CHARLES UNIVERSITY FACULTY OF SOCIAL SCIENCES

(1)

CHARLES UNIVERSITY FACULTY OF SOCIAL SCIENCES

Institute of Economic Studies

Impacts of various critical situations on aviation industry and parallel with the

COVID-19 pandemic

Bachelor’s thesis

Author: Milan Fedorik

Study program: Ekonomie a finance Supervisor: RNDr. Michal Červinka Ph.D.

Year of defense: 2021

(2)

The author hereby declares that he or she compiled this thesis independently, using only the listed resources and literature, and the thesis has not been used to obtain any other academic title.

The author grants to Charles University permission to reproduce and to dis- tribute copies of this thesis in whole or in part and agrees with the thesis being used for study and scientific purposes.

Prague, May 4, 2021

Milan Fedorik

(3)

Abstract

Civil aviation is expanding industry in the United States. The number of passengers had risen following the Airline Deregulation Act and kept rising until the 9/11 attacks hit the domestic aviation market. This unprecedented event resulted in a decline in the number of passengers. In January 2020 first rumors about disease caused by SARS-CoV-2 spread around the world.

New health threats made traveling more dangerous and unessential. In this study, the author copes with aviation history and how critical situations such as terrorist attacks or pandemics affect the civil aviation market. Additionally, this thesis presents methods of performing regression analysis on the primary U.S. domestic market and specific routes subsidized through the government program Essential Air Service.

Keywords aviation, COVID-19 pandemics, Essential Air Service, 9/11 attacks, OLS

Title Impacts of various critical situations on aviation industry and parallel with the COVID-19 pandemic

Abstrakt

Civilné letectvo je jedným z rastúcich priemyslov v Spojených štátoch americk- ých. Vďaka deregulácii amerického trhu leteckých spoločností, Airline Deregu- lation Act, počet cestujˇucich začal rásť a pokračoval až kým teroristické útoky z 9. septembra nezasiahli trh domácich leteckých spojení. Táto bezprece- dentná udalosť spôsobila dramatický pokles počtu cestujúcich. V januári 2020 sa svetom začali šíriť prvé informácie o chorobe spôsobenej vírusom SARS- CoV-2. Kvôli tejto novej zdravotnej hrozbe sa letectvo stalo nebezpečným a nepodstatným. Autor tejto štúdie analyzuje históriu letectva a dopady krit- ických udalostí akou sú teroristické útoky, či pandémie na letecký trh. Navyše v tejto práci prezentuje metódu regresnej analýzy na hlavnom domácom trhu Spojených štátov amerických a vládneho dotačného programu - Essential Air Service.

Kľúčové slová letectvo, pandémia COVID-19, program Essen- tial Air Serivce, útoky z 9. septembra, MNŠ Český názov práce Dopady vybraných historických krizí na letecký

prˇumysl a porovnání s pandemií COVID-19

(4)

The author is grateful especially to RNDr. Michal Červinka Ph.D.for his ad- vices during the preparation of this thesis. Thanks to comments and sugges- tions, the author prepared the study presented in this thesis.

Furthermore, the author would like to express gratitude to his closest friends and family, who always stand by his side and are supportive during his studies.

Typeset in L^ATEXusing the IES Thesis Template.

Bibliographic Record

Fedorik, Milan: Impacts of various critical situations on aviation industry and parallel with the COVID-19 pandemic. Bachelor’s thesis. Charles University, Faculty of Social Sciences, Institute of Economic Studies, Prague. 2021, pages 63. Advisor: RNDr. Michal Červinka Ph.D.

(5)

List of Tables

4.1 Available T-100 databases . . . 16

4.2 Number of domestic passengers by origin of carrier . . . 17

4.3 List of non-binary variables . . . 21

4.4 List of binary variables . . . 21

5.1 Descriptive statistics on revenue passenger mile (RPM) . . . 22

5.2 Descriptive statistics on available seat mile (ASM) . . . 23

7.1 Results from OLS estimation - Model 1 . . . 32

7.2 Results from OLS estimation - Model 2 . . . 34

7.3 Recalculated effect of COVID-19 in time . . . 34

7.4 Results from OLS estimation - Model 3 . . . 35 A.1 List of US States . . . I A.2 Correlation coefficients for variables included in the models . . . II

(8)

2.1 Distribution of communities in subsidy programs . . . 8 5.1 ASM and RPM on all U.S. domestic routes in first months of 2020 23 5.2 ASM and RPM on all EAS routes in first months of 2020 . . . . 24 5.3 Progress of load factor . . . 24 5.4 Labor market in the United States . . . 25 B.1 Subsidised EAS report for communities in Alaska - September

2020 (first page) . . . III B.2 Subsidised EAS report for communities in Alaska - September

2020 (second page) . . . IV B.3 Seasonality and trend on total RPM . . . IV B.4 Seasonality and trend on EAS RPM . . . V B.5 Z-score for descriptive statistics . . . V B.6 Load factor on all types of routes . . . VI

(9)

Acronyms

AEAS Alternative Essential Air Service

BTS Bureau of Transportation Statistics

CAB Civil Aeronautics Board

DOT Department of Transportation

EAS Essential Air Service

IATA International Air Transport Association

ICAO International Civil Aviation Organisation

LCC low-cost carrier

WHO World Health Organization

(10)

Author Milan Fedorik

Supervisor RNDr. Michal Červinka Ph.D.

Proposed topic Impacts of various critical situations on aviation industry and parallel with the COVID-19 pandemic

Research question and motivation The main research question of this bachelor thesis is to compare the main characteristics of different critical situations in past twenty years, what were the impacts of them on the aviation and the airline industry and how is this knowledge comparable to the COVID-19 pandemic as the prediction how the airlines should react to actual situation. The results of aviation and the airline industry are interconnected to the global economic conditions. Since 1978, when the Airline Deregulation Act was signed into law, the US aviation market became characterized by strong competition between hundreds of major international, medium-sized national and small regional air companies. This competition brought to the market low cost carriers (LCCs) and started ‘Southwest Effect’ which com- pletely changed the game and showed how the economic situation affects the aviation.

(Vasigh, Fleming, Tacker, 2013) However, during past twenty years the airlines have had to cope with extreme situation which they have not experienced ever before.

The September 11 attack caused increased fear of flying and more strict security policy which made the air travel less attractive and the costs for civil air transport were significant. (Ito, Lee, 2005) Similar effect could be seen in period of the ‘great recession’ (the years following the collapse of the housing market 2008-2013, Vasigh, Fleming, Tacker, 2013). During the stages of the economic decline, the potential passengers had less money and their incentive to travel was therefore lower. With higher unemployment rate it is expected that the group of business travelers stopped flying across the United States. In this thesis I intend to study similarities among terrorism, the ‘great recession’ and the pandemic of COVID-19 and provide a prediction for the consequences of ongoing crisis on aviation. The study would focus on

(11)

Bachelor’s Thesis Proposal xi

changes of travelling habits of the customers, how the passengers chooses the different type of airlines and how the crisis affects the whole industry, e.g. 80 per cent of April’s scheduled flights were cancelled. (Hollinger, 2020)

Contribution In this study I would like to find a scheme in reaction of aviation to the terrorism and the ‘great recession’ which could be the basic line for the following years after the COVID-19 pandemic. There are some articles which expect the decline in air transport to be like that after September 11 attacks and the economic recession which may follow the pandemic to be as much serious as it was in years 2008-2013.

These expectations may correctly predict the possible shocks on both sides of the markets - demand (e.g. the fear of getting the illness) and supply (e.g. new security measures). This knowledge could be helpful for other growth predictions and plans for the airlines during the next years. By working on this study, I hope to improve my analytical and data science skills beyond the standard curriculum at the Institute of economic studies, FSV UK.

Methodolody The inspiration for the methodology of the thesis is the model of Ito and Lee whose analysis is aimed on the impact of the September 11 attacks.

Their multiple regression model works with the different quantitative variables (e.g.

unemployment rate, the share of the LCC’) and the dummy variables (e.g. ongoing Iraq War, SARS epidemic) to explain RPK - revenue passenger kilometre (Ito, Lee, 2005). I would like to use this model as the foundation for the analysis of data from the Bureau of Transportation Statistics (https://www.bts.gov/) aiming on the years of the ‘great recession’. In the model I would need to include other possible variables which may have effect. Naturally, we may assume that the seasonality would have impact on the aviation. Therefore, it would need to be modelled with help of the dummy variables - monthly or by quarters (Vasigh, Fleming, Tacker, 2013.) The diverse variables may help to explain the changes on the market after COVID-19 pandemic.

Outline

1. Introduction 2. Literature Review 3. Theoretical part

• Impact of the terrorism

• The Great Recession and the aviation

(12)

• Defining relevant information for model during and after the COVID-19 pandemic

4. Empirical part - contracting data, analysis of data

• Contracting data

• Analysis of data

5. Results and identifications of parallels 6. Effects on the other regional markets 7. Conclusion

List of Academic Literature

Vasigh, B., Tacker, T., & Fleming, K. (2013). Introduction to Air Transport Economics: From Theory to Applications (2nd ed.). Farnham: Ashgate.

Harvey, G., & Turnbull, P. (2009). The Impact of The Financial Crisis on Labour in The Civil Aviation Industry. Geneva: International Labour Organi- zation.

Holinger, P. (2020, April 20). How coronavirus brought aerospace down to earth. Financial Times.

Harumi, I., & Lee, D. (2005). Comparing the Impact of the September 11th Terrorist Attacks on International Airline Demand. International Journal of the Economics of Business: pp. 225-249.

Lee, Darin. (2016). Competition Policy and Antitrust. Vol. 1, in Advances in Airline Economics, by Darin Lee. Amsterdam: Elsevier.

O’Connor, W. E. (2001). An Introduction to Airline Economics (6th ed.).

Westport: Praeger Publisher.

Wooldridge, J. M. (2009). Introductory Econometrics: A Modern Approach (4th ed.). Boston: Cengage.

Author Supervisor

(13)

Chapter 1 Introduction

In January 2020, the media published initial news concerning a new infection spreading in Hubei province in China. In the early stages, immunologists rec- ognized origin in SARS-CoV-2, the virus causing respiratory illness - similar to the first SARS epidemics in 2003. As of spring 2021, COVID-19 pandemics were the ninth deadliest pandemics in history, with approximately 3 million deaths worldwide. These pandemics disrupted the ordinary life of many.

COVID-19 pandemics left a mark on the global economy as well - air transportation including. Unprecedented fall in the number of passengers, numerous cancellations of flights, and new safety restrictions labeled months following the outbreak of COVID-19 pandemics as the biggest shock in aviation industries.

Because of its severity and its consequences on daily lives, aviation analysis started to compare COVID-19 pandemics to other past critical events - the 9/11 attacks and the ’great recession’ that affected aviation stunningly perma- nently.

The main objective of this thesis is to analyze the aviation market in the United States in the context of critical situations which occurred in recent years.

While Goodrich (2002) presented first uncovered effects of the 9/11 attacks on the U.S. economy, Roberts (2009) reported a more complex analysis on how these attacks affected economic growth in the long run. More specifically, Ito

& Lee (2005a) in their article covered what effect had terrorist attacks on the domestic aviation market. In this thesis, we used results of their study and gave them a new context - comparison with COVID-19 pandemics.

(14)

During our study, we have highlighted connections between market reactions to the 9/11 attacks and COVID-19 pandemics. Both these critical situations resulted in changes in both demand and supply of domestic travel - aviation included. As written by Lamb et al.(2020), COVID-19 pandemics changed the behavior of passengers. We used these connections as a relevancy condition for comparison of their effects on the aviation market. The outcome of this study approach is one of the main conclusions of this thesis.

Based on the book written by Vasigh et al. (2013) we selected all relevant information used for explaining activity on aviation markets. Similar to Ito &

Lee (2005a), we used revenue passenger miles (explained later in the thesis) as an explanatory variable in our regression analyzes. Apart from standard variables, which were used by other authors as well, we focused on all events which occurred since 2002. This period is consecutive for the period covered by Ito & Lee.

The second main topic covered by this thesis is the impact of COVID- 19 pandemics on subsidized air routes - governmental program Essential Air Service. After the literature review, we believe there is a limited range of academic articles in which their authors studied this program. To the best of our knowledge, there is no article in which its author performed regression analysis.

Hence, in this thesis, we present a method for analyzing revenue passenger mile for Essential Air Service routes. Collected data were used to perform alternative regression for the market of these subsidized routes.

The thesis is structured as follows: Chapter 2 introduces readers with basic information about aviation, U.S. aviation market, and subsidy programs;

Chapter 3 presents critical situations and their effect on the aviation market - 9/11 attacks and COVID-19 pandemics. In Chapter 4 are all details about data used in model summarised and in Chapter 5 is descriptive statistics presented;

Chapter 6 shows methodology, which was used, and conclusions and results are presented in Chapter 7.

(15)

Chapter 2 Introduction to civil aviation

Commercial aviation had started its history in the early years of the 20th century. More than ten years after a first successful flight made by the Wright brothers in Kitty Hawk, NC, SPT Airboat Line launched its inaugural commercial service on 1 January 1914 with 23-minutes long flight from St. Peters- burg, FL to Tampa, FL. In the following years, air transport had proved its importance in the United States as a fast and convenient alternative to cars and trains - reaching nearly 4.7 million passengers by 1944 (Modley 1945, p. 67).

The commercial aviation market had not become massive until the 60’s when Boeing introduced a new passenger jet model - Boeing 727. After years of recession, leisure traveling to domestic destinations helped airplane producers and airlines create a competitive market for the travel industry. Increase in the number of passengers backed division of different types of travel:

• Business travel - business trips;

• Leisure travel - holidays, vacation trips, family visits.

This definition is crucial for analyzing the development of the aviation market. Leiper et al. (2008) summarised examples of typical travelers. The most fundamental difference between these two groups is an obligation related to travel. While business travelers have to travel and usually can not choose a destination of their journey, the other group has an option to decide when, where, with whom, and how to travel. This mixed approach results in various reactions to changed conditions - e.g., price and policy changes, safety and health threats, and others.

(16)

However, this distinction describes possible reasons for the change of travel habits. Based on analyses and surveys made on groups of passengers, we know that business travelers account only for 12% of all airlines’ passengers. On the contrary, they are a more lucrative part of customers for air companies. Busi- ness passengers make for 75 percent of airlines’ profits. These companies used to adapt all their policies to make these profits even bigger - e.g., higher ticket fares close to the flight departure.

Alderighi et al. (2016) investigated price management of air companies on the case study of Ryanair. As they noted, the business model of major U.S.

airline Southwest Airlines was an inspiration for the founders of Ryanair - the dominant low-cost carrier in Europe. Apart from the case study, they performed regression analysis and estimated variables on which air ticket fares depend on business and leisure air routes. They concluded that business travelers are more likely to book a ticket shorter before travel while other travelers (leisure) plan their journey more in advance.

2.1 Airline Deregulation Act

In 1978, U.S. Government, led by Jimmy Carter, introduced and signed a new federal Airline Deregulation Act, which adopted new rules for the civil aviation market in the United States. The most relevant changes were:

• deregulation of airfares, i.e., airlines got power over the price of air tickets;

• removal of barriers for new carriers to enter the market, either as new airlines or on a specific route.

Previously, Civil Aeronautics Board (CAB) had responsibility for the civil aviation market in the United States. Kahn (1988), who was in 1977 appointed to be the chair of the Civil Aeronautics Board (CAB), identified fares and productivity to be the essential effects of deregulation. The average inflation ad- justed fare paid by passengers declined from 1976 to 1990 by approximately 30 percent. In their analysis, Morrison & Winston (1995) compared actual airfares in 1993 with derived regulated pricing from _CAB’s formula - formula used for calculation of air fare before deregulation. Based on their conclusion, passengers saved $12.4 billion dollars in 1993.

(17)

2. Introduction to civil aviation 5

Smith Jr. & Cox (2008) updated Kahn’s conclusion with additional consequence. They adapted productivity to the rise of low-cost carrier (LCC)s.

Vasighet al. (2013) introduced another interpretation through reduced consol- idation at United States’ airports. Based on their book, deregulation led to the immense competition when LCCs tried to introduce themselves at major airport hubs - e.g., Frontier Airlines at Denver Airport; while the biggest airlines launched regional brands - e.g., United Airlines founded United Express, American Airlines introduced American Eagle and others.

However, there were also negatives which U.S. Government Accountability Office and economists studied. Goetz & Vowles (2009) summarised results of their studies and uncovered the fact that air tickets on routes, which were dom- inated by more than 60% by a single carrier, had become more expensive. In contrast, the quality of services provided by airlines had declined - poor cus- tomer service, delays, and congestion. Based on Kahn (1988) it was even worse, with 93 small towns losing their last air connection following the deregulation.

Goetz & Vowles (2009) also presented"ugly" results of deregulation - namely, changes in industry structure. Financial problems lead to many "bankrupt- cies, termination, mergers, and acquisitions" between 2000 and 2008 - including "stalwarts Delta, Northwest, United, and U.S. Airways" which declared bankruptcy during these years. U.S. Airways, which represented 7.8% of the domestic market in 2010, bankrupted twice in this period. However, it re- covered and merged with American Airlines in October 2015. Manuela et al.

(2016) reported financial results of this merger. Based on their study, U.S Air- ways and American Airlines strengthen their position in U.S. domestic market.

Their passenger load factor increased together with the number of passengers indicating better efficiency of the performance.

Alternatively, Fan (2020) focused on the merger of Continental to United, which finished in March 2012. In this article, Fan covered the impact of the merger on three types of routes. Leisure routes (routes between two leisure destinations), hub market (hub is one of the primary airports of air carrier), and big-city routes (routes attractive for both - leisure and business travelers) separately. Fan concluded that United (merged companies branded as United) strengthen its position on the hub market, gaining a dominant position. Their legacy rivals benefited from the merger as well, and together with United, they

(18)

increased airfares.

2.2 Essential Air Service

Changes in the whole aviation industry, as one of the consequences of deregulation, were expected. U.S. Government was concerned with imminent cancellations of air routes to remote towns and communities of United States. For that reason, U.S. Congress codified a new federal program called Essential Air Service (EAS) in 49 U.S. Code. Essential Air Service was launched in 1978 together with the deregulation of the aviation market. Current applicable rules are in §41731 - 41748 (49 United States Code 2018). This code defines basic essential air service as scheduled air transportation of passengers and cargo to a hub airport. This hub airport has to have convenient air service to a reasonable number of destinations. Alternatively, for communities in Alaska and eligible places further than 400 miles from standard hub airport, the hub airport has to offer a connection to another small hub or non-hub airport.

The eligible place is defined as any place in the United States, which was determined to be eligible in the Federal Aviation Act of 1958 (under section 419). Also, it can not be listed as ineligible by the Department of Transporta- tion. This place has to have at least ten enplanements¹ per service day during the most recent fiscal year (in place since October 2012). Another condition is subsidy cap - on average $1,000 per passenger during the most recent fiscal year. The distance of the eligible place from the nearest medium or large hub has to be more than 70 miles. Finally, the eligible place can be agreed in specific notices - e.g., during ’great recession.’ These conditions differ for places in Alaska and Hawaii that need to have direct approval from the Secretary of Transportation and 50 percent participation on subsidy from the state, local government, or a person.

However, these conditions were repeatedly changed, and the Department of Transportation did not enforce them. Based on Congressional Research Ser- vice report 2018, the 70-mile rule, adopted in the Department of Transportation and Related Agencies Appropriations Act (106th Congress (1999-2000) 1999) (P.L. 106-69), prohibited subsidies to communities in 48 states and Puerto

1"Number of passengers enplaning, at an eligible place, on flights operated by the subsidized essential air service carrier."

(19)

2. Introduction to civil aviation 7

Rico which are within 70 highway miles to the nearest hub airport. As a result, Hagerstown, MD, and Lancaster, PA lost their eligibility to be served as EAS communities and had to request a review of their mileage determinants.

Criticism of this program was raised because of its inefficient financing.

101st Congress (1989-1990) initially approved The Dire Emergency Supple- mental Appropriations Act of 1990 (P.L. 101-45) which had set subsidy cap to $300 per passenger. Subsequently, six communities became ineligible for Essential Air Service. In following years, the cap was several times lowered to $200 per passenger resulting in new federal law from 1999 which confirmed cap to this level. Based on Appropriations Act from 1999 no subsidy shall be provided to communities in 48 states and Puerto Rico that require the rate of subsidy per passenger above $200. This cap does not apply for airports and communities further than 210 miles from the nearest large or medium hub airport.

The number of Essential Air Service (_EAS) communities is relatively stable.

There was an increase following the ’great recession’; however, some eligible places lost their eligibility due to high subsidy per passenger or breaking other conditions. These communities usually switched to an alternative program - Alternate Essential Air Service.

In 2003, U.S. Congress updated 49 United States Code and in Vision 100 – Century of Aviation Reauthorization Act, P.L. 108-176 and in its section 405 defined new Community and Regional Choice Programs. These programs allowed keeping air connection to communities even they did not satisfy all conditions. The basic principle of the Alternate Essential Air Service (AEAS) program that it is more adjustable than the_EAS program. The Department of Transportation does not pay the subsidy to air carriers but to the local government, responsible for assisting in either air carrier or person operating air connection - e.g., in the form of on-demand flights. The first community to be awarded a subsidy through this program was Manistee, MI, in March 2012 (Kotler 2012).

Figure 2.1. shows the progress in the number of communities entitled to one of the subsidy programs. Eligible places in Alaska are shown separately because of Department of Transportation (_DOT)’s a unique approach to them.

(20)

Figure 2.1: Distribution of communities in subsidy programs

(21)

Chapter 3 Literature review

Based on academic literature, analysts use several parameters or concepts on measuring activity results in commercial aviation. Each of these focuses on a different feature of the air industry. Financially, operating revenues represent all earnings made from air transportation of companies, operating expenses work vice versa. Profits (subtraction of expenses and revenues) and net profit margins (profit after interest and taxes were applied) then show financial results of air companies - same as in any other industry. However, financial results alone do not represent commercial aviation performance at large as the others determinants.

Vasigh et al. (2013) in their book presented the variable called revenue passenger mile (RPM) or revenue passenger kilometer (RPK). This variable combines the number of revenue passengers with the distance flown by them.

Revenue passengers are passengers who paid for their air tickets; typical non- revenue passengers are infants, authorized personnel by the air carrier, air marshals, and others. The exact definition of revenue and non-revenue passengers is in U.S. Federal Code (Department of Transportation 2002).

The formula for revenue passenger mile:

RPM = Number of revenue passengers·Number of miles flown.

This formula is mainly used for one non-stop flight, i.e., the RPM of one flight represents the number of revenue passengers on this flight times the distance between origin and destination on one leg. Flights with layovers are divided into segments; a layover airport is considered a destination for the first

(22)

flight and origin for the second or another flight. For flights with no revenue passengers onboard, RPM equal zero. This variable’s simplicity gives analysts the option to use it in various ways - measuring performance on one route, one specific air carrier, or the whole market. U.S. Department of Transportation reports the revenue passenger mile as one of the crucial determinants of activity.

Alternatively, the capacity of one flight is replaced by an available seat mile (ASM) or available seat kilometer (ASK). This variable is similar to RPM because of its combination with the distance between origin and destination. It helps to identify the total capacity of a specific route or market, and it can also give information about the load factor (explained below).

The formula for available seat mile:

ASM = Number of seats·Number of miles of route.

The available seat mile is crucial to show the attractiveness of air routes or markets - what is its total capacity. Demand on some routes is highly above supply, and most of the flights are sold out (until new rotations are launched);

there are also less attractive routes that need to be subsidized by the government (Department of Transportation 2017). ASM equalling RPM means that the flight was sold out. Even though ASM can be zero by definition, it makes no sense to assume aircraft with 0 capacity or air route with 0 distance.

To report usage of capacity, transportation analysts use passenger load factor (LF). The load factor is expressed as the percentage of occupied seats from the total number of available seats. Passenger load factor can be derived as follows:

LF = RPM

ASM ·100% = Number of revenue passengers

Number of available seats ·100%.

Load factor is an essential variable for all types of airlines. The motivation of air carrier planners is to achieve the highest load factor possible. Various factors affect the final load factor on the market. Apart from airfare Jenatabadi

& Ismail (2007) compared other factors in the United States and Iran. They reported from their model covering the period from 1997 to 2006 that on the Iranian domestic market was digital reservation system and level of subsidy key

(23)

3. Literature review 11

determinants. Contrary to that, the U.S.’s total passenger load factor depends more on advertising expenses and flight departures in a given year.

3.1 The 9/11 attacks and their impact on aviation

On 11 September 2001, four aircrafts operating domestic service routes were hijacked by 19 terrorists and used for one of the worst terrorist attacks in history. Two aircrafts departing from Boston, MA, United Airlines and Amer- ican Airlines, crashed at both World Trade Center towers in New York, NY.

The third aircraft, which took off from Washington D.C. serving the airport in Dulles, VA, then stuck in Pentagon building, and the last aircraft, crashed in Pennsylvania before reaching its unknown target (Bergen 2003).

Based on a summary from CNN Editorial Research, there were 2 977 vic- tims and more than 25 000 injured people (Stempel 2019). These figures make them the deadliest terrorist attacks ever. Apart from human and society im- plications, there was another consequence of the attacks. New York’s economy suffered from dramatic job losses. In the following three months, approximately 143 000 workers lost their job each month in New York, NY, leading U.S. to recession (Bureau of Labor Statistics 2004).Roberts (2009) analyzed what the effect on macroeconomic development in months after attacks was. According to his study, 9/11 had directly resulted in a 0.5% decrease in real GDP growth.

The economy and labor market were not the only segments hit by the 9/11 terrorist attacks. Due to the possible threat of another terrorist attack, U.S.

Government shut down the whole aviation industry in the United States for two days resulting in losses of over 100 million USD in sales revenues just from these two days (Goodrich 2002). One of many reasons for this disaster was insufficient security measures that allowed hijackers to board planes fully equipped. Safety concerns resulted in demand decrease - Cohen et al. (2002) reported that revenue passenger mile in September 2001 declined compared to September 2000 by 30 percent.

All security deficiencies were summarised by Seidenstat (2004) who grouped them into three categories - inadequate security design and mismanagement, underinvestment in security, and security screening provided by the private sec- tor. The Bush administration was forced to present new measures to prevent

(24)

new terrorist attacks. Therefore, they presented in 107th Congress (2001-2002) (2001) new Aviation and Transportation Security Act. This act took control over security from airports and transferred this responsibility to Transportation Security Administration (TSA), launched 100 percent baggage screening, sup- ported the Air Marshal security system, and enforced more secure bulletproof cockpit doors.

As a result of new measures, the number of air marshals increased from 33 before the TSA takeover to estimated 2,800 (Cohen et al. 2002). Based on Seidenstat (2004), the regulated system brought an increase in total costs for passenger and baggage screening - from 600 million USD to 3.3 billion USD.

Security became tighter and flying inconvenient - new restrictions required passengers to arrive earlier at the airport, and the list of not-allowed items expanded.

Structural changes lead to some unexpected situations. While four major airlines - United, Northwest, U.S. Airways and Delta were responsible for 63

% of total decline of RPM, small regional airlines and low-cost carriers like Aloha Airlines or Jet Blue get an option to boost their performance (Guzhva

& Pagiavlas 2004)

3.2 Further disruptive events following the 9/11 attacks

Airlines learned their lesson after the 9/11 attacks. The crucial part for sur- viving a critical situation is a fast reaction. As Franke & John (2011) noted, airlines grounded a reasonable portion of their fleet in a short time immedi- ately after the start of the ’great recession’ and stayed efficient. The private consumers were not hit that dramatically as was expected. However, in the

’great recession’, airlines had to cope with the decline of demand by business travelers. Back then, Franke & John forecasted that long-haul flights and the low-cost carrier would have been winners of the economic downturn.

Pearce (2012) presented a complex overview of the effects of the ’great recession’ on the aviation market. In this paper, the author concluded that air cargo returned to pre-crisis numbers earlier. Moreover, the number of premium class

(25)

3. Literature review 13

passengers declined more intensively, and the return of premium passengers to aircraft was slower than economy class passengers. Pearce reported that return to pre-crisis’s number of passengers lasted only 18 months thanks to adequate preparation by airlines.

3.3 COVID-19 pandemics and civil aviation

Suau-Sanchez et al. (2020) performed an early assessment of the COVID-19’s impact. They evaluated available seat kilometers in the first months of 2020.

Based on their observation, year-on-year change was more dramatic on international markets with the decline of capacity by more than 80%. Alterna- tively, they compared the effect on low-cost carriers and full-service operators.

Low-cost carriers are more typical on domestic routes. Full-service operators shortened their capacity earlier.

Additionally, Suau-Sanchez et al. questioned possible long-term effects on both leisure and business travelers. While leisure travelers are expected to return to traveling earlier, and their foremost concern would be disposable income and health concerns, this is a more complex issue for business travelers.

Business travelers travel for two reasons: keeping contacts with their clients or attending MICE (Meetings, Incentives, Conferencing, Exhibitions).However, in 2020 videoconferencing entered a new era and became more common, which raised new concerns that business travel might not return to normal. Denstadli et al. (2013) observed in their regression, the rise of video conferences in years following the ’great recession’ did not cause less interest in business trips. How- ever, with more advanced technologies, this might not be the case. A possible slight reduction of 5-10 % of business travel could severely decrease airlines’

yields as business travel makes their prime part.

Goessling (2020) came with the idea that civil aviation can not be considered as a victim of pandemics but also as a vector of the spread of viruses and diseases. In this paper, the author used examples of airport malaria and the spread of SARS and MERS on how aviation increases person-to-person trans- mission.

Furthermore, Goessling (2020) questioned the overcapacity of aviation and the possible vulnerability of this industry. Combining several factors, he com-

(26)

pared COVID-19 pandemics to climate change and its further issues. Consul- tants from McKinsey, Dichter et al. (2020) found the same connection, and additionally, they reported that changes that are needed in civil aviation might be motivated by pandemics.

In terms of return to pre-COVID-19 figures, Czerny et al. (2021) analyzed the second largest domestic aviation market - China. Chinese governments introduced several schemes of support to civil aviation. The most significant benefit of China was its fast recovery following the outbreak of the virus in Wuhan. Czernyet al. noted that a crucial element for stabilization of the aviation market is how the pandemics are under control. They reported that it took four months for domestic routes to return to 70-80% of pre-pandemics demand.

Zhanget al.(2020) studied the spread of the virus, and in their regression analysis, they confirmed that air connection helps the virus to spread. Moreover, Oum & Wang (2020) who performed analyses of lockdowns confirmed that if society returns to normal early, it might have severe consequences.

In terms of similarities with the 9/11 attacks, COVID-19 brought a new wave of biosecurity measures. As Macilree & Duval (2020) reported, there were conflicts between countries’ restrictions and international aviation organi- zations such as International Air Transport Association (IATA) or International Civil Aviation Organisation (ICAO). As they pointed, for better sustainability, their measures should be unified and enforced.

Lambet al.(2020) worked on several versions of how to estimate willingness to fly using various methods of regression analysis. They divided the market into two parts, leisure and business travelers, and presented its readers with the results and ideas on how to continue their work.

(27)

Chapter 4 Data

In this chapter, we analyze the data that we used in this thesis. We have described sources from which we have collected our dataset and presented reasons for selecting specific variables. All data and information on which we have performed our models and analyses were from official sources - governmental institutions in the U.S. In each section of this chapter, we have commented on details on sources and crucial statistics.

In this thesis, we study topics related to the U.S. aviation market. Our main topic is to assess and compare the effects of various critical situations on the U.S. aviation market. In the previous chapter, we presented the reader with the effects of the 9/11 attacks on aviation. Our secondary topic is to perform a similar analysis on a subdivision of the U.S. domestic market. In the second chapter, we introduced details on U.S. governmental subsidy program - Essen- tial Air Service. This thesis also includes an initial study on how the COVID-19 pandemic affected the market. Accordingly, our primary data source was the U.S. Bureau of Transportation Statistics which collects all available data about the transportation industry in the United States.

Bureau of Transportation Statistics (BTS) started publishing detailed aviation statistics in 1990. However, until October 2002, there were no available data for flights operated by aircraft with a capacity lower than 50 seats. Since

EAS is a program, which is usually provided by small, local air carriers with small capacity aircrafts, data would not have satisfied our needs for modeling a subsidized market. For that reason, we have used monthly data from October 2002 till January 2021 (T = 220). All other information that we used in our

(28)

model was from a given period.

4.1 Civil aviation data

The models we used in this thesis compare the effect of COVID-19 pandemics on the whole scheduled segment of civil aviation with the effect on subsidized routes in the United States. Trafic figures are collected by Bureau of Trans- portation Statistics (BTS) from all air carriers using form 41 - Traffic. All carrier statistics are in two groups by the origin of a carrier - either U.S. carrier or all carriers. Each air carrier statistics group consists of six separate T-100 databases:

Non-stop segment Market

Domestic T-100 Domestic Segment T-100 Domestic Market International T-100 International Segment T-100 International Market

Total T-100 Segment T-100 Market

Table 4.1: Available T-100 databases

A fundamental difference between statistics on segment and market is the variety of included variables. Market databases include fewer summaries than segment - flight capacity data and number of departures are not included. Al- ternatively,_BTSpublishes in segment databases the following details in databases T-100 for non-stop segments: number of scheduled and performed departures, cargo details, distance, airborne time, ramp to ramp time, type of aircraft, number of available seats, and number of passengers. Routes are identified with codes of airlines and IATA codes of origin and destination airport ¹.

We focused on the relevance of the distinction between in-US traffic and out-of-US traffic during the database selection. Based on BTS’s statistics, international passengers represent approx. 18−23% of all PAX in the United States in recent years. However, COVID-19 had a more dramatic impact on routes to destinations outside of the U.S. because of travel restrictions that were in place. Additionally, in 2020 only 15.57% of PAX had flown abroad.

Intuitively, we concluded that this decrease was the result of pandemics and lockdown measures. Another issue in terms of data is their availability. Data on international routes are published with a delay of 5 months, so our analyzed

1IATA code is a three-symbol location identifier. International Air Transport Association defines these codes for nearly all airports - only small private airports are excluded.

(29)

4. Data 17

period would have been more limited. Finally, this thesis is focused on the domestic EAS program, which does not apply to international destinations.

Therefore, we decided to omit international routes.

Another category that had to be determined is the origin of air carriers operating flights on which we performed our model for the civil aviation industry in the U.S.BTSpublishes two separate datasets T-100 for either U.S. carriers or all carriers. Table 4.2. shows the number of passengers who traveled on domestic routes by either U.S.-based airline or foreign company from 2001 to 2019.

Other carriers served in the last 20 years only a minor part of the U.S. domestic market, and their share decreased compared to 2001. Furthermore, none of the Essential Air Service routes are performed by foreign carriers. There is no particular rule in the federal code which would prohibit non-U.S. carriers from applying for subsidies on EAS routes. However, there is a general condition on reliability, experience (especially for Alaska), and interline arrangements of the carrier - i.e., cooperation with more giant carriers operating from hubs allowing transfer of passengers and cargo on one ticket.

Year U.S. carriers Other carriers Year U.S. carriers Other carriers 2001 579 353 033 1 093 794 2011 655 901 614 419 256 2002 571 067 942 844 685 2012 657 303 771 341 537 2003 603 342 159 718 046 2013 659 759 296 298 344 2004 650 318 701 860 527 2014 676 100 861 257 797 2005 675 874 491 818 000 2015 707 523 169 226 029 2006 676 083 887 728 997 2016 730 650 148 252 155 2007 697 663 783 684 398 2017 752 338 659 241 370 2008 669 211 085 538 544 2018 787 692 231 231 254 2009 634 375 431 522 387 2019 820 809 224 261 801 2010 645 946 704 534 868

Table 4.2: Number of domestic passengers by origin of carrier

Similar to international routes, statistics on foreign-based carriers are published with a longer delay, leading us to a shorter analyzed period. Therefore, our model and analyses were performed on domestic non-stop segments operated by exclusively U.S. carriers. However, there are no direct data on ASM and RPM; thus we decided to perform simple multiplication of either available seats, or the number of passenger and distance to get these figures.

Unlike other details, identification of whether the specific route was subsidized through the EAS program is not included in the T-100 database. Eli- gibility that we described in previous chapters does not imply that a specific

(30)

community or airport has _EAS route. On the contrary, not all communities which have subsidized air connection were eligible according to federal code.

E.g., there are ineligible airports that are using the Alternative Essential Air Service (AEAS) program, small airports in Alaska might be subsidized partly by local government or private person to keep occasional air connection to bigger airports.

Department of Transportation regularly publishes an actual list of subsidized air routes since 1987 in their reports. The sample report can be found in Appendix. Table reports include all details about the community, its annual subsidy rate, air operator, duration of the current contract, hub served, air operator, and aircraft type. Each community and hub have own IATA codes.

From these reports, we defined subsidized code pairs.

Not all communities are directly served by an air carrier from the hub airport. Federal code defines conditions on which air carrier might adjust their timetable and include more _EAS communities on one rotation from the hub.

Usually, this is a case for groups of airports in Alaska - e.g., Island Air serves 11 _EAS communities on Kodiak Archipelago to Kodiak’s main airport (ADQ) with various routings. Therefore, we decided to include code pairs of these routes to secure the accuracy of our results. Finally, we have assigned to each data point from database T-100 information whether this route was provided by EASor Alternative Essential Air Service (AEAS) support system or not.

4.2 Other data

For modeling the aviation market, we had to include all relevant aspects of the number of passengers and revenue passenger miles. The aviation industry is part of the national economy of the United States (as well as in any other state). Therefore, we have decided to include details which describe the situation on the market. Based on Mankiw (2009) there are three significant macroeconomic variables on which we can measure performance in a specific economy - gross domestics product (GDP), inflation rate, and unemployment rate.

Gross domestic product is computed to describe how one’s economy performed in a specific time period. Mankiw (2009) presented two views of GDP -

(31)

4. Data 19

either as the total income of all subjects in the economy oras the total expendi- ture of the economy’s output of goods and services. In our model, we had to use a figure which would simulate the actual situation in U.S. Economy. However, the Bureau of Economic Analysis computes GDP only every quarter, which is not compatible with our needs - we have monthly data.

The inflation rate shows the change in prices between two data points. Usu- ally, it comes from a formula using the consumer price index (CPI). Govern- mental offices, including the Bureau of Labour Statistics, compute the current level of various prices of goods and services - e.g., transportation. Mankiw (2009) presented CPI as an index for the cost of living, costs of transportation might have an impact on aviation figures - leisure passengers are more price- sensitive than business passengers for whom CPI might be interpreted as an attribute of the economy’s actual situation. Therefore, we collected monthly published Chained consumer price index for Transportation in U.S. city average with base in December 1999 as value 100. Hence, we derived a 1-year change in CPI.

The last important macroeconomics variable is the unemployment rate. Ito

& Lee (2005a) in their first article regarding the impacts of the 9/11 attacks, they used a combination of the national labor force and unemployment rate.

The unemployment rate is widely considered to be a sufficient business cycle indicator (Mankiw 2009), while labor force "control for the long-term growth of the overall economy." These two figures are published monthly by the U.S.

Bureau of Labor Statistics.

The main topic of this thesis is to identify how strong was the effect of COVID-19 pandemics on the aviation industry and its parts (e.g., _EAS). We have underlined three options how to include pandemics in our model:

• dummy variable for months since the start of pandemics;

• total number of confirmed cases by COVID-19 RT-PCR tests;

• total number of deaths caused by COVID-19.

We identified a possible issue with the number of confirmed cases and deaths caused by COVID-19 disease during the literature review. The first months of pandemics were shocking, and the number of new confirmed cases and deaths

(32)

were not reaching their maximums. The number of passengers decreased im- mediately after the start of pandemics. Later, travelers adapted to life in pandemics and became less sensitive. For that reason, we chose to continue with a variety of dummy variables - more details to be found in the following chapters.

To simulate supply, we used the producer price index (PPI) for Jet Fuel.

Based on Beers (2020), fuel costs represent 10% to 15% of operating expenses.

Ferguson et al. (2009) analyzed in their study effects of jet fuel price changes on the market of airport LaGuardia in New York, NY, which is used only for domestic flights. On a 4-year period, they concluded that a 131% increase in jet fuel price resulted in a 15% increase in fares, a 29% increase in airline rev- enue, and a 59% increase in operating costs. data on jet fuel PPI are published every month by the Bureau of Labour Statistics with base in 1982 as value 100.

In our model and estimations, we had to include all possible events which might affect our explained variables. Ito & Lee (2005b) already in their study covered the first months in our model in which they identified war in Iraq and SARS epidemic as possible variables in estimating RPM. They also identified an issue with the overlapping of these two events. In terms of our topic, SARS epidemics was not that strong as COVID-19 pandemics, and the number of cases is irrelevant in comparison with the number of confirmed cases of COVID-19.

Besides their study, we tried to point out all other extraordinary events between October 2002 and January 2021 and potentially affect civil aviation.

In the proposal, we identified ’great recession’ as a possible source of change in aviation. ’Great recession’ was a long-term crisis that impacted the economy of the United States. However, we chose not to use a separate dummy variable for this crisis because business cycle macroeconomic variables are explaining changes consequences of recession on aviation.

Apart from SARS epidemics, we included other severe pandemics in our model - H1N1 pandemics - i.e., Swine Flu. The first confirmed case of H1N1 was on 15 April 2009, and a level of 1000 cases was reached in the following month. Based on the 2009 H1N1 Pandemic Timeline presented by Centers for Disease Controls and Prevention (2010) second and last wave of disease peaked in October 2009, and the pandemic got weak in January 2010 with an end

(33)

4. Data 21

announced by World Health Organization (_WHO) in August 2010.

The U.S. aviation market experienced several events during a given period - two mergers lead to structural changes on the market. Continental Airlines merged in March 2012 to United Airlines, and U.S. Airways merged in October 2015 to American Airlines. Based on the summary from Mudde & Sopariwala (2014) both these companies accounted for approximately 7.5% of domestic RPM. Therefore, we used time after each of these mergers as a dummy variable in our model.

4.3 Summary of variables

In the following table, we have summarised all non-binary variables which we used in our final models. Variables in bold are aviation-related data, which can be sorted into different groups - e.g., U.S. domestic market, EAS routes.

Variable Definition Description LABOR Civilian Labor Force in US total number UNEM Unemployment rate in US share (0-1)

FUEL Jet Fuel PPI 1982 = 100

CAPAC Total capacity of aircrafts total number PAX Number of passengers total number ASM Available seat mile total number RPM Revenue passenger mile total number

LF Load factor share (0-1)

Table 4.3: List of non-binary variables

Apart from non-binary variables, we used binary or dummy variables as well. In the following table, we have summarised all binary variables used in our final models to describe the "1" value.

Variable Definition VARIABLE = 1

SEAS Seasons (summer, Christmas) for June-August, December

IRAQ Iraqi War for 2/2003-5/2003

SARS SARS outbreak ror 3/2003-7/2003

SWINEFLU Swine Flu outbreak for 5/2009-1/2010

COVID19 COVID-19 pandemics for 3/2020-1/2021

CONTI Continental Merger for 3/2012-1/2021

USAirw U.S.Airways Merger for 10/2015-1/2021

LEAP Leap years 0,75 for February in leap year, -0,25 for non-leap February

Table 4.4: List of binary variables

(34)

Descriptive statistics

We identified variables expected to be used in our model in previous sections, including our explained variable. In the third chapter, we presented revenue passenger mile (RPM) as the figure, which is widely used as a characteristic for civil aviation performance. In our case, we used two main types of RPM - one for all domestic routes and another for_EASprogram and_AEASprogram routes.

The distribution of these variables is summarised in the following table:

In milions Mean Min 10% 25% 50% 75% 90% Max

RPM_total 48 495 2 565 39 751 44 008 49 188 54 121 59 399 71 374 RPM_EAS 32.31 4.41 15.59 19.27 30.43 42.03 52.42 80.80

Table 5.1: Descriptive statistics on revenue passenger mile (RPM) From table 5.1, we concluded that there are probably outliers for total figures because of the difference between the minimum and 10% quantile. Data on_EASroutes are more standardly distributed. Apart from_EAS,_AEASprogram is used for a shorter time, and we decided to include it to _EAS statistics.

During the preparation of the model, we considered the possibility to use lag analysis - more details to be found in the following chapter. To better understand the data and reaction of the U.S. Aviation market, we performed descriptive statistics on available seat miles (ASM). Similar to RPM, we performed statistics on our three categories - all domestic routes and subsidized

EAS and _AEAS routes:

(35)

5. Descriptive statistics 23

In milions Mean Min 10% 25% 50% 75% 90% Max

ASM_total 60 508 18 020 54 043 56 722 60 577 64 293 70 674 80 268 ASM_EAS 69,51 40.12 45.36 49 69.01 83.07 95.94 112.90

Table 5.2: Descriptive statistics on available seat mile (ASM)

Compared to RPM, on ASM descriptive statistics, we observed a minor difference between minimum and 10%. To confirm our expectations regarding outliers of data, we computed the z-score for RPM_total, RPM_EAS, ASM_total, and ASM_EAS. On data for the U.S. domestic market, we identified three outliers with a z-score under -3 - April-June 2020. From this, we confirmed the presence of shock from COVID-19 pandemics - similar to the 9/11 attacks. On the other hand, data of EAS routes have only one outlier with a z-score over 3 on RPM - August 2019. This outlier was caused by the increase in attractiveness of EAS routes in recent years and strong seasonality.

We analyzed seasonality on RPM to confirm strong seasonality in the aviation market. Figure B.5 shows distribution of z-score and figures B.3 and B.4 display trends and seasonality. These plots can be found in appendix B. There are clear patterns of seasonality with more passengers traveling in summer.

Alternatively, we analyzed data in the first months of 2020 more deeply:

Figure 5.1: ASM and RPM on all U.S. domestic routes in first months of 2020

(36)

Figure 5.2: ASM and RPM on all EAS routes in first months of 2020 Figures 5.1. and 5.2. indicate that reaction to COVID-19 pandemics was different for all domestic routes and EAS routes - it is reasonable to compare coefficients of separate models. Another interesting fact is that while capacity remained stable until March 2020, the number of passengers started to decrease already in February 2020 - an indication that there was one month lag.

Another exciting trend is progress on a load factor of either U.S. domestic market or EAS routes:

Figure 5.3: Progress of load factor

(37)

5. Descriptive statistics 25

Figures 5.3. show how the load factor progressed on both - the U.S. domestic market and EAS routes. Until 2020, the load factor of the domestic routes was always above 60% with the expected effect of seasons. However, the seasons’ effect was weaker for load factor than for RPM, indicating that airlines offer more seats to satisfy demand. Load factor onEASroutes had pos- itive trend until 2020 and while in 2003 it was around 40%, in 2019 was load factor for around 60%. It suggests that not only the number ofEASroutes and level of subsidy were increasing but also the popularity of these air connections.

In section Other data, we presented macroeconomic variables expected to be used in our model. Two of them - labor force and unemployment rate represent the current situation in the labor market:

Figure 5.4: Labor market in the United States

Based on the first plot, we confirmed the expected upward trend of the labor force with a decrease in 2020. As was expected ’great recession’ had an impact on the number of unemployed people. The unemployment rate peaked in 2010 after the recession and started decreasing to pre-crisis levels. Subsequently, another peak of the unemployment rate was in 2020, which partly proved one of the consequences of COVID-19 pandemics - the shock on the labor market in the United States.

In this chapter, we identified possible issues in our model. We know that both revenue passenger mile (RPM) and labor force are trended upward. This trend might cause an issue of spurious correlation. However, the Johansen

(38)

test which we performed confirmed that there is co-integration related to the time trend. We challenged this conclusion in the following chapters. Another discovery is the possible need to use one month lag for COVID-19 variables.

(39)

Chapter 6 Model and methodology

In this chapter, we present the models which we used for our estimation. We defined data collected for our regression analysis in the previous chapter - time-series data (stochastic process). Inspiration for our methods comes from Wooldridge (2013) and studies made by Ito & Lee (2005a).

There are various models used for regression analysis. In our case, we decided to work with the ordinary least squares (OLS) method for time series data. We did not use panel data because we do not have cross-sectional parameters - our data are for the whole U.S. market. We rejected the possibility to use analysis by states because of the complexity of the issue - the majority of routes are between two states, and analysis by the state of origin or destination would divide our dataset into 51 parts. A simple static model for time-series data is defined as

yt=β₀+β₁zt+ut, t= 1,2, ..., n (6.1) in which y_t is explained variable, z_t is explanatory variable, u_t is error term with t as time parameter. OLS method estimates coefficients for all explanatory variables. It thus is preferable for our analysis in which we tried to compare the impact of COVID-19 on the U.S. market with the effect of the 9/11 attacks and then confirm or reject our hypotheses regarding different results on subsidized routes.

To have unbiased estimations, our models have to satisfy time series Gauss Markov assumptions. During our study, we had to perform further testing to verify whether these assumptions hold. In the previous chapter, we identified

(40)

a possible need to use one month lag for COVID-19 variables. This knowledge was reflected during the preparation of the model. Therefore we present a finite distributed lag model (FDL). FDL model of order one is defined as

y_t=α₀+δ₀z_t+δ₁zt−1+u_t. (6.2) In this case δ₀ represent immediate effect - i.e. immediate propensity. Perma- nent change is defined as δ₀+δ₁ - i.e. long-run propensity.

As we stated above, we planned to use more models and compare coefficients. The primary model for us is the model which explains RPM on U.S.

domestic routes operated by U.S. carriers. Parameter RPM was used in loga- rithmic form to avoid significant differences between high-value figures in the model. In our model, we used some explanatory variables in log-form as well.

Our second model is the model for RPM on EAS routes. During the model selection, we identified various versions of how to estimate COVID-19 as the explanatory variable.

For simplicity, we defined two sets of explanatory variables:

• X_t^′ represent all no-binary variables: log(LABOR)∗UNEM, FUEL and linear TREND;

• D^′_trepresent all dummy variables: CONTI, USAirw, SARS, IRAQ, LEAP, SEAS and SWINEFLU;

6.1 Model 1 - dummy variable with one month lag

The first type of model is the model in which is COVID-19 accounted as a simple dummy variable. This model expects that COVID-19 pandemics resulted in stable change with one month lag. However, it does not consider the actual situation (waves) or adaptation to life in pandemics. This model is therefore expected to be less accurate. Model is defined for the U.S. market as

log(RPM_total) =β₀+β_XX_t^′ +β_DD_t^′ +β_COV-19D_COV19+ +β_COV-19_t−1D_COV19_t−1 +u_t,

(6.3)

(41)

6. Model and methodology 29

and for the EAS rotes as

log(RPM_EAS) =β₀+β_XX_t^′+β_DD^′_t+β_COV-19D_COV19+

+β_COV-19_t−1D_COV19_t−1 +u_t. (6.4) In these models, D_COV19 has value 1 for all months from March 2020 and D_COV19_t−1 takes value 1 for all months from April 2020. Otherwise, these variables has value 0.

6.2 Model 2 - dummy variable, long-term

Pandemics of COVID-19 have been developing during this time. Not only its intensity but also the behavior of people changed in time. While in the first months, there was a shortage of personal protective equipment (Burki 2020) and some states initially imposed travel restriction between states, in following months, people got ready for pandemics, and traveling became safer and more accessible for the U.S. population.

In this version of the model, we decided to include such elements to dis- tinguish between short-term and long-term effects. Ito & Lee (2005a) in their studies used such variable for the effect of 9/11 attacks. For our analysis, we defined the effects of COVID-19

COV_estim =β_COV-19D_COV19+β_COV-19_1/m 1

COV19_1/m, (6.5) in whichD_COV19is a dummy variable equalling 1 for all months from April 2020 and COV19_1/m is the number of months since the start of pandemics - i.e., for April 2020 equals 1, for May 2020 equals 2. We considered using another, slower-growing version, which would expect to return to normal later. These other methods would be applicable if we would have an accurate estimate of when the pandemics would end.

Model for the U.S. domestic market is defined as

log(RPM_total) =β₀+β_XX_t^′+β_DD^′_t+ COV_estim +u_t, (6.6)

CHARLES UNIVERSITY FACULTY OF SOCIAL SCIENCES