Journal of Contemporary Accounting & Economics 12 (2016) 61–72

Contents lists available at ScienceDirect

Journal of Contemporary Accounting & Economics

j o u r n a l h o m e p a g e : w w w. e l s e v i e r. c o m / l o c a t e / j c a e

The effect of commercial banks’ internal control weaknesses

on loan loss reserves and provisions
Myojung Cho *, Kwang-Hyun Chung
Pace University, New York, USA

A R T I C L E I N F O A B S T R A C T

Article history: This study examines whether the material internal control weaknesses (ICW) of commer-
Received 9 June 2015 cial banks affect loan loss reserves and provisions. Bank regulators have been keen to improve
Received in revised form 1 February 2016 the internal control procedures of banks in order to obtain accurate estimates of loan loss
Accepted 1 February 2016
exposures. GAO (1991, 1994) reports that loan-loss reserves are often determined based
Available online 23 February 2016
on historical loss rates even for large loans, rather than individual loan impairment as-
sessments, and the reported loan loss reserves include substantial amounts of supplemental
Keywords:
reserves that are not linked to the loan loss exposure. We expect and find that banks with
Loan loss reserves
Loan loss provisions material ICW have, on average, higher loan loss reserves and provisions in years of ICW
Internal control weaknesses than those without ICW. We also find that ICW banks with successful remedial actions no
longer have higher levels of loan loss reserves or provisions in the next year, while banks
that report material ICW in both the current and following year continue to have signifi-
cantly higher amounts of loan loss reserves and provisions in the next year.
© 2016 Elsevier Ltd. All rights reserved.

1. Introduction

The effectiveness of banks’ internal control has been under the scrutiny of regulators since as early as 1991.1 GAO reports
of 1991 and 1994 document that many failed banks had inaccurate loan loss reserves and provisions due to poor internal
control procedures, distorting their financial statements. Loan loss reserves are the estimated amount of the bank’s loss ex-
posure to cover uncollectible outstanding impaired loans. Loan loss provisions are an expense necessary to adjust the loan
loss reserves. Loan loss reserves are the single largest component of bank accruals, and commercial bank loan portfolios
are typically ten to fifteen times larger than their equity (Altamuro and Beatty, 2010; Wahlen, 1994). Therefore, loan loss
reserves and provisions provide critical information about credit risk exposures to the clients and investors of banks. In line
with regulatory attention to internal control effectiveness and its relation to loan loss reserves, we investigate whether ma-
terial internal control weaknesses (ICW) are systematically associated with loan loss reserves and provisions of banks, and
whether the following remedial actions affect them in the year after the material ICW report.
After a number of banks failed in the 1980s, the Federal Deposit Insurance Corporation (FDIC) passed the Federal Deposit
Insurance Improvement Act (FDICIA) in 1991, requiring that the management of banks with more than $500 million in assets
evaluate the effectiveness of their internal controls processes of financial reporting and that the banks’ external auditors
attest to the effectiveness of the banks’ internal control structure. However, banks continued to suffer from internal control
problems related to corporate governance and financial reporting, and rocked the financial markets again in the early 2000s,

* Corresponding author. Pace University, 1 Pace Plaza, W414, New York, NY 10038, USA. Tel.: +1 212 618 6426; fax: +1 212 618 6605.
E-mail address: mcho@pace.edu (M. Cho).
1
Bank regulators include the Federal Deposit Insurance Corporation (FDIC), the Office of the Comptroller of the Currency (OCC), the Federal Reserve
Board (FRB), and the Office of Thrift Supervision (OTS).

http://dx.doi.org/10.1016/j.jcae.2016.02.004
1815-5669/© 2016 Elsevier Ltd. All rights reserved.
62 M. Cho, K.-H. Chung / Journal of Contemporary Accounting & Economics 12 (2016) 61–72

resulting in a significant loss of investor confidence in U.S. businesses and auditing firms.2 To regain investors’ trust, the
Securities and Exchange Commission (SEC) enacted the Sarbanes–Oxley Act (SOX) in 2002, imposing stricter corporate gov-
ernance measures on not only banks but also all other public firms with assets greater than $75 million. In particular, Section
404 (b) of the SOX requires public firms to evaluate and report on the effectiveness of their internal control attested by in-
dependent auditors, and as a result the ICW data have become readily available. More recently, the Federal Reserve has performed
a sweeping review of its bank supervision on internal control and how much loan loss reserves banks must maintain, after
the 2007–2008 financial crisis (Ryan, 2014). As a result, the Dodd–Frank Wall Street Reform and Consumer Protection Act
has been enacted in 2010.
Although there is a dearth of accounting studies on internal control issues and loan loss estimation among banks, Altamuro
and Beatty (2010) investigate whether the internal control regulations of the FDCIA improve the validity of loan loss pro-
visions. They compare banks affected by the FDCIA with total assets greater than $500 million to banks unaffected by the
regulation. However, it is possible that some banks with total assets greater than $500 million do not have material ICW.
Our study extends Altamuro and Beatty (2010) by comparing loan loss estimates between the banks that actually report
material ICW and those with no ICW in order to understand the role of material ICW in loan loss estimation. Therefore, our
study connects the current literature on ICW with the existing literature on loans loss.
One line of ICW literature focuses on the impact of ICW on the cost of capital (Ashbaugh-Skaife et al., 2009; Beneish
et al., 2008; Hammersley et al., 2008; Kim et al., 2011; Ogneva et al., 2007). Ogneva et al. (2007) find no significant rela-
tionship between ICW and the cost of equity. Ashbaugh-Skaife et al. (2009), however, report that ICW increases the cost of
capital and that the cost of capital significantly decreases after remedial actions for ICW. Another line of ICW literature in-
vestigates characteristics of ICW firms (Ashbaugh-Skaife et al., 2009; Chan et al., 2008; Doyle et al., 2007; Ge and McVay,
2005; Krishnan and Visvanathan, 2007).3 ICW firms tend to be younger, less profitable, and in the midst of rapid growth.
They tend to have more complex operations, recent changes in organizational structure, more accounting risk exposure,
fewer resources, or a lower return–earnings relationship, compared to their counterparts.
Prior studies on loan loss reserves usually focus on bank managers’ incentives to use loan loss estimation. Loan loss re-
serves may be used in income smoothing (Ahmed et al., 1999; Kanagaretnam et al., 2003, 2004; Kilic et al., 2013). When
earnings are poor, managers can draw down the allowances for loan loss to reduce loan loss provisions and, as a result, in-
crease earnings. Chen et al. (2005) document that determinants of loan loss reserves include nonperforming loans, loan interest
rates, and loan loss charge-offs. However, they do not look into the quality of the loan loss review process and the resulting
ICW reports as another determinant. To investigate these aspects, we posit that banks with material ICW have lower quality
internal loan loss review processes than their counterpart.
Banks have to follow the Statement of Financial Accounting Standards (SFAS) No. 114 Accounting by Creditors for Impair-
ment of a Loan and the SFAS No. 5 Accounting for Contingencies for their loan loss estimations (Chen et al., 2005). The SFAS
No. 114 requires that banks determine which loans are impaired and assess the degree of impairment according to their
internal evaluation process. The SFAS No. 5 provides the fundamental recognition criteria for loan losses. Due to the diver-
sity and subjectivity of internal loan loss evaluation methods, the reported loan loss reserves and provisions often lack reliability,
tainting the transparency of financial statements of banks. In our first hypothesis, we investigate whether banks with ma-
terial ICW are more likely to misrepresent loan loss reserves and provisions.
Regulators believe that banks need to use individual loan assessments especially for large impaired loans (GAO, 1994).
A high quality internal control system is imperative for accurate estimation of loan loss exposure, as individual loan as-
sessments require an in-depth understanding of loan payment history, collaterals, and other relevant information. However,
banks commonly use historical loss rates to evaluate loan loss risk and often keep excessive supplemental reserves. Supple-
mental reserves are set aside in addition to the reserves that are estimated from the loan loss assessment, and they are used
to avoid negative reserves adjustments, appease regulators, or follow historical loss rates (GAO, 1994).4 Unjustified supple-
mental reserves can be used to manage earnings and distort future loan loss estimation, resulting in more room for error
and manipulation.5 Thus, we expect that banks with material ICW tend to misrepresent their loan loss estimates and have
excessive supplement reserves. It is because they have less pressure to provide legitimate justifications of the estimates and
excessive reserves.6 In order to test the first hypothesis, we compare loan loss reserves and provisions between banks with
material ICW and banks without ICW. We construct our matching samples based on firm size and return on assets for each
year. We find that banks with material ICW indeed have significantly greater amounts of loan loss reserves and provisions
than the control sample. Our results remain robust even after controlling for the self-selection bias and endogeneity.

2 Examples of accounting fraud in the early 2000s include the cases of Enron, Tyco and WorldCom, among others.
3
Krishnan and Visvanathan (2007) report that ICW firms have a higher number of meetings of the audit committee, fewer “financial experts” on the
audit committee, more auditor changes, and a greater number of prior restatements, compared to the firms with no ICW. Chan et al. (2008) find that ICW
firms under Section 404 have more positive and absolute discretionary accruals than the control firms.
4
Supplemental reserves range from 30% to 70% of total reported loan loss reserves (GAO, 1994).
5
Although banks claim that supplemental reserves are established to cover possible losses from bad loans, loan grading errors, and economic condi-
tions, excessive supplemental reserves can also be used to distort the current risk exposure of a bank, as previously reserved amounts can absorb the
impairment of loan values in the current period without increasing loan loss reserves or provisions.
6
Because of the periodic examinations by regulators such as the FDIC and the State Banking Authorities, banks with material ICW are less likely to lower
loan loss reserves aggressively without proper justification.
 M. Cho, K.-H. Chung / Journal of Contemporary Accounting & Economics 12 (2016) 61–72 63

An alternative explanation of the greater loan loss reserves and provisions among ICW banks is private information sig-
naling. Prior studies report that banks use loan loss reserves and provisions to signal private information about positive future
prospects (Beaver and Engel, 1996; Kanagaretnam et al., 2004, 2005, 2009; Wahlen, 1994). However, Ahmed et al. (1999)
does not find private information signaling among banks. We investigate whether material ICW plays a significant role in
the banks’ decision to signal private information via loan loss estimates. If ICW banks increase loan loss estimates to signal
good news about future earnings, they would most likely enjoy better future earnings performances. Thus, we divide our
sample of ICW banks into two groups with high loan loss reserves (or provisions) and low loan loss reserves (or provi-
sions) and compare the future earnings performance of the two groups. We do not find a significant difference in future
earnings performances of the two groups, contrary to the signaling scenario.
In the second hypothesis, we investigate whether the excessive loan loss reserves would decrease in the following year
of ICW due to remedial actions. If banks with material ICW take successful remedial actions and improve their internal control
procedures and measurements of operating activities, we expect the excessive loan loss reserves to disappear in the next
year. To test the second hypothesis, we compare loan loss reserves and provisions between banks with material ICW and
banks without ICW in the year following the ICW reports. As expected, our results show that there is no significant differ-
ence in loan loss reserves and provisions between the two groups. However, it is possible that not all banks succeed in improving
their ICW problems in the following year. Thus, we divide our test sample into (1) banks with material ICW in the current
year but no ICW in the next year and (2) banks with material ICW in both the current and following years and construct
the matching samples for the two groups separately. We find that loan loss reserves and provisions of the banks that succeed
in remedying ICW are not significantly different from those of their counterparts in the next year. On the contrary, the banks
that report material ICW in both years continue to have significantly greater loan loss reserves and provisions in the next
year.
The remainder of this paper is organized as follows: Section 2 presents hypothesis development, Section 3 describes our
research design, and Section 4 discusses our data and empirical results with sensitivity tests. We present our conclusion in
Section 5.

2. Hypotheses development

Banks often pursue new lending opportunities with unfamiliar risks, especially when margins are under pressure.7 There-
fore, bank regulators have continuously endeavored to ensure high quality internal control procedures among banks for adequate
amounts of loan loss reserves and provisions with accurate credit risk assessments.
Regulators believe that the best approach to evaluating loss exposures on large impaired loans is through individual loan
assessments, while small loans can be evaluated on a group basis using historical loss rates (GAO, 1994). If historical losses
are used to evaluate the loss exposure for large loans or the individual loan assessments are not adequately performed, the
allowance for loan loss could be overstated or understated, misrepresenting the credit risk of the bank’s outstanding loans.8
In addition, excessive supplemental reserves can contribute to excessive amounts of loan loss reserves and provisions. While
supplemental reserves may be necessary to prepare for unidentified loss exposures, bank managers can also use them to
hide the true risk exposure of their impaired loans as a potential tool for earnings management. For example, previously
established supplemental reserves can mask increases in estimated losses in the current year. That way, a bank can avoid
increasing its current loan loss reserves and provisions, which would signal the impairment of their loan values. Thus, the
current loss provisions and charge-offs in existing reserves would not reflect the increased loan loss risk in its loans. GAO
(1994) reports that the majority of the banks in their review have supplemental reserves comprising more than 30% of the
total allowance of loan loss and that the supplemental reserves are often quantitatively unjustified to the reasonable esti-
mates of likely losses.9 Prior studies report that material ICW is significantly related to earnings management and low accruals
quality (Chan et al., 2008; Doyle et al., 2007). Therefore, we expect that banks with material ICW have higher supplemen-
tal reserves inflating the allowance for loan loss and loan loss provisions than those without ICW. If banks with ICW tend
to misrepresent loan loss estimates and stock up on supplemental reserves without supporting rationales due to the low
quality of their internal control processes, banks with material ICW are more likely to have higher loan loss reserves and
provisions than banks without ICW. Therefore, our first hypothesis is as follows:

Hypothesis 1. Banks with material ICW have higher loan loss reserves and provisions in the year of ICW report than those
without ICW.

7
See the statement of Walter Smiechewicz, a managing director in risk assurance at PricewaterhouseCoopers, in a CFO (2013, September) article at:
http://ww2.cfo.com/auditing/2013/09/bank-internal-auditors-warned-about-lending-risks/.
8 For example, using extensive interviews with commercial banks, GAO (1994) reports a case where a bank with $21.5 million loans with $25 million

collateral has reserved $4.3 million of additional allowance for the loan to meet the historical industry loss rate.
9
Reported rationales for the unnecessary supplemental reserves include: to adjust loss history for current conditions, to provide a cushion for future
uncertainties, to appease regulators, or to avoid a negative loan loss provision.
64 M. Cho, K.-H. Chung / Journal of Contemporary Accounting & Economics 12 (2016) 61–72

ICW banks may increase loan loss estimates to signal private information. For example, Beaver and Engel (1996) report
that investors interpret increases in discretionary loan loss reserves as a sign of earnings power to withstand the addition-
al decrease in earnings due to loan loss. Similarly, Wahlen (1994) report that banks increase discretionary loan loss provisions
when the prospects of future cash flows improve. However, Ahmed et al. (1999) does not find that loan loss provisions are
positively related to future earnings or current stock returns, contrary to the signaling results of the other studies. Thus,
we investigate private information signaling as an alternative explanation of excessive loan loss estimates among ICW banks.
If ICW banks increase loan loss estimates to signal private information about good future earnings news, ICW banks with
high loan loss estimates are more likely to have better earnings performance in the future than those with low estimates.
Next, we examine the impact of remedial actions for ICW on the level of loan loss reserves and provisions. One year after
material ICW reports or after remedial actions, the ICW firms show improvement in cost of debts, earnings management,
and corporate governance. Ashbaugh-Skaife et al. (2009) report that the cost of debt decreases after remedial actions of
ICW. Li et al. (2010) find that improvements in chief financial officer (CFO) quality are positively associated with ICW remediation.
Johnstone et al. (2011) report that in the year following the year of ICW reports, firms have improved corporate governance.10
If banks with ICW take remedial steps in the year after the material ICW reports and estimate their loss exposure of im-
paired loans with greater accuracy, we expect that loan loss reserves and provisions in the year following ICW reports will
be statistically indifferent between ICW banks and those without ICW. Therefore, our second hypothesis is as follows:

Hypothesis 2. Banks with material ICW do not have higher loan loss reserves and provisions than those without ICW in
the year following ICW report due to remedial actions.

3. Research design

Our first hypothesis investigates whether banks with material ICW have more loan loss reserves and provisions than
their counterparts. To test this hypothesis, we use the following regression model:
LLR t or LLPt = α 0 + α1ICWt + α 2NIMt + α 3NPATt + α 4ROA t + α 5SIZEt + ∑ t α 6,tFYt + ε t (1)

where LLRt = loan loss reserves in year t, as a percentage of total loans;

LLPt = loan loss provisions in year t, as a percentage of total loans;
ICWt = an indicator variable of material internal control weakness in year t, set as 1 if there is material internal control
weakness in year t, zero otherwise;
NIMt = net interest margin on loans in year t, as a percentage of total loans;
NPATt = nonperforming assets in year t, as a percentage of total loans;
ROAt = return on assets, calculated as income before extraordinary items in year t, divided by total assets;
SIZEt = firm size, calculated as the natural log of total assets in year t;
FYt = fiscal year t; and,
εt = an error term.

We suppress firm subscripts for simplicity. If banks with material ICW fail to accurately estimate loan loss exposure and
load up on supplemental reserves as a tool of earnings management or as a mask for true risk exposure, we expect that the
coefficient of ICWt would be significantly positive. We control for net interest margin on loans (NIMt), nonperforming assets
(NPATt), return on assets (ROAt) and firm size (SIZEt), which are known as factors of loan loss reserves and provisions in prior
studies (Chen et al., 2005).
As an additional test, we run the regression model (1) without ICWt and compare the model fit of the data between the
regression model without ICWt and the regression model (1). If the Fisher’s F-test result comparing the two regressions models
is significant, one can conclude that ICWt significantly contributes to explaining the variation of the dependent variable,
LLRt or LLPt, supporting Hypothesis 1.
Our second hypothesis predicts that loan loss reserves and provisions in the year following material ICW reports would
not be significantly different between the ICW banks and non-ICW banks. We run the following regression for Hypothesis
2:
LLR t+1 or LLPt+1 = β0 + β1ICWt + β 2NIMt+1 + β 3NPATt+1 + β 4ROA t+1 + β5SIZEt+1 + ∑ t β6,tFYt + ε t (2)

The variables are defined in regression (1). We suppress firm subscripts for simplicity. A significant coefficient of ICWt
indicates that banks with remedial actions have a significantly different level of loan loss reserves or provisions in year t+1
from their control group. Therefore, we expect an insignificant coefficient of ICWt for both LLRt+1 and LLPt+1.

10 They report that, in the year of material ICW, firms have an increase in the proportion of independent directors on the board, an increase in the per-

centage of independent directors who also serve on other boards, improvements in audit committee member financial expertise, and an increase in the
percentage of shareholdings of audit committee members, among others.
 M. Cho, K.-H. Chung / Journal of Contemporary Accounting & Economics 12 (2016) 61–72 65

It is possible that not all firms are successful in improving ICW problems in the following year. Therefore, we test Hy-
pothesis 2 in a more detailed manner by dividing our sample into (1) banks that have ICW in year t but fail to remedy the
ICW problems and continue to have ICW in year t+1; and (2) banks that have ICW in year t but succeed in remedying the
ICW problems and do not have ICW in year t+1, following Johnstone et al. (2011). We expect that banks who succeed in
improving ICW problems in the following year have similar levels of loan loss reserves and provisions in year t+1 to those
of the control group. On the other hand, we expect that banks who report ICW in both years of t and t+1 continue to have
greater loan loss reserves and provisions in year t+1 than the control group. We test the difference between firms with suc-
cessful remedial actions and firms with unsuccessful remedial actions using the following regression:
LLR t+1 or LLPt+1 = γ 0 + γ 1ICWt + γ 2NOREM ⋅ ICWt + γ 3NIMt+1 + γ 4NPATt+1 + γ 5ROA t+1 + γ 6 SIZEt +1 + ∑t γ 7,t FYt + ε t (3)

where NOREM = an indicator variable for no remedial action after ICW in year t, set as 1 if ICWt = 1 and ICWt+1 = 1, and zero
if ICWt = 1 and ICWt+1 = 0.
We expect the coefficient for ICWt to be insignificant and the coefficient of NOREM·ICWt to be significantly positive. An
insignificant coefficient of ICWt indicates that banks with successful remedial actions (NOREM = 0) no longer have greater
loan loss reserves and provisions than their control group. A significantly positive coefficient of the interaction term, NOREM·ICWt,
indicates that banks with unsuccessful remedial actions still have greater loan loss reserves and provisions than their control
group.

4. Empirical results

4.1. Data

Our pooled sample consists of 8167 firm-year observations of banks (SIC codes 6000–6189) with non-missing variables
of material ICW (data item: COUNT_WEAK) in the AuditAnalytics database and loan losses reserves (data item: RLL) and
loan loss provisions (data item: PLL) in the Compustat database for years 2002–2013. We exclude observations with missing
values of total loans (data item: LNTAL), nonperforming assets (data item: NPAT), net interest margin (data item: NIM), income
before extraordinary items (data item: IB), or total assets (data item: AT) from Compustat. We identify 235 observations
with material ICW in the fiscal year, out of the 8167 observations in the pooled sample. Therefore, there are 7932 obser-
vations with ICWt = 0 and 235 observations with ICWt = 1 in the pooled sample.
We also construct a matched sample based on firm size (SIZE) as natural log of total assets and return on asset (ROA) as
income before extraordinary items divided by total assets. Specifically, we calculate ROA and SIZE for all the observations in
the pooled sample and divide the sample into SIZE deciles groups for each year. For each of the 235 observations with ICWt = 1,
we search for a matching observation with ICWt = 0 within the same SIZE decile group, whose ROA value is closest to that
of the ICW observation of interest. The resulting matched sample has total 470 observations, 235 observations with ICWt = 0
and 235 observations ICWt = 1.
For Hypothesis 2, use loan loss reserves and provisions of year t+1 to investigate whether banks reporting ICW in year t
no longer have a significantly greater loan loss estimates in year t+1. We identify total 450 observations, 225 observations
with ICWt = 0 and 225 observations ICWt = 1 due to missing values of future loan loss estimates.
For the additional analysis for Hypothesis 2, we define ICW firms with successful remedial actions (NOREM = 0) as those
with ICWt = 1 and ICWt+1 = 0 and ICW firms with unsuccessful remedial actions (NOREM = 1) as those with ICWt = 1 and ICWt+1 = 1.
We divide our pooled sample into four groups to construct new matched samples based on remedial actions in addition to
ROA and SIZE. The first matched groups are ICW banks with successful remedial actions (NOREM = 0) with ICWt = 1 and ICWt+1 = 0
and their control group matched with (1) closest ROA values and (2) ICWt = 0 and ICWt+1 = 0 in the same SIZE decile group.
The second matched groups are ICW banks with unsuccessful remedial actions (NOREM = 1) with ICWt = 1 and ICWt+1 = 1
and their control group matched with (1) closest ROA values and (2) with ICWt = 0 and ICWt+1 = 1 in the same SIZE decile
group. Due to the additional restrictions in the matching process, we have in total 374 observations for the successful re-
medial action group and its control group (187 observations each) and a total of 72 observations for the unsuccessful remedial
action group and its control group (36 observations each).

4.2. Data description

Table 1A presents the descriptive statistics of the main variables. For the pooled sample, the loan loss reserves repre-
sent 1.595% of total loans on average and 1.333% for the median, while loan loss provisions are 0.723% of total loans on average
and 0.331% for the median. These statistics are bigger for the size-ROA-matched sample. Specifically, the loan loss reserves
for the matched sample are 2.254% on average and 1.608% for the median while loan loss provisions are 1.71% of total loans
on average and 0.712% for the median.
Table 1B shows that, in the pooled sample, loan loss reserves for ICW banks (mean 2.397%) are 0.826% greater than the
loan loss reserves of banks without ICW (mean 1.571%) on average, and the differences are significant at the 1% level for
both the t-test and the Wilcoxon Mann–Whitney nonparametric test. However, in the matched sample, the difference is
0.488%, and it is significantly so at the 1% level in Table 1C. The pooled sample shows that the loan loss provisions are 1.132%
66 M. Cho, K.-H. Chung / Journal of Contemporary Accounting & Economics 12 (2016) 61–72

Table 1
Descriptive statistics.

A: Descriptive statistics

Pooled sample Matched sample

Variable N Mean Median Std N Mean Median Std

LLRt 8167 1.595 1.333 1.009 470 2.254 1.608 1.863

LLPt 8167 0.723 0.331 1.155 470 1.710 0.712 2.270
NPATt 8167 2.355 1.178 3.218 470 4.619 2.222 6.170
NIMt 8167 3.518 3.550 0.952 470 3.479 3.485 0.863
ROAt 8167 0.005 0.007 0.012 470 −0.005 0.003 0.021
SIZEt 8167 7.475 6.947 1.982 470 7.491 7.244 1.345
LLRt+1 7540 1.526 1.297 1.040 450 2.049 1.442 1.972
LLPt+1 7540 0.689 0.298 1.196 450 1.211 0.465 2.001
NPATt+1 7540 2.337 1.142 3.238 450 4.235 1.925 6.224
NIMt+1 7540 3.349 3.460 1.059 450 3.229 3.335 1.066
ROAt+1 7540 0.002 0.007 0.017 450 −0.012 0.002 0.032
SIZEt+1 7540 7.353 6.889 2.061 450 7.288 7.091 1.434

B: Comparison between banks with material ICW and banks with no material ICW for the pooled sample

ICWt = 0 ICWt = 1 Differences

Variable N Mean (1) Median Std N Mean (2) Median Std (1)-(2) T-value Mann–Whitney Z stat

LLRt 7932 1.571 1.327 0.973 235 2.397 1.735 1.668 −0.826 −12.48 *** 7.89 ***
LLPt 7932 0.690 0.326 1.093 235 1.823 0.779 2.204 −1.132 −15.01 *** 8.08 ***
NPATt 7932 2.286 1.159 3.106 235 4.694 2.572 5.348 −2.408 −11.40 *** 7.63 ***
NIMt 7932 3.520 3.550 0.954 235 3.441 3.450 0.867 0.080 1.27 −1.96 **
ROAt 7932 0.005 0.007 0.011 235 −0.006 0.003 0.022 0.011 14.47 *** −9.17 ***
SIZEt 7932 7.475 6.932 2.000 235 7.468 7.261 1.256 0.007 0.05 3.10 ***
LLRt+1 7315 1.510 1.295 1.009 225 2.028 1.418 1.712 −0.518 −7.38 *** 3.14 ***
LLPt+1 7315 0.671 0.296 1.165 225 1.268 0.497 1.857 −0.597 −7.40 *** 3.60 ***
NPATt+1 7315 2.289 1.129 3.150 225 3.897 1.830 5.136 −1.608 −7.36 *** 3.54 ***
NIMt+1 7315 3.358 3.470 1.055 225 3.078 3.240 1.127 0.280 3.91 *** −4.07 ***
ROAt+1 7315 0.002 0.007 0.016 225 −0.014 0.000 0.029 0.016 13.94 *** −9.20 ***
SIZEt+1 7315 7.362 6.883 2.074 225 7.053 7.062 1.553 0.309 2.22 ** −0.29

C: Comparison between banks with material ICW and banks with no material ICW for the matched sample

ICWt = 0 ICWt = 1 Differences

Variable N Mean (1) Median Std Mean (2) Median Std (1)-(2) T-value Mann-Whitney Z stat

LLRt 235 2.010 1.519 1.709 2.499 1.735 1.980 −0.488 −2.86 *** −2.67 ***
LLPt 235 1.464 0.665 1.920 1.957 0.779 2.554 −0.493 −2.37 ** −1.78 ***
NPATt 235 4.088 1.725 5.212 5.150 2.572 6.969 −1.062 −1.87 * −1.66 **
NIMt 235 3.520 3.510 0.841 3.439 3.450 0.885 0.081 1.01 1.36 *
ROAt 235 −0.004 0.003 0.019 −0.007 0.003 0.024 0.002 1.18 0.39
SIZEt 235 7.515 7.232 1.430 7.467 7.261 1.258 0.048 0.39 0.08
LLRt+1 225 1.987 1.469 1.995 2.110 1.418 1.951 −0.123 −0.66 0.01
LLPt+1 225 1.127 0.428 1.898 1.295 0.497 2.099 −0.168 −0.89 −0.20
NPATt+1 225 4.288 1.994 6.108 4.182 1.830 6.351 0.106 0.18 0.52
NIMt+1 225 3.374 3.440 0.955 3.084 3.240 1.151 0.290 2.90 *** 2.76 ***
ROAt+1 225 −0.007 0.004 0.029 −0.017 0.000 0.034 0.009 3.08 *** 2.96 ***
SIZEt+1 225 7.411 7.138 1.449 7.165 7.062 1.412 0.246 1.83 * 1.63 *

*** , **, and * indicate the significance levels at the 1%, 5%, and 10%, respectively. The variables are defined as follows:
LLRt = loan loss reserves of year t, as a percentage of total loans;
LLPt = loan loss provisions of year t, as a percentage of total loans;
NIMt = net interest margin on loans of year t, as a percentage of total loans;
NPATRt = nonperforming assets of year t, as a percentage of total loans;
ROAt = return on assets, calculated as income before extra-ordinary items of year t, divided by total assets;
SIZEt = firm size, calculated as the natural log of total assets of year t.

greater for the ICW banks than banks with no ICW but in the matched sample ICW banks have greater loan loss provisions
than no ICW banks by 0.493%. The difference in the matched sample is smaller than in the pooled sample because the firms
are matched based on SIZE and ROA. Both samples’ differences are significant at the 1% level. The ICW banks have a signifi-
cantly lower ROA (mean −0.6%, median 0.3%) than the general population with no ICW (mean 0.5%, median 0.7%), according
to Table 1B.
The correlation coefficients in the matched sample in Table 2 show that ICWt is significantly and positively related to
LLRt (Pearson 0.131, Spearman 0.123) at the 1% level and to LLPt (Pearson 0.109, Spearman 0.082) at the 5% level. Thus, de-
scriptive statistics in Table 1 and correlation coefficients in Table 2 indicate that banks with ICWt =1 tend to have greater
loan loss reserves (LLRt) and provisions (LLPt), which is consistent with Hypothesis 1.
 M. Cho, K.-H. Chung / Journal of Contemporary Accounting & Economics 12 (2016) 61–72 67

Table 2
Correlation coefficients of the matched sample (N = 470).

ICWt LLRt LLPt NPATt NIMt ROAt SIZEt

ICWt 1 0.131 0.109 0.086 −0.047 −0.054 −0.018

(0.004) (0.018) (0.062) (0.311) (0.239) (0.699)
LLRt 0.123 1 0.772 0.566 −0.206 −0.579 0.144
(0.007) (<.0001) (<.0001) (<.0001) (<.0001) (0.002)
LLPt 0.082 0.710 1 0.693 −0.120 −0.791 −0.025
(0.075) (<.0001) (<.0001) (0.009) (<.0001) (0.592)
NPATt 0.077 0.704 0.762 1 −0.194 −0.671 −0.153
(0.096) (<.0001) (<.0001) (<.0001) (<.0001) (0.001)
NIMt −0.063 −0.032 −0.055 −0.117 1 0.317 −0.110
(0.175) (0.490) (0.236) (0.011) (<.0001) (0.017)
ROAt −0.018 −0.579 −0.728 −0.702 0.348 1 0.127
(0.696) (<.0001) (<.0001) (<.0001) (<.0001) (0.006)
SIZEt −0.003 −0.037 −0.078 −0.231 −0.005 0.243 1
(0.940) (0.422) (0.093) (<.0001) (0.914) (<.0001)

The Pearson correlation coefficients are above the diagonal, and the Spearman correlation coefficients are below the diagonal. The variables are defined
in Table 1.

4.3. Internal control weaknesses and excessive loan loss estimates

If banks with material ICW do not perform loan loss assessments properly or stock up on supplemental reserves without
legitimate needs, we expect that firms with ICW tend to have inflated loan loss reserves and provisions. In Table 3A, Model
2 reports that loan loss reserves (LLRt) have a significantly positive coefficient of ICWt (0.32) at the 5% level, indicating that
ICW banks indeed have greater loan loss reserves than the control sample, supporting Hypothesis 1. As an additional test
for Hypothesis 1, we run Regression (1) without ICWt in Model 1. We run a Fisher’s F-test and estimate the improvement of
model fit from Model 1 to Model 2. The F-statistics is 6.27, significant at the 1% level, indicating ICWt explains a significant
portion of the variation of LLRt.

Table 3
The effect of material ICW on loan loss reserves (LLRt) and provisions (LLPt).

Model 1 Model 2

Coefficient T-value Coefficient T-value

A: Dependent variable = LLRt

Intercept −1.049 −2.07** −1.232 −2.42**
ICWt 0.320 2.50**
NIMt 0.019 0.24 0.025 0.32
NPATt 0.107 7.60*** 0.104 7.46***
ROAt −32.548 −7.81*** −32.691 −7.89***
SIZEt 0.343 7.02*** 0.344 7.09***
Year effect Controlled Controlled
adj. R-sq. 45.13% 45.86%
N 470 470
Fisher’s test F-statistics 6.27***
B: Dependent variable = LLPt
Intercept −2.187 −4.83*** −2.335 −5.13***
ICWt 0.260 2.27**
NIMt 0.418 5.91*** 0.424 6.01***
NPATt 0.112 8.89*** 0.110 8.76***
ROAt −69.064 −18.56*** −69.180 −18.67***
SIZEt 0.207 4.75*** 0.208 4.79***
Year effect Controlled Controlled
adj. R-sq. 70.53% 70.85%
N 470 470
Fisher’s test F-statistics 5.19***

*** , **, and *indicate the significance levels at the 1%, 5%, and 10%, respectively. The variables are defined as:
LLRt = loan loss reserves of year t, as a percentage of total loans;
LLPt = loan loss provisions of year t, as a percentage of total loans;
ICWt = an indicator variable of material internal control weakness in year t, set as 1 if there is material internal control weakness in year t, zero otherwise;
NIMt = net interest margin on loans of year t, as a percentage of total loans;
NPATRt = nonperforming assets of year t, as a percentage of total loans;
ROAt = return on assets, calculated as income before extra-ordinary items of year t, divided by total assets;
SIZEt = firm size, calculated as the natural log of total assets of year t;
FYt = fiscal year t; and,
εt = an error term.
68 M. Cho, K.-H. Chung / Journal of Contemporary Accounting & Economics 12 (2016) 61–72

We report similar results for LLPt in Table 3B. The coefficient of ICWt (0.26) is significantly positive at the 5% level in Model
2. We run Regression (1) without ICWt with LLPt as our dependent variable in Model 1, and the resulting Fisher’s F-statistics
(5.19) shows that ICWt explains a significant portion of the variation of LLPt as well. Overall, the results in Table 3 show that
banks with material ICW have greater loan loss reserves and provisions than the matched sample without ICW, which sup-
ports Hypothesis 1.
An alternative explanation for the high loan loss reserves and provisions of banks is private information signaling. There-
fore, we investigate private information signaling of ICW banks in particular by dividing the 235 observations with ICWt =1
into two groups – above the median loan loss reserves (LLRt) or provisions (LLPt) and below the median. We match the two
groups with their control sample separately based on ROA and SIZE following the same matching steps described earlier
and obtain 470 observations combined. We test whether the ICW banks with high loan loss estimates (HiLLRt =1 or HiLLPt
=1) experience greater improvement of earnings in year t+1 (ΔEBTPt+1), consistent with private information signaling, in the
following regressions:
ΔEBTPt+1 = θ 0 + θ1ICWt + θ 2HiLLR t ⋅ ICWt + θ 3NIMt + θ 4NPATt + θ5ROA t + θ 6SIZEt + ∑t θ7,tFYt + ε t (4a)

ΔEBTPt+1 = θ 0 + θ1ICWt + θ 2HiLLPt ⋅ ICWt + θ 3NIMt + θ 4NPATt + θ5ROA t + θ 6SIZEt + ∑ t θ7,tFYt + ε t (4b)

where ΔEBTPt+1 = changes in earnings before tax and loan loss provisions in year t+1, following Wahlen (1994) and Ahmed
et al. (1999);
HiLLRt = set as 1 if LLRt is greater than its median, and zero otherwise; and
HiLLPt = set as 1 if LLPt is greater than its median, and zero otherwise.

We control for nonperformance loans (NPATt) following Wahlen (1994) and control for net interest margin on loans (NIMt),
return on assets (ROAt) and firm size (SIZEt).11 In this regression, we compare whether ICW banks have better future earn-
ings performances than non-ICW banks with a positive coefficient of ICWt and whether it is more so among ICW banks with
high loan loss estimates than ICW banks with low loan loss reserves with a positive coefficient of the interaction term
(HiLLRt·ICWt or HiLLPt·ICWt).
Table 4A reports an insignificant coefficient of ICWt. More importantly, the coefficient of the interaction term, HiLLRt·ICWt,
is not significantly different from zero, contrary to the signaling scenario. Table 4B presents similar results for loan loss pro-
visions (LLPt). Overall, we do not find evidence that ICW banks use loan loss estimates to signal private information.

4.4. Remedial actions of material internal control weaknesses

For Hypothesis 2, we run Regression (2) to see whether ICW firms continue to have greater loan loss reserves or provi-
sions in year t+1 after remedial actions. If banks reporting ICW in year t take remedial actions to improve their internal
control procedures, we expect that they no longer have greater loan loss reserves (LLRt+1) and provisions (LLPt+1) in year t+1.
The sample size of the matched sample is reduced to 450 observations from 470 due to the missing values of LLRt+1 and
LLPt+1. Table 5A reports that banks with ICW in year t indeed no longer have a significant coefficient of ICWt in year t+1 for
both loan loss reserves (LLRt+1) and provisions (LLPt+1), supporting Hypothesis 2.
However, it is possible that not all banks succeed in improving their ICW problems in the following year. We thus divide
our sample into (1) banks with material ICW in year t but no ICW in year t+1, i.e. the banks that succeed in remedying ICW
(NOREM = 0) and (2) the banks with material ICW reports in both years, i.e. the banks that fail to remedy ICW (NOREM = 1).
We match observations with successful remedies (NOREM = 0 where ICWt = 1, and ICWt+1 = 0) with those with no ICW in both
years (ICWt = 0, and ICWt+1 = 0) and observations with unsuccessful remedies (NOREM = 1 where ICWt = 1, and ICWt+1 = 1) with
those newly reporting ICW in year t+1 (ICWt = 0, and ICWt+1 = 1). We expect that firms with successful remedial actions
(NOREM = 0) no longer have a higher level of loan loss reserves or provisions than their control sample because both groups
do not report material ICW in year t+1. However, banks with no remedial actions (NOREM = 1) continue to have higher levels
of loan loss reserves or provisions in year t+1 because they report material ICW in both years while their control group starts
to report ICW in year t+1.
Table 5B presents the results of Regression (3), showing that both loan loss reserves (LLRt+1) and provisions (LLPt+1) in
year t+1 have insignificant coefficients of ICWt, indicating that banks with successful remedial actuations no longer have
higher loan loss reserves or provisions, supporting Hypothesis 2. The coefficient of the interaction term NOREM·ICWt is sig-
nificantly positive (0.644) for loan loss reserves (LLRt+1) at the 1% level, indicating that banks with unsuccessful remedies
continue to have greater loan loss reserves than their control group. These results support Hypothesis 2. However, it is not
significant for loan loss provisions (LLPt+1).

11 Our results remain qualitatively unchanged even when we use control variables of year t+1.
 M. Cho, K.-H. Chung / Journal of Contemporary Accounting & Economics 12 (2016) 61–72 69

Table 4
Private information signaling of ICW banks with loan loss estimates.

Coefficient T-value

A: predicting future earnings (ΔEBTPt+1) with loan loss reserves (LLRt) by ICW banks
Intercept −0.015 −3.36***
ICWt 0.000 −0.12
HiLLRt·ICWt 0.003 1.50
NIMt 0.002 2.62***
NPATt 0.000 −1.65*
ROAt 0.054 2.06**
SIZEt 0.001 1.45
Year effect Controlled
adj. R-sq. 7.70%
N 470
B: predicting future earnings (ΔEBTPt+1) with loan loss reserves (LLPt) by ICW banks
Intercept −0.015 −3.32***
ICWt 0.000 0.16
HiLLPt·ICWt 0.002 1.03
NIMt 0.002 2.59**
NPATt 0.000 −1.57
ROAt 0.054 2.03**
SIZEt 0.001 1.42
Year effect Controlled
adj. R-sq. 7.45%
N 470

*** , **, and * indicate the significance levels at the 1%, 5%, and 10%, respectively. The variables are defined as:
ΔEBTPt+1 = changes in earnings before tax and loan loss provisions in year t+1, following Wahlen (1994) and Ahmed
et al. (1999);
HiLLRt = set as 1 if LLRt is greater than its median, and zero otherwise; and
HiLLPt = set as 1 if LLPt is greater than its median, and zero otherwise.
Other variables are defined in Table 3.

Table 5
The effect of material ICW on loan loss estimates of year t+1 after remedial actions.

Dependent variable = LLRt+1 Dependent variable = LLPt+1

Coefficient T-value Coefficient T-value

A: Loan loss reserves (LLRt+1) and provisions (LLPt+1) in year t+1

Intercept −2.238 −4.29*** −3.686 −7.23***
ICWt 0.194 1.33 0.159 1.11
NIMt+1 0.002 0.03 0.299 3.81***
NPATt+1 0.179 15.07*** 0.174 14.95***
ROAt+1 −6.643 −2.26** −22.763 −7.91***
SIZEt+1 0.459 8.19 0.390 7.12***
Year effect Controlled Controlled
adj. R-sq. 40.42% 44.70%
N 450 450
B: Differences between firms with NOREM = 0 and NOREM = 1
Intercept −1.828 −4.29*** −4.016 −7.91***
ICWt 0.156 1.27 −0.030 −0.2
NOREM·ICWt 0.644 2.94*** 0.203 0.78
NIMt+1 0.246 3.77*** 0.415 5.32***
NPATt+1 0.173 18.19*** 0.173 15.25***
ROAt+1 −8.230 −3.09*** −29.923 −9.43***
SIZEt+1 0.285 6.22*** 0.390 7.13***
Year effect Controlled
adj. R-sq. 51.57% 50.26%
N 446 446

*** , **, and * indicate the significance levels at the 1%, 5%, and 10%, respectively.
The variables are defined as:
NOREM = an indicator variable for banks unsuccessful remedial actions after ICW in year t, set as 1 if ICWt = 1 and ICWt+1 = 1
and the control firms with ICWt = 0 and ICWt+1 = 1; and zero if ICWt = 1 and ICWt+1 = 0 and the control firms with ICWt = 0
and ICWt+1 = 0.
The other variables are defined in Table 3. The number of observations with non-missing LLRt+1 and LLPt+1 variables is
450, of which 225 observations are with ICWt =1 and 225 observations with ICWt =0. For the results presented in B,
we have 446 observations due to more restrictions imposed in the matching process.
70 M. Cho, K.-H. Chung / Journal of Contemporary Accounting & Economics 12 (2016) 61–72

Table 6
Heckman’s two stage estimation of the effect of material ICW on loan loss reserves (LLRt).

Coefficient T-value

A: The first stage estimation: propensity of ICW (dependent variable = ICWt)

Intercept 0.472 1.01
LLPt 0.168 3.49***
NIMt −0.130 −1.74*
NPATt 0.001 0.06
ROAt 12.325 2.52**
SIZEt −0.033 −0.74
Year effect Controlled
Chi- Sq. 188.16***
N 470
B: The second stage estimation: the impact of ICW on loan loss reserves
(dependent variable = LLRt)
Intercept −1.203 2.54**
ICWt 0.323 2.54**
NIMt 0.013 0.16
NPATt 0.112 7.68***
ROAt −32.037 −7.72***
SIZEt 0.342 7.07***
INVMILLS 0.000 −1.81*
Year effect Controlled
adj. R-sq. 45.55%
N 470

*** , **, and * indicate the significance levels at the 1%, 5%, and 10%, respectively. The
variables are defined in Table 3.

Overall, Hypothesis 2 is supported and we conclude that banks reporting ICW are more likely to succeed in improving
the ICW problem in the following year and banks with successful remedial actions no longer have greater loan loss re-
serves or provisions in the following year.

4.5. Sensitivity tests

4.5.1. Self-selection problem

While we find that banks with material ICW are more likely to have higher levels of loan loss reserves and provisions
than the control group, our tests may have a self-selection bias. It is possible that banks with relatively more loan loss re-
serves and provisions have a greater propensity to report ICW because such banks are more likely to manipulate earnings
(Ahmed et al., 1999; Kanagaretnam et al., 2004). To control for this self-selection bias, we estimate the coefficient of ICWt
by running a Heckman’s two-stage estimation. The first-stage probit model estimates the propensity of reporting material
ICW that is explained by loan loss reserves (LLRt) or provisions (LLPt). The second-stage model estimates the coefficient of
ICWt in explaining loan loss reserves or provisions with the inversed Mills ratio (INVMILLS) as an additional control variable.
Table 6A reports the probit model results on the propensity of ICW. It shows that banks with greater loan loss provi-
sions indeed tend to report more ICW in that year with a significantly positive coefficient (0.168) at the 1% level. Table 6B
shows that our main results remain unchanged in the second-stage model. Even if we control for the self-selection bias,
ICW continues to play a significant role in explaining loan loss reserves (LLRt) with a positive coefficient (0.323) at the 5%
level, consistent with the results in Table 3.
We perform a Heckman’s two-stage estimation for loan loss provisions (LLPt) with ICW to control for the self-selection
bias and present the results in Table 7. Table 7A shows that ICW banks tend to have greater loan loss reserves with a sig-
nificantly positive coefficient (0.280) at the 1% level in the first-stage probit model. Nonetheless, our main results still hold
according to Table 7B, in that the coefficient of ICWt is significant and positive (0.269) at the 5% level in explaining the vari-
ation of loan loss provisions (LLPt) even after controlling for the self-selection bias.

4.5.2. Endogeneity problem

While banks with material ICW tend to have greater loan loss reserves and provisions, it is also possible that banks with
greater loan loss reserves and provisions are more likely to report material ICW. To control for this endogeneity problem,
we perform the simultaneous two-stage least squares estimation using the probit model used in the Heckman’s two-stage
estimation. Our results (untabulated) remain qualitatively unchanged even if we control for the endogeneity problem.

4.5.3. Tests in the pooled sample

Our results are based on the SIZE-and-ROA-matched sample. As a sensitivity test, we run our tests using the pooled sample
to examine whether our results hold in the general population. Table 8 reports that ICW banks (235 observations) have sig-
 M. Cho, K.-H. Chung / Journal of Contemporary Accounting & Economics 12 (2016) 61–72 71

Table 7
Heckman’s two stage estimation of the effect of material ICW on loan loss provisions
(LLPt).

Coefficient T-value

A: The first stage estimation: propensity of ICW (dependent variable = ICWt)

Intercept 0.122 0.27
LLRt 0.280 7.40***
NIMt −0.065 −0.93
NPATt 0.001 0.06
ROAt 6.350 1.56
SIZEt −0.072 −1.57
Year effect Controlled
Chi-Sq. 410.14***
N 470
B: The second stage estimation: the impact of ICW on loan loss provisions
(dependent variable = LLPt)
Intercept −2.284 −5.13***
ICWt 0.269 2.41**
NIMt 0.394 5.70***
NPATt 0.126 9.88***
ROAt −68.160 −18.76***
SIZEt 0.206 4.86***
INVMILLS 0.000 −4.63***
Year effect Controlled
adj. R-sq. 71.78%
N 470

*** , **, and * indicate the significance levels at the 1%, 5%, and 10%, respectively. The
variables are defined in Table 3.

Table 8
The effect of ICW on loan loss reserves and provisions in the pooled sample.

Dependent variable = LLRt Dependent variable = LLPt

Coefficient T-value Coefficient T-value

Intercept −0.255 −4.55*** −0.914 −17.27***

ICWt 0.284 5.46*** 0.218 4.44***
NIMt 0.161 16.07*** 0.206 21.86***
NPATt 0.158 49.00*** 0.104 34.29***
ROAt −14.820 −15.46*** −60.144 −66.45***
SIZEt 0.132 28.70*** 0.130 29.85***
Year effect Controlled Controlled
adj. R-sq. 41.09% 60.39%
N 8167 8167

*** , **, and * indicate the significance levels at the 1%, 5%, and 10%, respectively. The variables are defined in Table 3.

nificantly greater loan loss reserves (the coefficient of ICWt = 0.284) and provisions (the coefficient of ICWt =0.218) compared
to banks with no ICW (7932 observations) at the 1% level, strongly supporting our main results.

5. Conclusion

This study examines whether the internal control effectiveness of commercial banks affects their loan loss reserves and
provisions. The credit risk exposure of banks is reflected in loan loss reserves and thus loan loss provisions, and inaccurate
loan loss estimates can serve as a tool for fraudulent conduct by bank management. We expect and find that firms with
ICW tend to have inflated loan loss reserves and provisions, indicating that internal control effectiveness is an important
factor of loan loss estimates. We do not find evidence of private information signaling among ICW banks with respect to
loan loss estimates. We also find that ICW firms continue to have greater loan loss reserves and provisions in the following
year if their material weaknesses are not remedied. However, if banks succeed in improving their ICW problems, they no
longer have greater loan loss reserves or provisions in the next year. Our results hold after controlling for the self-selection
bias and the endogeneity problem.
In conjunction with the increase in the regulatory supervision over commercial banks after the 2007–2008 financial crisis,
our study has significant contributions to the accounting literature by providing timely and relevant insights on banks’ in-
ternal control effectiveness. Specifically, we investigate the role of ICW as a determinant of loan loss reserves and provisions
while prior studies have not considered the quality of the loan loss evaluation process reflected in the ICW reports as one
of the determining factors of loan loss estimates. Our findings show that ICW is directly related to the banks’ loan loss re-
72 M. Cho, K.-H. Chung / Journal of Contemporary Accounting & Economics 12 (2016) 61–72

serves and provisions. In addition, we show that banks that take remedial actions of ICW cease to stack unnecessary loan
loss reserves. This result indicates that the regulatory requirements of ICW successfully prompt the improvement of inter-
nal control procedures among banks. Lastly, unlike prior studies based on proxy measures of ICW, our results are based on
the actual material ICW reports among banks.

Acknowledgements

The authors would like to thank Samir El-Gazzar, Donald Kent, and Dan Simunic (the editor), the anonymous referee
and the seminar participants at the 2015 American Accounting Association Northeast Regional Meetings for their com-
ments. The financial support of the Lubin School of Business of the Pace University is gratefully acknowledged.

References

Ahmed, A., Takeda, C., Thomas, S., 1999. Bank loan loss provisions: a reexamination of capital management, earnings management, and signaling effects.
J. Account. Econ. 28, 1–25.
Altamuro, J., Beatty, A., 2010. How does internal control regulation affect financial reporting? J. Account. Econ. 49 (1), 58–74.
Ashbaugh-Skaife, H., Collins, D.W., Kinney, W.R., Jr., LaFond, R., 2009. The effect of SOX internal control deficiencies on firm risk and cost of equity capital.
J. Account. Res. 47 (1), 1–43.
Beaver, W.H., Engel, E.E., 1996. Discretionary behavior with respect to allowances for loan losses and the behavior of security prices. J. Account. Econ. 22,
177–206.
Beneish, M.D., Billings, M.B., Hodder, L.D., 2008. Internal control weakness and information uncertainty. Account. Rev. 83 (3), 665–703.
Chan, K.C., Farrell, B.R., Lee, P., 2008. Earnings management of firms reporting material internal control weaknesses under Section 404 of the Sarbanese-Oxley
Act. Auditing 27 (2), 161–179.
Chen, G.T., Chung, K.-H., El-Gazzar, S., 2005. Factors determining commercial banks’ allowance for loan losses. Com. Lending Rev. 20 (2), 25–32.
Doyle, J., Ge, W., McVay, S., 2007. Determinants of weaknesses in internal control over financial reporting. J. Account. Econ. 44 (1&2), 193–223.
GAO, 1991. Failed banks: accounting and auditing reforms urgently needed. United States General Accounting Office – Report to Congressional Committees.
GAO, 1994. Depository institutions: divergent loan loss methods undermine usefulness of financial reports. United States General Accounting Office – Report
to Congressional Committees.
Ge, W., McVay, S., 2005. The disclosure of material weaknesses in internal control after the Sarbanes–Oxley Act. Account. Horiz. 19 (3), 137–158.
Hammersley, J.S., Myers, L.A., Shakespeare, C., 2008. Market reactions to the disclosure of internal control weaknesses and to the characteristics of those
weaknesses under Section 302 of the Sarbanes–Oxley Act of 2002. Rev. Account. Stud. 13 (1), 141–165.
Johnstone, K., Li, C., Rupley, K.H., 2011. Changes in corporate governance associated with the revelation of internal control material weaknesses and their
subsequent remediation. Contemp. Account. Res. 28 (1), 331–383.
Kanagaretnam, K., Mathieu, R., Lobo, G., 2003. Managerial incentives for income smoothing through bank loan loss provisions. Rev. Quant. Finance Account.
20, 63–80.
Kanagaretnam, K., Yang, D., Lobo, G., 2004. Joint tests of signaling and income smoothing through bank loan loss provisions. Contemp. Account. Res. 21,
843–884.
Kanagaretnam, K., Yang, D., Lobo, G., 2005. Determinants of signaling by banks through loan loss provisions. J. Bus. Res. 58, 312–320.
Kanagaretnam, K., Krishnan, G., Lobo, G., 2009. Is the market valuation of banks’ loan loss provision conditional on auditor reputation? J. Bank. Finance
33, 1039–1047.
Kilic, E., Lobo, G., Ranasinghe, T., Sivaramakrishnan, K., 2013. The impact of SFAS 133 on income smoothing by banks through loan loss provisions. Account.
Rev. 88, 233–260.
Kim, J., Song, B., Zhang, L., 2011. Internal control weakness and bank loan contracting: evidence from SOX Section 404 disclosures. Account. Rev. 86 (4),
1157–1188.
Krishnan, G.V., Visvanathan, G., 2007. Reporting internal control deficiencies in the post-Sarbanese-Oxley era: the role of auditors and corporate governance.
International J. Audit. 11 (2), 73–90.
Li, C., Sun, L., Ettredge, M., 2010. Financial executive qualifications, financial executive turnover, and adverse SOX 404opinions. J. Account. Econ. 50 (1),
93–110.
Ogneva, M., Raghunandan, K., Subramanyam, K.R., 2007. Internal control weakness and cost of equity: evidence from SOX Section 404 certification. Account.
Rev. 82 (5), 1255–1297.
Ryan, T. 2014. Fed launches review of practices for supervising big banks: fed review comes amid accusations central banks stifles dissent. The Wall Street
Journal, Nov 21, 2014.
Wahlen, J., 1994. The nature of information in commercial bank loan loss disclosures. Account. Rev. 69 (3), 455–478.