Differences between
the financial
characteristics of
interest-free banks and
conventional banks
M.M. Metwally
The author
M.M. Metwally is a Professor in the Department of
Economics, University of Wollongong, Wollongong,
Australia.
Abstract
Uses logit, probit and discriminant analysis to test for
structural differences between the financial characteristics
of interest-free banks and conventional banks. The
analysis extends to various financial dimensions which
evaluate performance, namely: liquidity, leverage, credit
risk, profitability and efficiency. Covers 15 interest-free
banks and 15 conventional banks. The statistical evidence
suggests that the two groups of banks may be differentiated in terms of liquidity, leverage and credit risk, but not in
terms of profitability and efficiency.
92
M.M. Metwally
Interest-free banks were established to conform with Islamic law which prohibits interest
on all types of loans (personal, commercial,
agricultural, industrial, etc.) whether these
loans are made to friends, private or public
companies, government or any identity.
Those opposed to interest have recognized
that modern banks perform a useful function
which cannot simply be ignored. The supporters of interest-free banking believe that
the best solution is to permit banks to function but to specify a code under which they
should avoid the payment and receipt of
interest. Such a code already exists and is
being refined and modified constantly to meet
the rapidly changing needs of ever more
sophisticated businesses. The essential principle of interest-free banking is profit/loss sharing. By this, it means both the supplier for the
capital and the borrower share the risks; both
prosper when returns are favourable and
suffer together when returns are poor. This is
the basis for what became known as interestfree banks.
Currently, there are more than 100 interest-free banks active in 45 different countries.
The Al Baraka group, which began at the start
of the 1980s, is the fastest growing interestfree banking group with branches in the
Middle East, Africa and Europe. The Al Rajhi
group, which was developed by a group of
Saudi money changers in 1978, is also
expanding into the UK. It, together with Al
Baraka group and the Kuwait Financial
House, established the Islamic Banking System International which has founded a separate Islamic bank in Denmark (IBID) the
first such European-based bank designed to
encourage trade and development participation between Scandinavia and the Middle
East.
Very few Western business people deal with
interest-free banks, however, as they are
regarded as a strange kind of institution which
functions in unusual ways which few can
understand, although this is now beginning to
change as such institutions penetrate European financial services and set up branches
there (Wilson, 1985). Normally, however, the
first indirect contact most Western business
people have with these banks is when
importers in the Islamic world arrange their
own credit through Islamic banks. It is clearly
important for the Western exporter to have
M.M. Metwally
It
1
2
D = discriminant score
bs = discriminant coefficients or weights
Xs= prediction or independent variables.
The coefficients b are calculated to maximize
, by solving:
b Bb
max =
b Wb
exp (t 2 / 2)dt
At maximum, we have:
( B W ) b = 0
I j = 0 + 1 X ij + 2 X2 j + + n X nj
where Xij is the value of the ith independent
variable related to the jth bank. We let Y = 1,
if the bank is interest-free and Y = 0 if it is a
traditional bank. It is assumed that, for each
bank, there is a critical or threshold level of
the index, say I*j, such that if Ij exceeds I*j the
bank will become interest-free; otherwise it
will not.
In the logit model, an equivalent index can
be defined but the logistic function is used to
model the dependent variable (see Judge
et al., 1988; Kramer, 1991; Maddala,
1983):
1
Pt = F ( I t ) = F ( X t ) =
(1 + exp ( X t ))
The data
Almost every Middle East country and a
number of African and Asian countries run
what becomes known as a dual banking
system, where interest-free banks operate
side-by-side with conventional banks. However, the banks of some Muslim countries,
namely Iran, Pakistan and the Sudan, have all
converted to interest-free financial organizations. It was possible to collect data on the
financial characteristics of 30 banks, half of
which are interest-free. Forty-two conventional banks and 19 Islamic banks operating in
various parts of the world, where a dual banking system (or a sole interest-free banking
system) operates, were requested to provide
information on their total assets, equity,
deposits, loans, gross income before tax,
operating expenses, return on deposit and
cash reserves for the three years 1992, 1993
and 1994. Replies were received from 21
conventional banks and 18 interest-free financial institutions. However, only 15 banks of
each group provided enough details for the
purpose of this study.
The data provided by the sample banks
were used to test for structural differences
between the two groups of banks in terms of
the five financial dimensions suggested by
Wood and Porter (1979), namely: liquidity,
leverage, credit risk, profitability and efficiency. Liquidity refers to the bank ability to meet
deposit withdrawals, maturing liabilities and
loan requests without delay. There are several
financial ratios which reflect a bank liquidity
{Yt 1n[ F ( X t )]
t =1
+ (1 Yt )1n[1 F ( X t )]}.
The estimates of the parameters from the two
methods (the probit and the logit) are not
directly comparable. Amemiya (1981) suggests that, if we multiply the logit estimates by
0.625, we will obtain a close approximation
between the logistic distribution and the
distribution function of the standard normal.
The discriminant analysis model involves
linear combinations of the following form
(Lachenbrach, 1975):
D = b0 + b1X1 + b2 X 2 + + bk X k
where:
94
M.M. Metwally
Statistical results
The results of both the logit and the probit
models are given in Table I. The logit coefficients are reported in their original form. The
figures in parentheses under each coefficient
represent the t values. The logit model gives
marginally better results (in terms of R2 and t
values) than the probit model. The variables
EA, CD and RL are statistically significant
and carry a positive sign. This suggests that
95
M.M. Metwally
Independent variables
Constant term
Equity capital plus reserves/total assets (EA)
Total deposits/total assets (DA)
Operating expenses/total assets (XA)
Cash/total deposits (CD)
Gross income/total assets (YA)
Average return on deposits (PD)
Loans for durables/total loans (DL)
Direct investment/total loans (RL)
Personal loans/total loans (PL)
Log of likelihood function
R-squared
Per cent correct predictions
Logit
Probit
157.4
(0.884)
234.5*
(2.098)
250.2**
(2.359)
3885.1
(0.979)
267.0**
(2.482)
4280.2
(0.932)
439.7
(0.776)
371.8
(0.789)
148.0**
(2.288)
138.9**
(2.951)
5.012
0.830
0.967
83.1
(0.774)
142.2*
(2.059)
164.6**
(2.355)
2504.3
(0.875)
168.8**
(2.249)
2643.0
(0.881)
287.6
(0.764)
234.6
(1.042)
91.3**
(2.2601)
84.8**
(2.662)
4.338
0.827
0.967
Notes:
**Significant at the 0.05 level
*Significant at the 0.10 level
DA
DA
EA
DL
PD
PL
CD
RL
XA
YA
EA
DLPD
PL
CD
RL
XA
1.00000
0.19975 1.00000
0.16359 0.06276 1.00000
0.29806 0.37231 0.01544 1.00000
0.36000 0.19168 0.11277 0.22286 1.00000
0.25852 0.40088 0.20039 0.22510 0.42794 1.00000
0.16217 0.04096 0.28644 0.09292 0.16643 0.243543 1.00000
0.09890 0.02356 0.26717 0.00439 0.37311 0.09712 0.17623
0.26147 0.12156 0.20841 0.17650 0.17532 0.10049 0.12075
YA
1.00000
0.20212
1.00000
M.M. Metwally
Table III Results of the discriminant analysis: Wilks lambda *U-statistic) and univariate F-ratio with 1 and 28 degrees of
freedom
Variable
DA
EA
DL
PD
PL
CD
RL
XA
YA
Wilks lambda
Significance
0.62961
0.69692
0.92938
0.93395
0.41917
0.48682
0.66593
0.99884
0.94330
16.4721
12.2605
2.1276
1.9803
38.7986
29.5156
14.0467
0.0324
1.6831
0.0004
0.0011
0.1558
0.1704
0.0000
0.0000
0.008
0.8585
0.2051
Percentage
of
Cumulative Canonical
Function Eigenvalue variance percentage correlation
After
function
Wilks
lambda
Chisquare
df Significance
0.274645
30.368
:
1*
2.6411
100.00
100.00
0.8517
Note:
*Marks the 1 canonical discriminant functions remaining in the analysis
Function 1
DA
EA
DL
PD
PL
CD
RL
XA
YA
0.50808
0.40649
0.17292
0.04512
0.66647
0.58971
0.14099
0.19225
0.03965
0.0004
M.M. Metwally
Group
Function 1
0
1
1.57003
1.57003
Actual
group
Group 0
Group 1
No. of
cases
15
15
References
1
14
93.3%
Note:
Per cent of grouped cases correctly classified: 93.33%
Conclusions
98