# Comparing Company’s Performance to Its Peers: A Data Envelopment Approach

**Tihana Skrinjaric**

University of Zagreb

## Introduction

A company needs to re-evaluate its performance and constantly compare itself to others (Soboleva et al. 2018). This is due to the pressure continuously rising in all business industries, constant changes in the market, demand, technology, sustainability demands, and other factors influencing the business itself. Thus, it is important for a company to know where it stands compared to others. In order to do this, an objective approach needs to be made, where important and relevant variables and factors are taken into consideration. The management cannot make good decisions for future business if the meaningful analysis is not made (Narkuniene and Ulbinaite 2018). Concept of business performance is now commonly used not only within academic literature but in circles of professional managers (Yildiz and Karaka 2012). Furthermore, adequate mathematical modelling should be applied and used in order to obtain objective results. It is easily seen how the whole process of obtaining such results needs cooperation and interaction between the firm’s management, quantitative modellers, and financial experts as well. As Matthias et al. (2017:41) state that analysis, which is not critical and done on poorly understood data, cannot generate new knowledge. Such analysis has become a necessity (Sarlija and Jeger 2011). On the other hand, such comparisons can be made from the (potential) investor’s side, which is looking at investment possibilities for their portfolios. Dynamic changes in the stock markets and in the businesses themselves force the investors to constantly re-evaluate the financial assets and companies in their portfolios. Again, an objective approach of comparisons and ranking system are needed, which can facilitate investment decision-making. Furthermore, business banks also have benefits in using such data when making decisions on approving new loans and constructing credit scoring (Roje 2005; Chan-Lau 2006; Demerjian 2007).

Recent decades have experienced the development of many different models, methods, and techniques within mathematics, statistics, and econometrics in order to answer specific questions. Some classifications of different approaches were made in Granger (1989), Ho et al. (2002), Taylor and Allen (1992), and Wallis (2011). On the other hand, evaluating business performance has become more complex, due to many different aspects of the business itself which need to be taken into consideration. The financial ratios are often used in order to compare businesses one to another (Yalcin et al. 2012; Neely et al. 1995; Marshall et al. 1999; Najmi and Kehoe 2001), as they give insights into the profitability, productivity, liquidity, and other relevant aspects of the business. Due to many financial ratios that need to be evaluated for many companies, such problems lie within big data analytics. Big data can be very useful in the decision-making process (Li et al. 2016; Matthias et al. 2017), but the complexity of analyzing so many financial ratios can be seen in Beaver (2010), MySkova and Hajek (2017), or Laitinen (2018).

One of the popular approaches includes some of the models from Data Envelopment Analysis (DEA) (Feroz et al. 2003; Kohers et al. 2000; Cummins et al. 2000; Yu et al. 2013), as a branch of Operations Research (OR), a set of mathematical models, and methods which are used to evaluate relative efficiency of alternatives which are being compared. This is not a new approach in comparing the businesses one to another. However, previous related research mostly focuses on basic models (e.g., Charnes-Cooper-Rhodes and Banker-Charnes-Cooper models), which have several drawbacks. Furthermore, research often compares a relatively smaller number of firms and even starts with a small number of financial ratios in the analysis. That is why this research focuses on a bigger sample with respect to the number of firms and financial ratios, in order to reflect real-life problems in such analysis. Furthermore, this research employs the SBM (slacks-based measure) model within DEA methodology, as this approach has several advantages compared to basic models. The SBM model does not depend on the data translation, measures of each variable, and is nonradial (for details, please see the methodology section). Thus, it could provide better results in terms of reliability. Another contribution of this research is that robustness checking of the results is conducted. This is often ignored in the literature. The robustness testing will be performed by another methodological approach, multiple criteria decision-making (MCDM) model. This model belongs to another branch of OR, which is used in constructing a ranking system of different alternatives based on several criteria. Thus, the main goals of this research include the following. First, a comprehensive literature overview will be given, in order to obtain as many insights as possible. Second, a detailed empirical analysis will be provided so that all those interested could make similar research in the future, with straightforward interpretations. Third, the mentioned robustness checking will be performed, so that the results obtain greater reliability.

The rest of this research is structured as follows. The second section deals with the literature overview which is most related to this study. The methodology used in this study is described in the third section. The fourth section deals with the empirical analysis, while the final, fifth section concludes the research with recommendations for future research.

## Previous Related Research

By observing the existing literature which examines the financial ratios in business comparisons, it can be seen that probably all of the ratios have been used in empirical analysis at some point. Commonly used methodologies include econometric techniques (regression analysis, time series analysis methods, and models) (e.g., Penman et al. 2007; Jordan et al. 2007; Dempsey 20Ю) in which the stock returns are modelled and forecasted based on financial ratios; nonparametric approaches such as the DEA (Zamani et al. 2014; Skrinjaric 2014), Gray Relational Analysis (GRA, Fang-Ming and Wang-Ching 2010; Huang et al. 2015; Skrinjaric and Sego 2019), and Analytic Hierarchy Process (AHP, Li et al. 2010) in which the ranking systems are made based on the financial ratios criteria.

Earlier work includes testing the Efficient Market Hypothesis with book-to-market ratio as a proxy for the value premium of stock returns (Fama and French 1992, 1993). The focus was made on more developed markets: Japan in Chan et al. (1991), USA in Kothari and Shanken (1997), and a set of developed countries (UK, USA, Belgium, Germany, France, etc.) in Fama and French (1998). Another popular financial ratio is the dividend yield, which has its roots in Lintner (1956), Brennan (1970), and Litzenberger and Ramaswamy (1979). Today, different dividend policies are being considered in theory and empirical applications (stable dividend policy, Leary and Michaely 2011; residual policy, Baker and Smith 2006, etc.). Since the early work of Basu (1977, 1983), the P/E (price-to-earnings) ratio has been included in the asset pricing models as it increases the forecasting power of such models (see Noda et al. 2015; Alcock et al. 2011). Earnings per share (EPS), return on assets (ROA), liquidity indicators, and receivable turnover ratio have been found as the most useful indicators about the business in Wu (2000). ROA has been in focus in Muhammad and Scrimgeour (2014), as it is a proxy for firm’s performance and management’s efficiency to generate profits from assets, whereas others (Jablonsky and Barsky 2001) argue that ROIC (return on invested capital) is a better measure compared to ROA and ROE, due to ROIC being calculated based on value above the average cost company pays for its equity capital and debt. The econometric approach of estimation has been very popular over the decades. Some of the recent research includes Lee and Lee (2008), Dempsey (2010), Gregoriou et al. (2017), etc. Other approaches include chi-square test and ratio analysis (Damjihabi 2016), or factor analysis (Hornungova and FrantiSek 2016). Banking systems have also been often observed in the literature as well (see Maradin et al. 2018). However, the approaches in such research are not linked closely to this one. The second group of research utilizes similar approaches. Thus, the focus will be made more on them.

The DEA and related approaches are found in the following papers. Powers and McMullen (2002) have focused on 185 American and British stocks in evaluating their efficiency by including EPS, market betas (capital asset pricing model), standard deviation, and other risk and return indicators. 230 American firms have been examined in Edirisinghe and Zhang (2007), where authors developed a generalized DEA indicator (RFSI - relative financial strength indicator). A simulation was included in the study, where authors compared trading strategies based on including the RFSI variable in the consideration or not. Some of the papers which focus mostly on creating trading strategies from the investor’s point of view include Chen (2008), who has observed three years of data (2004-2007) on the Taiwanese market; Lopes et al. 2008), who have examined the Brazilian stocks; Lim et al. (2013), who have used a large amount of market and financial ratios data in their analysis; Zamani et al. (2014) have focused on the Mumbai stocks; Gardijan and Skrinjaric (2015), who have examined the Croatian stock market, etc. However, such research uses stock market characteristics, financial ratios, or combination of both. Such analysis does not provide some answers into the (in)efficiencies of businesses; its purpose is to provide rankings of stocks based on some criteria which is important for the investor so that he can re-evaluate his portfolio over time and rebalance it accordingly.

Emrouznejad and Cabanda (2010) examined the General Non-Parametric Corporate Performance via DEA model on a sample of 27 UK industries and 6 performance ratios. This research was more technical (comparing methodological approaches and changes) and did not focus on questions as this research does. A combination of DEA model with other methodologies is found in Rosini and Gunawan (2018) where authors combined DEA with TOPSIS (Technique for Order of Preference by Similarity to Ideal Solution); Ding and Sickles (2018) where authors combined DEA, SFA (stochastic frontier analysis), and panel two-step GMM (generalized method of moments) on US banks; Fang-Ming and Wang-Ching (2010) where the DEA was compared to GSD (Grey Systems Decision); Huang et al. (2015) combined a two-level DEA and the GRA approaches, etc.

Oberholzer (2012) compared 55 manufacturing companies on Johannesburg Stock Exchange with several input and output variables within the DEA model (sales, dividend payouts, tangible assets, etc.) and found usefulness in such models to detect relative (in)efficiencies when comparing firms. Demerjian (2017) is more focused on methodological aspects of using DEA models in financial analysis. This research is extensive but provides detailed insights into the robustness of such models in particular applications. Lin et al. (2010) focused on the shipping industry combined DEA with ABC (activity-based costing) methodologies. The sample included only 14 firms with 6 variables. Siew et al. (2017) examined financial companies in Malaysia (for period 2010-2015), with basic calculations and interpretations (ranking of the companies, comments on the efficiency, and input and output weights in the optimization process). Malaysian stocks have been examined in Arsad et al. (2018) as well, where SFA and DEA results were compared for 115 companies in 2015. The comparison showed that rankings differ based on the chosen methodology, but no explanations were given on why.

A lot of researchers do not state why they do or not include specific financial ratios and other variables in the analysis. The majority of research uses yearly data due to the nature of the variables, as companies release balance sheets and other relevant financial statements on a quarterly and yearly basis. Furthermore, a lot of research utilizes basic DEA models, which forces the researchers to use those variables which are suitable for such models (due to the model assumptions). Thus, the contribution of this research includes explanations and rationale on why some ratios are used in the empirical part or not, with extensions of basic models so that the model is closer to reality.