Is increasing aircraft size common practice of airlines at congested airports?

Journal of Air Transport Management xxx (2015) 1e9

Contents lists available at ScienceDirect

Journal of Air Transport Management journal homepage: www.elsevier.com/locate/jairtraman

Is increasing aircraft size common practice of airlines at congested airports? Peter Berster, Marc C. Gelhausen*, Dieter Wilken German Aerospace Center (DLR), Institute of Air Transport and Airport Research, Linder Hoehe, 51147 Cologne, Germany

a r t i c l e i n f o

a b s t r a c t

Article history: Available online xxx

If the overall demand for air transport grows, but additional airport capacity is not available at congested airports, we could assume that airlines will offer flights with more seats in order to cope with the demand. An analysis of frequency and average seat capacity developments at congested, and not yet congested airports, has shown that the hypothesis of bigger aircraft being used in congested situations is valid in most instances, although not at all airports. The objective of this paper is to report on an analysis of the development of average seat capacity at congested airports, in contrast to the situation at not yet congested airports, and to find out the reasons for airlines increasing the number of seats at congested airports, by means of a statistical model using variables including the degree of airport congestion and average flight distance. © 2015 Elsevier Ltd. All rights reserved.

Keywords: Airport capacity Airport utilisation Airport congestion Average seat capacity Causal factors of average seat capacity development

1. Introduction Capacity constraints analysis of the global airport system has shown that in 2008 only a small number of airports were congested, including important airports such as London Heathrow, New York La Guardia and Paris Charles de Gaulle. Overall, 10 airports were identified as highly capacity critical airports (Gelhausen et al., 2013): -

San Diego (SAN), Shanghai (SHA), Shenzhen (SZX), London Heathrow (LHR), Mexico City (MEX), New York La Guardia (LGA), Barcelona (BCN), Charlotte (CLT), Frankfurt/Main (FRA), and Paris Charles de Gaulle (CDG).

* Corresponding author. E-mail address: [email protected] (M.C. Gelhausen). 1 Capacity utilisation index is defined by the ratio of average daytime hourly flight volume and the 5% peak hour volume in 2010 for each airport.

These airports have high peak-hour traffic volumes and values of the so-called capacity utilisation index (CUI) of more than70.1 These 10 airports handle about 6% of all flights worldwide, which indicates that the great majority of flights operate under unconstrained conditions. According to the market forecasts of the aircraft manufacturers and international organizations such as ICAO, we have to assume that the number of flights will increase in the future, although probably not as much as passenger traffic measured in revenue passenger kilometres (RPK). However, given a long-term growth rate of around 5% for traffic development (Airbus, 2010; Boeing, 2012; Teyssier, 2010), a growth rate of 3% for the flight volume does not seem implausible (ICAO, 2005; ICAO, 2012; Eurocontrol, 2008). Therefore, the number of flights will grow by about 30e40% in 10 years. Airport capacity analysis has demonstrated that traffic conditions at airports, which were still favourable in 2008, will soon deteriorate since many more airports will suffer from bottleneck situations. The majority of flights in the airport network worldwide will be affected by capacity constraints. The impact of capacity constraints on flight activities can be mitigated by capacity enhancing measures, such as new runways, or demand management measures. Investment options, such as new infrastructure, are increasingly subject to public opposition, especially in Europe, since the residents in the vicinity of airports are against higher levels of noise pollution due to increased aircraft movements. Interconnecting high-speed and regional

http://dx.doi.org/10.1016/j.jairtraman.2015.03.012 0969-6997/© 2015 Elsevier Ltd. All rights reserved.

Please cite this article in press as: Berster, P., et al., Is increasing aircraft size common practice of airlines at congested airports?, Journal of Air Transport Management (2015), http://dx.doi.org/10.1016/j.jairtraman.2015.03.012

2

P. Berster et al. / Journal of Air Transport Management xxx (2015) 1e9

trains with airports is another investment option that relieves the pressure on short distance flights. However, this solution is also becoming more opposed by the affected population. In the meantime, demand management is gaining more acceptance as an alternative or a complementary measure to capacity expansion and enhancement through reorganizing traffic patterns. The following provides a brief discussion of some potential demand management approaches: - Pricing schemes, for example, peak-period pricing or congestion pricing. However, such pricing schemes are often difficult to realise effectively due to price regulations imposed on the airports (e.g., Charlton, 2009). - Using off-peak times more intensively: at hub airports this is only feasible to a small degree, since inbound and outbound flights are generally closely coordinated (e.g., Button, 2002). - Based on technological progress, changing air traffic control (ATC) rules in order to augment the throughput of aircraft movements: this measure will raise capacity probably to a degree which corresponds to the traffic growth of just a few years. - Substitution of short-distance air travel by high-speed trains: this measure is only effective for short-haul routes between areas of dense population like Europe and is of limited potential to mitigate congestion at airports. - Diverting traffic to less congested airports: is often contradictory to the demands of passengers and hub operations (e.g., Dennis, 1994). - Using aircraft with a higher seat capacity: this may lead to additional congestion at terminals or on the apron. Strengthening the high-speed network between large urban agglomerations is a measure used to ease the pressure on airports and reduce the level of unaccommodated demand at congested hub airports. The continued improvement of the high-speed train network in Europe will lead to a reduction in demand for air travel, and consequently the number of flights by 0.6% in Europe by 2035, depending on the efforts taken by the member countries (Eurocontrol, 2013). High-speed trains have had significant impacts for some city-pairs. For example, with the opening of the ICE train service between Cologne main station and Frankfurt airport (a distance of 177 km, travelled in 70 min), air services between Cologne and Frankfurt airport were discontinued. Since the opening of the Madrid-Barcelona high-speed service in 2008, 40% of the traffic between those two cities was captured by the high-speed train service (European Commission, 2010). Nevertheless, as a measure to mitigate airport congestion, high-speed trains are only of limited value, typically because it is limited to short- to medium-distance continental travel. Indeed, high-speed train access at an airport tends to reduce short-haul feeder flights, but also increases the catchment area of an airport at the trip origin or destination, respectively (Gelhausen et al., 2008). Diverting traffic to less congested airports and a “division of work” between various airports in a metropolitan area is not optimal from the view of hub operations (Dennis, 1994), but this measure is employed quite frequently. In many cases, traffic distribution among the airports in a particular metropolitan area is historically set, such as the cases of Paris Orly and Charles de Gaulle in Paris or Haneda and Narita in Tokyo. Other examples include Frankfurt and Munich, London Heathrow and Gatwick, New York JFK and La Guardia, as well as the airports of Rio de Janeiro and Sao Paulo. Often, the “newer” airport in the area has a longer runway system and focuses on long-haul international traffic, whereas the “older” airport handles more short-haul domestic/continental traffic. This is reflected by the average stage length of flights to/from these airports. For example, average stage length of flights is

3010 km at London Heathrow and 1668 km at London Gatwick (OAG, 2012). Of course, the extent of differences in stage length varies, but is consistent throughout. Congestion is typically quite high at both types of airports, but not necessarily at the same level. For example, congestion at New York La Guardia is considerably higher than at New York JFK (CUI value of 0.76 at LGA vs. 0.67 at JFK). Diverting traffic to secondary airports becomes less attractive if the secondary airport is too far away from the origins of demand, as the examples of London Stansted and London Luton show: these airports are mainly served by low-cost and charter carriers but not by British Airways. Despite the distance between the airports, Frankfurt and Munich are a different story, as both airports have a strong catchment area. Nevertheless, Lufthansa has started to relocate international flights from Munich back to Frankfurt, after the opening of the fourth runway at Frankfurt airport. Using bigger aircraft and aircraft with a higher seat density are measures that airlines use to varying degrees depending on factors such as the level of airport congestion, fleet, network structure, competition with other airlines, etc. This paper focuses on this measure and reports on a statistical analysis regarding the development of average aircraft size, i.e., the number of seats per flight at constrained and unconstrained airports worldwide, as well as in world regions. A model has been developed that relates average aircraft seat capacity to causal factors, such as the degree of airport congestion and average flight distance. The results indicate that airport capacity is primarily constrained by the airside, especially the runway system, and not the landside, i.e. the terminal ground access, etc. This stems from the fact that, in many cases, the most critical element of airport capacity is the runway system, since airport expansion planning often requires the involvement of the public, who are most likely to oppose to new runways to protect their neighbourhood against increased noise emissions (Wilken et al., 2011). The paper is organised as follows: Chapter 2 describes the methodology for the selection of constrained and unconstrained airports to be included in the analysis. Chapter 3 analyses the development of average seat capacity at capacity constrained airports versus unconstrained airports from a global perspective and from the point of view of different world regions. Chapter 4 takes a look at airport-specific developments of average seat capacity. Chapter 5 addresses factors that lead to increasing average seat capacity per flight, and an econometric model of average seat capacity, along with some sensitivity analyses follow in Chapter 6. The paper finally concludes with a brief summary and some conclusions. 2. Selection of sample constrained and unconstrained airports A working hypothesis at the outset of the analysis was: airlines that want to serve a growing market increase their capacity by offering more seats on existing routes, as well as new routes (ICAO, 2008). At congested airports, airlines would do so by deploying bigger aircraft and at uncongested airports they would first increase the number of flights. Capacity constraints would hinder airlines from increasing frequencies, whereas at airports with a capacity surplus, airlines would prefer to offer more flights in order to better comply with the needs of travellers, in particular, the needs of business travellers. Our analysis of the development of the average seat capacity of flights offered should then differentiate between congested and uncongested airports; however, it would not necessarily have to include all airports worldwide. The global air traffic network consists of several thousand airports, most of which handle only small

Please cite this article in press as: Berster, P., et al., Is increasing aircraft size common practice of airlines at congested airports?, Journal of Air Transport Management (2015), http://dx.doi.org/10.1016/j.jairtraman.2015.03.012

P. Berster et al. / Journal of Air Transport Management xxx (2015) 1e9

numbers of aircraft movements. An analysis of traffic distribution in the global network of around 2400 airports, which are part of the international scheduled air traffic, has shown that traffic is concentrated on a relatively small number of airports (Gelhausen et al., 2013). In 2008, just 100 airports (corresponding to 4% of all airports) handled 50% of total number of flights, and the top 1000 airports (41%) handled 95% of all flights. Conversely, there were 1400 airports with volumes so low as to account for just 5% of the total volume. For the analysis of the seat capacity of flights, we have selected the airports that reached the threshold of 70,000 air transport movements (ATMs) in 2010. This threshold was decided by the fact that single runway airports with 70,000 ATMs have a 5% peak hour volume of around 20 ATMs, which corresponds to about 50% of the hourly capacity of a runway under IFR conditions. Airports below this threshold are considered as airports of regional importance without any capacity problems in the near future. In 2010 there were 178 airports worldwide with traffic volumes exceeding 70,000 ATMs, handling about two thirds of the total flight volume. Based on OAG data, we have calculated the average daytime hour volume and the 5% peak hour volume in 2010 for each airport. The ratio of these two volumes is defined as the capacity utilisation index (CUI) (Reichmuth et al., 2011). Following the same approach as in Gelhausen et al. (2013), the constrained airports in the sample of 178 airports were identified based on CUI and the 5% peak hour volume. A high traffic volume in the 5% peak hour indicates congestion in peak times of the day; a high CUI value indicates that congestion also occurs during normal traffic hours of the day. If the values of these indicators exceed certain threshold values, then the airport can be regarded as an airport with congestion problems over longer operating hours of the year, the duration of which depends on the value of the thresholds. In terms of the 5% peak hour volume, the threshold is dependent on the capacity class of the airport (single runway, two parallel runways, etc.). According to Bubalo and Daduna (2011), ultimate capacity is determined mainly by runway configuration, even if traffic mix and demand patterns differ. In this paper, capacity classes are inevitably defined broadly for airports with more than two runways, to achieve adequate sample sizes per capacity class. Nevertheless, CUI is considered as the main criterion for identifying congested airports, as congestion during normal traffic hours of the day typically implies congestion in peak times of the day. The definition of capacity classes is not critical, as the 5% peak hour volumes are secondary indicators that primarily serve to eliminate airports with high CUI values but very low capacity utilisation. This is typically the case if the average daytime hour volume and 5% peak hour volume are close together, but the 5% peak hour volume is far below the hourly runway capacity. This often happens at smaller regional airports. In brief, the CUI is the average slope of a ranking curve between 5% peak hour volume and average daytime hour volume. Research has shown that the CUI and the average slope of the ranking curve, respectively, are reasonable indicators of airport congestion that are comparable between different airports (Reichmuth et al., 2011; Wilken et al., 2011). The chosen CUI approach mitigates the problem that airports of especially higher capacity classes are usually very different with regard to runway configuration, traffic mix and demand patterns. Airport congestion is not a clear-cut phenomenon; the term encompasses the spectrum between constrained flow conditions at some peak hours to dense traffic conditions with high delays for each aircraft over longer periods of time. For the purposes of differentiating airports with and without capacity problems in this analysis, we have defined as threshold values a capacity utilisation of around 75% in the 5% peak hour time and a CUI of 65%. In the case of a single runway airport, this means that the airport is regarded as

3

constrained if the 5% peak hour volume exceeds 30 ATMs and the normal day hour utilisation exceeds 65% of the 5% peak hour volume, assuming that the 5% peak hour volume of 30 or more ATMs reflects a near capacity volume. As a result, airport congestion is purely defined by actual traffic patterns. Another option is to define slot-controlled airports as congested airports; however, slotcontrol policies are very different throughout the world (Ball et al., 2010), rendering the analysis in this paper potentially inconsistent. Due to different policies, slot-control is used in fewer cases in the USA than, for example, in Europe irrespective of the actual congestion situation. Even within Europe, Level-3 slot-control is employed very heterogeneously and thus not an unambiguous indicator of airport congestion as analysed in this paper. For example, Spain has 13 coordinated airports, compared to five in France, six in Germany, six in the UK, 15 in Italy and 22 in Greece (WWACG, 2014). Nevertheless, a constrained airport in our analysis is typically schedule facilitated or coordinated, but not vice versa. Based on the aforementioned criteria, we identified 43 congested airports (135 uncongested airports) in the sample of 178 airports worldwide, with a traffic volume of more than 70,000 ATMs in 2010. About 50% of congested airports had runway systems of three and more runways; only seven single runway airports were regarded as constrained airports. Examples of the high volume airports with capacity constraints were: London Heathrow (LHR), Charlotte (CLT), Washington R. Reagan (DCA), Newark (EWR), Istanbul (IST) and Beijing (PEK). Among the single runway airports were San Diego (SAN), Geneva (GVA), and a number of airports in China. Other airports with capacity problems were: Guangzhou (CAN), Munich (MUC) and Jakarta (CGK), which are in the category of airports with two independent parallel runways. Seattle (SEA), Mexico City (MEX) and London Gatwick (LGW) have two dependent parallel runways; New York La Guardia (LGA), Delhi (DEL) and Melbourne (MEL) have two crossing runways. 3. Development of average seat capacity at constrained and unconstrained airports worldwide and in world regions Fig. 1 shows that average seat capacity per flight has grown in the global network of 178 selected airports by almost 10% from 126 seats in 2006 to 137 seats in 2012. At the same time, the total number of aircraft movements at those 178 airports has increased by 5.5% from 34 to approximately 36 million; the number of passengers has grown by 19% from 3.7 to 4.4 billion. The number of flights has grown less than the passenger volume, partly due to the fact that, on average, more seats per flight have been offered. In addition, airlines have been disciplined with capacity, so that load factors have increased as well. In 2013, the average load factor worldwide was almost 80% (IATA, 2014). The trend towards bigger aircraft had already begun before 2006: in 2000 the average seat capacity was about 105 seats per flight in the global network. Since 2000, the average seat capacity has grown by almost 20% worldwide. When we differentiate between airports with and without capacity problems (with traffic volumes of more than 70,000 ATMs in 2010), we see that the average seat capacity has increased at both groups of airports. Average seat capacity has grown by 8.5% from 130 to 141 seats at the 43 constrained airports and by 9% from 123 to 134 seats per flight at the 135 unconstrained airports. That is, airlines have deployed bigger aircraft into the market in general, regardless of the constraint situation at airports. As shown in Fig. 1, air traffic at constrained airports has grown faster than at unconstrained airports. At unconstrained airports, traffic fluctuated but did not change in volume very much between 2006 and 2012. The average traffic volume of constrained airports in 2012 was about 370,000 ATMs, whereas the corresponding volume of unconstrained airports was less than 150,000 ATMs. Traffic has

Please cite this article in press as: Berster, P., et al., Is increasing aircraft size common practice of airlines at congested airports?, Journal of Air Transport Management (2015), http://dx.doi.org/10.1016/j.jairtraman.2015.03.012

4

P. Berster et al. / Journal of Air Transport Management xxx (2015) 1e9

Fig. 1. Average seat capacity and number of aircraft movements in the global network of 178 selected (43 constrained & 135 unconstrained) airports 2006e2012 (OAG, 2006e2012).

concentrated more at hubs and other high volume airports, partly because of the concentration of intercontinental traffic on relatively few airports and the need to feed traffic into these hub airports. Total traffic at the 43 constrained airports amounted to 16 million ATMs in 2012, while the unconstrained airports handled a total volume of 20 million ATMs. The 43 constrained airports accounted for less than 2% of the airports worldwide. However, they handled one quarter of all movements. The traffic share of the 135 unconstrained airports was about 31%, so that, altogether, the selected airports handled more than 50% of total air traffic. The development of the average seat capacity has been analysed at constrained and unconstrained airports in the world regions as well. World regions have been defined as: -

Africa Asia Europe North America South America South West Pacific (i.e., essentially Australia)

Table 1 shows the development of average seat capacity per flight at constrained and unconstrained airports by world region. The picture of average seat capacity development becomes more diversified if we look at world regions. The average number of seats offered per flight has grown at congested airports worldwide. However, the growth is much more significant in the Middle East and South West Pacific than in Asia and Europe, and only marginal in North and South America. No airport was classified as constrained in Africa. Traffic volumes have increased in all regions except in North America, where the number of flights fell by 10% between 2006 and 2012. Traffic growth was strongest in the Middle East (þ68%) and Asia (þ55%). The highest growth of average seat

capacity was also in the Middle East (þ18%), whereas the decline of traffic in North America was accompanied by almost no change in the average aircraft size. In addition, the average seat capacity in North America was 112 seats per flight. This was much smaller than in other regions, in particular, smaller than the Middle East where the average number of seats offered was 236. It should be noted, however, that the category of congested airports in the Middle East is represented by just one airport, Dubai (DXB). The corresponding values of Asia and Europe were 171 and 162, respectively. The low capacity of flights in North America is partly due to the fact that most of these flights are domestic flights and domestic routes are typically characterised by higher frequencies (with somewhat smaller aircraft) than international routes. The average seat capacity also increased at uncongested airports in Europe, North and South America, Africa and in South West Pacific, whereas in Asia and the Middle East average seat capacity per flight decreased by about 5% between 2006 and 2012. Air traffic at uncongested airports has grown in all world regions, except in North America, where traffic was down by 18%. With the exception of Asia, the average seat capacity at unconstrained airports was lower in all world regions than at constrained airports, particularly in Europe, the Middle East and South West Pacific. Average seat capacity per flight was rather low in North America, with 111 seats. The highest was in Asia with 176 seats on average. 4. Airport specific developments of average seat capacity At an individual airport level, we see an even more diversified picture. Among the constrained airports, we identified 34 (79%) out of 43 constrained airports where the average number of seats offered per flight increased between 2006 and 2012. Conversely, 9 (21%) airports' average seat capacity declined. Both the average seat capacity and the number of ATMs increased at 22 airports, while

Please cite this article in press as: Berster, P., et al., Is increasing aircraft size common practice of airlines at congested airports?, Journal of Air Transport Management (2015), http://dx.doi.org/10.1016/j.jairtraman.2015.03.012

P. Berster et al. / Journal of Air Transport Management xxx (2015) 1e9

5

Table 1 Average seat capacity and traffic volumes at congested and uncongested airports in world regions for 2006 and 2012 (OAG, 2006 & 2012). World region

Asia

Africa

Europe

North America

Middle East

South America

Southwest Pacific

Constrained Airports Average Seat Capacity: 2012 2006

171 162

e e

162 148

112 111

236 200

133 132

170 153

Growth 2012/2006 (%) No. of ATMs (106): 2012 2006 Growth 2012/2006 (%) Unconstrained Airports Average Seat Capacity: 2012 2006

5.5

e

9.5

0.9

3.24 2.09

e e

3.66 3.53

7.64 7.93

e

3.7

55

176 186

Growth 2012/2006 (%) No. of ATMs (106): 2012 2006

5.7

Growth 2012/2006 (%)

43

3.87 2.71

150 143 4.9 0.40 0.32 25

10.4

135 120

111 103

12.5

7.45 8.77 17.8

the average aircraft size increased as the number of ATMs decreased at 12 airports. The regional distribution of constrained airports where on average flights have grown in seat capacity was as follows: -

11 airports in North America; 11 airports in Europe; 9 airports in Asia; 2 airports in South West Pacific; 1 airport in the Middle East.

There were five congested airports with decreasing seat capacity in North America, two airports in South America and two airports in Asia. None of the congested airports, where the average number of seats offered per flight decreased, were found in Europe, the Middle East or the South West Pacific. Among the constrained airports in North America with growing seat capacity were Atlanta (ATL), the largest airport in terms of number of passengers, New York (JFK), Denver (DEN) and Los Angeles (LAX). Washington National (DCA), Chicago O'Hare (ORD) and San Diego (SAN), the busiest single runway airport, were examples of constrained airports with decreasing average seat capacity. Chicago O'Hare had the second highest traffic volume worldwide, but the lowest average number of seats per flight of all constrained airports (94 seats in 2012). The average number of seats offered at London Heathrow (LHR) was 199 seats (in 2012), more than twice as high as that at Chicago O'Hare. The high seat capacity correlates well with the capacity constraint situation at LHR; no other airport is more constrained. LHR's capacity utilisation index (CUI) value of 85% is the highest worldwide. Apart from capacity constraints, the increase in aircraft size at ATL is, in part, due to the fact that the airport has been actively increasing international destinations. Washington National (DCA), however, is subject to perimeter restrictions and does not serve any international destinations, apart from flights to three Canadian destinations. Thus, the potential for employing larger aircraft at DCA is rather limited due to the destination structure. Many airports undertook capacity expansion and enhancement projects during the study period, ranging from building new terminals or expanding existing terminals, such as London Heathrow

0.32 0.19

0.8 0.6 0.45

11 0.52 0.44

68

33

18

173 181

133 118

134 120

4.6

7.8

5.55 5.54 0.2

18

0.63 0.37 70

12.7 1.47 1.24 18.5

11.7 0.57 0.51 11.8

(Terminal 5 in 2008) and Beijing Capital International Airport (Terminal 3 in 2008) to building new runways, such as Frankfurt airport (fourth runway in 2011) and Beijing Capital airport (third runway in 2007). Furthermore, at some airports regulated capacity restrictions were partially eased from time to time, such as Dusseldorf airport, resulting in a significant capacity boost (for more details in this case see Reichmuth et al., 2011). However, at most airports with sufficient demand potential, the congestion situation improved only for a short time, if at all. CUI values declined temporarily at airports like Beijing or Dusseldorf, but not at Frankfurt airport after the fourth runway was added. This is noteworthy, given the large capacity increase at Frankfurt airport. Lufthansa used Munich airport as a second hub when running into capacity problems at Frankfurt airport, but now move more and more international flights back from Munich to Frankfurt after the capacity increase. The same is true for average seat capacity per flight: a small decline tends to occur only temporarily after adding airport capacity, if at all. In the long run, average seat capacity per flight tends to rise, as well as CUI values, if the capacity constrained airport is backed up by sufficient demand potential. Nevertheless, there are reasons besides congestion for employing larger aircraft as well, as discussed later in this paper. There are examples of congested airports with capacity expansion but congestion levels considerably below London Heathrow, like Chicago O'Hare, for example, where CUI values have increased from 0.63 to 0.67, but the average number of seats per flight has decreased from 103 to 94 between 2006 and 2012. Average stage length has remained rather constant with values between 1420 km and 1450 km. Altogether, 135 airports in our sample (all over the 70,000 ATMs threshold) were identified as unconstrained airports. Of them, 107 airports (almost 80%) have been served by flights with increasing seat capacity per flight. At 28 airports, the average seat capacity per flight decreased between 2006 and 2012. At 63 airports with growing seat capacity per flight the passenger traffic volume has increased as well, whereas at 44 airports traffic has decreased. The regional distribution of unconstrained airports is as follows: - 48 airports in North America, of which 39 airports had increasing and 9 airports had decreasing seat capacity per flight.

Please cite this article in press as: Berster, P., et al., Is increasing aircraft size common practice of airlines at congested airports?, Journal of Air Transport Management (2015), http://dx.doi.org/10.1016/j.jairtraman.2015.03.012

6

P. Berster et al. / Journal of Air Transport Management xxx (2015) 1e9

- 40 airports in Europe, of which 39 airports had increasing and 1 airport had decreasing seat capacity per flight. - 25 airports in Asia, of which 14 airports had increasing and 11 airports had decreasing seat capacity per flight. - 5 airports in South West Pacific, of which 5 airports had increasing seat capacity per flight. - 5 airports in the Middle East, of which 1 airport had increasing and 4 airports had decreasing seat capacity per flight. - 9 airports in South America, of which 7 airports had increasing and 2 airports had decreasing seat capacity per flight. - 3 airports in Africa, of which 2 airports had increasing and 1 airport had decreasing seat capacity per flight. Traffic at most of the Asian airports has grown strongly in the past, whereas most of the North American airports reported traffic declines. The average number of seats per flight increased at all but one European airport, while in the Middle East most airports were served by flights with decreasing seat capacity, although traffic increased strongly at these airports. In summary, we can state that in almost 80% of all constrained airports, airlines have increased the average number of seats per flight in recent years. The reasons for rising seat capacity may have varied, depending on the airport specific situation. One reason was the lack of airport capacity, so increasing seat capacity per flight was one option to cope with demand in bottleneck conditions. On the other hand, we have found that also in 80% of all unconstrained airports, airlines have chosen the option to offer more seats per flight. Clearly, lack of airport capacity did not play a direct role here but may have played an indirect role when the other airport of the flight stage was constrained. Other reasons must have played a decisive role as well, since at only 20% of all airports, airlines had scheduled flights with declining seat capacity. 5. Factors causing airlines to raise the seat capacity of flights So far we have only indirectly seen the relationship between seat capacity per flight and slot availability, by examining the average seat capacity at congested airports and seeing that it is positively correlated with the traffic volume, as Fig. 1 displays. If we were able to show that seat capacity is functionally related with a measure of congestion, we could directly demonstrate that there is also a causal relationship between aircraft size and airport congestion. As discussed earlier, the capacity utilisation index (CUI) is used as an indicator of airport congestion. Fig. 2 shows the positive correlation between average seat capacity and the CUI in the 43 constrained airports. As numerous examples of airports have shown, particularly in North America, there is an incentive on economic grounds for airlines to offer more seats in bigger aircraft, without increasing the frequency. Unit costs of bigger aircraft are lower than that of smaller aircraft. Based on data from the US air traffic market in 2006, as provided in The Airline Monitor (ESG Aviation Services, 2007), one can show that costs per seat mile fall sharply from about 16 US Cents to 6 US Cents, when the average seat capacity per flight rises from about 40 seats to about 150 seats. If aircraft get bigger still, unit costs go down only marginally. Fig. 3 shows the development of the average seat capacity and Fig. 4 illustrates the development of the load factors of four US full service network carriers and Lufthansa. The four US carriers were merged into two mega carriers in 2010, and Lufthansa is one of the largest European flag carriers. They are contrasted with their lowcost counterparts, Southwest and Ryanair, respectively. For the US carriers, we have included all flights from US airports; and for reasons of comparability, we only have included flights within Europe for Lufthansa and Ryanair.

Ryanair increased their average aircraft size up to 189 seats per flight as a result of changing their aircraft fleet to Boeing 737e800 aircraft. This process was finished in 2006. Southwest Airlines kept their average aircraft size rather constant over the study period: they are the world's largest operator of Boeing 737e300 and 737e700 fleet. After Boeing 737e800 entered services in 2012, the average number of seats per flight increased to 143 seats in 2013. The average seats per flight offered by the other four US carriers did not increase significantly over the study period and remained in the range of 82 for Continental Airlines in 2005 and 109 for Delta Airlines in 2013. Furthermore, the mergers between Delta and Northwest Airlines, as well as United and Continental Airlines had no visible effect on the development of average aircraft size. Lufthansa has managed to increase average aircraft size by 30 seats per flight since 2008, as they increasingly had to compete with lowcost-carriers in Europe. Furthermore, they ceased their codeshare agreements with some regional carriers (Lufthansa Regional), such as Contact Air in 2012 and Air Dolomiti and Augsburg Airways in 2013. On the other hand, Fig. 4 shows that there was an increasing trend in average load factors since 2002. Nevertheless, load factors tend to fluctuate from year to year, but they were trending towards the 80% mark, except for Lufthansa. However, because of the fierce competition from low-cost-carriers, Lufthansa started in 2012 to transfer their intra-European flights to Germanwings, with the exception of flights to and from their main hubs, Frankfurt and Munich airports. The aforementioned mergers between the US airlines did not seem to affect the average load factors of those airlines. In summary, increasing aircraft size as a result of airport congestion is not that evident at the airline level as at the airport level. Besides taking note of the fact that the airline discussion is only based on a few example airlines, the airport and airline level discussion are not inconsistent with each other. If we look closely, by analysing different airlines as a whole, we have automatically included many smaller airports that are not congested. Nevertheless, decreasing unit costs is becoming more and more important for airlines, but we cannot expect to see bigger aircraft across the board as a result. Still, using larger aircraft is an important strategy to achieve lower unit costs, as shown by many successful low-costcarriers. Furthermore, increasing average aircraft size of a fleet usually takes time, as we can see in the case of Ryanair. Fig. 3 also underlines regional differences in aircraft size between the US and European markets at the airline level. 6. Model of average seat capacity To quantify our hypotheses about causal factors of average seat capacity development, we estimated a linear model that takes the form (e.g., Greene, 2011):

AVGSEATSi ¼ CONST þ

X

bj *xij

j

The dependent variable AVGSEATSi is the average number of seats per flight at airport i. The independent variables comprise a constant term, CUI, average flight length and several region-specific constants. Observations were weighted by the number of flights at the airport. Table 2 shows the final model estimation results. AVGFL describes the average flight length in km at an airport (OAG, 2012). NA (North America), EUR (Europe), ASIA (Asia) and MEAST (Middle East) are binary variables that take a value of 1 if an airport belongs to the region and 0 otherwise. However, coefficients of binary variables for the regions of Africa or South America were not significantly different from zero. CUI is the capacity utilisation index and

Please cite this article in press as: Berster, P., et al., Is increasing aircraft size common practice of airlines at congested airports?, Journal of Air Transport Management (2015), http://dx.doi.org/10.1016/j.jairtraman.2015.03.012

P. Berster et al. / Journal of Air Transport Management xxx (2015) 1e9

Fig. 2. Average number of seats per flight and capacity utilisation index (CUI) in the global network of 43 congested airports 2006e2012 (OAG, 2006e2012).

Fig. 3. Average number of seats per flight of selected US and European airlines (OAG, 2002e2013).

Please cite this article in press as: Berster, P., et al., Is increasing aircraft size common practice of airlines at congested airports?, Journal of Air Transport Management (2015), http://dx.doi.org/10.1016/j.jairtraman.2015.03.012

7

8

P. Berster et al. / Journal of Air Transport Management xxx (2015) 1e9

Fig. 4. Average load factors of selected US and European airlines (Sabre Airport Data Intelligence, 2002-2013).

Table 2 Estimation results (final model). Variable

Coefficient

R2

CONST CUI AVGFL NA EUR ASIA MEAST

57.6845812 *** 61.8579267 *** 0.02539032 *** 23.9042594 *** 9.89872101 ** 30.4363804 *** 37.6541034 ***

75.82%

*** Significant at the  1% level. ** Significant at the  5% level.

theoretically takes values between 0 and 1, but in real world applications CUI is typically limited to values between 0.4 and 0.85 (London Heathrow). A likelihood-ratio-test shows that discriminating the CUI coefficient between constrained airports and unconstrained airports (“test model”) was not significant (2*(LogL(final model) e LogL(test model)) ¼ 0.134). All variables were significant at the  1% level, except for the EUR variable (<¼5% level). Model fit (R2) was 75.8%. This is a reasonable value given the complex task of explaining average aircraft size with a rather simple model. However, we have to consider that there remains a significant part of variation in the data observed that this model cannot explain. Nevertheless, the model is sufficient for the problem studied in this paper. The number of passengers (PASS) was significant at the 5% level, but not at the 1% level. The test value was 5.44 and the critical values were 3.84 (5%) and 6.63 (1%), respectively. However, we decided to omit PASS from the final model setup because of the high correlation with the variables CUI and AVGFL at 0.5445 and 0.5817, respectively. Including PASS increased the explanatory power only marginally (DR2 ¼ 0.58%), yet the variance of the parameter estimates increased significantly.

The estimation results in Table 2 show that the average number of seats offered per flight increased by around 6 seats for a 0.1 increase in CUI value for a given region with stage length being constant. That means, for a constant flight length the average number of seats offered per flight increased by about 27 if the CUI value increased from 0.4 (very low CUI value of an uncongested airport) to 0.85 (London Heathrow, highly congested). Average flight length has significant effects on the average aircraft size: the average number of seats offered per flight increased by around 2.5 seats for every 100 km increase in average flight length at an airport. Finally, there were important regional differences: the average number of seats per flight tended to be lower in North America than in Europe, or even in Asia and the Middle East. There was more domestic traffic with smaller aircraft at higher frequencies in North America and more long-haul and intercontinental traffic with larger aircraft in Asia and the Middle East. Europe was somewhere between those two types. Table 3 shows the estimation results of different models that include only a subset of the explanatory variables and of the final model (#7) of Table 2. Here, we have to keep in mind that the estimators in models #1 to #6 are potentially biased because some

Table 3 Comparison of models. Model#

Model name

R2

Log-likelihood

1 2 3 4 5 6 7

CUI AVGFL CUI þ AVGFL REGIONS REGIONS þ CUI REGIONS þ AVGFL FINAL MODEL

12.22% 31.86% 37.69% 52.08% 56.34% 74.32% 75.82%

1124.478 1096.118 1086.086 1056.693 1046.253 986.802 980.061

Please cite this article in press as: Berster, P., et al., Is increasing aircraft size common practice of airlines at congested airports?, Journal of Air Transport Management (2015), http://dx.doi.org/10.1016/j.jairtraman.2015.03.012

P. Berster et al. / Journal of Air Transport Management xxx (2015) 1e9

explanatory variables are missing. Table 3 displays the share of linear variance (R2) in the estimation data set that is explained by the (sub) models. All models include a constant term (CONST). Model #1 comprises a constant term and the CUI variable. Model #2 is composed of a constant term and the AVGFL variable, model #3 is composed of a constant term and the CUI and AVGFL variables. Model #4 includes a constant term and the region-specific binary variables (REGIONS). Equally, models #5 and #6 are comprised of a constant term, region-specific variables, the CUI variable (#5) or the AVGFL variable (#6). Table 3 illustrates that the largest part of the linear variance (52.08%) in the estimation data set is explained by the region-specific variables NA, EUR, ASIA and MEAST, if we do not account for overlapping of explanatory power of variables. Accordingly, AVGFL and CUI account for 31.86% and 12.22%, respectively, of the linear variance that is explained by the submodels. However, if we take a look at the last three models in Table 3 we see a rather large increase of R2 if we include the AVGFL variable (þ22.24 percentage points), but only a small increase in R2 if we include the CUI variable (þ4.26 percentage points and þ1.50 percentage points, respectively). From this observation we concluded that there is a rather small part of linear variance in the estimation data set that is exclusively explained by the CUI variable. The most important variable for explaining the average number of seats per flight at an airport in this model is the region-specific characteristics (NA, EUR, ASIA and MEAST), followed by the average flight length at an airport (AVGFL) and then the capacity utilisation at the airport (CUI). Nevertheless, despite the rather small share of the linear variance in the estimation data set that is exclusively explained by the CUI variable, a likelihood-ratio-test shows that the CUI variable is still highly significant. A possible explanation for this finding could be the (potential) competition between airlines, leading them to employ smaller aircraft to keep slots and deter potential entrants (Dresner et al., 2002; Fukui, 2012; Givoni & Rietveld, 2009). 7. Summary and conclusions Airlines vary aircraft size, and thus seat capacity, depending on factors such as the level of airport congestion and flight length. The results from our study confirm the expected increase in average aircraft size predominantly at capacity-constrained airports. The results also indicate that average aircraft size has increased at most uncongested airports as well. Overall, the average seat capacity per flight rose at the 178 sample airports by almost 10% from 126 seats in 2006 to 137 seats in 2012. At the same time, the total number of aircraft movements at those airports increased by 5.5% from 34 to about 36 million, while the number of passengers increased by 19% from 3.7 to 4.4 billion. When sufficient levels of load factor and frequency are reached and demand continues to grow, airlines have an economic interest to operate aircraft with a higher seat capacity at lower unit costs, rather than to increase the frequency of flights. Furthermore, bigger aircraft types tend to be employed on longer routes for technical and economic reasons. From the statistical analysis, we have seen that the average flight length has increased in the past, at both congested and uncongested airports. An increase in average flight length may, therefore, be regarded as a factor describing the tendency of employing aircraft with a higher seat capacity at a lower unit cost. Other factors may cause airlines to schedule otherwise, depending, for instance, on local conditions, aircraft availability and airspace and airline regulations. Our model analysis has shown that the average number of seats offered per flight increased by around 6 seats for a 0.1 increase in CUI value, and by around 2.5 seats for a 100 km increase in average flight length at an airport. Furthermore, there were regional differences: the average number of seats per flight tended to be lower

9

in North America (24) than in Europe (þ10), or in Asia (þ30) and the Middle East (þ38). Finally, the average flight length at an airport was the single most important factor explaining average aircraft size, followed by regional differences. The explanatory power of increasing airport congestion is highly significant, but nevertheless, much smaller than the aforementioned factors. The prevailing trend of operating larger aircraft has implications for both airports and airlines. Airports may have to adjust their infrastructure to accommodate larger aircraft in the future. This affects not only the runway system, but terminals and the apron as well, to avoid congestion elsewhere. From an economic perspective, adjusting landing charges might be necessary to accommodate changes in aircraft size and maximise an airport's revenues. However, increasing aircraft size actually threatens airports and some airlines of low-demand markets, because these markets are becoming less and less attractive for airlines with increasing aircraft size. Acknowledgements The authors would like to thank the anonymous reviewers for their insightful comments and suggestions for improvements. References Airbus, 2010. Global Market Forecast 2010e2029. Airbus, Blagnac. Ball, M., Dupont, M., Hansen, M., 2010. Structural differences in the US and European systems and their impact on airline scheduling practices. In: NEXTOR Symposium January 2010, FAA Headquarters Washington DC. Boeing, 2012. Current Market Outlook 2011e2031. Boeing, Seattle. Bubalo, B., Daduna, J.R., 2011. Airport capacity and demand calculations by simulation e the case of Berlin-Brandenburg International Airport. Netnomics 12, 161e181. Button, K., 2002. Debunking some common myths about airport hubs. J. Air Transp. Manag. 8, 177e188. Charlton, A., 2009. Airport regulation: does a mature industry have mature regulation? J. Air Transp. Manag. 15 (3), 116e120. Dennis, N., 1994. Airline hub operations in Europe. J. Transp. Geogr. 2 (4), 219e233. Dresner, M., Windle, R., Yao, Y., 2002. Airport barriers to entry in the US. J. Transp. Econ. Policy 36 (2), 389e405. ESG Aviation Services, 2007. The Airline Monitor 20(3). ESG Aviation Services (Ponte Vedra Beach). Eurocontrol, 2008. Eurocontrol Long-term Forecast: Flight Movements 2008e2030. Eurocontrol, Brussels. Eurocontrol, 2013. Challenges of Growth 2013. Eurocontrol, Brussels. European Commission, 2010. High-speed Europe, a Sustainable Link Between Citizens. European Commission, Luxembourg. Fukui, H., 2012. Do carriers abuse the slot system to inhibit airport capacity usage? evidence from the US experience. J. Air Transp. Manag. 24, 1e6. Givoni, M., Rietveld, P., 2009. Airline’s choice of aircraft size e explanations and implications. Transp. Res. Part A 43 (5), 500e510. Gelhausen, M.C., Berster, P., Wilken, D., 2008. Airport choice in Germany and the impact of high-speed intercity train access: the case of the Cologne region. J. Airpt. Manag. 2, 355e370. Gelhausen, M.C., Berster, P., Wilken, D., 2013. Do airport capacity constraints have a serious impact on the future development of air traffic? J. Air Transp. Manag. 28, 3e13. Greene, W.H., 2011. Econometric Analysis. Pearson Education, Harlow. International Air Transport Association (IATA), 2014. Passenger Demand Maintains Historic Growth Rates in 2013. IATA, Montreal. International Civil Aviation Organization (ICAO), 2005. The World of Civil Aviation. ICAO, Montreal. International Civil Aviation Organization (ICAO), 2008. FESG CAEP/8 Traffic and Fleet Forecasts, CAEP-sg/20082-ip/02. ICAO, Seattle. International Civil Aviation Organization (ICAO), 2012. ICAO Annual Report of the Council. ICAO, Montreal. Official Airline Guide (OAG), 2002e2013. Market Analysis. Reed Travel Group, Dunstable. Reichmuth, J., Berster, P., Gelhausen, M.C., 2011. Airport capacity constraints: future avenues for growth of global traffic. CEAS Aeronaut. J. 2 (1e4), 21e34. Sabre Airport Data Intelligence, 2002-2013. Data Based on Market Information Intelligence Tapes (MIDT). Sabre Airline Solutions, Southlake. Teyssier, N., 2010. Aviation Statistics & Data: a Vital Tool for the Decision Making Process. ICAO, Montreal. Wilken, D., Berster, P., Gelhausen, M.C., 2011. New empirical evidence on airport capacity utilisation: relationships between hourly and annual air traffic volumes. Res. Transp. Bus. Manag. 1, 118e127. WorldWide Airport Coordination Group (WWACG), 2014. Airport List & Their Coordination Status. http://www.wwacg.org (accessed 14.10.14.).

Please cite this article in press as: Berster, P., et al., Is increasing aircraft size common practice of airlines at congested airports?, Journal of Air Transport Management (2015), http://dx.doi.org/10.1016/j.jairtraman.2015.03.012