Understanding efficiency levels is crucial for understanding the competitive structure of a market and/or segments of a market. This study uses two artificial neural networks (NN) and a traditional statistical classification method to classify the relative efficiency of top listed Egyptian companies. Accuracy indices derived from the application of a non-parametric data envelopment analysis approach are used to assess the classification accuracy of the models. Results indicate that the NN models are superior to the traditional statistical methods. The study shows that the NN models have a great potential for the classification of companies’ relative efficiency due to their robustness and flexibility of modeling algorithms. The implications of these results for potential efficiency programs are discussed.
The competitiveness of a country derives from the efficiency of its enterprises. While competitiveness at the national level is reflected in the performance of the country, it is reflected in the size of the market share at the company level (Porter, 1998). Both notions highlight the importance of efficiency and performance evaluation. Efficiency evaluation and benchmarking are widely used methods to identify the best practices as a means to improve the performance and increase productivity (Barros, 2004). Measuring efficiency levels has become an important issue for managers and investors alike (Galagedera & Silvapulle, 2002). Consumers also benefit from efficient resource usage and allocation because this may mean lower prices and more professional service (Anderson, Fok, Zumpano, & Elder, 1998).
Gandjour, Kleinschmit, Littmann, and Lauterbach (2002) concluded that many quality and efficiency indicators used by executives are lacking in general validity. Using a recognized and valid measure of efficiency is critical for managers seeking to increase the effectiveness of their organizations. Over the past two decades, data envelopment analysis (DEA) has become a popular methodology for evaluating the relative efficiencies of decision making units (DMUs) within a relatively homogenous set (e.g. Sun & Lu, 2005). DEA is an approach to estimate the production function of organizations and organizational units and enables the assessment of their efficiency.
Although widely employed to evaluate efficiency across industries (e.g. Rickards, 2003), DEA can hardly be used to predict the performance of other DMUs (Wu, Yang, & Liang, 2006). As a result, neural network models (NN) were introduced recently to complement DEA in estimating efficiency frontiers of DMUs (Wang, 2003). Wang (2003) showed formally that neural network find data envelopes based on the entire data set, rather than some extreme data points. Athanassopoulos and Curram (1996) were first to combine NN and DEA for classifying and predicting efficiency in bank branches. A comprehensive search through several databases yielded no studies dealing with companies’ efficiency using a DEA-NN approach. This confirms Santin, Delgado, and Valino (2004, p. 630) claim that NN models “have no theoretical studies in efficiency analysis and few applications have been made in this field.” We, going beyond the conventional methods, have attempted to merge both methodologies to evaluate the relative efficiency of the top listed companies in Egypt. The paper also contributes methodologically through the comparison of various parametric and non-parametric techniques, which results in considerable information for business analysis. More specifically, the purpose of this research is twofold:
•
To assess the market performance of the top listed companies in Egypt; and
•
To benchmark the performance of NN models against traditional statistical techniques.
This paper is organized as follows. The next section summarizes the methodology used to conduct the analysis. The subsequent section presents empirical results of the efficiency levels of Egyptian companies. After a brief preliminary data analysis, this section first set out efficiency scores derived from estimating the basic DEA models; it also presents sensitivity analysis of DEA-NN derived efficiency scores as a rough validity check on the results. Next, the paper sets out some managerial and policy implications of the analysis. The final section of the paper deals with the research limitations and explores avenues for future research.
The simple DEA model is based on constant returns to scale (CRS), implying that the size of a company is not relevant when assessing efficiency. However, it is likely that the size of the company will influence its ability to produce goods and services more efficiently. As the CRS totally ignores the scale of operations and will possibly lead to an identification of very unrealistic benchmarks (Munksgaard, Pade, & Fristrup, 2005), a variable return to scale model (VRS) is used in this study. A VRS frontier allows best practice level of outputs to inputs to vary with size of company. A DEA model can be analyzed in two ways, an input-orientation or an output-orientation. An input-orientation provides information as how much proportional reduction of inputs is necessary while maintaining the current levels of outputs for an inefficient company to become DEA-efficient. On the other hand, an output-orientation analysis provides information on how much augmentation to the levels of outputs of an inefficient company is necessary while maintaining current input levels for it to become DEA-efficient. Since it is well known that, in competitive markets, the DMUs are output-oriented (Barros & Athanassiou, 2004), we use the output maximization assumption in this study.
To ensure the validity of the DEA model specification, an isotonicity test (Avkiran, 1999) was conducted. An isotonicity test involves the calculation of all inter-correlations between inputs and outputs for identifying whether increasing amounts of inputs lead to greater outputs. As positive inter-correlations were found (see Table 1), the isotonicity test was passed and the inclusion of the inputs and outputs was justified.
