This paper examines the productivity of the 100 larger Italians banks for the period
2001-2002. Inputs and outputs are used as nominal values (millions of euros) and as the
natural logarithms of these values. The mean error between the actually total factor
productivity and the estimated one is calculated according to both approaches. Moreover,
the weighted arithmetic mean of the Malmquist productivity index is calculated in addition
to the geometric mean. Also, the correlation coefficient and the ranking correlation
coefficient are computed to shed more lights to the relationship between bank’ size and its
performance. The empirical results revealed that the use of natural logarithms and neural
networks regression reduces the errors in the estimates. Finally, there is rather an inverse
relationship between size and productivity growth, in contrast to the literature. However,
this relationship is not statistically significant for our sample firms.