BANK AUDIT AND BANK FAILURE:WHICH CAUSES WHICH?

Rebel A. ColeDePaul University / U.S. IRS

Drew DahlUtah State University / U.S. Treasury Dept.

Presentation for the2014 Annual Meetings of the Midwest Finance Association

Orlando, FLMarch 6, 2014

Summary

We examine the decision of a bank to hire an external auditor and the decision of a bank regulator to close a bank.

In univariate tests, we find that audited banks failed at 3X the rate of unaudited banks.

Using a bivariate probit model to control for endogeneity of the audit decision, we also find that this disparity disappears once we control for differences in ex ante and ex post measures of risk.

However, we find no evidence that external audit reduces the probability of failure; in fact, we find no significant effect.

Summary

From this evidence, we conclude that the incidence of external audit has no impact on the likelihood of bank failure.

In contrast, our evidence supports our hypothesis that high-risk managers choose to be audited in order to benefit from the certification effect of audits. Being subject to external audit enables high-risk

bank managers to push back against regulators who question their high-risk asset and liability portfolios.

Our findings with respect to the determinants of audit and failure are generally consistent with the extant literature on these two topics. More complex firms are more likely to choose to be

audited. CAMEL proxies explain the likelihood of bank failure.

Introduction

There are widespread beliefs that bank external auditors: enhance market confidence, improve the quality of information relied on by banking

supervisors, mitigate internal control weaknesses, identify problems unknown to supervisors and help correct

them U.S. General Accounting Office (GAO), 1989. Basel Committee of Bank Supervision (BCBS), 2008. Commercial Bank Examination Manual, 2011.

Such beliefs have been confirmed, to varying degree, in studies by Dahl et al. (1998) and Gunther and Moore (2003) indicating that auditors affect discretionary accounting decisions in banking.

Introduction

Yet bank failures during the recent financial crisis raise questions concerning external auditing practices in the banking industry.

Why did auditor reports issued in early 2009 on financial institutions receiving government bailouts fail to contain any “red flags” when compared to the very same audit reports in 2006, at the peak of the business cycle (U.S. Senate, 2011)?

Did audit inadequacies play a role in the inability of regulators to mitigate bank risk in a timely manner during the crisis period (FRB, 2011)?

Introduction

In this study, we shed new light on this issue. We examine whether the incidence of external

audit affects the likelihood of bank failure. The only previous study of bank audit and

failure (Jin et al. JBF 2010) examines the intensity of audit. Are banks audited by bank specialists? They find that banks audited by specialists

are significantly less likely to fail. However, that study is riven by

methodological problems that render its findings meaningless.

Data:Sources

Our bank data come from two sources. Failure data come from the FDIC’s website. Audit status and financial data from the

U.S. FFIEC, which collects quarterly financial data on all U.S. banks on behalf of the three primary regulators (FDIC, FRB, and OCC).

These financial data are available quarterly from 1980 – 2010 from the FRB-Chicago website.

We focus on year-end data.

Data:Explanatory Variables

The explanatory variables we create for inclusion in our audit and failure models follow the existing literatures on audits and failures. The incidence of bank audit is modeled

as a function of firm complexity. The likelihood of bank failure is

modeled as a function of proxies for the regulatory CAMELS component ratings.

Data:Explanatory Variables

AUDIT_1_2 i,t = a0 + a1 LNSIZE i,t-1 + a2 LNAGE i,t-1

+ a3 SCORP i,t-1 + a4 PUBLIC i,t-1

+ a5 FFE_EXP i,t-1 + a6 TTE i,t-1 + a7 TLGROWTH i,t-1

+ a8 GOODWILL i,t-1 + a9 INSIDER i,t-1 + e i,t

FAILURE,t+1 = a0 + a1 AUDIT_1_2 i,t

+ a2 TTE i,t + a3 LLR i,t

+ a4 NPA i,t + a5 ROA i,t + a6 SEC i,t

+ a7 RECOM i,t + a8 RECON i,t + a9 BD i,t + e i,t

Data:Bank Audit Status

Banks with $500 million or more in total assets are required to be externally audited.

Banks with less than $500 million in assets are encouraged, but not required to be externally audited.

In practice, about 1 in 3 small banks choose not to hire an external auditor.

Data:Bank Audit Status (1) independent audit of the bank conducted in accordance with generally

accepted auditing standards by a certified public accounting firm which submits a report on the bank;

(2) independent audit of the bank’s parent holding company conducted in accordance with generally accepted auditing standards by a certified public accounting firm which submits a report on the consolidated holding company (but not on the bank separately);

(3) attestation on bank management’s assertion on the effectiveness of the bank’s internal control over financial reporting by a certified public accounting firm;

(4) director’s examination of the bank conducted in accordance with generally accepted auditing standards by a certified public accounting firm;

(5) director’s examination of the bank performed by other external auditors; (6) review of the bank’s financial statements by external auditors; (7) compilation of the bank’s financial statements by external auditors; (8) other audit procedures; and (9) no external audit work.

Data:Bank Audit Status

We combine (1) and (2) into a classification for being externally audited.

We combine (3) – (9) into a classification for not being externally audited.

These two classification define our audit status variables AUDIT_1_2 and ZEROAUDIT .

We collect annual audit status as of 2005 – 2008 and classify a bank as audited if it reports classification 1 or 2 during any of these four years.

In practice, very few banks change audit status.

Data:Bank Failure Status

We classify banks that were closed by their primary regulator during 2009 as failures.

We also classify banks that reported year-end 2009 NPLs greater than 200% of equity as failures.

We classify all remaining banks as survivors. We only include in our analysis banks with

less than $500 million in assets as of year-end 2008.

We collect our explanatory variable for our failure model as of year-end 2008.

Data:Endogeneity of Audit and Failure

One important issue we face is the endogeneity of bank audit and failure: Small banks receiving a composite

CAMELS rating of 4 or 5 (the two worst rating, indicating serious problems) must hire an external auditor.

This induces a positive correlation between audit and subsequent failure.

Data:Endogeneity of Audit and Failure

We deal with this potential endogeneity using two approaches:

First, we employ the bivariate probit methodology, This enables us to jointly model the bank

audit decision and regulatory closure decision.

The methodology controls for endogeneity of failure and audit by ensuring that the error terms in the two equations are uncorrelated. [See Wooldridge (2002), pp. 477-478.]

Data:Endogeneity of Audit and Failure

Second, we lag our classification of audit status by up to four years—back to 2005—in our model of 2009 failure. It is unlikely that audit status in 2005 is

correlated with incidence of failure in 2009. Moreover, only 52 banks were rated

CAMEL 4 or 5 back in 2005—the smallest number in 36 years—and no banks failed in 2005 or 2006. However, we do not know their identity

because CAMELS are confidential, so we cannot exclude them.

Univariate Results:Descriptive Statistics by Audit Status

Variable Mean Std.Error Mean Std.Error Mean Std.Error Difference t-Stat

LNSIZE 11.29 0.01 11.48 0.01 10.92 0.02 0.56 24.7 aLNAGE 4.03 0.01 3.82 0.02 4.43 0.01 -0.61 -31.3 aSCORP 0.308 0.006 0.231 0.007 0.452 0.011 -0.221 -16.7 aPUBLIC 0.095 0.004 0.142 0.006 0.007 0.002 0.135 22.2 aFFE_EXP 0.004 0.000 0.004 0.000 0.004 0.000 0.001 7.7 aTLGROWTH 0.178 0.005 0.229 0.008 0.084 0.003 0.144 17.3 aGOODWILL 0.002 0.000 0.003 0.000 0.002 0.000 0.001 3.4 aINSIDER 0.014 0.000 0.016 0.000 0.011 0.000 0.005 10.3 a

All Banks Audited Banks Unaudited Banks Difference in Means

Univariate Results:Descriptive Statistics by Failure Status

Variable Mean Std.Err. Mean Std.Err. Mean Std.Err. Difference t-Stat

ZEROAUDIT 0.348 0.006 0.355 0.007 0.121 0.025 0.235 9.14 a

TTE 0.106 0.001 0.107 0.001 0.076 0.002 0.031 13.71 aNPA 0.030 0.000 0.027 0.000 0.117 0.006 -0.090 -14.69 aROA 0.006 0.000 0.007 0.000 -0.020 0.002 0.027 11.94 aSEC 0.211 0.002 0.214 0.002 0.107 0.006 0.107 16.51 a

RECOM 0.150 0.002 0.147 0.002 0.244 0.009 -0.097 -10.36 aRECON 0.063 0.001 0.058 0.001 0.193 0.009 -0.135 -15.17 aBD 0.035 0.001 0.031 0.001 0.130 0.010 -0.099 -9.57 a

All Banks Surviving Banks Failing Banks Difference in Means

Multivariate Results:Bivariate Probit ZEROAUDIT Eq.

Coefficient Standard ApproxVariable Estimate Error t Value Pr > |t|

Audit EquationIntercept 3.375 0.305 11.08 <.0001LNSIZE -0.502 0.024 -21.07 <.0001LNAGE 0.515 0.031 16.53 <.0001SCORP 0.420 0.041 10.32 <.0001PUBLIC -1.094 0.130 -8.43 <.0001FFE_EXP -65.715 10.167 -6.46 <.0001TLGROWTH -0.356 0.084 -4.23 <.0001GOODWILL -5.758 2.574 -2.24 0.0253INSIDER -3.292 1.344 -2.45 0.0143

Multivariate Results:Bivariate Probit FAILURE Eq.

Coefficient Standard ApproxVariable Estimate Error t Value Pr > |t|

Failure EquationIntercept -1.716 0.317 -5.41 <.0001ZEROAUDIT 0.104 0.468 0.22 0.8238TTE -17.639 2.431 -7.26 <.0001NPA 9.772 1.046 9.34 <.0001ROA -6.644 2.532 -2.62 0.0087SEC -0.165 0.552 -0.3 0.7644RECOM 2.217 0.490 4.53 <.0001RECON 3.309 0.579 5.71 <.0001BD 0.972 0.421 2.31 0.0208

Correlation of error terms -0.086 0.313 -0.27 0.7840

Summary & Conclusions

Our purpose has been to explain the apparent paradox by which auditors are thought to constrain risk-taking by banks, yet during the recent financial crisis, audited banks were 3x more times likely to fail than unaudited banks.

We analyze a sample of 5,568 small commercial banks, of which 3,630 were audited and 1,938 were unaudited; and of which 174 failed and 5,394 survived.

Summary & Conclusions

We find that a strong and positive relation between incidence of audit and likelihood of failure disappears once we control for bank risk-taking.

Specifically, we find that adding controls for CRE lending, C&D lending, and brokered-deposit funding to a simple CAMELS-based model renders the incidence of audit insignificant in explaining failure.

Summary & Conclusions

This supports our hypothesis that high-risk managers choose to be audited in order to obtain “certification” benefits, and is at odds with our hypothesis that audits reduce bank risk-taking.

Our results conflict with those of Jin et al. (2010), who conclude that auditors are better able to “constrain the tendency…to engage in aggressive reporting or fraud.”

Our results also conflict with the belief that auditors can help identify, and mitigate, incipient financial problems.