0
Do incoming bank CEOs take “big baths”?
Master Thesis
Andrei Budescu, student number 5746361 August 2014
1
Introduction
This paper aims to investigate the extent of opportunistic earnings management in the periods surrounding commercial banks’ CEO changes. The key question is if new bank CEOs' are inclined to take an “earnings bath” during their first year after having been appointed, by increasing loan loss provisions (LLPs) and implicitly decreasing the banks’ profitability in order to reverse LLPs during future periods. The rationale for doing so would be that the new bank CEOs can blame the company's poor performance on the previous CEO and subsequently take credit for the future years’ performance improvements.
The motivation for this paper initially stems from an article that appeared in the Romanian media in 2009, announcing that eight out of the 40 banks operating in the country had changed their CEOs during the previous year alone. Subsequent changes continued to take place, providing anecdotal evidence of new CEOs engaging in “earnings baths”. However, there is a lack of empirical evidence on the topic, not only with respect to Romanian banks, but with respect to banks in general. Existing literature has found evidence of “earnings baths” for (non-financial) firms in general; at the moment when this paper had been started no such study had been performed for banks, as financial institutions are excluded from the samples of existing studies focusing on commercial companies due to differences in accrual processes. One would expect that, faced with similar incentives as non-bank CEOs, bank CEOs would also engage in earnings management, but the question of whether or not they do so remains pertinent not only due to the methodological differences in the accruals models used in performing such analyses, but also due to banks having more disclosure and reporting requirements and being part of a highly regulated industry, which although may not fully prevent them from engaging in earnings management in the longer term, could deter a new CEO from doing so upon taking office.
With CEO turnover being a recurrent event for banks, similarly to all other firms, we believe that it is important to understand if new bank CEOs are also inclined to and engage in “earnings baths”. Due to data constraints surrounding Romanian banks, this paper investigates this issue using data for US commercial banks. In 2014, two research papers also investigating earnings management surrounding bank CEOs turnover have recently been released: Sarkar, Subramanian
2
and Tantri (2014) and Bornemann, Kick, Pfingsten and Schertler (2014) focus on banks in India and Germany respectively and both find evidence in favor of earnings baths done by incoming bank CEOs. Unlike these papers, while we do find evidence in favor of new bank CEOs recording abnormal loan loss provisions, our results are not statistically significant and suggest that they do not do so to engage in “earnings baths” or for income smoothing purposes, but that the abnormal loan loss provisions are driven by the level of existing loan loss allowances. If bank CEOs do increase loan loss provisions as a result of being risk adverse (and not necessarily for the purpose of creating a buffer income), this could also partially explain why outsider CEOs or CEOs appointed as a result of a non-routine turnover may record higher abnormal loan loss provisions than insider CEOs / appointed as a result of a routine turnover: given that the first are less familiar with the bank than the latter, their incentive for booking higher abnormal loan loss provisions is also higher, as they are not as able to fully commensurate the risks currently borne by the bank’s loan portfolio.
The paper is organized as follows: Section 1 provides an overview on earnings management in general and on the existing literature on earnings baths surrounding CEO turnovers for (non-financial) firms. Section 2 reviews the existing literature on earnings management done by banks, followed by Section 3, which presents the hypotheses. Section 4 develops an econometric model for the estimation of the nondiscretionary portions of the allowance for loan loss provisions and presents the sample selection. Section 5 presents the empirical results and concluding remarks are provided in Section 6.
Section 1. Earnings management
This paper falls under the broader literature focusing on earnings management. Earnings management has become an issue central to accounting research with a large number of studies starting to investigate the topic starting with the 1970s and 1980s. Several definitions have been brought to the term1 but for the purpose of this paper we will refer to that of Healy and Wahlen (1999, p. 368): “Earnings management occurs when managers use judgment in financial reporting and in structuring transactions to alter financial reports to either mislead some
1
Please refer to Beneish (2001) for an overview of other Earnings Management definitions suggested by existing literature
3
stakeholders about the underlying economic performance of the company or to influence contractual outcomes that depend on reported accounting numbers.”
One first key note is that there is a distinction between accruals based earnings management (referred to above as the use of judgment “in financial reporting”) and real earnings management (“structuring transactions” in order to bring about desired reported outcomes). Existing literature has found evidence of real earnings management – i.e. manipulating a company’s real activity, for example by offering discounts in order to expand sales. As this paper is focused on earnings management in commercial banks during CEO turnovers, we believe that real earnings management is less relevant. This is due to the fact that commercial banks’ main activity – lending – is typically more complex as it involves several steps before the finalization of the lending transaction (client analysis, loan credit approval etc.), which are done by several bank departments. As lending is driven by clients’ specific funding needs, which also include specific timelines, it is more difficult to deliberately alter the exact timing of finalizing the lending transaction, i.e. – it would be difficult for an incoming CEO to purposely alter the timing of these transactions. And as importantly – as commercial banks typically have large, granular portfolios, they are less prone to being influenced by a reduced number of transactions. As such, in line with the existing literature on earnings management done by banks, our focus is on accruals-based earnings management, linked to accounting choice and the extent to which managers alter reported earnings for their own benefit, through unexpected accruals.
Incentives for accrual based earnings management
Healy and Wahlen (1999) perform a broad literature review of (accruals based) earnings management and identify three main incentives for managers, in general: (i) capital market motivations, (ii) regulatory motivations and (iii) contracting motivations. These continue to represent key areas of research for non-financial companies and have also been mirrored in the existing earnings management literature focused on banks. This literature is extensive and in the following paragraphs we broadly present the results of the main areas of research on earnings management done for commercial companies and of papers which are linked to or served as the basis for studies focused on earnings management done by banks, which will then be further detailed in Section 2.
4
(i) Capital market expectations and valuation
Existing literature has found evidence of unexpected accruals being recorded in relation to capital market transactions, via income reducing unexpected accruals before a management buyout (Perry and Williams 1994) or income increasing unexpected accruals being recorded before initial public offers (Teoh, Welch, and Wong 1998) or before stock-financed acquisitions (Erickson and Wang 1998). Such studies provide evidence indicating that, under specific circumstances, managers may be more prone to engage in earnings management.
This stream of literature has also identified a different rationale for companies to engage in earnings management: to meet analysts’ forecasts. Burgstahler and Eames (1998) found evidence of income increasing unexpected accruals undertaken in order for companies not to report earnings below forecasted levels. In relation to this, one other note-worthy result is that of Hirst and Hopkins (1998). In their experiment aimed at testing the conditions under which analysts are more likely to identify strategic timing of realized gains on investment securities, they found that a clear disclosure of the companies’ income components resulted in more accurate valuations than footnote disclosures. This result could suggest that increased transparency in financial disclosures reduces the scope for earnings management. One question raised by several papers is the extent to which earnings management is perceived by investors. In their literature overview, Healy and Wahlen (1999, p. 367) note that “Several recent studies indicate that there are situations in which investors do not see through earnings management. In other cases, notably in the banking and property-casualty industries, it appears that investors do see through earnings management. One explanation for these apparently conflicting findings is that, as a result of regulation, investors in banking and insurance firms have access to extensive disclosures that are closely related to the key accruals.'' As will be noted in Section 2, heightened regulation and the degree of financial disclosure represent factors identified by the literature as deterring banks from engaging in earnings management or at least making it less likely for bank CEOs to engage in earnings management, as compared to their non-financial companies’ counterparts.
(ii) Antitrust or other government regulation
Studies focused on commercial companies have found evidence that companies engage in earnings management as a response to various anti-trust and government regulations: Watts and
5
Zimmerman (1978) found that managers of companies exposed to anti-trust investigations are more prone to engaging in income decreasing earnings management in order to seem less profitable. Similar evidence was found by Key (1997) for cable television companies involved in Congressional hearings focused on whether the industry should be deregulated. Jones (1991) found evidence suggesting that companies requesting import relief decrease reported income during the year of application.
Earnings management as a result of government regulation has also received significant attention from papers focused on banks’ earnings management – particularly with respect to capital requirements - the results of which will be broadly outlined in Section 2.
(iii) Contracts written in terms of accounting numbers
Researchers also looked at whether the various contracts into which companies enter may also provide incentives for earnings management. Such studies find evidence of companies engaging in income-increasing practices in order to avoid borrowing contracts covenant violations or reduce the likelihood of future breaches (DeFond and Jiambalvo (1994) and Sweeney (1994)). A stream of literature more relevant to our paper looks into management compensation contracts and finds that managers are more prone to defer reported income when they will not be able to meet their bonus plans, or when they have already reached their maximum bonus (Guidry et al. 1999). Other studies, such as Healy (1985) and Holthausen et al. (1995), find that companies which cap bonuses are more likely to record income deferring accruals when the caps are reached than comparable companies which do not have caps in place. Management compensation contracts and income deferring behavior are also closely linked to what has been referred to in accounting literature as "big bath accounting” or “earnings baths”.
Section 1.2. CEO Turnover and “earnings baths”
“Earnings baths” refers to the phenomena of incoming managers reducing reported income during the initial part of their tenure and has established itself as an important explanation for earnings management by incoming CEOs, starting with the work of Moore (1973). Moore (1973) found that companies which had newly appointed CEOs exhibited a greater incidence of
income-6
reducing accounting changes than other companies and that the new CEOs were much more “conservative” in asset-valuation and engaged in income-reducing policies. Such “earnings baths” are done with the view to blame the previous managers for the firm’s current performance, while at the same time reducing the likelihood that such charges will be made on future periods’ income. Given that performance is usually not measured in absolute terms, but as compared to a benchmark (Kahneman and Tversky (1979)), incoming managers can thus lower their own performance benchmark. Further evidence in favor of “earnings baths” by incoming CEOs was found by subsequent studies from Strong and Meyer (1987) and Elliot and Shaw (1988) who found a significant association between senior management turnover and large, income-reducing discretionary write-offs for their samples of US companies. Cotter et al. (1998) identified the same association in Australian firms.
“Earnings baths” were not found to be inherent to management turnover. In studies focusing on decreases in total accruals (as opposed to those examining specific accruals/ write-offs), Pourciau (1993) found a significant decrease in total accruals to be reported only by companies which experienced non-routine CEO turnovers, while Murphy and Zimmerman (1993) found the same relationship to hold for non-routine CEO changes, but only for below-median performing companies. In his study focused on the largest 100 listed Australian companies, Wells (2002) found an insignificant relationship between income decreasing total accruals and CEO turnovers (for both routine and routine turnovers), though identified a weak relationship between non-routine CEO changes and earnings management via extraordinary items and abnormal charges. Another key question raised by the literature and highly relevant to this paper is what happens to these discretionary accruals following the year of the CEO turnover. If abnormal income decreasing accruals were recorded to create a buffer income for new CEOs, such accruals would then be reversed during subsequent periods. Pourciau (1993) found evidence of income-increasing earnings management in the first year following the CEO turnover and similar results were found by Godfrey et al. (2003), using data for Australian companies. At the same time, Wells (2002) and Murphy and Zimmerman (1993) also predict income-increasing earnings management in the first year following the CEO appointment, but find continued significant income-reducing accruals during the period. These results have been interpreted as mixed, but we believe that following initial income-reducing earnings management, the discretionary
7
accruals do not necessarily need to be reversed in the first year immediate after the CEO appointment, but can be hold on to as reserves to be used in subsequent periods. While unfortunately not addressing it in this paper, we note that the existing literature does not analyze the determinants and timing of subsequent discretionary accruals reversals.
Section 2 Earnings Management by banks
The need to separate the earnings management literature focused on banks from the general earnings management literature stems from the fact that banks “have fundamentally different accrual processes that are not likely to be captured well by total accrual models”, which are typically used in earnings management research (Peasnell, Pope and Young, 2000, p. 318). For this reason, studies focused on non-financial companies exclude banks from their samples. Apart from the differences in the accrual process, as also noted above, there are other significant differences which set apart banks from commercial companies: banks face heightened regulatory scrutiny and extensive reporting and disclosure requirements. Even more importantly, the levels of banks’ main accruals: loan loss provisions are crucial for ensuring their financial health. For these reasons, the extent of earnings management, for which evidence has been found for industrial companies, may differ for banks and financial companies in general and the literature on earnings management done by banks and other financial institutions has evolved into a separate stream.
Such literature is also extensive and dates back to the eighties when initial studies like that of Ma (1988) found that there is a disconnect between LLP levels and the quality of banks’ loan portfolios and that management tended to increase these accruals when operating income was high and reduce them when income was low. Similarly, Greenwalt and Sinkey (1988), after controlling for the quality of banks’ portfolios and economic contexts, concluded that loan loss provisions were used to smooth income.
While initial papers considered total LLP as the dependent variable, subsequent studies split loan loss provisions into non-discretionary LLPs, required to cover expected losses, and discretionary LLPs, the remaining amount. A similar approach is also used in this paper. Following the dichotomization of LLPs, some studies identified evidence in favor of LLPs being used for
8
earnings management (Collins et al. (1995), Bhat (1996), and Hasan and Hunter (1999)), whereas others (Wetmore and Brick (1994), Beatty, Chamberlain, and Magliolo (1995), and Ahmed et al. (1999)) found the opposite. In a broad literature review on earnings management by banks, which included the above mentioned studies, Anandarajan et al. (2005, p. 60) concluded that “overall, the results are conflicting. Thus, one cannot come to a definitive conclusion regarding the role of LLPs as a tool for earnings management”.
Nonetheless, subsequent studies continued to find a broad array of evidence in favor of banks using LLPs as an instrument for earnings management and have started to focus on factors influencing income smoothing through loan loss provisions. Shen and Chih (2005) used earnings benchmark tests to document that most banks in their sample managed their earnings. They also showed that stronger investor protection and greater transparency in accounting disclosure reduce a bank’s incentives to manage earnings. Fonseca and González (2008) focus on factors influencing income smoothing through loan loss provisions and find that income smoothing is lower in jurisdictions with greater bank regulation and supervision. Cornett et al (2009) also find evidence of earnings smoothing through LLPs, and that some corporate governance mechanisms (e.g., board independence) constrain earnings management whereas others (e.g., CEO pay-for-performance) induce it. Altamuro and Beatty (2010) found that mandated internal control requirements increased loan-loss provision validity – i.e. reduced earnings management. Kanagaretnam et al. (2010) found that auditor reputation constrain income-increasing earnings management. In a more recent study, Balboa, Espinosa and Rubia (2013) identified non-linear patterns between LLPs and banks’ income suggesting that the type of earnings management employed by bank managers differs as a result of the magnitude of earnings. While their results rejected the hypothesis of an overall predominance of earnings management, they contained evidence in support of banks engaging in income-smoothing for larger earnings and also identified evidence in favor of the following alternative earnings management strategies: (1) ‘‘big-bath’’ – income reducing for negative earnings, (2) income increasing for positive earnings and (3) ‘‘cookie-jar’’ accounting: income decreasing – provisions used to smooth earnings when these are positive and substantial. Such non-linear patterns can affect the results from standard analyses and may explain mixed results previously identified in the existing literature.
9
Section 2.2. Bank CEO Turnover and “earnings baths”
As mentioned at the beginning of this paper, since we started investigating this topic, two papers focused on earnings management done by banks surrounding CEO turnovers have recently been released. Sarkar, Subramanian and Tantri (2014) look at Indian public sector banks over the period 2001-2013 and analyze the exogenous variation generated by age-based CEO retirement policies in Indian public sector firms. They find evidence in support of the view that bank CEO turnover leads to earnings management and of the “big bath” hypothesis, where the incoming CEOs reduce reported earnings at the beginning of their tenures in order to lower their benchmark and subsequently reflect improved performance.
The second paper, by Bornemann, Kick, Pfingsten and Schertler (2014), uses data for a sample of German savings banks over the period 1993-2012, and also finds evidence in favor earnings baths. The paper distinguishes between turnovers where the succeeding CEOs are from inside the banks and those with CEOs coming from outside the bank – an approach which, as the authors note, is rather similar to a routine versus non-routine dichotomization of CEO turnovers mainly as most routine bank CEO turnovers result in an insider taking the office. The authors also find evidence in favor of “big baths” and their results indicate that outside CEOs take larger earnings baths than insiders, which they account as possibly resulting from social ties between incoming insider CEOs and the previous CEOs or to the fact that insider CEOs most likely occupied significant executive positions and a deterioration of the bank’s financial condition as reflected by increased accruals could also negatively reflect on their previous activity and skills. An important contribution that this paper brings is to show that the results are robust when controlling for the level of existing risk provisions – i.e. to show that the abnormal/ discretionary provisions are not driven by shortages in provisioning. This approach – of separating banks into above and below median levels of provisioning - is also used in this paper.
Section 2.3. Note regarding discretionary changes in LLPs for capital management purposes
Before presenting our hypotheses, we briefly note an alternative explanation for loan loss provisions manipulation, which has received significant attention in the literature. This is capital management - the use of LLPs in order for banks to remain compliant with minimum capital adequacy ratios. As noted previously, one stream of literature which stemmed from that focused
10
on earnings management in relation to government regulation, investigated whether banks also use LLPs in order to favorably alter their capital adequacy ratios. This is due to the fact that in the United States, prior to 1989, loan loss reserves were directly included in the formula used for calculating capital adequacy ratios as part of the numerator. The results for studies investigating the issue using pre-1989 were found to be conflicting. Collins et al. (1995) found that banks made use of other items such as loan charge-offs rather than loan loss provisions for capital management. Still, several studies had found that LLPs recorded abnormal variations for the purpose of managing capital (Beatty et al., 1995; Moyer, 1990; Scholes, Wilson, & Wolfson, 1990).
But, the change in capital adequacy regulations in 1989 eliminated loan loss provisions from the numerator (though indirectly LLPs still had a reduced influence) and implicitly decreased incentives for banks to use these as a tool for managing capital. For the purpose of our paper we note the conclusion reached by Anandarajan et al. (2005) following a review of the related literature looking at post 1989 data: “After 1989, the studies are conclusive. There is no significant association between LLPs and capital management”. Given that our sample analyzes post 1989 data (due to availability of related data in Execucomp our data starts with 1992) – we assert that manipulations of LLPs / abnormal LLPs are not be due to capital management and as such, in line with more recent literature, we do not include capital ratios in the modeling of discretionary/ abnormal LLPs.
Section 3. Hypotheses:
The first hypothesis which we need to confirm in order to test whether incoming CEOs do have a propensity to engage in earnings baths is the following:
H1: In the period of a bank CEO change (T0), earnings management is used to decrease reported income (through the increase of loan loss provisions).
In line with existing literature, we expect that if the new CEO is appointed as a result of a ‘routine change’, the extent for earnings baths may be limited as he will basically be continuing the work done by his predecessor and the turnover will most likely translate into ‘business as usual’ for the bank. As such, an extension of this first hypothesis will also be tested:
11
H2: In the period of a CEO change (T0), CEOs appointed as part of a non-routine change undertake more income decreasing earnings management than CEOs whose appointment is routine.
Two of the papers discussed above distinguished between ‘routine’ and ‘nonroutine’ CEO changes and report results contingent on this distinction (Pourciau, 1993; Wells, 2002). The identification of ‘non-routine’ changes in these papers requires subjective assessment of company announcements and media reports of the circumstances surrounding CEO change. Of particular importance to these papers is whether these sources consistently described the departure of the outgoing CEO as the result of a ‘retirement’ and if the incoming CEO is an insider or outsider. The treatment of cases for which there is disagreement across sources with regard to the circumstances surrounding the CEO change necessarily requires subjective judgment.
A confirmation of H2 might indicate that there have been financial problems with the company which led both to the non-routine change and to the abnormal increase in provisions – i.e. that poor economic performance has been the driver of both the CEO turnover and of the unexpected increase in provisions. If so – we would expect that the unexpected provisions were needed and that they will not be reversed in subsequent periods. As such, the following hypothesis will be tested (in line with Wells (2002)):
H3: In the periods subsequent to a CEO change (interval between T+1 and T+5), earnings management increases reported income. (i.e. the unexpected provisions will be reversed by the end of the new CEO’s term, within the next five years – for those CEOs who are still in office during this period)
H4: In the periods subsequent to a CEO change (interval between T+1 and T+5), earnings management increases reported income more for CEOs appointed as part of a non-routine change than for CEOs whose appointment is routine.
Note: Whereas Wells (2002) checks whether the earnings management subsequently increases reported income in T+1 or T+2, we are uncertain as to the specific timing of the unexpected provisions will be reversed (and this may vary from bank to bank based on specific
12
circumstances). To rephrase – while we do expect that following the CEO appointment (during which abnormal income decreasing loan loss provisions will be reported) the overall continuing trend will be of an income increase, we are uncertain as to whether it will manifest in a clear trend from year to year.
Section 4. Model and data
4.1 Estimating the extent of earnings management
As mentioned previously, due to differences in the accrual processes between financial institutions and non-financial companies, the total accruals models typically used in earnings management literature would not be appropriate for commercial banks. In this paper, we use a proxy for abnormal loan loss provisions based on Kanagaretnam et al., 2010, which in turn is based on prior banking research on loan loss provisions (Wahlen, 1994; Kanagaretnam et al., 2004):
LLP = λ0 + λ1BEGGLA+ λ2LCO+ λ3CHLOANS+ λ4LOANS+ λ5NPL+ λ6 (LOAN
CATEGORIES) + λ7 (YEAR CONTROLS) + Ɛ (1)
Where:
LLP - provisions for loan losses deflated by beginning total assets BEGLLA - beginning loan loss allowance deflated by beginning total assets LCO - net loan charge-offs deflated by beginning total assets
CHLOANS - change in total loans outstanding deflated by beginning total assets LOANS - total loans outstanding deflated by beginning total assets
NPL - nonperforming loans deflated by beginning total assets
LOAN CATEGORIES - commercial loans (COMM), consumer loans (CONS),
hire-purchase/lease (LEASE), mortgages (MORT), loans to banks (BK) and other loans (OTH), all deflated by beginning total assets
13
Ɛ - the error term from equation 1; the residuals of this regression are commonly referred to as the abnormal or discretionary component of LLP, short ALLP.
We use Kanagaretnam’s model as we find it to be most consistent with existing literature. Other studies (such as Balboa et. al. 2013) also include Tier 1 Capital in the above equation/ regression. Based on Ahmed et al. (1999), the 1990 change in USD bank capital adequacy regulations substantially limited the use of loan loss reserves as regulatory capital in two ways. First, loan loss reserves do not count as part of Tier 1 or primary capital under the new regulations. Second, loan loss reserves only counted as part of Total capital up to 1.25% of risk weighted assets. This diminished the costs for banks associated with managing earnings through loan loss provisions and as such we no longer expect capital ratios to represent a significant factor in determining loan loss provisions, given that our sample will cover the period 1992-2014.
The above model accounts for the differences between banks and companies with regards to underlying variables, but does not account for differences in the accrual process per se and we could not identify any references in the existing literature with respect to these. For commercial banks, loan loss provisions are largely the sum of provisions for individually assessed loans and collectively impaired loans. As the name suggests, provisions for individually assessed loans are done on a loan by loan basis and the review of a bank’s portfolio can take up to a few months prior to the reporting date. Based on this, it could be argued that it could be more difficult for an incoming CEO to alter these analyses immediately before the reporting date, than it would be for an incoming CEO of a non-financial company which could arguably only need to modify a few large accruals. We try to account for such a potential difference by accounting for the moment when the new bank CEOs were appointed – i.e. for CEOs appointed during the last three months of the year we considered the CEO turnover (T0) to have taken place during the next financial year. The results were broadly in line with those to be reported in the paper. The rationale for this may partially lie in the fact that, similarly to large accruals available to commercial companies, banks also have a third LLP component: allocations based on general economic conditions and risk factors which can easily be used to significantly influence the level of provisioning. We would have expected that, given that such reserves need to be disclosed, this would reduce the level to which they can and would be used by CEOs and would have expected to have ‘stronger’
14
results by accounting for the moment of appointment, but this was not the case. Given that this falls slightly beyond the scope of this paper – which is to determine if bank CEOs also engage in earnings baths – for the moment we emphasize that the differences between the accrual processes (and not just underlying variables) of financial institutions and commercial companies can constitute an area for further research.
Section 4.2. Sample selection and description
We used EXECUCOMP to obtain information regarding CEO turnover for the period 1992-2014. We selected the information with respect to commercial banks – SIC Code 6020 – for which we identified information regarding 304 commercial bank CEOs. 86 of these refered to CEOs appointed prior to 1 Jan 1992 and were excluded from our sample and for another 76 we could not identify information with respect to the previous CEO and implicitly with respect to the reason for the turnover. Three other CEOs had tenures which spanned for less than one year and for the purpose of having one CEO turnover type per company financial year we also eliminated these and considered the information for the subsequent CEOs that followed them. We were left with a series of 139 turnovers.
We then obtained the financial information needed to run the regression (loan loss provisions, loan charge-offs etc.) from COMPUSTAT – similarly for commercial banks, SIC Code 6020, for the period 1992 – 2014. After merging the two databases we were left with a sample of 85 commercial bank CEO turnovers between 1992 and 2014, reported by 51 commercial banks. Of these, two companies reported four CEO changes during the period, seven reported three CEO changes, fourteen reported two CEO changes and with a remaining of 28 companies each reporting one CEO change. A distribution of the CEO changes across years is provided in Table 1, indicating that the CEO changes are relatively evenly spread over the period – with a minimum of zero in 1992 and a maximum of ten turnovers in 2007.
15 Table 1 - Summary of CEO changes
Financial year ended No. % No. %
1992 0 0.0% 0 0.0% 1993 3 3.5% 3 0.7% 1994 4 4.7% 7 1.6% 1995 7 8.2% 14 3.2% 1996 3 3.5% 16 3.6% 1997 2 2.4% 15 3.4% 1998 2 2.4% 15 3.4% 1999 2 2.4% 14 3.2% 2000 7 8.2% 20 4.5% 2001 8 9.4% 26 5.9% 2002 3 3.5% 28 6.3% 2003 3 3.5% 30 6.8% 2004 2 2.4% 27 6.1% 2005 4 4.7% 25 5.7% 2006 4 4.7% 24 5.4% 2007 10 11.8% 26 5.9% 2008 4 4.7% 25 5.7% 2009 3 3.5% 26 5.9% 2010 7 8.2% 26 5.9% 2011 4 4.7% 24 5.4% 2012 1 1.2% 24 5.4% 2013 2 2.4% 26 5.9% Total 85 100.0% 441 100.0%
CEO changes Firm/Years
Execucomp contains information on the date when the new CEO was appointed, the departure of the previous CEO (both from his function and from the company altogether) as well as information on the reason of departure - ‘REASON LEFT COMPANY’ with one of the following entries: 1) resigned 2) retired 3) deceased and 4) unknown. When we set out to do the analysis, we initially planned to consider ‘routine’ changes as those in which the previous CEO left due to retirement and ‘non-routine’ as those leaving due to resignation or due to decease and to either exclude the ‘unknown’ turnovers or assess them based on media reports of the circumstances surrounding CEO changes. The issue is that the information contained in the above field does not refer to the reasons for which the CEOs left their function, but to the reasons
16
they left the company – and these two do not necessarily conincide. Due to the fact that by selecting a random sample of ten CEO turnovers we have identified discrepancies between the reasons as per Execucomp and as per media reports, we decided, similarly to Wells 2002 for non-financial companies, to base the routine versus non-routine turnovers entirely on media reports/ review of news articles which we performed for all of our final sample. As noted by Pourciau (1993) and Wells (2002), classifying CEO turnovers as routine versus non-routine is not necessarily clear cut and requires judgement. We used Pourciau (1993)’s description of routine vs non-routine changes as a starting point2. Similarly to the intuition that we subsequently identified in Bornemann, Kick, Pfingsten and Schertler (2014) the key drivers for non-routine changes were whether the incumbent CEO resigned (or deceased – i.e. did not retire) and whether the new CEO was an insider or an outsider. In order to identify the first piece of information we used search engines including alternative search strings (CEO Name + “resigned” OR “retired” OR “deceased” OR “fired” OR “Quit” etc.) and reviewed the first pages(s) of query results (with a focus on non-retirements as sometimes companies would release in their press releases that a certain CEO had retired though media articles noted that he was fired or had resigned). In order to see whether the new CEO was an insider or an outsider – if no information was included in the press release announcing his tenure (e.g. “has been working with the company since..”, “previously a CRO” etc), we searched for the CEO’s CV or again used search engines to try to derive the information. Information for tenures beginning in the 1990s were more difficult to find, but the number of these cases was relatively smaller. A summary description of CEO changes based on information from these articles is provided in Table 2. For the purpose of this paper we did not discriminate between ‘Fired’ and ‘Resigned’ as per Wells 2002 due the fact that sometimes it was difficult to distinguish between the two and because for the purpose of our paper both were considered as non-routine changes.
2 Pourciau (1993, p 318): “‘Routine’ executive changes are those in which the company structures an orderly, well-planned
process of turnover, as described by Vancil (1987). Routine changes typically conclude with the retirement of the top executive, who often remains a member of the board of directors. As discussed in section 2, the structure of the routine executive change reduces the incentives and opportunities for earnings management.
In ‘nonroutine’ executive change, the company is not in a position to plan an orderly process of executive succession, due to inadequate time and or insufficient opportunity to select and groom a successor CEO with the support of the incumbent. Nonroutine changes include most resignations, both voluntary and nonvoluntary. It is suggested here that the environment surrounding nonroutine executive changes provides incentives and opportunities for earnings management.
Nonroutine executive changes are often unplanned, making it difficult for the directors and stockholders to structure the turnover in a way that minimizes the opportunities and incentives for earnings management.”
17 Table 2 - Nature of CEO changes
Description Number of changes %
Retirement 69 81.2%
Resignation/ Fired 14 16.5%
Deceased 2 2.4%
Total 85 100.0%
As per above table, 69 changes have been classified as retirements, the most commonly cited reason for CEO, in turn subsequently considered as routine changes accounting for 81.2% of the total turnovers. Non-routine changes are resignations or cases in which the previous CEO has been fired or has deceased, a total of 16 non-routine changes, 18.8% of the total.
As mentioned previously, financial statements information was obtained from Compustat. While most variables were available for all firm years corresponding to the initially identified 139 turnovers, the main variable that reduced our sample size to 85 turnovers was Loan Charge Offs, which we computed as Loan Loss Written Off minus Recoveries. Descriptive statistics for banks reporting CEO changes are presented in Table 3. Panel A reports unscaled values, while Panel B provides corresponding information scaled by lagged total assest. Poor performance is evident, with a mean pretax income divided by beginning total assets (which should be thus higher than the return on average assets) of only 1.35%.
18
Table 3
Descriptive statistics on sample firms in year of CEO change Panel A
Unscaled values - All amounts in $million
Mean Min Q1 Med Q3 Max
Total Assets 131,540 4,139 18,407 47,946 150,374 2,264,909
Pretax income 1,292 (3,678) 173 702 1,443 12,215 Loan Loss Provions 941 (60) 22 115 700 28,435 Beginning loan loss allowances 1,387 31 148 405 992 37,200 Non-performing loans 1,170 4 96 242 1,139 32,664 Loan Charge-Offs 922 (3) 34 99 573 34,334 Total Loans 65,318 2,580 9,906 24,443 73,069 940,440 Commercial 21,237 409 3,436 9,265 22,540 225,655 Consumer 17,997 - 790 3,420 13,910 372,369 Leases 2,010 - - 271 1,558 21,942 Mortgages 22,056 - 3,942 9,856 22,535 317,884 Loans to banks 403 - - - - 11,100 Other loans 1,473 - - - 482 25,093
Change in total loans 3,752 (67,242) (1) 856 4,523 62,304
Panel B
Values scaled by lagged total assets
Mean
Standard
deviation Min Q1 Med Q3 Max
Total Assets 1.1022 0.2024 0.8381 1.0136 1.0573 1.1227 2.2091 Pretax income 0.0135 0.0164 (0.0677) 0.0097 0.0164 0.0210 0.0344 Loan Loss Provions (LLP) 0.0070 0.0101 (0.0011) 0.0018 0.0035 0.0081 0.0631 Beginning loan loss allowances (BEGLLA) 0.0109 0.0059 0.0005 0.0078 0.0101 0.0132 0.0390 Non-performing loans (NPL) 0.0106 0.0138 0.0001 0.0037 0.0076 0.0113 0.1032 Loan Charge-Offs 0.0057 0.0079 (0.0001) 0.0017 0.0033 0.0063 0.0460 Total Loans 0.6476 0.2200 0.0529 0.5700 0.6770 0.7333 1.4746 Commercial (COMM) 0.2168 0.1062 0.0186 0.1426 0.2061 0.2851 0.5259 Consumer (CONS) 0.1196 0.0826 - 0.0586 0.1108 0.1701 0.4872 Leases (LEASE) 0.0195 0.0273 - - 0.0079 0.0298 0.1387 Mortgages (MORT) 0.2749 0.1598 - 0.1430 0.2707 0.4091 0.6225 Loans to banks (BK) 0.0019 0.0068 - - - - 0.0362 Other loans (OTH) 0.0138 0.0326 - - - 0.0107 0.2111 Change in total loans (CHLOANS) 0.0515 0.1215 (0.1299) (0.0000) 0.0276 0.0766 0.7824
19 Table 4
Regression in estimating abnormal loan loss provisions
Variable Coeff Estimate
Intercept λ0 -0.000 (0.37) BEGGLA λ1 -0.575 (11.47)*** LCO λ2 1.213 (17.26)*** CHLOANS λ3 -0.004 (2.19)** LOANS λ4 0.049 (2.20)** NPL λ5 0.129 (3.70)*** COMM λ6 -0.043 (1.95) CONS λ7 -0.039 (1.77) LEASE λ8 -0.023 (1.06) MORT λ9 -0.044 (1.96) BK λ10 -0.033 (0.94) OTH λ011 -0.042 (1.91)
Year dummies included without being reported
N 441
R2 0.94
We report the results for the following stage one regression model:
LLP = λ0 + λ1BEGGLA+ λ2LCO+ λ3CHLOANS+ λ4LOANS+ λ5NPL+ λ6YEAR DUMMIES + (LOAN CATEGORIES) + Ɛ (ALLP)
Where LLP is the provision for loan losses deflated by beginning total assets; BEGLLA - beginning loan loss allowance deflated by beginning total assets; LCO - net loan charge-offs deflated by beginning total assets; CHLOANS - change in total loans outstanding deflated by beginning total assets; LOANS - total loans outstanding deflated by beginning total assets; NPL - nonperforming loans deflated by beginning total assets; LOAN CATEGORIES - commercial loans (COMM), consumer loans (CONS), hire-purchase/lease (LEASE), mortgages (MORT), loans to banks (BK) and other loans (OTH), all deflated by beginning total assets; e- the residuals from Eq. (1) are the abnormal or discretionary component of LLP, referred to as ALLP and Year controls.
20
Table 4 reports the results of the first-stage regression for estimating abnormal LLP. Consistent with prior studies (e.g., Kanagaretnam et al., 2009), BEGLLA is negatively associated with LLP since a higher initial loan loss allowance will require a lower LLP in the current period. As expected, LCO, LOANS and NPL are positively associated with LLP, at 1% or 5% levels of significance, consistent with the evidence reported in Kanagaretnam et al. 2004 and 2010. The residuals from Eq. (2) represent the abnormal component of LLP. While Kanagaretnam 2010 finds different levels of significance also for the loan categories (COMM, CONS, LEASE, MORT, BK, OTH) – in our sample these are not statistically significant. This may be due to the smaller sample size (441 bank-years in our sample compared to 7680 in Kanagaretnam’s sample), sample which exhibits a high degree of corellation, particularly between LLPs and LCOs, driving up the R2 to a high value of 94%.
Section 5. Evidence of earnings management
Similarly to Wells (2002), a precursor to any evaluation of earnings management is an assessment of whether the estimates of income manipulation are economically significant. Table 5 (Panel A) reports evidence of the economic significance of abnormal loan loss provisions scaled by lagged total assets. The results show that unexpected accruals are generally small relative to total assets. For example, unexpected accruals as a percentage of lagged total assets range from a mean of −0.054 per cent (in the fifth year after the CEO change) to 0.062 per cent (in the year of the CEO change). However, while unexpected accruals may be small relative to lagged total assets, the mean value of pretax income over lagged assets is also reduced: of 1.35 per cent (Table 3, Panel B). Nonetheless, in absolute terms, abnormal loan loss provisions represent a mean of 27.4 per cent of pretax income (11.7 per cent weighted average), confirming that they are substantial relative to the income reported by the banks.
There is evidence that for CEO changes classified as non-routine, mean abnormal loan loss provisions are more economically significant (and income increasing) in periods T0 – T2 than for the corresponding partitions of routine CEO changes. Over these periods mean unexpected accruals are 0.123 per cent, 0.088 per cent and 0.065 per cent of total assets respectively for the non-routine CEO changes, while for routine CEO changes the corresponding values are 0.048 per cent, 0.017 per cent and 0.004 per cent respectively. This would indicate the presence of clear income decreasing abnormal loan loss provisions during the year of the CEO change – i.e.
21
an “earnings bath”, but as well a persistence of income increasing abnormal loan loss provisions, though with a declining trend, up to and including the second year following the appointment of the new CEO.
Table 5
Tests of earnings management through loan loss provisions Panel A
Raw abnormal loan loss provisions over lagged total assets
Full sample n Routine n Non-routine n
T0 Mean 0.0006217 85 0.0004799 69 0.0012335 16 Median 0.0004098 0.0002237 0.0007165 T1 Mean 0.0002984 77 0.0001695 63 0.0008786 14 Median 0.0001306 0.0001004 0.0008675 T2 Mean 0.0001603 72 0.0000425 58 0.0006481 14 Median 0.0000279 -0.0000193 0.0004458 T3 Mean -0.0002927 61 -0.0002808 50 -0.0003469 11 Median -0.0004222 -0.0004814 0.0000539 T4 Mean -0.0004687 43 -0.0001999 38 -0.0025113 5 Median -0.0004661 -0.0003500 -0.0026121 T5 Mean -0.0005441 32 -0.0003711 28 -0.0017555 4 Median -0.0005311 -0.0004528 -0.0017453
The different trends between routine and non-routine ALLPs may have to do with information asymmetries and may indicate that incoming outsider bank CEOs manage loan loss provisions also for reasons other than earnings baths or income smoothing – such as due to risk averseness or being more cautiouss. If this is the case and some bank CEOs also want to set aside provisions to make sure that NPLs are adequately covered, due to the fact that, following a non-routine change, the incoming CEO did not have as much time to prepare for his new position (as described by Pourciau and reflected in the turnovers in our sample), he may be less familiar with the bank and its portfolio and could realize that in T1 or in T2, after having become more familiar with the bank and its portfolio, believes that additional provisions still need to be set aside. On the other hand, for CEOs which have had time to prepare for their new position and are more familiar with the bank, after initially having set up buffer provisions, it is easier to plan a gradual reduction of income decreasing/ buffer ALLPs. This type of information assymetry may also partially explain why the coefficients for ALLPs for non-routine changes are considerably higher than for routine changes during the CEO turnover year and the two following years.
22
In Table 5, Panel B, we standardize the values of raw abnormal loan loss provisions over lagged total assets3. In the period of the CEO change (T0), both routine and non-routine CEO changes report income decreasing abnormal loan loss provisions (means of 0.244 and 0.627 respectively). While these results are in the direction to that hypothesised, they are not significant and the difference between the Routine and Non-routine turnovers samples is not significant at conventional levels (at the 10 per cent level). These results are in line with those of Wells (2002) for tests of earnings management through accruals for non-financial companies.
Table 5
Tests of earnings management through loan loss provisions Panel B
Standardized abnormal loan loss provisions over lagged total assets
Sign Full sample St dev n Routine St dev n Sign Non-routine St dev n Difference
T0 Mean H1+ 0.316073 0.002534 85 0.243951 0.002678 69 H2+ 0.627099 0.001714 16 -0.383147 Median H1+ 0.208332 0.113724 H2+ 0.364250 -0.250527 T1 Mean H3\ 0.151704 0.001920 77 0.086158 0.001961 63 H4\ 0.446661 0.001665 14 -0.360504 Median H3\ 0.066394 0.051041 H4\ 0.440989 -0.389948 T2 Mean H3\ 0.081481 0.001919 72 0.021622 0.001765 58 H4\ 0.329470 0.002478 14 -0.307849 Median H3\ 0.014192 -0.009803 H4\ 0.226608 -0.236412 T3 Mean H3\ -0.148813 0.001732 61 -0.142754 0.001703 50 H4\ -0.176354 0.001946 11 0.033600 Median H3\ -0.214635 -0.244705 H4\ 0.027402 -0.272107 T4 Mean H3\ -0.238255 0.001655 43 -0.101623 0.001396 38 H4\ -1.276658 0.002195 5 1.175035 Median H3\ -0.236952 -0.177905 H4\ -1.327923 1.150018 T5 Mean H3\ -0.276627 0.001088 32 -0.188651 0.000943 28 H4\ -0.892462 0.001413 4 0.703812 Median H3\ -0.269997 -0.230166 H4\ -0.887239 0.657073
Z tests (for differences in medians) and t tests (for differences in means); the “\” next to H3 and H4 indicates that we expect a decreasing trend for abnormal loan loss provisions following the turnover year, but that we are uncertain as to the sign and to the exact timing then after which reversing abnormal loan loss provisions will become negative; *** significant at the 1% level; **significant at the 5% level; *significant at the 10% level;
Up to and including the second year following the appointment of the new CEO – apart from the median value for routine changes in T2, abnormal loan loss provisions are reported to be income decreasing, but again, the differences between the two samples are not statistically significant. In year four and five the results would indicate that abnormal loan loss provisions have a clear
3
ALLPs/lagTA have been standardized by using the following formula for the sample: Standardized ALLP/lagTAi = (ALLPi/lagTAi – sample avg(ALLP/lagTAi))/sample stdev(ALLP/lagTAi)
23
income increasing trend, as hypothesized (H3) and that these income increasning ALLPs are more pronounced for non-routine changes (as per H4). The results would indicate that, in our sample, there seems to be a clear trend for recording and then reversing abnormal loan loss provisions: income decreasing ALLPs are recorded during the CEO appointment year as well as during the following two years and these then start to be reversed during year 3. While we did expect to identify evidence of income decreasing abnormal loan loss provisions during the year of the CEO turnover (T0) – i.e. creating an income buffer to be reversed at a later time and that these would be more significant for the non-routine changes (resignations, fired and deceased), we did not expect to identify such a clear trend of a gradual switch to income increasing abnormal loan loss provision – towards which most bank CEOs seem to gradually move to starting with their third year of tenure. The reason why we did not expect such a clear trend to be is because while we assess that there may be a general propensity to incur income decreasing abnormal loan loss provisions during the year of the CEO turnover and revert these subsequently, we expected that the strategy based on which CEOs do reverse these abnormal loan loss provisions may differ according to specific banks circumstances which may vary between banks (such as changes in profitability targets). We find partial support for the anticipated differences in strategies for reversing ALLPs in the results for the non-routine turnovers for year three: mean standardized abnormal loan loss provisions over lagged total assets become income increasing (-0.176), while the median remains income decreasing (0.027); A difference between median and mean values is also recorded for routine turnovers in year two: for this sample while median standardized ALLPs are income increasing (-0.010) , the mean is income decreasing (0.022).
Overall, such results seem to reject the hypothesis that poor performance is what drives both the CEO turnovers and the abnormal loan loss provisions. Firstly – even if a bank would be operating poorly – this would not automatically imply that its provisioning levels are out of line with its fundamentals and would need to be corrected via abnormal loan loss provisions. The presence of such a disconnect between provisioning and fundamentals could indeed indicate that the previous CEO did not record sufficient loan loss provisions in order to artificially improve the bank’s results. But if this would be the case, it would be sufficient to record higher abnormal loan loss provisions during the CEO turnover year to correct for such disconnects and to realign loan loss provisions to their fundamentals. As such, we believe that the above results would
24
indicate that indeed, new bank CEOs are inclined to take a big bath shortly after taking office, but consistent with the idea that an immediate reversal would be too obvious, continue to create reserves, at a lower pace, to be reversed further down the line. The clear income increasing abnormal loan loss provisions indentified in years 4 and 5 above seem to strongly support such a view, but again we note that the results are not statistically significant.
Finally, in order verify whether performance is driving abnormal loan loss provisions, similarly to Bornemann, Kick, Pfingsten and Schertler (2014), we introduce control variables, which contain measures of the bank's nondiscretionary income, credit risk and risk provisioning and run the following stage 2 regression:
ALLP = β0 + β1D_TURN+ β2POSNDI+ β3NEGNDI+ β4CHLOANS+ β5NPLt-1+ β6CHNPL +
β7RPROVt-1+ β8LNTA+ β9ALLPt-1+ Ɛ (ALLP) (2), where
ALLP - abnormal or discretionary component of LLP (the residuals regression 1)
D TURN - a binary variable equaling 1 if a CEO turnover occurs in bank i in year t, and 0 otherwise.
POSNDI - equals non-discretionary income (net income less loan loss provisions) (NEGNDI) if positive (negative) as a percentage of end-of-year total assets (TA)
CHLOANS - the change in the volume of the overall loan portfolio from year t-1 to t as a percentage of TA
NPLt-1 - is the volume of non-performing loans as a percentage of TA, year t-1
CHNPL - is the change in the volume of non-performing loans from year t - 1 to t as a percentage of TA at year t – 1
RPROVt-1 - denotes the aggregate stock of risk provisions (i. e. the loan loss allowance) as a percentage of TA at year t-1
LNTA - the natural logarithm of TA.
In line with Bornemann, Kick, Pfingsten and Schertler (2014), we separated non-discretionary income into positive (POSDI) and negative (NEGNI) in order to capture bank’s income
25
smoothing behaviour, as adjustements to positive non-discretionary income may be different from the one to negative non-discretionary income. Bornemann, Kick, Pfingsten and Schertler (2014, p 17): “Given highly positive non-discretionary income, managers incur high amounts of discretionary expenses. When non-discretionary income is low or even negative, managers incur low amounts of discretionary expenses or even realize discretionary gains.”
Table 6
Panel A - CEO Turnover effects - loan loss allowances subsamples
Variable Coeff Exp All High Low
D_TURN β1 + 0.001 0.000 0.002 (3.07) *** (0.35) (2.60) ** POSNDI β2 + 0.037 0.001 0.067 (1.90) (0.04) (1.98) ** NEGNDI β3 + 0.033 0.116 0.005 (0.62) (1.55) (0.06) CHLOANS β4 + 0.000 0.002 -0.004 (0.24) (2.46) ** (1.29) NPLt-1 β5 ? -0.089 -0.193 -0.062 (4.47) *** (3.46) *** (1.83) CHNPL β6 + 0.015 -0.141 0.027 (0.80) (2.02) ** (0.94) RPROVt-1 β7 - -0.185 -0.159 -0.204 (5.76) *** (4.17) *** (3.57) *** LNTA β8 ? 0.000 -0.000 0.000 (0.52) (0.38) (1.48) ALLPt-1 β9 + 0.179 0.200 0.103 (3.40) *** (2.92) *** (1.25)
Year dummies included without being reported
N 389 216 173
R2 0.23 0.26 0.17
We report the results for the following stage two regression model:
ALLP = β0 + β1D_TURN+ β2POSNDI+ β3NEGNDI+ β4CHLOANS+ β5NPLt-1+
β6CHNPL + β7RPROVt-1+ β8LNTA+ β9ALLPt-1+ Ɛ (ALLP)
t-Statistics are reported in Paranthesis; *** significant at the 1% level; **significant at the 5% level; *significant at the 10% level; High (Low) considers only observations with an above-median (below-median) stock of risk provisions.
26
We use the same approach as the authors and, based on a bank's aggregate stock of risk provisions as a percentage of its non-performing loans, we first estimate the baseline model for the full sample and then for the subsamples of banks with high and low risk provisions.
In Table 6, we present the results, estimated for the full sample (marked as All ) as well as for subsamples of banks with high and low stock risk provisions separately. Our results differ significantly from those of Bornemann, Kick, Pfingsten and Schertler (2014): firstly while the coefficients for POSNDI and NEGNDI are positive, as expected – meaning that an additional unit of non-discretionary income would lead to additional discretionary expenses, they are not significant at any relevance level. As such, we do not find evidence for income smoothing behavior by the banks in our sample. What our results indicate as driving abnormal loan loss provisions is the level of prior year loan loss allowances – RPROVt-1, which is significant at the 1% level for the enitre sample (All), as well as for the high and low provision sample: the higher the level of loan loss allowances as a percentage of total assets in year t-1, the lower the level of abnormal loan loss provisions in year t. While the CEO turnover (D_TURN) is found to be positively associated with higher ALLPs for the full sample, significant at the 1% level, this relationship is not statistically significant for banks with above median loan loss allowance levels and is relevant only at the 5% level for banks with below median loan loss allowance levels. This result could suggest that past loan loss allowance levels may be driving both the current abnormal loan loss provisions as well as CEO turnovers – with turnovers not influencing the level of abnormal provisions for above median provisioned banks.
Past higher levels of nonperforming loans (NPLt-1) seem to be associated with lower levels of current abnormal loan loss provisions – significant at the 1% level for both the full sample and for banks with above median allowances, but this relationship is not significant for the “Low” allowances sample. Surprisingly, the change in non-performing levels (CHNPL) leads to a decrease in abnormal loan loss provisions for above median provisioned banks, while an increase in the loan portfolio (CHLOANS) does lead to an increase of abnormal loan loss provisions for these banks, as expected – but similarly to CHNPL – the relationship is significant only at the 5% level and only for banks with above median allowances. Past levels of abnormal loan loss provisions (ALLPt-1) also lead to higher current levels of ALLPs suggesting that there is a
27
decreasing, lagged effect of abnormal loan loss provisions, in line with the results presented in table 5, particularly for routine turnovers.
In table 6 Panel B, we present the results from re-running the above equation 2, this time for the entire sample and for subsamples based on the type of CEO turnover. Surprisingly, while the CEO turnover is significant at the 1% level for the Routine changes subsample, it is not significant for non-routine turnovers. Similarly, the results do not find any evidence of income smoothing for the two subsamples – routine and non-routine changes, with prior period loan loss allowances (RPROV t-1) driving ALLPs, together with the changes and prior levels of NPLs (CHNPL and NPLt-1). While the results do find evidence in favor of the CEO turnover leading to abnormall loan loss provisions for the overall sample, as anticipated by our first hypothesis, these results are not statistically significant for the non-routine turnovers and thus does not find significant evidence to support our second hypothesis that in the period of a CEO change (T0), CEOs appointed as part of a non-routine change undertake more income decreasing earnings management than CEOs whose appointment is routine.
28 Table 6
Panel B - CEO Turnover effects - CEO turnover type subsamples
Variable Coeff Exp All Routine Non-routine
D_TURN β1 + 0.001 0.001 0.001 (3.07) *** (2.07) ** (1.85) POSNDI β2 + 0.037 -0.17 -0.072 (1.90) (0.80) (1.43) NEGNDI β3 + 0.033 -0.046 -0.195 (0.62) (0.74) (1.20) CHLOANS β4 + 0.000 0.000 0.005 (0.24) (0.21) (1.31) NPLt-1 β5 ? -0.089 -0.057 -0.149 (4.47) *** (2.30) ** (3.58) *** CHNPL β6 + 0.015 0.061 -0.124 (0.80) (2.67) *** (3.43) *** RPROVt-1 β7 - -0.185 -0.144 -0.282 (5.76) *** (3.90) *** (2.78) *** LNTA β8 ? 0.000 0.000 0.000 (0.52) (0.16) (1.32) ALLPt-1 β9 + 0.179 0.152 0.220 (3.40) *** (2.64) *** (1.38)
Year dummies included without being reported
N 389 330 59
R2 0.23 0.20 0.75
We report the results for the following stage two regression model:
ALLP = β0 + β1D_TURN+ β2POSNDI+ β3NEGNDI+ β4CHLOANS+ β5NPLt-1+
β6CHNPL + β7RPROVt-1+ β8LNTA+ β9ALLPt-1+ Ɛ (ALLP)
t-Statistics are reported in Paranthesis; ** significant at the 5% level; *significant at the 10% level; Routine (Non-routine) considers only observations for routine (non-(Non-routine) CEO changes.
29
Section 6. Concluding remarks
We set out to investigate whether incoming bank CEOs take “an earnings bath” – i.e. incur significant income decreasing behaviour during their first year of tenure. Contrary to other studies, while we did find evidence in favor of abnormal loan loss provisions being recorded during the year of CEO turnovers, our results does not support the theory that these are booked for the purpose of managing income – i.e. taking “an earnings bath” by recording abnormally high, income decreasing loan loss provisions during and after the CEO turnover in order to blame the previous CEO and create a buffer income to be subsequently reversed in future periods. While our univariate results found differences between the routine and non-routine CEO turnover samples, which would support the “big baths” hypothesis, these results were not statistically significant. Furthermore, when controlling for other factors – particularly for the level of provisioning – the results were at best, not conclusive. One potential explanation that our paper offers is that bank CEOs are intrinsically risk adverse and tend to raise provisions upon taking office and that the key determinant to the extent that they record abnormal loan loss provisions is the level of existing loan loss allowances. We would not say that our study contradicts the two recent studies focused on bank CEO turnovers which have identified evidence supporting the ‘income smoothing’ and ‘earnings baths’ hypotheses, but simply that our results do not fully support it. The different results could be due to differences in samples or country regulations, but there are several statistical issues surrounding our results which we must admit to – such as the small final sample size, abnormally high R2 or the potential impact of outliers in our data.
Nonetheless, if bank CEOs do increase loan loss provisions as a result of being risk adverse (and not necessarily to create a buffer income), this could also partially explain why outsider CEOs or CEOs appointed as a result of a non-routine turnover may record higher abnormal loan loss provisions than insider CEOs / appointed as a result of a routine turnover: given that the first are less familiar with the bank than the latter, their incentive for booking higher abnormal loan loss provisions is also higher, given that they are not as able to fully commensurate the risks borne by the bank’s loan portfolio. We believe that the issue of “big baths” surrounding bank CEO turnovers deserves further attention from researchers: in particular the rationale behind large abnormal accruals after a new CEO takes office and, as importantly, the determinants and timing
30
List of references
Ahmed, A.S., Takeda, C. and S. Thomas, 1999, Bank loan loss provisions: a reexamination of capital management, earnings management and signaling effects, Journal of Accounting and Economics, vol. 28, pp. 1-25.
Altamuro, J. and Beatty, A., 2010, How does internal control regulation affect financial reporting, Journal of Accounting and Economics, vol. 49 pp. 58–74
Anandarajan, A., Hasan, I. and A. Lozano-Vivas, 2005, Loan loss provision decisions: An empirical analysis of the Spanish depository institutions, Journal of International Accounting, Auditing and Taxation, vol. 14, pp. 55–77.
Balboa, M. Espinosa, G. and Rubia, A., 2013, Nonlinear dynamics in discretionary accruals: An analysis of bank loan-loss provisions, Journal of Banking & Finance, no. 37
Beatty, A., Chamberlain, S.L. and J. Magliolo, 1995, Managing Financial Reports of Commercial Banks: The Influence of Taxes, Regulatory Capital, and Earnings, Journal of Accounting Research, vol. 33, no. 2, pp. 231-261.
Beneish M.D, (2001) "Earnings management: a perspective", Managerial Finance, Vol. 27, No. 12, pp.3 – 17
Bornemann, S., Kick, Thomas K., Pfingsten, A. and Schertler, A., Earnings Baths by Bank CEOs During Turnovers (June 18, 2014). Deutsche Bundesbank Discussion Paper, No. 05/2014.
Burgstahler, D. and I. Dichev. 1997. Earnings management to avoid earnings decreases and losses. Journal of Accounting and Economics, no. 24, pp. 99-126.
Bhat, V.N., 1996, Banks and income smoothing: an empirical analysis, Applied Financial Economics, vol. 6, pp. 505–510.
31
Collins, J.A., Shackelford, D.A. and J.M. Wahlen, 1995, Bank Differences in the Coordination of Regulatory Capital, Earnings and Taxes, Journal of Accounting Research, vol. 33, no. 2, pp. 263-291.
Cornett, M.M., McNutt, J.J. and H. Tehranian, 2009, Corporate governance and earnings management at large U.S. bank holding companies, Journal of Corporate Finance vol. 15
Cotter, J., D. Stokes, and A. Wyatt, 1998, An analysis of factors influencing asset writedowns, Accounting and Finance, vol. 38, pp. 157–180.
DeFond, M. L. and J. Jiambalvo. 1994. Debt covenant effects and the manipulation of accruals. Journal of Accounting and Economics vol. 17, pp. 145—176.
Elliott. J. and W. Shaw. 1988. Write-offs as accounting procedures to manage perceptions, Journal of Accounting Research, Suppl.. vol. 26.
Erickson, M. and S-w. Wang. 1999. Earnings management by acquiring firms in stock for stock mergers. Journal of Accounting and Economics vol. 27, pp. 149-176.
Fonseca, A. R., Gonzalez, F., 2008, Cross-country determinants of bank income smoothing by managing loan-loss provisions, Journal of Banking & Finance vol. 32, pp. 217–228
Godfrey, J., P. Mather, and A. Ramsay, 2003, Earnings and impression management in financial reports: the case of CEO changes, ABACUS 39, pp. 95–123.
Greenawalt, M.B. and J.F. Sinkey Jr., 1988, Bank Loan-Loss Provisions and the Income-Smoothing Hypothesis: An Empirical Analysis, 1976-1984, Journal of Financial Services Research, vol. 1, pp. 301-318.
Guidry, F., A. Leone, and S. Rock. 1999. Earnings-based bonus plans and earnings management by business unit managers. Journal of Accounting and Economics vol. 26, pp. 113-142.
Hasan, I. and W.C. Hunter, 1999, Income-smoothing in the depository institutions: An empirical investigation, Advances in Quantitative Analysis of Finance and Accounting, vol. 7, pp. 1-16.
