Article Preview
TopIntroduction
The banking sector has experienced a deep and essential changes in the world such as liberalization of financial activities, increasing economy globalization and a sharp decline in the financial intermediation role of banks.
These changes have caused an increasing pressures of competitiveness. These results also lead to an increase in banks risk, and difficulty to make profits. Moreover, it calls for an increased risk of financial solvency and survival of banks. In this context, banks become increasingly concerned by controlling and analyzing both costs and revenues.
Widespread literature has been examined banks efficiency level using different methods for the mentioned purposes. Analytical methodologies of efficient frontier, such as the stochastic frontier analysis methods are sophisticated. They lead financial institutions to determine the level of efficiency and their position with regard to the competition.
The previous work applied the stochastic frontier analysis method using two ways. The first is by using cobb-Douglas cost function, and the other way is a multi-product cost function trans-logarithmic, which is adopted to compare banks efficiency in our sample.
Several studies have used the latter method (Ferrier and Lovell, 1990 Mester, 1933, Berger and Mester, 1997 Berger and Humphrey, 1997 Dietsch and Lozano, 2000). “However, their efficiency measures ignore the inclusion of certain decision variables that are dependent on the technological process of each institution, these variables can be unobservable directly, but it is possible to measure or observe their effects.”
In this group of studies of banking efficiency applied the standard approach of the stochastic frontier, we find that we face the homogeneity assumption of observations. Meanwhile, firms can use different technologies. In this case, the estimation of a common frontier for the entire sample is not correct where the estimated frontier does not probably represent the real technology, (maximum adopted). (Berger and Mester, 1997 Berger and Humphrey, 1997).
Therefore, to take possibility of heterogeneity between the observations into consideration, we propose initially latent class analysis. This mentioned method classifying observations in similar groups. Thus, estimation of technology using a sample of firms is performed in two steps. First of all, observations are classified into different groups. Furthermore, analyzes are performed for each class. Therefore, we can consider the use of a latent class model, which can model the heterogeneity between the observations through a framework of random parameters.
However, the latent class model has some weaknesses. Indeed, this model does not take into account the time variable and thus has a certain persistence in the movement of a group / system to another. In this context, a stochastic frontier model with system structural changes can be proposed, which is Markov method. This model takes into consideration simultaneous heterogeneity, temporal heterogeneity from a side, and heterogeneity between observations from another side.
The advantage of this approach is to allow a recent exploitation of data. In other words, it allows the analysis of banks efficiency by adopting a data classification method that classify data in categories that takes under consideration the possibility of movement between classes.