Content uploaded by Elias Raad
Author content
All content in this area was uploaded by Elias Raad
Content may be subject to copyright.
INTERNATIONAL JOURNAL OF BUSINESS, 9(3), 2004 ISSN: 1083-4346
The Performance of Banks in Post-war Lebanon
David Peters
a
, Elias Raad
b
, and Joseph F. Sinkey, Jr.
c
a
Faculty of Social Sciences, University of Western Ontario, dwpeters@uwo.ca
b
School of Business, Lebanese American University, eraad@lau.edu.lb
c
Terry College of Business, The University of Georgia, jsinkey@terry.uga.edu
ABSTRACT
This paper analyzes the performance and balance-sheet characteristics of banks in post-
war Lebanon for the years 1993 to 2000. Although we find that Lebanese banks are
profitable, most of them had accounting return on assets (ROA) greater than one
percent over most of our test period, they are not as profitable as a control group of
banks from five other countries located in the Middle East. Bank safety and soundness
in Lebanon has increased as leverage was reduced (capital adequacy improved) and a
risk index indicates lower probabilities of book-value insolvency. We attribute this
improved bank performance and safety to better management and to three external
factors: political (cessation of war), economic (lower inflation), and regulatory (BIS
capital requirements). We employ regression models that relate bank profitability ratios
to various explanatory variables. We find, for example, that ROA is positively
associated with lagged growth in real GDP, spread or net interest margin, and holdings
of Lebanese T-bills but negatively related to bank size as measured by the natural log of
total assets. As a policy implication, we recommend that Lebanese banks increase their
lending to the private sector to achieve a more efficient allocation of resources and to
stimulate economic growth. To help achieve this objective, Banque du Liban, the
central bank, should abandon its practice of setting T-bill rates above market levels,
which provides a disincentive to bank lending.
JEL: G1, G2, L1, O1
Keywords: Lebanese banks; Financial performance; Bank capital; Safety and
soundness
Copyright2004 by SMC Premier Holdings, Inc. All rights of reproduction in any form reserved.
260 Peters, Raad, and Sinkey
I. INTRODUCTION
Profitability in the banking sector has been extensively examined in developed
countries, especially in North America and Europe. Evidence from these studies shows
that bank profitability depends on several factors. Economic and financial data
published by Banque du Liban (the central bank of Lebanon) and Lebanese banks
indicate that the Lebanese banking industry has been among the most profitable
economic sectors in Lebanon in the post civil-war era. Since no academic work
investigating the performance of the banking sector in Lebanon has been published, this
paper fills that void. We examine the performance of Lebanese banks during the years
1993 through 2000 and for a control group of banks from five other countries in the
Middle East for the years 1995 through 1999. Paucity of information prohibits an exact
data matching for the two groups. The return-on-equity accounting model (market data
are not available) and a risk-index proposed by Hannan and Hanweck (1988] provide
two of the methodological foundations for the univariate analysis. We estimate multiple
regression models to attempt to explain the profitability of Lebanese banks. To provide
background for the investigation, we also describe the characteristics of the Lebanese
economy and its financial sector before, during, and after the civil war in Lebanon.
At the end of the civil war in Lebanon (1990), Charbaji, Mikdashi, and Chebaro
[1994] describe Lebanese banks as having “suffered a severe decline in activity and
profitability and dissolution of financial capital” (p. 86). Since then, we find that
Lebanese banks are profitable but that the high ROEs in the first part of our test period
are due more to leverage than profitable use of assets. In addition, we report that 52
control banks across five Middle East countries have greater asset efficiency (higher
ROAs) and use less leverage. Various risk measures indicate that bank safety and
soundness has improved in Lebanon as, for example, probabilities of book-value
insolvency have declined. We attribute this improvement in bank safety to greater
recognition by management of the importance of bank capital adequacy and to three
external factors: political (cessation of war), economic (lower inflation), and regulatory
(e.g., BIS capital requirements). Behind these performance measures, we document and
analyze the portfolio characteristics that drive these results. Lebanese banks tend to
follow a practice of using the money they receive in Lebanese pound deposits to invest
in Lebanese treasury bills, and using the money they receive in US dollar deposits to
make US dollar loans.
The paper proceeds as follows. The next section briefly describes the Lebanese
economy and banking sector before and during the civil war. The third section focuses
on these factors during the post-war period. The fourth section presents the data and our
methodological approach. The fifth section analyzes our findings while the sixth section
summarizes and concludes.
II. BANKING AND ECONOMIC ACTIVITY BEFORE AND DURING THE
CIVIL WAR IN LEBANON
1
Prior to the civil war, specifically between the 1950s and mid-1970s, real GDP in
Lebanon grew at an average annual rate of six percent and the annual inflation rate was
INTERNATIONAL JOURNAL OF BUSINESS, 9(3), 2004 261
a modest three percent. Over this period, the Lebanese banking industry flourished due
to a strict banking secrecy law, which was introduced in 1956, and the flow of the
petrodollars from the Arab world, a free-market economy, and a free exchange-rate
system. In fact, Lebanon was the banking center of the Middle East. At one time, 100
domestic and foreign banks operated in Lebanon in the mid-1960s, mainly in Beirut,
which also flourished as the social and cultural center of the region under the rubric
“Paris of the Middle East.”
The civil war, which started in April of 1975 and ended in October of 1990, had
devastating effects on the Lebanese economy. A substantial proportion of the Lebanese
labor force was killed or left the country. The Lebanese industrial, agricultural, and
tourism sectors were devastated. Nominal per capita income was around $820 in 1990,
equivalent, in real terms, to about one-third of the per capita income in 1975.
Table 1−Panel A
Economic and financial data
ECONOMIC GROWTH IN LEBANON (1991-2000)
Year
GDP
(Trillions of
Lebanese
Pounds)
GDP
(Equivalent
in Billions of
US$)
Average
Exchange
Rate (Cost of
US$ in LP)
Real GDP
Growth
Inflation Rate
(GDP
Deflator)
1991 4.1 4.5 928 38.2% 51.5%
1992 9.5 5.5 1,713 4.5% 120.0%
1993 13.1 7.5 1,741 7.0% 29.1%
1994 15.3 9.1 1,680 8.0% 8.0%
1995 18.0 11.1 1,621 6.5% 10.6%
1996 20.4 13.0 1,571 4.0% 8.9%
1997 22.9 14.8 1,539 4.0% 7.7%
1998 24.5 16.2 1,516 3.0% 4.0%
1999 24.8 16.4 1,508 1.0% 0.2%
2000 24.8 16.4 1,508 0% 0%
Source: Banque du Liban, "Financial Markets Handbook"
The economic consequences of the civil war extracted a heavy price from the
Lebanese banking sector as well. Specifically, hyperinflation and severe depreciation in
the value of the currency led to disintermediation and declining loan values. For
example, the average exchange rate depreciated from US$ = 2.29 LP in 1975 to US$ =
701.76 LP in 1990. Since 1998, the exchange rate has been pegged around US$ = 1,508
LP (Table 1, Panel A). Many bank customers pulled their funds out of Lebanese banks
and reinvested them in banks and other financial institutions abroad (disintermediation).
With bank borrowers under stress, bank profits turned to losses as loan default rates
soared. Capital resources of Lebanese banks, the final line of defense for any bank,
262 Peters, Raad, and Sinkey
were reduced dramatically and most banks became undercapitalized. In this
environment, some banks went out of business. Nevertheless, the majority of Lebanese
banks managed to survive the civil war and some of them even managed to open
branches in foreign countries. Not surprisingly, foreign banks, such as Bank of
America, Citibank, First National Bank of Chicago, and Bank of Nova Scotia withdrew
from Lebanon during the civil war.
Table 1−Panel B
Economic and financial data
Summary of Lebanese Pound (LP) / US Dollar Exchange Rates, Average Yields on Various Financial
Instruments and Inflation Rates, 1993-2000
Variable 1993 1994 1995 1996 1997 1998 1999 2000
LP/US$ Exchange Rate at end of year 1,711 1,647 1,596 1,552 1,527 1,508 1508 1508
Average Rates on 3 months t-bills (in
LP) %
18.70 15.09 18.88 15.19 13.42 12.70 11.57 10.88
Average Rates on 6 months t-bills (in
LP) %
19.94 17.21 20.65 16.93 14.30 13.78 12.74 11.43
Average Rates on 12 months t-bills
(in LP) %
21.42 18.67 24.59 17.88 15.25 15.17 14.38 11.84
Average Rates on 24 months t-bills
(in LP) %
25.32 19.23 23.36 22.79 16.83 16.72 15.89 14.14
Average Rate on all T-bills (in LP) % 21.35 17.55 21.87 18.2 14.95 14.59 13.65 12.07
Average Rates On Bank Deposits in
LP %
12.78 13.16 15.14 14.71 12.68 12.97 11.94 10.68
Average Discount and Loans Rates in
LP %
28.53 23.88 24.52 25.21 20.29 20.24 19.48 18.15
Spread Between Average Rates on t-
bills and Rates on Deposits (in LP) %
8.57 4.39 6.73 3.49 2.27 1.62 1.71 1.39
Spread Between Average Rates on
Discount and Loans and Average
Rates on Deposits in LP %
15.75 10.72 9.38 10.50 7.61 7.27 7.54 7.47
Repo Rate (in LP) % 31.61 30.00 43.75 27.83 27.08 30.00 27.92 20.83
Inter-Bank Rate in LP % 6.60 7.33 34.88 11.19 13.00 11.23 7.46 7.58
Average Discount and Loans Rates in
US$ %
NA NA 12.03 11.91 11.76 11.54 10.95 11.19
Average Rates On Bank Deposits in
US$ %
NA NA 5.29 5.41 5.72 5.89 5.60 5.92
Spread Between Average Rates on
Discount and Loans and Average
Rates on Bank Deposits in US$ %
NA NA 6.74 6.5 6.04 5.65 5.35 5.27
Rate of Inflation % 29.1 8.0 10.6 8.9 7.7 4.0 0.2 0.0
Note: A few outliers exist in the data. The average inter-bank rates were 275% and 43% in May and August,
1995, respectively. The average inter-bank rates were 45% and 36% in September 1997 and December 1998,
respectively.
Sources: Banque du Liban website, Banque du Liban, "Financial Markets Handbook"
INTERNATIONAL JOURNAL OF BUSINESS, 9(3), 2004 263
During the war, the Lebanese economy became dollarized and those customers
that continued to support Lebanese banks tended to keep their funds in US dollar
deposits rather than Lebanese pound deposits. With slack loan demand and the
government trying to fund the war effort, Lebanese banks increased substantially the
proportion of their assets held in Lebanese treasury bills, which coincidently offered
high yields. Although foreign investments in the country were reduced, a substantial
amount of funds kept flowing into the country, either from Lebanese living and
working abroad or from parties and countries that were helping to finance the militias
that were fighting each other.
By the end of the civil war, although total assets of the Lebanese banking sector
had increased, in nominal terms, from about $4.7 billion in 1975 to $5.7 billion in 1990,
they had declined in real terms because of inflation. Regarding personnel and facilities,
a paucity of both existed as Lebanese bank employees lacked modern training and
facilities lagged substantially behind those in developed countries.
After the cessation of the civil war in October of 1990, some of the foreign
banks that had left Lebanon during the war returned, including Citibank, Bank of Nova
Scotia, and ING Bank. In addition, new foreign banks opened branches in Lebanon,
e.g., Commerzbank and Berliner Bank. The resurgence of the banking sector, including
a substantial amount of foreign investment, provided the funds needed to begin
rebuilding Lebanon. As vital as financial capital is to a banking system, the restoration
of reputational capital (confidence) also was critical. With confidence in the banking
sector growing, the future of Lebanon began to look brighter. We now turn to a look at
Lebanon during the post-war period.
III. THE LEBANESE ECONOMY AND BANKING ENVIRONMENT DURING
THE POST-WAR PERIOD
With a population of nearly four million and a gross domestic product equivalent to
US$16.4 billion (1.7% of US GDP), Lebanon had a per capita income of roughly
$4,100 in 2000 compared to $35,000 in the US. Four sectors form the backbone of the
economy: agriculture, trade, services, and tourism. Although the civil war in Lebanon
ended in 1990 and the political and social situation in Lebanon improved during the
1990s, the overall environment has not been conducive to economic growth. This
shortcoming restricted the economy’s potential, especially with respect to tourism,
which until the late 1990s was depressed because of uncertainty regarding safe travel
conditions.
2
Recently, however, Beirut has been attempting to reclaim its title as the
“Paris of the Middle East.”
Panel A of Table 1 shows some basic data for the Lebanese economy for the
years 1991 to 2000. Nominal GDP (in US$) grew at an annual average rate of 12
percent over this period but inflation cut real growth considerably to 4.2 percent per
annum (excluding 1991). During 2000, although the inflation rate was down to zero
percent (from 120 percent in 1992), real GDP growth also was zero percent. During the
1990s and continuing today, the Lebanese economy faces challenges with respect to (1)
government-budget deficits, (2) current-account deficits, (3) high interest rates, and (4)
high unemployment. We address these four areas next.
264 Peters, Raad, and Sinkey
A. Budget Deficits
The Lebanese public infrastructure was badly damaged during the civil war. In the
years following the conflict, the government incurred large capital expenditures,
financed mainly by issuing treasury bills and notes rather than by taxes. Since 1994,
deficits have been around LP 3 trillion (equivalent to US$2 billion) every year. Even
though capital expenditures have been reduced, the government is making substantial
interest payments on its accumulated debt. The current net total debt of Lebanon stands
at LP 45.4 trillion (equivalent to US$30 billion), more than 100 percent of the nation's
gross domestic product.
B. Current-Account Deficits
Lebanon has had large trade deficits in recent years. In 2000, for example, exports of
goods amounted to the equivalent of US$714 million while imports of goods amounted
to the equivalent of US$6.2 billion. Lebanon's exports are primarily to countries in the
Middle East while its imports are primarily from European countries. The Lebanese
economy has a substantial surplus in services, primarily derived from tourism.
3
Despite
having substantial current-account deficits, a positive balance of payments has existed
in the post-war period. This positive balance has been propped up by external capital
inflows and from remittances by Lebanese living outside Lebanon to family members
in Lebanon.
4
C. High Interest Rates and High Unemployment
Large government borrowings during the post-war period have put substantial pressures
on financial markets. To induce investors to buy treasury bills and notes, the Lebanese
government offers high interest rates on its debt obligations. Panel B of Table 1 shows
that the yield on three-month treasury bills has averaged 14.6 percent while the yield on
two-year treasury notes has averaged 19.29 percent for the years 1993 to 2000. The
average rate on all treasury obligations has declined, however, dropping from 21.4
percent in 1993 to 12.07 percent in 2000. The combined effects of high interest rates,
rising wages, and a currency that has risen vis-a-vis its major trading partners has
dampened real economic growth over the past few years. Although inflation has slowed
down considerably, the economy is still experiencing substantial unemployment.
INTERNATIONAL JOURNAL OF BUSINESS, 9(3), 2004 265
D. Financial Markets and Interest Rates
Financial markets in Lebanon are underdeveloped compared to industrialized nations in
Europe or standards in North America. Lebanon has one stock exchange, the Beirut
Stock Exchange, but only 14 companies with a total market capitalization of
approximately $3 billion were listed on the exchange at the end of 2000. The stocks of
six additional companies were traded over the counter. Although no bond market exists
in Lebanon, some large Lebanese banks issue bonds in European markets. In addition,
there is no money market in Lebanon. Moreover, since the Lebanese government's
treasury bills and notes are only quasi-marketable securities, no secondary market exists
where investors can trade treasury securities. The only liquidity for treasury securities is
provided by the central bank, which buys back its obligations but only at a substantial
discount.
Panel B of Table 1 shows the average yields on three-month and 24-month
treasury obligations and average interest rates on Lebanese pound loans and deposits
for the years 1993 to 2000. Overall, interest rates were high and volatile compared to
those in other industrialized countries throughout the period. For example, the yield on
the three-month treasury bills averaged 18.7 percent in 1993, fell to 15.1 percent in
1994, went up to 18.9 percent in 1996, but then fell to 10.9 percent in 2000. The yields
on treasury bills of other maturities followed a similar pattern.
The average interest rates on discounts and loans denominated in Lebanese
pounds were exceptionally high from 1993 to 2000. For example, the average LP loan
rate was 28.5 percent in 1993 but fell to 18.2 percent in 2000.
5
In contrast, the average
LP deposit rate was 12.8 percent in 1993, went as high as 15.1 percent in 1995, but
dropped to 10.7 percent in 2000. The data show that the interest rates paid on LP loans
are less volatile than the yields on Lebanese treasury bills. For purposes of comparison,
Panel B of Table 1 also presents the inflation rate for each year.
Lebanese banks offer loans and deposits denominated in US dollars (USD) as
well as those denominated in Lebanese pounds. The average interest rates on USD
loans and deposits are much lower than those denominated in LP, but are still higher
than the yields on financial instruments in the US. Panel B of Table 1 shows the
average USD loan and deposit rates for the years from 1995 to 2000.
6
Rates on deposits
ranged from 5.29 percent (1995) to 5.92 percent (2000) while loan rates ranged from
12.03 percent (1995) to 11.19 percent (2000).
E. Business Structure
Most businesses in Lebanon are family-owned organizations that mainly finance
themselves by issuing stock to family and friends and by obtaining loans from banks.
As a result, banks should play an important role in commercial finance but, as we
document below, they come up short in their pivotal role of allocating resources more
efficiently. The main problem is the high yield on Lebanese treasury obligations, which
discourages lending to the private sector. The high yields are costly to investors,
however, as Lebanese treasury obligations do not have well-developed secondary
markets and the central bank only liquidates them at a substantial discount.
266 Peters, Raad, and Sinkey
F. Banking Structure and Regulation
There are five major types of financial intermediaries in Lebanon: (1) commercial
banks, (2) banks specializing in medium and long-term credit, (3) money dealers, (4)
financial institutions that would be regarded as investment banking firms, and (5)
brokerage firms. In addition to these institutions, a number of foreign banks have
representative offices in Lebanon. As of year-end 2000, a total of 59 commercial banks
existed in Lebanon, 17 of which were subsidiaries of foreign banks. Seven banks
specialize in medium- and long-term credit by providing residential mortgage loans or
long-term loans to businesses secured by property. Money dealers are primarily small
retail foreign-exchange dealers and do not provide loans or take deposits. At the end of
2000, there were 28 financial institutions in Lebanon that would be referred to as
investment banks in North America; these firms are relatively small in size and assist
firms in obtaining long-term financing and also trade securities on behalf of customers.
In addition, there were five brokerage institutions that trade securities on behalf of
customers but do not underwrite securities.
Banks are required to have shareholders capital of at least 10 billion Lebanese
pounds (US$6.7 million) plus 250 million Lebanese pounds (US$167,000) for each
branch.
7
Banks are required to keep reserves of 13 percent, but some of these reserves
can be invested in treasury bills. Banks are also required to comply with the BIS risk-
based-capital (RBC) requirements.
IV. DATA AND METHODOLOGY
We draw the initial sample of banks for this study from issues of Bilanbanques
8
from
1993 to 2001. Similar to other countries, individual banks furnish the financial data for
these reports. All banks in Lebanon are included in our sample except those (1) owned
by the government, (2) classified as investment banks, or (3) do not report income-
expense and balance-sheet data in Bilanbanques.
9
The number of banks included in our
sample varies by year and ranges from 66 to 54 with 65 in 1993, 65 in 1994, 66 in
1995, 64 in 1996, 62 in 1997, 52 in 1998, 59 in 1999, and 54 in 2000. Overall, our
sample captures approximately 92 percent of all commercial banks in Lebanon.
10
We
compare our sample banks with a control group of 52 banks from five countries,
including United Arab Emirates (18 banks), Saudi Arabia (10), Kuwait (8), Bahrain (9),
and Oman (7). Our comparisons, however, are limited because we only have two
profitability measures and a leverage factor for our control banks.
A. Descriptive Methods
We use several approaches to analyze the performance of our sample banks, including
univariate (descriptive) tests and multivariate analyses. First, we employ the accounting
return-on-equity (ROE) model to investigate profitability and leverage, that is, ROE =
ROA x EM, where ROA = return on assets and EM equals the equity multiplier. Three
variables, total assets, total equity, and profits, define these three ratios with ROE =
profits/equity, ROA = profits/assets, and EM = assets/equity. In addition, we analyze
INTERNATIONAL JOURNAL OF BUSINESS, 9(3), 2004 267
the components of bank profitability by focusing on net interest margin (NIM = net
interest income/total assets) and operating expenses. Since the portfolios of assets and
liabilities that banks hold drive these cash flows, we also examine these balance-sheet
characteristics.
To investigate bank riskiness and the probability of book-value insolvency, we
employ the risk index suggested by Hannan and Hanweck [1988] and used by various
other researchers, for example, Liang and Savage [1990], Eisenbeis and Kwast [1991],
Sinkey and Nash [1993], Nash and Sinkey [1997], and Blasko and Sinkey [2004]. The
empirical version of the risk index we employ is calculated as follows:
ROA
1
/)EMAOR(RI σ+=
−
(1)
where ROA bar is average return on assets,
11
EM
-1
is the reciprocal of EM or the ratio
of shareholders' equity to total assets, and σ
ROA
is the standard deviation of ROA.
Hannan and Hanweck [1988] derive the upper bound probability of book value
insolvency (p) and show that it equals 1/[2(RI)
2
]. Thus, for example, if a bank has RI =
2.0, its probability of book-value insolvency would be 12.5 percent; in contrast, if RI =
50, then p = 0.02 percent.
B. Regression Framework
To complement the univariate methods described above, we employ multiple regression
analysis to attempt to explain the profitability of Lebanese banks over the years 1993-
2000. Our models build on the existing empirical literature including Bourke (1989),
Molyneux and Thornton (1992), Berger (1995), Goldberg and Rai (1996), Neely and
Wheelock (1997). Using ROE and ROA as alternative dependent variables, we test
various regression models in which independent variables attempt to capture
relationships between profitability and the level of interest rates, capital structure,
concentration, government ownership and changes in per capita income. In addition, we
test the relationship between bank profitability and size,
12
asset portfolio composition,
off-balance sheet items, ownership by a foreign bank, and the ratio of employment to
assets. We do not test the relationship between profitability and changes in per capita
income or concentration, since we lack detailed data to test regression models that
include our control group of banks from outside Lebanon. In addition, we do not test
the relationship between profitability and government ownership because the few
government-owned banks in Lebanon are special-purpose banks that are not included in
our sample of banks.
V. EMPIRICAL FINDINGS
This section first presents our univariate or descriptive findings followed by our
regression results.
268 Peters, Raad, and Sinkey
A. Balance-Sheet Characteristics
Lebanese banks had average total assets of 1,316 billion Lebanese pounds
(approximately US$869 million) at the end of 2000. Total assets grew at a compound
annual rate of 25 percent from 1993 through 2000. In contrast, since total equity capital
grew at a compound rate of almost 48 percent, our sample banks reduced their leverage
substantially. All other things being equal, Lebanese banks became safer (i.e., the
traditional notion of bank capital as a buffer or cushion for absorbing losses) over our
test period as EM declined from 46.7 (1993) to 15 (2000) based on ratios of the
aggregate data. Averages of individual bank’s EMs show a similar improvement as EM
(= ROE/ROA) declined from 79.69 (1993) to 7.8 (2000). Based on either measure, we
attribute this improvement in bank safety to four factors: political (cessation of war),
economic (lower inflation), regulatory (BIS capital requirements), and managerial
(greater recognition of the importance of capital adequacy and risk management).
The assets of banks in Lebanon can be classified into four broad categories: (1)
treasury bills denominated in Lebanese pounds (LP), (2) loans denominated in LP, (3)
loans denominated in foreign currencies (USD), and (4) other assets. Table 2 shows the
average proportion of total assets invested in each of these asset-classes for the banks in
our sample.
The macroeconomic environment and local traditions have some interesting
effects on the portfolio composition of Lebanese banks. Because of huge government
deficits and the decline of the Lebanese pound during the civil war, many Lebanese
people distrust their own currency. Table 2 shows loan and deposit data that capture
these effects. For example, at year-end 2000, the average deposit-to-asset ratios were 27
percent in LP but 49 percent in USD. The highest LP deposit-to-asset ratio was 30
percent (1996) while the highest USD deposit-to-asset ratio was 55 percent (1993). The
loan-to-asset ratios reveal an even stronger preference for USD loans over LP loans.
For example, at year-end 2000, the ratios were 26 percent and 4.7 percent, respectively.
A loan-to-asset ratio of only 30.7 percent would be considered small (and unacceptable)
in industrialized nations. Bank lending in Lebanon is primarily to business firms. The
residential mortgage market is small in Lebanon and banks have only a small
proportion of their assets invested in such loans; traditionally, young people in Lebanon
buy homes by obtaining financing from relatives rather than borrowing from financial
institutions.
13
If Lebanese banks have less than 31 percent of their assets in the form of loans,
what do they do with their funds? The high yield on LP treasury securities (Table 1,
Panel B) makes them an attractive investment for Lebanese banks. Table 2 shows that
our average sample bank had 21 percent of its total assets invested in LP treasury
securities at the end of 2000. Over our test period, this ratio ranged from 21 percent
(1993) to 29 percent (1996).
The composition of total assets also varied among banks as shown by the
standard deviations of the balance-sheet ratios. The most obvious trend over time has
been the relative increase in the proportion of total assets invested in treasury bills and
the relative decrease in the proportion of assets invested in other assets.
INTERNATIONAL JOURNAL OF BUSINESS, 9(3), 2004 269
Table 2 Statistical description of selected balance sheet ratios. Total Equity to Total Assets (Equity/Assets), Investment in Lebanese Pounds
Treasury Bills to Total Assets (LPtbills/Assets ), Lebanese Pounds Discounts and Loans to Total Assets (LPloans/Assets), Foreign Currencies
Discounts and Loans to Total Assets (FCloans/Assets ), Lebanese Pounds Deposits to Total Assets (LPdeposits/Assets), Foreign Currencies
Deposits to Total Assets (FCdeposits/Assets), Other Assets to Total Assets (OtherAssets/Assets), and Off-Balance Sheet Assets to Total Assets
(Off-balance /Assets).
1993 1994 1995 1996 1997 1998 1999 2000
Variable Mean Stand Mean Stand Mean Stand Mean Stand Mean Stand Mean Stand Mean Stand Mean Stand
N=65 Dev N=65 Dev N=66 Dev N=64 Dev N=62 Dev N=59 Dev N=59 Dev N=54 Dev
Equity/Assets
1
0.036 0.05 0.042 0.04 0.10 0.14 0.086 0.053 0.118 0.105 0.115 0.10 0.099 0.072 0.094 0.063
LPtbills/Assets 0.21 0.10 0.255 0.11 0.25 0.11 0.29 0.12 0.235 0.116 0.24 0.12 0.30 0.14 0.214 0.11
LPloans/Assets 0.043 0.56 0.056 0.064 0.06 0.077 0.057 0.058 0.061 0.062 0.052 0.065 0.056 0.074 0.047 0.056
FCloans/Assets 0.26 0.12 0.25 0.10 0.26 0.12 0.25 0.095 0.25 0.10 0.26 0.10 0.27 0.11 0.26 0.11
LPdeposits/Assets 0.23 0.10 0.27 0.11 0.254 0.11 0.30 0.12 0.24 0.12 0.233 0.125 0.32 0.15 0.27 0.144
FCdeposits/Assets 0.55 0.17 0.47 0.15 0.446 0.16 0.427 0.14 0.46 0.14 0.45 0.16 0.44 0.15 0.494 0.131
O
therAssets/Assets
2
0.48 0.15 0.44 0.15 0.426 0.154 0.40 0.15 0.454 0.14 0.45 0.16 0.374 0.14 0.48 0.107
O
ff-balance/Assets
3
0.10 0.09 0.11 0.10 0.11 0.10 0.11 0.12 0.11 0.13 0.12 0.13 0.094 0.077 0.09 0.08
N
is the sample size.
1
The equity-to-asset ratio for 1995 excludes an outlier (new bank) with a ratio of 0.98.
2
Other Assets are equal to (Total assets - Lebanese Pounds T-bills - Lebanese Pounds Loans - Foreign Currencies Loans). This residual category
includes 15 items: (1) Cash and Deposits with Central Bank, (2) Bonds and Fixed-Income Investments (other than Lebanese Treasury Securities), (3)
Other Marketable Securities (other than Lebanese Treasury Securities), (4) Deposits with banks and other financial institutions, (5) Deposits with
Affiliates, (6) Bank Acceptances, (7) Investments and Loans to Related Parties, (8) Investments in Related Parties, (9) Tangible Fixed Assets, (10)
Intangible Fixed Assets, (11) Other Assets (12) Regularization Accounts and Miscellaneous Debto
r
Accounts, (13) Revaluation of Other Fixed Assets,
(14) Goodwill, and (15) Investments in Foreign Currencies T-bills.
3
Off-balance sheets activities for Lebanese banks include guarantees and standby letters of credits (the largest of the items), documentary and
commercial letters of credit, fiduciary deposits, and interes
t
-rate swaps.
270 Peters, Raad, and Sinkey
We calculate “other assets” as a residual item, which equals total assets net of
loans (in LP and USD) and treasury bills (in LP and USD). As defined, other assets
include 15 items, none of which dominate the category. Table 2 provides a list of the
components of this asset class. Over our sample period, the ratio of other assets to total
assets ranged from a high of 48 percent in 1993 to 37.4 percent in 1999. At the end of
our sample period, it was 48 percent.
Off-balance sheets activities (OBSAs), which for Lebanese banks include
guarantees and standby letters of credits (the largest of the items), documentary and
commercial letters of credit, fiduciary deposits, and interest-rate swaps, have not kept
pace with the growth of total assets. As a percent of total assets, OBSAs have varied
between 10 and 12 percent over the period.
Lebanese banks have two major sources of funds: deposits denominated in
Lebanese pounds (LP) and deposits denominated in US dollars (USD). Table 2 presents
the average proportion of total assets financed by LP and USD. The data show that the
proportion of assets financed by foreign currency deposits has decreased since 1993.
The proportion financed by Lebanese pound deposits increased between 1993 and
2000. The data also show that liability-management practices differ among banks.
B. ROE and ROA Profit Measures
Table 3 presents ROE and ROA for our sample banks. Both measures were computed
using average shareholders’ equity and average total assets. For each of our eight
sample years, the means and standard deviations are shown. Since ROE = ROA X EM,
ROE mixes profitability and leverage. Thus, ROA is the preferred accounting measure
of a firm’s profitability from the perspective of asset efficiency. The difference between
ROE and ROA arises from the leverage factor, EM, which can be derived as
ROE/ROA. Using a benchmark ROA of one percent as a minimum performance
standard,
14
our sample banks pass this test, except in 1993 (ROA = 0.64%), 1999 (ROA
= 0.70%), and 2000 (ROA = 0.59%). Since 1996, ROA has been declining, dropping
about 55 percent from 1.3 percent in 1996 to 0.59 percent in 2000. The high ROEs in
1993 (51%) and 1994 (52%) were driven by high degrees of financial leverage as EM
was 79.69 in 1993 and declined to 47.27 in 1994. The reduced leverage
15
was more
than offset, however, by the jump in ROA from 0.64 percent to 1.1 percent, hence ROE
increased.
INTERNATIONAL JOURNAL OF BUSINESS, 9(3), 2004 271
Table 3 Profitability and Risk Measures including Return on Average Equity (ROE), Return on Average Assets (ROA), Net Interest Margin (NIM), Risk
Index [RI, Blasko & Sinkey], [RI, Eisenbeis & Kwast], and the probability of book-value insolvency (Probability).
1993 1994 1995 1996 1997 1998 1999 2000
Variable Mean Stand Mean Stand Mean Stand Mean Stand Mean Stand Mean Stand Mean Stand Mean Stand
N=65 Dev N=65 Dev N=66 Dev N=64 Dev N=62 Dev N=59 Dev N=59 Dev N=54 Dev
ROE
1
0.51 1.01 0.52 0.95 0.23 0.27 0.20 0.21 0.153 0.18 0.143 0.14 0.077 0.13 0.046 0.27
ROA
2
0.0064 0.018 0.011 0.021 0.010 0.012 0.013 0.016 0.011 0.018 0.012 0.012 0.007 0.011 0.0059 0.011
N
IM
3
0.041 0.019 0.039 0.017 0.042 0.017 0.051 0.076 0.037 0.021 0.032 0.015 0.028 0.014 0.028 0.011
Coefficien
t
2.81 NA 1.91 NA 1.20 NA 1.23 NA 1.64 NA 1.00 NA 1.57 NA 1.86 NA
of Variation
4
RI (Blasko & 8.44 9.23 8.29 6.97 14.43 13.05 14.37 13.29 17.43 15.00 15.72 12417 15.48 12.08 15.70 1211
Sinkey)
5
Probability
7
0.044 0.1 0.044 0.105 0.032 0.018 0.01 0.02 0.013 0.044 0.0093 0.023 0.012 0.032 0.019 0.073
RI (Eisenbeis & 2.36 NA 2.52 NA 8.67 NA 6.19 NA 7.17 NA 10.58 NA 9.63 NA 9.08 NA
Kwast)
6
Probability
7
0.09 NA 0.079 NA 0.007 NA 0.013 NA 0.010 NA 0.0044 NA 0.0054 NA 0.0064 NA
N is the sample size.
1
Three observations in 1993 and one other observation in 1994 are eliminated when we calculated ROE because in each of those four cases the bank equity is
negative.
2
ROE and ROA are equally weighted Return on Average Equity and Return on Average Assets, respectively. The weighted ROEs (total profits / total equities of
all banks in the sample) from 1993 through 2000 are respectively, 45%, 42%, 16.42%, 20%, 18.9%, 17.4%, 14.5%, and 12.8%. The weighted ROAs (total
p
rofits / total assets of all banks in the sample) from 1993 through 2000 are respectively, 1.0%, 1.12%, 1.0%, 1.20%, 1.32%, 1.20%, 1.0%, and 0.86%.
3
NIM is equal to (Interest and similar income - Interest and similar charges) divided by average assets.
4
Coefficient of Variation is a relative measure of variation and is equal to standard deviation of ROA / mean of ROA.
5
RI (Blasko & Sinkey) is a risk or safety index used by Blasko & Sinkey, which is equal to (average of ROA + Equity/Assets) / Standard Deviation of ROA] for
each bank. RI is calculated for each bank and then averaged for each year.
6
RI (Eisenbeis & Kwast) is an alternative risk or safety index following Eisenbeis & Kwast who used cross-sectional components of ROA, Equity/Assets), and
standard deviation for each year, which gives one RI for each year. Safe or low-risk banks have high RIs while high-risk banks have low RIs.
7
Probabilitiy is equal to 1/[2(RI)
2
].
272 Peters, Raad, and Sinkey
Table 4
The return-on-equity model based on pooled data for 52 control banks across five
middle east countries: United Arab Emirates (18 banks), Saudi Arabia (10), Kuwait (8),
Bahrain (9), and Oman (7)
Year Return on equity
(ROE)
Return on assets
(ROA)
Equity multiplier
(EM)
1995 0.1218 0.0163 7.47
1996 0.1348 0.0190 7.08
1997 0.1498 0.0213 7.05
1998 0.1284 0.0182 7.04
1999 0.1232 0.0181 6.80
Grand mean 0.1316 0.0186 7.09
Standard deviation 0.0114 0.0018 0.24
Memo: Grand risk index (Eisenbeis and Kwast) = 88.8 [= (0.0181 + 1/7.09)/0.0018] with a probability of
book-value insolvency of 0.01 percent. Appendix A shows ROE and ROA by country for the banks that
comprise our control group.
Source: Various Publications and Databases of the Research Unit of the Institute of Banking Studies (IBS),
Kuwait. This source contains the following disclaimer: “Please note that, information given here was
prepared by the Research Unit of the IBS from various sources. Every effort has been taken to collect,
classify, analyze and present the information given here, as accurately as possible. However, the Institute or
its Researchers would not be responsible for any human or mechanical errors, if any in this respect.”
Table 4 presents ROE, ROA, and EM for our control group of 52 banks. These
aggregate data (see Appendix A for the data by country) reveal that the banks in these
five countries have greater asset efficiency (the grand mean for ROA is 1.86%) but
employ less leverage (EM is only 7.1 on average). Using 1999 to illustrate, the
following ROE profiles are insightful:
Lebanese banks: ROE = ROA x EM = 0.0770 = 0.0070 x 11.0
Control group: ROE = ROA x EM = 0.1232 = 0.0181 x 6.8
In terms of risk exposure, measured by the variability of ROA, our control banks
are much safer as the standard deviation of ROA (1995-1999) was only 0.18%
compared to 1.4% for our sample banks over the same period. On balance, while
Lebanese banks are profitable in their own right, they pale in comparison to our control
banks. We suspect a major reason for this difference is the fact that the majority of our
control banks are located in countries with steady oil revenues while Lebanon is an oil-
dependent country.
C. Net Interest Margin
Since traditional banking is the business of funding loans with deposits, a bank’s net
interest margin [NIM = (interest income – interest expense)/average assets] is a key
performance measure that drives ROA. Table 3 presents NIMs for our sample banks.
They averaged about 4 percent for the years 1993 to 1995. They peaked in 1996
INTERNATIONAL JOURNAL OF BUSINESS, 9(3), 2004 273
(NIM = 5.1 percent), but since then they have declined steadily, dropping to 2.8 percent
in 2000. The culprit in the declining NIM has been interest expense per dollar of assets,
which rose from 6.0 percent in 1993 to 7.1 percent in 2000, reaching a peak at 7.7
percent in 1996. Of course, as NIM goes, so goes ROA, which, all other things being
equal, accounts for the declining ROA. To maintain ROA in the face of declining
interest-rate spreads, banks must increase noninterest income or improve operating
efficiency or both.
D. Efficiency Measures
Table 5 presents several measures of operating efficiency for Lebanese banks. To
cushion the effect of declining NIM on ROA, Lebanese banks have been reducing their
operating expense per dollar of assets, which declined from 4.1 percent (1993) and 4.2
percent (1994) to 2.52 percent (2000). An alternative measure of operating efficiency,
the ratio of total cost to income, has shown similar improvement dropping from 83
percent in 1993 to 71 percent in 1998 but then increased in 1999 (82 percent) and 2000
(84 percent).
Except for the latest Internet-only banks (none of which exist in Lebanon as of
this writing), banks have labor-intensive operations. Therefore, it is important to
analyze the “staff expenses” of Lebanese banks. Table 5 shows that the ratio of staff
expenses relative to total operating expenses had a shallow inverted U-shape over our
test period, rising from 52 percent (1993) to a peak of 60 percent (1997) before
dropping to 56 percent (1999). However, in 2000 this ratio rose again to 62 percent.
Over this same period, staff expenses per employee more than doubled rising from
LP16.47 million (1993) to LP39.52 million (2000), or in USD from roughly $9,600 to
$26,100.
E. Loan Quality
Since information on loan quality is scarce in Lebanon, we have little to report on this
topic. The only measure we have is the “provision for loan loss” (PLL), which in
Lebanon is not the noncash outlay but the actual reserve account, known in the US as
the loan-loss reserve or allowance for loan and lease loss. As a percent of total loans,
the loan-loss reserve was 17.4 percent in 1994 but dropped to 14 percent in 1998 and
rose again to 16 percent in 2000. By US standards, where two percent is a benchmark,
the ratio is high. However, the relatively low level of lending by Lebanese banks
accounts for this substantial difference. On balance, Lebanese banks do not make many
loans but they keep a high level of reserves relative to the loans that they do make.
Perhaps, the political uncertainty of the country leads to this conservative behavior.
274 Peters, Raad, and Sinkey
Table 5 Efficiency measures: statistical description of selected expense ratios.
1993 1994 1995 1996 1997 1998 1999 2000
Variable Mean Stand Mean Stand Mean Stand Mean Stand Mean Stand Mean Stand Mean Stand Mean Stand
N=62 Dev N=63 Dev N=64 Dev N=62 Dev N=61 Dev N=58 Dev N=58 Dev N=53 Dev
General Ope
r
ating 0.041 0.026 0.042 0.025 0.039 0.02 0.036 0.016 0.031 0.013 0.027 0.013 0.026 0.013 0.022 0.010
Expenses / Total
Assets
Cost / Income
1
0.83 0.18 0.73 0.28 0.75 0.26 0.75 0.27
0.71 0.24 0.71 0.27 0.82 0.38 0.84 0.51
Staff Expenses / 0.52 0.11 0.57 0.098 0.58 0.10 0.60 0.11
0.60 0.093 0.58 0.11 0.56 0.093 0.62 0.10
Operating Expenses
Staff Expenses / 16.47 6.06 23.39 13.06 26.85 8.71 32.46 11.06 33.77 12.78 36.93 13.93 38.04 12.82 39.52 12.55
N
umber of
Employees
Interest Expenses / 0.06 0.025 0.070 0.028 0.074 0.023 0.077 0.021 0.072 0.019 0.066 0.022 0.068 0.02 0.071 0.024
Total Assets
Loan Loss NA NA 0.174 0.17 0.15 0.15 0.15 0.14 0.14 0.14 0.14 0.13 0.14 0.12 0.16 0.13
Provisions /
Total Loans
N
is the sample size.
1
Cost = (general operating expenses + allocation to provision & depreciation of fixed assets + net allocation to provisions on financial fixed assets) and
Income = net financial income.
INTERNATIONAL JOURNAL OF BUSINESS, 9(3), 2004 275
F. Risk Measures
Table 3 also shows the risk index (RI) we calculated.
16
The probability of book-value
insolvency (p) equals 1/[2(RI)
2
]. The average RI for a censored sample of banks, which
excludes low outliers, increased steadily from 8.44 (p = 4%) in 1993 to 15.7 (p = 1.9%)
in 2000. On balance, the conclusion is that, on average, the safety of banks in Lebanon
increased and their associated probability of book-value insolvency declined over the
years 1993 to 2000. In contrast, our control banks (see Table 4) have a risk index of
88.8 (p = 0.01%). The higher ROA and lower variability of ROA for the control banks
account for the substantial difference between the risk indices. Thus, on a cross-country
basis Lebanese banks do not look as strong as our control banks; on average, however,
the trend in their financial strength has been favorable, although slipping a bit recently.
Another measure of risk we employ is the variability of ROA as measured by its
standard deviation. Table 3 shows that this cross-sectional sigma has ranged from 2.1
percent (1994) to 1.1 percent (2000), another indicator of the reduced riskiness of
Lebanese banks. This favorable trend, however, can be put in perspective by noting that
our control banks (Table 4) show much less variability of ROA (0.18%). The
coefficient of variation (standard deviation of ROA/mean ROA) is a relative measure of
dispersion. Over our sample period, it has declined from 2.81 in 1993 to 1 in 1998, and
then has risen to 1.86 in 2000. Although we do not have a year-by-year estimate for the
coefficient of variation for our control group, we note that it was 0.1 for the period 1995
to 1999.
To recap, all three measures of risk, the risk index, the standard deviation of
ROA, and the coefficient of variation of ROA, indicate that Lebanese banks became
safer in 2000 than in 1993 and 1994 but this safety has started to decline since 1997
when it reached its peak. Compared to our control banks, however, Lebanese banks are
not as profitable or as safe.
G. Differences due to Size
The average Lebanese bank, as measured by total assets, nearly quadrupled its size
between 1993 and 2000; the distribution of total asset size, however, is highly skewed
as the standard deviation of asset size is larger than the mean average for every one of
the six years.
17
To test for the effects of size on the performance and characteristics of
Lebanese banks, we split our sample at the median asset size and examine the groups
above and below the median. We find several interesting results.
18
First, with respect to
profitability, no significant differences exist for ROA. However, the ROE of the larger
banks is significantly higher in each of the years because they employ greater leverage
than smaller banks. In contrast, the group of larger banks has, on average, significantly
lower NIM than smaller banks in seven of the eight years. This profile of larger banks
having lower capital ratios and lower NIMs is similar to the one observed for US banks
(Basset and Zakrajsek, 2000). The risk indices across the two groups are not
significantly different as the group of larger banks has, on average, ROAs that are more
stable to offset its lower ratio of capital to assets.
276 Peters, Raad, and Sinkey
The similar ROAs between the two size groups come about differently. The
group of larger banks has significantly higher staff expenses per employee but
significantly lower operating expenses per dollar of assets. However, the bigger banks
have higher interest expenses per dollar of assets across all eight years.
Behind the cash-flow differences described above, the following portfolio
differences exist between large and small banks. First, the bigger banks have
significantly higher USD deposits as a percent of total assets for all eight years, e.g., 54
percent vs. 45 percent at year-end 2000. This difference probably accounts for the
higher interest expenses per dollar of assets for the bigger banks. For the loan portfolio,
the average larger bank makes more USD loans as a percent of assets (significantly so
for the years 1995 and 1996) but significantly less LP loans as a percent of assets for
the entire period.
H. Regression Models and Estimates
We establish and test possible explanations for variations in bank profitability in this
section. Using our pooled cross-sectional, time-series sample of banks in Lebanon from
1993 to 2000, we estimate regression equations for ROE and ROA models.
19
The total
number of banks included in the sample is 484. Table 6 presents the Pearson correlation
matrix for the variables used in the regression analysis and a discussion of the variance-
inflation factor (VIF) tests used to detect multicollinearity among our regressors.
Our first regression equation attempts to explain the variation in ROE:
ROE
it
= B
0
+ B
1
(SPREAD)
t
+ B
2
(GDPGROWTH-1) + B
3
(TBILLS/EQUITY)
it
+ B
4
(FCDE/EQUITY)
it
+ B
5
(LNASSETS)
it
+ B
6
(FOREIGN)
it
+ B
7
(EMPLOYEE/EQUITY)
it
+ B
8
(OFFBALANCE/EQUITY)
it
+ ε
it
(2)
where ROE
it
is the return on average equity for bank i in year t; SPREAD
t
is the
difference in the average month-end interest rate paid on T-bills denominated in
Lebanese pounds and the average month-end interest rate paid on deposits in Lebanese
pounds in year t; GDPGROWTH-1 is a lagged independent variable which represents
the real growth rate in gross domestic product in year t-1; ASSETS
it
is the average total
assets for bank i in year t; EQUITY
it
is the average total equity for bank i in year t;
TBILLS
it
is the average investment in treasury bills for bank i in year t; FCDE
it
is the
average amount of foreign currency deposits for bank i in year t; LNASSETS
it
measures the natural log of average assets; FOREIGN
it
is a dummy variable with a
value of 1 when a bank i is at least 50 percent owned by a foreign bank in year t;
EMPLOYEE
it
is the number of employees for bank i in year t; OFFBALANCE
it
is the
average balance sheet assets for bank i in year t; and ε
it
is the error term.
INTERNATIONAL JOURNAL OF BUSINESS, 9(3), 2004 277
Table 6. Pearson Correlation Matrix. Coefficients of Correlation among the independent variables used in the regression equation when ROA is used as the
dependent variable. The spread between rates on t-bills and rates on deposits in Lebanese pounds (SPREAD); the annual growth in real GDP in the previous
year t-1 (GDPGROWTH-1); Lebanese pound t-bills to total assets (TBILLS/ASSETS );Foreign Currencies Deposits to Total Assets (FCDE/ASSETS); the natural
logarithm of total assets (LNASSETS); a dummy variable (FOREIGN) which is equal to zero if the bank is domestically owned and one if it is an affiliate of a
foreign bank, the ratio of the number of employees to assets (EMPLOYEE /ASSETS), and the ratio of off balance sheet assets to total assets (OFFBALANCE /
ASSETS).
VARIABLE SPREAD GDPGROWTH-1 TBILLS/ FCDE / LNASSETS FOREIGN EMPLOYEE / OFFBALANCE /
ASSETS ASSETS ASSETS ASSETS
SPREAD 1.0
GDPGROWTH-1 0.57
b
1.0
TBILLS / ASSETS -0.07 0.08
1.0
FCDE / ASSETS 0.14
b
-0.07
-0.34
b
1.0
LNASSETS -0.32
b
-0.26
b
0.24
b
0.21
b
1.0
FOREIGN -0.05 -0.05 -0.34
b
0.29
b
0.06 1.0
EMPLOYEE / ASSETS 0.43
b
0.23
b
-0.034 -0.21
b
-0.57
b
-0.18
b
1.0
OFFBALANCE / ASSETS 0.013 0.065 -0.20
b
0.11
b
0.097
0.234
b
-0.153
b
1.0
a
Correlation is significant at the 0.05 level (2-tailed).
b
Correlation is significant at the 0.01 level (2-tailed).
Notes: Since the existence of high pairwise correlations among the independent variables in regression models creates multicollinearity problems, we used the
variance-inflation factor (VIF) to test for this adverse effect. VIFs, which measure the effect of multicollinearity on the variances of the regression coefficient
estimates, are calculated for each independent variable as VIF = 1/(1-R
2
), where R
2
is the coefficient of determination obtained when each of the independent
variables is regressed on the remaining independent variables. A large VIF (> 5) caused by a large R
2
indicates multicollinearity. Our results show that none of
the VIFs is high enough to cause any concern about multicollinearity.
278 Peters, Raad, and Sinkey
Table 7
Results for the following OLS regression:
Performance
it
= B
0
+ B
1
(X1)
t
+ B
2
(X2)
it
+ B
3
(X3)
it
+ B
4
(X4)
it
+ B
5
(X5)
it
+ B
6
(X6)
it
+
B
7
(X7)
it
+ B
8
(X8)
it
+ ε
it
1
ROE
it
= the return on average shareholders equity in percent for bank i in year t; ROA
it
= the return on
average total assets in percent for bank i in year t; SPREAD = Spread between rates on t-bills and rates on
deposits in Lebanese pounds; GDPGROWTH-1 = annual growth in real GDP in the previous year t-1; TBILLS
it
= the amount invested in treasury bills for bank i in year t ;FCDE
it
= the amount of foreign currency deposits
for bank i in year t; ASSETS
it
= the average total assets of bank i in year t; FOREIGN
it
is a dummy variable
with a value of 1 when a bank i is at least 50% owned by a foreign bank in year t and zero otherwise;
EMPLOYEE
it
= the number of employees for bank i in year t; OFFBALANCE
it
= off balance sheet assets for
bank i in year t; EQUITY
it
= the average shareholders' equity for bank i in year t; LNASSETS
it
= the natural
logarithm of the average total assets of bank i in year t;
Independent
Variables
Dependent
Variable
ROE
t-statistic
Independent
Variables
Dependent
Variable
ROA
t-statistic
Constant 14.61 0.77
Constant 1.57 1.99
c
SPREAD 2.18 2.15
c
SPREAD 0.03 0.88
GDPGROWTH-1 115.55 1.14 GDPGROWTH-1 8.60 2.25
c
TBILLS/EQUITY 2.02 5.25
a
TBILLS/
ASSETS
2.51 4.458
a
FCDE/EQUITY 1.40 10.53
a
FCDE/ASSETS 0.67 1.64
d
LNASSETS -2.07 -1.48
LNASSETS -0.136 -2.34
b
FOREIGN 2.45 0.58 FOREIGN 0.34 2.02
c
EMPLOYEE/EQUITY -831.91 -10.04
a
EMPLOYEE/
ASSETS
-722.62 -6.932
a
OFFBALANCE
/ EQUITY
0.87 1.66
d
OFFBALANCE/
ASSETS
-0.079 -0.513
Model adjusted R
2
0.463 Model adjusted R
2
0.153
Model F-Statistic 52.51 Model F-Statistic 12.1
Model significance
level
0.000 Model significance
level
0.000
Sample size = 484
a
significant at the 0.01 level
b
significant at the 0.02 level
c
significant at the 0.05 level
d
significant at the 0.10 level
We use SPREAD
t
and GDPGROWTH-1 to test for the respective effects of
interest margin and growth in the economy on bank profitability. We expect both of
these variables to be positively related to bank profit. Banks that hold relatively more
TBILLS may not be as profitable as banks that invest more heavily in loans or other
assets. But in Lebanon, T-bills pay high interest rates compared to those in developed
countries. We use FCDE
to test for the effects of foreign-currency deposits on bank
INTERNATIONAL JOURNAL OF BUSINESS, 9(3), 2004 279
profitability. To control for foreign affiliation we use the binary variable, FOREIGN.
EMPLOYEE is used to test for labor efficiency while OFFBALANCE is used to test
for the effect of investing in off-balance sheets assets on profitability. We employ
LNASSETS as a control for bank size.
We also use ROA as a dependent variable. In this regression, average total assets
replaces average total equity to scale the independent variables derived from the
balance sheet.
Table 7 presents the results for estimating our two regression equations. The first
regression equation attempts to explain the variation in return on average total equity,
ROE. With the balance-sheet variables standardized by average total equity capital, this
equation explains 46.3 percent of the variation in ROE. Five of the eight independent
variables have significant effects on ROE. Four of the five variables, SPREAD,
TBILLS/EQUITY, FCDE/EQUITY AND OFFBALANCE/EQUITY have significant
positive associations with ROE. In contrast and as expected, banks that have more
employees (per dollar of equity) have lower ROEs. The coefficients of GDPGROWTH-
1 and FOREIGN are not statistically significant in determining ROE.
20
The control for
bank size (LNASSETS), which has some marginal significance (t = -1.48), suggests
that bigger banks have lower ROEs.
Our second regression equation attempts to explain the variation in return on
average total assets, ROA. Although this equation has lower overall explanatory power
than the ROE model (15.3% vs. 46.3%, see Table 7), the expected signs and statistical
significance of the regressors are similar across the two equations. Two differences,
however, are worth mentioning: (1) the reduced significance of FCDE in the ROA
model, where the t-statistic drops from 10.5 to 1.64, and (2) the stronger negative
association between bank size and profitability (t = -2.34). When we test for the
separate effects of SPREAD and GDPGROWTH-1, as described in footnote 20, we
find similar results to those reported in the experiments described there.
The relatively low R
2
for the ROA regression based on OLS and the nature of
our time-series data suggest that unique differences and cross-sectional variation
between banks may play an important role in explaining variations in ROA among
banks. To account for this possibility, we estimate a model suggested by Greene (1990).
It assumes that differences across banks can be captured by differences in the constant
term by using a “fixed-effects model”. This estimation procedure includes a dummy
variable for each bank while dropping the traditional constant term from the equation.
For this version of the ROA model, the dummy variable for affiliation with a foreign
bank was not included. The regression equation is:
ROA
it
= B
1
(SPREAD)
t
+ B
2
(GDPGROWTH-1)
t
+ B
3
(TBILLS/ASSETS)
it
+B
4
(FCDE/ASSETS)
it
+ B
5
(LNASSETS)
it
+ B
6
(FOREIGN)
it
(3)
+B
7
(EMPLOYEE/ASSETS)
it
+ B
8
(OFFBALANCE/ASSETS)
it
+ COEFFICIENTS AND DUMMY VARIABLES FOR EACH BANK + ε
it
280 Peters, Raad, and Sinkey
Table 8
Fixed-effects model for ROA
MODEL*
ROA
it
= B
1
(SPREAD)
t
+ B
2
(GDPGROWTH-1)
t
+ B
3
(TBILLS/ASSETS)
it
+ B
4
(FCDE/ASSETS)
it
+ B
5
(LNASSETS)
it
+
B
6
(EMPLOYEE/ASSETS)
it
+ B
7
OFFBALANCE/ASSETS)
it
+ COEFFICIENTS AND DUMMY VARIABLES FOR
EACH BANK + ε
it
ROA
it
= − 0.018 (SPREAD)
t
+ 10. 53(GDPGROWTH-1)
t
+ 1. 33(TBILLS/ASSETS)
it
+ 0.018 (FCDE/ASSETS)
it
(−0.50) (3.06)
a
(1.73)
b
(0. 32)
+ 0.05(LNASSETS)
it
− 129.89 (EMPLOYEE/ASSETS)
it
+ 0 . 10 (OFFBALANCE/ASSETS)
it
(0 .63) ( −0.97) (0.67)
Sample size = 484 pooled, cross-sectional observations, 1993-2000
Model Adjusted R
2
= 0.57
Model F-Statistic = 9.4
Model Significance Level = 0.0001
Figures in parentheses are t-statistics.
a
Significant at 0.01 level
b
Significant at 0.10 level
*
ROA
it
= the return on average total assets in percent for bank i in year t; SPREAD
t
= Spread Between Rates
on Treasury Bills and Rates on Deposits in Lebanese Pounds in year t; GDPGROWTH-1 = annual growth in
real GDP in the previous year t-1; (TBILLS/ASSETS)
it
= the amount invested in treasury bills divided by
average total assets for bank i in year t; (FCDE/ASSETS)
it
= the average amount of foreign currency deposits
divided by average total assets for bank i in year t; LNASSETS
it
= the natural logarithm of the average total
assets of bank i in year t; (EMPLOYEE/ASSETS)
it
= the ratio of the number of employees to average total
assets in millions of Lebanese pounds for bank i in year t; (OFFBALANCE/ASSETS)
it
= the ratio of off
balance sheet assets to average total assets for bank i in year t.
Table 8 shows the estimates for the fixed-effects model, which explains 57
percent of the variation in ROA. Although the estimated coefficients for the 70 banks
dummy variables are not shown (for practical reasons), the important point is whether
they differ across banks.
21
The null hypothesis is that the coefficients of the dummy
variables are zero, which is rejected at the one-percent level of significance. Aside from
the significant effects of the bank dummy variables, GDP growth (1% level of
significance) and t-bills per dollar of assets (10% level of significance) are the only
other significant variables in the fixed-effects model. On balance, the results suggest
that cross-sectional variation among banks play a major role in explaining ROA.
I. Discussion
Given the statistical significance of our univariate findings and regression estimates,
what can we say about our results from a financial/economic perspective? First, banks
in Lebanon are profitable, although not as profitable as banks in other Arab Gulf
countries, and appear to be much safer than they were at the start of our sample period,
INTERNATIONAL JOURNAL OF BUSINESS, 9(3), 2004 281
but not as safe as our control banks that benefit from petrodollar economies. In general,
the cessation of war, reduced inflation, and the worldwide focus on bank capital
adequacy are clear reasons for an improved banking environment in Lebanon.
Specifically, our models show a strong association between economic growth (or
spread management) and bank profitability, whether measured by ROE or ROA. From
a bank financial-management perspective, it seems that investing in Lebanese t-bills has
a positive association with bank profitability. Rather than gather deposits and mainly
make (nongovernment) loans, the traditional banking function, Lebanese banks gather
deposits but come up a bit short on the lending side as they invest heavily in t-bills.
This portfolio strategy is not irrational, of course, as Lebanese t-bills offer high returns.
It comes at a potentially high social cost, however, as loans to the private sector should
provide a greater stimulus for economic growth. The basic function of financial
intermediaries − to channel savings into productive investments − is being
circumvented in Lebanon. If the low loan-to-asset ratios found in Lebanese banks can
be increased through productive lending, this improvement would help jump-start the
sluggish Lebanese economy.
VI. SUMMARY AND CONCLUSIONS
At the end of the civil war in Lebanon (1990), Charbaji, Mikdashi, and Chebaro (1994)
describe Lebanese banks as having “suffered a severe decline in activity and
profitability and dissolution of financial capital” (p. 86). Since then, we find bank
profitability and capital are quite strong in Lebanon but not as strong as a control group
of banks from five other Arab Gulf countries. However, in terms of better performing
the traditional bank intermediation function, Lebanese banks have considerable room
for improvement. This finding provides our main policy implication for the banking
sector in Lebanon: Quite simply, do what banks are supposed to do and make more
loans, which will stimulate the economy and promote economic growth. A related
policy recommendation is for Banque du Liban, the central bank, to discontinue
offering high returns on t-bills, which provides a major disincentive for banks not to
engage in lending to the private sector. On balance, if banks in emerging markets do not
do their jobs, they restrict the ability of their countries to grow and their markets to
develop more fully.
NOTES
1. This section draws on Charbagi (2001), Shehadi and Schneider-Sickert (1998), and
Schneider-Sickert and Iskandar (1998).
2. During July 1997, the US lifted its travel ban on its citizens traveling to Lebanon.
Prior to that de facto entry was achieved by entering Lebanon with a visa separate
from one’s passport. The events of September 11, 2001 have renewed concerns
about safe travel around the world.
3. The total value of tourism was estimated to be about the equivalent of US$600
million for 1993. By year-end 1998, however, the figure had more than doubled to
282 Peters, Raad, and Sinkey
US$1.3 billion. From 1993 through 1998, the number of visitors to Lebanon and
tourism revenues have grown at annual rates of 24% and 21%, respectively. See
Ladki, et al. [2001].
4. About 12 million people of Lebanese origin live elsewhere in the world, most of
them in economies more prosperous than Lebanon. The Ministry of Finance
estimated the amount of remittances at the equivalent of US$939 million in 1996,
but this figure is quite unreliable. Errors and omissions in the Ministry of Finance
data amounted to over US$4 billion in their reconciliation of the balance of
payments.
5. Approximately 90 percent of loans made by Lebanese banks are denominated in
US dollars. Lebanese banks tend to follow a practice of using the money they
receive in Lebanese pound deposits to invest in Lebanese treasury bills, and using
the money they receive in US dollar deposits to make US dollar loans.
6. The average interest rate on US dollar denominated loans and deposits were not
available for 1993 and 1994.
7. One aspect of capital-structure regulation in Lebanon is that even the smallest
banks must have at least 10 billion Lebanese pounds in shareholders' equity, a
substantial constraint for small banks. However, although all banks are subject to
BIS capital-structure rules, they tend to be more binding for larger banks.
8. Bilanbanques is published annually by Bank Data Financial Services in
collaboration with the Association of Banks in Lebanon and is sponsored by
Banque du Credit Libanais. Each issue of Bilanbanques reports data for the two
years immediately preceding the issuance year.
9. As of 2001, the Ministry of Finance requires all Lebanese banks to follow
International Accepted Accounting Standards. Before this, most banks were
preparing financial statements according to US Generally Accepted Accounting
Principles. The Banking Control Commission of the Lebanese Central Bank
supervises closely the activities and operations of Lebanese banks according to
strict rules and regulations. Lebanese banks are highly regulated. Financial
statements of Lebanese commercial banks are audited by major accounting firms
(e.g., Deloitte and Touche and Price Waterhouse Coopers).
10. The number of banks varied between 1993 and 2000 because some banks started
operations during this period while other banks disappeared because of
insolvencies and mergers.
11. In their theoretical exposition, Hannan and Hanweck used expected ROA. We
employ average ROA over a number of years as a proxy for expected ROA.
12. Although we specifically test the relationship between bank profitability and size,
we do not test for economies of scale.
13. Awkward foreclosure laws in Lebanon as well as a lack of liquidity in residential
real-estate markets also make mortgage lending an unattractive investment for
Lebanese banks.
14. This standard is based mainly on the US experience, excluding the ten largest
banks. For example, during the 1980s a bank that did one percent or better on
assets was a high-performance bank. After the early 1990s, banks that were not
INTERNATIONAL JOURNAL OF BUSINESS, 9(3), 2004 283
doing at least one percent on assets were underperforming. For example, see
Bassett and Zakrejsek [2000].
15. The simultaneous increase in ROA and decrease in ROE were due to large
increases in the ratio of shareholders equity to total assets for most banks. Banque
du Liban put pressure on banks to increase equity capital and introduced capital-
structure rules on banks during this period.
16. Our risk indices in Table 3 are calculated as “typical”, as described by Eisenbeis
and Kwast [1992], which means that they are based on cross-sectional estimates of
the components of the risk index. Since Blasko & Sinkey [2004] show that RIs
based on individual bank, time-series standard deviations of ROA produce
different results, we also employ their method of calculation. The findings,
however, are not qualitatively different in this case (Table 3). On balance, bank
safety in Lebanon has improved over our test period.
17. At the end of 2000, the largest bank was Banque du Liban et d'Outre-Mer with
total assets of 8,736 billion Lebanese pounds ($5.8 billion) while the smallest bank
was Rafidain Bank with total assets of 15 billion Lebanese pounds ($10 million).
18. A three-page table of these finding is available from the authors.
19. We acknowledge that pooling the observations across years introduces some
dependence in the error terms of the model because the same firms are represented
multiple times.
20. Although VIF estimates (see the notes to Table 6) indicate that multicollinearity is
not a problem, we still are concerned about the high pairwise correlation (0.57)
between SPREAD and GDPGROWTH-1. Therefore, we reestimated the ROE
regression without SPREAD and find that GDPGROWTH-1 has the expected
positive sign and is statistically significant (t = 2.76). When we exclude
GDPGROWH-1, the t-statistic on spread increases to 3.3 from 2.15. Thus, when
both variables are included in the ROE regression, their separate effects become
blurred. The estimated relationships of the other variables showed only minor
difference in these experiments.
21. In the F-test, the unrestricted model is the fixed-effects model while the restricted
model is a pooled model with a single constant term.
REFERENCES
Bassett, W. and Zakrejsek, E., 2000, “Profit and Balance Sheet Developments at U.S.
Commercial Banks in 1999,” Federal Reserve Bulletin, June, pp. 367-395.
Berger, A. N., 1995, "The Relationship between Capital and Earnings in Banking,"
Journal of Money, Credit and Banking, Vol. 27, No.2, May, pp. 432-456.
Bilanbanques, Association of Banks in Lebanon, various issues.
Blăsko, M. and Sinkey, J.F. Jr., 2004, “Bank Asset Structure, Real-Estate Lending, and
Risk-Taking,” Quarterly Review of Economics and Finance, (forthcoming).
Bourke, P., 1989, “Concentration and Other Determinants of Bank Profitability in
Europe, North America and Australia,” Journal of Banking and Finance, Vol. 13,
pp. 65-79.
284 Peters, Raad, and Sinkey
Charbagi, A., 2001, "Developing a Model to Restructure the Overpopulated Banking
Industry in Lebanon," Managerial Auditing Journal, Vol. 16, No. 1, pp. 28-35.
Charbaji, A., Mikdashi, T., and Chebaro, H., 1994, “Qualitative Techniques Transfer in
Lebanon: Implications for the Financial Sector,” International Journal of
Commerce and Management, Vol. 4, No. 4, pp. 85-94.
Eisenbeis, R.A. and Kwast, M.L., 1991, “Are Real Estate Depositories Viable?
Evidence from Commercial Banks,” Journal of Financial Services Research,
March,. 5-24.
Financial Markets Handbook, Banque du Liban.
Goldberg, L.G. and Rai, A., 1996, "The Structure Performance Relationship in
European Banking," Journal of Banking and Finance, Vol. 20, No. 4, May, pp.
745-771.
Greene, W. H. Econometric Analysis, New York: MacMillan, 1990.
Hannan, T.H. and Hanweck, G.A., 1988, "Bank Insolvency Risk and the Market for
Large Certificates of Deposit," Journal of Money Credit and Banking, Vol. 20, No.
2, May, pp. 203-211.
Ladki, S., Mikdashi, T., Fahed, W., and Abbas, H., 2001, “Arab Tourists and the
Lebanese Vacation Ownership Industry: A Quality of Life Perspective,”
International Journal of Hospitality Management.
Liang, J.N. and Savage, D.T., 1990, “The N onbank Activities of Bank Holding
Companies,” Federal Reserve Bulletin, May, pp. 280-292.
Monthly Interest Rates, Banque du Liban, (found on website: www.bdl.gov.lb)
Molyneux, P. and Thornton, J., 1992, "Determinants of European Bank Profitability: A
Note," Journal of Banking and Finance, Vol. 16, No. 6, December, pp. 1173-1178.
Nash, R.C. and Sinkey, J.F., Jr., 1997, “On Competition. Risk, and the Hidden Assets in
the Market for Bank Credit Cards,” Journal of Banking and Finance, 21, pp. 89-
112.
Neely, M.C. and Wheelock, D.C., 1997, "Why Does Bank Performance Vary Across
States?" Federal Reserve Bank of St. Louis Review, Vol. 79, No. 2, March-April,
pp. 27-40.
Schneider-Sickert C. and Iskandar, M., 1998, “Banking and Growth” Lebanon, the
Annual Business, Economic and Political Review, Oxford Business Group, pp. 71-
81.
Shehadi, K. and Schneider-Sickert, C., 1998, “Re-emerging Into Uncertainty”,
Lebanon, the Annual Business, Economic and Political Review, Oxford Business
Group, pp. 37-43.
Sinkey, J.F. Jr., 1998, Commercial Bank Financial Management. Upper Saddle River,
NJ: Prentice Hall, pp. 616-624.
Sinkey, J.F. Jr., Nash, R.C., 1993, “Assessing the Riskiness and Profitability of Credit-
Card Banks,” Journal of Financial Services Research, June, Volume 7, Number 2,
127-150.
INTERNATIONAL JOURNAL OF BUSINESS, 9(3), 2004 285
APPENDIX A
This appendix shows ROE, ROA, and derived relations from these data from 1995 to
1999 by country for our control group of 52 banks from five Arab Gulf Countries. The
group contains all the banks in these countries and the data are the only available
information that we could find.
Country UAE Saudi
Arabia
Kuwait Bahrain Oman
Year Performance N= 18 N= 10 N= 8 N= 9 N= 7
ROA % 1.84 1.38 1.16 1.82 1.95
1995
ROE % 11.0 14.0 8.73 8.30 18.89
ROA % 2.66 1.43 1.54 2.07 1.82
1996
ROE % 14.98 13.39 11.27 9.87 17.80
ROA % 2.78 1.41 1.79 2.56 2.09
1997
ROE % 16.14 14.13 12.63 13.65 18.52
ROA % 2.56 1.52 1.40 1.68 1.96
1998
ROE % 14.37 15.14 8.94 10.27 15.48
ROA % 2.48 1.41 1.54 1.99 1.64
1999
ROE % 13.27 13.73 9.66 11.98 12.89
ROA % 2.46 1.43 1.49 2.02 1.89
MEANS
ROE % 13.96 14.06 10.24 10.82 16.72
Standard
Deviation
ROA % 0.37 0.05 0.23 0.33 0.17
Capital
Ratio(%)
Inverse of
ROE / ROA
17.65 10.17 14.51 18.71 11.32
Risk Index Eisenbeis &
Kwast
54.90 217.30 69.50 61.80 77.60
Solvency P(BVE) < 0 % 0.0166 0.0010 0.0103 0.0131 0.0083
Notes: P(BVE) < 0 stands for the probability of book value of equity less than zero. It equals the reciprocal of
2(RI)
2
, where RI = risk index. See Table 3 for additional notes.
Source: See Table 4.
286 Peters, Raad, and Sinkey
