بهبود بهره وری از یک دورنمای چندگانه: یک برنامه کاربردی برای صنعت بانکداری ژاپنی

The current study focuses on efficiency improvement for banking systems from multiple perspectives, which have different definitions of input/output about various attributes of a banking system. In this research we utilize data envelopment analysis (DEA) and Nash bargaining game (NBG) theory to improve inefficient banks in order to: (1) Make the inefficient bank be the state of Pareto Optimality for multiple perspectives, which can avoid discontentment of some perspectives. (2) Improve a bank by changing its attributes and provide various improving schemes for decision makers. A numerical case study of Japanese banks is also given to show the results of equilibrium solution from multiple perspectives.

The concept of data envelopment analysis (DEA) was first introduced by Farrell [1], and developed by Charnes et al. [1] and [2] who bring forward an epochal CCR model to consummate DEA theory. Being a popular nonparametric methodology to evaluate and compare the efficiencies of peer entities, viz. “decision making units” (DMUs), DEA is widely used in a variety of research fields, including management and finance, as well as nonprofit organizations. The main use of a DMU is transforming inputs into outputs. DEA estimates efficiency scores of DMUs by their relative locations with respect to the efficient frontier, which is composed by all efficient DMUs in the system. In traditional DEA research, efficiency analysis is based on a single perspective, namely, a unique input/output classification scheme about the attributes of DMU. Generally the classification scheme used to determine whether an attribute should be considered as an input or an output is determined by the perspective before efficiency analysis. If the value of an attribute is considered the more the better from the perspective, it is determined as an output. On the contrary, it is considered to be an input. In the case of multiple perspectives an attribute may play different roles from different perspectives. For example, in the present study the DMU is a complicated banking system in which the same attributes may be interpreted differently based on the multiple perspectives of various stakeholders. One attribute that is considered to be an input from one perspective but may be considered to be an output from another perspective is “profitability” of a bank. Profitability of a bank is usually considered to be an input from the perspective of the customer, because higher profitability of banks is achieved at their expense in the form of higher fees and charges [3]. However, from the perspective of management, profitability may be defined as an output, since higher profitability leads to higher salary and bonuses for them. Given this, the efficiency score of DMUo estimated by the CCR model varies with different perspectives, as they would use different input/output classifications. In the case of a single perspective, the method to effectively improve DMUo (assume DMUo is not efficient) is to move DMUo to a point located on the efficient frontier, which is the linear combination of the points in its reference set. Basically the movement of DMUo can be summarized as either decreasing inputs while keeping the status quo for outputs (for input-oriented CCR model) or increasing outputs and keeping inputs (for output-oriented CCR model) [4]. But for the case of multiple perspectives, as an attribute considered to be input from one perspective may be considered to be output from another, it is difficult to determine whether to increase or decrease its value in order to improve DMUo. The main challenge is figuring out how to measure the efficiency score of DMUo in the case of multiple perspectives, and how to adjust its attributes in order to satisfy multiple perspectives to the maximum extent. As multiple perspectives have different input/output classifications which cause different preferences in adjusting attributes of DMUo, the opinions of adjusting attributes of DMUo from these perspectives are always conflictive. For example, the management may expect to benefit from the higher profitability, but the customer may insist on decreasing profitability. Also they have different opinions in adjusting the attributes of credit quality and efficiency of a bank. If we also consider other perspectives like shareholder, employee and so on, the situation becomes much more complicated. Thus we need an appropriate method to determine how to adjust the attributes of DMUo in such a conflictive case under multiple perspectives. Nash bargaining game (NBG) [5] and [6] is a popular method in dealing with equilibrium solutions to problems involving multiple players. In the current study, each perspective of a stakeholder is defined as a player. We use NBG (1) to determine the appropriate value for each attribute, namely, whether to increase or decrease an attribute of DMUo and to what extent to improve an attribute; (2) to improve DMUo in different directions which can provide multiple improving schemes for decision maker. The proposed game mode of NBG is cooperative. Multiple perspectives negotiate for a higher efficiency score in fixing the appropriate value of an attribute of DMUo. There exist various other approaches, like weighted sum or weighted mean of efficiency scores of multiple perspectives. But these methods cannot ensure the equilibrium for multiple perspectives, and the results may invoke discontentment of some perspectives. For a complicated banking system, it is not an advisable method to sacrifice the profit of some perspectives. The remainder of this article is structured as follows. In Section 2, we briefly review the previous studies related to our research. In Section 3, we introduce the main process of efficiency improvement in the case of multiple perspectives. A numerical case study of 65 Japanese banks is given to show the actual application of the method in Section 4. The paper ends with conclusion and a discussion about the results.

This paper illustrates a method about how to select an appropriate scheme to improve the efficiency of a bank from multiple perspectives. As different perspectives have different preferences in increasing or decreasing an attribute of DMUo, we use NBG value to describe efficiency score of DMUo for multiple perspectives. Thus the NBG value is an equilibrium solution which can avoid incurring discontentment of some perspectives. The data set of 65 Japanese banks is used to demonstrate the model we proposed. NBG values for all attributes of DMUo are calculated by Eq. (7), and improving schemes are provided for different conflictive attributes: NIA/SE, ROAE, C/I and DPS. The detailed improvement process along selected attribute is also discussed from the aspect of finance in section four. As the main methodological section of the study, Eq. (7) can be modified in some parameters to get more significant results. (1) wj denoting market weight of perspective j is set as “1” in the current study, which means each perspective has the same market status. For other studies which might have perspectives with different market weights, the model is still adaptive by replacing the value of wj. (2) Breakpoints of all perspectives are set as the lowest efficiency scores along each improving direction. In future study, we may have special request about some perspectives, for example, a bank may request its efficiency score for shareholder to be above 0.9 in the process of NBG. In such case, the breakpoints of according perspectives should be added into Eq. (7). (3) The improving directions in our current research are the directions of conflictive attributes. We also performed the model in many other directions which we will not give further description due to the limit of the space. The research follows the input/output classifications and the concept of multiple perspectives that have been frequently referred to in the recent DEA literature. More important is we give an improvement scheme based on the results of efficiency evaluation from multiple perspectives, which may be a new method for other researchers who are interested in performing DEA from multiple perspectives.