Loan loss provisioning in the commercial banking system of Barbados: practices and determinants
Loan
loss provisioning in the commercial banking system of Barbados: practices and
determinants
Loan Loss
Provisioning in the Commercial Banking System of Barbados: Practices and
Determinants
By
Roland Craigwell 1
Professor of Economics
University
of the West Indies
Cave
Hill Campus And
Research
Associate
Central
Bank of Barbados
And
Wayne Elliott
Undergraduate
Student
University
of the West Indies
Cave
Hill Campus
And
Summer
Intern
Central
Bank of Barbados
November
2010
Loan Loss
Provisioning in the Commercial Banking System of Barbados: Practices and
Determinants
Roland Craigwell and Wayne Elliott
Abstract
The purpose of this paper is to investigate
the process of loan loss provisioning within the commercial banking system of
Barbados. It uses questionnaires and interviews to ascertain how banks set
their provisional standards and levels. In addition, the results from this
approach reveal, for the first time in Barbados, the individual banks‟
procedures for loan loss provisioning. An evaluation of the impact of
macroeconomic and bank specific factors on commercial banks‟ provisions
utilising panel dynamic ordinary least squares is also undertaken. Both sets of
factors are found to influence the level of provisions. In particular, loan
loss provisions are heavily dependent upon the performance of the real economy
and competition in international markets is shown to have serious implications
for the banking sector in both the short and long run. Moreover, this study
asserts that larger banks in Barbados are better able to screen loans and avoid
defaults.
Keywords: Loan Loss Provisioning;
Banking System; Loan Classification;
JEL: G21, G28, M41
Introduction
A loan loss provision is a charge to commercial banks‟
profit and loss statements that creates a reserve on their balance sheets. It can be viewed as a cushioning mechanism
which may ensure that banks do not unexpectedly lose their entire outstanding
loan balances. Without this adjustment, the amount of loans and advances on the
balance sheets of banks would include possible future losses. Furthermore,
regulators, creditors and investors could be misled by overstated capital
figures.
One of the major consequences pertaining to the financial
crisis of 2007 is to raise concerns about the need for closer control and
supervision on lending institutions. In Barbados, the Central Bank has started
to review the regulatory framework of these activities with the goal of
minimising the adverse effects of future crises. International regulators
(International Monetary Fund, World Bank, and the Bank for International
Settlements) have suggested several reforms, and programs aimed at
strengthening the international banking and financial system. One key element
of these reforms highlighted by the regulators is to assess the process of loan
loss provisioning.
Given the above, this study evaluates the loan loss
provisioning process of the banking system of the small open economy of
Barbados. It is the first of its kind to be undertaken on this country and
possibly, on the wider Caribbean. Two methodologies are used: interviews with
the main objective of ascertaining how banks provide for loan losses, and an
econometric model which will explore the main factors driving the level of
provisions.
The structure of the paper is as follows: Section 2 gives some
background information on loan loss provisions.
Section 3 presents the methodologies while section 4 describes the
results derived from employing the various methods. Section 5 concludes with a
summary of the findings, including limitations and policy implications.
2.
Background
On account of many different factors, the principle and
interest that banks agree upon with their creditors are not always collected.
As a result, they have to apply a technique to soften the impact of such
losses: loan loss provisioning. To make the provisioning process consistent
across different jurisdictions, banks use many different means, of which a
major one is utilising loan loss provisioning standards.
2.1
Loan Loss Provisioning Standards
The Basel 1 Capital Accord was released in 1988 seeking to
create a competitive international banking system by strengthening its
stability and setting up fair and consistent standards. It consists of three
mutually reinforcing pillars - minimum capital requirements, the supervisory
review process and market discipline - which together contribute to the safety
and soundness of the financial system.
The Basel Committee (1999) emphasizes that there are no
uniformed loan classifications, nor is there a standard procedure to assess
loan risk across all jurisdictions. It further posits that several concepts of
loan loss provisioning are susceptible to different interpretations. To deal with these problems, diverse systems
are utilised in different countries (Laurin and Majnoni, 2003). Also, there was
a move to introduce more suitable standards, for example, The International
Accounting Standards (IAS), to resolve some of these
issues. According to Laurin and Majnoni (2003) these standards are yet to give
detailed guidance for loan loss provisioning. Due to the shortcomings of these
accounting standards, countries that implement the IAS still have different
loan loss provisioning regulatory frameworks. For instance, sometimes banks
need to apply more complex internal classification systems, while more
standardised systems are required by bank regulators for reporting purposes.
In concurrence, the Central Bank of Barbados, as a
regulator, uses the Financial Institutions Act 1998 which is an adaptation of
Basel Accord 1. Prior to the current Act, financial institutions were governed
by the Financial Intermediaries Regulatory Act (1992) which was predated by the
Banking Act (1978). The Financial
Institutions Act (1998) governs all financial institutions in Barbados. It is
the benchmark for loan loss provisioning regardless of the standards the
individual commercial banks choose to employ. When reporting to the Central
Bank of Barbados, banks must ensure that all information concerning loan loss
provisioning is in accordance with the Financial Institutions Act 1998. However, the flexibility given to commercial
banks through their classification processes, limits the use of penalties and
sanctions that could be utilized to regulate inappropriate classification and
provisioning (Laurin and Majnoni, 2003). Regulators therefore rely mainly on
moral suasion and the threat of sanctions rather than specific penalties to
enforce the regulations.
2.2
Classification Criteria
The provisioning criteria employed to estimate losses in
the loan portfolio are one of the key ways to assess the adequacy of loan loss
provisioning levels and reserves. An inaccurate classification of individual
loans destabilizes the process of provisioning, which, in turn, distorts the
figures on the balance sheet and overstate capital and capital ratios
(Angklomkliew et al, 2009). In support of this view, Song (2002) adds that
there ought to be one set of well-known rules governing the provisioning
process. He states that the classification of outstanding loans should be based
on a comprehensive assessment of whether or not the borrower is able to service
the debt, rather than on the outstanding loan value or the collateral provided.
The Basel Committee (1999) supports this viewpoint but also believes that the
current value of collateral and the ability to realize this value should be
included with other criteria that banks deem necessary.
The main method of classifying loans is by the time that
has elapsed since the last payment was due. The longer the time past the last
due date, the less likely it is for commercial banks to recover the entire
balance and the greater the banks‟ provision for the outstanding balance. This
procedure is used mainly as a trigger, after which further evaluation is undertaken.
Relying only on the „time past the due date‟ approach would cause losses to be
recognized later in the loan analysis rather than earlier.
Many countries have become aware of the inefficiencies of
the „time past due‟ methodology and agree that more forward-looking criteria -
borrower‟s cash flow and repayment capacity - should be taken into account.
Consequently, some commercial banks now look at the borrower‟s behavior as a
proxy for their willingness to repay a loan. An example of this is a case where
a borrower misses a few payments, the commercial bank contacts him but he gives
no suitable reason for his delinquency. This indicates that a default may be in
the near future and that a provision should be made.
Another major difference in loan loss provisioning
practices across jurisdictions, as pointed out by Cortavarria et al (2000), is
the classification of restructured loans. Restructured loans are loans that
have been modified to better ensure the loan balance is recovered. In the Czech
Republic, for instance, bankers may characterize restructured loans as
substandard while in
Thailand they may be placed in the „special mention‟ or
pass category. However, the reclassifications of the restructured loans may not
be implemented immediately but may be undertaken over a period of time when
bankers are satisfied that borrowers are meeting the terms of the new
contracts. Bearing this in mind, credit analysts should ensure that
restructuring is based on sound underwriting standards because the cost to
recover collateral is time sensitive. In other words, if the collateral was
collected early, and the option of restructuring not issued, then the bank may
have recovered more of the outstanding loan balance (Cortavarria et al, 2000).
Just as there are different ways to characterize loans and
various categories for loans to be classified into, there are several types of
loan loss provisions. The type of provisioning in one economic community may
not be suitable in another.
2.3
Types of Loan Loss Provisioning
Balla and Mckenna (2009) highlight two broad categories of
loan loss provisioning: the dynamic loan loss provisioning procedure and the
traditional incurred loss method.
Dynamic provisioning is a statistical method that utilizes the
historical data of various asset classes. It determines the event driven
provisioning periodically. It is a deliberate way to construct the loan loss
reserve in good economic times. The built up reserve then eases pressures on
earnings and capital by absorbing loan losses during an economic downturn.
Conversely, the incurred loss method delays provisioning until the economy
fails to grow; it may magnify the bust because most bad loans will only reveal
themselves during recessions. In essence,
the key difference between the two categories of loan loss provisioning is not
the level of provisioning but its timing (Balla and Mckenna, 2009).
Bouvatier and Lepetit (2006) also distinguished between
non-discretionary and discretionary loan loss provisioning. The
non-discretionary component is designed to cover expected loan losses in the
banks‟ loan portfolio. The authors state that this component drives the
cyclicality of loan loss provisioning and it leads to a misevaluation of
expected credit losses. The discretionary component is caused by management‟s
use of loan loss provisioning for its own objectives.
Bouvatier and Lepetit (2006) give at least three functions
for which banks‟ management utilises loan loss provisioning (see also Bikker
and Metzemakers, 2002). The first function is the practice of earnings
management where banks reserve more in good years to cover for bad years. This
effectively raises and lowers income, and by extension, profits and dividends,
as desired. The second is the management of the capital ratio which is possible
on account of loan loss provisions being a part of the regulatory capital
depending on the stipulations of the territory.
The third function is tax evasion and this is common
because provisions are tax deductible in most countries.
In addition, Bikker and Metzemakers (2002) argue that the
propensity to use discretionary loan loss provisioning to smooth income is
greater for banks with good performance relative to banks with moderate
performance. The authors concluded that, during upswing phases, weak specific
provisions encourage the expansion of credit. With sudden downturns, the
identification of problem loans constrains the banks to make provisions,
thereby reducing their ability to provide new credit. Poorly capitalized banks
appear more restricted by provisioning. Dziobek et al (2000) also suggest that,
if loan loss provisions count as regulatory capital and are tax deductible,
management has a greater incentive to use loan loss provisions which should
lower the tax burden. Loan loss provisions could also inadvertently signal
financial strength. A bank‟s level of loan loss provisions and reserves may
indicate that it is strong enough to withstand a financial blow to the loan
portfolio.
Using the theoretical model by Cavallo and Majnono (2002),
Bikker and Metzemakers (2002) examined balance sheet data of 29 Organization
for Economic Co-operation and Development (OECD) countries and found that banks
that hold a greater amount of risky loans act somewhat prudently and provision
more. There was also evidence in support of the capital management hypothesis
that banks lend more when their capital ratios are low. In addition, it emerged
that provisions depend strongly on credit risk. In harmony with most previous
authors (for example, see Laeven and Majnoni, 2003; Floro, 2010; Jiménez and Saurina,
2006; Wezel, 2010),
the evidence confirmed that provisions depend significantly on the business
cycle. However, Bikker and Metzemakers (2002) could not support the hypothesis
that an expansion in provisioning during successive years of economic boom
resulted in higher reserves after the year of recession. There was also no
evidence in favour of the hypothesis that the erosion of reserves after years
of consecutive recession was due to increased provisioning in years of economic
boom.
2.4
How much to Provide for Loan Losses?
Another key question that needs to be answered in order to
complete the provisioning process is how much to provide for each category of
loans. As described by the Financial Institutions Act 1998, following the
annual review of the loan portfolio, each loan classification category should
be assigned minimum levels of provisions.
As a result, the substandard, doubtful and loss categories are considered
as classified debt and are provided for whilst pass, special Mention and
mortgages (up to six months) are not. Therefore, reserves are comprised of the
provisions on total categorical classified debt; 10% for the total substandard
loans value, 50% for total doubtful and 100% for loss loan amounts. There is also a 1% provision that is made for
the balance of the loan portfolio which was not reviewed in the past 12 months.
One of the criteria, which is critical during the analysis and categorisation
of loans, is the security that backs the loan.
2.5
Collateral
Collateral in its simplest definition is a form of security
to a lender in case the borrower fails to repay a loan. It plays an important
role in the financial sector, as it is a means of covering potential losses.
The Basel Committee (1999) recognizes the importance of collateral and advises
that the loan classification process should take the ability to realize
collateral into consideration.
There are many different approaches concerning whether and
how collateral should be viewed when classifying loans and determining their
appropriate provisions (Laurin and Majnoni, 2003). All regulatory frameworks do
not guarantee the acceptance of some forms of collateral. This contributes to
the problems in comparing the loan loss provisioning processes across different
jurisdictions. Once there is acceptance of some form of security, Song (2002)
posits that when classifying a troubled loan, it is reasonable that a
conservative value of the collateral be taken into account, instead of just the
value of the collateral. This conservative value represents a truer picture of
the amount that can be recovered, after taking the cost of collection into
consideration. Jokivuolle and Peura (2003), in an attempt to provide insight
into the estimation of the credit risk quantity, highlight the sensitivity of
using collateral as a source of recovery. For example, if a company is unable
to meet its debts and goes into default, its assets, which are not high enough
to cover its liabilities, are usually the same ones that would have been
pledged as collateral.
Collateral also plays a major role when making decisions
about the amount of provision to be made on an impaired loan. The question that
arises is whether or not to „net off‟ the collateral value against the impaired
loan amount before a provision is made. Arguments against „netting off‟ refer
mainly to deriving the collectible value for collateral. There are difficulties
with valuing, limited marketability and legal impediments of liquidating
collateral. Legal procedures are usually prolonged, expensive and may include
extra costs for court and sales fees. For these reasons, it is believed that
collateral values should not be deducted from impaired loan values. The case
for the inclusion of collateral in the calculation is simpler. If the
collateral is liquid, of high quality or is a marketed government issued
security then it can be easily recovered and appraised. Consequently, only
collateral that can be reliably measured are likely to be incorporated in the
loan provisioning process.
The monitoring of the value of collateral is another issue
that banks are expected to grapple with. Banks may establish a mechanism
periodically to have collateral appraised. Assets that cannot be seized,
possessed or foreclosed should not be considered capital. Continuously
monitoring capital could steer a bank clear of tremendous losses in the loan
portfolio. As witnessed in the early 1990s, neglecting to monitor collateral
values may be perilous to financial institutions (Song, 2002).
3.
Methodology
3.1
Questionnaires and Interviews
One of the primary interests of this paper is to ascertain
the loan loss provisioning practices of commercial banks in Barbados. The key
participants interviewed are therefore the employees of the banks who are
directly involved with or managed the processes of loan loss provisioning.
Regulatory staff and relationship officers from the Central Bank of Barbados
who deal specifically with the commercial banks were also interviewed to gain
their perspectives of the processes.
The questionnaire- available on request- was designed to
acquire information from all six commercial banks in Barbados during the interview.
It is structured similar to the literature review and consists of twenty
questions within the following sections: the standards that are followed, loan
loss provisioning practices including the treatment of collateral and how the
banks‟ management feel about their levels of provisions and reserves.
3.2
Econometric Models and Data
Econometric Models
Finding out how the commercial banks provide for loan
losses is important but this does not give sufficient information to describe
the major determinants of loan loss provisioning in Barbados, which is required
because provisions and reserves lessen the impact of non-performing loans that
are sometimes major leading indicators of banking crises. As a result, this
paper applied the panel dynamic ordinary least squares (PDOLS) methodology
proposed by Kao and Chiang (2000) and adopted by Mark and Sul (2003) to estimate regression models aimed at
assessing the factors that influence loan loss provisioning in Barbados in the
short and long run. PDOLS is based on
the single equation DOLS procedure pioneered by Saikkonen (1991) and
generalized by Stock and Watson (1993) and has the following similar features:
it allows for the direct estimation of a mixture of I(I) and I(0) variables,
performs well in small samples and avoids the issue of endogeneity. Further, statistical inference on the
parameters of the co-integrating vector is facilitated by the fact that the
t-statistics of the estimated coefficients have an asymptotic normal
distribution, even with endogenous regressors (Stock and Watson, 1993).
The application of PDOLS involves three steps. First, panel unit root tests are employed to
check the stochastic nature of the variables. The procedures used are due to
Levin, Lin and Chu (LLC) (2002), and Breitung (2002) [which have a common unit
root process as their null hypothesis], Im, Pesaran and Shin (IPS) (2002), the
Augmented Dickey Fuller - Fisher Chisquare (ADF) [whose null assumes individual
unit root processes] and the Hadri z-statistic which has a null hypothesis of
no unit root.
If the variables are non-stationary, the next step is to
test for co-integration. The methods
employed are the residual based panel and group statistics tests of Kao
(1999). The more popular Pedroni (1999)
co-integration tests were attempted but the data set lacked enough observations
for estimation. Finally the short run and long run factors influencing loan
loss provisioning is determined.
The following long run model is
estimated:
There X is a vector of all explanatory variables, X1
is a subset of I(1) variables of X, is a vector
of long –run coefficients and e is a well behaved error term. The leads and
lags of the first differenced I (1) regressors are included to deal with the problems
of endogeneity and autocorrelation. In
this paper, the models begins with five leads and lags and as is customary the general to specific
methodology is executed in order to obtain a parsimonious representation of the
regression equations. Therefore, only statistically significant variables are
retained in the model.
In addition, a general Error Correcting Model was done so
that the short run coefficients could be obtained. This is given as
Changes in the loan loss provisions on total loans ratio
are a function of its past, lagged first difference non-stationary variables (X1),
lagged stationary variables (Z) and the lagged error correcting term. The short run effects are captured by 1, 2and 3 while the rate at
which loan loss provisions readjusts to steady state, after disequilibrium have
occurred, is given by.
The independent variables used in this study to explain the
natural logarithm of loan loss provisions (Lllp) combines variables from Khemraj
and Pasha (2009) and Anglomkliew et al (2009) and can be broadly categorized
into two types: macroeconomic and bank specific. The macroeconomic variables include: the
natural logarithm of real gross domestic product (lGdplev) which is expected to have a negative
association with Lllp. It is
anticipated that as the economy expands, loan payment defaults are less likely
to occur because of increased income, thus, a reduced provisioning level based
on a person‟s ability to service their debt. lPrices represents the
logarithm of the Retail Price Index and a positive relationship is expected
between this variable and Lllp since
inflation makes it more expensive to service debts, which may cause the
provision for defaults to rise. The
association between the logarithm of
the real effective exchange rate (Reer) and Lllp is likely to be positive following the argument that for
countries with fixed exchange rates and major export industries, it is
anticipated that if the currency appreciates those local exports would become
relatively more expensive when compared to foreign products. This implies that
since a substantial number of loans go into exporting enterprises that an
expected appreciation of this variable should have a positive impact on Lllp.
With respect to the bank specific variables lEarn represents the logarithm of
profits before tax and provisions divided by total assets for the individual
bank. An ambiguous sign between Lllp
and earnings is anticipated. If banks
are using provisions to smooth earnings a positive relationship is likely. A
direct link between the logarithm of total loans divided by total assets of the
banks (lLoAs) and Lllp is expected. Total loans to total
assets describe the amount of credit risk. Therefore, if a bank has a high loan
to asset ratio it may provide more provisions for the credit risk it is
undertaking. lSiz is a proxy for the
size of the bank and represents the logarithm of the bank‟s assets divided by
the total assets of all the banks. The a priori sign for this variable is
ambiguous. It is believed that the sign will be negative if banks are better
able to manage credit risk efficiently during the credit rationing stage
because they have better screening processes - due to their size. A positive
association is likely between the logarithm of the total loans for the banks (ltloans)
and Lllp. Rapid credit growth implies higher defaults and by extension, the
larger the loan portfolio the greater the risk of default. During the
interviews, commercial bankers said that they use historical default
information when making decisions on current loan loss provisions. For this
reason, a positive relationship is expected between the logarithms of
non-performing loans to total loans ratio (lNpl1).
Higher interest rates imply that borrowers should find it harder to service
their debt, thus, leading to more defaults and higher provisions. In essence, a
direct link between (Rir) and Lllp is anticipated.
Data Issues
The data utilised in this paper cover the quarterly period
from 1996 to 2009 and were obtained from the Central Bank of Barbados. The dependent variable is loan loss
provisions on total loans ratio (Lllp)
and the independent variables are as follows: real gross domestic product (lGdplev),
the Retail Price Index (lPrices), the real effective exchange
rate (Reer), profits before tax and
provisions divided by total assets (lEarn),
total loans as a ratio of total assets (lLoAs),
the bank‟s assets divided by the total assets (lSiz), total loans (ltloans), nonperforming loans divided by total
loans (lNpl1) and the real interest rate (Rir)
given by the total weighted average loan rate minus inflation. The
descriptive statistics of the series are shown in Table 1.
Over the period of interest there was one amalgamation and
a number of acquisitions. The amalgamation occurred in 2002Q4. There was
further restructuring of this new venture in 2006Q1 when one bank increased its
share capital by buying a part of the other bank‟s share. As a result of this
amalgamation and the fact that one bank now prevails the data prior to 2002Q4
was summed as if this bank existed as it appears today, over the entire study
period. Although this summation does not truly represent the structure of the
banking sector from 1996 to 2002Q3 it was necessary to maintain a consistent
dataset.
Table 1: Descriptive Statistics
|
|
Lllp
|
lEarn
|
lGdplev
|
lLoAs
|
lPrices
|
lSiz
|
lTloan
|
Reer
|
Rir
|
LNpl1
|
|
Mean
|
0.485
|
-4.880
|
5.526
|
3.894
|
4.732
|
1.330
|
12.926
|
96.561
|
10.945
|
1.542
|
|
Median
|
0.648
|
-4.824
|
5.516
|
3.993
|
4.729
|
2.240
|
12.966
|
96.371
|
10.714
|
1.511
|
|
Maximum
|
3.065
|
-3.239
|
5.694
|
4.323
|
4.932
|
3.682
|
14.299
|
104.540
|
12.192
|
3.830
|
|
Minimum
|
-1.664
|
-7.346
|
5.362
|
3.112
|
4.589
|
-6.397
|
11.421
|
88.974
|
9.642
|
-16.118
|
|
Std. Dev.
|
1.047
|
0.824
|
0.085
|
0.293
|
0.092
|
2.756
|
0.798
|
4.078
|
0.783
|
1.295
|
|
Skewness
|
-0.070
|
-0.354
|
0.130
|
-0.726
|
0.318
|
-1.574
|
-0.204
|
0.123
|
0.040
|
-7.480
|
|
Kurtosis
|
2.253
|
2.815
|
2.055
|
2.321
|
2.483
|
3.924
|
1.867
|
2.035
|
1.612
|
103.932
|
|
Jarque-Bera
|
8.077
|
7.491
|
13.449
|
35.942
|
9.409
|
150.638
|
20.290
|
13.886
|
27.063
|
145753
|
|
Probability
|
0.018
|
0.024
|
0.001
|
0.000
|
0.009
|
0.000
|
0.000
|
0.001
|
0.000
|
0.000
|
|
Sum
|
163.106
|
-
1639.680
|
1856.654
|
1308.216
|
1589.946
|
446.910
|
4343.155
|
32444.64
|
3677.594
|
518.179
|
|
Sum Sq. Dev.
|
367.479
|
227.468
|
2.425
|
28.689
|
2.807
|
2544.865
|
213.372
|
5570.053
|
205.357
|
562.087
|
|
Observations
|
336
|
336
|
336
|
336
|
336
|
336
|
336
|
336
|
336
|
336
|
The first acquisition during the period occurred in 2003Q4
and the second, in 2004. The last one came in 2008Q2 but this did not disrupt
the data series as separate reporting continued up to the end of the review
period. All in all this presents a full data set of six banks spanning the
entire period for the banking system that prevails today.
In Barbados, the behaviour of loan loss provisions to total
loans ratio is different for each bank (Figure 1). For instance, one bank
showed a steady decline in this ratio from 1.62% in 1996Q2 to its lowest level
(0.21%) in 2006Q1 followed by a steady increase to 0.49% at the end of 2009. On
the other hand, the provisions ratio for another bank was 0.19% at the
beginning of 1996 and expanded over the next seven years to it maximum of 2.03%
in 2003Q2. Subsequent to this, the ratio recorded a steady decline during the
following four years before it began to rise with the onset of the financial
crisis in 2007 when it reached 0.70% in 2009Q2.
The main reason for this discrepancy in the behaviour of banks is the
difference in the credit levels and types of credit issued by each bank.
Certain loans within particular industries and sectors have been historically
proven to be unlikely to default and would carry a lower actual provision. For
example, a bank with mainly mortgages in its portfolio would have a lower actual
provision because people seldom default on mortgage payments since they place a
high priority on keeping their
properties.
4.
Results
4.1
Findings of the Questionnaire and Interviews
The structure of the loan provisioning process in Barbados
is such that the Central Bank regulates loan loss provisions by setting minimum
levels on a categorical system. The Bank provides the commercial banks with the
criteria for loan classification and minimal provisioning levels. It then
conducts credit reviews of the individual bank‟s loan portfolios to ensure
compliance with the criteria and categories of the Financial Institutions Act
(1998). The purpose of this Act is to level the playing field across banks and
other financial institutions, making regulations more enforceable.
An issue that is highlighted by regulators is the
discrepancy between required and actual provisions. Required provisions are the
provisions calculated by the Central Bank of Barbados based on their
classifications as determined by The Financial Institutions Act 1998 while
actual provisions reflect the amount that banks set aside to cover losses
within their loan portfolios. The study asked
commercial bankers whether or not they believed that their provision and
reserve levels were adequate and all responded in the affirmative.
From Figure 2 it is observed that there are many small
fluctuations in the data as commercial banks are constantly adjusting their
provisions based on the economic climate and previous loan default behaviours.
It can also be seen that regulatory provision and actual provision levels move
very closely together but commercial banks provided more than required by their
regulator, the Central Bank of Barbados. As of the 2006Q2 there was an upward
trend in actual provisions with an opposite movement in the level of regulatory
provisions for the subsequent four periods. With the onset of the financial
crisis of 2007, one can recognise that the actual and regulatory provisions
begin to go in the same direction. What is of greater significance is that
these two trends vary significantly in the recent period between 2007Q2 and
2009Q4. The regulatory provisions are
noticeably lower than the actual provisions because the former are based solely
on a set percentage of the loans within the different non-performing categories
and allows for no discretionary influences.
Also, the financial expectations in Barbados were dismal thus, banks
would have been trying to raise their reserves as the economic climate was and
continues to be very unstable.
The underlying cause of the discrepancy between regulatory and actual provisions is because commercial banks have different internal rating systems for classifying their loans. The interviews highlight that commercial banks use the same primary indicator as the Central Bank the time past the due date - and incorporate other indicators such as internal credit scores, the borrower‟s significant assets, the value and type of collateral as well as other borrower specific characteristics. The reason for the difference in classification is that the commercial banks utilize The International Financial Reporting Standards (IFRS) or the Generally Accepted Accounting
Principles (GAAP) which do not align perfectly with the
Central Bank‟s Regulations on provisioning. However, the hierarchy is such that
no matter what standards the commercial banks decide are best for their
personal use, their provisions and reserves must comply with the
Central Bank‟s regulations. Commercial banks usually employ
the IFRS because they are more flexible. This flexibility is desired because it
allows the banks to take the current valuation of assets and collateral into
consideration. On the other hand, under the Central Bank‟s regulation, the
historical value of the loan is used to estimate the provision.
Due to the differences in the classification ratings for
loans, regulators usually face the problem of synchronization. They must seek
to decide, based on the criteria of the commercial banks‟ loan categories,
which one(s) match the criteria set out in the regulations. This process is
necessary in determining which provisioning level is being employed in the
calculations and is usually well documented by the banks.
In was revealed during the interviews that the provisioning
levels in some commercial banks are computed using hindsight while in the
others only current information is taken into consideration. Also, some local
banks have no authority in determining their own provision levels and take the
level passed on by their international head offices. Sometimes, the
provisioning levels decided by the commercial banks are less than those of the
Central Bank and have to be corrected by the creation of an additional
account.
The primary trigger that commercial banks in Barbados use
to determine if a loan is to be classified as non-performing is the „time past
the due date‟. In addition to timing, there are other criteria that can be
employed such as the borrower‟s cash flow, the type of collateral and the
probability of collection of the debt. The amount and purpose of the loan do
not impact on the level of provision applied to the loan but may become
relevant when the decision to issue credit is being made.
Other important information derived from the interviews are
(1) commercial banks in Barbados partition their loan loss provisioning into a
general provision and a loan specific provision related to the individual loan
amounts; (2) Central Bank‟s regulations do not consider collateral while some
banks do; (3) the most common type of securities in Barbados is property - land
and houses – which are usually valued by third party appraisals but this value
is sometimes adjusted by the bank‟s management to reflect other characteristics
about these securities; and (4) regulatory monitoring of commercial banks is of
a continuous nature, with reporting to the Central Bank of Barbados done on a
weekly, monthly, quarterly and annual basis. To supplement these reports,
officials also conduct on-site inspections to test different key areas of
concern which may range from grievances with a particular process to the
traditional high-risk areas such as capital management and credit risks.
4.2
Econometric Results
The results of the panel unit root tests mentioned above
revealed that all the variables, except the logarithm of loans to assets ratio
(lLoAs), logarithm of operating
profit on total assets ratio (lEarn)
and the logarithm of non-performing loans divided by total loans ratio (lNpl1), are integrated of order 1 [I(1)], that is, they need to be
differenced once to become stationary (see Table 2). lLoAs,
lEarn and lNpl1 are all
stationary in levels I(0).
Table 2: Panel Stationary Tests
|
Variable
|
|
Levin, Lin
|
Breitung
|
Im, Pesaran and
|
ADF - Fisher
|
PP - Fisher Chi-
|
|
Lllp
|
Level
|
-1.78997**
|
2.25073
|
-1.38674*
|
16.4544
|
15.5911
|
|
1st Difference
|
-17.8331***
|
-6.04304***
|
-17.4028***
|
213.590***
|
215.380***
|
|
|
lEarn
|
Level
|
-3.59849***
|
-2.43416***
|
-4.37206***
|
51.3037***
|
103.556***
|
|
1st Difference
|
1.00358
|
-3.14124***
|
-10.4063***
|
121.558***
|
135.691***
|
|
|
lprices
|
Level
|
2.84057
|
0.40571
|
2.86434
|
1.51318
|
2.07558
|
|
1st Difference
|
-15.0916***
|
-13.8678***
|
-13.5521***
|
162.483***
|
161.502***
|
|
|
lLoAs
|
Level
|
-.91547***
|
0.87629
|
-3.02976***
|
31.4518***
|
30.4265***
|
|
1st Difference
|
-17.4912***
|
-11.4741***
|
-16.5494***
|
199.925***
|
204.049***
|
|
|
lTloan
|
Level
|
-0.06293
|
0.52110
|
3.04777
|
4.68405
|
6.32559
|
|
1st Difference
|
-15.5133***
|
-10.6867***
|
-13.5692***
|
163.121***
|
168.525***
|
|
|
lgdplev
|
Level
|
-2.38532***
|
-0.44658
|
-0.72176
|
10.6070
|
16.9463
|
|
1st Difference
|
15.3195
|
-2.69677**
|
-3.25119***
|
29.8734***
|
304.050***
|
|
|
RIR
|
Level
|
1.43881
|
-1.49582*
|
2.34140
|
2.09488
|
2.38555
|
|
1st Difference
|
-22.1595***
|
-17.2348***
|
-18.3591***
|
228.123***
|
227.148***
|
|
|
lSiz
|
Level
|
-1.78875**
|
-3.34322***
|
-1.73913**
|
21.5210**
|
49.6274***
|
|
1st Difference
|
-14.9972***
|
-13.1326***
|
-16.0338***
|
195.809***
|
247.959***
|
|
|
REER
|
Level
|
-0.73426
|
-2.85402***
|
-2.34624***
|
22.2960**
|
12.0016
|
|
1st Difference
|
-15.4053***
|
-12.0439***
|
-12.3879***
|
145.579***
|
141.504***
|
|
|
lNpl1
|
Level
|
-3.27370***
|
-0.08771
|
-4.70317***
|
57.9441***
|
58.1823***
|
|
1st Difference
|
-20.8006***
|
-8.39757***
|
-20.9672***
|
245.288***
|
223.927***
|
Note: *** **
and * are the critical values for the rejection of the null hypothesis of a
(common or individual) unit root at the 1%, 5% and 10% level respectively. All
tests were done using an intercept with an automatic selection of lags except
for the Breitung Common Root Test which utilises an intercept with a trend.
Lags were selected automatically using Schwarz information criterion.
The long run determinants of the PDOLS model are presented
in Equation (3). It is derived from a general model and reduced to a
parsimonious form using the general to specific methodology a la Hendry. The
leads and lags are omitted to save space.
Lllp=12.88967*** -
0.10641 LEarn *** +
0.10574*LNpl1- 0.71419***LTloan- 0.46973***LSiz
(9.8285) (-2.8962) (1.7218) (-7.3808) (-5.2328) (3)
+ 0.023564*** Reer-
0.2151***Rir- 0.65675***LPrices
(4.6764) (-7.3859) (-2.2739) R2 0.918408
Note: t- statistics of regressors are shown in
parentheses. ***, ** and * indicates significance at the 1, 5 and 10% level of
testing, respectively. R 2 is the coefficient of determination. The
standard errors in brackets are heteroskedasticity consistent a la White
(1980).
Seven significant variables can be found in this long run
model and they explained 91.8% of the variation in loan loss provisions (Lllp). The operating profit on total
loans ratio (lEarn) has an adverse
relationship with provisions. This result is consistent with Angklomkliew et al
(2009) and it suggests that the commercial banks in Barbados are not using Lllp to smooth income in the long run.
The real effective exchange rate (Reer)
has a weak positive significant relationship with the Lllp, implying that the impact that the real economy has on
provision levels is minuscule. This was also found to be the case in Pasha and
Khemraj (2009). The level of classified
debt on total loans (LNpl1) is only significant at the 10% level and has the a priori sign,
indicating that as the level of classified debt or non-performing loans
increases Lllp also advance.
The relative market share (lSiz) of commercial banks has an adverse sign with Lllp. This suggests that, unlike Guyana (Pasha and Khemraj, 2009),
the larger banks in Barbados are more able to screen loans through better risk
management strategies thus providing a stronger loan portfolio. Contrary to the
results seen in Jimenez and Saurina (2005) a negative link between provisions
and the real interest rate (Rir)
is observed. The retail price index (LPrices)
gives evidence that the price level affects Lllp
in the long run. However, it should be noted that the a priori sign was
positive. The parameter on the total number of loans (LTloan) has an adverse sign. This indicates that as total loans
increases Lllp will decrease. This
anomaly between a priori and actual signs can be explained if one considers
that bigger banks in Barbados are the ones with the largest loan portfolios and
are better able to screen possible defaults. Hence it is expected that as total
loans expand then Lllp should
contract.
The results of the PDOLS error correcting
model (ECM) are reported in Equation (4).
Lllp=1.683799*** +
0.101463*** d LDep( 1 )+ 0.074894 (( 3) d LDep1 )-( 4)
1.20615*** d(LTloan)
(4.1007) (2.3817) (1.5618) (-4.8949)
+ 0.660077*** d L( Prices( 3) ) -
0.445547 d(L** Gdplev)
+ 0.026876** d LEarn( ( 5)
) (4)
(5.6323) (-2.4593) (1.6451)
- 0.065133 d(Rir ) + 0.013059** (-3) *** d(Reer( 4) )- 0.03217 d(lSiz** )- 0.129188*** ECM(-1)
(-1.9606) (2.9136) (-1.7649) (-4.0764)
R2 0.247186
Note: t- statistics
of regressors are shown in parentheses.. ***, ** and * indicates significance
at the 1, 5 and 10% level of testing, respectively. R 2 is the
coefficient of determination. The standard errors in brackets are heteroskedasticity
consistent a la White (1980). „d‟ represents the first difference operator.
The error correcting term in this model is negatively
signed, has the a priori magnitude (less than one) and is very significant; its
coefficient suggests that given an exogenous shock to Lllp it returns to equilibrium with 12.9% of the initial shock
dissipating in the ensuing quarter. This finding for the error correcting term
supports the earlier result that the variables cointegrate.
Except for LGdplev
and LNpl1 all the other variables are
significant in both the long run and ECM formulations. In addition, the signs
of the parameters in both specifications are similar excluding LPrices and LEarn. With respect to the ECM all the variables have the a priori
sign except for the real interest rate (Rir(-3))
and total loans (LTloan). The
variable for total loans has a very significant negative relationship with Lllp and suggests that as the loan
portfolio grows Lllp should decrease.
In support of this is the relative market share variable (LSiz) which is significant at the 10% level and carries a negative
sign, implying that larger banks are better at risk management. The gross
domestic product (LGdplev) has the
expected relationship with Lllp,
indicating that a one unit rise in LGdplev
causes a 44.6% reduction in Lllp.
This is consistent with the results obtained in the interviews where bankers
state that as incomes increase creditors appear better able to service their
debt and avoid default.
To shore up the effect of the real economy on loan loss
provisions the real effective exchange rate from four periods prior (Reer(-4)) is significant and has the
expected sign; if it advances by one unit a 1.3% change in Lllp in the same direction results. The retail price index from
three periods before (Lprices(-3))
also proved to have a influential effect on Lllp.
As prices expand by one unit Lllp are
expected to rise by 66% as it would now be harder to service debt. Finally, the
model suggests that the short run expectation is quite small with a value of
17.36%. The signs obtained on these lagged parameters (Lllpt-3 and Lllpt-4) are positive and imply that the higher the
value of these parameters, the more banks must provide for current Lllp. This persistence in Lllp implies that a period of hardship
is most likely to be followed by another one until the cycle breaks.
Information gathered from the interviews found a similar result, namely, that
commercial banks usually use historical default data along with current and
previous business cycle information when determining their current levels of
provisioning.
5.
Conclusion
This paper examined the components of the loan loss
provisions within the banking system of Barbados. It combines information
acquired through the use of interviews along with data from the regulator- The
Central Bank of Barbados. The study also used an econometric model to explore
the main factors which drive loan loss provisions. The results revealed, for
the first time in Barbados, and possibly regionally, a greater appreciation for
the individual procedures of the commercial banks as it pertains to loan loss
provisioning. Additionally, the paper discovered that all of the factors
affecting loan loss provisions were significant in both the long run and short
run; non-performing loans ratio and real gross domestic product were only
significant in the long run and the short run, respectively.
This article showcased the persistence of loan loss
provisions and revealed how current and forecasted economic trends could
influence the levels of reserves that provide safety for depositors. A heavy
dependence of loan loss provisions on the real economy emerged and is evidenced
in the significant relationships between loan loss provisions and the real effective
exchange rate, the retail price index and real gross domestic product. This
implies that regulators and policy makers‟ need to keep abreast of leading
economic indicators both locally and internationally as competition could
influence the level of defaults and, by extension, loan loss provisions.
In particular, this research proposes that commercial banks
are using loan loss provisions to smooth income in the short run. This means
that as income falls, profits will decline, and tax revenue, which is
calculated based on profits, will follow suit. Consequently, it is very
important for policy makers in Barbados to take the relationship of loan loss
provisions and tax revenue into consideration.
In addition, evidence from the model suggested that the
larger commercial banks were better able to screen borrowers to avoid future
defaults in both the long and short run, probably, through the use of more
efficient risk management strategies.
One implication of the study is that banks must hold enough
reserves to lessen the impact of adverse shocks to the economy. It would appear
that the shortcomings of the current regulatory framework could curtail this
development, therefore, considerable groundwork should continue to be done to
propel the Barbadian Banking System to a better regulatory framework: Basel
Accord 2.
In Barbados, the current regulatory methods are proving to
be successful possibly because there are few commercial banks. Their adherence
to the reporting regulations and their willingness to incorporate the
suggestions made by the regulator improves the long standing relationships and
communication channels between banks and the Central Bank. Therefore, there is
normally no need to apply any sanctions or penalties. However, there is not sufficient
flexibility of regulative authority to prevent the major disparity between
regulatory and actual provisions during undesirable economic times, as
evidenced in the recent financial crisis.
In light of this belief, this paper argues that an
additional x-factor that could be controlled by the Central Bank be included.
This would give regulatory staff the much needed flexibility to influence the
regulated provision and to compensate for the deficiencies in the current
historical category value method until a better alternative can be fully
adapted.
There are also limitations of this study that provide scope
for future research. Variables such as the loss given default and the
probability of default were contemplated but were not included because of the
large amount of information required to construct appropriate proxies. Also,
given the importance of the offshore banking sector to Barbados, future work
could be directed towards these institutions. Such a study could be aimed at
making a comparison of their loan loss provisioning practices with the local commercial banks‟.
References
Angklomkliew, S.,
George, J., and Packer, F. (Dec 2009) “Issues and Developments in Loan Loss
Provisioning: The Case of Asia.” BIS Quarterly Review.
Balla, E. and
McKenna, A., (2009). “Dynamic provisioning: a countercyclical tool for loan
loss reserves.” Economic Quarterly, Federal Reserve Bank of Richmond, issue
fall, pages 383-418.
Basel Committee on
Banking Supervision (1999) “A New Capital Adequacy Framework.” Bank for
International Settlements
Bikker, J.A. and
Metzemakers, P.A.J., (2002). “Bank provisioning behaviour and procyclicality.”
Research Series Supervision 50, Netherlands Central Bank, Directorate
Supervision.
Bouvatier, V. and
Lepetit, L., (2008). “Banks' procyclical behavior: Does provisioning matter?”
Journal of International Financial Markets, Institutions and Money, Elsevier,
vol. 18(5), pages 513-526, December.
Breitung, J. (2002),
“Nonparametric Tests for Unit Roots and Cointegration.” Journal of Econometrics
108, 343-364.
Cortavarria, L.
Kanaya, A., Song, I. and Dziobek, C.H, (2000) “Loan Review, Provisioning, and
Macroeconomic Linkages.” IMF Working Papers 00/195, International Monetary
Fund.
Floro, D, (2010)
“Loan Loss Provisioning and the Business Cycle: Does Capital Matter? Evidence
from Philippine Banks.” Bank for International Settlements.
Greenidge, K. and
Grosvenor, T.,(2010), “Forecasting Non-Performing Loans in Barbados.” Business.
Finance and Economics in Emerging Economies Vol 5 No.1 2010.
Im, K.S, Pesaran, M.H.
and Shin, Y. (2003) “Testing for Unit Roots in Heterogeneous Panels.” Journal
of Econometrics 115(1): 53-74.
Jiménez, G. and Saurina, J., (2006), “Credit cycles, credit
risk, and prudential regulation” International Journal of Central Banking (2),
65-98.
Jokivuolle, E., and
Peura, S. (2003). “Incorporating Collateral Value Uncertainty in Loss Given
Default Estimates and Loan-to-value Ratios. European Financial Management.”
9(3), 299-314. Retrieved from E-Journals database.
Khemraj, T., and
Pasha S. (2009) “The determinants of non-performing loans: an econometric case
study of Guyana.” Presented at the Caribbean Centre for Banking and Finance
Biannual Conference on Banking and Finance, St. Augustine, Trinidad.
Kao, C., (1999). “Spurious Regression and
Residual-Based Tests for Cointegration in Panel
Data.” Journal of Econometrics, 90, 1-44.
Kao, C., and Chiang,
M. (2000) “On the estimation and inference of a cointegrated regression in
panel data.” Advances in Econometrics 15, 179-222.
Laeven, L. and
Majnoni, G., (2003). “Loan loss provisioning and economic slowdowns: too much,
too late?” Journal of Financial Intermediation, Elsevier, vol. 12(2), April.
Laurin, A. and
Majnoni, G. (2003) “Bank loan
classification and provisioning practices in selected developed and emerging
countries.” World Bank working
paper series; no. 1, Report Number 26056
Levin, A., Lin, C.
and Chu, S,. (2002) “Unit root tests in panel data:Asymptotic and finitesample
properties.” Journal of Economics 108: 1-24.
Mark, N., and Sul,
D., (2003). “Cointegration Vector Estimation by Panel DOLS and Long-run money
Demand.” Oxford Bulletin of Economics and Statistics 65(5), 655-80.
Pedroni, P., (1999).
“Critical Values for Cointegration Tests in Heterogeneous Panels with Multiple
Regressors.” Oxford Bulletin of Economics and Statistics 61, 653-678.
Saikkonen, P.
(1991), “Asymptotically Efficient Estimation of Cointegration Regressions.”
Econometric Theory, 7, 1-21.
Song, I. (2002). “Collateral in Loan
Classification and provisioning.” IMF- Monetary and Exchange Affairs
Department. EconLit with Full Text, EBSCOhost(accessed June 10,2010).
Stock, J., and
Watson, M., (1993). “A Simple Estimator of Cointegrating Vectors in High Order
Integrated Systems.” Econometrica,61, 783-820.
Wezel, T., (2010) “Dynamic Loan Loss
Provisions in Uraguay: properties, Shock Absorption
Capacity and Simulations Using
Alternative Formulas.” IMF Working Papers, Vol., pp. 1-22, 2010.
White, H. (1980). “A
Heteroskedasticity- Consistent Covariance Matrix Estimator and a Direct Test
for Heteroskedasticity.” Econometrica, vol. 48, issue 4, 817-838.
No comments