Productivity and Efficiency Measurement of Airlines: Data Envelopment Analysis using R

By Boon L. Lee

In today’s competitive environment, airlines are doing everything they can to improve efficiency and productivity. Productivity and Efficiency Measurement of Airlines: Data Envelopment Analysis using R identifies and explains sources of airline efficiency and helps achieve these goals through the use of state-of-the-art measurement techniques.

Each chapter measures airline performance through the data envelopment analysis (DEA) model and other DEA variants. This book thoroughly discusses topics such as cost and revenue efficiency performance, carbon emissions performance management, and complex airline data analysis, employing appropriate models for each. Model methodologies are also discussed. The in-depth coverage is useful for all audiences, including students with a basic understanding of models, researchers and airline operators and management.

Productivity and Efficiency Measurement of Airlines: Data Envelopment Analysis using R provides R codes to help readers generate results and quantify efficient practices. These results provide airline decision-makers with the essential information they need to create better policies and avoid underperforming practices.

• Thoroughly summarizes key DEA measurement models for productivity and efficiency ofairlines
• Guides users in generating airline performance results using DEA model and its variants
• Features R codes useful for generating empirical results, and best practices, promoting qualitypolicy and management decisions

• Language: English
• Publisher: Elsevier Science
• Release date: Feb 24, 2023
• ISBN: 9780128126974

Boon L. Lee

Boon Lee is Associate Professor in economics at the Queensland University of Technology, Faculty of Business and Law, in Brisbane, Australia. His primary research interest focuses on eco-efficiency, efficiency and productivity analysis with a focus on key service industries, such as the airline industry and higher education. He is experienced in the application of data envelopment analysis (DEA) in measuring efficiency of firms and has published several articles on efficiency and productivity of airlines, including in Elsevier’s Journal of Air Transport Management. He has widely published in international journals including Journal of Development Economics, Energy Economics, Tourism Management, OMEGA, Economics of Education Review, Journal of Productivity Analysis, Agricultural Economics, Ecological Economics, Land Use Policy and Journal of Transport Economics and Policy. He is a certified facilitator of the LEGO® SERIOUS PLAY® method and materials. He is Senior Fellow of the Higher Education Academy (HEA) and Associate Fellow of the HEA (Indigenous).

Book preview

Preface

This book is designed to provide researchers and academics a range of R programming scripts and the use of R packages available from CRAN (the Comprehensive R Archive Network) to measure efficiency and productivity via data envelopment analysis (DEA). This book was inspired from the authors' experience and anecdotal comments on why there does not exist a book that provides explicit instructions via step-by-step use on R programming scripts to measure efficiency and productivity. While this book provides a wide range of R programming scripts and the use of specific R packages on DEA, it is by no means complete as it is impossible to consider all DEA models.

This book may be used as a textbook or as a reference text to accompany other DEA books. This book is extremely helpful as it provides the application side to the understanding of DEA. Undergraduates and postgraduates can use the R programming scripts or R packages detailed in the book and work through the models and generate efficiency and productivity estimates using the sample data provided in the book.

This book is not designed to provide a thorough theoretical background of the various DEA models and not designed to compete with the existing excellent DEA books. There are already plenty of such books available. For example Färe et al. (1985), Färe et al. (1994), Coelli et al. (1998), Cooper et al. (2000), Coelli et al. (2005), Cooper et al. (2007) and O'Donnell (2018). In fact, this book aims to complement all other DEA studies, theoretical and applied, by providing a step-by-step guide on the use of specific DEA models with the associated R programming scripts.

Another purpose of this book is to contextualize the efficiency and productivity performance of international airlines through the use of DEA models. The focus on measuring the efficiency and productivity of airlines is because of deregulation of the global airline industry with the aim of maximizing efficiency. By employing DEA (i.e. standard DEA and its variants), this book aims to identify and demonstrate the adoption of the appropriate DEA model in order to measure the efficiency and productivity of airlines and explain their performance since deregulation.

We believe this book offers readers something unique like a one-stop shop (or book) because the book provides a range of explicit R programming scripts and a step-by-step use on the measurement of efficiency and productivity. We hope that this book also provides a point of departure for others to further develop and improve the existing DEA models.

Chapter 1: Introduction

Abstract

This chapter introduces the reader to the two reasons for writing this book. First, to provide the reader easy-access and user-friendly R programming scripts for data envelopment analysis (DEA) and its variants to measure efficiency and productivity. Second, to measure the efficiency and productivity of airlines since deregulation. This chapter provides a brief history of the evolution of the airline industry since deregulation and a brief historical development of DEA models. By familiarizing the various DEA models, it allows DEA practitioners to appropriately measure the efficiency and productivity performance of airlines since deregulation.

Keywords

Data envelopment analysis; Deregulation; R programming

1.1. Introduction

The purpose of this book is to provide data envelopment analysis (DEA) practitioners DEA models written in the R programming script to measure the efficiency and productivity of airlines (or any other industry for that matter). This book is not meant to provide a thorough theoretical background of the various data envelopment analysis (DEA) models as there are already plenty of such books available, for example, Färe et al. (1985), Färe et al. (1994), Coelli et al. (1998), Cooper et al. (2000), Coelli et al. (2005), Cooper et al. (2007) and O'Donnell (2018). This book aims to complement all other DEA studies, theoretical and applied, by providing a step-by-step guide on the use of specific DEA models with the associated R programming scripts, and interpretation of the results therefrom. By making it user-friendly, we hope that DEA practitioners are able to apply the models with minimal fuss. As the number of DEA models is extensive, this book does not cover all of them. Only the key DEA models are covered to meet the needs of most DEA practitioners.

Another purpose of this book is to measure the productivity and efficiency performance, albeit using a sample of international airlines. The focus on measuring the efficiency and productivity of airlines is because of deregulation of the global airline industry with the aim of maximizing efficiency. By employing DEA (i.e., standard DEA and its variants), this book aims to identify and demonstrate the adoption of the appropriate DEA model to measure the efficiency and productivity of airlines and explain their performance since deregulation.

This chapter is divided into four sections. Following the introduction in Section 1.1, we provide a brief comment on the evolution of the global airline industry since deregulation in Section 1.2. Section 1.3 presents a brief history of the developments of the DEA model and evolution of variants of DEA. This chapter concludes in Section 1.4 with an outline of the chapters of this book.

1.2. Evolution and deregulation of the global airline industry—a brief comment

The Airline Deregulation Act of 1978 (ADA) and its international counterpart, International Air Transportation Act of 1979 (IATA), were major turning points in the US airline industry. Prior to the ADA, the airline industry was heavily regulated, and leading economists, such as Alfred E. Kahn, had argued that regulation creates inefficiencies and leads to higher costs. The ADA opened up domestic competition in the United States with airlines adopting competitive and operating strategies such as price competition (i.e., fare reductions) and adopting hub-and-spoke route systems. The ADA also opened up competition in the form of entrance of low-cost carriers (LCCs) into the market. The ‘Open Skies’ agreement was also introduced during this period, and it was only in 1992 when the first European country, the Netherlands, signed the first Open Skies agreement with the United States. Morrison and Winston (1986) noted that the financial outcomes from the deregulation of the US airline industry were also an impetus to the European Union's air transport liberalization. As observed in Button (2001, p.255), ‘by 1987 passenger enplanements had risen by 88% compared with 1976, employment in the industry had risen from 340,000 to 450,000, scheduled passenger miles were up by 62% and seat availability had grown by 65%’.

The deregulation process of the European Union (EU) air transport began in 1987 and comprised three packages (Button, 2001). The ‘First Package’ started in January 1988 introduced a degree of flexibility in the aviation sector. The ‘Second Package’ in November 1990 included further flexibility on fares, capacity restrictions and market access. The ‘Third package’ in January 1993 removed the remaining barriers of aviation, thus culminating in the formal EU airline liberalization.

Besides the EU, many nations also began deregulating their airline industry since the deregulation of the US airline industry. Canada began deregulating in 1979 and was fully deregulated from 1988 (Oum et al., 1991). Deregulation in Japan began in 1986 with the enactment of the revised Aviation Law in February 2000 culminating in full liberalization (Ida and Tamura, 2005). Australia began deregulating the airline industry in the late 1970s and was fully deregulated in 1990 after the repeal of the ‘Two Airline agreement’ under the Airline Agreement Termination Act, which saw the abandonment of fare setting and limits on entry to the industry (Quiggin, 1997). China began airfare deregulation in 1997 through the adoption of price discrimination (Wang et al., 2016). Airline deregulation began in Korea with the Airline Deregulation Act in 2008 (Kim, 2016).

The process of air transport deregulation in ASEAN (Association of South-East Asian Nations) progressed with each milestone of signing of agreement/declaration towards full liberalization. This began with the Bangkok Summit Declaration of 1995, which included the development of an Open Sky Policy (Forsyth et al., 2006). In 2001, the 7th Air Transport Ministers' Meeting resulted in the regional initiative for the progressive and phased liberalization of air services in ASEAN. In 2004, the 10th Air Transport Ministers' Meeting formed the Action Plan for ASEAN Air Transport Integration and Liberalization 2005–2015 (Laplace and Latgé-Roucolle, 2016). In 2009, the ASEAN Multilateral Agreement on Air Services was signed in Manila, Philippines. This was followed up with the 2010 Multilateral Agreement for the Full Liberalization of Passengers Air Services signed in Brunei Darussalam. These agreements culminated with the operation of the ASEAN Single Aviation Market in 2016, also known as the ASEAN Open Skies Agreement (Heriyanto and Putro, 2016).

The airline industry also witnessed the development of alliances among airlines, mainly to overcome various regulatory and financial obstacles (Kottas and Madas, 2018). Since 1997, the three key global airline alliances, Oneworld, SkyTeam and Star alliance, have dominated the global airline industry. The growth of these alliances is shown in Fig. 1.1.

As a result of the events that have occurred in the global airline industry, namely deregulation leading to intense competition with the entrance of LCCs; and the creation of the global alliances, the aforementioned events provide a unique opportunity to apply the nonparametric model, DEA and its variants to measure the efficiency and productivity of airlines in the era of deregulation.

1.3. A brief history of developments in data envelopment analysis

The theory behind deregulation is to achieve maximum efficiency (Keeler, 1984; Peltzman et al., 1989). To that end, this book uses the nonparametric model, DEA, to measure the efficiency and productivity performance of airlines. DEA, based on linear mathematical programming, constructs a piecewise frontier from a given data set. It essentially measures the efficiency of decision-making units (DMUs) relative to the frontier. The frontier efficiency concept was first elucidated in Farrell's (1957) seminal work, but the term DEA was first introduced in Charnes et al. (1978), henceforth CCR. The CCR model, under constant returns to scale assumption, was further developed by Banker et al. (1984), by introducing a variable returns-to-scale DEA model. Since then, DEA has been employed in numerous studies covering a range of industries. For a bibliography of DEA studies, we refer readers to Seiford (1997), Tavares (2002), Gattoufi et al. (2004a, 2004b), Cook and Seiford (2009), Emrouznejad and Yang (2018) and Contreras (2020).

Figure 1.1  Number of airlines per global airline alliance, 1997–2022. Source: Trend diagram based on data drawn from Wikipedia. https://en.wikipedia.org/wiki/Oneworld, https://en.wikipedia.org/wiki/SkyTeam https://en.wikipedia.org/wiki/Star_Alliance.

While DEA gained popularity because of its simplicity and applicability, it nonetheless has several limitations (Stolp, 1990; Coelli et al., 1998). Most of these limitations can be overcome by following the protocols described in Dyson et al. (2001) and Cook et al. (2014). There are, however, two limitations that are much more serious. As noted in Schmidt (1986), there is no error term in DEA indicating any error may well be attributed to inefficiency. Also DEA efficiency scores have no statistical significance due to its nonparametric nature (Grosskopf, 1996). The lack of statistical inference has since drawn considerable interests among practitioners of DEA to overcome these challenges by incorporating sensitivity analysis using statistical tests (Sengupta, 1987; Valdmanis, 1992) and bootstrapping (Simar and Wilson, 1998).

DEA has evolved to accommodate and reflect a range of production models that the standard DEA is unable to. For example, variants of DEA models were developed to reflect the production models that incorporate both good (desirable) and bad (undesirable) outputs. Halkos and Petrou (2019) provide a critical review of a list of DEA studies that incorporate bad output and Section 2.5, which focuses only on airline studies. Then, there are other DEA variants that reflect a network process whereby outputs produced in the first stage of production become inputs in the following stage of production. Such models were developed because the standard DEA treats the production process as a ‘black box’ and simply transforms inputs into outputs and neglects any possible intervening processes, including dissimilar series or parallel functions. This problem is overcome by using a network DEA (NDEA) model developed by Färe (1991) and Färe and Grosskopf (1996, 2000). For a list of bibliography on NDEA, we refer readers to Koronakos (2019) and Section 2.6, which focuses only on airline studies.

DEA has also evolved as an applied tool for consulting purposes or for policy analysis (Grosskopf, 1996). This involved a two-stage procedure whereby DEA efficiency scores estimated in the first stage are regressed against a host of independent variables (sometimes referred to as explanatory variables or environmental variables) in the second stage intended to explain inefficiency. The most common regression models used in DEA studies are ordinary least squares (OLS), Tobit and Simar and Wilson (2007) bootstrap truncated regression. Lovell et al. (1995) argued that OLS was inappropriate due to censoring issues, and this led to Tobit regression being widely used in DEA two-step analysis. Simar and Wilson (2007) found that many two-stage studies suffer from the problem of serial correlation and the lack of a coherent data-generating process. Hence, they extended this area of research by proposing a bootstrap truncated regression model based on the maximum likelihood