• No results found

Hypotheses on behavioural factors affecting levels of loan loss provisions

While the previous section identified a variety of estimation philosophies tailored to the needs of investors and regulators, a bank’s reported loan-loss provisions are ultimately largely under its managers’ control, and managers are likely to use any available discretion to attain their own goals. In this respect, research has suggested a range of motivations and, in line with the

regard to its application to financial instruments without observable market prices arising from active

markets (see for example BCBS, 2004a, p. 5).

categorization suggested in Lobo & Yang (2001), they will be discussed under the following general headings (subsections 2.4.1 to 2.4.4): income smoothing, capital regulation, signalling and, finally, tax considerations. In essence, the literature looking at such behavioural aspects which determine level of loan loss provisions has mostly been researched with data on US banks and there is a relative paucity of studies using data from other banking systems. The final

subsection 2.4.5 nevertheless reviews some of these articles.

2.4.1 Income Smoothing

In common accounting textbooks like Wild, Bernstein & Subramanyam (2001) income smoothing is described as a common form of earnings management, that is, managers may decrease or increase reported earnings to reduce volatility. The method used to smooth income involves not reporting part of earnings in profitable years through ‘hidden reserves’ or ‘earnings banks’, with these stored earnings reported in less profitable years. Given that loan loss

provisions are generally the largest accrual expense item in a bank’s P&L statement, they are thus likely to play a significant role in a manager’s income smoothing strategy.

While Buckmaster (1992; 1997) documents the existence of income smoothing literature as far back as 1898, the hypothesis that banks will use loan-loss provisions for income smoothing purposes was first explored for the US by Schreiner (1981)8 and in a more formal way by

Greenawalt & Sinkey (1988) again with US data covering a period of 1976-1984. Over the years, considerable empirical evidence supporting this hypothesis has been accumulated. Scholes, Wilson, & Wolfson (1990, p. 646) show that loan-loss provisions are used in income smoothing in conjunction with unrealized securities gains. Collins & Shackelford (1995) document income smoothing but just for profitable banks while contrary to these findings, Bhat (1996) characterize an income smoothing bank “as a small one with high risk loans and poor financial conditions”.

8 As reported in Buckmaster (2001, p.179, 234), Schreiner (1981, p.123) concludes that “in general, banks do not appear to use the loan loss provision as a device to smooth net income”.

Looking at bank capitalization, Niswander & Swanson (2000) find income smoothing solely for banks above a certain critical capital adequacy threshold.

Some studies, however, fail to find evidence of income smoothing through loan loss provisions. Examples are Wetmore & Brick (1994), Beatty, Chamberlain, & Magliolo (1995, 5.1.2, p. 254), and Ahmed, Takeda, & Thomas (1999).

Most of above studies do not attempt an in-depth interpretation of the patterns of income smoothing. Explanatory hypotheses are however explored in Kanagaretnam, Lobo, & Yang (2000) who argue that the level of current performance relative to the industry median is a key determinant of managers’ decisions to smooth income. In a later paper, the same lead authors (Kanagaretnam, Lobo, & Mathieu, 2003) explore bank specific factors that explain income smoothing and conclude that the need to obtain external financing and the managers’ job security concerns9 appear to be a significant driver of income smoothing behaviour.

No such motives are identified in an earlier more technical explanation by Kim &

Santomero (1993) who see income smoothing as a consequence of Bayesian models used by banks when forecasting loan losses. These models update projected loan losses as a function of new information obtained from the new audit and the historical variance of loan loss rates over the bank's previous history. A series of good years will thus mean that provisions get smaller. On the other hand, successive bad outcomes reduce the bank's prior belief in the historical

distribution. The corresponding provisions on average get larger because the bank becomes more sure that it is drawing outcomes from a distribution with higher average loss rates and


9 Job security concerns are analysed in a framework developed by Fudenberg & Tirole (1995) which suggests that when current performance is poor, relative to other banks, managers have an

incentive to shift future earnings into the current period to reduce the chance of dismissal or interference.

Alternatively, when future relative performance is expected to be poor, managers have an incentive to shift current earnings to the future to reduce the likelihood of poor future performance.

2.4.2 Capital Regulation

The capital management incentive has its root in the fact that regulators monitor the banks accounting based capital ratios which are affected by provisioning decisions. Basel I capital adequacy rules, for instance, allow loan loss provisions, subject to certain upper limits, to be counted as a component of regulatory capital (BCBS, 1988, items 18-21, p. 5-6). Capital management through loan loss provisions is also addressed later in this chapter in what we classify as literature with a macro prudential and bank regulatory focus in section 2.5. At this stage, this subsection will review the development of the capital management argument through time.

The first paper that explicitly posits capital management through loan loss provisions is Moyer (1990, 3.1, p. 129-131). Her research looks at data before the introduction of the Basel regime with US regulations based on ‘primary capital adequacy ratios’ which allowed banks to prop up their ratios by inflating loan loss provisions. Accordingly, she tests and confirms the capital management hypothesis that predicts that the capital ratio is negatively related to discretionary loan loss provisions, i.e. the lower the capital ratio the greater the incentive to report higher provisions. Other researchers who confirmed her results also for sample periods before the introduction of the Basel capital adequacy regime are Collins, Shackelford & Whalen (1995) and Beatty, Chamberlain & Magliolo (1995).

A later study by Kim & Kross (1998) studies the impact of the introduction of the Basel I rules in 1989 which brought about limitations on the use of provisions as part of a bank’s regulatory capital. The authors confirmed that the incentive of low capitalized banks to report provisions at high pre-1989 levels was indeed reduced. Kim & Kross’ findings were confirmed by Ahmed, Takeda, & Thomas (1999) who conclude that the capital management motivation is the most important aspect in setting discretionary provisions, much more important than earnings management or signalling.

A more recent article by Luengnaruemitchai & Wilcox (2004) on capital management by US and Japanese banks looks at patterns in the use of discretionary provisions and charge-offs through time, in particular through what the authors call ‘troubled’ times. They firstly argue that in difficult times banking regulators are more lenient in the enforcement of capital requirements for fear of systemic repercussion (credit crunch, widespread bank failures) which in turn allows the banks to exercise more discretion, i.e. report lower provisions and charge offs when the banking system is in a troubled state. The authors secondly hypothesize that since supervisors are more likely to close ‘atypical’ banks, one should observe clustering of reporting behaviour when financial institutions seek ‘safety in similarity’. As to the results, Luengnaruemitchai & Wilcox find some evidence for their hypotheses when capital ratios were low in the banking crisis of the late 1980s but no systematic relation of capital ratios among peer banks in the generally healthier times of the late 1990s when these ratios were generally higher. Such behavioural patterns, the authors conclude, would help mitigate procyclical effects of Basel II as they are sometimes feared in the literature (e.g. in Borio, Furfine, & Lowe, 2001).

2.4.3 Signalling

The need for signalling arises when managers, who possess information indicating that bank values are higher than those assessed by the market, wish to have market values revised upward. Because of an adverse selection problem as described in Akerlof (1970) and the

accepted wisdom that well-informed agents can improve their market outcome by signalling their private information to poorly informed agents (Milgrom, 1981; Spence, 1973), bank managers could likewise employ signalling tools to communicate concerns about stock undervaluation resulting from information asymmetry.

Well known is the so-called dividend signalling theory which has given rise to an extensive literature going back to seminal articles by Bhattacharya (1979), Miller & Rock (1985) and John

& Williams (1985). This theory explains excess returns observed following announcements by firms of an increase in dividend.

Less famous are papers that explore potential signalling effects of discretionary accounting items like loan loss provisions. One potential hypothesis is that a bank increases the loan loss provision to signal that it is strong enough to absorb future potential losses. Research by Beaver, Eger, Ryan, & Wolfson (1989, p. 169 and Table 2, p. 170) suggests that an increase in loan loss provisions is indeed interpreted as ‘good news’ in that management indicate the "the earnings power of the bank to be sufficiently strong that it can withstand a ‘hit to earnings’ in the form of additional loan loss provisions." In some respects, signalling may be related to income

smoothing or also capital management activities of bank managers. If a bank engages in earnings management, this might well be used as a signalling device. In this respect, Wahlen (1994) provides evidence that bank managers increase the discretionary component of unexpected loan loss provisions when future cash flow prospects improve. In his view, too, increased unexpected loan loss provisions could be interpreted a ‘good news’ consistent with above results.

An obvious way to test whether the provisioning is interpreted as ‘good’, respectively

‘bad’ news is to observe the reaction by the capital markets as was done in Grammatikos &

Saunders (1990) in a case study for Citicorp Group which, at the time, had a very substantial LDC loan exposure. Beaver & Engel (1996) follow a more sophisticated approach. They

hypothesize that the capital market perceives the stock of loan loss provisions to be comprised of two components: a nondiscretionary component which is negatively priced and a discretionary component whose incremental pricing coefficient is positive. In their study they model the non-discretionary portion of provisions as a function of subsequent loss experience and other factors while the remaining provision stock then becomes the discretionary portion. When regressing the market value of the bank with these two components, Beaver & Engel indeed find the predicted coefficients. This means market participants clearly interpret discretionary provisions as positive signals, and so bid up the share price.

Other hypotheses tested with regard to what makes provisioning good or bad news can be found in Liu, Ryan, & Wahlen (1997) who posit that increased loan loss provisions are good news only for banks with apparent loan default risk problems based on prior information. The data then confirm that loan loss provisions are good news only for such ‘at risk’ banks and bad news for ’not at risk’ banks.

One recent example of researching loan loss provisions in the light of the signalling theory are Kanagaretnam, Lobo, & Yang (2005). They look at bank specific factors that determine signalling with loan loss provisions and find that signalling differs across banks based upon the degree of information asymmetry. It varies negatively with bank size and positively with earnings variability, future investment opportunities, and degree of income smoothing.

2.4.4 Taxation Management

The use of loan loss provisions in tax minimization strategies is comparably less

researched but there is generally broad agreement among researchers that tax considerations do have an impact on levels of reported loan-loss provisions. In the US (see for example in Collins, Shackelford, & Wahlen, 1995, pp. 268-270) and many other countries, tax authorities will only accept specific provision or actual loan write-offs as a tax deductible item. General provisions, on the other hand, do not reduce taxable income in many instances. The typical research design will thus test whether the marginal corporate tax rate affects the level of loan write-offs while discretionary non-deductible loan loss provisions should not be affected.

Earlier studies that find evidence for tax optimization strategies being followed by banks are Scholes, Wilson, & Wolfson (1990) and Collins & Shackelford (1995). More recently, Niswander & Swanson (2000) document such behaviour for well capitalized banks, i.e. where there is a true discretion in setting loan loss reserves.

No evidence of tax planning is however found in (Beatty, Chamberlain, & Magliolo, 1995, 5.1.3, p. 254). The authors attribute the result to their “poor choice” of crude proxies for

marginal tax rates.

2.4.5 Studies on behavioural aspects using non-US data

The previous four subsections have reviewed hypotheses mostly developed for the US banking market which provides large and homogeneous data samples over long time periods.

This is important since such behavioural effects might be quite weak. Even though banks in other countries may have different accounting rules, regulation and supervision, and possibly different incentives, only few researchers have looked at behavioural factors driving the discretionary loss provisions outside the US.

One example is Hasan & Wall (2004) who test for signs of earnings management firstly with a sample of international banks and then especially with a smaller sample of Japanese and Canadian banks using Bankscope data from 1993 to 2000. They model levels of total loan loss provisions as a function of proxies for non-discretionary and discretionary components. The proxies for the discretionary component are the capital ratio at the beginning of the period, to test for potential signs of capital management, and the (pre-provision, pre-tax) return on assets (earnings ratio), to test for earnings management in general. Only effects of earnings management are found to be significant while the evidence for capital management is

inconclusive. The coefficients on the earnings ratio are found to be positive and significantly lower for U.S. banks than for the non-U.S. banks. Hasan & Wall argue that these differences may reflect differences in the financial market benefits of managing earnings or in the flexibility that management has to manage earnings (i.e., the cost of managing earnings). They state that ”in at least some cases, the results may also reflect the banks’ determination to artificially boost reported net income, such as by realizing capital gains, in those periods where they need to increase their bad debt provisions.”

Data on the 50 largest EU banks are used by Valckx (2004) which, among mainly macroeconomic aspects, also looks at behavioural issues of loan loss provisioning. The time period covered is 1997 to 2001 only but is extended back to 1988 for a smaller panel of 21 EU banks. The author of this working paper concludes that ”the data are broadly consistent with the

hypothesis that income smoothing takes place in the EU but the findings are more mixed with respect to capital management”.