Delays and fares - A case study of John Wayne Airport

115
Delays and fares - A case study of John Wayne Airport * Elsa Delpech Sophie Leroy ENAC ENAC January 30, 2012 Abstract In this study, we focus on a very regulated airport, John Wayne, in California. We merge data from the U.S. Department of Transportation on delays and fares for each quarter between 1999 and 2010 in order to study the impact of delays on fares at this airport. So, we find that price fall by $1.23 for each additional minute of delay; and in a competitive market, prices do not decrease. * We thank Chantal Latg´ e-Roucolle and Steve Lawford, our project tutors. We also thank Yannick Jestin for helping us with Python loading. We are grateful to Pierre-Henri Chabrouty and Adam Ouammou for providing us with parsed DB1B Coupon and Ticket data and for their help with some code problems. And we thank the support of the Reasearch and Innovative Technology Administration for information about the data. Correspondence can be addressed to Elsa Delpech ([email protected]) or Sophie Leroy ([email protected]) at ENAC, 7 avenue Edouard Belin, BP 54005, 31055, Toulouse, Cedex 4, France.

Transcript of Delays and fares - A case study of John Wayne Airport

Page 1: Delays and fares - A case study of John Wayne Airport

Delays and fares - A case study of John Wayne Airport∗

Elsa Delpech Sophie Leroy

ENAC ENAC

January 30, 2012

Abstract

In this study, we focus on a very regulated airport, John Wayne, in California. We mergedata from the U.S. Department of Transportation on delays and fares for each quarter between1999 and 2010 in order to study the impact of delays on fares at this airport. So, we find thatprice fall by $1.23 for each additional minute of delay; and in a competitive market, prices donot decrease.

∗We thank Chantal Latge-Roucolle and Steve Lawford, our project tutors. We also thank Yannick Jestin forhelping us with Python loading. We are grateful to Pierre-Henri Chabrouty and Adam Ouammou for providing uswith parsed DB1B Coupon and Ticket data and for their help with some code problems. And we thank the supportof the Reasearch and Innovative Technology Administration for information about the data. Correspondence canbe addressed to Elsa Delpech ([email protected]) or Sophie Leroy ([email protected]) at ENAC, 7avenue Edouard Belin, BP 54005, 31055, Toulouse, Cedex 4, France.

Page 2: Delays and fares - A case study of John Wayne Airport

1

Contents

1 Introduction 4

1.1 The airport and its region . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

1.1.1 History . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

1.1.2 Orange County . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

1.1.3 Activities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

1.1.4 Traffic and statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

1.2 Regulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

1.2.1 Traffic regulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

1.2.2 Noise Regulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

2 Fares 16

2.1 Fare data description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

2.2 First analysis of fare data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

2.2.1 SNA, a regional airport? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

2.2.2 Influence of class and distance on price . . . . . . . . . . . . . . . . . . . . . . 21

2.2.3 Influence of competition on price . . . . . . . . . . . . . . . . . . . . . . . . . 23

2.3 Descriptive linear regressions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

3 Delays 31

3.1 Description of data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

3.2 Delays’ distribution for all flights . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

3.2.1 Quantile analysis of departure delays . . . . . . . . . . . . . . . . . . . . . . . 35

3.2.2 Quantile analysis of arrival delays . . . . . . . . . . . . . . . . . . . . . . . . 37

3.3 Delay distribution by carrier . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

3.4 Delays by time slot . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40

Page 3: Delays and fares - A case study of John Wayne Airport

2

3.5 Model of departure delays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42

3.5.1 Model for all departure delays . . . . . . . . . . . . . . . . . . . . . . . . . . . 43

3.5.2 Model of departure delays above the 90% quantile . . . . . . . . . . . . . . . 51

3.5.3 Model of departure delays below the 10% quantile . . . . . . . . . . . . . . . 56

3.6 Other explanatory factors of delay . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59

4 The merge 61

4.1 2010:2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61

4.1.1 Data description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61

4.1.2 Descriptive linear regressions . . . . . . . . . . . . . . . . . . . . . . . . . . . 63

4.2 Fares and delays from 1999 to 2010 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66

4.2.1 Departure delays on time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67

4.2.2 Baseline model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68

4.2.3 Variation of prices over time . . . . . . . . . . . . . . . . . . . . . . . . . . . 70

4.2.4 Delay effect on competitive market . . . . . . . . . . . . . . . . . . . . . . . . 73

4.2.5 Model including the influence of the carrier . . . . . . . . . . . . . . . . . . . 74

4.2.6 Model with a maximum of explanatory variables . . . . . . . . . . . . . . . . 77

4.2.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78

5 Final conclusion 80

A Appendix 81

B List of IATA codes 85

B.1 Airports . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85

B.2 Airlines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85

Page 4: Delays and fares - A case study of John Wayne Airport

3

C Code 86

C.1 Gathering months together for each quarter . . . . . . . . . . . . . . . . . . . . . . . 86

C.2 Merging all the quarters and gathering quarters together . . . . . . . . . . . . . . . . 90

Page 5: Delays and fares - A case study of John Wayne Airport

4

1 Introduction

California’s John Wayne Airport (SNA), serving nearly 9 million passengers per year, is one of

the most heavily regulated in the United States. The restrictions on departures, passengers, and

noise-levels, have been modified a number of times since they were first imposed in 1985. This

project will attempt to measure and explain the impact of changing regulation on passenger levels,

fares and delays, at John Wayne Airport, from the 1990s to the present day. The results will be

used to examine links between service quality and fares: are passengers willing to pay less for air

travel when delays increase?

1.1 The airport and its region

John Wayne Airport is a commercial airport which is owned and operated by Orange County, in

Southern California (SoCal), in the United States (cf Figure 1). Unlike most population centers

in the U.S., Orange County contains no major cities and is mostly suburban. The largest town is

Anaheim, with 365,463 inhabitants. The city of Santa Ana serves as the county seat, providing

local government services to residents of the county; and is the site of John Wayne Airport.

1.1.1 History

Constructed in 1923, the airport was initially a flying school. During the Second World War,

it served as a military base, after which the federal government returned it to the County with

the stipulation that it remain open to all kinds of aviation use. The airport was only served by

one airline (Bonanza Air Lines, which operated flights between Los Angeles and Phoenix, via San

Diego) until the building of a new terminal in 1967, which permitted the treatment of increased

numbers of passengers, and the creation of regular routes. John Wayne Airport was originally

known as Orange County Airport, and was renamed in 1979 to honor the actor John Wayne, who

resided near Newport Beach and died that year. In the late 1990s and early 2000s, a new larger

Page 6: Delays and fares - A case study of John Wayne Airport

5

airport was proposed for the nearby site of El Toro Marine Corps Air Station, which had recently

closed. The cities in the immediate vicinity of El Toro created a joint powers authority, the El Toro

Reuse Planning Authority (ETRPA), in order to oppose the project.

The battle between pro and anti-airport groups dominated Orange County politics for much of

the period. Airport proposals were defeated in two hotly-contested ballot initiatives, and further

challenges took place in the courts. A series of demonstration commercial flights from the airport in

June 1999 resulted in much public outcry, due to the use of different flight paths from the military,

and continued circling over nearby cities when lining up for another approach. The proposal was

eventually defeated in March 2002, and no new airport was built.

Today, John Wayne serves nearly 9 million passengers, while Los Angeles and Ontario serve almost

59 million and 7 million passengers respectively.

1.1.2 Orange County

The “Mediterranean” climate of the county, and its eleven kilometers of beaches attract millions

of tourists every year.1 Huntington Beach is a famous surf spot, known as “Surf City”, and is the

location of many competitions. Other tourist destinations include the Disneyland theme park, the

Disney California Adventure park in Anaheim (with almost 16 million visitors in 2010) and Knott’s

Berry Farm in Buena Park (3.6 million visitors in 2010). The Anaheim Convention Center is the

largest complex of its type on the West Coast. Little Saigon is another important tourist center, and

hosts the largest concentration of Vietnamese outside of their country. Famous buildings include

the Crystal Cathedral in Garden Grove, which is the largest place of worship in California, and the

Balboa Pavilion in Newport Beach. The Pavilion is one of California’s last surviving waterfront

recreational pavilions from the turn of the century. It continues to serve the public today as a

marine recreational facility, and is Newport Beach’s most famous landmark, as well as its oldest

standing building. Some of the most expensive and selective suburbs of the United States are

Page 7: Delays and fares - A case study of John Wayne Airport

6

located in the county, especially near the coast (the Orange Coast). Other attractions include

Huntington Beach harbor. The county has a number of large shopping centers, including the Irvine

Spectrum Center, South Coast Plaza in Costa Mesa, Fashion Island in Newport Beach, and The

Block at Orange. The main tourist centers are Mission San Juan Capistrano and the Richard

Nixon Library & Birthplace in Yorba Linda: this is the only private presidential library in the

country since the government started their construction in 1939. The Nixon Home is a National

Historic Landmark. Finally, there is the Edison Field, which is the home ballpark to Major League

Baseball’s Los Angeles Angels of Anaheim. This stadium attracts approximately 100,000 visitors

to the area over the Harvest Crusades three-day weekend, according to the OC Register.

(a) Orange County,California.

(b) Orange County with his county seat SantaAna in pink in the center and his main touristdestination, Anaheim, in yellow in the north.

Figure 1: Geography of Orange County, on the Pacific Ocean. Source: Wikipedia.

1.1.3 Activities

John Wayne Airport is located roughly 56 kilometers south of Los Angeles, between the cities of

Costa Mesa, Irvine and Newport Beach. The airport is very close to two major commercial airports,

Los Angeles International (LAX) and Ontario International (ONT), both of which are within 60

Page 8: Delays and fares - A case study of John Wayne Airport

7

kilometers of John Wayne (SNA) and are owned by Los Angeles World Airports (cf Figure 2).

Figure 2: Location of John Wayne Airport and major nearby airports, denoted by their three-letterIATA codes. Source: Google Maps.

John Wayne is primarily a local service airport, serving cities on the West Coast and other

hub airports to the East, like Ontario International, whereas Los Angeles International is served

by international flights.

John Wayne makes an effort to attract tourists by providing good transportation services. The

John Wayne Shuttle Association links the airport to other featured locations in the Los Angeles

/ Orange County area, with collection and delivery at Disneyland, Universal Studios, the Los

Angeles Convention Center, the Anaheim Convention Center, the Long Beach Convention Center,

the Staples Center, the Los Angeles Coliseum, Dodger Stadium and Edison Field. With these

shuttles, John Wayne reaches as many tourist destinations as Los Angeles International .

Page 9: Delays and fares - A case study of John Wayne Airport

8

1.1.4 Traffic and statistics

Table 1: Total number of passengers to and from John Wayne Airport since 1991, and variationon the previous year. Source : John Wayne Airport Press Kit [13].

Table 2: Total number of passengers to and from John Wayne Airport, and aircraft operations,and variation between 2009 and 2010. Source : John Wayne Airport Press Kit [13]

Page 10: Delays and fares - A case study of John Wayne Airport

9

Since 2008, the number of passengers at John Wayne has fallen considerably (cf Table 1). Similarly,

Los Angeles International had more than 62 million passengers in 2007 and 59 million in 2010, while

Ontario International had 7 million passengers in 2007 and 4 million in 2010. As the number of

passengers and the volume of cargo decrease, all other aircraft operations decrease too except

military usage (cf Table 2).

Two runways serve commercial and private aircraft: a 5,700-foot main runway and a 2,887-foot

general aviation runway. The biggest aircraft that the airport is able to accomodate is the Boeing

757 (up to 239 passengers). Some gates are built to handle planes up to the size of a Boeing 767 (up

to 375 passengers), which could operate with payload or fuel load restrictions. John Wayne Airport

is very important for general aviation: approximately 500 general aviation aircraft are based there,

and are supported by facilities offering aircraft fueling, hangar space, maintenance services, aircraft

sales, charter flights and flight instruction.

Ten commercial, three commuter and two all-cargo airlines operate at John Wayne Airport.

In 2011, 85% of the routes into and out of John Wayne Airport were national. (Source: John

Wayne Airport website http://www.ocair.com/ ). For comparison, Ontario International has

eight commercial airlines and Los Angeles International has fifty-eight commercial airlines.

Page 11: Delays and fares - A case study of John Wayne Airport

10

Type Carrier Destination cities (regular non stop)

Commercial

Alaska Airlines Portland, Seattle-Tacoma

American Airlines Chicago-O’Hare, Dallas-Fort Worth

Continental AirlinesChicago-O’Hare,

Houston-Intercontinental, Newark

Delta AirlinesAtlanta, Minneapolis-St. Paul, Salt Lake

City

Frontier Airlines Denver

Southwest Airlines

Chicago-Midway, Denver, Las Vegas,

Oakland, Phoenix, Sacramento, San

Francisco, San Jose

United Airlines Chicago-O’Hare, Denver, San Francisco

US Airways Phoenix

US Airways Express Phoenix

WestJet Vancouver (Seasonal: Calgary)

Commuter

Delta Connection Salt Lake City

SkyWest Airlines Salt Lake City, San Francisco

United ExpressDenver, San Francisco (Seasonal, winter:

Mammoth Lakes)

Cargo FedEx

UPS

Table 3: Airlines serving John Wayne Airport and non-stop destinations. Source: John WayneAirport website (http://www.ocair.com/airlines/).

Page 12: Delays and fares - A case study of John Wayne Airport

11

1.2 Regulation

1.2.1 Traffic regulation

The aircraft access and noise level regulations at John Wayne Airport are among the strictest in

the world, mainly because of its proximity to residential areas. The Airport is regulated by the

John Wayne Aiport Settlement Agreement, which was established in 1985. The settlement was

the outcome of a lawsuit in federal court between Orange County (the owner of the airport) and

the following group: the City of Newport Beach, the Airport Working Group of Orange County,

Inc., and a local non-profit environmental organization, Stop Polluting Our Newport. The citizens’

Airport Working Group checks that aircraft access and noise restrictions are enforced at the airport.

The Settlement Agreement limited [6]2

• operations from 7 am to 11 pm on weekdays and Saturdays, and from 8am to 11pm on Sundays

(except in the case of emergencies; and delays due to mechanical problems, air traffic control

issues or weather);

• the number of gates to 14, with no more than 9 serving Boeing 767 size aircraft;

• the number of commercial flights to 73 departures a day

• the number of cargo flights to 2 a day;

• the terminal size to 337,400 square feet;

• traffic volume to 8.4 million passengers per year.

In 1990, the U.S. congress voted on the Airport Noise and Capacity Act that prohibits cities from

limiting aircraft operations without prior approval from the Federal Aviation Administration.3 In

2003, some extensions on aircraft access were made to the Settlement Agreement to enable increased

operations:

Page 13: Delays and fares - A case study of John Wayne Airport

12

• the maximum number of gates was raised from 14 to 20;

• the maximum number of commercial flights was increased from 73 to 85 per day;

• the maximum number of cargo flights was increased from 2 per day (before 2003) to 4 per

day thereafter;

• planning of a third terminal allowed (opened on 14 November 2011)4

• the maximum annual number of passengers was increased by 30%, to 10.8 million (this has

not yet been reached, since there are currently roughly 9 million passengers per year).

The Federal Aviation Administration approved the extensions of the Settlement Agreement on

the grounds that they were not a constraint for operations, and because the safety level remained

acceptable. The three original parties agreed to extend the Settlement until 31 December 2015.

1.2.2 Noise Regulation

The Access and Noise Office, which is particular to John Wayne Airport, is in charge of monitoring

and enforcing the regulation with the assistance of the noise monitoring stations, the locations

of which are displayed in figure 4 in the appendix5. These stations enable the Office to follow

aircraft noise in real time. They also have at their disposal an internet flight tracking system, the

“AirportMonitor”.6 It shows the traffic within a range of 129 kilometers. The staff also responds

to calls about noise complaints, community concerns and information requests.7

Two forms of noise regulation are applied at John Wayne Airport: one for commercial aviation,

and the other for general aviation. General aviation refers to “all flights other than military and

scheduled airline and regular cargo flights, both private and commercial. General aviation flights

range from gliders and powered parachutes to large, non-scheduled cargo jet flights”.8 It includes

private planes.

Page 14: Delays and fares - A case study of John Wayne Airport

13

Commercial carriers must comply with the Phase 2 Commercial Airline Access Plan and Regulation

that is applicable from 1 October 1990 until 31 December 2015. This plan is issued by the County

of Orange as the owner and certified operator of John Wayne Airport. It implements the applicable

regulation at John Wayne airport, including the 1985 Settlement agreement and 2003 amendment.

It deals with operational capacity, hours of operations, and noise levels. On the noise contour map,

in figure 5, we observe the different geographic limits of noise levels. For example, within the 70

CNEL (Community Noise Equivalent Level) contour, the average noise level over a 24-hour period

(from midnight to midnight) must not exceed 70 decibels. For an idea of the severity of the different

levels of noise, see table 6 in the appendix. For instance, television and radio typically reach 65

decibels. The noise level in CNEL refers to an average noise over one day. There is also a noise

limitation for every flight. In table 5, we report the noise level limits in decibels SENEL (Single

Event Noise Exposure Level methodology) of class A and class E aircraft at different monitoring

stations. The noise level in SENEL refers to the noise of one aircraft.

Table 4: Noise limits at John Wayne Airport, as of February 2011. Source: Boeing, Airport Noiseand Emission Regulations (http://www.boeing.com/commercial/noise/john_wayne.htm).

The class of an aircraft depends on its route. An aircraft like a B737 can be in both class A

and E: E if it flies from Northern Califonia to Southern California, A if it flies from California to

East Coast. Class E aircrafts usually correspond to short-haul flights. They need less fuel, and

Page 15: Delays and fares - A case study of John Wayne Airport

14

are generally lighter aircraft; and so produce less noise, unlike class A aircraft that usually serve

medium and long-haul routes.9 The penalty for not complying with these limits can be10

• disqualification of aircraft in that class;

• a penalty of up to $500,000;

• disqualification of the carrier from the airport.

The General Aviation Noise Ordinance regulates the operations of general aviation. Table 5

gives the noise level limits in decibels SENEL for general aviation at different monitoring stations.

Table 5: General aviation noise limits at John Wayne Airport in February 2011. Source:Boeing, Airport Noise and Emission Regulations (http://www.boeing.com/commercial/noise/john_wayne.html).

In case of violation of these limits, a “Notice of Violation” is issued and is applied to the aircraft

owner, the aircraft operator and the aircraft, for 3 years after the date of violation. If three notices

are registered in a 3-year period, the aircraft operator and the aircraft cannot operate at John

Page 16: Delays and fares - A case study of John Wayne Airport

15

Wayne Airport for a further 3 years.

In order to comply with these noise rules, pilots are required start their takeoffs at almost full

power, and then incline quickly and steeply, to gain altitude, before cutting down engine power,

so that the aircraft is sufficiently quiet when it flies over Newport Beach and the surrounding

communities. It makes John Wayne one of the 10 scariest airports in the world according to two

pilots and some SmartTraveler readers, who gave their opinion about difficult takeoffs and landings

to the travel site Smarter Travel.11 In 1993, new Federal Aviation Administration regulations

gave pilots a longer full-power climb out of John Wayne, increasing the altitude at which they are

required to throttle back from 500 to 800 feet. But the ascent is still regarded as an especially

demanding one.

Page 17: Delays and fares - A case study of John Wayne Airport

16

2 Fares

We begin with a first description and analysis of the data we have. Indeed, the objective is to

analyse the potential influence of delays on prices by merging two tables of data: the first one with

the fares and the second one with the delays. But we would like to describe separately these two

tables in a first time in order to better understand the influence of each parameter on fares on one

hand, and on delays on the other hand.

2.1 Fare data description

With the parsed DB1B Coupon and Ticket data, we have the fares of 10% of all the tickets for

direct routes sold in the United States for each quarter from 1999 to 2010. We retain by route for

each quarter:

• The origin of the route, identified by its three letter IATA code;

• The destination of the route, identified by its three letter IATA code;

• The length of the route, in miles;

• The airline which operated the flight, identified by its two letter IATA code;

• The market share that the airline has on this route in percentage;

• The ticketclass;

• The number of passengers;

• The mean price by passenger.

To illustrate, we focus on 2010:2, and routes to or from John Wayne Airport because this is the

interest of our study. We remove all tickets with fare below $20, which correspond to frequent flyer

Page 18: Delays and fares - A case study of John Wayne Airport

17

programs; and above $10 000.

We identify the carriers that operate at John Wayne Airport during this period of time. There are

38 routes into SNA and 31 routes out of SNA.

Table 6: Routes to and from SNA, for each carrier in 2010:2

Table 6 shows that carriers in John Wayne share a lot of routes. So, we can wonder how they

share the market on these routes and whether the price is a lot different between the carriers.

On the John Wayne airport website, it is said that there are also Skywest and Westjet but they do

not appear in our data. We assume that this is because code-sharing is not taken into account in

our data whereas, for example, Skywest operates some flights for Delta or United. On the other

hand, in the data we have Virgin America, which does not appear on the SNA website because

Virgin ceased operations at SNA on May 2010.

2.2 First analysis of fare data

2.2.1 SNA, a regional airport?

We look at carriers’ market share in terms of number of passengers:

Page 19: Delays and fares - A case study of John Wayne Airport

18

Figure 3: Carriers’ market share (number of pax) in SNA (2010:2).

Southwest has the highest presence at SNA; and the other carriers share the remaining half of

the market. Note that the market share of Virgin America is inaccurate because it left the market

in the middle of the quarter.

Moreover, we have some evidence that John Wayne Airport is a local or regional airport. Indeed,

short-haul routes have a majority of economy class and carriers operating at John Wayne do not

have a lot of first or business classes:

Page 20: Delays and fares - A case study of John Wayne Airport

19

Table 7: Ticket fare class distribution for each carrier operating at SNA (2010:2).

Not all the carriers offer all the classes and the price distributions are very different between

them because of that. Indeed, we can observe modes in the price distribution corresponding to the

mean price of each class. Figure 4 shows Continental and Southwest price distributions. Continental

has three classes, and we see in its price distribution a first mode at approximately $500, which

correponds to the majority of ticket fares for economy class, and a second mode, more smeared

between $1500 and $2500, which corresponds to both first and business classes. As for Southwest,

it only offers economy class, and its price distribution only has one mode at appromimately $300.

Page 21: Delays and fares - A case study of John Wayne Airport

20

(a) Price distribution of Continental (2010:2). (b) Price distribution of Southwest (2010:2).

Figure 4: Example of price distribution of carriers operating at SNA

By plotting the mean distance covered by each carrier during this quarter (cf figure 5), we

confirm that John Wayne is a local airport. Indeed, it is commonly said that a long haul flight is

approximately above 7000 km. So all carriers cover short or medium haul. Moreover, Southwest has

a big weight at John Wayne regarding Continental (cf figure 3) and cover the shortest routes. So,

when we calculate the mean distance per passenger, all carriers included, we find 1284 kilometers.

Page 22: Delays and fares - A case study of John Wayne Airport

21

Figure 5: Mean distance per passenger for each carrier (km) (2010:2).

2.2.2 Influence of class and distance on price

We take the directional route from John Wayne to Newark operated by Continental as an example

in order to see the repartition of classes and the difference in price between the classes. Economy

class represents on this route 99% and first and business classes share the last percent. Nevertheless,

ticket fares for first class is three times as much as ticket fares for economy class.

Page 23: Delays and fares - A case study of John Wayne Airport

22

Figure 6: Mean price for each class (in US$) (2010:2).

Indeed, the class has a big influence on the price. Figure 7 shows that there are two regression

lines: one for the economy class and the second one for the two other classes. Moreover, we observe

that the price of the business class or the first class increases more quickly with the distance than

the economy class.

Page 24: Delays and fares - A case study of John Wayne Airport

23

(a) Scatter plot of the mean price on distance (b) Scatter plot of the mean price on distance times the

indicator of the class (we observe here only the price of

the tickets from economy class

Figure 7: Influence of the class and the distance on the price of the ticket (2010:2).

2.2.3 Influence of competition on price

We took the route for which there is the greatest competition, John Wayne to Denver and Denver

to John Wayne, in order to see the market share of each carrier on this route and the difference in

price between the carriers. All the carriers present on these routes operate their flight with only

economy class.

Page 25: Delays and fares - A case study of John Wayne Airport

24

(a) Marketshare on the route SNA to DEN (b) Marketshare on the route DEN to SNA

Figure 8: There are 4 carriers operating on SNA/DEN: Frontier (F9), United (UA), US Airways(US) and Southwest (WN) (2010:2).

Figure 9: Mean price per carrier between Denver and John Wayne in US$ (2010:2).

Figure 9 shows that there is a big difference in prices between US Airways and the other carriers

sharing the market. Indeed, Frontier, United Airlines and Southwest align their prices between $200

Page 26: Delays and fares - A case study of John Wayne Airport

25

and $300, whereas US Airways has a price twice as much as the three others. But, that may be a

consequence, figure 8 shows that US Airways’ market share is really low (lower than 1%), whereas

Southwest, United and Frontier share the market almost equally.

That is surprising that carriers do not put the same price for the route from Denver to John Wayne

and the route from John Wayne to Denver.

2.3 Descriptive linear regressions

We regress the mean price on a constant, distance, a dummy variable which indicates if the ticket

corresponds to economy class or not, another dummy which indicates if there is competition on the

route and eight other dummies corresponding to the operating carrier, basing our regression upon

American Airlines.

Page 27: Delays and fares - A case study of John Wayne Airport

26

Table 8: Linear regression number one (2010:2).

For the classes, we split them into two categories: economy class and business and first classes

together. Indeed, the difference of the price between first class and business class is not very big

compared to the difference of the price between economy class and the other classes. Moreover, the

number of passengers in economy class is much higher than in the two other classes. That is why

we put just one dummy “eco” basing the regression on first and business together.

Page 28: Delays and fares - A case study of John Wayne Airport

27

There are only 168 observations for the regression, whereas the sample goes from 1 to 184, because,

after taking off the ticket fares below $20, there are some routes with no passenger, and then, it

was impossible to calculate a mean price.

The R-squared and the F-statistic indicate that the model is globally significant. The coefficient of

the variable ECO shows that economy class is probably much less expensive than other classes and

the coefficient of the variable DISTANCE shows that the ticket fare increases with the lenght of

the route. Indeed, all the other variables constant, the ticket for an economy class will be $775 less

than a ticket for another class. The variable COMPET seems not significant at usual level, so, we

can not interpret it very well. Looking at the coefficients for the dummies indicating the carrier, we

could say that only Continental, United and Southwest are more expensive than American Airlines.

This conclusion is very surprising for Southwest and when we look at their probabilities, we can

guess that those variables are non significant. We confirm that with a Fisher test and we make the

same regression without the carrier dummies.

Page 29: Delays and fares - A case study of John Wayne Airport

28

Table 9: Linear regression number two (2010:2).

But, we guess the big influences of the distance and the economy class on the price hide the

influence of the operator. Moreover, there is a link between the operator and the distance. So, we

decided to make a regression only with the dummies indicating the operator.

Page 30: Delays and fares - A case study of John Wayne Airport

29

Table 10: Linear regression number three (2010:2).

In this case, the variables corresponding to the operator seem to be more significant and the

coefficients are more logical. Indeed, only Continental is more expensive than American Airlines

and Southwest is much cheaper this time. To conclude, we can say that the operator has an

influence on the price of a ticket but much weaker than the influences of the class or the distance.

Page 31: Delays and fares - A case study of John Wayne Airport

30

In order to assess the influences of the class and the distance, we regress the logarithm of the

mean price on a constant, the dummy for the economy class, the distance and the dummy for

competition:

Table 11: Linear regression number four (2010:2).

This model shows that if the distance increases by 100 miles, the price will increase by 2.1%

and if the ticket is a economy class ticket, it is 89% cheaper than a ticket for other classes. The

dummy competition is still insignificant.

Page 32: Delays and fares - A case study of John Wayne Airport

31

3 Delays

3.1 Description of data

For the delays, we use monthly U.S. data from the U.S. Department of Transportation. The

variables in the database are:

• Year;

• Month;

• Carrier: code assigned by IATA and commonly used to identify a carrier;

• Origin airport, IATA code;

• Destination airport, IATA code;

• Actual departure time (local time: hhmm);

• Departure delay: difference in minutes between scheduled and actual departure time (early

departures show negative numbers);

• Arrival time: actual arrival time (local time: hhmm);

• Arrival delay: difference in minutes between scheduled and actual arrival time (early arrivals

show negative numbers);

• Cancelled: cancelled flight indicator (1=Yes) ;

• Diverted: diverted flight indicator (1=Yes);

• Air time: flight time, in minutes;

• Distance between airports, in miles.

Page 33: Delays and fares - A case study of John Wayne Airport

32

In this section, we will see which kind of those variables departure and arrival delays depend

on and we will establish a model of departure delays.

3.2 Delays’ distribution for all flights

In the following parts, we focus on the delays in January 2011 for flights to/from John Wayne

Airport. We have 6862 observations.

Figure 10: Carrier shares in terms of operated flights at John Wayne Airport on January 2011.

Ten carriers operate at John Wayne Airport in January 2011, compared to nine in the second

quarter of 2010 (see figure 3 in fare analysis). Virgin America left SNA in May 2010. SkyWest

Airlines has code sharing agreements with United Airlines and Delta, and Mesa with US Airways,

that is why both carriers do not appear in the fare data. In figure 10, we see that Southwest

Airlines operates more than one third of the traffic from/to John Wayne in January 2011. It is

difficult to make a comparison with figure 3 of analysis fare because the market shares are in terms

of passengers and not of operated flights, then we do not know the shares of SkyWest and Mesa in

the second quarter of 2010. But we observe that Southwest is in the lead, in both market analysis.

Page 34: Delays and fares - A case study of John Wayne Airport

33

Figure 11: Kernel density of departure delays and arrival delays for all departure/arrival flightsto/from John Wayne Airport in January 2011.

We see on figure 11 that the lower tail of arrival delays is to the left to that of the departure

delays. Indeed, pilots can save time in the air, depending on weather conditions. But the right

tails of the density curves seem similar. In figure 12, the regression line of arrival delay against

departure delay, for flights whose departure delay is above the 90% quantile, is almost a bisector.

Then, the big departure delays seem to be reflected in arrival delays.

Page 35: Delays and fares - A case study of John Wayne Airport

34

Figure 12: Scatterplot of arrival delay against departure delay for flights whose departure delay isabove the 90% quantile.

Table 12: Summary statistics for departure and arrival delays at John Wayne Airport in January2011.

Mean Minimum 10% quantile Median 90% quantile Maximum

Departure delays 4.8 -27 -9 -1 24 259 (4h19min)

Arrival delays -0.25 -56 -20 -5 23 259 (4h19min)

The 10% quantile of departure delays is -9 minutes, that means that 10% of flights to or

from John Wayne Airport left 9 minutes or more early. The 90% quantile of arrival delays is 23

minutes, that means that 90% of arrival delays are below 23 minutes. As the mean, the minimum

is lower for arrivals than for departures. The minimum departure delay corresponds to a flight

operated by Frontier Airlines from John Wayne to Denver whereas the arrival delay minimum

is for a flight operated by United Airlines from Chicago-O’Hare to John Wayne. The maximum

for both departure and arrival delays are the same. This concerns the same flight, operated by

SkyWest Airlines, from John Wayne to Denver. Those observations reinforce the conclusions from

Page 36: Delays and fares - A case study of John Wayne Airport

35

the density curve analysis, about saving time during the flight.

3.2.1 Quantile analysis of departure delays

(a) Proportion of carriers below the 10% quantile

of departure delays in January 2011.

(b) Proportion of carriers above the 90% quantile

of departure delays in January 2011.

Figure 13: Carrier shares in the smaller and larger departure delays in January 2011.

Although Southwest Airlines operates 38% of flights to/from John Wayne Airport, the carrier has

no flights among the 10% of flights which leave the earliest, whereas 35% of those flights are operated

by Alaska Airlines, whose share of total traffic on January 2011 is 11%. So there is a big difference

between the proportion of all operated flights by a carrier and its proportion of flights in the 10%

quantile (that is to say, the proportion of the 10% of flights that leaves the earliest, in comparison

with the scheduled departure time). As for the 90% quantile, the difference of proportion is also

Page 37: Delays and fares - A case study of John Wayne Airport

36

important for both carriers. Southwest’s share above the 90% departure quantile is 55%, against

3% for Alaska Airlines.

First, Southwest, as a low-cost carrier, has a very short turnaround time, the scheduled length

to empty, clean and reboard a plane for takeoff. So delays propagate more than if the carrier

had more time for turnaround. Then, the two destinations of Alaska Airlines from John Wayne

are Seattle/Tacoma International Airport and Portland International, which are two hubs of the

airline. So Alaska Airlines may have an important control at those both airports, which allows the

carrier to have lower delays or leave earlier. Besides, 5 out of 8 destinations of Southwest from John

Wayne are in California (Las Vegas, Oakland, San Francisco, San Jose, Sacramento). So distance

may also be an explanatory factor. In figure 14, we observe that the regression line slope of the

scatterplot departure delay against distance is negative. Lower distance involves higher departure

delays.

Figure 14: Scatter plot of departure delay against distance for all flights to/from John Wayne inJanuary 2011.

Page 38: Delays and fares - A case study of John Wayne Airport

37

3.2.2 Quantile analysis of arrival delays

(a) Proportion of carriers below the 10% quantile

of arrival delays in January 2011.

(b) Proportion of carriers above the 90% quantile

of arrival delays in January 2011.

Figure 15: Carrier shares in the smaller and larger arrival delays in January 2011.

We observe the same pattern as for departure delays, for Southwest and Alaska. Southwest has

few flights below the 10% quantile: 11%, against 21%, for Alaska. And Southwest operates 48%

of flights above the 90% quantile, while Alaska operates 3%. Besides, United Airlines operates

24% of flights among the 10% which arrive the earliest, whereas it operated 9% of flights to/from

John Wayne in January 2011. However, in the first 10% quantile of departure delays, it operates

12% of flights. So the earlier arrival of United Airlines are not necessarily the consequence of early

departures. It seems that air time has been saved. United Airlines connects John Wayne to San

Franciso International, Chicago O’Hare International and Denver International. The largest share

of flights below the 10% quantile of arrival delays are flights to/from Chicago and Denver.

Page 39: Delays and fares - A case study of John Wayne Airport

38

3.3 Delay distribution by carrier

When we look at the delay distributions by carrier, we see the same trends as the distribution

of delays for all carriers. The left tail of arrival delays is to the left to that of the departure

delays. The right tails for both departures and arrivals are similar and flat. The arrival delays are

more scattered than the departure delays. However, on the first distributions, for all carriers, the

two right tails are taken together whereas it is not as clear-cut when we only consider flights of

one carrier. Arrival delays’ density is below the departure delays on the right part of the curves.

Departure delays are reduced at arrival.

We see that SkyWest Airlines has very extended left tails. The carrier has big delays in term of

length.

Page 40: Delays and fares - A case study of John Wayne Airport

39

(a) Distribution of arrival/departure delays in

January 2011 for American Airlines.

(b) Distribution of arrival/departure delays in

January 2011 for Alaska Airlines.

(c) Distribution of arrival/departure delays

in January 2011 for Continental Airlines.

(d) Distribution of arrival/departure delays

in January 2011 for Delta Airlines.

(e) Distribution of arrival/departure delays

in January 2011 for Frontier Airlines.

(f) Distribution of arrival/departure delays in

January 2011 SkyWest Airlines.

Figure 16: Delay distribution by carrier.

Page 41: Delays and fares - A case study of John Wayne Airport

40

(a) Distribution of arrival/departure delays

in January 2011 for United Airlines.

(b) Distribution of arrival/departure delays

in January 2011 for US Airways.

(c) Distribution of arrival/departure delays

in January 2011 for SouthwestAirlines.

(d) Distribution of arrival/departure delays

in January 2011 for Mesa Airlines.

Figure 17: Delay distribution by carrier (next carriers).

3.4 Delays by time slot

In this part, we analyse the delays on the scheduled departure time. We add in the variables:

scheduled departure time, that is deduced from the actual departure time and the departure delay.

We see on figure 18 that there are no scheduled departure between 22h and 23h, especially from

John Wayne, although departures are allowed until 23h. The flights scheduled between 23h and 1h

Page 42: Delays and fares - A case study of John Wayne Airport

41

leave from Kahului Airport, Hawaii. Departures and arrival are not allowed before 7am on weekdays

at John Wayne Airport. However, among flights in the slot 6h-7h, we can see several scheduled

departures from John Wayne since 6h45. It must be legal because the scheduled departure time

must be the time when the doors of the aircraft are closed, not when the plane takes off.

Figure 18: Histogram of scheduled departure times at SNA in January 2011.

Figure 19: Mean departure delay by time slot in January 2011.

Page 43: Delays and fares - A case study of John Wayne Airport

42

Figure 20: Maximum departure delay by time slot in January 2011.

The maximum of departure delays between 00h and 1h is the only one that is negative. But

this value is not very significant since there are only 2 flights in this slot. Similarly, there are only

7 flights in the slot of 23h-00h. In the slots 6h-7h and 8h-9h, the mean of departure delays are

negative but there are respectively 377 and 425 flights in those slots. Besides the two figures show

departure delays increasing during the morning, then decreasing in early afternoon before a rise in

the slot 17h-18h, and then they decrease again. The two histograms are quite different from the

one of scheduled departure times’ distribution (figure 18). In the slot 9h-10h, there are many fewer

flights than in the slot 8h-9h but the mean and the maximum departure delay are slightly higher

(figures 19&20). So departure delays do not directly depend on concentration of flights in the slot.

3.5 Model of departure delays

In this part, the objective is to find what affects departure delay and model it as a linear function

of its explanatory variables.

Page 44: Delays and fares - A case study of John Wayne Airport

43

3.5.1 Model for all departure delays

We now develop a model where we suppose that departure delays depend on carrier, time slot and

distance.

We define time slot dummies variables: timeslot 67 , timeslot 78, ..., timeslot 2122, timeslot 2324,

timeslot 01, that correspond to one slot of one hour among those : 6h-6h59, 7h-7h59, ..., 21h-21h59,

23h-23h59, 00h-00h59. Their value is equal to 1 if the scheduled departure time of the flight is in

the slot and 0 otherwise.

We also use carrier dummies : dum aa, dum as, etc... The value of the carrier dummy is equal to

1 if the flight is operated by the carrier and 0 otherwise.

To avoid collinearity between slot dummies and between carrier dummies, we have to omit one

variable among the time slot dummies and one of the carrier dummies, as references. We remove

dum aa and timeslot 1213 from the explanatory variables and we run a linear regression of departure

delays on distance, time slot dummies and carrier dummies (see table 13, at the end of the subpart).

The coefficient of distance is significant. If the distance increases by 1000 miles, departure delays

increases by 1.6 minute.

The coefficients of time slots between 6h and 12h and between 20h to 22h are significant. Besides,

if a flight leaves at a scheduled time in those slots instead of 12h-13h, departure delay decreases.

For instance, if a flight leaves at a scheduled time between 6h and 7h instead of leaving between

12h and 13h, departure delay decreases by 8.7 minutes.

Among the coefficient carrier dummies, dum as, dum oo, dum us, dum wn are significant:

• if a flight is operated by Alaska Airlines instead of American Airlines, the reference carrier,

departure delay decreases by 7.6 minutes;

• if a flight is operated by SkyWest instead of American Airlines, departure delay increases by

8.0 minutes;

Page 45: Delays and fares - A case study of John Wayne Airport

44

• if a flight is operated by US Airways instead of American, departure delay decreases by 2.6

minutes;

• if a flight is operated by Southwest instead of American, departure delay increases by 7.5

minutes.

As for Alaska and Southwest, it reinforces our observations from the 10% and 90% quantile analysis.

But we did not find that SkyWest had more important delays. In figure 16.b), we can see that

SkyWest Airlines operates 8% of the 10% flights that leave the latest, while the carrier operates

5% of flights to/from John Wayne Airport (figure10). So its share in the biggest delays, in term of

number of delays, is a little more important than its traffic share. However, it is the carrier which

has the biggest departure delays in term of length. It has the biggest mean of departure delays.

Besides, in the 10% flights that leave the latest, the mean of departure delays for all carriers is equal

to 52.60 whereas the mean of departure delays for SkyWest Airlines is equal to 74.38. As for US

Airways, its traffic share is 5% (figure10) but its share of the 10% of flights that leave the earliest

is higher, 8%. Moreover, its mean departure delay is negative. US flights leave earlier than the

scheduled time on average. In conclusion, the coefficients dum as, dum oo, dum us and dum wn

are coherent.

R2 = 0.0787, so we explain with this model about 8% of the variability of departure delays. Then

the F-stat is good. We conclude the model explains something.

To improve the model, we gather some slots together and we take as new time dummies:

• dum 612 =1 if scheduled departure between 6h and 11h59;

• dum 1220 =1 if scheduled departure between 12h and 19h59;

• dum 2022 =1 if scheduled departure between 20h and 22h;

• dum 2301 =1 if for scheduled departure between 23h and 1h.

Page 46: Delays and fares - A case study of John Wayne Airport

45

The dummies are equal to 0 in the other cases. We run another linear regression of departure

delays on distance, those time dummies, except dum 1220 as a reference and carrier dumies (table

14). The coefficient of distance is still significant. If the distance increases by 1000 miles, departure

delays increases by 1.8 minute. The coefficient dum 612 and dum 2022 but not dum 2301. So if

a flight is scheduled at departure between 6h and 12h, instead of 12h and 20h, departure delay

decreases by 4.8 minutes. And if a flight is scheduled at departure between 20h and 22h, instead

of 12h and 20h, departure delay decreases by 3.8 minutes. Then, the carrier dummies’ coefficients

which are significant are the same : dum as, dum oo, dum us and dum wn. The influence of those

carriers on departure delays are nearly the same as the previous model. The R-squared decreased

a little to 7.3%.

To have a model that explains more the variability of departure delays. We add another explanatory

variable, a competition dummy: dum comp =1 if more than one carrier operates that route and 0

otherwise . The coefficient of this dummy variable is significant (see table 15). If several carriers

operate the same route, departure delay increases by 1.7 minute. With this new variable, the

carrier dummies’coefficients are more significant. As for distance and time slot, the results do not

change very much. Then, the R-squared number is a little improved, the model explains 7.4% of the

variability of departure delays. Why more delay on routes with competition? Maybe, competition

is a clue that the concerned airports are busy.

Page 47: Delays and fares - A case study of John Wayne Airport

46

Figure 21: Mean departure delays by carrier in January 2011 for flights to/from John Wayne.

Page 48: Delays and fares - A case study of John Wayne Airport

47

Table 13: Regression of departure delays in January 2011, on distance, time slot dummies, carrierdummies.

Page 49: Delays and fares - A case study of John Wayne Airport

48

Table 14: Regression of departure delays in January 2011, on distance, gathered time slot dummies,carrier dummies.

Page 50: Delays and fares - A case study of John Wayne Airport

49

Table 15: Regression of departure delays in January 2011, on distance, gathered time slot dummies,carrier dummies and competition dummy.

Page 51: Delays and fares - A case study of John Wayne Airport

50

Table 16: Regression of departure delays in January 2011, on distance squared, time slot dummies,carrier dummies and competition dummy.

Page 52: Delays and fares - A case study of John Wayne Airport

51

3.5.2 Model of departure delays above the 90% quantile

In this part, we only keep observations whose departure delay is above the 90% quantile. In those

observations, scheduled departure time is between 6h and 22h. The dummy variables timeslot 01

and timeslot 2324 are not useful here. We run a linear regression of departure delay on a constant,

distance, time slot dummies except timeslot 1213 and carrier dummies except dum aa (see table

17). The model explains 9.1% of the variability of departure delays above the 90% quantile. The

coefficient of distance is significant, if the distance increases by 100 miles (161 km), the departure

delay increases by 6.7 minutes. The two significant coefficients of carrier dummies are those of

SkyWest and United Airlines:

• if a flight is operated by SkyWest instead of American Airlines, departure delay increases by

24.1 minutes;

• if a flight is operated by United instead of American Airlines, it increases by 12.7 minutes.

All coefficients of time slot dummies, except dum 2021 and dum 2122, are not significant. So we

run a Wald test: all those coefficients equal to zero. We do not reject this hypothesis at level 80%.

We run another linear regression (table 18) where time slots have been grouped together in 3

slots: 6h-11h59, 12h-16h59 and 17h-22h. The dummy variable dum 1217 is the time slot that has

been removed as a reference. The model explain 7.3% of the variability of departure delays above

the quantile 90%. The coefficient of distance is still significant, if the distance increases by 100

miles (161 km), the departure delay increases by 7.1 minutes. The coefficient of time slots are

significant. If a flight has a departure scheduled between 6h and 12h or between 17h and 22h,

instead of 12h and 17h, departure delay decreases by about 5.6 minutes. The carrier dummies’

coefficients of SkyWest and United Airlines are also the only significant ones:

• if a flight is operated by SkyWest instead of American Airlines, departure delay increases by

Page 53: Delays and fares - A case study of John Wayne Airport

52

22.9 minutes;

• if a flight is operated by United instead of American Airlines, it increases by 10.3 minutes.

On the third regression (table 18), a competition dummy was added. This variable is significant.

If a route is operated by several carriers instead of one, departure delay increases by 8.3 minutes.

We now explain 8.5% of departure delays above the 90% quantile, instead of 7.2% without the

competition dummy. However, the significance of carrier dummies decreases except for Alaska.

The coefficients and significance of the other variables are similar.

Page 54: Delays and fares - A case study of John Wayne Airport

53

Table 17: Regression of departure delays in January 2011 above the 90% quantile, on distance,time slot dummies, carrier dummies.

Page 55: Delays and fares - A case study of John Wayne Airport

54

Table 18: Regression of departure delays in January 2011 above the 90% quantile, on distance,grouped time slot dummies, carrier dummies.

Page 56: Delays and fares - A case study of John Wayne Airport

55

Table 19: Regression of departure delays in January 2011 above the 90% quantile, on distance,grouped time slot dummies, carrier dummies.

Page 57: Delays and fares - A case study of John Wayne Airport

56

3.5.3 Model of departure delays below the 10% quantile

In this part, we only keep observations whose departure delay is below the 10% quantile. In

comparison with the scheduled departure times from the 90% quantile data, there are a few flights

in the slot 0h-1h in addition. As for carriers, Southwest operates no flight in thoses observations.

For the regression on table 8, we do not use the dummies timeslot 1213 and dum aa as references.

The model explains 1.3% of the variability of departure delays below the 10% quantile. In this

case, the coefficient of the distance is non significant and near null. The significant time slot

dummies are: timeslot 67, timeslot 78, timeslot 89, timeslot 1112, timeslot 1314, timeslot 1314,

timeslot 1415, timeslot 1516, timeslot 2021. If a flight is scheduled at departure in those slots,

except 20h-21h, instead of the slot 12h-13h, departure delay increases. The only significant carrier

coefficients are the ones of dum as and dum f9. If a flight is operated by Alaska Airlines, instead

of American Airlines, departure delay decreases by 1.1 minute. If a flight is operated by Frontier

Airlines, instead of American, departure delays decreases of 0.8 minute.

In the second regression, we aggregated slots in this way : 6h-12h, 12h-17h and 17h-1h. The

reference slot is 12h-17h. The only significant time dummy is timeslot 1701. If a flight is scheduled

in this slot instead of the slot 12h-17h, departure delay decreases by 0.9 minute. The R-squared

number is lower : 10.8 % instead of 16%.

Page 58: Delays and fares - A case study of John Wayne Airport

57

Table 20: Regression of departure delays in January 2011 below the 10% quantile, on distance,time slot dummies, carrier dummies.

Page 59: Delays and fares - A case study of John Wayne Airport

58

Table 21: Regression of departure delays in January 2011 below the 10% quantile, on distance,time slot dummies, carrier dummies.

Page 60: Delays and fares - A case study of John Wayne Airport

59

3.6 Other explanatory factors of delay

Other factors that can be taken into account are12:

• precipitation;

• season;

• weekday;

• arrival time;

• spacing between 2 successive arriving flights (if it increases then delay decreases);

• turnaround chain...

Here are the different causes of delay listed by the Bureau of Transportation Statistics:

• Air carrier (maintenance or crew problems, aircraft cleaning, baggage loading, fueling, etc.);

• Extreme Weather: Significant meteorological conditions (actual or forecasted) that, in the

judgment of the carrier, delays or prevents the operation of a flight such as tornado, blizzard

or hurricane;

• Late-arriving aircraft: A previous flight with same aircraft arrived late, causing the present

flight to depart late;

• Delays and cancellations attributable to the national aviation system that refer to a broad

set of conditions, such as non-extreme weather conditions, airport operations, heavy traffic

volume, and air traffic control;

• Security: Delays or cancellations caused by evacuation of a terminal or concourse, re-boarding

of aircraft because of security breach, inoperative screening equipment and/or long lines in

excess of 29 minutes at screening areas.

Page 61: Delays and fares - A case study of John Wayne Airport

60

The share of each delay cause from 2003 to 2010 are shown in the following figure and table.

Page 62: Delays and fares - A case study of John Wayne Airport

61

4 The merge

4.1 2010:2

4.1.1 Data description

We merge the two databases, fares and delays, in order to analyse the possible influence of delays

on ticket fares. One observation corresponds to one route with one class, operated by one carrier.

We merge the databases for the 2nd quarter of 2010. In this new database, we have the variables:

• Origin airport, identified by its three letter IATA code;

• Destination airport, identified by its three letter IATA code;

• The length of the route, in miles;

• The carrier which operated the flight, identified by its two letter IATA code;

• The marketshare that the carrier has on this route, in percentage;

• The ticketclass;

• The number of passengers;

• The mean price by passenger;

• The mean departure delay;

• The 90% quantile of departure delays;

• The mean arrival delay;

• The 90% quantile of arrival delays.

Page 63: Delays and fares - A case study of John Wayne Airport

62

We did a preliminary work on each database before merging. For fares’ database, we had to

gather fare classes together because each class has two possible letters:

• Economy class: ’X’ and ’Y’, we keep ’X’;

• Business class: ’F’ and ’G’, we keep ’F’;

• First class: ’C’ and ’D’, we keep ’C’.

So, we added the number of passengers and averaged the price, weight by the number of passengers.

As for the delay database, we computed for each route by carrier: the mean and the 90% quantile

for both departure and arrival delays.

After this preliminary work, we merge the databases for each route operated by each carrier. With

the merge, we lose a lot of observations. Some routes were in delay database and not in fare database

and conversely. Some carriers do not appear in fare database at all: that is the case for SkyWest

Airlines and Mesa Airlines because these carriers operate their flights for other carriers whereas fare

database does not consider code-sharing. At the contrary, some routes which are in fares database,

are not in delay database. After asking to the U.S. Department of Transportation, we reach the

conclusion that it was a problem of definition of “direct route” between databases. Indeed, the

fare database may take two flights with the same flight number as a direct route whereas the delay

database does not. That is not a big issue after all because, even if we lose 52% of observations, that

represents only 19400 passengers in comparison with a total of 450000 passengers, namely 4.3%.

More over, there are no data for the carrier Virgin America in the delay database, which only

includes data of carriers that represent at least 1% of the domestic, scheduled-service passenger

revenue.

Page 64: Delays and fares - A case study of John Wayne Airport

63

4.1.2 Descriptive linear regressions

As for the previous regressions, we use a dummy variable for economy class, called here ”dumx”

and dummies for carriers, identified by the two letter carrier IATA code. We run a regression of

price on a constant, distance, dumx, carrier dummies and Q90 depdelay.

Table 22: Regression of price on a constant, distance, dumx and carrier dummies on quarter 22010.

The significant coefficients are the ones of the constant, the economy class dummy and Alaska

Page 65: Delays and fares - A case study of John Wayne Airport

64

Airlines’ dummy. So we can make an interpretation of those coefficients. As for the coefficient of

economy dummy, if the ticket is an economy class one instead of business or first, then the price

decreases by $1046. If the distance increases by 100 miles, the ticket price increases by 20.8 dollars.

Then, if a flight is operated by Alaska instead of American Airlines, the price decreases by 516

dollars.

There is little evidence that the other carrier dummies are significant relative to American. But

as the signs of coefficients are not the ones we expected. Indeed, in this regression, United and

Southwest would be more expensive than American whereas Southwest is a low-cost. Moreover,

for example, US Airways would be more expensive than Frontier.

The Q90 delay variable has the sign that we would expect, i.e. greater delays associated with lower

fares; quantitatively, an additional 10 minutes of Q90 (big) delay would reduce the average fare

by $46, controlling for class, distance and carrier. Forbes, who examines delays on flights out of

LaGuardia (LGA) using different methods[11], finds that prices fall by $1.42 on average for direct

passengers for one minute in additionnal delay, namely an additional 10 minutes of delay would

reduce the average fare by $14.2. The order of magnitude is not very similar but here we just have

2010:2 whereas Forbes considers all data before 2007 and he take an average of all delays whereas

we consider only big delays.

The variable 90% quantile of departure delay has a p-value of 22%, which is little high to prove

the significance of the variable. There is negative autocorrelation, so we do a Breusch-Godfrey test

whose result is: we reject the hypothesis of no autocorrelation in residuals of order 1 at a level 95%.

We run again the regression with Newey-West standard errors and the coefficient of Q90 depdelay

is now significant.

R squared is 82%. As we saw with the analyse of tables from 8 to 10, this statistical ”power”

comes from the economy dummy and distance whereas carrier dummies have a very small influence.

Page 66: Delays and fares - A case study of John Wayne Airport

65

Table 23: Regression of price on a constant, distance, dumx, carrier dummies and Q90 depdelaywith Newey-West standard errrors on quarter 2 2010.

Page 67: Delays and fares - A case study of John Wayne Airport

66

Table 24: Regression of price on a constant, distance, dumx and Q90 depdelay on quarter 2 2010.

4.2 Fares and delays from 1999 to 2010

In this part, we study the impact of delays on fares from 1999 to 2010. The databases of fares and

delays were merged for each quarter from 1999 to 2010, except quarter 4 of 2010, as we did for the

quarter 2 of 2010 (see the PYTHON code of the merge in annex). We have 4905 observations. Each

observation still corresponds to one route by carrier and fareclass. We have the same variables as in

the study of quarter 2 2010 with an additional variable: TIME. TIME corresponds to one quarter,

it takes value between 1 and 47.

Page 68: Delays and fares - A case study of John Wayne Airport

67

4.2.1 Departure delays on time

On figure 22, we observe the departure delays above the 90% quantile against time. We see there

are not delays above 30-40 minutes from the observations 17 to 20, which correspond to the year

2003. In that year, the amendment to the Settlement Agreement in 2003 enabled the airport to

increase its capacity and its operations. So the aiport has more means to operate but the trafic did

not increase in one shot. After 2003, higher and higher 90% quantiles of departure delays appear.

That is because the traffic has increased during years. It would be interesting to see this graph in

few years to see the impact of the new terminal opening in November 2011.

Figure 22: 90% quantile of departure (minutes) delay against time (quarters).

To check wether if the off-peak is significant or not, we make a regression of 90% quantile

of departure above 20 minutes delays on a constant, time and time2. We are able to determine

Page 69: Delays and fares - A case study of John Wayne Airport

68

the impact of the regulation on substantial delays. We plot the fitted values of the regression

superimposed on the previous scatterplot, without values below 20 minutes (see figure 23)

Figure 23: 90% quantile of departure (minutes) delay against time (quarters) withoutQ90 dep delay less than 20 minutes.

4.2.2 Baseline model

We run a first regression of prices on a constant, compet, distance, eco, q90 depdelay and time. On

this regression, the Durbin-Watson value is 2.119535. So we do a Breusch-Godfrey test with null

hypothesis : no autocorrelation in errors of order 1. obs ∗R2 = 19.16 where as chi-square(1)=3.841

at level 95%, so we reject the null hypothesis. As there is autocorrelation, we run another regression

with NeweyWest standard errors (see table 25):

Page 70: Delays and fares - A case study of John Wayne Airport

69

PRICEi = 435.26 + 170.70COMPETi + 0.39DISTANCEi − 513.72ECOi − 1.23Q90 DEP DELAYi

− 2.62TIMEi (1)

All coefficients of explanatory variables are significant. If a route is operated by several carriers,

the price increases by 170.70 dollars. Here the competition does not make the price decrease as we

might expect. The carriers may want to operate highly demanded routes. Although several carrier

share a route, the offer may still be less than the demand. A reason of that could be the restriction

of operation access at SNA.

If the distance on a route increases by 100 miles, the price inceases by 39 dollars.

If the flight ticket corresponds to an economy class one, the price decreases by 513.72 dollars.

If the 90% quantile increases by one minute one minute, the price decreases by 1.23 dollars.

The price decreases by 2.62 dollars in the time by quarter. The aviation sector observes a trafic

growth years after years, it corresponds to the democratization of the travel by plane.

According the F-statistic, the model explains something. And by looking at the R-squared value,

the model explains 39 % of the variability of the price.

Page 71: Delays and fares - A case study of John Wayne Airport

70

Table 25: Regression of prices on a constant, compet, distance, eco, q90 depdelay and time withNewey-West standard errors.

4.2.3 Variation of prices over time

In this part, we are interested in the evolution of price over time. We run a regression of price

on a constant, compet distance, eco, Q90 dep delay and time dummies for each quarter taking the

first quarter as the reference (cf Table 26). Most of time dummies are significant and F-statistic

Page 72: Delays and fares - A case study of John Wayne Airport

71

is good. In order to see the variation of price, we plot the coefficients of time dummies from the

regression against time.

Figure 24: Coefficients of time dummies against time.

With regard to quarter 1 1999, the prices decrease. There is a peak-off at quarter 12, the

quarter after September 11, 2001. After this event, the variations are not constant until 2004

(quarter 22). Prices decrease significantly in 2004. In 2003, the second Gulf War and the SRAS

crises caused a drop of the trafic. Severals carriers including American Airlines, United and US

Airways, were almost in bankrupcy. So these important events prevent us from measuring the

influence of deregulation of 2003. It is only in 2007 that their financial status improved. We see

that after 2007, prices increase steadly. In 2009, the subprime crisis makes the prices fall (peak-off

on quarter 43). Then the prices increase again with the increase of oil. In 2010, the price is quite

the same of 1999.

Page 73: Delays and fares - A case study of John Wayne Airport

72

Table 26: Regression of prices on a constant, compet, distance, eco, q90 depdelay and time dummiesfor each quarter with Newey-West standard errors.

Page 74: Delays and fares - A case study of John Wayne Airport

73

4.2.4 Delay effect on competitive market

In this part, we want to measure the impact of delays on fares in a competitive market. We run a

regression of price on a constant, compet distance, eco, Q90 dep delay, time and Q90 dep delay*compet

(cf Table 27).

All coefficients are significant and their sign did not change in comparison with our first model.

Indeed, the quantitative results on compet, distance, eco, q90 dep delay and time are quite the

same.

As for the new variable Q90 dep delay*compet, the effect of delays does not get larger in competitive

market as Forbes found for La Guardia airport[11]. When we are in a monopoly market, one minute

of delays involves a decrease by 4.25 dollars of price. And on competitive market, if the 90% quantile

increases by one minute, price still increases by 0.75 dollar.

Page 75: Delays and fares - A case study of John Wayne Airport

74

Table 27: Regression of prices on a constant, compet, distance, eco, q90 depdelay, time andQ90 dep delay*compet with Newey-West standard errors.

4.2.5 Model including the influence of the carrier

In this part, we study the influence of the carriers. We run a regression on a constant, distance, eco,

Q90 dep delay, time and carrier dummies, American Airlines as the reference (cf Table 28). The

coefficients of the first model are still quantitatively the same. All coefficients of carrier dummies,

except US Airways and Southwest ones, are significant. All the carriers are cheaper than American

Airlines,except Continental Airlines. That is difficult to interpret because they do not operate the

Page 76: Delays and fares - A case study of John Wayne Airport

75

same route and they are not present in SNA on the same quarters.

Page 77: Delays and fares - A case study of John Wayne Airport

76

Table 28: Regression of prices on a constant, compet, distance, eco, q90 depdelay, time and carrierdummies with Newey-West standard errors.

Page 78: Delays and fares - A case study of John Wayne Airport

77

4.2.6 Model with a maximum of explanatory variables

We run a regression of price on a cosntant, compet, distance, distance2, eco, Q90 dep delay,

Q90 dep delay*compet, time dummies and carrier dummies (cf Table 29). The coefficient of compet

changes in comparison with our baseline model. It is not significant anymore. On the other hand,

in this model we see that departure delay has an impact only in monopoly market. Indeed, in a

monopoly market, when the 90% quantile of departure delays increases by on minute, the ticket

fare decreases by 4.57 dollars, whereas in a competitive market, the coefficients of Q90 dep delay

and Q90 dep delay*compet nullify according to a Wald test, delays have no impact.

Page 79: Delays and fares - A case study of John Wayne Airport

78

Table 29: Regression of prices on a constant, compet, distance, distance2, eco, Q90 dep delay,Q90 dep delay*compet, time dummies and carrier dummies with Newey-West standard errors.

4.2.7 Conclusion

All the linear regressions strenghten the robustness of our baseline model. In the survey of Forbes

for La Guardia, price falls by $1.42 for direct passengers when departure delay increases by one

minute. In our model for John Wayne, when the 90% quantile of departure delay increases by one

Page 80: Delays and fares - A case study of John Wayne Airport

79

minute, the ticket fare decreases by $1.23. By including the data for 47 quarters, we gain precision.

Page 81: Delays and fares - A case study of John Wayne Airport

80

5 Final conclusion

We would like to analyse the possible variations of price or delays with the changes of regulation

at John Wayne Airport. However, the amendment of the regulation between 1999 and 2010 was in

2003, which matches with major external factors such as Gulf War. So we have focused the study

on the impact of delays on price. Our result of the price decreasing regarding the delay is similar

to Forbes’s academic paper[11]. On the other hand, our result about the impact of competition

is different. Unlike Forbes, the competition does not make decrease fares, maybe because John

Wayne has one of the strictest regulation in the world. It could be interesting in a few years to see

the impact of the new terminal at SNA.

Page 82: Delays and fares - A case study of John Wayne Airport

81

A Appendix

Figure 25: The location of John Wayne Airport in Orange County, California. Source: John WayneAirport website, Map of Orange County (http://www.ocair.com/directions/).

Page 83: Delays and fares - A case study of John Wayne Airport

82

Figure 26: Noise Monitoring Station map at John Wayne Airport in 2008. Source: Boeing,Airport Noise and Emission Regulations (http://www.boeing.com/commercial/noise/john_wayne.html).

Page 84: Delays and fares - A case study of John Wayne Airport

83

Figure 27: Annual CNEL (Community Noise Equivalent Level) Noise Contours map in2010. Source: John Wayne Airport website (http://www.ocair.com/reportspublications/AccessNoise/default.aspx).

Page 85: Delays and fares - A case study of John Wayne Airport

84

Table 30: Relative severity of noise levels. Sources: Aeroport&Environnement course, D.Vergezand Le portail citoyen sonores (http://www.moinsdebruit.com/le-bruit/lechelle-du-bruit.html).

Page 86: Delays and fares - A case study of John Wayne Airport

85

B List of IATA codes

B.1 Airports

ATL (Atlanta), BDL (Bradley), BOI (Boise), CVG (Cincinnati), DAL (Dallas-Love Field), DCA

(Washington), DEN (Denver), DFW (Dallas-Fort Worth), DTW (Detroit), EWR (Newark), GEG

(Washington-Spokane), HOU (Houston-Hobby), HNl (Honululu), IAH (Houston-Intercontinental),

JFK (John F. Kennedy International Airport of New York), LAS (Las Vegas), LAX (Los Angeles

International), MIA (Miami), MCO (Orlando), MDW (Chicago-Midway), MHT (Boston), MKE

(Milwaukee), MSP (Minneapolis-St Paul), OAK (Oakland), OMA (Omaha), ONT (Ontario International),

OGG (Kahului), ORD (Chicago-O’Hare), PDX (Portland), PHL (Philadelphia), PHX (Phoenix),

RDU (Raleigh-Durham), RNO (Reno-Tahoe), SAT (San Antonio), SEA (Seattle-Tacoma), SJC

(San Jose), SFO (San Francisco), SLC (Salt Lake City), SMF (Sacramento), SNA (John Wayne

Airport), STL (St Louis), TUS (Tucson), YVR (Vancouver), YYC (Calgary).

B.2 Airlines

AA (American Airlines), AQ (Aloha Airlines), AS (Alaska Airlines), CO (Continental Airlines), DL

(Delta Airlines), FX (FedEx), F9 (Frontier Airlines), HP (America West Airlines), NW (NorthWest

Airline), OO (SkyWest Airlines), TW (Trans World Airlines), UA (United Airlines), US (US

Airways), WN (Southwest Airlines), WS (WestJet), YV (Mesa Airlines),5X (UPS).

Page 87: Delays and fares - A case study of John Wayne Airport

86

C Code

C.1 Gathering months together for each quarter

de f parseDelayData ( year , month) :

import csv

v a r i a b l e L i s t =[ ’YEAR’ , ’QUARTER’ , ’MONTH’ , ’CARRIER ’ , ’ORIGIN ’ ,\

’DEST ’ , ’DEP TIME ’ , ’DEP DELAY ’ , ’ARR TIME ’ , ’ARR DELAY ’

,\

’CANCELLED’ , ’DIVERTED ’ , ’AIR TIME ’ , ’DISTANCE ’ ]

f i l ename=’ / Users / e l s a d e l p e c h /Desktop/ de lay data / de l ay ’+s t r ( year )+’

’+s t r (month)+’ . csv ’

f=open ( f i l ename , ’ rb ’ ) ; dataReader=csv . reader ( f , d e l i m i t e r=’ ; ’ )

f o r header in dataReader : break

a l l V a r i a b l e s L i s t=header [ 0 ] . s p l i t ( ’ , ’ )

a l l V a r i a b l e s L i s t =[ ]

f o r var in a l l V a r i a b l e s L i s t :

var =var . s p l i t ( ’ ” ’ )

i f l en ( var )==3 and var [0]== var [−1]== ’ ’ : v a r =var [ 1 ]

e l i f l en ( var )==1: v a r =var [ 0 ]

Page 88: Delays and fares - A case study of John Wayne Airport

87

e l s e : pass

a l l V a r i a b l e s L i s t . append ( v a r )

headerDict={}

f o r i in range ( l en ( a l l V a r i a b l e s L i s t ) ) :

i f a l l V a r i a b l e s L i s t [ i ] != ’ ’ : headerDict [ a l l V a r i a b l e s L i s t [ i ] ]= i

output=’ ’

f o r name in v a r i a b l e L i s t :

output+=name

output+=’ \ t ’

output+=’ \n ’

f o r row in dataReader :

rowList=row [ 0 ] . s p l i t ( ’ , ’ )

rowList . pop (17)

rowList . pop (24)

i f rowList [ headerDict [ ’ORIGIN ’ ]]== ’ ”SNA” ’ or rowList [ headerDict

[ ’DEST ’ ]]== ’ ”SNA” ’ :

i f rowList [ headerDict [ ’DEP TIME ’ ] ] ! = ’ ”” ’ and rowList [

headerDict [ ’ARR TIME ’ ] ] ! = ’ ”” ’ :

f o r v a r i a b l e in v a r i a b l e L i s t :

output+=rowList [ headerDict [ v a r i a b l e ] ]

output+=’ \ t ’

output+=’ \n ’

Page 89: Delays and fares - A case study of John Wayne Airport

88

f . c l o s e ( )

g=open ( ’ / Users / e l s a d e l p e c h /Desktop/ output / de l ay ’+s t r ( year )+’ ’+s t r

(month)+’ . txt ’ , ’w ’ )

g . wr i t e ( output ) ; g . c l o s e ( )

re turn None

de f c o l l e r m o i s ( year ) :

f o r i in range (1 , 13 ) :

parseDelayData ( year , i )

f o r j in range (1 , 5 ) :

f i l ename=’ / Users / e l s a d e l p e c h /Desktop/ output / de l ay ’+s t r ( year )+’

’+s t r (3∗ j−2)+’ . txt ’

f=open ( f i l ename , ’ r ’ )

header=f . r e a d l i n e ( ) . s t r i p ( ) . s p l i t ( )

d a t a d i c t j=d i c t ( [ x , [ ] ] f o r x in header )

l i g n e s=f . r e a d l i n e s ( )

f o r k in range ( l en ( l i g n e s ) ) :

l=l i g n e s [ k ] . s t r i p ( ) . s p l i t ( ”\ t ” )

f o r i in range ( l en ( header ) ) :

i f l [ i ]== ’ ’ : d a t a d i c t j [ header [ i ] ] . append ( ’NA’ )

e l s e : d a t a d i c t j [ header [ i ] ] . append ( eva l ( l [ i ] ) )

Page 90: Delays and fares - A case study of John Wayne Airport

89

f . c l o s e ( )

f o r month in range (3∗ j −1 ,3∗ j +1) :

f i l ename=’ / Users / e l s a d e l p e c h /Desktop/ output / de l ay ’+s t r (

year )+’ ’+s t r (month)+’ . txt ’

f=open ( f i l ename , ’ r ’ )

f . r e a d l i n e ( )

l i g n e s=f . r e a d l i n e s ( )

f o r k in range ( l en ( l i g n e s ) ) :

l=l i g n e s [ k ] . s t r i p ( ) . s p l i t ( ”\ t ” )

f o r i in range ( l en ( header ) ) :

i f l [ i ]== ’ ’ : d a t a d i c t j [ header [ i ] ] . append ( ’NA’ )

e l s e : d a t a d i c t j [ header [ i ] ] . append ( eva l ( l [ i ] ) )

f . c l o s e ( )

z=open ( ’ / Users / e l s a d e l p e c h /Desktop/ output / de l ay ’+s t r ( year )+’

qua r t e r ’+s t r ( j )+’ . txt ’ , ’w ’ )

z . wr i t e ( ’YEAR \ t QUARTER \ t MONTH \ t CARRIER \ t ORIGIN \ t DEST

\ t DEP TIME \ t DEP DELAY \ t ARR TIME \ t ARR DELAY \ t

CANCELLED \ t DIVERTED \ t AIR TIME \ t DISTANCE \n ’ )

f o r i in range ( l en ( d a t a d i c t j [ ’ORIGIN ’ ] ) ) :

z . wr i t e ( ’%s \ t %s \ t %s \ t ”%s ” \ t ”%s ” \ t ”%s ” \ t ”%s ” \ t

%s \ t ”%s ” \ t %s \ t %s \ t %s \ t %s \ t %s \n ’%(

d a t a d i c t j [ ’YEAR’ ] [ i ] , d a t a d i c t j [ ’QUARTER’ ] [ i ] ,

d a t a d i c t j [ ’MONTH’ ] [ i ] , d a t a d i c t j [ ’CARRIER ’ ] [ i ] ,

Page 91: Delays and fares - A case study of John Wayne Airport

90

d a t a d i c t j [ ’ORIGIN ’ ] [ i ] , d a t a d i c t j [ ’DEST ’ ] [ i ] ,

d a t a d i c t j [ ’DEP TIME ’ ] [ i ] , d a t a d i c t j [ ’DEP DELAY ’ ] [ i ] ,

d a t a d i c t j [ ’ARR TIME ’ ] [ i ] , d a t a d i c t j [ ’ARR DELAY ’ ] [ i ] ,

d a t a d i c t j [ ’CANCELLED’ ] [ i ] , d a t a d i c t j [ ’DIVERTED ’ ] [ i ] ,

d a t a d i c t j [ ’AIR TIME ’ ] [ i ] , d a t a d i c t j [ ’DISTANCE ’ ] [ i ] ) )

z . c l o s e ( )

re turn None

C.2 Merging all the quarters and gathering quarters together

de f p e r c e n t i l e (N, percent , key=lambda x : x ) :

import math

import f u n c t o o l s

i f not N:

re turn None

k = ( l en (N)−1) ∗ percent

f = math . f l o o r ( k )

c = math . c e i l ( k )

i f f == c :

re turn key (N[ i n t ( k ) ] )

d0 = key (N[ i n t ( f ) ] ) ∗ ( c−k )

d1 = key (N[ i n t ( c ) ] ) ∗ (k−f )

r e turn d0+d1

Page 92: Delays and fares - A case study of John Wayne Airport

91

de f mean( numberList ) :

i f l en ( numberList ) == 0 :

re turn f l o a t ( ’ nan ’ )

floatNums = [ f l o a t ( x ) f o r x in numberList ]

r e turn sum( floatNums ) / l en ( numberList )

de f f u s i o n q u a r t e r ( year , quarte r ) :

f i l ename=’ / Users / e l s a d e l p e c h /Desktop/ output / de l ay ’+s t r ( year )+’

qua r t e r ’+s t r ( quarte r )+’ . txt ’

f=open ( f i l ename , ’ r ’ )

header=f . r e a d l i n e ( ) . s t r i p ( ) . s p l i t ( )

d a t a d i c t=d i c t ( [ x , [ ] ] f o r x in header )

l i g n e s=f . r e a d l i n e s ( )

f o r k in range ( l en ( l i g n e s ) ) :

l=l i g n e s [ k ] . s t r i p ( ) . s p l i t ( ”\ t ” )

f o r i in range ( l en ( header ) ) :

t ry :

i f l [ i ]== ’ ’ : d a t a d i c t [ header [ i ] ] . append ( ’NA’ )

e l i f l [ i ]== ’NA’ : d a t a d i c t [ header [ i ] ] . append ( ’NA’ )

e l s e : d a t a d i c t [ header [ i ] ] . append ( eva l ( l [ i ] ) )

except NameError :

Page 93: Delays and fares - A case study of John Wayne Airport

92

d a t a d i c t [ header [ i ] ] . append ( ’NA’ )

f . c l o s e ( )

d de l ay keys =[ ’YEAR’ , ’QUARTER’ , ’CARRIER ’ , ’ORIGIN ’ , ’DEST ’ , ’

MEAN DEP DELAY ’ , ’Q90 DEP DELAY ’ , ’MEAN ARR DELAY’ , ’Q90 ARR DELAY ’

]

d de lay merge=d i c t ( [ x , [ ] ] f o r x in d de l ay keys )

l i s t e r o u t e =[ ]

f o r k in range ( l en ( d a t a d i c t [ ’ORIGIN ’ ] ) ) :

l i s t e r o u t e . append ( [ d a t a d i c t [ ’ORIGIN ’ ] [ k ] , d a t a d i c t [ ’DEST ’ ] [ k

] , d a t a d i c t [ ’CARRIER ’ ] [ k ] ] )

l i s t e r o u t e . s o r t ( )

n b f l i g h t s =[ ]

f o r k in range ( l en ( d a t a d i c t [ ’ORIGIN ’ ] ) ) :

n b f l i g h t s . append ( l i s t e r o u t e . count ( l i s t e r o u t e [ k ] ) )

k=0

j=0

whi le k<l en ( n b f l i g h t s ) :

f o r i in range ( n b f l i g h t s [ k ]−1) :

l i s t e r o u t e . remove ( l i s t e r o u t e [ j ] )

k=k+n b f l i g h t s [ k ]

j=j+1

Page 94: Delays and fares - A case study of John Wayne Airport

93

f o r k in range ( l en ( l i s t e r o u t e ) ) :

d de lay merge [ ’YEAR’ ] . append ( year )

d delay merge [ ’QUARTER’ ] . append ( quarte r )

d de lay merge [ ’ORIGIN ’ ] . append ( l i s t e r o u t e [ k ] [ 0 ] )

d de lay merge [ ’DEST ’ ] . append ( l i s t e r o u t e [ k ] [ 1 ] )

d de lay merge [ ’CARRIER ’ ] . append ( l i s t e r o u t e [ k ] [ 2 ] )

l i s t d e p d e l a y =[ ]

l i s t a r r d e l a y =[ ]

f o r i in range ( l en ( d a t a d i c t [ ’ORIGIN ’ ] ) ) :

i f d a t a d i c t [ ’ORIGIN ’ ] [ i ]== l i s t e r o u t e [ k ] [ 0 ] and d a t a d i c t [

’DEST ’ ] [ i ]== l i s t e r o u t e [ k ] [ 1 ] and d a t a d i c t [ ’CARRIER ’ ] [ i

]== l i s t e r o u t e [ k ] [ 2 ] :

l i s t d e p d e l a y . append ( d a t a d i c t [ ’DEP DELAY ’ ] [ i ] )

l i s t a r r d e l a y . append ( d a t a d i c t [ ’ARR DELAY ’ ] [ i ] )

l i s t d e p d e l a y . s o r t ( )

l i s t a r r d e l a y . s o r t ( )

f o r i in range ( l i s t d e p d e l a y . count ( ’NA’ ) ) :

l i s t d e p d e l a y . remove ( ”NA” )

f o r i in range ( l i s t a r r d e l a y . count ( ’NA’ ) ) :

l i s t a r r d e l a y . remove ( ”NA” )

d delay merge [ ’MEAN DEP DELAY ’ ] . append (mean( l i s t d e p d e l a y ) )

d de lay merge [ ’MEAN ARR DELAY’ ] . append (mean( l i s t a r r d e l a y ) )

d de lay merge [ ’Q90 DEP DELAY ’ ] . append ( p e r c e n t i l e ( l i s t d e p d e l a y

, percent =0.9) )

Page 95: Delays and fares - A case study of John Wayne Airport

94

d delay merge [ ’Q90 ARR DELAY ’ ] . append ( p e r c e n t i l e ( l i s t a r r d e l a y

, percent =0.9) )

re turn d delay merge

de f wi th index ( seq ) :

f o r i in xrange ( l en ( seq ) ) :

y i e l d i , seq [ i ]

de f r e p l a c e a l l ( seq , obj , replacement ) :

f o r i , elem in with index ( seq ) :

i f elem==obj :

seq [ i ]= replacement

de f ouv r i r ( year , quarte r ) :

import cPick le , copy

path=”/ Users / e l s a d e l p e c h /Desktop/ Pro j e t 3A/ data / Data Fina l ”+

s t r ( year )+” ”+s t r ( quarte r )+” . bin ”

f=open ( path , ’ r ’ )

d=cP i ck l e . load ( f )

f . c l o s e ( )

Page 96: Delays and fares - A case study of John Wayne Airport

95

d2={}

f o r k in d : d2 [ k ] = [ ]

f o r i in range ( l en (d [ ’ Or ig in ’ ] ) ) :

i f d [ ’ Or ig in ’ ] [ i ]== ’SNA ’ or d [ ’ Dest ’ ] [ i ]== ’SNA ’ :

f o r k in d : d2 [ k ] . append (d [ k ] [ i ] )

d3=copy . deepcopy ( d2 )

f o r j in range ( l en ( d2 [ ’ Or ig in ’ ] ) ) :

d4 =[ ]

f o r l in range ( l en ( d2 [ ’ I t i nFar e ’ ] [ j ] ) ) :

i f d2 [ ’ I t i nFar e ’ ] [ j ] [ l ]>20 and d2 [ ’ I t i nFar e ’ ] [ j

] [ l ]<10000:

d4 . append ( d2 [ ’ I t i nFar e ’ ] [ j ] [ l ] )

d3 [ ’ I t i nFar e ’ ] [ j ]=d4 [ : ]

demande1={}

demande1 [ ’ Or ig in ’ ] = [ ]

demande1 [ ’ Dest ’ ] = [ ]

demande1 [ ’ Distance ’ ] = [ ]

demande1 [ ’ OpCarrier ’ ] = [ ]

demande1 [ ’ MarketShare ’ ] = [ ]

demande1 [ ’ FareClass ’ ] = [ ]

demande1 [ ’Pax ’ ] = [ ]

Page 97: Delays and fares - A case study of John Wayne Airport

96

demande1 [ ’ Pr i ce ’ ] = [ ]

demande1 [ ’ Or ig in ’ ]=d2 [ ’ Or ig in ’ ]

demande1 [ ’ Dest ’ ]=d2 [ ’ Dest ’ ]

demande1 [ ’ Distance ’ ]=d2 [ ’ Distance ’ ]

demande1 [ ’ OpCarrier ’ ]=d2 [ ’ OpCarrier ’ ]

demande1 [ ’ MarketShare ’ ]=d2 [ ’ MarketShare ’ ]

demande1 [ ’ FareClass ’ ]=d2 [ ’ FareClass ’ ]

f o r i in range ( l en ( d2 [ ’ Or ig in ’ ] ) ) :

demande1 [ ’Pax ’ ] . append (10∗ l en ( d3 [ ’ I t i nFar e ’ ] [ i ] ) )

f o r i in range ( l en ( d3 [ ’ Or ig in ’ ] ) ) :

i f l en ( d3 [ ’ I t i nFar e ’ ] [ i ] ) ==0:

demande1 [ ’ Pr i ce ’ ] . append ( d3 [ ’ I t i nFar e ’ ] [ i ] )

e l s e :

demande1 [ ’ Pr i ce ’ ] . append ( f l o a t (sum( d3 [ ’ I t i nFar e

’ ] [ i ] [ : ] ) ) / f l o a t ( l en ( d3 [ ’ I t i nFar e ’ ] [ i ] ) ) )

d i co={}

dico [ ’ Or ig in ’ ] = [ ]

d i co [ ’ Dest ’ ] = [ ]

d i co [ ’ Distance ’ ] = [ ]

d i co [ ’ OpCarrier ’ ] = [ ]

d i co [ ’ MarketShare ’ ] = [ ]

Page 98: Delays and fares - A case study of John Wayne Airport

97

d ico [ ’ FareClass ’ ] = [ ]

d i co [ ’Pax ’ ] = [ ]

d i co [ ’ Pr i ce ’ ] = [ ]

r e p l a c e a l l ( demande1 [ ’ Pr i ce ’ ] , [ ] , 0 . 0 )

i=0

whi le ( i<l en ( demande1 [ ’ Or ig in ’ ] )−1) :

i f demande1 [ ’ Or ig in ’ ] [ i ]==demande1 [ ’ Or ig in ’ ] [ i +1] and

demande1 [ ’ Dest ’ ] [ i ]==demande1 [ ’ Dest ’ ] [ i +1] and

demande1 [ ’ OpCarrier ’ ] [ i ]==demande1 [ ’ OpCarrier ’ ] [ i +1]

and demande1 [ ’ FareClass ’ ] [ i ]== ’X ’ and demande1 [ ’

FareClass ’ ] [ i +1]== ’Y ’ :

i f demande1 [ ’Pax ’ ] [ i ]+demande1 [ ’Pax ’ ] [ i +1]==0:

d i co [ ’ Or ig in ’ ] . append ( demande1 [ ’ Or ig in ’ ] [ i ] )

d i co [ ’ Dest ’ ] . append ( demande1 [ ’ Dest ’ ] [ i ] )

d i co [ ’ Distance ’ ] . append ( demande1 [ ’ Distance ’ ] [ i

] )

d i co [ ’ OpCarrier ’ ] . append ( demande1 [ ’ OpCarrier ’ ] [

i ] )

d i co [ ’ MarketShare ’ ] . append ( demande1 [ ’

MarketShare ’ ] [ i ] )

d i co [ ’ FareClass ’ ] . append ( demande1 [ ’ FareClass ’ ] [

i ] )

Page 99: Delays and fares - A case study of John Wayne Airport

98

d ico [ ’Pax ’ ] . append ( demande1 [ ’Pax ’ ] [ i ]+demande1 [

’Pax ’ ] [ i +1])

d i co [ ’ Pr i ce ’ ] . append (0 )

e l s e :

d i co [ ’ Or ig in ’ ] . append ( demande1 [ ’ Or ig in ’ ] [ i ] )

d i co [ ’ Dest ’ ] . append ( demande1 [ ’ Dest ’ ] [ i ] )

d i co [ ’ Distance ’ ] . append ( demande1 [ ’ Distance ’ ] [ i

] )

d i co [ ’ OpCarrier ’ ] . append ( demande1 [ ’ OpCarrier ’ ] [

i ] )

d i co [ ’ MarketShare ’ ] . append ( demande1 [ ’

MarketShare ’ ] [ i ] )

d i co [ ’ FareClass ’ ] . append ( demande1 [ ’ FareClass ’ ] [

i ] )

d i co [ ’Pax ’ ] . append ( demande1 [ ’Pax ’ ] [ i ]+demande1 [

’Pax ’ ] [ i +1])

var=f l o a t ( demande1 [ ’ Pr i ce ’ ] [ i ]∗ demande1 [ ’Pax ’ ] [

i ]+demande1 [ ’ Pr i ce ’ ] [ i +1]∗demande1 [ ’Pax ’ ] [ i

+1]) / f l o a t ( demande1 [ ’Pax ’ ] [ i ]+demande1 [ ’Pax ’

] [ i +1])

d i co [ ’ Pr i ce ’ ] . append ( var )

i=i+2

cont inue

Page 100: Delays and fares - A case study of John Wayne Airport

99

e l i f demande1 [ ’ Or ig in ’ ] [ i ]==demande1 [ ’ Or ig in ’ ] [ i +1] and

demande1 [ ’ Dest ’ ] [ i ]==demande1 [ ’ Dest ’ ] [ i +1] and

demande1 [ ’ OpCarrier ’ ] [ i ]==demande1 [ ’ OpCarrier ’ ] [ i +1]

and demande1 [ ’ FareClass ’ ] [ i ]== ’F ’ and demande1 [ ’

FareClass ’ ] [ i +1]== ’G’ :

i f demande1 [ ’Pax ’ ] [ i ]+demande1 [ ’Pax ’ ] [ i +1]==0:

d i co [ ’ Or ig in ’ ] . append ( demande1 [ ’ Or ig in ’ ] [ i ] )

d i co [ ’ Dest ’ ] . append ( demande1 [ ’ Dest ’ ] [ i ] )

d i co [ ’ Distance ’ ] . append ( demande1 [ ’ Distance ’ ] [ i

] )

d i co [ ’ OpCarrier ’ ] . append ( demande1 [ ’ OpCarrier ’ ] [

i ] )

d i co [ ’ MarketShare ’ ] . append ( demande1 [ ’

MarketShare ’ ] [ i ] )

d i co [ ’ FareClass ’ ] . append ( demande1 [ ’ FareClass ’ ] [

i ] )

d i co [ ’Pax ’ ] . append ( demande1 [ ’Pax ’ ] [ i ]+demande1 [

’Pax ’ ] [ i +1])

d i co [ ’ Pr i ce ’ ] . append (0 )

e l s e :

d i co [ ’ Or ig in ’ ] . append ( demande1 [ ’ Or ig in ’ ] [ i ] )

d i co [ ’ Dest ’ ] . append ( demande1 [ ’ Dest ’ ] [ i ] )

d i co [ ’ Distance ’ ] . append ( demande1 [ ’ Distance ’ ] [ i

] )

Page 101: Delays and fares - A case study of John Wayne Airport

100

d ico [ ’ OpCarrier ’ ] . append ( demande1 [ ’ OpCarrier ’ ] [

i ] )

d i co [ ’ MarketShare ’ ] . append ( demande1 [ ’

MarketShare ’ ] [ i ] )

d i co [ ’ FareClass ’ ] . append ( demande1 [ ’ FareClass ’ ] [

i ] )

d i co [ ’Pax ’ ] . append ( demande1 [ ’Pax ’ ] [ i ]+demande1 [

’Pax ’ ] [ i +1])

d i co [ ’ Pr i ce ’ ] . append ( f l o a t ( demande1 [ ’ Pr i ce ’ ] [ i

]∗ demande1 [ ’Pax ’ ] [ i ]+demande1 [ ’ Pr i ce ’ ] [ i +1]∗

demande1 [ ’Pax ’ ] [ i +1]) / f l o a t ( demande1 [ ’Pax ’ ] [

i ]+demande1 [ ’Pax ’ ] [ i +1]) )

i=i+2

cont inue

e l i f demande1 [ ’ Or ig in ’ ] [ i ]==demande1 [ ’ Or ig in ’ ] [ i +1] and

demande1 [ ’ Dest ’ ] [ i ]==demande1 [ ’ Dest ’ ] [ i +1] and

demande1 [ ’ OpCarrier ’ ] [ i ]==demande1 [ ’ OpCarrier ’ ] [ i +1]

and demande1 [ ’ FareClass ’ ] [ i ]== ’C ’ and demande1 [ ’

FareClass ’ ] [ i +1]== ’D’ :

i f demande1 [ ’Pax ’ ] [ i ]+demande1 [ ’Pax ’ ] [ i +1]==0:

d i co [ ’ Or ig in ’ ] . append ( demande1 [ ’ Or ig in ’ ] [ i ] )

d i co [ ’ Dest ’ ] . append ( demande1 [ ’ Dest ’ ] [ i ] )

Page 102: Delays and fares - A case study of John Wayne Airport

101

d ico [ ’ Distance ’ ] . append ( demande1 [ ’ Distance ’ ] [ i

] )

d i co [ ’ OpCarrier ’ ] . append ( demande1 [ ’ OpCarrier ’ ] [

i ] )

d i co [ ’ MarketShare ’ ] . append ( demande1 [ ’

MarketShare ’ ] [ i ] )

d i co [ ’ FareClass ’ ] . append ( demande1 [ ’ FareClass ’ ] [

i ] )

d i co [ ’Pax ’ ] . append ( demande1 [ ’Pax ’ ] [ i ]+demande1 [

’Pax ’ ] [ i +1])

d i co [ ’ Pr i ce ’ ] . append (0 )

e l s e :

d i co [ ’ Or ig in ’ ] . append ( demande1 [ ’ Or ig in ’ ] [ i ] )

d i co [ ’ Dest ’ ] . append ( demande1 [ ’ Dest ’ ] [ i ] )

d i co [ ’ Distance ’ ] . append ( demande1 [ ’ Distance ’ ] [ i

] )

d i co [ ’ OpCarrier ’ ] . append ( demande1 [ ’ OpCarrier ’ ] [

i ] )

d i co [ ’ MarketShare ’ ] . append ( demande1 [ ’

MarketShare ’ ] [ i ] )

d i co [ ’ FareClass ’ ] . append ( demande1 [ ’ FareClass ’ ] [

i ] )

d i co [ ’Pax ’ ] . append ( demande1 [ ’Pax ’ ] [ i ]+demande1 [

’Pax ’ ] [ i +1])

Page 103: Delays and fares - A case study of John Wayne Airport

102

d ico [ ’ Pr i ce ’ ] . append ( f l o a t ( demande1 [ ’ Pr i ce ’ ] [ i

]∗ demande1 [ ’Pax ’ ] [ i ]+demande1 [ ’ Pr i ce ’ ] [ i +1]∗

demande1 [ ’Pax ’ ] [ i +1]) / f l o a t ( demande1 [ ’Pax ’ ] [

i ]+demande1 [ ’Pax ’ ] [ i +1]) )

i=i+2

cont inue

e l s e :

d i co [ ’ Or ig in ’ ] . append ( demande1 [ ’ Or ig in ’ ] [ i ] )

d i co [ ’ Dest ’ ] . append ( demande1 [ ’ Dest ’ ] [ i ] )

d i co [ ’ Distance ’ ] . append ( demande1 [ ’ Distance ’ ] [ i ] )

d i co [ ’ OpCarrier ’ ] . append ( demande1 [ ’ OpCarrier ’ ] [ i ] )

d i co [ ’ MarketShare ’ ] . append ( demande1 [ ’ MarketShare ’ ] [

i ] )

d i co [ ’ FareClass ’ ] . append ( demande1 [ ’ FareClass ’ ] [ i ] )

d i co [ ’Pax ’ ] . append ( demande1 [ ’Pax ’ ] [ i ] )

d i co [ ’ Pr i ce ’ ] . append ( demande1 [ ’ Pr i ce ’ ] [ i ] )

i=i+1

dico [ ’ Or ig in ’ ] . append ( demande1 [ ’ Or ig in ’ ] [ l en ( demande1 [ ’ Or ig in ’

] ) −1])

d i co [ ’ Dest ’ ] . append ( demande1 [ ’ Dest ’ ] [ l en ( demande1 [ ’ Or ig in ’ ] )

−1])

Page 104: Delays and fares - A case study of John Wayne Airport

103

d ico [ ’ Distance ’ ] . append ( demande1 [ ’ Distance ’ ] [ l en ( demande1 [ ’

Or ig in ’ ] ) −1])

d i co [ ’ OpCarrier ’ ] . append ( demande1 [ ’ OpCarrier ’ ] [ l en ( demande1 [ ’

Or ig in ’ ] ) −1])

d i co [ ’ MarketShare ’ ] . append ( demande1 [ ’ MarketShare ’ ] [ l en ( demande1

[ ’ Or ig in ’ ] ) −1])

d i co [ ’ FareClass ’ ] . append ( demande1 [ ’ FareClass ’ ] [ l en ( demande1 [ ’

Or ig in ’ ] ) −1])

d i co [ ’Pax ’ ] . append ( demande1 [ ’Pax ’ ] [ l en ( demande1 [ ’ Or ig in ’ ] ) −1])

d i co [ ’ Pr i ce ’ ] . append ( demande1 [ ’ Pr i ce ’ ] [ l en ( demande1 [ ’ Or ig in ’ ] )

−1])

re turn d ico

de f merge ( year , quarte r ) :

de lays=f u s i o n q u a r t e r ( year , quarte r )

t a r i f s=ouvr i r ( year , quarte r )

merge={}

merge [ ’ Or ig in ’ ] = [ ]

merge [ ’ Dest ’ ] = [ ]

merge [ ’ Distance ’ ] = [ ]

merge [ ’ OpCarrier ’ ] = [ ]

merge [ ’ MarketShare ’ ] = [ ]

merge [ ’ FareClass ’ ] = [ ]

Page 105: Delays and fares - A case study of John Wayne Airport

104

merge [ ’Pax ’ ] = [ ]

merge [ ’ Pr i ce ’ ] = [ ]

merge [ ’ Mean dep delay ’ ] = [ ]

merge [ ’ Q90 dep delay ’ ] = [ ]

merge [ ’ Mean arr delay ’ ] = [ ]

merge [ ’ Q90 arr de lay ’ ] = [ ]

merge [ ’ Quarter ’ ] = [ ]

merge [ ’ Year ’ ] = [ ]

f o r i in range ( l en ( t a r i f s [ ’ Or ig in ’ ] ) ) :

f o r j in range ( l en ( de lays [ ’ORIGIN ’ ] ) ) :

i f t a r i f s [ ’ Or ig in ’ ] [ i ]==de lays [ ’ORIGIN ’ ] [ j ] and t a r i f s [ ’

Dest ’ ] [ i ]==de lays [ ’DEST ’ ] [ j ] and t a r i f s [ ’ OpCarrier ’ ] [ i

]==de lays [ ’CARRIER ’ ] [ j ] :

merge [ ’ Or ig in ’ ] . append ( t a r i f s [ ’ Or ig in ’ ] [ i ] )

merge [ ’ Dest ’ ] . append ( t a r i f s [ ’ Dest ’ ] [ i ] )

merge [ ’ Distance ’ ] . append ( t a r i f s [ ’ Distance ’ ] [ i ] )

merge [ ’ OpCarrier ’ ] . append ( t a r i f s [ ’ OpCarrier ’ ] [ i ] )

merge [ ’ MarketShare ’ ] . append ( t a r i f s [ ’ MarketShare ’ ] [ i ] )

merge [ ’ FareClass ’ ] . append ( t a r i f s [ ’ FareClass ’ ] [ i ] )

merge [ ’Pax ’ ] . append ( t a r i f s [ ’Pax ’ ] [ i ] )

merge [ ’ Pr i ce ’ ] . append ( t a r i f s [ ’ Pr i c e ’ ] [ i ] )

merge [ ’ Mean dep delay ’ ] . append ( de lays [ ’MEAN DEP DELAY ’

] [ j ] )

Page 106: Delays and fares - A case study of John Wayne Airport

105

merge [ ’ Q90 dep delay ’ ] . append ( de lays [ ’Q90 DEP DELAY ’ ] [ j

] )

merge [ ’ Mean arr delay ’ ] . append ( de lays [ ’MEAN ARR DELAY’

] [ j ] )

merge [ ’ Q90 arr de lay ’ ] . append ( de lays [ ’Q90 ARR DELAY ’ ] [ j

] )

merge [ ’ Quarter ’ ] . append ( de lays [ ’QUARTER’ ] [ j ] )

merge [ ’ Year ’ ] . append ( de lays [ ’YEAR’ ] [ j ] )

r e turn merge

de f f i n ( ) :

m={}

m[ ’ Or ig in ’ ] = [ ]

m[ ’ Dest ’ ] = [ ]

m[ ’ Distance ’ ] = [ ]

m[ ’ OpCarrier ’ ] = [ ]

m[ ’ MarketShare ’ ] = [ ]

m[ ’ FareClass ’ ] = [ ]

m[ ’Pax ’ ] = [ ]

m[ ’ Pr i ce ’ ] = [ ]

m[ ’ Mean dep delay ’ ] = [ ]

m[ ’ Q90 dep delay ’ ] = [ ]

m[ ’ Mean arr delay ’ ] = [ ]

m[ ’ Q90 arr de lay ’ ] = [ ]

Page 107: Delays and fares - A case study of John Wayne Airport

106

m[ ’ Quarter ’ ] = [ ]

m[ ’ Year ’ ] = [ ]

c l e s f i n a l =[ ’ Or ig in ’ , ’ Dest ’ , ’ Distance ’ , ’ OpCarrier ’ , ’ MarketShare ’ , ’

FareClass ’ , ’Pax ’ , ’ Pr i ce ’ , ’ Mean dep delay ’ , ’ Q90 dep delay ’ , ’

Mean arr delay ’ ,

’ Q90 arr de lay ’ , ’ Quarter ’ , ’ Year ’ ]

f o r year in range (1999 ,2010) :

f o r quarte r in range (1 , 5 ) :

p r i n t ’ par s ing ’ , year , quarte r

d i co=merge ( year , quarte r )

f o r j in range ( l en ( d ico [ ’ Or ig in ’ ] ) ) :

f o r x in c l e s f i n a l :

m[ x ] . append ( d ico [ x ] [ j ] )

z=open ( ’ / Users / e l s a d e l p e c h /Desktop/ in te rmed ia te . txt ’ , ’w ’ )

z . wr i t e ( ’ Origin , Dest , Distance , OpCarrier , MarketShare ,

FareClass , Pax , Price , Mean dep delay , Q90 dep delay ,

Mean arr delay , Q90 arr de lay , Quarter , Year\n ’ )

f o r i in range ( l en (m[ ’ Or ig in ’ ] ) ) :

z . wr i t e ( ’”%s ” , ”%s ” , %s , ”%s ” , %s , ”%s ” , %s , %s , %s , %s

, %s , %s , %s , %s \n ’%(m[ ’ Or ig in ’ ] [ i ] ,m[ ’ Dest ’ ] [ i ] ,m[ ’

Distance ’ ] [ i ] ,m[ ’ OpCarrier ’ ] [ i ] ,m[ ’ MarketShare ’ ] [ i ] ,

m[ ’ FareClass ’ ] [ i ] ,m[ ’Pax ’ ] [ i ] ,m[ ’ Pr i ce ’ ] [ i ] ,m[ ’

Page 108: Delays and fares - A case study of John Wayne Airport

107

Mean dep delay ’ ] [ i ] ,m[ ’ Q90 dep delay ’ ] [ i ] ,m[ ’

Mean arr delay ’ ] [ i ] ,m[ ’ Q90 arr de lay ’ ] [ i ] ,m[ ’ Quarter

’ ] [ i ] ,m[ ’ Year ’ ] [ i ] ) )

z . c l o s e ( )

z=open ( ’ / Users / e l s a d e l p e c h /Desktop/ f i n a l f i l e . txt ’ , ’w ’ )

z . wr i t e ( ’ Origin , Dest , Distance , OpCarrier , MarketShare , FareClass , Pax ,

Price , Mean dep delay , Q90 dep delay , Mean arr delay , Q90 arr de lay ,

Quarter , Year\n ’ )

f o r i in range ( l en (m[ ’ Or ig in ’ ] ) ) :

z . wr i t e ( ’”%s ” , ”%s ” , %s , ”%s ” , %s , ”%s ” , %s , %s , %s , %s , %s , %s

, %s , %s \n ’%(m[ ’ Or ig in ’ ] [ i ] ,m[ ’ Dest ’ ] [ i ] ,m[ ’ Distance ’ ] [ i ] ,m[

’ OpCarrier ’ ] [ i ] ,m[ ’ MarketShare ’ ] [ i ] ,m[ ’ FareClass ’ ] [ i ] ,m[ ’Pax

’ ] [ i ] ,m[ ’ Pr i ce ’ ] [ i ] ,m[ ’ Mean dep delay ’ ] [ i ] ,m[ ’ Q90 dep delay ’

] [ i ] ,m[ ’ Mean arr delay ’ ] [ i ] ,m[ ’ Q90 arr de lay ’ ] [ i ] ,m[ ’ Quarter

’ ] [ i ] ,m[ ’ Year ’ ] [ i ] ) )

z . c l o s e ( )

re turn None

Page 109: Delays and fares - A case study of John Wayne Airport

108

List of Figures

1 Geography of Orange County, on the Pacific Ocean. Source: Wikipedia. . . . . . . . 6

2 Location of John Wayne Airport and major nearby airports, denoted by their three-

letter IATA codes. Source: Google Maps. . . . . . . . . . . . . . . . . . . . . . . . . 7

3 Carriers’ market share (number of pax) in SNA (2010:2). . . . . . . . . . . . . . . . 18

4 Example of price distribution of carriers operating at SNA . . . . . . . . . . . . . . . 20

5 Mean distance per passenger for each carrier (km) (2010:2). . . . . . . . . . . . . . . 21

6 Mean price for each class (in US$) (2010:2). . . . . . . . . . . . . . . . . . . . . . . . 22

7 Influence of the class and the distance on the price of the ticket (2010:2). . . . . . . 23

8 There are 4 carriers operating on SNA/DEN: Frontier (F9), United (UA), US Airways

(US) and Southwest (WN) (2010:2). . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

9 Mean price per carrier between Denver and John Wayne in US$ (2010:2). . . . . . . 24

10 Carrier shares in terms of operated flights at John Wayne Airport on January 2011. 32

11 Kernel density of departure delays and arrival delays for all departure/arrival flights

to/from John Wayne Airport in January 2011. . . . . . . . . . . . . . . . . . . . . . 33

12 Scatterplot of arrival delay against departure delay for flights whose departure delay

is above the 90% quantile. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

13 Carrier shares in the smaller and larger departure delays in January 2011. . . . . . . 35

14 Scatter plot of departure delay against distance for all flights to/from John Wayne

in January 2011. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

15 Carrier shares in the smaller and larger arrival delays in January 2011. . . . . . . . . 37

16 Delay distribution by carrier. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39

17 Delay distribution by carrier (next carriers). . . . . . . . . . . . . . . . . . . . . . . . 40

18 Histogram of scheduled departure times at SNA in January 2011. . . . . . . . . . . . 41

19 Mean departure delay by time slot in January 2011. . . . . . . . . . . . . . . . . . . 41

Page 110: Delays and fares - A case study of John Wayne Airport

109

20 Maximum departure delay by time slot in January 2011. . . . . . . . . . . . . . . . . 42

21 Mean departure delays by carrier in January 2011 for flights to/from John Wayne. . 46

22 90% quantile of departure (minutes) delay against time (quarters). . . . . . . . . . . 67

23 90% quantile of departure (minutes) delay against time (quarters) without Q90 dep delay

less than 20 minutes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68

24 Coefficients of time dummies against time. . . . . . . . . . . . . . . . . . . . . . . . . 71

25 The location of John Wayne Airport in Orange County, California. Source: John

Wayne Airport website, Map of Orange County (http://www.ocair.com/directions/). 81

26 Noise Monitoring Station map at John Wayne Airport in 2008. Source: Boeing,

Airport Noise and Emission Regulations (http://www.boeing.com/commercial/

noise/john_wayne.html). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82

27 Annual CNEL (Community Noise Equivalent Level) Noise Contours map in 2010.

Source: John Wayne Airport website (http://www.ocair.com/reportspublications/

AccessNoise/default.aspx). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83

Page 111: Delays and fares - A case study of John Wayne Airport

110

List of Tables

1 Total number of passengers to and from John Wayne Airport since 1991, and variation

on the previous year. Source : John Wayne Airport Press Kit [13]. . . . . . . . . . . 8

2 Total number of passengers to and from John Wayne Airport, and aircraft operations,

and variation between 2009 and 2010. Source : John Wayne Airport Press Kit [13] . 8

3 Airlines serving John Wayne Airport and non-stop destinations. Source: John

Wayne Airport website (http://www.ocair.com/airlines/). . . . . . . . . . . . . . 10

4 Noise limits at John Wayne Airport, as of February 2011. Source: Boeing, Airport

Noise and Emission Regulations (http://www.boeing.com/commercial/noise/john_

wayne.htm). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

5 General aviation noise limits at John Wayne Airport in February 2011. Source:

Boeing, Airport Noise and Emission Regulations (http://www.boeing.com/commercial/

noise/john_wayne.html). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

6 Routes to and from SNA, for each carrier in 2010:2 . . . . . . . . . . . . . . . . . . . 17

7 Ticket fare class distribution for each carrier operating at SNA (2010:2). . . . . . . . 19

8 Linear regression number one (2010:2). . . . . . . . . . . . . . . . . . . . . . . . . . . 26

9 Linear regression number two (2010:2). . . . . . . . . . . . . . . . . . . . . . . . . . . 28

10 Linear regression number three (2010:2). . . . . . . . . . . . . . . . . . . . . . . . . . 29

11 Linear regression number four (2010:2). . . . . . . . . . . . . . . . . . . . . . . . . . 30

12 Summary statistics for departure and arrival delays at John Wayne Airport in

January 2011. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

13 Regression of departure delays in January 2011, on distance, time slot dummies,

carrier dummies. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

14 Regression of departure delays in January 2011, on distance, gathered time slot

dummies, carrier dummies. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48

Page 112: Delays and fares - A case study of John Wayne Airport

111

15 Regression of departure delays in January 2011, on distance, gathered time slot

dummies, carrier dummies and competition dummy. . . . . . . . . . . . . . . . . . . 49

16 Regression of departure delays in January 2011, on distance squared, time slot

dummies, carrier dummies and competition dummy. . . . . . . . . . . . . . . . . . . 50

17 Regression of departure delays in January 2011 above the 90% quantile, on distance,

time slot dummies, carrier dummies. . . . . . . . . . . . . . . . . . . . . . . . . . . . 53

18 Regression of departure delays in January 2011 above the 90% quantile, on distance,

grouped time slot dummies, carrier dummies. . . . . . . . . . . . . . . . . . . . . . . 54

19 Regression of departure delays in January 2011 above the 90% quantile, on distance,

grouped time slot dummies, carrier dummies. . . . . . . . . . . . . . . . . . . . . . . 55

20 Regression of departure delays in January 2011 below the 10% quantile, on distance,

time slot dummies, carrier dummies. . . . . . . . . . . . . . . . . . . . . . . . . . . . 57

21 Regression of departure delays in January 2011 below the 10% quantile, on distance,

time slot dummies, carrier dummies. . . . . . . . . . . . . . . . . . . . . . . . . . . . 58

22 Regression of price on a constant, distance, dumx and carrier dummies on quarter 2

2010. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63

23 Regression of price on a constant, distance, dumx, carrier dummies and Q90 depdelay

with Newey-West standard errrors on quarter 2 2010. . . . . . . . . . . . . . . . . . 65

24 Regression of price on a constant, distance, dumx and Q90 depdelay on quarter 2

2010. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66

25 Regression of prices on a constant, compet, distance, eco, q90 depdelay and time

with Newey-West standard errors. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70

26 Regression of prices on a constant, compet, distance, eco, q90 depdelay and time

dummies for each quarter with Newey-West standard errors. . . . . . . . . . . . . . . 72

Page 113: Delays and fares - A case study of John Wayne Airport

112

27 Regression of prices on a constant, compet, distance, eco, q90 depdelay, time and

Q90 dep delay*compet with Newey-West standard errors. . . . . . . . . . . . . . . . 74

28 Regression of prices on a constant, compet, distance, eco, q90 depdelay, time and

carrier dummies with Newey-West standard errors. . . . . . . . . . . . . . . . . . . . 76

29 Regression of prices on a constant, compet, distance, distance2, eco, Q90 dep delay,

Q90 dep delay*compet, time dummies and carrier dummies with Newey-West standard

errors. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78

30 Relative severity of noise levels. Sources: Aeroport&Environnement course, D.Vergez

and Le portail citoyen sonores (http://www.moinsdebruit.com/le-bruit/lechelle-du-bruit.

html). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84

Page 114: Delays and fares - A case study of John Wayne Airport

113

Notes

1See http://www.ocparks.com : 7 miles of beaches

2Source: “John Wayne Airport Settlement Agreement - Past And Future”, http://www.aviationairportdevelopmentlaw.

com

3Sources: http://jmpostings.posterous.com/moorlach-update-newport-beach-patch-march-212 and http:

//www.eltoroairport.org/issues/JWA.htm

4Source: http://www.ocair.com

5http://www.ocair.com/communityrelations/accessandnoise.aspx

6We can see it on the website www4.passur.com/sna.html

7Access & Noise Office: (949) 252-5185, [email protected]

8Source: http://en.wikipedia.org/wiki/General_aviation

9Source: ”Press Kit” by Alan L. Murphy, Director of John Wayne Airport, 26 September 2011

10http://www.boeing.com/commercial/noise/john_wayne.html

11Sources: OCWeekly, “John Wayne Airport Named Among 10 Scariest” by Matt Coker, 5 August 2010, http:

//articles.latimes.com/1998/apr/13/local/me-38832 and “World’s Scariest Airports” by Jaclyn Liechti and Kate

Sitarz, 31 July 2010 http://www.smartertravel.com/photo-galleries/editorial/worlds-scariest-airports.

html?id=32&photo=8736

12http://www.bts.gov/help/aviation/html/understanding.html

Page 115: Delays and fares - A case study of John Wayne Airport

114

References

[1] Airport monitor at john wayne, www4.passur.com/sna.html.

[2] http://www.aviationairportdevelopmentlaw. com.

[3] http://www.eltoroairport.org/issues/jwa.htm.

[4] http://www.moinsdebruit.com/le-bruit/lechelle-du-bruit.html.

[5] John moorlach’s postings, http://jmpostings.posterous.com/moorlach-update-newport-beach-

patch-march-212.

[6] John wayne airport settlement agreement-past and future.

[7] Orange county airport website, http://www.ocair.com.

[8] wikipedia.

[9] Matt Coker. John wayne airport named among 10 scariest. OCWeekly, 2010.

[10] D.Vergez. Conference aeroports & environnement.

[11] Silke J. Forbes. The effect of air traffic delays on airline prices. International Journal of

Industrialization Organization, page 15, 2007.

[12] Jaclyn Liechti and Kate Sitarz. Worlds scariest airports. Smart Traveler, 2010.

[13] A. L. Murphy. John wayne airport press kit. 2011.