Recovery Determinants of Distressed Banks
Regulators, Market Discipline, or the Environment?
  • 1 https://isni.org/isni/0000000404811396, International Monetary Fund
  • | 2 https://isni.org/isni/0000000404811396, International Monetary Fund

Contributor Notes

Based on detailed regulatory intervention data among German banks during 1994-2008, we test if supervisory measures affect the likelihood and the timing of bank recovery. Severe regulatory measures increase both the likelihood of recovery and its duration while weak measures are insignificant. With the benefit of hindsight, we exclude banks that eventually exit the market due to restructuring mergers. Our results remain intact, thus providing no evidence of "bad" bank selection for intervention purposes on the side of regulators. More transparent publication requirements of public incorporation that indicate more exposure to market discipline are barely or not at all significant. Increasing earnings and cleaning credit portfolios are consistently of importance to increase recovery likelihood, whereas earnings growth accelerates the timing of recovery. Macroeconomic conditions also matter for bank recovery. Hence, concerted micro- and macro-prudential policies are key to facilitate distressed bank recovery.

Abstract

Based on detailed regulatory intervention data among German banks during 1994-2008, we test if supervisory measures affect the likelihood and the timing of bank recovery. Severe regulatory measures increase both the likelihood of recovery and its duration while weak measures are insignificant. With the benefit of hindsight, we exclude banks that eventually exit the market due to restructuring mergers. Our results remain intact, thus providing no evidence of "bad" bank selection for intervention purposes on the side of regulators. More transparent publication requirements of public incorporation that indicate more exposure to market discipline are barely or not at all significant. Increasing earnings and cleaning credit portfolios are consistently of importance to increase recovery likelihood, whereas earnings growth accelerates the timing of recovery. Macroeconomic conditions also matter for bank recovery. Hence, concerted micro- and macro-prudential policies are key to facilitate distressed bank recovery.

I. Introduction

The International Monetary Fund (IMF) estimates that aggregate write-downs due to the financial crisis borne by banks could reach the staggering amount of $2.6 trillion by 2010 (IMF, 2009b). The deterioration of asset prices paired with only slowly ceasing gridlock in money markets implies liquidity shortages, and ultimately insolvency threats, for many banks in the global financial system. Deleveraging stressed banks requires according to the IMF, among other things, to swiftly re-capitalize stressed banks. Hoshi and Kashyap (2008) add as a lesson from the Japanese banking crisis in the 1990s that re-capitalization also has to be large enough to be effective and the IMF estimates that around $375 billion of capital injections are required for euro area banks to ensure gross equity ratios of 4 percent.2At the same time, an active role for the supervisor in terms of assessing the viability of impaired institutions as well as imposing corrective actions is advocated as one of the key elements of a global bank stress resolution strategy.

In light of these recommendations, we ask in this paper whether active supervisory intervention affects the repayment (or recovery) from such capital support schemes. Numerous studies analyze recovery determinants of non-financial firms. Acharya and others (2007) emphasize that industry-wide distress reduces recovered present values of failed firms. This is particularly relevant for the banking industry where systemic distress is more likely due to higher interdependency of financial institutions, for instance through interbank exposures (Upper and Worms, 2004; Liedorp and van Lelyveld, 2006). The lack of evidence on the determinants of bank recovery in general and the role of supervisory measures in particular are therefore surprising. While the determinants of bank distress are well understood (see, for example, Lane and others, 1986; Cole and Gunther, 1995; Estrella and others, 2000; DeYoung, 2003), only Dahl and Spivey (1995) analyze determinants of bank recovery and its duration albeit without assessing the role of supervisory interventions. The present paper seeks to fill this gap. We investigate to what extent severe regulatory measures that restrict the scope of the managerial choices of banks having received capital assistance and unscheduled supervisory audits influence recovery from distress. We test simultaneously whether regulatory covenants increase the likelihood of repaying capital injections and if the duration of repayment is reduced once regulators impose certain restrictions on the banking firm.

In Europe, banking supervision is nationally fragmented (IMF, 2009a). One approach is to provide distressed banks with capital support so as to permit independent recovery and to avoid disruptions of confidence in the financial system that may follow from outright closure of banks (Garcia and Nieto, 2005).3 To the extent that we estimate the likelihood and the duration of repaying such capital support measures, our paper relates to literature that analyzes different modes of dealing with insolvent non-financial firms: re-organization (Chapter 11) or liquidation (Chapter 7). Bris and others (2006) find that the cost of failure resolution by means of the former are on average smaller.4 But Kalaya and others (2007) also report that the benefits of re-organization accrue primarily to distressed firms rather than acquiring firms, competitors, or the public. This might be different in banking, where the social cost of bank failures due to negative externalities, possible spillovers to other industries, and risks of contagion are presumably larger compared to non-financial firms (Dell’Ariccia and others, 2005).5

Partly because of the importance of highly leveraged financial institutions for the real economy, banks are heavily regulated and closely supervised in industrialized economies. On- and off-site monitoring of banks by supervisors generates information both for micro (Flannery and Houston, 1999; Berger and others, 2000) and macro prudential policy purposes (Peek and others, 1999). Such additional information generation compared to financial markets and other stakeholders that monitor the bank (e.g., depositors) may enable independent supervisors to neutralize inevitable moral hazard problems inherent to the provision of additional equity capital to distressed banks (see Kane, 1989; Cobos, 1989). Insider knowledge could allow supervisors to identify more precisely banks with large recovery potential and/or systemic relevance and to impose adequate constraints on the management to align behavior with shareholder and depositor interests, i.e., avoid excessive risk-taking. We coin this the “regulatory insider” hypothesis and test if additional audits and regulatory covenants levied upon banks that receive capital injections increase the likelihood of recovery.

Alternatively, the policy of (conditionally) injecting capital resembles the forbearance practice pursued in the U.S. during the S&L crisis of the 1980s.6 Numerous empirical studies document the lacking success of this strategy (see, for example, DeGennaro and others, 1993; DeGennaro and Thomson, 1996; Brinkmann and others, 1996; Guo, 1999; Gupta and Misra, 1999). Two key impediments to smooth “selfhealing” forwarded by Kane (1989, 1990) are perverted incentives to bank managers and capture of regulators. First, a flat rate deposit insurance scheme induces excessive risk taking of bank managers because of no risk-adjusted pricing of the insurance. Second, numerous academics emphasized the interest of federal regulators to camouflage the true state of thrifts and banks so as to ensure their own reputation and career prospects (see next to Kane also McKenzie and others, 1994; Cole, 1993). Instead, already Benston and others (1986) advocated prompt corrective actions (PCA) to avoid the participation of weak institutions for too long in the financial system (see also Benston, 1994; Kaufman, 1995; DeYoung, 2007).7 In essence, PCA requires regulators to swiftly liquidate a troubled bank or merge it with a healthy institution. One of the main insights from the US savings and loan crisis was thus to avoid unconditional support of distressed banks and to replace supervisory discretion with rules.8 From this follows our alternative “regulatory capture” hypothesis, which implies that the probability of repayment is negatively affected by regulatory actions.

In addition to the likelihood of repayment, we also test the effect of regulatory measures on the duration of capital support. We define recovery as the complete repayment of received capital support, since successfully restructuring the bank should enable it to generate enough (retained) earnings so as to independently reach sufficient capitalization levels. It remains a priori unclear whether the time required to reach sustainable levels of capital is positively or negatively affected by regulatory intervention. On the one hand, intervention may substitute for poor governance due to a lack of a functioning market for corporate control among the mostly non-listed small German universal banks. Active intrusions into the business of the bank, for example restrictions to take deposits or distribute profits, may then facilitate a more efficient choice of production plans, which ultimately should speed up recovery. Alternatively, measures that aim to discipline previous mismanagement and induce thorough restructuring of the bank are likely to take a considerable amount of time. Moreover, banks that have received capital injections may be exactly those that are most severely troubled. Therefore, more time might be required until portfolios are set straight and processes are re-designed so as to ensure sustainable banking business. Given this ambiguity regarding the relation between regulatory measures and recovery time, we test empirically these competing “acceleration” and “sustainability” hypotheses. Table (1) summarizes these four hypotheses.

Table 1.

Description of Hypotheses

article image

We use a unique dataset on equity injections in which German universal banks received capital injections in 473 cases from banking-sector specific insurance funds during the period 1994 to 2008.9 We distinguish weak and severe interventions issued by the German Financial Supervision Authority (Bundesanstalt fuer Finanzaufsicht, BaFin) and control for unscheduled supervisory audits, too. This information is augmented by financial account and audit report data, all of which has been collected systematically by the German Central Bank, the Bundesbank.

Our results support the “regulatory insider” and the “sustainability” hypotheses. Severe regulatory measures increase the likelihood of recovery while additional audits have no effect. Both unscheduled audits and severe regulatory measures increase the duration until capital injections are fully repaid. Regulators seem to prefer diligent, but potentially time-consuming restructuring efforts. While possibly effective tool for less severely distressed banks without capital support, weak regulatory measures, such as official warning letters by banking supervision, have no significant impact on repaying injected capital. Results are robust also after accounting for a possible selection bias of very troubled banks, to which regulators direct more (severe) measures or audits. With the benefit of hindsight, we identify most severely troubled institutions as those that were forced to merge with another bank or closed due to a moratorium without changes to our results. Market discipline, approximated by an indicator whether the bank is publicly incorporated,10 and thus subject to more stringent reporting standards, is only weakly related to recovery odds and not at all to recovery duration. This could reflect the limited role of capital markets for corporate governance in Germany. It might also reflect some crowding out of monitoring bank managers by equity and debt holders by a fairly active regulator. Finally, (regional) macroeconomic factors (interest spreads and real income) consistently affect both recovery odds and duration. This suggests that prudential supervision policies have both a macro and a microeconomic component that should both be considered.

The remainder of the paper is organized as follows. We start by describing the regulatory framework in Germany. Next, we introduce the data on capital injections and recovery timing. We then describe the split-population duration method to disentangle probabilities from duration determinants of recovery and discuss further covariates. Finally, we present and discuss estimation results before concluding with the main findings.

II. Regulatory Background of Capital Measures

The core principles issued by the Basel Committee of Banking Supervision that apply to Europe are largely in line with the PCA requirements: independence of supervisory body from the political and judicial systems, access to a broad range of supervisory measures, provision of supervisors with adequate resolution procedures, and access to accurate and timely financial information on banks’ financial condition (Nieto and Wall, 2006). However, the European regulatory environment continues to differ considerably from that of the U.S. in many respects and disclosure of distress among banks as well as resolution procedures are much more opaque.

Also, supervisory responsibilities rest with national rather than one pan-European authority (Garcia and Nieto, 2005). Even within countries, multiple institutions often participate in the bailout process, as is also the case for Germany (IMF, 2009a).

In Germany, privately-owned financial institutions are required by law to participate in a legal deposit insurance (DI) scheme. The DI scheme guarantees 90 percent of each customer’s deposits up to a maximum amount of 20,000 Euro. Voluntary DI schemes of banking associations supplement legal DI.11 Voluntarily DI schemes guarantee substantially larger deposit volumes. Both DI schemes aim to protect private persons and (in particular small) firms. In general, each scheme is based on the insurance principle. Support of distressed banks is financed by contributions from member institutions as well as re-payments from granted capital injections in previous years. Moreover, the fund receives returns from capital investments, and there are administrative expenses to cover.

Exceptions from the legal deposit insurance scheme are government-owned savings banks12 and credit cooperatives, which are protected by banking sector-specific insurance schemes. The Insurance protection by the Federal Association of Cooperative Banks (“Bundesverband der Deutschen Volksbanken und Raiffreisenbanken, BVR”) comprises two elements: (1) deposit insurance and (2) institutional warrants (“Institutssicherung”). Purpose of the deposit insurance is to save member banks (which are currently distressed or threatened to become distressed) by means of capital injections or warrants by the fund. This centralized federal fund is financed by risk-oriented annual contributions by the member banks, repayment of previous capital injections, or returns from capital investments. Second, institutional warrants (by member banks) constitute a further reaching insurance against insolvency in the cooperative banking sector. These warrants are, however, limited to a predefined maximum amount per member bank. Condition for a distressed member bank to make use of institutional warrants is an expected recovery within the next five years.

The protection by the German Savings and Giro Association (“Deutscher Sparkassen und Giroverband, DSGV”) is differently financed. The DSGV-system relies on a pre-defined, risk-oriented funds volume that is financed by direct payments into the fund (at least one third) and by the obligation for additional contributions (up to two thirds). Reductions of the fund due to bank rescues imply an obligation to immediately restock the fund, resulting in a fairly constant fund size over time. The DSGV-system comprises 11 regional funds to insure regional savings banks and one fund respectively for central savings and mortgage savings banks. In case of the rescue of a savings bank the according regional fund is initially liable but recourse with the other regional as well as both central and mortgage savings bank funds is possible, too.13

III. Capital Injections and Recovery

Table 2 provides descriptive statistics for the recovery and injection data obtained from the Bundesbank. The sample comprises up to 2,165 observations on 473 banks that received at least one capital injection during the 1994–2008 period. We define recovery as full repayment of injected capital.14

Table 2.

Capital Injections, Recovery, and Exit 1994–2008

This table presents descriptive statistics for capital injections, intervention, and exit patterns in German banking between 1994 and 2008. Injections are the number of capital preservation measures taken by banking-sector specific insurance schemes. Strong interventions are the number of active intrusions of the regulator at the going concern of the bank comprising four specific types: (i) Threat of measures pursuant to §36 of the Banking Act; (ii) Prohibition of profit distribution; (iii) Measure pursuant to §46 of the Banking Act; and (iv) Threat of a measure pursuant to §46 of the Banking Act. Weak measures comprise official letters or warnings sent by supervisory authorities. Audits are the number of unscheduled audits according to §44 of the Banking Act. Exit is the frequency of forced closures by means of moratoria by the Federal Agency for Financial Market Supervision (BaFin) or participation in restructuring mergers. Recoveries indicate the number of banks that repaid capital preservation measures in full and censored shows banks exiting the sample before recovering.

article image

Capital injections are a fairly frequently used instrument to deal with troubled banks, also in non-crisis times. Each year, between 3 and 42 of all active universal banks, which comprise commercial, savings, and cooperative banks,15 received either cash injections or lines of equity according to annual audit reports compiled by the Bundesbank, which we consider here jointly.16 From the perspective of receiving banks, the size of capital injections is substantial, too. Mean capital support measures account for approximately 83 percent of total gross equity at year end when injections occurred. However, for the industry as a whole capital support measures seem to have been geared especially to small banks since the mean share of equity of supported banks in the total industry does not exceed 2 percent.

Since our sample is dominated by the years prior to the 2008 banking crisis, recovery estimates obtained here represent an optimistic scenario for repayment behavior since sampled banks did not face further adverse industry effects emphasized by Acharya and others (2007) to reduce recovery odds further.17

On average, 62 percent of all banks that received support eventually settled their scores with respective insurance schemes during the sample period. The mean duration until recovery is 4.6 years. This corroborates earlier conjectures by, for example, BIS (1999) and Hoshi and Kashyap (2008) that turning around distressed banks takes time. Banks that did not repay by the end of 2008 or exited the sample either voluntarily or due to ordered restructuring mergers and closure are censored. By the end of our sample period, 50 banks did not yet repay received capital injections but might still do so. About 30 percent of banks ultimately exit due to restructuring mergers or closure enforced by the regulator despite capital support (see column Merger & Moratoria). We control for this group of particularly troubled banks because restructuring mergers and closures represent in fact alternative policy tools in line with the PCA philosophy to ensure smooth but swift exit of banks considered too weak to remain in the banking system. Regulatory measures might have a negative (positive) impact on recovery probabilities (duration) if especially very weak banks attract regulatory attention. Therefore, we compare the below baseline estimates with those that (i) include an indicator for banks that ultimately are subject to enforced exit and (ii) are based on a sample excluding banks censored for these two reasons. Thereby, we test (indirectly) whether regulators focus their measures on, with hindsight, very weak banks.18

Figure 1 illustrates the activity stance of German supervisors, taking 1,737 regulatory measures since 1993.19 The scope of these measures varies widely and most have an early warning character, for example official hearings or formal letters.

Figure 1.
Figure 1.

Top Ten Regulatory Measures Administered to All German Banks (1994-2008)

Citation: IMF Working Papers 2010, 027; 10.5089/9781451962499.001.A001

For the present sample of capital supported banks, a total of 440 measures were issued. Only 4.5 percent were severe interventions as shown in column four of Table 2.20 We construct one dummy variable comprising 63 weak incidences of sending standardized warning letters and another indicator that takes value one if one out of four severe measures was issued to a bank that received a capital injection:21

  1. Threat of dismissal of managers pursuant to §36 of the Banking Act.

  2. Prohibition of profit distribution.

  3. Actions pursuant to §46 of the Banking Act comprising:

    • a. issue instructions on the management of the institution’s business,

    • b. prohibit the taking of deposits or funds or securities of customers and the granting of loans,

    • prohibit proprietors and managers from carrying out their activities,

    • or limit such activities, and appoint supervisors.

  4. Threat of the former measures.

In addition, we specify an indicator if non-scheduled on-site audits were conducted. Column Audit in Table (2) highlights that this tool is used in 41 percent of the cases, corroborating that supervisors monitor supported banks actively. Since on-site audits are exactly the source of inside information available only to regulators, we specify it below as a separate covariate to predict recovery spells and likelihood. We turn next to the determinants of recovery probabilities and duration.

IV. Methodology and Covariates

A. Methodology

To predict simultaneously the likelihood of repaying received capital injections and the duration required to repay capital support measures, we use the split-population duration model suggested by Schmidt and Witte (1989). Cole and Gunther (1995) and DeYoung (2003) employed this model to predict the failure of (de novo) banks and Dahl and Spivey (1995) to predicted recovery probabilities and duration of capital support in the U.S.. Especially the latter study is close to ours but differs in an important respect since it does not test to what extent regulatory covenants influence recovery likelihood and/or timing. Given the ample evidence on failure to facilitate recovery when supporting banks without any restrictions regarding their risk-taking, this is key in our analysis. We test to what extent conditions tied to capital injections affect the recovery probability of troubled institutions.

The split population duration model distinguishes between banks that recover (i.e., repay the injected capital) and banks that do not. Let Q be an unobservable binary variable that takes value 1 for banks that recover at time t and value 0 for banks that do not recover. The model takes into account that the recovery can take place either during the sample period, t ∈ [1, T], or thereafter, t > T. The probability that bank i will eventually recover (δi) is modeled using a logistic model:

δi=exp(Xi)1+exp(Xi)(1)

where Xi are variables determining the recovery probability of bank i, and a is the vector of coefficients. Larger coefficients imply a higher likelihood of recovery. Denote f(t) and F(t) as the probability density and cumulative density functions of the time until recovery, respectively, conditional on the actual recovery taking place (Q = 1). The contribution of banks that recover during the sample period to the total density function is δif(t). The contribution of banks that do not recover during the sample period is a sum of two components: the probability of no recovery, (1 − δi), and probability that recovery will take place after the sample ends, δi(1 − F(T)). The total likelihood function can then be expressed as (Schmidt and Witte, 1989):

L=i=1N[δif(t)]Qi×[(1δi)+δiS(T)](1Qi),(2)

where N is the total number of banks that received capital injections, and S(T) = 1 − F(T) is the survival function.

We parameterize equation (2) using the log-logistic distribution. The main advantage of this distribution is that it generates a non-monotonous hazard function that first rises and then declines, which corresponds to the empirical regularities observed in banking data (Cole and Gunther, 1995). The survival and density functions of the log-logistic distribution are given by:

S(T)=1[1+(λt)p](3)
f(t)=λp(λt)(p1)[1+(λt)p]2(4)

where p and λ are parameters to be estimated. p > 0 is the parameter governing the shape of the distribution: p > 1 implies unimodal hazard function, which dispersion increases with the magnitude of the parameter, while p ≤ 1 suggests monotonically decreasing hazard. For p > 1, the mean of the distribution is proportional to the inverse of the parameter λ.22 Therefore, to model the timing of recovery for bank i as a function of corresponding bank-specific and regional covariates Xi, the following specification is used:

λi=exp(Xiβ)(5)

Positive coefficients β in equation (5) imply a longer time required for the recovery.

B. Covariates and Expectations

We distinguish three sets of covariates: governance measures including supervisory action and incorporation, bank-specific, and environmental variables capturing the (regional) macroeconomic conditions. Following earlier duration analyzes, we use the mean of each covariate over the time of the spell.23

The first panel in Table 3 shows that the time to recovery varies considerably across banks having received capital support. We construct a dummy for regulatory measures if banks were subjected to one of the four severe measures detailed in subsection 3, which applies to 4.5 percent of the sample. Unscheduled audits are conducted at 41 percent of banks that received equity injections. In line with Flannery and Houston (1999), this could indicate efforts of regulators to generate additional information by closely monitoring supported banks. Corrective actions by regulators could also reflect failure of other governance systems that are supposed to ensure prudent behavior of managers. Usually, financial markets ensure to align principal and agent’s interests. Therefore, we specify a dummy equal to one for those few banks in Germany that incorporated as private or public limited companies. While not all of these banking firm’s stocks are traded freely, stricter publication requirements should proxy for the influence of outsider share- and stakeholders to assess the recovery efforts of the bank.24

Table 3.

Summary Statistics on Regulatory, Bank, and Environmental Covariates

The table shows descriptive statistics for 473 banks that received capital injections. All data are obtained from the Bundesbank. Shares, ratios, and growth rates are measured in percent unless noted otherwise. Core capital ratio: Tier I capital to risk-weighted assets; Hidden reserve ratio: hidden equity reserves according to §340f commercial code (HGB) to total assets; NPL share: non-performing loans to total assets; RoE (growth): (growth during the injection spell of) return on equity; RWA growth: growth of risk-weighted assets during the spell; Customer loan growth: growth of customer loans during the spell; GDP: log of real gross domestic product per county; Interest spread: difference between 10-year and 1-year federal government bonds.

article image

Panel two depicts bank-specific determinants of recovery.25 The accumulation of core capital is key to allow repayment. Therefore, we specify both Tier I capital ratios as well as hidden reserves. Higher mean capital ratios during the spell of support serve as a stabilizing signal to the market and should thus enhance recovery odds. Troubled banks typically exhibit high shares of non-performing loans (NPL) (see e.g., Cole and Gunther, 1995; Dahl and Spivey, 1995) and we expect a negative relation with recovery odds. Severely distressed banks are characterized by high losses (see e.g., Gan, 2004). Successfully improving earnings as measured by return on equity (RoE) should therefore have a positive effect on recovery odds. The dispersion in both variables also indicates that numerous banks manage to improve their profitability and to rectify the quality of their loan portfolio to conventional levels. But other supported banks apparently maintain very risky credit portfolios.

Thus, we test if regulatory and other disciplining mechanisms can predict eventual recovery of banks choosing fairly different production plans. To explain the duration until repayment, we focus on the growth of three variables during the period when capital injection was in place. First, re-establishing a strong earnings base is central to accumulate retained earnings. We expect faster ROE growth to reduce the time-to-recovery. Second, banks that received capital presumably hold risky portfolios. To mitigate threats of undercapitalization, deleveraging the bank can be achieved by reallocating portfolios to less risky asset classes. Therefore, we specify growth of risk-weighted assets as a determinant and expect a negative relation with recovery duration. Third, and related to deleveraging distressed banks, we control for possible attempt to shrink the balance sheet (see also Bandopadhyaya and Jaggia, 2001, for evidence of higher recovery probabilities of low-growth firms). At the same time, generating revenues is a necessary condition to generate (retained) earnings. The effect of customer loan activity, the main line of business of banks in our sample, is therefore ambiguous a priori.

To explain the duration until repayment, we focus on the growth of three variables during the period when capital injection was in place. First, re-establishing a strong earnings base is central to accumulate retained earnings. We expect faster ROE growth to reduce the time-to-recovery. Second, banks that received capital presumably hold risky portfolios. To mitigate threats of undercapitalization, deleveraging the bank can be achieved by reallocating portfolios to less risky assets. Therefore, we specify growth of risk-weighted assets as a determinant and expect a negative relation with recovery duration. Third, and related to deleveraging distressed banks, we control for possible attempt to shrink the balance sheet (see also Bandopadhyaya and Jaggia, 2001, for evidence of higher recovery probabilities of low-growth firms). At the same time, generating revenues is important for repaying injected capital. The effect of customer loan activity, the main line of business of banks in our sample, is therefore ambiguous a priori.

Recent banking crisis studies, such as Hoshi and Kashyap (2008) or Quagliariello (2008), and corporate finance literature on non-financial firm failure (Acharya and others, 2007), emphasize the importance of macroeconomic conditions for the effectiveness of regulatory policy. To control for regional macroeconomic conditions, we specify the log of real GDP at the county (Kreis) level. Higher real income indicates larger demand for financial services, which ceteris paribus should increase recovery probabilities and ease banks’ efforts to repay capital injections. Since one of the key intermediation functions of banks includes maturity transformation, we specify also the spread between long- and short-term government bond rates. Larger spreads should permit banks to earn higher margins on their transformation function and thus enhance recovery probabilities and reduce the time until repaying injected capital.

V. Results

Table 4 shows results from three specifications, each depicting in the first column estimates of the probability of recovery and the duration of recovery thereafter. For all specifications [1] through [3], the shape parameter γ = 1/p is significant, thus supporting the split-population model over a conventional duration specification that would neglect the possibility of not recovering. Given that γ < 1, the value of p is larger than unity, implying unimodal distribution of the log-logistic distribution and existence of the mean, which is modeled as a function of covariates.

Table 4.

Split-Population Duration on Injection Recovery Between 1994 and 2008

The table shows parameter estimates without (column pairs [1] and [2]) and with (column pair [3]) controls for bad banks. Standard errors are in brackets. Regulatory measures comprise four severe measures described in Section 3. Variables are defined as in Table 2. *; **; *** denote significance at the 1, 5, and 10 percent levels, respectively.

article image

Top panel results in Table 4 regarding recovery probabilities support the “regulatory insider” hypothesis. Imposing severe regulatory measures increases the likelihood of recovery by around four times relative to the likelihood of not intervening. This suggests that German regulators had superior information at their disposal to take adequate action that facilitated bank recovery.

An important caveat concerns endogenous regulatory intervention at banks that are in more severe distress, and thus simply need more time for recovery. The positive effect of regulatory measures on recovery duration might therefore merely reflect that banks are asked by the regulator to thoroughly, i.e., over a longer period of time, restructure operations and “clean” balance sheets. To control for this possible selection bias, we use the benefit of hindsight and define banks that ultimately exit the sample due to ordered restructuring mergers or moratoria as the most severely troubled institutes. If measures are mainly applied to banks exhibiting higher degree of distress, the effect of regulatory intervention proxies should vanish and other factors should determine recovery. Specification [2] reports results for the subsample excluding most severely troubled banks. Specification [3] is based on the full sample but includes an indicator variable for supported banks that eventually were forced to a distressed merger or to exit the market. Both specifications corroborate the main conclusion in favor of the “regulatory insider” hypothesis.

The evidence for recovery duration supports the “sustainability” hypothesis. Dealing with weak banks apparently requires patience. The estimated coefficient for regulatory action of 0.51 entails that recovery is delayed by 1.9 years compared to “laissez-faire.” The alternative policy instrument to generate more information about supported banks by means of unscheduled audits also lengthens the spell of supported banks. The likelihood of recovery, in turn, is barely significant for the sample of banks that are not eventually forced to cease as a going concern and vanishes entirely when controlling explicitly for most severely distressed banks. Hence, additional scrutiny alone does not enhance the odds of recovery but seems to primarily serve as a conduit to generate necessary information that facilitates thorough and sustainable restructuring efforts.

Conventional monitoring mechanisms that should ensure that managers having received a bailout do not shirk are only of limited importance among German banks. The indicator variable for incorporated banks, and associated stricter publication requirements, exhibits a comparable impact on both recovery odds and duration compared to the one of regulatory intervention. To some extent this result is in line with the finding by Berger and others (2000) of a complementary role of both financial markets and supervisors to discipline distressed banks. But statistical significance turns weak once most severely troubled banks are excluded. Market based governance therefore seems primarily of importance for substantially distressed banks that receive capital support.

The middle panel of Table 4 highlights that, in contrast to conventional failure rate studies, only few bank-specific variables affect recovery. We find in line with expectations that rebuilding capitalization and the reduction of non-performing loans is of central importance to increase the chances of recovery. In contrast, profitability and alternative capital reserves have no significant effect. The time of capital support spells, in turn, is shortened for all supported banks if earnings grow faster. For the sub-sample excluding most severely troubled banks, we provide evidence that increasing risk-weighted assets lengthen the spell. This could be because higher capital requirements following from larger risk-weighted assets postpone or even preclude repayment. In contrast, expanding the volume of banks core business, customer loans, appears to provide less severely distressed banks with additional scope to accumulate retained earnings and therefore reduce support duration.

Positive regional economic conditions generally ease recovery as indicated by a negative coefficient for the log of real GDP for the duration equation. Likewise, the positive (negative) effect of interest rate spreads on recovery probability (duration) indicate that next to regulatory and bank-specific measures, a favorable macroeconomic environment is important to facilitate bank recovery. But relying on favorable business conditions would be an insufficient policy to foster recovery since it does not affect the likelihood of recovery significantly.

In sum, severe regulatory measures and market discipline increase recovery probabilities. Unscheduled audits, in turn, lengthen the duration of recovery spells together with regulatory measures, corroborating the conjecture by various academics and policy makers that a thorough turnaround of troubled banks takes time. While different in magnitude, these effects appear robust to the separation of most troubled banks. Market imposed discipline, in turn, does not affect the time until repayment once we control for the most troubled institutions. This underlines the relatively limited role of financial markets for corporate governance purposes in Germany’s bank-based financial system (Levine, 2002).

A. Severity of Measures

As indicated in Section 3, the vast majority of regulatory measures is weak and has an early warning character. In Table 5 we show that only the coefficient for the few severe interventions has a significant effect on both the recovery probability and the duration of capital support spells. Specification [2], in contrast, highlights that weak measures by the regulator, such as warning letters, have no statistically significant impact on the recovery pattern of banks.

Table 5.

Separating Different Severity of Interventions

The table shows parameter estimates for weak regulatory measures (column pairs [1]), severe regulatory measures (column pairs [2]), and both types of regulatory measures (column pairs [3]) including a dummy to control for bad banks. Standard errors are in brackets. The individual measures are described in Section 3. Variables are defined as in Table 2. *;**;*** denote significance at the 1, 5, and 10 percent levels, respectively. 2,165 observations.

article image

This result is confirmed in non-reported regressions specifying an indicator of all measures issued by the authorities. Note that these results do not preclude that such measures can be effective for less severely distressed banks, i.e., those without need to tap equity support schemes. Our results show, however, that once banks received capital, only stern regulatory action has a significant effect.

In line with sub-sample results in Table 3, market governance as measured by the indicator variable for incorporation is eliminated once we control for weaker measures taken by regulators. Potentially, monitoring by equity and debt holders of distressed, incorporated banks is crowded out by regulatory attention independent of whether actions taken effectively facilitate recovery of the bank or not.

B. Time Span Sensitivity

Our sample is subject to two further limitations. First, available capital support data does not permit identification of the exact timing of injections prior to 1994. Second, some results might partly be driven by including the year 2008, which is when the global financial crisis fully unfolded and induced the government to directly or indirectly support banks by providing guarantees or acquiring common equity. Here, we measure capital injections by banking-sector specific insurance funds, which have been tapped less frequently under these extreme conditions. However, low frequency of such “conventional” support schemes is deceptive since it does not indicate healthy banks per se but might rather reflect substitution of capital preservation measures by federal government actions. Ideally, we would consider all types of capital preservation activities by any government or non-government agency explicitly, but such data is unavailable. Therefore, we limit our robustness check in Table 6 to the exclusion of the uncertain starting year 1994, or the crisis year 2008, or both.

Table 6.

Exclusion of Uncertain and Crisis Years

The table shows robustness check results for the following subsamples of data: excluding 1994(column pairs [1]), excluding 2008 (column pairs [2]), and excluding both 1994 and 2008 (column pairs [3]). All specifications include a dummy to control for bad banks. Standard errors are in brackets. The individual measures are described in section 3. Variables are defined as in Table 2. *;**;*** denote significance at the 1, 5, and 10 percent levels, respectively.

article image

The results corroborate the main conclusion that only stern regulatory action significantly increases recovery odds but also lengthens the duration of capital support schemes. As such, the “regulatory insider” hypothesis as well as the “sustainability” hypothesis are consistently accepted. Across specifications in Tables 4 through 6, signs, significance, and magnitudes of bank-specific effects point consistently into the same direction.

The ability of banks to clean credit portfolios and reduce the share of risky loans paired with the accumulation of capital buffers increases the likelihood of recovery. The duration of capital support spells, in turn, primarily depends on increasing earnings growth. Independent of controlling for different sample years and the presence of particularly bad banks, an increase of ROE growth by one percentage point reduces the time until recovery by around a year. Evidence for reduced support durations due to asset or credit growth is weak since significant results are not only sensitive to the specification of multiple types of measures, but also time spans.26

VI. Conclusion

We test if and how regulatory intervention affects the recovery probability and duration of re-capitalized banks, respectively. We use detailed data of a total of 440 supervisory measures and 192 non-scheduled audits applied to all 473 universal banks that received capital support from banking-sector specific insurance schemes in Germany. The first hypothesis tested is that regulators generate and process additional information that mitigates moral hazard after capital support (“regulatory insider” view) and therefore facilitate recovery versus the hypothesis that regulators are entrenched with their subjects and therefore delay necessary, but potentially painful decisions, ultimately hindering recovery (“regulatory capture” view). Given the probability that a bank recovers, we also hypothesize that regulatory intervention affects the duration of such capital support spells. We test whether supervisory prudence implies thorough restructuring (“sustainability” hypothesis) or enables the bank to recover faster by enforcing necessary restructuring measures (“acceleration” hypothesis). We estimate conditional recovery probability and duration using the split-population duration model for the period 1994 to 2008.

Estimation results provide support for “regulatory insider” and “sustainability” hypotheses. Severe regulatory measures that can limit the scope of managerial choices substantially enhance recovery probabilities but also lengthen the time until repayment. Unscheduled audits, in turn, only lengthen the capital support spell and thus seem to primarily serve the purpose of generating additional insider information to permit thorough and sustainable restructuring. Regulatory intervention lengthening the time until repayment corroborates the conjecture by various academics and policy makers that a thorough turnaround of troubled banks takes time (BIS, 1999; Hoshi and Kashyap, 2008; IMF, 2009b).

A first important qualification of our main result is the absence of significant effects on recovery patterns by weak regulatory action. Official warning letters do neither reduce recovery time nor increase likelihood of recovery. Because the majority of regulatory actions in fact resemble weaker measures, this result bears the important policy implication that once banks are sufficiently distressed to receive capital support, only stern regulatory action has a statistically significant influence on recovery. Weak, frequently atomized measures in turn might only be suited for less severely troubled banks that did not tap capital support schemes.

Second, we find only limited evidence of effective market-imposed discipline, measured by incorporation indicators and the associated stricter publication requirements. Positive effects on recovery odds turn insignificant once we control with the benefit of hindsight for the presence of severely distressed banks, defined as those that are ultimately forced to merge or closed due to moratoria. Hence, market governance appears to be of relatively limited importance in Germany’s bank-based corporate governance culture.

Bank’s ultimate production choices matter, too. Capitalization and non-performing loan share levels are the main driver of recovery odds, while earnings growth is the prime determinant for recovery duration. In addition to these bank-specific factors, both regional real GDP as well as federal interest rate spreads are economically significant co-determinants of recovery probabilities and duration. This highlights the importance to coordinate both macro- and micro-prudential supervision policies to foster the recovery of supported banks.

Recovery Determinants of Distressed Banks: Regulators, Market Discipline, or the Environment?
Author: Michael Koetter, Mr. Tigran Poghosyan, and Thomas Kick