سنجش کارایی سیستم های شبکه ای: تاثیر فن آوری اطلاعات (IT) بر عملکرد شرکت

A recent development in DEA (data envelopment analysis) examines the internal structure of a system so that more information regarding sources that cause inefficiency can be obtained. This paper discusses a network DEA model which distributes the system inefficiency to its component processes. The model is applied to assess the impact of information technology (IT) on firm performance in a banking industry. The results show that the impact of IT on firm performance operates indirectly through fund collection. The impact increases when the IT budget is shared with the profit generation process.

One of the most useful methodologies for measuring the relative efficiency of a set of decision making units (DMUs) which utilize multiple inputs to produce multiple outputs is data envelopment analysis (DEA) developed by Charnes et al. [15]. This methodology has been applied to research related to decision support systems [32]; for example, facilitating the agent's intelligent behavior in agent-based merchandise management [50], selecting learning cases to improve the forecasting accuracy of neural networks [51], assessing the contribution of knowledge to business performance [1], measuring the efficiency in Internet companies [55], selecting enterprise resource planning (ERP) software [7], performing group evaluation of production units [59], and evaluating data warehouse operations [44]. Advances in DEA methodology will further aid research and applications in decision support systems. In a production system, the input usually goes through several processes before it becomes the output. Traditional DEA models treat the system as a whole unit, disregarding the interactions of the processes in the system when calculating the efficiency. The first paper discussing this idea was prepared by Charnes et al. [13], which found that the army recruitment had two processes: the first created awareness through advertisement, and the second created contracts using other recruitment resources. Separating large operations into detailed processes helps identify sources of inefficiency and the real impact of factors. Many empirical studies have successfully applied this idea to real world problems. However, it has been frequently observed that, for some DMUs, the system is efficient while the component processes are not. For this reason, Färe and Grosskopf [27] proposed the idea of network DEA, taking the operation of component processes into consideration in calculating the efficiency of the system. Several models for measuring the efficiency of network systems have been proposed [17], [27], [35], [39], [52], [61] and [68]. They can be classified into three groups. The first is an independent approach which recognizes the existence of the processes in the system, yet the efficiencies of the system and all processes are calculated independently. The second is a connected approach, in that interactions between processes are taken into account in calculating the system efficiency. There are several variations on this approach; some are able to calculate the system efficiency and process efficiencies in the same mathematical program, while others need to rely on the conventional DEA model to calculate the process efficiencies separately. The third is a relational approach; its underlying concept is that some kind of mathematical relationship exists between the system efficiency and the component process efficiencies; for example, simple multiplications [37] and weighted average [18]. In this paper, we discuss a model which provides a unified mathematical relationship between the system efficiency and process efficiencies for all types of network structure. For illustration, the problem of assessing the impact of information technology on the performance of a firm as discussed in Wang et al. [65] and Chen and Zhu [21] is revisited. Moreover, a model for measuring the efficiency when certain resources are being shared is used to obtain a better assessment. The rest of this paper is organized as follows. Section 2 introduces the idea of the relational approach using an example. Section 3 discusses how this approach is used to model series and parallel systems. The problem of assessing the impact of IT on bank performance is discussed in Section 4. Finally, some conclusion is given in Section 5 based on the discussion of the results.

Conventional DEA models for measuring the efficiency of a system treat the system as a black box, disregarding its internal structure. More representative and informative results can be obtained if interactions of the component processes within the system are taken into account. The independent network approach calculates the system and process efficiencies independently, which sometimes produces inconsistent results between the system and process efficiencies. The connected network approach does not have this problem, but the relationship between the system efficiency and process efficiencies cannot be obtained. This paper discusses a relational network approach which took into account interactions between component processes when calculating the system efficiency. The system and process efficiencies can be calculated at the same time and the relationship between them can be obtained: the system slack is the sum of process slacks. Based on this relationship, the system inefficiency can be distributed to processes according to the proportion of process slack in the system slack. The key processes, which cause the system to be inefficient, are thus consistently identified. The problem of “the IT impact on firm performance” was discussed. The impact of IT on firm performance is indirectly through an IT-produced product, namely deposits. This conclusion is consistent with those of previous studies. For the bank system, approximately 56% and 44% of the system inefficiency were attributed to the two component processes, fund collection and profit generation, respectively. By allowing a portion of the IT budget to switch from the fund collection operations to profit generation operations, the bank performance improved approximately 10.2%. This result also provides a better estimate of the bank performance. The model discussed in this paper was a CCR-type model with the underlying assumption of constant returns to scale. It should not be difficult to extend it to the BCC-type model [3] to accommodate situations of variable returns to scale. The economies of scale of the system and component processes can then be discussed. With the flexibility of the relational model for modeling general network systems, the dynamic performance of the bank system in the long run can also be measured to obtain a clearer idea of the IT impact on firm performance. Recently, the efficiency of banking firms has received increased interest due to bad performing loans. Bad performing loans, if they exist, cannot be ignored in efficiency evaluation because they increase managerial risk. Since conventional DEA models fail to account for risk factors, the resulting efficiency measures will be biased if a banking firm has bad performing loans. The network DEA model can deal with risk factors by incorporating them as undesirable output. This concern provides a direction for future study.