A network DEA aeronautical and non-aeronautical production model: an application to South Korea airports

There have been numerous attempts to capture the airport performance efficiency in various methods. Total factor productivity index (TFP), data envelopment analysis (DEA) and stochastic frontier analysis (SFA) were mainly used to measure the efficiency. Among nonparametric methods, DEA has been the most popular method.

In the former studies, these methods were adapted by numerous papers: Abbott and Wu (2002), Hooper and Hensher (1997), Oum et al. (2003), Parker (1999), Pels et al. (2001) and Yoshida and Fujimoto (2004). Following these papers, airport efficiency was measured through these methods: Abrate and Erbetta (2010), Adler et al. (2013), Barros and Sampaio (2004), Barros (2008a, b), Gitto and Mancuso (2012a, b), Malighetti et al. (2007), Merkert et al. (2012), Merkert and Mangia (2013) and Nicola et al. (2013). Moreover, there has been a number of researches on Asia-pacific region, measuring different aiport efficiencies in a different timeline (Fung et al. 2008a; Lam et al. 2009; Ha et al. 2010; Chow and Fung 2012; Fan et al. 2014; Li 2014). Oum et al. (2006) and Adler and Liebert (2014) adjusted all financial data according to the tool of Purchasing Power Parity (PPP) when collecting a sample from different Asia regions. Furthermore, after conducting DEA analysis, in recent studies, the usage of regression analysis has become a general method to determine the factors that affect the levels of efficiency in an airport company (Adler et al. 2013; Barros and Sampaio 2004; Chi-Lok and Zhang 2009; Coto-Millan et al. 2014; Ha et al. 2013; Merkert and Mangia 2014; Tsui et al. 2014). Especially, there has been a numerous attempts to explore the additional determinants of DEA estimates using censored regression from Tobin (1958). However, McDonald (2009) proposed using Panel Data regression model to find the influential factors of efficiency for the airport companies. Especially, Chang et al. (2013) found that the more airlines and destinations, the higher aeronautical efficiency, that is, the bigger the airport, the higher chance that the airports are aeronautically efficient. The above studies mainly focus on aeronautical efficiency, which place the number of passengers, cargos and aircraft movement as output variables and runway, terminal, operation cost as input variables. The input of runway was represented by number of runways or runway length or runway area; similarly, the inputs of terminal were expressed by cargo terminal size, passenger terminal size or total size (Chang et al. 2013; Chow and Fung 2012; Fung et al. 2008a, 2008b; Ha et al. 2013; Lam et al. 2009; Merkert and Assaf 2015; Tsui et al. 2014, etc). In addition, the operation cost usually was represented as soft cost input, as used by Oum et al. (2003), Oum and Yu (2004), Oum et al. (2007). Only Oum et al. (2003), Oum and Yu (2004), Oum et al. (2007) and Yu (2004) considered commercial services revenue and airport revenue to output variables. Above all, most studies considered an airport company as a whole by exploring the overall efficiency and productivity change with initial inputs and final outputs. This method treats an airport industry as a black box, neglecting intermediate products of linking activities in an airport company or failing to distinguish the sources from different sub-processes resulting in airport company efficiency (Tone and Tsutsui 2009; Tsui et al. 2014). This issue could not provide corresponding improvement strategies to managers.

In recent studies, Puls and Lentz (2018) indicated that airports in Europe improve their non-aeronautical revenue from travelers. And many articles focused on the non-aeronautical revenue for airport sustainability (Chuchu et al. 2018; Kidokoro and Zhang 2018; Fuerst and Gross 2018).

Except for the recent attempts to grapple with the black box issues in regard to DEA (Yu 2010; Lozano et al. 2013; Maghbouli et al. 2014), the recent studies were underpinned by exploring efficiency and productivity changes in accordance with initial inputs and final outputs. Airport operation was divided into two sub-processes, including aircraft movement process and aircraft loading movement process. Then, they, respectively, measured the efficiencies of each sub-process. They, however, used input and output variables, which were lopsided to aeronautical efficiency. Liu (2016) tried to capture commercial efficiency using NDEA and regression analysis. For the input in the first sub-process, runway area, staff costs and other operating costs were employed as other paper did (Adler et al. 2013; Coto-Millan et al. 2014; Fung et al. 2008a; Gillen and Lall 1997). In addition, as an intermediate product, aircraft movement was employed to link the two sub-processes. (Yu 2010; Abrate and Erbetta 2010; Lozano et al. 2013; Maghbouli et al. 2014; Pels et al. 2003) As the output, passengers and cargo and operating revenue were utilized. As the output of aeronautical business and other business includes the number of passengers and amount of cargo and the operating revenue reflects the results of business, these factors were included as the outputs. This is comparable to most studies (Ahn and Min 2014; Ha et al. 2013; Tsui et al. 2014; Yu et al. 2008). Then, influencing factors for the sub-process were measured using regression analysis.

Airport is a complex and dynamic place where not only does planes take off and land at, but also travelers enjoy shopping, relax while waiting for transfer or have a nice farewell meal with their family or beloved ones.

Airport production could be divided into two parts: non-aeronautical and aeronautical process. Aeronautical production includes any activities directly related to aircraft taking off and landing. Airport invests on quasi-fixed infrastructure such as runway and terminal capacity. Moreover, labor forces are required to manage the aircraft movement. These are the inputs for measuring aeronautical efficiency. Utilizing these inputs directly leads to aircraft movement. If inputs were efficiently utilized, the first process would turn out to be efficient. Aircraft movement results in aeronautical revenue that is generated from aircraft landing and taking off. In addition, cargos and passengers are the result of the intermediate input. Table 1 shows that how we are going to apply NDEA model to measure different efficiencies. Variables for input, intermediate and output are chosen carefully with the help of previous researches (Lozano et al. 2013;  Maghbouli et al. 2014; Wanke 2013; Yu 2010) and by considering the availability of data.

Non-aeronautical production per se is very complex due to the fact that airport is no longer just a place where people go travel by airplane. The non-aeronautical business, that is also known as commercial service usually consists of franchise-based operations and self-operations, such as ground handling agent service supplied for airliners, in-flight catering services, duty free and other retail shops in the terminals, leasing of advertising space inside and outside the terminals of the airport, provision of goods warehousing, cargo handling agent companies (Liu 2016; Adler and Liebert 2014; Oum et al. 2003, 2007; Oum and Yu 2004). The airport is rightly a place for airplane but the business expanded to non-aeronautical side due to the increase in air traffic globally. Studying financial statements for the airport and counseling with airport managers, among inputs for non-aeronautical business, top three contributing inputs were chosen: restaurant size, duty-free store size and parking lot capacity. Inefficiency occurs when aircraft movement is low given the non-aeronautical inputs. Aircraft movement is chosen as intermediate product as airports’ main focus is aircraft taking off and landing. From managerial perspective, decreasing the size of non-aeronautical inputs should be considered if aircraft movement is low; that is, not many people visit airport but excessive investment on building large airport should be reconsidered. For the second sub-process for the non-aeronautical production, if the small number of aircraft movement leads to higher non-aeronautical and the large number of passengers, it could be efficient. This could give us an insight that passengers visiting the airport might spend relatively more money than other airports or a number of visitors visit the airport or the schedule for the aircraft movement is tight. Overall, we want to see whether the inefficiency or efficiency for the sub-processes might have an impact upon the overall efficiency by calculating the overall efficiency.

The major disadvantage of DEA is that these processes are treated as ‘black box.’ NDEA, however, separates the whole box into smaller boxes which could give insights to decision makers by investigating the inefficiency or efficiency for the sub-process. This also provides which sub-process could generate inefficiency.

Table 2 shows the efficiency ranking of airports in South Korea using NDEA and DEA. And the (#) means the number of same ranks in the DMU group. Top ranking three airports are ICN, GMP and CJU. The results could be easily expected as the number of air traffic for these airports is high due to it is located at the hub of South Korea. The high aeronautical efficiency is inevitable since it is highly related to the traffic.

Table 3 shows airport ranking based upon non-aeronautical efficiency scores. Non-aeronautical efficiency gives an interesting result. Tracing the top three airports that show the high efficiency in aeronautical efficiency, rank 2 ICN in aeronautical efficiency is placed at rank 10. Looking at the sub-processes efficiency, it can be inferred that the first stage causes the whole inefficiency; that is, given the significant input of non-aeronautical, it does not generate sufficient aircraft movement. This indicates that ICN invests relatively a lot compared to the aircraft movement. However, the second stage of ICN is ranked 1, which means that ICN generates enough non-aeronautical revenue and passengers. However, the overall efficiency is still ranked 10, which is derived from inefficiency occurred in the first stage. The combined efficiency, which includes aero and non-aero efficiency, should be analyzed if this inefficiency might have a significant impact upon combined efficiency. If this inefficiency can be overcome with excellent performance of aero side of business, ICN will be placed top 3.

As the Table 4 shows, ICN is placed at rank 3. This is a bit disappointment as it is ranked 2 in aeronautical efficiency and ICN is a representative for South Korea airports. As discussed in non-aero parts, ICN rank does reflect upon the fact that the inefficiency caused in the first process of the non-aero sides can be overcome but not just enough to beat GMP. Still, it performs relatively well but through our research, inefficiency arises in the first stage of non-aero side of business.

TAE, however, is still placed at rank 12, which gives us similar results to non-aeronautical efficiency. This indicates that it is non-aeronautical side of business that TAE’s strategic plan should focus on.

Running 3 NDEA models on data provided us a number of different perspectives toward airport industry. As illustrated, separating production processes into aeronautical and non-aeronautical efficiency enabled us to look at the performance in various ways. In contrast, traditional DEA model notably shows the issues of ‘black box,’ which neglects the importance of sub-process. Lastly, combining the non-aero and aero provided the driving factors in terms of overall efficiency.