Abstract: Can financial sector firms learn to be rational in their business activities? The answer depends on the institutionally bounded constraints to learning. From an evolutionary perspective the functionality of learning to be rational creates strong incentives for such learning without, however, guaranteeing that each member of the particular economic species actually achieves increased fitness. We investigate this issue for a particular economic species, namely, commercial banks. The purpose of this study is to present two main contributions. First we review the topic of creditworthiness of financial sector firms, with emphasis on Neural-Network (NN) models. Second, we provide an empirical analysis by using an NN model with back propagation mechanism for creditworthiness decision. The data is taken from three of the South Asian countries; Bangladesh, India and Pakistan. Our empirical findings were provided for three NN architectures by applying training and testing samples constructed from data of the firms that applied for credit in regional banks of South Asian countries for the period 1999 - 2005. To study the effect of proportion between the number of firms that obtained and did not obtain credit, three proportions of the training and testing set compositions were created: A, B, C, finally, the empirical methodology used in our analysis evaluates the classification accuracy in terms of errors made by the NN.