1. Introduction
Aviation can be considered essential within the ongoing process of (cultural, social and economic) globalization, also from a sustainability point of view [
1]. Actually, any increase in economic activity, industrial production or trade relations unavoidably causes greater need for transport [
2] and an increasing level of externalities, such as air pollution [
3,
4].
There are many studies that have focused on airport management efficiency, but they have seldom analyzed the correlation between regional airport sustainability and airport management efficiency.
Through studying the connection between the catchment area of Italian regional airports and their management quality, in this paper we try to ascertain when regional airports are economically unsustainable due to problems related to their catchment area dimension and when this happens because of management deficiencies. In order to obtain this result, we use the DEA method that appears particularly adequate because it allows the obtaining of scores measuring overall technical, pure technical, and scale efficiency.
In the European Union air transport system currently plays a fundamental economic role, particularly after that the three “deregulation packages” have become effective. Until that period, the aviation market was characterized by the supremacy of the domestic carriers, or “flag carriers”, and by bilateral agreements between nations. Over the years there has been a growing need of strengthening the market mechanism in the air transport sector, since up to that moment it was regarded as overregulated and with a high incidence of public monopolies. Transport deregulation and privatization, traditionally, has been thought of as something that should lower the level of pollution as a consequence of propensity to innovate of private firms [
5,
6] and of an improvement in fuel efficiency and load factors [
7].
Moreover, the recent wave of deregulation and liberalization has determined an increase of airline competition, which translated into higher competition among airports; therefore, it appears important to assess the operational performance of airports, especially their efficiency. As a matter of fact, airport benchmarking might be useful for comparing the various structures and gauge their level of efficiency, with the additional aim of driving them toward the “best practices”.
The deregulation process in Europe has followed four steps [
8,
9,
10], which have led to a unique domestic market throughout the countries. The last step has taken place in 2008 when, due to the Regulation (EC) No. 1008/2008, which reviewed the regulatory rules of 1992, European Institutions have accomplished the process of liberalization, so that EU routes are now “freely” reserved to the community air carriers. Actually, henceforward European countries are compelled to accept the entry of all “authorized” carriers into their airspace.
This situation has clearly led to phenomena such as increased competition among carriers, decreased average fares, increased frequency, and new route services [
11,
12]. Airlines have become more “footloose”, having a greater freedom to choose where they fly to and from, and in general to set fares, frequencies, capacities and routes according to commercial consideration [
12]. This has been quite important for providing opportunities for airports to grow through attracting new routes but also challenging the existing ones (actually, around 2500 new routes were opened in 2011). Moreover, air transport produces relevant sustainability issues in both economic and environmental terms, even if our analysis is focused on the economics aspects and more specifically on the positive effects produced by the European policy of transports privatization. It is widely recognized in economic literature that an efficient management of regional airports produces positive effects on the environment both for the congestion reduction in the bigger airports and for a better use of existing infrastructures reducing soil consumption [
13,
14]. One of the main results of the deregulation process has been the entrance and the development of low-cost carriers (LCCs). These new players have greatly stimulated a part of demand, which was “neglected” by bigger carriers. Highly price sensitive customers are willing to receive a low profile service (‘no frills’) so, thanks to an aggressive pricing policy, LCCs have attracted millions of passengers. They have also developed a network called “point-to-point”, which indicates a connection of pairs of destinations with a high frequency gain. This strategy, by maximizing the number of passengers boarded per way, leads to reduced congestion and increased environmental performance of the whole system [
15]. By doing so, LCCs have focused especially on secondary airports, due to the lower level of charges, increasing the chance for competition among airports. It is important to notice that LCCs played a key role in the aviation market because they have changed the traditional business relationship between airports and airlines. Particularly, the capability of LCCs to move high volumes of passengers has created an asymmetry between airlines and airports, with more market power in the hands of the airlines [
16,
17].
In Europe airport industry was traditionally characterized by public ownership and national requirements [
18]. As soon as domestic air transport market was liberalized, a number of governments in Europe began to transfer the ownership or operation of larger airports to the private sector. Many smaller airports in Europe are still publicly owned, but the majority is now operated by corporatized entities. From a theoretical perspective, airport privatizations improve the efficient use of infrastructures, enhancing long-term sustainability [
19].
For decades, major airports around the world have predominantly served passenger markets, thus their operations and infrastructure were designed primarily to meet the needs of people, with their remaining capacity serving air cargo. Such airports are also referred to as “gateway airports” [
20]. This phenomenon can be attributed to the fact that the volume of air cargo is not sufficiently large to reach a critical mass. To a great extent, air cargo plays a complementary role for passengers, filling the excess capacity of aircraft.
However, the recent growth of global air cargo traffic has instilled great concern in policy makers and airport planners. This feature is interesting also when examining sustainability and efficiency of European airports, and the Italian ones among them. Particularly, we now aim at evaluating the impact of relevant external factors, such as the size of the airport and the presence of LCCs, on airports’ efficiency and environmental sustainability [
21,
22].
In the recent years, privatization and restructuring processes have affected also the Italian airport industry, with likely spillovers on their overall efficiency. Within this dynamic context, in our opinion it is worthwhile to assess whether local airport characteristics have also had a role in influencing efficiency. Actually, the latter is likely to depend not only on exogenous features on which airport management has a limited direct control (unless they act on local public policy makers and/or vectors in order to attract new demand), but also on factors that can be directly handled by local managers, particularly those pertaining the size and the business mix.
Unlike previous papers regarding Italian airports, we consider the effect of cargo traffic on airports’ efficiency and sustainability, which is expected to exhibit a higher factor productivity and a lower environmental impact because “handling cargo is capital intensive and therefore more productive than handling passengers” [
23]. However, we take into account also the impact of other important external factors, particularly the size of airport and the presence of LCCs.
The producers performances are often affected by external or environmental factors which may affect the production process—being responsible for differences in the performances of the data management units (DMUs)—but, unlike the inputs and the outputs, they are not under the control of production units: we refer mainly to quality indicators, regulatory constraints, market conduct (competitive vs. monopolistic), type of ownership (private-public or domestic-foreign), environmental features. Such factors can be however included in a model as exogenous variables, thus helping to explain the efficiency differentials and even to improve policymaking.
The paper is structured as follows. After introduction, in
Section 2 we give a concise review of literature dealing with both the measurement of airport efficiency and the assessment of its determinants. In
Section 3 we present the first stage of our procedure and, by means of the DEA (Data Envelopment Analysis) method, estimate and examine the scores measuring overall technical, pure technical, and scale efficiency of Italian airports over the period 2006–2016. In
Section 4, we make use of a Tobit model and regress such efficiency scores on three explanatory variables for the same airports: airport size, the share of LCCs passenger, and the share of cargo traffic, also providing some discussion on the results.
3. First Step: Estimating Efficiency Scores through DEA Analysis
3.1. The Model
The economic theory underlying efficiency analysis is based on Debreu (1951) [
60] and Farrell (1957) [
41], who made the first efforts on measuring the efficiencies of a set of observed production units. Within this context, the DEA original model, introduced by Charnes et al. (1978) [
44], represents an improvement on those seminal works.
The Data Envelopment Analysis is also the basic method that has been used in order to assess the performance of transportation infrastructure management. Originally, the development of DEA aimed at solving problems that were hard to deal with other approaches. This difficulty was due to the complex (frequently unknown) nature of the relations among the multiple inputs and outputs involved in the activities [
43].
In DEA, the basic premise is homogeneity, that is, the DMUs must perform similar activities and produce comparable products and/or similar services, so that it can be set as a common range of products [
61].
Each DMU’s score is individually optimized through mono-objective linear programming, comparing the resources used (inputs) and the quantities produced (outputs) to the levels of other units. The result is the construction of an efficient frontier. The DMUs lying on it are efficient (score of 100%) while the others are inefficient (score of less than 100%).
Besides efficiency scores, the envelope formulation of DEA models provides targets and a reference set for the inefficient DMUs. The targets are the levels that the inputs and outputs of those inefficient units must achieve in order to be efficient. The reference set represents the efficient DMUs (benchmarks), i.e., those that will be used as references for good management practices. A linear combination of these benchmarks provides the targets for each inefficient DMU. Such targets are, in most cases, virtual, as they do not characterize a real efficient DMU.
Depending on the industry characteristics, there are different DEA models: input-oriented, output-oriented, or both. The input orientation focuses on proportional decrease of the input vector; the output orientation adjusts the proportional increase of the output vector; the output/input orientation does not discriminate the importance of possible increase of output or decrease of input.
In the air transport sector, the output-oriented model has been considered as more suitable, due to the fact that it is not possible to recover investments in infrastructure that normally are made well in advance. On the contrary, the goal of the manager is to expand the demand as much as possible and to use airport facilities as intensively as possible, since production factors are fixed or semi-fixed. In terms of returns to scale model, there are three basic DEA models: constant returns to scale (CRS), variable returns to scale (VRS), and additive. They can be used to assess which of DMUs determine the frontier of the envelopment surface. Units that do not lie on the frontier are inefficient, and the measurement of the grade of inefficiency is determined by the selection of the model.
Our paper focuses on both constant and variable returns to scale models, while it makes use of an output-oriented approach in order to analyze the financial and operating performance of Italian airports. Particularly, for the
jth airport (out of
n airports) the output-oriented technical efficiency under constant return to scale (
CRS) is obtained by solving the following linear programming problem:
where:
X and
Y are the input and output vectors, respectively;
θCRS = 1/
θjCRS is the technical efficiency of airport
j under
CRS; and
λ is a
n × 1 vector of weights. The non-negative weights
λ measure the contribution of the selected efficient airports to define a point of reference for the inefficient
jth airport. In general, 0 ≤
θjCRS ≤ 1, where
θjCRS = 1 if the airport is producing on the (production) frontier and hence is technically efficient. When
θjCRS < 1, the airport is technically inefficient.
For the case of variable returns to scale, technical efficiency θjVRS is obtained by adding the convexity constraint Σλj = 1 into the maximization problem.
3.2. Data
We employ data for a cross section of Italian airports that differ in ownership, financing and operational characteristics. In particular, our sample considers all 34 Italian airports certified by ENAC (the Italian Civil Aviation Authority) for the period 2006–2016. Such time interval considers the very last deregulation steps linked to the Regulation (EC) No. 1008/2008, but it especially takes into account the revitalization of air traffic occurred after 2011 (i.e., after the attack on the Twin Towers).
Table 1 portrays some characteristics of Italian airports for 2016.
Some interesting insights can be drawn from its data. First, for the workload units (WLU) and the total number of passengers the standard deviation is higher than the mean, which indicates that the sample is not very homogeneous. This is largely due to the presence of both small and large airports in the sample, a characteristic that reflects the local population density. Second, the lowest ratios of the rate of turnover of invested capital are found for small airports, most of which are managed by a capital company totally in public hands. The highest values belong to the airports of Genova (where the rate of change is greater than 1), Pantelleria, Alghero, Lamezia Terme and Trieste. Such airports are mostly managed by partially privatized company, which foresees that privatized or partially privatized airports achieve higher levels of efficiency. Third, none of the listed airports is completely privatized: the majority is characterized by a joint management system (public-private), whereas the remaining airports, mostly small-sized, are totally public managed.
To measure airport productivity through DEA, we need to identify both the outputs that an airport produce and the inputs used in producing those outputs. In statistical analysis, the most common airport output measure used is the number of passengers served, as most airports serve mainly passenger traffic, so in our study we will follow the same line. Air cargo, however, is becoming more and more important for many airports. Therefore, we consider it as a separate additional output.
While passengers and cargo handling are usually considered as the outputs of airport landside operations (also considered as final outputs of an airport), aircraft movements are regarded as an output of airside operations generating revenues for airports in the form of landing and aircraft parking charges, although they can be also seen as an intermediate outputs (meaning that they carry passengers and cargo that generate additional revenues in airports’ landside operations). They represent therefore our third output.
Moreover, airports revenues arise also from concessions, car parking, and many other services that are not directly related to aeronautical activities in a traditional sense, but nonetheless are becoming increasingly more important for airports around the world. Thus, additional outputs in our analysis are the revenues from aeronautical, handling and commercial services. From the above, it is clear that the airport industry emerges as a paradigmatic case of joint production [
62,
63].
As for inputs, we consider three categories: labor (measured by the number of employees who work directly for an airport operator), capital (consisting of the various infrastructure and facilities, and expressed by the book assets value), and all expenses not directly related to capital and personnel (which allow us to take into account the effects of airports’ operation strategies with respects to outsourcing activities on production).
Summing up, for each airport we have information on six output variables: the number of passengers movements (APM), the ton of cargo (CAR), the yearly number of aircraft movements (AAM), revenues from aeronautical activities (AR), revenues from handling services (HR), and revenues from commercial activities (CR). Input variables are: labor costs (LC), invested capital (IC), and other expenses (
OC) (
Table 2).
Our choice regarding inputs and outputs meets various DEA requirements. Actually, in the DEA methodology there exists a direct correlation between the number of variables used (inputs and outputs) and the number of observations considered “efficient.” As Seiford and Thrall (1990) [
64] show, a low ratio of observations to the number of inputs and outputs weakens the discriminatory power of DEA models, and most DMUs could be rated efficient. Liebert and Niemeier (2013) [
65] suggest that this should not be considered a flaw of the methodology but rather a direct result of the dimensionality of the input/output space (
m inputs +
s outputs), relative to the number of observations (
n). In this case, too few inputs and outputs would reduce the capacity of the DEA to select the efficient airports.
With respect of the minimum set of data points in the evaluation set, in the DEA literature there are a couple of guidelines. A first rule states that the sample size should be greater than the product of the number of inputs and outputs, while a second rule declares that the number of observation in the data set should be at least three times the sum of the number of input and output variables. However, the size of our sample exceeds the desirable size as suggested by these rules, so we trust we are able to obtain sufficient discriminatory power.
As before, it comes out that the Italian airports are relatively heterogeneous, being the standard deviation higher than the average for all the considered variables.
3.3. Empirical Results
The results were obtained using the Open Source DEA software for both the CRS (constant return to scale) and VRS (variable return to scale) models and following the output-oriented approach (i.e., we assume that airports aim to maximize the profits resulting from their activity, and try to determine whether an airport is able to produce the same level of output with fewer inputs).
It should be noted that the DEA technique provides a “relative” efficiency measure, therefore its value refers only to the context under investigation. Changing the characteristics of the sample (e.g., increasing the number of units) or varying the model (type of returns to scale, orientation) generate different “efficient” units, or different efficiency values. The relative performance of an airport is defined as the ratio of the weighted sum of its outputs to the weighted sum of its inputs. The weights are not predetermined, but rather allocated by the model, avoiding bias resulting from subjectively assigned weights.
By using the CRS model via an output-oriented approach, it is possible to obtain each unit’s technical efficiency score that represent its capacity to produce a certain amount of output, given a set of input.
For each airport,
Table 3 shows the values of the inputs and outputs used for the DEA analysis; they are averaged over the 2006–2016 period.
Table 4 reports the efficiency results obtained by the application of output-oriented
CRS model to the 34 Italian airports. Values still refer to the 2006–2016 period.
As told, an efficiency value less than 1 indicates that the unit is inefficient. In our case we find 16 inefficient airports out of 34: Aosta, Bologna, Bolzano, Comiso, Florence/Pisa, Olbia, Palermo, Parma, Pescara, Reggio Calabria, Salerno, Turin, Verona/Brescia. Efficient units (for which the estimated value of efficiency is equal to 1) are: Alghero, Ancona, Bari/Brindisi/Foggia/Taranto, Bergamo, Cagliari, Catania, Cuneo, Elba, Genoa, Grosseto, Lamezia Terme, Lampedusa, Milan Linate/Milan Malpensa, Naples, Pantelleria, Perugia, Rome Ciampino/Rome Fiumicino, Trapani, Treviso, Trieste, Venice.
Table 5 shows the performance indicators’ breakdown. The output-oriented
CRS model allows measurement of the overall efficiency of each DMU (both technical efficiency and scale efficiency: third column, corresponding to the last column of
Table 4), while the pure technical efficiency is measured through a VRS output-oriented model (fourth column). Hence, the efficient DMUs (which assume value 1) are fewer than those obtained taking into account the same set with the VRS model. The latter values reveal that technical efficiency (or managerial efficiency) is not influenced by the firms’ size. As already seen, the efficiency of scale value (fifth column) is obtained through the ratio between the overall technical efficiency (
CRS) and pure technical efficiency (VRS). Finally, the sixth column shows the type of returns to scale (constant, increasing or decreasing:
CRS, IRS or DRS, respectively) exhibited by each DMU.
The efficiency results help to discover some interesting features of the Italian airport system. All technically efficient airports under the assumption of constant returns to scale (16 over 34, see above) are technically efficient even under the assumption of variable returns to scale (VRS), which means that the dominant source of efficiency is the efficiency of scale. Looking at the VRS efficiency scores, i.e., those concerning pure technical efficiency due to management skills, 24 airports out of 34 (70.6%) are efficient over the time period here considered. Therefore, In Italy only 10 airports appear inefficient from an operational point of view, which makes us conclude that overall Italians airports are well managed with regard to the pure technical efficiency.
Considering the scale efficiency, 26 of 34 units analyzed reach the value of 1. As the efficiency score has to be considered as an average value during the period, it suggests that approximately 71% of Italian airports reveal pure technical efficiency, but some of them (4 units, 16.7%) do not exhibit efficiency of scale.
Finally, 4 airports exhibit increasing returns to scale (they are either small- and medium-sized), while 10 have decreasing returns to scale (many of them are small-sized). The units characterized by increasing returns to scale (IRS) may improve efficiency by increasing the productive dimension; conversely, units characterized by decreasing returns to scale (DRS) could gain in efficiency just by reducing the size of production. For smaller airport, this could call for a reconsideration of their activity. The units characterized by constant returns to scale (CRS) work, instead, in optimal production conditions and the size of their pure technical efficiency is equal to one, given the concurrence of overall efficiency and scale efficiency.
4. Second Step: Assessing the Efficiency Drivers through Tobit Analysis
4.1. The Model
The Tobit model is also called censored regression model, because the latent variable cannot always be observed while the independent variable is observable. It is an alternative to OLS regression, and is employed when the dependent variable is bounded from below or above or both (as it happens when dealing with efficiency scores, normally lying between 0 and 1). Particularly, we make use of a random effect Tobit model, which assumes that the unobservable effects are uncorrelated with the observed exploratory variables (on the contrary, a fixed effect model assumes that they are correlated).
The model can be expressed in a general way as follows:
with:
In the above, N is the number of observation, Yi* is an unobserved latent variable, Yi is the dependent variable (DEA scores), Xi is a vector of independent variables, β is a vector of unknown coefficients to be estimated, and εi is an independently distributed error term, assumed to be normal with zero mean and constant variance σ2.
As said, we employ the Tobit regression in order to identify the factors that are more significant in influencing airports efficiency. The importance of our analysis stays in providing information to airports managers about the major determinants requiring more attention in the course of business activities.
We again consider the 34 Italian airports, and employ the first-stage DEA efficiency scores (technical efficiency, overall efficiency, and scale efficiency) as the dependent variable, while the explanatory variables are a number of other important factors that may affect the production efficiency: the number of Work Load Units, the percentage of passengers handled by low-cost carriers, and the percentage of cargo traffic. In our model, we suppose that estimated coefficients are independent of time, so that, the greater the value of efficiency, the more efficient the airport. In case the regression coefficient comes out to be positive, we deduce that they positively affect efficiency.
Our explanatory variables (airport size, the share of LCC passengers, and the share of cargo traffic) have been selected among those that have been widely used within the international literature on the topic. Hence, we estimate the following relationship:
Here, Yi are the efficiency scores (overall technical, pure technical, and scale efficiency scores, respectively), while the three environmental variables (measured for each airport) are: the number of Work Load Units, or WLU (SIZE); the percentage of passengers handled by low-cost carriers (LCC); and the percentage of cargo traffic relative to total WLUs (CARGO).
4.2. Empirical Results
In
Table 6 we show the empirical results of our three Tobit estimations for the sample of 34 Italian airports. Particularly, we report the marginal effect of each regressor on the dependent variables. The Likelihood Ratio test indicates that in all of them the variables included in the model have a statistically significant effect on the dependent variables. All parameters exhibit a significant sign (at least at the 5% level), and their sign always reveals a positive influence of the corresponding explanatory variable on the dependent (efficiency) variable.
The coefficient of
SIZE (airport size) indicates that larger airports are expected to have higher overall technical efficiency, pure technical efficiency, and scale efficiency scores compared to smaller airports. This evidence is in line with previous similar research. For example, Perelman and Serebrisky (2012) [
66] perform a DEA analysis for Latin American airports and conclude that larger airports have higher technical efficiency than smaller ones. The same result characterizes the analyses by Pels et al. (2003) [
39], Martín et al. (2009) [
67], and Coto-Millan et al. (2014) [
63].
The presence of LCCs is also positively associated with all types of efficiency. Actually, the introduction of LCCs has generated a substantial growth of demand in the Italian airports, especially in small- and medium-sized ones. This is also what Cavaignac and Petiot (2017) [
68] find in their survey on papers applying DEA to the transport sectors. Moreover, our results are in line with those obtained for British airports by Bottasso et al. (2013) [
69], who conclude that LCCs’ entry on European markets has stimulated airports productivity improvements, which in turn has positively affected total factor productivity. Likewise, Coto-Millan et al. (2014) [
63] find that the share of LCC passengers has had a positive effect on the efficiency of Spanish airports.
Finally, the cargo traffic (
CARGO) positively impacts all types of efficiency as well. Hence, airports with a higher proportion of cargo traffic are again able to get higher levels of efficiency [
23]. Once more, this confirms the evidence of previous works, for which airports with a large proportion of cargo traffic are expected to disclose a higher variable factor productivity, since handling cargo is capital intensive and therefore more productive than handling passengers.
5. Conclusions
The European policies on air transport liberalization have had positive effect on environment inasmuch as they have improved the natural resources use. Notwithstanding, management systems designed to achieve the maximum efficiency of regional airports do not seem to pay particular attention to the issue of the economic and environmental sustainability.
In the air transport related literature little attention has been devoted to the aspects involved in the sustainability issue of peripheral airports and this might seem surprising given the relevance of the problem.
In our opinion, a viable path for exploring the influence of exogenous factors in airports’ efficiency is to employ a two-stage procedure. Specifically, from a pure efficiency perspective, we estimate technical efficiency of Italian airports over a group of variables that may affect it.
This paper, starting from general considerations about the literature concerning DEA studies in the air transport, focuses attention on the overall technical, pure technical, and scale efficiency of Italian airports. Analyzing several factors and their impact on efficiency level, we investigate the economic and environmental performance of 34 Italian airports in the period 2006–2016. In order to provide a clear-cut answer, we use the DEA method that appears particularly adequate because it allows us to get scores measuring overall technical, pure technical, and scale efficiency. Such analysis highlights some interesting features of the Italian airport system. First, scale economies affect the efficiency outcome in a significant way, so that, in presence of increasing returns to scale, the airport managers should aim at extending the catchment area of the airport or developing strategies that are able to attract new carriers and new business activities, such as cargo activity. Second, the majority of Italian airports are well managed with regard to the pure technical efficiency. In a second step, we have applied a Tobit regression model, with the aim of investigating whether airport size, LCCs presence and cargo traffic have a significant influence on technical and scale efficiency of Italian airports. The results suggest a positive association between overall technical efficiency and these variables, confirming the key role of market in achieving a more sustainable and efficient airport economy. In other words, a regional airport can become economically sustainable if the airport managing authority acts on the market so as to improve both cargo demand and LCCs demand. This confirms the theoretical hypothesis according to which regional airports are economically sustainable if there is an intensive use of the infrastructure, otherwise they would become a problem—and not a resource—for its territory [
70].
The analysis of airport performances is carried out following the idea that political determinants are very influential on management system.
Moreover, our findings seem of major interest not only for airport managers and operators, but also for policy makers and regulators. Actually, they shed light on how airports’ performance and environmental sustainability could be improved as well as on the influence that airport size and traffic distribution between passengers and cargo can exert on airport efficiency.
Finally, in the light of the analysis’ findings reported in this paper, we conclude that air transport privatization and deregulation is able to positively affect the transport environmental efficiency and sustainability. This is to say that our results are helpful for better understanding the role of the market mechanism in achieving transport sustainability. However, the topic needs to be further and carefully investigated as regulation, traditionally, is considered crucial in terms of consumer benefits and environmental performance. Therefore, the empirical analysis of the effects of privatization and deregulation is recommended for evaluating such political programs.