Review of Islamic Economics, Vol. 8, No. 2, 2004
the key variables for a comparative evaluation. Salama (2003) in his
comments has also raised some pertinent questions concerning the
issue.7
The above noted blemishes have led to many faux pas in the
argument. For example, Iqbal finds the overall cost∕income ratio for
Islamic banks as 55.9% during 1990-1994, falling further to 53.4%
in 1994-1998. These ratios were much lower than the benchmark
fixed at 65% for the top 1000 conventional banks, leading to the
claim of overall better performance by the Islamic banks. But the very
next moment, the author dilutes his conclusion by revealing that the
weighted average was pushed below the benchmark due to the ratio
of the largest bank in the sample being too low for special reasons (p.
378)/ He turns to the simple averages of the ratios for the sample,
and finding them higher than the benchmark in both the periods,
reverses his claim quoted earlier to “conclude that in general Islamic
banks are not working in a cost effective manner” (p. 378).
Finally, it may be noted that ratio analysis as a measure of bank
efficiency has some serious limitations. The reliance on benchmark
ratios is, for example, a ticklish issue. Fixing the benchmark could be
arbitrary and misleading, as we find it in Iqbal. Also, ratios fail to
capture the long-run dynamics of business; they aggregate many
aspects of performance relating to operations, marketing, and finance
at the expense of vital details. The conclusions drawn on their basis
are invariably taken with a grain of salt.
IV. Econometric Models: Theoretical Framework
In recent years ratio analysis has been fast losing ground to the
frontier analysis techniques for measuring the efficiency of financial
institutions.9 The vast majority of their application has focused on the
cost effects of scale and scope economies. Nevertheless, despite the
growing volume of research in the area, there is still no consensus on
the best method for measuring the efficiency of financial institutions.
Of the various approaches so far employed, the most widely used are
the econometric or Stochastic Frontier Approach (SFA) models and
the non-parametric linear programming or the Data Envelopment
Analyses (DEA).