The ECB
2 An example for a national independent incentive to make supervision standards laxer, according to
Dell’Arricia (2015), is that a country can increase profitability and therefore international competitiveness for domestic banks with lower regulation standards and therefore a lower burden for the bank to apply with these standards. The Nash equilibrium in this case would result in a race to the bottom.
demands higher regulatory capital requirements from the banks it supervises (ECB, 2018a), suggesting that banks might increase their loan loss provisions and therefore be more conservative in their accounting. Kandrac and Schlusche (2017) also attribute to this awe theory, by finding in their natural experiment that the examined bank increased their risk-taking when the level of supervision decreased.
Additionally, Gebhardt and Novotny-Farkas (2011) add another way to argue that the introduction of the SSM might have a positive effect on accounting conservatism. They state that the primary goal of bank supervisors is financial stability. Supervisors use financial statements for determining the equity ratios required by their capital adequacy regulations. Therefore, strict supervisors can prefer and demand more conservative loan loss provisioning by its supervised banks, since according to Leventis et al. (2013) this is preferred by financial statement users to embed managerial opportunism. The assumption that supervisors can prefer more conservatism in the loan loss accounting of banks is supported by Novotny-Farkas (2016). He describes that the Basel Committee’s Guidelines reflect its preference for a higher degree of conservatism in a bank’s loan loss provisioning. That can explain why also the introduction of the SSM might lead to more accounting conservatism since supervisors seem to prefer this.
Also, Benston and Wall (2005) state in their article that bank supervisors expect reserves to at least cover expected losses, suggesting that the bank supervisors are excessively
conservative. Altogether, these arguments predict that the introduction of the SSM had a positive effect on loan loss accounting conservatism. Therefore, I present the following hypotheses:
H₀(NULL HYPOTHESIS): The introduction of the Single Supervisory Mechanism had no effect on the level of conservatism that banks in the Eurozone applied to their loan loss accounting.
H₁(AWE HYPOTHESIS): The introduction of the Single Supervisory Mechanism had a positive effect on the level of conservatism that banks in the Eurozone applied to their loan loss accounting.
4 Sample selection and descriptive statistics
In the following sections, I describe how the sample for this thesis was selected and what the motivation was behind the sample selection. Also, I present the descriptive variables of the sample.
4.1 Sample selection
The bank data I use for my sample is collected from the Orbis database from Bureau van Dijk, which has financial data of over 300 million companies worldwide. This database provides financial information retrieved from financial statements for the Eurozone banks, including most banks that fall from a certain point under the SSM. I also received data from the ECB by requesting lists of banks that were introduced in the SSM and in which year (see Appendix A). Combining these sources and removing banks with incomplete relevant data, a sample of 201 bank-years between 2012 and 2015 with complete information remains from 78 different Eurozone banks that fall from a certain point under the SSM. Before dropping these, 138 different Eurozone banks were in the sample with 414 bank-years between 2012 and 2015.
To be able to control for environmental changes in the Eurozone, in example other changes in regulation, I also added a control group of banks that did not fall under the SSM in the analysed period. This control group was created by using the data from the top 500 biggest banks in the Eurozone, based on total assets. From this top 500, all banks which did fall under the SSM from a certain point were excluded. Also, the banks from the remainder which did not have complete relevant data were also dropped. 309 Eurozone bank-years in which a bank did not fall under the SSM remained, with 105 different banks. This control group is chosen because these banks are in the same countries as the SSM group, are also the biggest banks measured at total assets and therefore should be following the same
environmental changes, except for the introduction of the SSM.
To complete the data, financial information from 2011 and 2016 was used to create certain forward-looking and backwards-looking variables, which are described below. I used information of the years between 2011 and 2016 since these are the years around the
introduction of the SSM by the ECB for most banks, namely in 2014. 2011 is the first year that Orbis provides complete information for most banks for and 2016 is the most recent year that is available at the moment of writing.
The downside of the combination of different sources of data, namely Orbis and the ECB, is that both sources do not use the same names or other identifiers for the same company. Lacking any other option, I was forced to combine the ECB bank data with the Orbis bank data by hand, which was done by searching for the company name and country provided by the ECB in Orbis. However, not all mentioned banks on the lists of the ECB were available in Orbis. Therefore, some SSM banks were not included in the sample. Other banks from the ECB lists that were available in Orbis did not have complete information on all relevant variables for my test. Also, these banks were excluded from the sample. The remaining SSM bank sample is still sufficient for testing, as it still reflects a significant portion (17.2 trillion total assets in 2014) of the total population (25 trillion total assets in 2014) of SSM banks.
4.2 Descriptive statistics
I present in Table 1 the descriptive variables for the sample with the Eurozone banks that fall under the SSM at a certain point in the sample period (hereafter: SSM group/banks) and banks that do not fall under the SSM in the period of analysis (hereafter: control
group/banks). The descriptive variables are described by their mean, minimum, maximum, the 25th and 75th percentile, standard deviation, and Skewness. The variables were also transformed by logging and then Winsorizing them, in order to control for outliers. This brought the Skewness to more acceptable levels in most cases. Only the Skewness for the control group in the variables of the loan loss provisions and the future change in
nonperforming loans is somewhat high. These statistics were used for the testing in section 5. A more brief description of the statistics before the transformation of the variables can be found in Appendix B, in order to be able to analyse the absolute differences between the two groups.
The selection of statistics was made based on the variables used in the model for loan loss provisions in the article of Nichols et al. (2009) and Kanagaretnam et al. (2013), since these articles also compared one or more effects on banks loan loss provision accounting to test for differences in accounting conservatism by using a dummy variable. However, some
descriptive statistics were not available in Orbis or any other available database and were therefore not used3.
The variables were constructed in the following fashion: the Assetst are the total
assets of a bank on the balance sheet; the loan loss provisions (LLPt) is the amount of current
year loan loss provisions on the balance sheet, divided by the beginning of the current year total loans; the current year change in nonperforming loans (ΔNPLt) is the difference between
the current year end nonperforming loans (the sum of borrowed money upon the debtor has not made the scheduled payments for a period of time of at least 90 days) and the beginning of the current year nonperforming loans, divided over the current gross loans; the same has been done for the difference in nonperforming loans for the previous year (t - 1) and the upcoming year (t + 1); the net charge-offs (NCOt) are the current year realised net
charge-offs (realised loan losses) divided over the gross loans; the same has been done for the upcoming year (t + 1); the loan loss reserves (LLRt-1) are the last year end loan loss reserves
on the balance sheet divided over the last year gross loans and last, the growth of the loan portfolio (LNGROt) is the current year end of the total loans, divided over previous year end
total loans.
The reason for the dividing of variables over total loans or total assets is to control for the size of the banks in to make the variables comparable over the total sample. The
backwards-looking variables (t - 1) and forward-looking variables (t + 1) have been constructed on the realised figures of the year before the analysed year and the year after. I expect that these figures are used by the management in the current year to look back and predict future figures so that they can decide in the current year on what choices they will make relating to loan loss provisions. The change in nonperforming loans is a predicting value for loan loss provisions since a positive change means that more loans are failing and therefore the management has to respond by recognising more provisions for potential losses. The same applies to net charge-offs since these are realised losses made on loans, indicating that the bank might need to recognise more provisions.
I added two other control variables, following Nichols et al. (2009) and Kanagaretnam et al. (2013): the last year loan loss reserves (LLR), because Ryan (2007) states that over-reserved banks will recognise fewer provisions in the following year. The core difference between loan loss provisions and loan loss reserves is that a reserve is recognised based on
3 Databases used in the mentioned articles of Nichols et al. (2009) and Kanagaretnam et al. (2013) did not
contain Eurozone banks or were not available anymore at the moment of writing, due to the discontinuation of the BankScope database.
manager’s discretion to have a buffer for future years, usually in years in which the company is profitable. Loan loss provisions are recognised when there is an indicator of a possible future loss, in example a lender that has lost his job and might suffer in income because of this. The second control variable I added is the growth in the loan portfolio (LNGRO) since a bank with more loans needs to recognise more loan loss provisions under IFRS 9.
The descriptive variables in Appendix B show that in both groups (SSM and control) the sample contains a huge difference in total assets between the smallest and largest bank, even after being transformed by logging and Winsorizing. The fact that the lowest amount of assets of 2.6 billion is still in the SSM sample and therefore a significant SSM bank, even though that does not meet the significance requirement related to size (more than €30 billion assets), can be explained by the fact that it probably met one of the three other significance requirements (b, c or d) of the ECB to fall under the SSM, as mentioned in section 2.3. It is also shown that the mean of the assets of the SSM group is almost triple the amount of the mean of the control group, indicating that the SSM banks really are a big part of the Eurozone bank sector (82%, based on total assets, according to the ECB, 2018a). SSM banks also have relatively higher loan loss provisions, net charge-offs, and loan loss reserves. This might suggest that they already account more conservative since they seem to prefer to acquire more reserves and provisions and recognise loan losses quicker through the net charge-offs than the control group. The control group banks have a relatively higher change in
nonperforming loans, which might suggest that they are performing relatively worse by selecting loans with a higher credit risk than the SSM banks, or that SSM banks tend to borrow their money relatively more to trustworthy lenders. This might already be a consequence of being supervised by the ECB, who can demand a lower credit risk of their supervised banks.
Table 2 presents the correlation matrix between the different independent variables used for analysing conservatism in loan loss provisions. It shows that there is a significant positive correlation between loan loss provisions and last and current year changes in nonperforming loans. This seems not surprising, as managers tend to increase their loan loss provisions when loans turn nonperforming to cope with possible future losses. Loan loss provisions correlate positively with last and current year net charge-offs, which would assume that managers increase their loan loss provisions when more loan losses are realised. This could lead to more expected future losses on the loan portfolio. Thus an increase in loan loss provisions is necessary. The significant positive correlation between loan loss provisions and loan loss reserves and the negative correlation with growth in the loan portfolio is
surprising, as the literature (Ryan, 2007; Kanagaretnam et al., 2003) suggests the opposite. The significant negative correlation with the total assets suggests that bigger banks relatively recognise fewer loan loss provisions than smaller banks. A possible explanation for this can be that smaller banks relatively are more vulnerable for future loan losses than larger banks and therefore need to recognise more provisions to cope with them. Lastly, the significant negative correlations between loan loss provisions and the control dummy and the interacting variables with the control dummy would make one assume that the control group seems to recognise relatively lower loan loss provisions than the SSM group. This supports the
assumption made in the previous section about the difference in loan loss provisions between the SSM group and the control group.
Other nominable correlations are the significant positive correlations between last, current and future year changes in nonperforming loans. This would mean that a higher change in nonperforming loans usually would also be accompanied by a higher change in the adjacent years. These variables also each have a significant negative correlation with the net charge-off variables, remarkably hinting that more nonperforming loans relate to lower recognition of loan losses through realised net charge-offs. Also remarkable is that these nonperforming loan variables, except the future year variable, do not have a significant correlation with their interacting variables with both dummy variables, which could mean that generally, a change in nonperforming loans would not automatically lead to a change in these when interacting with their dummy variables. Current year net charge-offs do not correlate significantly with other interacting variables with the SSM dummy variable, other than its interaction with the dummy, while future year net charge-offs do. This hints that the SSM dummy variable only has a significant effect with future year charge-offs, however not strongly (only at a 0.05 level). Last, last year loan loss reserves and total assets correlate significantly with almost every other variable, making these a variable that fluctuates strongly with all the other variables. The growth of the loan portfolio does not, which might be
explained by the fact that almost all the variables are divided over the total loan portfolio, eliminating the correlation with this variable.
To conclude, all interacting variables unsurprisingly have a significant correlation with each other since they are all explained by either the SSM dummy variable or the control dummy variable. And since when the SSM dummy variable is always a 1 when the control dummy variable is a 0 and the other way around, these variables should have a strong correlation.
5 Empirical tests and results
In the next section, I describe a fixed effects model regression between the introduction of the SSM and the amount of loan loss provisions banks have recognised, and subsequently I describe the refined model I follow for predicting banks’ loan loss provisions of Nichols et al. (2009) and Kanagaretnam et al. (2013). After that I run a regression analysis on these models. In the second part of this section, I discuss the outcomes of the tests and provide some
interpretations for them. 5.1 Testing and results
The central question of this thesis is what impact the introduction of the SSM had on the accounting conservatism through loan loss provisions of the Eurozone banks. To answer this question, I predict the following fixed effects model (1) to measure loan loss provisions:
where LLPt denotes the loan loss provisions divided by beginning of year total loans; Dssmt
denotes the dummy variable for a year in which a bank falls under the SSM; COMPANY denotes the individual fixed company effects and YEAR denotes the individual fixed year effects. If the introduction of the SSM on itself would have made a direct impact on the loan loss provisioning, the dummy variable Dssm should have a significant coefficient (φ1).
The results of the simple linear regression on the model (1) is presented in Table 3. The high adjusted R square together with both an F-score of 9,206 and a Durbin-Watson score of 2,08 indicates that the model explains the dependent variable well, is compatible with the data and is free of autocorrelation. The regression shows that there is no significant direct effect of the dummy variable for the SSM on the recognition of loan loss provision, controlling for fixed company and fixed year effects. This would result in the first assumption that the introduction of the SSM has not affected accounting conservatism of loan loss
provisions for Eurozone banks. However, further analysis could lead to other results on how banks might be more or less sensitive to triggers for increasing their loan loss provisions, like a change in nonperforming loans, under the regime of the ECB through the SSM. The next section refines this analysis, by using the model of Nichols et al. (2009) and Kanagaretnam et al. (2013).
The above-mentioned articles refine their analysis by examining the effect of the relation between loan loss provisions and changes in nonperforming loans. They discuss that changes in nonperforming loans represent exogenous and relatively nondiscretionary
indicators of possible future credit losses. I adjust their model to measure for my independent variable, the influence of the SSM. This way, I can compare differences in the timeliness of bank-years under the SSM and bank-years not under the SSM in loan loss recognition by analysing the associations between loan loss provisions and current year, last year and future year’s nonperforming loans coefficients. It can be argued that managers not only look at the current year figures when assessing the amount of loan loss provisions which is necessary to be recognised, but this assessment is also made based on last year figures (ΔNPLt-1) and
expected future year figures (ΔNPLt+1). Even after recent introduction to the SSM, a
manager’s opinion on last year figures can be influenced due to the change in supervision.4
While performing this test, I also control for bank’s size, last year’s loan loss reserves, growth in the loan loss portfolio and net charge-offs.
To show how banks are more or less sensitive for changes in their nonperforming loans, net charge-offs and loan loss reserves under the SSM in their accounting for loan loss
4 In example, when a bank has been introduced in 2014 to the SSM, a manager still can look back at the figures
of 2013 in which it did not fall under the SSM to make decisions for 2014. It can be argued that even though the bank was not under the SSM in 2013, the opinion of the manager in 2014 on these figures can be influenced by the fact that it now falls under SSM in 2014.
provisions, I estimate the following adjusted model (2) of Nichols et al. (2009) and Kanagaretnam et al. (2013):
where LLPt denotes the loan loss provisions divided by beginning of year total loans; ΔNPLt
denotes the change in nonperforming loans between the current year and last year divided by the gross loans; NCOt denotes the net charge-offs divided by the average gross loans; Dssmt
denotes the dummy variable for a year in which a bank falls under the SSM; LLRt-1 denotes
the loan loss reserves at last year’s end; LNGROt denotes the growth in loans by dividing
current year total loans by last year’s total loans; Assetst denotes the total assets of current
year and Dcontrolt denotes the dummy variable for the control group of banks that do not fall
under the SSM in the analyzed period.
This adjusted model (2) of Nichols et al. (2009) reflects the five stages of loan loss recognition timing. According to them, managers base their expectations for loan loss provisions on last year’s loans that turned into nonperforming (ΔNPLt-1), on the current year
(ΔNPLt), and on the expected loans that turn into nonperforming in the future year (ΔNPLt+1).
Current year net charge-offs (NCOt), or realised loan losses, and future expected net
charge-offs (NCOt+1) also relate directly to loan loss provisions. Therefore I expect those five
variables to have positive coefficients (β1, β2, β3, β4, and β5), just like Nichols et al. (2009)
found in their article. Since these variables do not include the dummy variable for the SSM, their coefficients show the associations between these variables and non-SSM bank-years.
Next, I make these five variables interact with the dummy variable for the SSM, conform the tests of Nichols et al. (2009) and Kanagaretnam et al. (2013). However, according to Ryan (2007), an increase in loan loss reserves of last year (LLRt-1) could also
impact the amount of loan loss provisions in the current year, because an over-reserved bank might want to recognise less provisions in the next year. This effect might also be interesting to analyse under the SSM, as this also might change under the new supervisor who prefers high reserves. Therefore, I also interact variable LLRt-1 with the SSM dummy variable. This
way I can analyse differences between bank-years that fall under the SSM and bank-years that do not. The coefficients for Dssmt* ΔNPLt-1, Dssmt * ΔNPLt and Dssmt* ΔNPLt+1 (β11,
β12 and β13) are expected to be positive, since I expect that banks that fall under the SSM
SSM, controlling for size and endogeneity. Coefficients for Dssmt* NCOt and Dssmt*
NCOt+1 (β13 and β14) are expected to be positive as well since higher losses could push
managers to recognise more loan loss provisions under the SSM to seem more conservative to the central supervisor.
Next, I add interacting variables of the five stages of loan loss recognition timing and loan loss reserves (LLRt-1) with the control group dummy (Dcontrol). This way I will be able
to analyse the difference in loan loss recognition timing between the control group and the banks that are going to fall under the SSM in the analysed period. I have no prediction for these coefficients (β18, β19, β20, β21,β22, and β23), since the SSM group does not fall under the
SSM for the whole period, only partly, and I predict them only to be more conservative when they fall under the SSM. This ambiguity would make a prediction too hard to make.
Following the articles mentioned above, I include the control variable LNGROt to
control for differences across banks in expected loan loss provisions. LNGROt is expected to
have a positive correlation with LLPt, since a bigger loan portfolio will likely lead to higher
loan loss provisions (Kanagaretnam et al., 2003) which is also required under IFRS 9. The results of the linear regression on the model (2) is shown in Table 4. It should be mentioned that the Variance Inflation Factor was relatively high for most coefficients, which indicates multicollinearity. This is not surprising when accounted for the number of
interacting variables. Due to this high multicollinearity, variables Dssmt, Dssmt* NCOt+1,
Dcontrolt and, Dcontrolt * NCOt were excluded from this model5. The individual dummy
Dssmt was already tested on itself in model 1 however, so this should not be a problem. The
high adjusted R square together with both an F-score of 58,379 and a Durbin-Watson score of 2,13 indicates that the model explains the dependent variable well, is compatible with the data and is free of autocorrelation.
Coefficients β2, β3, β4 and β5 on ΔNPLt, ΔNPLt+1, NCO and NCOt+1 are positive as
predicted, implying that generally banks recognise loan loss provisions in a timelier manner relative to last year changes in nonperforming loans, implicating a degree of accounting conservatism (Kanagaretnam et al., 2013). However, surprisingly, the coefficient of last year changes in nonperforming loans (β1) is not significant, implying that banks do not generally
recognise loan loss provisions timelier relative to last year changes in nonperforming loans. Coefficients (β6 and β7) of last year loan loss reserves and growth in the loan portfolio are
both significant. However the predicted signs are the opposite of the estimated direction of the coefficients found by the regression analysis.
When interacting the dummy variable for an SSM year with the change in nonperforming loans of current year and next year, it does have a significant estimated positive effect (β11 and β12) on loan loss provisions. This would suggest that banks do
recognise timelier and larger loan loss provisions under the SSM when there is a change in nonperforming loans for current or future years, confirming my predictions. Here again, however, last year changes in nonperforming loans interacting with the dummy variable for the SSM do not have a significant effect on loan loss provisioning. Also current year net charge-offs under a SSM year seems to have a significant effect (β13) on loan loss provisions.
However, this effect is in the opposite direction than I predicted. I refer to the next section for implications of these findings.
5.2 Implications for results
The results of the tests in the previous section have multiple outcomes for our analysis. The regression on the first Model (1) did not have any significant results, suggesting that the mere introduction of the SSM on itself did not directly impact the accounting conservatism through the way that banks account for their loan loss provisions. It seems probable that a bank does not directly increase its loan loss provisions, just because it was introduced to a new
supervisory regime. This on itself has no signals that a bank needs to adjust its loan loss provisions since the loan portfolio and the credit risk of the bank has not changed solely due to the introduction of the SSM. However, the introduction of the central supervisor might have impacted the way that a bank is sensible to different kind of triggers, like an increase in nonperforming loans. A more refined analysis through the second adjusted Model (2) of Nichols et al. (2009) and Kanagaretnam et al. (2013) shows how banks are more or less sensitive for changes in their nonperforming loans, net charge-offs and loan loss reserves under the SSM in their accounting for loan loss provisions. This model was chosen for the analysis of accounting conservatism since it describes several triggers for a manager to change its loan loss provisions and measures the sensibility of a bank to those triggers resulting in timelier or larger recognition of loan loss provisions.
The regression on the second Model (2) confirmed my prediction that changes in nonperforming loans for the current and last year have generally a significant positive effect on loan loss provisioning of banks. This implies that banks generally recognise loan loss provisions in a timelier manner relative to current and future year changes in nonperforming loans, implicating a degree of accounting conservatism (Kanagaretnam et al., 2013). The same applies to the net charge-offs in current and future year. Interestingly, the stand-alone recognition of last year loan loss reserves has a positive effect on loan loss provisions, contradicting my prediction based on the effect of Ryan (2007). This might implicate that banks with higher reserves are in their origin more conservative in their loan loss
provisioning. When interacting with the SSM dummy variable, however, it does have a significant negative effect, conform the effect of Ryan (2007). It seems that only banks under the SSM are willing to lower provisions after recognising more loan loss reserves in the previous year, maybe because they already account more conservative and therefore can lower their loan loss provisions if they raised their reserves last year. Combining this explanation with the result of coefficient β22, which interacts loan loss reserves with the
cannot generally conclude the difference between the SSM group not under the SSM, the SSM group under the SSM and control group on how they account for their loan loss provisions after an increase in last year loan loss reserves.
The fact that the coefficient (β8) for assets is insignificant shows that generally bigger
banks do not recognise less or more loan loss provisions. This is surprising since one could argue when banks grow in size it could have more or fewer loans relative to their other assets and therefore have more or less loan loss provisions relative to its total assets. We might assume that when banks grow in size, some banks increase their loan portfolio relatively, while other banks increase relatively other assets since the coefficient of assets has no significant effect. The results also show that the growth in the loan portfolio (LNGROt) has
the strongest coefficient (β7), which is also in the opposite direction of my expectation. It
seems assumable that the increase/decrease in the number of loans is the strongest influencer for a bank to acquire/lessen its loan loss provisions, since this leads to more/fewer loans which are exposed to credit risk. However, the fact that an increase in the loan portfolio leads to a decrease in loan loss provisions is unexpected. This can probably be explained by the fact that the LLP variable is constructed by the total amount of loan loss provisions, divided over the total loans of the bank. When there is a growth in the number of loans in the portfolio, the denominator for the LLP variable increases. Even though the bank has to recognise some loan loss provisions under IFRS 9, this would relatively be a lower provision since it is only a provision for the expected losses in the upcoming twelve months (Novotny-Farkas, 2016), which is rarely higher than the total expected credit losses over the whole loan portfolio6. Therefore, the numerator in LLP over total loans increases less than the
denominator, resulting in a lower LLP variable after an increase in the LNGRO variable. After interacting the SSM dummy variable with the variables for current year and future year changes in nonperforming loans, I find significant positive coefficients (β11 and
β12) on loan loss provisions. This confirms my predictions that banks that fall under the SSM
account timelier and larger loan loss provisions and are thus more conservative when there is a change in nonperforming loans in the current or future years. It seems, however, that an increase in current year net charge-offs will result in a bank lowering its loan loss provisions, which I did not predict. It is possible that a bank under the SSM that realised more loan losses
6 In example, a big Dutch bank had in 2016 in their retail department for the Benelux 4,7% of their loan
portfolio categorised as NPL (nonperforming loans), based on their published financial statement of 2016. The credit risk for these nonperforming loans is significantly higher than generally for loans in the first twelve months, hence the total credit risk over the total loan portfolio is higher than loans that just have been recognised.
through its net charge-offs had to get rid of the relating nonperforming loan and therefore could write-off the loan loss provision. Another explanation is that the bank expects to successfully adjust their lending choices as a response to these current year losses for the next year and thus does not need more loan loss provisions.
It seems surprising that the coefficients for both under and not under the SSM of the variable measuring the change in nonperforming loans of last year have no significant effects on the amount of loan loss provisions a bank recognises. An explanation for this can be the fact that this coefficient in this sample measures mostly years that were still not under the SSM. The t - 1 for 2014 is 2013, a year in which no banks fell under the SSM, and the t - 1 for 2015 is 2014, a year in which most but not all banks were introduced to the SSM. This would mean that the coefficient (β10) measures partly the change in nonperforming loans in
non-SSM years and partly the change in nonperforming loans in SSM years and is therefore not very reliable. It is not possible at the moment of writing to control for this and make it so that this coefficient only measures last year changes in nonperforming loans under the SSM since it has been too short of a period since the introduction of the SSM and figures for 2017 and 2018 are not available yet. Future research should be able to eliminate this problem.
Based on the outcomes of this regression, it can be concluded that banks under the SSM recognise loan loss provisions in the current year larger and more timely when there is an increase in nonperforming loans in the current and future year. An increase in net charge-offs in the current year, however, results in a bank under the SSM lowering their provisions, indicating a lower level of accounting conservatism. This results in the fact that it remains unclear if the situation described by Bushman and Williams (2012), in which the way banks account for loan losses is in direct conflict with the goals of supervision described by Rochet (2005), has completely changed after the introduction of the SSM. The positive coefficients (β11 and β12) relating to current year change in nonperforming loans support the statement that
there has been an improvement in the way banks account for their loan losses related to the goals of supervision. However, some ambiguity appears with the negative coefficient (β13) of
the variable Dssmt * NCOt, since this might imply the opposite.
Gormley, Kim, and Martin (2012) found in their paper that a change in accounting conservatism in the banking industry of India had an impact on the way firms apply
conservatism to their accounting. It is thinkable that this also might happen in the Eurozone since the reasons for these adjustments in the other sectors were not culture or nationality related. Companies in the Eurozone might also want to change their accounting policy to attract more capital for a lower cost from these banks that adjusted their accounting
conservatism. The idea behind this is that conservative banks might prefer to invest in conservative companies since the accounting conservatism embeds managerial opportunism and this is reflected in the annual reports of the company. If a company fulfils this desire of the bank, the cost of capital might decrease. This could mean that a change to the centralised supervision of Eurozone banks, which in turn changed the accounting conservatism of banks in the Eurozone, could theoretically significantly impact the economy of the Eurozone by influencing the cost of capital for companies. Future research could look further into this effect.
Besides supervisors (Turner, 2010) and lenders and borrowers (Gormly et al., 2012) being interested in an increase in accounting conservatism, also other parties might benefit according to García Lara, Garcia Osma and Penalva (2014). They refer to outside equity-holders, which can be interested in the limitation of earnings management and a more complete information environment that a conservative reporting system provides. This, in turn, could mean that it would be easier to raise equity for the more conservative companies, like the Eurozone banks in our sample.
The next section provides a summarising conclusion of the results described in this section, some limitations of this thesis and gives recommendations for further research on how to solve these limitations.
6 Conclusion
The main research question of this thesis was to find out if banks are more conservative in their loan loss provisioning when they were introduced to the SSM, which was launched in 2014. I addressed this question by analysing a sample of 78 Eurozone banks which fell under the SSM from a certain point and 105 Eurozone banks which did not fell under the SSM in the period 2012-2015, for a total of 510 analysed bank years. For this, I used the financial data from these banks in the period from 2011-2016.
The empirical results from this thesis indicate that banks are positively sensitive to a change in nonperforming loans in the current or next year while under the SSM in their accounting for loan loss provisions. This indicates more accounting conservatism since Turner (2010) mentioned that central supervisors prefer conservatism through loan loss provisioning. However, there is an opposite sensitivity of banks when there are more net charge-offs recognised in the current year under the SSM in their loan loss provisioning. This might suggest that banks account less conservative. Concluding, it remains a bit ambiguous if the situation described by Bushman and Williams (2012), in which the way banks account for loan losses is in direct conflict with the goals of supervision described by Rochet (2005), is changed after the introduction of the SSM. This is due to the fact that it is not absolutely clear if this introduction improved accounting conservatism of banks.
My thesis contributes to existing literature in several ways. It adds to the academic research by describing motivations of banks for when they account more or less conservative. The outcome can also be interesting for regulators and supervisors, in how their new
regulations might impact the accounting for banks since they seem to prefer banks to account conservative for their loan loss provisions (Turner, 2010).
This thesis and its outcome might be limited in a few ways: First, there are only a few years of data available since the introduction of the SSM. Future studies should be able to analyse more years after the introduction of the SSM and therefore might see different results since banks could be more adjusted to the new supervisor and the change in supervision. Second, I have only tested the change in conservatism in loan loss provision accounting. There are multiple ways to test for accounting conservatism which this thesis did not test. However, it is argued that loan loss provisioning is a good measure for a bank’s accounting conservatism (Turner, 2010). Future research could also implement other ways of measuring accounting conservatism, like the timing of recognition of increases and decreases in
since they might have access to databases used by the supervisors themselves. Future research might solve the mentioned limitations. Also, it might be interesting to research the effect of the change in accounting conservatism of banks on other companies in the
Eurozone, similar to the results that Gormley et al. (2012) found in India.
References
Barth, J. R., Dopico, L. G., Nolle, D. E., & Wilcox, J. A. (2002). Bank Safety and Soundness and the Structure of Bank Supervision: A Cross‐Country Analysis. International Review of Finance, 3(3‐4), 163-188.
Barth, M. E., & Landsman, W. R. (2010). How did financial reporting contribute to the financial crisis?. European accounting review, 19(3), 399-423.
Basel Committee (2017). High-level summary of Basel III reforms. Retrieved from: https://www.bis.org/bcbs/publ/d424_hlsummary.pdf
Basu, S. (1997). The conservatism principle and the asymmetric timeliness of earnings. Journal of accounting and economics, 24(1), 3-37.
Beaver, W. H., & Ryan, S. G. (2005). Conditional and unconditional conservatism: Concepts and modeling. Review of accounting studies, 10(2-3), 269-309.
Beetsma, R., & Eijffinger, S. (2009). The restructuring of financial supervision in the EU. European View, 8(1), 3-12.
Benston, G. J., & Wall, L. D. (2005). How should banks account for loan losses. Journal of Accounting and Public Policy, 24(2), 81-100.
Bushman, R. M., & Williams, C. D. (2012). Accounting discretion, loan loss provisioning, and discipline of banks’ risk-taking. Journal of Accounting and Economics, 54(1), 1- 18.
Calomiris, C. W. (1999). Building an incentive-compatible safety net. Journal of Banking & Finance, 23(10), 1499-1519.
Chang, M. (2015). The rising power of the ecb: the case of the single supervisory
mechanism. In biennial conference of the European Union Studies Association (pp. 5- 7).
Dell’Ariccia, G. (2015). Governance and Policy Challenges of Forming and Running a Supervisory and Regulatory Union: A Theoretical Perspective 23. European Economy, (3), 71.
ECB (2008). Monthly Bulletin. May. Retrieved from: https://www.ecb.europa.eu/pub/pdf/mobu/mb200805en.pdf
ECB (2015). The impact of the CRR and CRD IV on bank financing. Eurosystem response to the DG FISMA consultation paper, December 2015.
ECB (2018a). Single Supervisory Mechanism. Retrieved from
ECB (2018b). Banking Union. Retrieved from
https://www.bankingsupervision.europa.eu/about/bankingunion/html/index.en.html ECB (n.d.-a). National supervisors. Retrieved from
https://www.bankingsupervision.europa.eu/organisation/nationalsupervisors/html/ index.en.html
ECB (n.d.-b). Criteria for determining significance. Retrieved March 16, 2018, from https://www.bankingsupervision.europa.eu/banking/list/criteria/html/index.en.html European Council (2012). European Council 13/14 December 2012 Conclusions. EUCO
205(12).
Furfine, C. H. (2001). Banks as monitors of other banks: Evidence from the overnight federal funds market. The Journal of Business, 74(1), 33-57.
García Lara, J. M., Garcia Osma, B., & Penalva, F. (2014). Information consequences of accounting conservatism. European Accounting Review, 23(2), 173-198.
Gebhardt, G. U., & Novotny‐Farkas, Z. (2011). Mandatory IFRS adoption and accounting quality of European banks. Journal of business finance & accounting, 38(3‐4), 289-333.
Gormley, T. A., Kim, B. H., & Martin, X. (2012). Do firms adjust their timely loss recognition in response to changes in the banking industry?. Journal of Accounting Research, 50(1), 159-196.
Hovakimian, A., Kane, E. J., & Laeven, L. (2003). How country and safety-net
characteristics affect bank risk-shifting. Journal of financial services research, 23(3), 177-204.
Hunt, A., Moyer, S., & Shevlin, T. (2000). Earnings volatility, earnings management, and equity value. Unpublished working paper. University of Washington.
Kanagaretnam, K., Lim, C. Y., & Lobo, G. J. (2013). Influence of national culture on accounting conservatism and risk-taking in the banking industry. The Accounting Review, 89(3), 1115-1149.
Kanagaretnam, K., Lobo, G. J., & Mathieu, R. (2003). Managerial incentives for income smoothing through bank loan loss provisions. Review of Quantitative Finance and Accounting, 20(1), 63-80.
Kandrac, J., & Schlusche, B. (2017). The Effect of Bank Supervision on Risk Taking: Evidence from a Natural Experiment. Retrieved from SSRN:
