Ebook Commercial Bank Loan Loss Recoveries
The stability of the banking sector is of major importance for economic outcomes. Banks form the backbone of modern economies and instability in the banking sector can pose problems to the economic system as a whole. Credit losses, or more generally, asset quality problems, have repeatedly been identified as a key trigger of bank failures, e.g. Graham and Horner (1988), Caprio and Klingebiel (1996). Accordingly, much research effort has gone into developing methods for assessing credit risk both at a systemic and bank-specific level.
Two major components determine the extent of a credit loss suffered: first, the probability of a default (PD) and, second, the loss given default (LGD), which equals one minus the recovery rate in the event of default. Most credit risk literature has focussed on estimating PD; much less attention has been devoted to estimating characteristics of LGD. We address this lack of research by analyzing the determinants of LGD (or, more specifically, the recovery rate) using a comprehensive sample of Australian banks.
The paucity of research on LGD has been changing in recent years as the new Basel Capital Accord (Basel II) enables banks to employ their own proprietary risk models to calculate capital adequacy. Schuermann (2004) and Altman (2006) provide a comprehensive review of these LGD studies, many of which are authored by rating agencies (e.g. Gupton et al., 2000; Verde, 2003; Keisman, 2004; Moody's Investors Service, 2008).
These studies analyse recovery patterns of defaulted corporate bonds; likewise there is research into the recovery of syndicated corporate loans (Asarnow and Marker, 1995; Emery et al., 2004). Results for both corporate bonds and syndicated loans are similar with seniority of the claim and collateralization (secured vs. unsecured lending) impacting the rate of recovery. Given that syndicated loans typically represent senior lending, have stricter covenants and are also mostly secured in the samples studied, their median level of recovery is generally found to be much higher than unsecured bonds.
A related stream of research investigates the correlation between PD and LGD. The consensus of these studies is that there is a negative correlation between the two variables with low (high) recovery rates in times of high (low) defaults (Frye, 2000; Altman et al., 2003; Hu and Perraudin, 2006).
The research into characteristics of LGD is for the most part based on price data of defaulted bonds and in some cases of traded bank loans because such data is readily available. Accordingly, these analyses are not based on comprehensive samples of non-traded loans which represent the bulk of assets for many banks and where there is no market value of distressed debt shortly after default. Nor are they based on realized recoveries but rather assume that market prices are an efficient reflection of the present value of all future recoveries on these claims.
Research on actual recoveries exists for traditional, non-traded bank loans but is typically limited to smaller proprietary bank specific data samples (e.g Asarnow and Edwards, 1995; Eales and Bosworth, 1998; Araten et al., 2004; Miu and Ozdemir, 2006). In the context of the Basel II implementation, banks have started building their LGD databases but these are not in the public domain. Moreover, anecdotal information indicates that these data typically cover just a few years, and data have not been collected in a consistent and standardized fashion through time.
This paper presents an alternative method for gaining insights into the dynamics of recovery rates for distressed bank lending over longer periods of time, i.e. through economic cycles. Since the late 1980s, banks of most developed countries have reported on the level of loans and other assets considered impaired from a credit risk perspective. Moreover, banks not only report the gross book value of these assets but typically also their expected realizable value thus providing a point in time estimate of overall recovery rates of their total distressed asset portfolio.
These values can be interpreted as a proxy for expected recoveries by bank management just as the distressed price based methods represent market expected recovery values of corporate bonds. The main benefit of the method is that recovery estimates are for a representative composition of bank distressed credit exposures rather than the specific bond portfolios of the traditional bond LGD literature. It also enables analysis over longer periods and mirrors outcomes for the whole system, not just a single bank.
We apply the method to explore determinants of recovery rates for a comprehensive panel dataset of 18 Australian banks for the period 1989 to 2005. We consider two groups of explanatory variables. One includes idiosyncratic factors specific to the bank and its risk profile, the other aggregate macroeconomic drivers with systemic impact on asset recoveries. We find that the macroeconomic factors act in the opposite direction for bank loans relative to their effect on corporate bonds found in the traditional distressed price based literature. This latter result implies that banks may use reported recovery rates as a method to smooth their earnings over the cycle. Our estimation utilises three different methods to test robustness of key results.
The paper proceeds as follows. Section 2 describes the data used in the study, including some background on the Australian banking system. Section 3 formulates the hypotheses, and presents the test methodology, empirical results and robustness tests; section 4 concludes.
Posted in :
