Вы находитесь на странице: 1из 13

Measuring the Sectoral Distribution of Lending to Irish

Non-Financial Corporates
by Rory McElligott and Rebecca Stuart1

ABSTRACT
There has been a rapid increase in lending to the Irish Non-Financial Corporate (NFC) sector in recent times accompanied by a
significant shift in the sectoral distribution of lending. At first glance, the effect of this appears to have been to increase the
concentration of the NFC loan portfolio. However, there are a number of measures of concentration and it is possible that different
measures would yield different results. In the first part of this paper we use a number of measures of concentration to determine
whether Irish banks’ loan portfolios have become more or less concentrated in recent times. Two distinct groups of concentration
measures are used: traditional measures, such as the Hirschman-Herfindahl Index and the Gini coefficient, and distance measures,
which benchmark individual banks’ loan portfolios against proxies of the structure of the economy. Our findings reveal that, for
the most part, both the traditional and distance measures give the same results in the Irish context. Most banks in our sample
showed an increase in concentration in the NFC loan portfolio.

In the second part of this paper, we take a closer look at sectoral concentration. First, we review the literature in this area. We
find that the literature is divided on whether concentration or diversification is the best strategy for banks. Historical experience
shows that concentration of credit risk in asset portfolios has been one of the major causes of bank distress. However, the
monitoring costs involved in operating in distinct areas and the sectoral expertise gained by concentrating may make it preferable
for a bank to concentrate their lending to a small number of sectors. Secondly, we present a European comparison of the sectoral
distribution of the NFC loan book which shows that the distribution of the Irish NFC loan book is not dissimilar to the European
position. Finally, we list some possible mitigating factors that may apply specifically to the Irish NFC loan book.

1. Introduction of measures to determine whether Irish banks’ loan


portfolios have become more or less concentrated in
There has been a rapid increase in lending to the Irish
recent times. We use traditional measures of
Non-Financial Corporate (NFC) sector in recent times.
concentration, such as the Hirschman-Herfindahl Index
Lending to NFCs as a proportion of total private-sector
(HHI) and the Gini coefficient, and distance measures,
credit increased to almost 40 per cent in June 2006
which benchmark individual banks’ loan portfolios
compared with approximately 30 per cent in 2000.
against proxies for the structure of the economy in
Indeed, in 2007Q2 the annual growth rate in lending to
Ireland. These two sets of measures have different
the NFC sector was in excess of 30 per cent. Alongside
underlying assumptions about what a diversified loan
this rapid growth, there have been significant shifts in the
portfolio looks like. The first group, traditional measures,
sectoral distribution of lending. The proportion of
assume that perfect diversification2 should mean equal
lending accounted for by certain sectors has changed
exposure by the lender to every sector, whereas the
substantially over this time period. In particular, the second group, distance measures, allow one to use a
property-related sector (consisting of the Construction predefined benchmark — such as the economic structure
and Real Estate sectors) accounted for almost 67 per — as an indicator of diversification.
cent of NFC lending by June 2007, a marked increase
from 2000 when it accounted for less than 35 per cent. Our findings from this part of the analysis reveal that, for
the most part, both the traditional and distance measures
Given this recent increase in the level of NFC lending, it give the same results in the Irish context. Most banks in
is an opportune time to examine the structure of the loan our sample showed an increase in concentration in the
book and, in particular, changes in the level of NFC loan portfolio.
concentration. However, different measures of
concentration may give different answers to this Having examined the overall trend in concentration over
question. In the first part of this paper, we use a number recent years, in the second part of this paper we examine
1
The authors are, respectively, economists in the Statistics Department and the Monetary Policy & Financial Stability Department. The views expressed
in this paper are the personal responsibility of the authors and are not necessarily those held by the Central Bank & Financial Services Authority of
Ireland or by the ESCB. The authors would like to thank their colleagues within the CBFSAI for their assistance in completing this paper.
2
Throughout the literature in this area ‘diversification’ is used simply as the opposite of ‘concentrated’, and does not carry the connotation, as it does
in finance literature, of reducing the risk of a portfolio by taking into account the degree to which the return on assets in a portfolio can be expected
to move together. We follow this example, using ‘diversification’ merely in the sense of being ‘unconcentrated’.

Financial Stability Report 2007 115


sectoral trends more closely. Firstly, the economic personal sector, NFCs and other financial intermediaries.
literature is divided on which of two competing Our study is confined to the non-financial corporate
hypotheses of concentration or diversification is the best sector. The eleven NFC sub-sectors are listed in Table 1
strategy for banks. Historical experience shows that of the Annex. It is on the concentration of banks’
concentration of credit risk in asset portfolios has been lending to these eleven NFC sub-sectors that this paper
one of the major causes of bank distress. This is true for focuses.
individual institutions as well as for banking systems as a
whole. In a sector-concentrated loan portfolio, the bank Although our dataset covers all resident banks in Ireland,
is lending to a portfolio of corporations that operate in we restrict our analysis to banks whose business is
the same sector and thus are likely to be facing similar primarily orientated towards the Irish economy. The
operating conditions. The implication for banks is that inclusion of banks whose business is internationally
the performance (or non-performance) of the loans in its focused would bias the results, as their domestic loan
portfolio will tend to be highly correlated. In the face of book may not be a true representation of their
negative economic conditions in a specific sector, the diversification policy. This potential bias would be
bank would face declining asset quality in a large share particularly strong in Ireland given the importance of the
of its loan book. However, there is a second hypothesis international financial services sector. We also adjust the
in the literature which suggests an alternative view on panel for mergers and splits in the banking population.
the link between concentration and financial stability. For example, if one bank splits into two or more banks
The monitoring costs involved in operating in distinct during the study period, we combine the new and old
areas may make it preferable for a bank to concentrate banks and treat them as though they were a single bank
its lending on a small number of sectors. This throughout the study period. A list of the thirteen banks
concentration of sectoral expertise may improve the covered in the study is included in Table 2 of the Annex.
bank’s credit risk decisions. The dataset covers the period from February 1999 to
June 2007 and this provides 34 observations across time
Secondly, we present an international comparison of the for each bank. Our sample of banks covers on average
sectoral distribution of the NFC loan book. The data 93 per cent of the total stock of loans to the NFC sector
have not previously been available on a cross-country in each quarter. This coverage has increased over time
basis. The data in this section benchmarks the Irish from 85 per cent in February 1999 to 97 per cent in
sectoral distribution against that of some of our June 2007.
European counterparts and shows that the distribution of
the Irish NFC loan book is not dissimilar to the European Chart 1: Aggregate Concentration Measures
position.
0.7 0.35

Finally, we list some possible mitigating factors that may Gini (LHS) HHI (RHS)
apply specifically to the Irish NFC loan book, including
within-sector dispersion that may reduce the correlation 0.6 0.30

of loan performance; the higher likelihood of loans being


collateralised in the largest sectors; and narrowed scope
of the data in dealing only with the NFC portion of the 0.5 0.25

loan book.

2. Data 0.4 0.20

We employ a bank-level dataset collected by the Central


Bank & Financial Services Authority of Ireland on a
0.3 0.15
quarterly basis. All credit institutions resident in Ireland
are required to report lending to Irish residents
categorised by economic activity of the borrower. The
0.2 0.10
aggregated sectoral data are published in Table C8 of
1999 00 01 02 03 04 05 06 07
the CBFSAI Quarterly Bulletin and a commentary on the Feb June
sectoral developments is published in the quarterly Source: CBFSAI, Authors' calculations
‘‘Sectoral Developments in Private-Sector Credit’’. The
data are separated into fifteen sub-sectors including the

116 Financial Stability Report 2007


3. Lending to NFCs loan book. By June 2007 Manufacturing had declined to
Prior to investigating changes in individual banks’ just 6 per cent of the loan book and Real Estate
sectoral lending trends, it is useful to first look at dominated the NFC loan book of Irish banks with 50
per cent of total loans. Alongside Real Estate, only the
developments in the total banking sector over the
Construction sector increased its share of the loan book
period. The sectoral distribution of loans to Irish NFCs
in the time period. The Real Estate and Construction
has changed considerably over the study period. The
sectors — including the residential real estate sector, the
concentration level as measured by the Hirschman-
commercial real estate sector, property companies
Herfindahl Index (HHI) has increased since February
investing overseas, and infrastructure construction
1999 from a level of 0.134 to 0.301 in June 2007. The
companies — are closely related and now account for a
Gini coefficient also shows a change in the structure of
combined 67 per cent of all loans by Irish banks to Irish
the aggregate loan portfolio towards a more
NFCs (against 24 per cent in February 1999). All other
concentrated portfolio. As shown in Chart 1,
sectors now account for a proportionately lower level of
concentration increased slowly up to mid-2002, after
the banking book than in early-1999 with the main losers
which the pace of increase accelerated.
being Manufacturing, detailed above, Agriculture and
Forestry (13 per cent to 3 per cent) and Wholesale/Retail
The data in Chart 2 facilitate a more detailed analysis of
Trade and Repairs (15 per cent to 8 per cent).
the change in the sectoral distribution of loans to the
NFC sector over the study period. In 1999 there was a Table 1 presents data on the stock of loans and growth
relatively balanced loan book with significant exposure rates for each sector from February 1999 to June 2007.
to a number of sectors including Manufacturing, Real The total stock of loans to Irish NFCs increased from
Estate, Agriculture and Forestry and Wholesale/Retail \22.75 billion in 1999 to \142.98 billion in 2007, an
Trade and Repairs. It is clear from the 2007 data that increase of 528 per cent. The Real Estate sector
banks’ exposure to the Real Estate sector has grown accounted for 57 per cent of the total NFC increase,
hugely in the eight-year period. In February 1999, the growing by just under 1,900 per cent since 1999.
Real Estate sector accounted for 15 per cent of the Manufacturing and Agriculture and Forestry recorded
outstanding stock and this was the second largest sector the slowest rate of increase in the period at just 80 and
after Manufacturing, which comprised 19 per cent of the 63 per cent, respectively.

Chart 2: Sectoral Distribution of Advances, February 1999 & June 2007


February 1999 June 2007
1% 1% 1% 1%
2%
6% 3%
5%
15%
13%
Real Estate 6%

Construction

9% Wholesale/Retail
Trade & Repairs 7%
9% Hotels & Restaurants
50%
Manufacturing

Business
8%
Agriculture
15%
Transport, Storage &
Communications
Electricity, Gas & Water
19% Supply 17%
12%
Other

Source: CBFSAI, Authors' calculations


Note: 'Other' combines Fishing with Mining and Quarrying, which individually represent less than 1 per cent of total advances in both 1999 and
2007.

Financial Stability Report 2007 117


Table 1: Sectoral Breakdown of NFC Loans, 1999 and 2007

Stock (\ million) Change

1999 2007 \ million per cent

Agriculture & Forestry 2,888 4,702 1,814 62.8


Fishing 111 406 295 265.8
Mining & Quarrying 196 447 251 128.1
Manufacturing 4,498 8,095 3,597 80.0
Electricity, Gas & Water Supply 281 1,127 846 301.1
Construction 1,962 24,351 22,389 1,141.1
Wholesale/Retail Trade & Repairs 3,331 12,103 8,772 263.3
Hotels & Restaurants 2,677 10,537 7,860 293.6
Transport, Storage & Communications 1,273 2,939 1,666 130.9
Real Estate 3,594 71,833 68,239 1,898.7
Services 1,946 6,437 4,491 230.8

NFC Total 22,757 142,977 120,220 528.3

4. Concentration Measures is segment i’s share of bank b’s total NFC lending
Having reviewed the data and presented graphical portfolio. Let yit denote the share of loan segment i in the
evidence suggesting an increasing sectoral benchmark loan portfolio NFC loans.
concentration in the aggregate Irish loan book, we now a) The traditional measures of dispersion are the
turn to statistical measures of concentration, and their Hirschman-Herfindahl Index and the Gini coefficient.
evolution over time. In this section we outline a number n

of concentration levels, while in Section 5 we analyse i) Hirschman-Herfindahl Index (HHI) HHI(x) =  (xi)2
i=1
the developments in these measures in recent years. n

 (2j − n − 1)Xi
b,t
ii) Gini coefficient
G=
j=1
4.1 Traditional and Distance Measures ,
2n2 
Concentration measures can be separated into two
where n is the number of values observed, j is the rank
distinct classes: traditional measures and distance
of values in ascending order and  is the average of
measures. The implicit assumption in traditional
the observed euro values of exposure, Xb,i t.
measures is that a perfectly diversified3 loan book should
have equal exposures to all sectors of the economy. A low value in the above measures means a low level of
However, true diversification is unlikely to require an concentration (the individual exposure of the bank to the
equal weighting of all sectors, but a weighting that is different sectors is low), and a high value means a high
similar to the relative size of each economic sector in level of concentration. The HHI values range from a
the overall economy. Distance measures attempt to minimum of n1 to a maximum of 1. The Gini coefficient
overcome the assumption of equal weighting by values range from a minimum of 0 to a maximum of
measuring the distance of a bank loan portfolio from a (n−1)
. The HHI and Gini coefficient are widely used
n
predefined benchmark portfolio that captures the measures of market concentration.
relative size of different sectors. Details of the measures
utilized in our study are presented below. We use two b) The five distance measures are:
traditional measures of concentration and five distance DM1: Maximum Absolute Difference
measures.
D1(x, y) = max x1 − y1
i

For all measures, we represent bank exposures in the


DM2: Normalised Sum of Absolute Differences
following way: let Xb,i t be the euro value of exposure of
Xb,t 1
n

bank b at time t to loan segment i. Then xb,i t = n i b,t D2(x, y) =  xi − yi
j=1 Xj 2 i=1

3
As noted above, we use ‘diversification’ in the same sense as it is used throughout the literature in this area: simply as the opposite of ‘concentrated’.
In this sense, it does not carry the connotation, as it does in finance literature, of reducing the risk of a portfolio by taking into account the degree
to which the return on assets in a portfolio can be expected to move together.

118 Financial Stability Report 2007


DM3: Normalised Sum of Squared Differences equally. However, as argued by Kamp et al, distance
measures, which weight industries by their importance
n
1 in the overall economy, take account of this kind of
D3(x, y) =  xi − yi
2

2 i=1 specialisation, providing a more accurate picture. In the


second study, Pfingsten and Rudolph (2002) used data
DM4: Average Relative Difference
on four categories4 of banks to estimate the distance
between the loan book of each category and the total
1 xi − yi
n

D4(x, y) =  x +y loan book of all four categories over the period 1970 to
n i=1 i i
2001. They suggest that a benefit of distance measures
DM5: Average Squared Relative Difference is that they abstract from any kind of risk or return
specification. They argue that this is beneficial as it is

 
2
x − yi
n
1 unclear which measure of risk best captures loan
D5(x, y) =  i
n i = 1 xi + yi default risk.

All the above distance measures are normalised to the 4.3 Benchmarks used in this Study
interval [0,1] and the higher the level the higher the In our current study we wish to measure the
concentration. The first distance measure D1 simply concentration of bank NFC lending when benchmarked
reports the distance of the sector in the loan portfolio against the relative size of each sector in the economy.
which is the furthest from the benchmark for that sector. We have a number of benchmarks or proxies for each
This measure is therefore insensitive to all but this one sector’s share of the Irish economy available to us.
sector and may not pick up significant changes in
The first benchmark we use is employment by sector.
concentration. D2 is the normalised average of the
These data are available on a quarterly basis in the
absolute difference across all sectors. The result should
Quarterly National Household Survey compiled by the
be read as the proportion of a bank’s loan portfolio that
Central Statistics Office. A breakdown is available by
must be rearranged to achieve the benchmark portfolio
NACE sector and is, therefore, directly comparable to
structure. Measure D3 is similar but assigns a higher
the sectoral classification used in the lending data. Other
weight to sectors with a large deviation from the
advantages of these data are its availability on the same
benchmark. Measures D4 and D5 are based on the
timescale and frequency as the bank-level data. Though
relative difference of a loan portfolio and the benchmark
we believe that employment figures are a good proxy
portfolio. These two measures have the property that the
for economy structure, it is possible that there may be
deviation from the benchmark is seen relative to the size
some sectors that are not labour intensive but provide a
of the sector. This has the effect of giving a lower weight
high value-added component to the economy.
to a 1 per cent divergence in a sector representing 90
per cent of the market than a 1 per cent deviation in a Our second benchmark is each sector’s contribution to
sector representing just a 10 per cent share. It is worth the Gross Value Added (GVA) of the economy. The use
noting a disadvantage of the latter two methods — when of this benchmark along with the employment
a sector is not present in a bank lending portfolio (e.g., benchmark presented above is that it overcomes the
x1 = 0) it will still contribute n1 towards the distance main risk of the employment benchmark of giving too
measure. high a weight to labour-intensive industries. The major
drawback with the GVA benchmark is the annual, rather
4.2 Distance Measures in Other Studies
than quarterly, nature of the observations, and the time
Two previous studies have used distance measures over lag in the release of the data. These data are available
traditional measures of concentration. Kamp, Pfingsten on an annual basis with the latest data relating to 2005.
and Porath (2005), carried out a similar study, using data To address the problem of lack of observations we
on 2,231 individual German banks’ loan books to ‘quarterise’ the data using the Denton estimation
estimate loan portfolio concentration over the period method to produce quarterly sectoral GVA figures.
1993 to 2002. They argue that the benefit of distance
measures over traditional measures, such as the HHI, The third benchmark is each sector’s share of the total
arises from their ability to take account of size difference NFC lending portfolio for all banks in Ireland. This is the
in each sector. For instance, if a bank has a similar most widely used benchmark in previous studies (Kamp
nominal exposure to two sectors — one a relatively big et al (2005) and Pfingsten et al (2002)). Like the data for
sector within the economy, and one a relatively small the employment figures benchmark, these data have the
sector — the bank’s loan book might be said to be very advantage of being on the same timescale and frequency
specialised in the smaller sector. Such specialisation will as the bank-level data. A major drawback with this
not be picked up by a HHI which weights each sector benchmark is that it is endogenous: it is the sum of the
4
The four categories are: commercial banks, banks under public law, cooperative and other banks.

Financial Stability Report 2007 119


individual bank lending portfolios and its movement is, measures, the closer to 1 that they are, the more
therefore, a function of the individual banks’ lending concentrated the loan book. Therefore, any increase in
decisions. We believe that due to the small size of the the traditional measures indicates an increase in
banks in Ireland such a benchmark would be unsuitable concentration. Table 2a shows that there has been an
and we therefore concentrate on the first two increase in the average HHI and Gini coefficient over
benchmarks where more robust results should be found. the period of analysis. As the distance measures show
the distance between the individual loan books and the
5. Results benchmarks, the closer to 1 that the distance measures
Section 3 presented graphical evidence of the changing are, the greater the difference between each individual
distribution and concentration of the NFC loan book in bank’s loan book and economy structure. A decrease in
recent years. In this section the focus is on the trend in the average distance measure indicates that banks’
individual banks’ loan books over the same time period. lending behaviour more closely reflects the structure of
Specifically, we use the concentration measures outlined the economy as a whole. Table 2a shows the distance
measures using employment in each sector as a
in Section 4 to examine whether statistical techniques
benchmark. It is clear from Table 2a that all measures
show that loan books are becoming more concentrated
suggest the banking system’s NFC loan portfolio has
over time. In their study of German banks, Kamp et al
been moving away from the economy structure in recent
(2005) find that some distance measures gave conflicting
years. Using the other benchmark — Gross Value Added
results on concentration and it is not clear which
— similar results are obtained. Pair-wise correlation
measure should be selected. Due to these findings, we
coefficients presented in Table 2b show that there is a
will present our results for all measures rather than select
strong significant correlation between the distance
a single measure at the outset.
measures using the same sectoral employment
benchmark. However, while the correlation between the
5.1 Summary Statistics distance measures is strong, it is not perfect, and we
Summary statistics for both the traditional and distance therefore present results for all the measures in the
measures are provided in Table 2a. For the traditional analysis below.

Table 2a: Summary Statistics: Concentration Averages, 1999-2006

HHI Gini DM1 DM2 DM3 DM4 DM5

Feb ’99 0.276 0.595 0.300 0.443 0.104 0.537 0.412


Dec ’99 0.257 0.586 0.290 0.447 0.095 0.530 0.403
Dec ’00 0.254 0.584 0.289 0.455 0.094 0.532 0.400
Dec ’01 0.285 0.605 0.366 0.484 0.120 0.519 0.389
Dec ’02 0.282 0.611 0.369 0.483 0.118 0.516 0.382
Dec ’03 0.287 0.629 0.409 0.501 0.130 0.543 0.413
Dec ’04 0.309 0.641 0.428 0.513 0.140 0.561 0.423
Dec ’05 0.323 0.650 0.456 0.522 0.149 0.558 0.422
Dec ’06 0.359 0.674 0.493 0.559 0.172 0.578 0.443

Note: ‘DM’ denotes distance measure.

Table 2b: Summary Statistics: Distance Measure Correlation Coefficients

Measure 1 Measure 2 Measure 3 Measure 4 Measure 5

Measure 1 1
Measure 2 0.9978 1
(0.000)
Measure 3 0.9936 0.991 1
(0.000) (0.000)
Measure 4 0.9543 0.9595 0.9639 1
(0.000) (0.000) (0.000)
Measure 5 0.9541 0.9535 0.9769 0.9839 1
(0.000) (0.000) (0.000) (0.000)

Note: Significance test p-values are in brackets.

120 Financial Stability Report 2007


5.2 Empirical Results The bank-level regression analysis shows that the panel
Regression analysis is used to estimate the change in the estimates are not being driven by a small number of
sectoral make-up of banks’ loan portfolios over time. banks. With the exception of DM5, the majority of banks
First, a fixed-effect panel regression is run to estimate have significant positive results for all measures. These
the time trend in each of the measures.5 The estimates results support the conclusion, drawn from the panel
obtained from this model provide the average behaviour estimates, that banks’ loan portfolios are becoming
of all the banks in the sample. Second, individual more concentrated.
regressions are run for each bank.6 This provides
estimates of the behaviour of individual banks. This The bank-level results also show that a number of banks
analysis is first carried out on the distance measures. The for each measure have significant negative results,
traditional measures of diversification (HHI and Gini suggesting these banks’ lending behaviour is moving
coefficient) are then used to cross-check the results. closer to the benchmark — i.e., becoming less
concentrated. However, these results are largely from
5.2.1 Distance Measures using Employment Figures banks that would have had a very concentrated loan
Benchmark book at the start of the period of analysis. The results
are, therefore, likely to have resulted from their shifting
The results from the analysis using employment by sector
into new sectors over the period of analysis.
as a benchmark are presented in Table 3. The panel
analysis results are positive and significant for all of the
5.2.2 Distance Measures using GVA Benchmark
distance measures, implying that the ‘distance’ between
the benchmark and banks’ loan portfolios is increasing. GVA (Gross Value Added) by sector is the second proxy
As the distance measures take the benchmark — in this for the economy structure used as a benchmark. The
instance employment by sector — as the level of perfect panel data estimates for the GVA benchmark show
diversification, significant positive results such as these significant positive results for all distance measures7.
imply that, on the whole, banks’ NFC loan books have These results correspond with those for the employment
become more concentrated over the period of by sector benchmark in showing a movement towards
analysis. more concentrated loan portfolios.

Table 3: Empirical Results for Employment Benchmark Distance Measures

DM1 DM2 DM3 DM4 DM5

Panel estimates

Panel Coefficeint 0.00678 0.00342 0.00248 0.00181 0.00157


P-value 0.000 0.000 0.000 0.000 0.000

Bank-level regressions

No. banks with trend


Significant @ 5% level 1 2 2 3 3
Diversifying
Significant @ 10% level — — 1 1 —
No Change 2 1 — 2 4
Significant @ 10% level — — — — —
Concentrating
Significant @ 5% level 10 10 10 7 6

Note: ‘DM’ denotes distance measure.

5
Each measure of diversification is regressed on a linear trend variable:
DMtb =  + .time + tb,
where DMtb is the diversification measure of bank b at time t.
6
The trend model used is:
DMtb = b + b.time + tb.
7
DM3 is significant at the 10 per cent level.

Financial Stability Report 2007 121


Table 4: Empirical Results for GVA Benchmark Distance Measures

DM1 DM2 DM3 DM4 DM5

Panel estimates

Panel Coefficeint 0.00174 0.00132 0.00047 0.00150 0.00070


P-value 0.000 0.002 0.075 0.000 0.020

Bank-level regressions

No. banks with trend


Significant @ 5% level 3 3 3 3 4
Diversifying
Significant @ 10% level — — — — —
No Change 3 3 3 2 2
Significant @ 10% level 2 — — 1 —
Concentrating
Significant @ 5% level 10 7 7 7 7

Note: ‘DM’ denotes distance measure.

In the bank-level regressions the majority of banks have 5.2.3 Traditional Measures
significantly positive results. Again, this is similar to the Traditional measures are used to cross-check the results
results from the employment-by-sector benchmark, and from the distance measures above. The results for the
indicates increased concentration in the loan portfolios traditional measures are indeed in line with the findings
of most banks. For each measure, there are also a for the distance measures. Table 2a has already shown
number of banks that have significantly negative results. that the mean of both traditional measures appear to
As above, these are largely banks that had highly have increased over time. This is borne out by a more
concentrated loan portfolios at the start of the sample formal regression analysis results presented in Table 5.
period. The movement of these banks from a very small The panel estimation shows that there was a significant
number of sectors, into a somewhat wider variety, would positive trend in both traditional measures over time.
account for these results. The final point of note arising Similarly, the bank-level regressions show that the
from these results is that a larger number of banks have majority of banks’ loan books have significant positive
insignificant results than for the employment benchmark. results over time. Only two banks show reduced
It is felt this is largely due to drawbacks in the data, concentration. These are, again, banks with very
specifically the annual nature of the data and the need concentrated loan portfolios at the beginning of the
to ‘quarterise’ it, as detailed in Section 4.3. period of analysis.

Table 5: Empirical Results for Traditional Measures

HHI Gini Coefficient

Panel estimates

Panel Trend 0.00349 0.00292


P-value 0.000 0.000

Bank-level regressions

No. banks with trend


Significant @ 5% level 1 1
Diversifying
Significant @ 10% level 1 1
No Change — 1
Significant @ 10% level — —
Concentrating
Significant @ 5% level 11 10

122 Financial Stability Report 2007


6. A Closer Examination of Sectoral Results from Acharya et al are broadly in line with
Concentration Winton, finding that bank returns in general, measured
So far, the focus of this paper has been on using various by a number of indicators, are decreased by a diversified
measures of concentration to determine the trend in the loan book, most particularly when risk (and therefore
distribution of the loan book in recent years. In this part monitoring costs) are high. Acharya et al conclude that
of the paper, we explore a number of pros and cons diversification is not a guarantee of superior
associated with sectoral concentration in loan books in performance or greater bank safety. The effects of
general and in the Irish case in particular. In dealing with concentration on United States financial services firms
sectoral concentration in particular, we review existing are studied by Elyasiani and Deng (2007). They analyse
literature in this area which shows that there are benefits sectoral, geographical and activity diversification. It is
associated with both diversified and concentrated loan found that all types of diversification lead to decreasing
books. For Ireland in particular, we use new international returns as well as decreasing income. Elyasiani and Deng
data to benchmark the sectoral distribution of the Irish find that this relationship is stronger the harder it is for
loan book against our European counterparts, and we banks to monitor their loans. Overall, they suggest that
also present some possible factors that may mitigate risks banks trade off returns for reduced risks.
associated with a concentrated NFC loan book in the
Irish context. On the other hand, if a bank’s loan book is concentrated
in a specific sector, or a number of highly correlated
6.1 Literature on Sectoral Concentration sectors, then a shock affecting debtors’ ability to repay
their loans may cause the bank financial distress. In such
A survey of the existing literature in this area shows that
a situation, banks must hold enough capital to allow for
a tension exists with regard to whether it is in the best
the sectoral concentration of their loan books. Düllman
interests of banks to hold a concentrated or a
and Masschelein (2006) estimate the effect of sector
‘diversified’ loan book. On the one hand, diversification
concentration on economic capital using a sample of
should reduce the risk of financial losses as the bank’s
German banks. They find that, in comparison with a
portfolio encompasses projects across a number of
benchmark portfolio reflecting the concentration of the
unrelated sectors [Diamond (1984)]. On the other hand,
aggregate loan book of German banks, the sector
the monitoring costs associated with lending to a
concentration observed in certain individual banks can
number of sectors may make it cheaper to maintain a
increase economic capital by 37 to 50 per cent. For the
more concentrated loan book [Stomper (2004)].
United States, Heitfield et al (2005) simulate the
Winton (1999) uses a theoretical model to analyse the distribution credit losses for real syndicated loan
trade-off between the benefits of close monitoring and portfolios. They find that sector concentration risk is the
the risks arising from a high level of concentration. It main contributor to economic capital for all portfolios.
appears that holding a diversified loan book is the best
strategy when loans have moderate exposure to sector 6.2 International Comparison of Lending to NFCs
downturns. When downside risks are high, monitoring is
Placing the sectoral distribution of Irish NFC lending in
important, and diversification, by reducing the ability of
an international context also allows us to benchmark the
banks to monitor, has a negative effect on bank
sectoral distribution against some of our European
performance. This implies that monitoring costs have an
counterparts. Cross-country sectoral distribution data
impact on the credit risk of loan portfolios.
have not previously been available on a cross-country
Further evidence of the impact of monitoring costs on basis, and is still not available for all European countries.
a diversified loan book arises from the ‘winner’s curse’ However, for the data available, our international
problem whereby banks experience higher monitoring comparison shows that the structure of the Irish NFC
costs in sectors that they have newly entered. Winton loan book is not dissimilar from our European
(1999) finds some evidence of this, and further counterparts.
theoretical evidence is provided by Dell’Ariccia (2001),
Dell’Ariccia et al (1999), Gehrig (1998) and Marquez To facilitate cross-country comparison we combine
(1997). Schaffer (1998) finds supporting empirical the following sectors: Agriculture and Forestry with
evidence. Fishing, and Mining and Quarrying with Manufacturing.
Also, in the Irish data above, the NACE sector Real
Acharya et al (2002) use data on 105 Italian banks to Estate Renting and Business Activities is separated
test a number of hypotheses put forward by Winton. into Real Estate Activities and Services. It is not possible

Financial Stability Report 2007 123


to make this separation for the other national data However, we do have these data for Ireland and Real
and, therefore, we present the combined sectors in Estate Activities accounts for 90 per cent of the
Chart 3. composite category. The Manufacturing/Mining and
Quarrying sector is the second largest category for the
EA5 accounting for 19 per cent of NFC lending. Of the
Chart 3: Sectoral Distribution of Advances nine countries where data are available, Ireland has the
Comparison, end-2006
lowest proportion of lending to this sector at just under
per cent
100
6 per cent of total NFC lending. One potential reason
for this low level may be due to the large presence of
90
multi-national corporations which may source there
80 funding from parent companies or from non-domestic
banks.
70

60
Lending by Irish banks to the Construction sector, at 16.3
per cent of the NFC loan book, is somewhat above the
50
EA5 average at 12.9 per cent of the total NFC loan book.
40 The variation of this category is large across countries.
Banks in Spain and Portugal have the largest exposure
30
to this sector, 18.2 and 20.6 per cent, respectively.
20 Sweden (2.5 per cent), Germany (4.2 per cent) and the
10 United Kingdom (5.9 per cent) all have relatively low
exposure.
0
GR IT EA5 EA5 PT SP UK IE DE MT SE Despite the large decline in the importance of
(Weighted) (Ave)
Real Estate & Other Business Activities Construction Agriculture to Irish banks’ NFC loan books in recent
Wholesale/Retail Trade & Repairs Hotels & Restaurants years (described above), the share of Agriculture is still
Manufacturing, Mining & Quarrying Agriculture, Fishing & Forestry
marginally above the euro area average. Ireland has a
Transport, Storage & Communications Electricity, Gas & Water Supply
higher exposure to the Hotels and Restaurants sector
Source: Bank of Greece, Banca D’Italia, Banco de Portugal,
Banco de Espana, Bank of England, CBFSAI, Deutsche than any other country with the United Kingdom having
Bundesbank, The Central Bank of Malta, Sveriges Riksbank, the second largest. The EA5 Wholesale and Retail Trade
Authors’ calculations
Note: 'EA 5 (Weighted)' and 'EA 5(Ave)' refer to the weighted lending is proportionally much larger than in Ireland.
and simple average of the 5 euro area countries on the chart Ireland also has a below average exposure to the
respectively.
Transport, Storage and Communications sector.

6.3 Implications in Ireland


We present the data for 9 European countries (6 euro
As outlined above, sectoral concentration may have
area, including Ireland) where data are available in Chart
positive or negative financial stability implications. The
3. Using the five other euro area countries (Germany,
high level of sector concentration may increase the
Italy, Portugal, Spain and Greece) we also calculate a
correlation between the performance (or non-
euro area 5 (EA5) average sectoral distribution, both
performance) of loans in the banks’ portfolio. If the
weighted and unweighted. The unweighted data are
shock is large enough to lead to non-performance of
useful as Germany and Italy account for a very large loans, then banks may find that their level of non-
weight in the weighted data and hence strongly bias performing loans is sensitive to a shock to these sectors.
the results. Thus, in the event of a shock to the Construction and
Real Estate sectors, banks may find that the performance
With the exception of Greece, the diverse sector of Real of these loans might be correlated and asset quality
Estate, Renting and Business Activities is the largest in could deteriorate in many loans. An additional risk is that
each country. Ireland’s exposure to this category is the rates of increase in prices across different segments
higher than the European average, however, it is below of the Irish property market have tended to move
that for Germany and Sweden. As mentioned above, for upwards and downwards over the last thirty years,
the vast majority of countries it is not possible to particularly during those periods of slower economic
separate Real Estate from Services in this category. growth.8 This includes commercial property and
8
See Kearns, A. and M. Woods, (2006).

124 Financial Stability Report 2007


residential property, which is the dominant exposure of Our findings reveal that, for the most part, both the
banks to the household sector. traditional and distance measures give the same results
in the Irish context. Using both measures, most banks
However, there are several possible mitigating factors showed an increase in concentration in the NFC loan
that may limit the risks in Ireland. Firstly, there is some portfolio. Both sets of measures show a small number of
within-sector dispersion that may reduce the correlation banks moving to a more diversified portfolio, however;
of loan performance. For instance, the Real Estate and these shared the characteristics of beginning the period
Construction sectors comprise the residential real estate as very specialised banks and they have realigned their
sector, the commercial Real Estate sector, property portfolios more in line with the market.
companies investing overseas, and infrastructure
construction companies. Correlation would be lowered In the second part of this paper we examined a number
by construction/real estate companies borrowing to
of pros and cons associated with the current sectoral
invest abroad in markets that may be subject to a
distribution of the Irish NFC loan book. Firstly, a review
different business cycle to the domestic market.
of the existing literature in this area is divided with regard
Infrastructure companies may encounter a positive
to whether, from a financial stability perspective,
business climate even in the event of a slowdown in the
diversification or concentration is a better strategy for
real estate sector. In the particular case of Ireland, a large
banks. Secondly, when a comparison is made with other
infrastructure spending programme in the National
European countries, it is found that the structure of the
Development Plan is scheduled for the coming years that
Irish NFC loan portfolio is not significantly different from
could partly guard against any decline in the other
subsectors such as residential construction. that of other countries. The share of NFC loans to the
Irish Construction sector is significantly higher than in
the UK and Germany, but less than Spain and Portugal.
Secondly, the Construction and Real Estate sectors
account for an increasing share of the loan book where Ireland’s Manufacturing sector accounts for the lowest
there may be a higher likelihood of loans being share of any NFC loan book where data was available.
collateralised than is perhaps the case in other sectors. Finally, there are a number of possible mitigating factors
This will have the effect of acting as an increased buffer in the case of Irish banks: within-sector dispersion that
to banks. may reduce the correlation of loan performance, the
higher likelihood of loans being collateralised in the
Finally, it should be kept in mind that our study is largest sectors, and narrowed scope of the data in
confined to a section of the Irish banking system’s total dealing only with the NFC portion of the loan book.
loan book — lending to Irish non-financial corporations.
We therefore exclude lending to other sectors such as References
public sector, financial sector and households as well as
lending to non-Irish residents. Acharya, V. V., I. Hasan and A. Saunders, (2002), ‘‘The
Effects of Focus and Diversification on Bank Risk and
Return: Evidence from Individual Bank Loan
7. Conclusion
Portfolios’’, March 2002.
The recent rapid increase in credit growth to the Irish
NFC sector has been accompanied by a shift in the
Central Bank & Financial Services Authority of Ireland,
sectoral distribution of lending. A comparison of the
(2007), ‘‘Sectoral Developments in Private-Sector
distribution of lending in 1999 and 2006 showed that at
Credit, June 2007’’, www.centralbank.ie.
the aggregate level, only two sectors, Construction and
Real Estate, have increased their share of the NFC loan
portfolio over time. These gains have primarily come at Dell’Ariccia, G., (2001), Asymmetric Information and the
the expense of the Agriculture and Manufacturing Structure of the Banking Industry’’, European
sectors. In the first part of this paper, we used a number Economic Review, 45, pp. 1957-1980.
of measures of concentration to determine whether Irish
banks’ loan portfolios have become more or less Dell’Ariccia, G., E. Friedman and R. Marquez, (1999),
concentrated in recent times. The measures employed ‘‘Adverse Selection as a Barrier to Entry in the Banking
fall into two distinct groups (traditional and distance Industry’’, RAND Journal of Economics, 30, pp. 515-
measures) that assume different theories on the ideal 534.
structure of lending portfolios.

Financial Stability Report 2007 125


Deng, S. and E. Elyasiani (2007), ‘‘Diversification and Kearns, A. and M. Woods, (2006), ‘‘The Concentration
Performance of the U.S. Commercial banks’’, April in Property-Related Lending — a Financial Stability
2007. Perspective’’, Financial Stability Report 2006, CBFSAI,
pp. 133-144.
Diamond, D., (1984), ‘‘Financial Intermediation and
Delegated Monitoring’’, Review of Economic Studies, Marquez, R., (1997), ‘‘Lending Capacity and Adverse
51, pp. 393-414. Selection in the Banking Industry’’, MIT, Working
Paper.

Düllmann, K., and N. Masschelein, (2006), ‘‘Sector


Pfingsten, A. and K. Rudolph, (2002), German Banks’
Concentration in Loan Portfolios and Economic
Loan Portfolio Composition: Market Orientation Vs
Capital’’, National Bank of Belgium, Working paper
Specialisation’’, Institut für Kreditwesen, Münster,
research No. 105, November 2006.
Tech. Rep. 02-02.

Gehrig, T., (1998), ‘‘Screening, Cross-Border Banking, Schaffer, S., (1998), ‘‘The Winner’s Curse in Banking’’,
and the Allocation of Credit’’, Research in Economics, Journal of Financial Intermediation, vol. 7, issue 4, pp.
52, pp. 387-407. 359-392.

Heitfield, E., S. Burton and S. Chomsisengphet, (2005), Stomper A., (2004), ‘‘A Theory of Banks’ Industry
‘‘The Effects of Name and Sector Concentration on Expertise, Market Power, and Credit Risk’’, University
the Distribution of Losses for Portfolios of Large of Vienna.
Wholesale Credit Exposures’’, October 2005.
Winton, A., (1999), ‘‘Don’t Put All Your Eggs in One
Kamp, A., A. Pfingsten and D. Porath, (2005), ‘‘Do Banks Basket? Diversification and Specialization in Lending’’,
Diversify Loan Portfolios? A Tentative Answer Based Finance Department, University of Minnesota,
on Individual Bank Loan Portfolios’’, Deutsche Minneapolis.
Bundesbank, Discussion Paper, Series 2: Banking and
Financial Studies, No. 03/2005.

126 Financial Stability Report 2007


Annex
Table 1: Sub-sectors included in the Non-Financial Table 2. Banks Covered in Study
Corporate Sector
ACC Bank plc
Allied Irish Bank plc
NACE code Sub-sector
Anglo Irish Bank Corporation plc
A Agriculture & Forestry The Governor and Company of the Bank of Ireland
B Fishing Bank of Scotland (Ireland) Ltd
C Mining & Quarrying Danske Bank A/S
D Manufacturing EBS Building Society
E Electricity, Gas & Water supply First Active plc
F Construction ICS Building Society
G Wholesale/Retail Trade & Repairs IIB Bank plc
H Hotels & Restaurants Irish Nationwide Building Society
I Transport, Storage & Communications Irish Life & Permanent plc
K1 Real Estate Ulster Bank Ireland Ltd
K minus K1 Services

Financial Stability Report 2007 127

Вам также может понравиться