Performance Analysis of Listed Companies in the UAE - Using DEA Malmquist Index Approach

Data Envelopment Analysis (DEA) is becoming an increasingly popular tool for assessing the relative performance of industries and companies. By applying DEA theory to the non-financial sector, the relative efficiency of 27 listed corporations in the United Arab Emirates (UAE) has been analyzed in this paper. The focus of the study has been on the impact of the financial crisis and the recovery thereafter. Further, the productivity change was decomposed into technical efficiency change and technological change by using the non-parametric Malmquist Productivity Index (MPI) over the period from 2007 to 2014. Based on Malmquist analysis, we find that the most efficient industries during the post-crisis period were food and beverages, telecommunication and pharmaceuticals. In contrast, the sectors that were adversely affected by the crisis were services, real estate, construction and cements. The break-up of the TFP indicated that the efficiency indices in the top performing industries were driven by technological improvements or frontier effects. The top-performing companies in the UAE during the 2007-14 period demonstrated innovation- led growth, aided by the use of better technology, investments in capital equipment, and adoption of new production processes.


Introduction
The global financial crisis during [2008][2009]  In the face of the economic slowdown, the UAE government began to pursue its diversification strategy into non-oil high-value manufacturing and service sectors. They adopted various industrial policies for sustainable development in The present study looks into the impact of the global recession on the nonfinancial sectors in the UAE. We apply data envelopment analysis (DEA) techniques to analyze productivity growth for the publicly listed companies in the UAE. The study adopts the efficient frontier approach, by using Malmquist's [3] productivity index, based on DEA. The index breaks down changes into two mutually exclusive and exhaustive components: namely, changes in technical efficiency over time and shifts in technology over time.
The remainder of this paper is organized as follows. Section 2 introduces DEA and its extensions to the TFP growth through computation of the Malmquist Index. Section 3 describes the data and empirical findings. Section 4 discusses the sector-wise analysis of the empirical results. Section 5 provides conclusions and possible future directions.

Data Envelopment Analysis
In this paper, we use performance measures of decision-making units (DMUs) using data envelopment analysis (DEA) techniques. Estimation of frontier production and measures of efficiency began with Farrell [4]. He adopted an econometric approach and measured technical efficiency through actual deviations from an idealized frontier isoquant. Following him, there were several studies that looked into the production function estimation through different specifications of the function itself [5] [6] [7] because the choice of functional form brings a series of implications with respect to the shape of the implied isoquants and the concepts of efficiency frontiers.
Electronic copy available at: https://ssrn.com/abstract=3345138 The econometric approach was subsequently taken over by the mathematical programming approach using Data Envelopment Analysis (DEA). DEA is an optimization method that uses linear programming for assessing the efficiency and productivity of DMUs in terms of a proportional change in inputs or outputs [8]. Technical efficiency is achieved with the maximization of outputs, from the use of a given set of inputs [9].
Suppose a decision making unit (DMU) generates the outputs To measure the efficiency of DMU "p", Charnes, Cooper, and Rhodes [9] defined the fraction utilizing the total factor productivity rates as The Data Envelopment Analysis program utilizes this rate of total factor productivity and maximizes the performance of the DMU "p" relative to the performance of the other units. We can transform the fractional programming model into a linear programming model [8] [10] as follows. Electronic copy available at: https://ssrn.com/abstract=3345138

Malmquist Productivity Index (MPI)
The Malmquist Productivity Index (MPI) consists of distance functions representing multi-output and multi-input technologies based on the input and output quantities. The output distance function is used to consider a maximum proportional expansion of the output, given the inputs and the MPI measures the Total Factor productivity (TFP) growth change between two data points by calculating the ratio of the distances of each data point relative to a common technology.
The MPI represents Total Factor Productivity (TFP) growth of a DMU and reflects the increase or decrease in efficiency with progress or regress of the frontier technology over time under multiple inputs and multiple outputs framework.
The TFP index is used to estimate the productivity change, which is decomposed into technical efficiency change and technological change.
The framework employed in the current study can be illustrated by Figure 1 following Fare et al. [11] [12], Hjalmarsson and Veiderpass [13], Berg, Førsund and Jansen [14], and Price and Weyman-Jones [15]. Figure 1 shows two observations on the input x and output y domain, at time t and t + 1. The objective is to measure the productivity growth between t and t + 1, in terms of the change from ( ) z t to ( ) 1 z t + and this is done by imposing a potential production frontier, as in Figure 1. The frontiers represent the efficient levels of output "y" that can be produced from a given level of input " x ". In order to make production technically efficient, the bundle ( ) z t can be reduced by the horizontal distance ratio (ON/OS). Again, in order for z(t + 1) to be efficient in period (t + 1), it must be reduced by the horizontal distance (OP/OQ). Malmquist [3] based his index on the output distance function and Caves et al. [16] extended the measure by a multi-input, multi-output index. Färe et al. [11] measured the frontier shift as the relative distance between the frontiers at t and t + 1 and further measured the MPI as a geometric mean of such indices which was factorized into the product of technical efficiency change and technological change. The relative movement of a particular DMU over time will depend on both its position relative to the corresponding frontier (technical efficiency) and the position of the frontier itself (technological change).
In order to calculate these indices it is necessary to solve several sets of linear programming problems. If there are N companies and each uses varying amounts of K different inputs to produce M different outputs, then we would have ( K N × ) input matrix and ( M N × ) output matrix. This would help us construct a nonparametric envelopment frontier over the data points such that all observed points lie on or below the production frontier. The input distance functions that have been used to construct the Malmquist indices in this paper are the reciprocals of Farrell's [14] input-orientated technical efficiency measures, following the DEA model of Charnes, Cooper and Rhodes [8].
The DEA measure of efficiency is relative in that it is calculated in relation to all other DMUs in the sample. The efficiency score for each company is compared with technologies from the same time period and then compared to technologies from different time periods. The Malmquist Productivity Index constructs an efficient frontier based on the data and each company is then compared to that frontier. The closeness of a company to the frontier is "catching up" while the shifting of the frontier itself is the "innovation". The product of these two components yields a frontier version of productivity change, which is the Malmquist TFP Index.
The Malmquist productivity index measures the total factor productivity change between two data points over time, by calculating the ratio of distances of each data points relative to a common technology. Färe et al. [12] define the change of productivity between period t and t + 1 as where M is the Malmquist productivity index in time t + 1 using the "t + 1" technology relative to the production in time t using the "t" technology. This equation represents the Malmquist productivity or TFP growth between the time period t and t + 1 as a geometric mean of the efficiency measures of each period.
Moreover, the efficiency measure in a particular time period is given by the distance function relative to the frontier technology. A value greater than one will indicate a positive TFP growth from period t to period t + 1 while a value lesser than one will indicate a decrease in TFP growth relative to the previous year. We now write the Malmquist Productivity Index as Electronic copy available at: https://ssrn.com/abstract=3345138 where the first term defines changes in technical efficiency (E) from period t and t + 1 and the second indicates changes in technology (P), i.e., a shift in the frontier from period t to period t + 1. So, we now have our equations as follows: If an organization fails to achieve an output combination on its production possibility frontier, and fails beneath this frontier, it can be said to be "technologically inefficient". Over time the level of output an organization is capable of producing will increase due to technological changes that affect the ability to optimally combine inputs and outputs. These technological changes cause the production possibility frontier to shift upward, as more outputs are obtainable from the same level of inputs. Thus, for any organization in an industry, productivity improvements over time may be either technical efficiency improvements (catching up with their own frontier) or technological improvements (the frontier is shifting up over time) or both.
Further, the variable returns to scale (VRS) are incorporated by introducing convexity constraints into the linear program. Using these models, and the Fare et al. [12] approach, it is thus possible to provide four efficiency/productivity indices for each firm and a measure of technical progress over time. These are: 1) technical efficiency change (E) (relative to a CRS technology); 2) technological change (P); 3) pure technical efficiency change (PT) (relative to a VRS technology); 4) scale efficiency change (S); and 5) total factor productivity (M).
We can then calculate scale efficiency from the technical efficiency measures for VRS technologies in the following way: Technical Efficiency Index VRS = Technical Efficiency Index VRS × Scale Efficiency Index Fare et al. [12] decomposed the catching up effect into "pure" technical efficiency change and "scale" efficiency change. That is, If M > 1, then productivity gains occur, and if M < 1, productivity losses occur.

Technical efficiency increases (decreases) if and only if E is greater (less) than
one and technological progress (regress) has occurred if P is greater (less) than one. An assessment can also be made of the major sources of productivity gains/losses by comparing the values of E and P. If E > P then productivity gains are largely the result of improvements in efficiency, whereas if E < P productivity gains are primarily the result of technological progress. Moreover, if PT > S then the major source of efficiency change is improvement in pure technical efficiency, whereas if PT < S the major source of efficiency is an improvement in scale efficiency. Further details on the interpretation of these indices may be found in Charnes et al. [17].

Literature Review
The first empirical application of the decomposition of productivity changes into technical and technological changes was used to measure hospital productivity in Sweden by Nishimizu and Page [18]. The enhanced decomposition model to identify between pure technical efficiency and scale efficiency was carried out by Färe et al. [12] while analyzing the productivity growth in 17 OECD countriesthe United States was found to consistently shift the frontier driven by technical change but the productivity growth in Japan was above average mainly due to technological changes. Following this there have been numerous empirical studies that followed this methodology to measure productivity changes at the country level; industry level as well as firm level. Most of the studies have concentrated on the financial sector, especially banking, and the literature is quite scant in the non-financial sector. Since the present study looks into the non-financial sector of the UAE, this section will concentrate of a review of existing literature relating to the non-financial industries.
Some of the countries that have been included in the research for performance evaluation of non-financial sector in the Middle Eastern region include Israel by Friedman and Stern [19]; Jordan by Shammari [20] and Turkey by Ulucan [21].
A world-wide study of Fortune 500 companies was done by Zhu [22] [26] studied US automakers for the period 1955-1979 and found a decreasing productivity growth for Chrysler. In fact a comparative trend showed that the big three Japanese producers (Toyota, Nissan, and Mazda) had achieved higher labor productivity during the 1970s than their US Big three counterparts Lieberman [27].
DEA approach has also been applied for various other industries across the globe. Performance evaluation of the pharmaceutical sector using the non-parametric DEA approach has been carried out in India and output efficiency of firms has been studied by Majumder [9], Saranga, and Phani [28] and Mazumdar and Rajeev [27]. Keramidou et al. [29]  The performance evaluation of small and medium enterprises has also been of particular interest, especially in the Far Eastern countries. Batra and Tan [35] found [37]. Singh [38] found the performance of the private sector sugar mills in North India to be better performers as compared to cooperative and government sectors. Higher inefficiencies of public sector firms were also found for pharmaceutical firms in India [9]. For the state owned Electric utilities in India, only 24 percent of the firms were efficient [39].
Therefore, we find that the non-parametric productivity analyses have been

Research Methodology
We evaluated the overall efficiency of 27 firms in the United Arab Emirates by We have taken a balanced panel and the data was collected from individual audited annual reports from companies for each year. Table 1 gives the classification of the data in terms of the industry.
Our study included two outputs and three inputs. Ideally, the analysis of efficiency should include the physical volume of outputs and inputs. However, in the absence of data following standard practice [40], we have used values of outputs and inputs according to the financial statements of the companies.
In this study there are 27 DMUs with three inputs: Labour (measured in terms of the wages and salaries of workers); Capital (measured in terms of the net book   Table 2 and Table 3   crisis and found it difficult to recover during this period were services, real estate, construction and cements (Figure 2).    Electronic copy available at: https://ssrn.com/abstract=3345138   There are three listed companies in the Foods and Beverages sector in the Electronic copy available at: https://ssrn.com/abstract=3345138  Table 5 shows that Foods and Beverages sector showed the highest overall TFP score with Dubai Refreshments and Foodco being the two top performers (Table 6). All the three listed food companies-Foodco, Dubai Refreshments and Agthia-showed positive technological change during the 2007-2014 period. Dubai Refreshments leads the carbonated soft drinks sector and made huge investments in capital projects for building new capabilities. The company's relocation to its new state of the art distribution facility at the Dubai Investment Park enhanced their production capacity and operational efficiency. Investments in 2013 also went up with Agthia's new water bottling line and a new mega distribution center along with the completion of their research and development laboratory.

Food and Beverages
Although the food and beverages sector showed the highest level of efficiency, it was driven by technological advances without any commensurate efficiency improvements in the internal management of the firms. Foodco was the only company that showed a combination of technological and technical efficiency.  However, such human resource incentives of Foodco to improve internal technical expertise of the company were not evident across the industry. Despite the technological improvements and capital expansions, both Agthia and Dubai Refreshments showed decline in technical efficiency, implying positive investments without a balance of the new inputs versus outputs.

Telecommunications
The UAE telecommunications sector is among the strongest and most advanced in the world and was served by a monopolist provider, Emirates Telecommuni-

Pharmaceuticals
Healthcare has been a top priority for the government and the sector received strong incentives to encourage local manufacturer through favorable pricing structures of generics; reduction of imports; policies relating to mandatory health insurance. The pharmaceutical industry remained strong and stable during the recession years owing to mandatory health insurance and the growth of Gulf Pharmaceuticals mainly drove the growth of the industry through its entry into the high-technology field of biotech as it began production of bulk human insulin crystals. Despite the recessionary pressures, the company continued to launch new products to diversify its market presence and successfully began off-shore investments in Saudi Arabia, Ethiopia and Algeria. However, the major hurdle for local manufacturers was the capital intensive nature of operations and the shortage of knowledge and skilled manpower which led to the decline in the technical efficiency of the sector. Further, the companies are highly dependent on imports of equipment, ingredients, and medicines for end use which makes the industry vulnerable to foreign exchange fluctuations, leading to a fall in the Malmquist index since 2011.

Services, Construction and Cements
Significant decline in tourism, widespread closures and exodus of expatriates from the UAE led to the downtrend in these sectors. The services industry recovered the quickest owing to international conferences/exhibitions by multinational companies; leisure tourism from the emerging Asian economies; niche markets like spa tourism, halal tourism, cultural tourism and budget tourism.

Conclusions
Performance measurement is crucial for any firm in any sector, not only for determining its own efficiency and achievement but also by benchmarking itself in comparison with its peers. This research paper uses the Malmquist productivity index which has been widely used across the globe, to evaluate the productivity Therefore, this study makes a major contribution to the existing literature of efficiency measures using non-parametric approach.
There are some limitations of this research since the DEA approach determines only relative efficiency and also does not identify the factors that give rise to inefficiency. So, we could only highlight those units in which inefficiency exists and those that require attention and the inefficiency index is only relative to the most efficient frontier. More importantly, we would be able to gain more insight by incorporating a model of structural change within this framework-so that we can identify the difference in trends during the crisis years (2007)(2008)(2009)(2010) and the recovery period (2010-2014).