Developing an ‘Airline Rating Model’

Rijksuniversiteit Groningen

Master of Science of Business Administration

International Financial Management

Developing an

‘Airline Rating Model’

to indicate an airline’s expected short term

general performance and its performance

at Amsterdam Airport Schiphol

Schiphol Group

Aviation Statistics & Forecast

Developing an

‘Airline Rating Model’

to indicate an airline’s expected

short term general performance

and its performance at

Amsterdam Airport Schiphol

Joyce C. Gardner

Rijksuniversiteit Groningen

International Financial Management

May - August, 2009

Schiphol Group Supervisors:

Preface

This paper is not only the finalization of my internship within Schiphol Group and the Master International Financial Management; it also marks the end of my study at the Rijksuniversiteit Groningen. During my time as a student I have learned a lot and developed myself both personally and academically. I would like to thank the people that contributed to this unforgettable period, which I deeply treasure.

Special thanks go out to my parents John and Everdien, who endlessly support and encourage me. My boyfriend Roché Sulley has always been very supportive, giving me feedback and energy in both a personal and academic way.

I’d like to thank Daniel dos Reis Miranda for giving me the opportunity to write my thesis at Amsterdam Airport Schiphol, and Hylke ter Beek for his guidance, support and encouragement throughout the process of developing this model. My internship at Schiphol has been a very instructive, fun and interesting experience.

Management Summary

An airport’s dependency on airlines is one of its largest operational risks. Forecasting an airline’s performance and knowing what effect it might have on the airport’s operations can prepare airports for possible flight cancellations, it creates the possibility to reinforce client-relations at the right time and it aids in making more accurate forecasts.

The objective of this research is to answer the following management question for Amsterdam Airport Schiphol:

Is it possible to define and predict an airline’s performance and the impact of its airport-specific performance on Amsterdam Airport Schiphol in the short term?

The answer to this question lies in the Airline Rating Model; a tool with which an indication about an airline’s expected performance in general and in airport-specific terms can be made. Previous research resulted in specific models predicting one aspect of an airline’s performance, either financial or non-financial performance. This is the first prediction model that combines all factors influencing an airline’s performance. Moreover, it is the first study that tries to predict an airline’s performance at a specific airport. Amsterdam Airport Schiphol was used as a case study.

The model is constructed with data from the year 2008 of 45 of the largest airlines operating at Amsterdam Airport Schiphol. By using information mainly derived from annual reports, it ranks airlines on their external, financial, non-financial and airport-specific performance. The independent variables were regressed (using manual stepwise multiple regression) on the dependent variables; scores assigned to each aspect of an airline’s performance by Schiphol-experts.

The external performance is influenced by whether or not the airline is related to an alliance and by the political stability of its home country. The latter is to such an extent correlated to the country’s economic situation that GDP or inflation figures are not included in the External Model.

The financial health of an airline is influenced by its turnover, revenues, leverage, liquidity and its average margin. The Financial Model (Pilarski and Dinh, 1999) estimates the bankruptcy risk of an airline and correlates with the payment behavior of airlines signaled by Dun & Bradstreet.

The Non-Financial model is based on the conceptual foundation of Gudmundsson (2002) and combines labor productivity, traffic related ratios, governmental influence1 _{and fleet characteristics.}

The latter involves the average age of an airline’s fleet and the average aircraft size. This research resolves the ongoing discussion about the most efficient aircraft size; a fleet mainly consisting of wide-body aircraft is likely to have a higher non-financial (operational) performance.

The performance of an airline at a specific airport is tested with Amsterdam Airport Schiphol as a case example. Variables influencing this performance are an airline’s average weekly frequency per destination, its load factor and its punctuality. With a prediction accuracy of 85% it indicates the probability of an airline ceasing its operations in the subsequent year.

Index

1.

I

NTRODUCTION

... 8

1.1 Problem Statement ... 8

1.2 Amsterdam Airport Schiphol... 9

1.3 Introduction to the theoretical framework ... 10

2.

C

ONCEPTUAL AND

T

HEORETICAL

B

ACKGROUND

... 12

2.1 External Variables ... 12

2.1.1 Economic situation ... 12

2.1.2 Political situation ... 13

2.1.3 National health ... 14

2.1.4 Perception of safety ... 14

2.1.5 Member of an alliance ... 15

2.2 Financial Variables ... 16

2.2.1 Operating revenues/total assets ... 17

2.2.2 Retained earnings/total assets ... 17

2.2.3 Equity / debt ... 18

2.2.4 Current assets / current liabilities ... 18

2.2.5 EBIT / operating revenues... 18

2.3 Non-Financial Factors ... 18

2.3.1 Fleet age... 20

2.3.2 Variety of aircraft brands operated ... 20

2.3.3 Departures per aircraft ... 20

2.3.4 Load factor ... 20

2.3.5 Passengers carried per departure ... 20

2.3.6 Hours flown per pilot ... 21

2.3.7 Employees per aircraft ... 21

2.3.8 Governmental influence ... 21

2.3.9 Economic factors ... 21

2.3.10 Aircraft size ... 21

2.3.11 Aircraft ownership and lease contract flexibility ... 22

2.3.12 Hedging strategy ... 23

2.4 AMS Specific Variables ... 24

2.4.1 Load factor ... 24

2.4.2 Frequencies per route ... 24

3.3 Methodology ... 34

4.

R

ESULTS

... 38

4.1 Dependent Variables ... 38

4.2 The Airline Rating Model ... 39

4.2.1 General Model ... 39

4.2.2 External Model ... 40

4.2.3 Financial Model ... 41

4.2.4 Non-Financial Model ... 42

4.2.5 AMS-Specific Model ... 44

5.

O

UTCOME

:

T

HE

A

IRLINE

R

ATING

R

EPORT

... 47

6.

V

ALIDITY AND

R

ELIABILITY

... 50

7.

C

ONSLUSION

... 56

7.1 Implications for the existing literature ... 56

7.2 Implications for airports ... 57

7.3 Limitations ... 58

7.4 Recommendations for future research ... 59

8.

R

EFERENCES

... 61

8.1 Websites ... 61

8.2 Annual reports ... 62

8.3 Literature ... 62

A

PPENDIX

I

S

AMPLE

... 65

A

PPENDIX

II

A

IRLINE

R

ATING

F

ORM

... 66

A

PPENDIX

III

E

XTERNAL

V

ARIABLES

... 67

III.1 Economic Situation ... 67

III.2 Political Situation ... 67

III.2a Political Stability and Absence of Violence /Terrorism ... 67

III.2b Political Risk ... 68

III.2c Terrorism Threat ... 69

III.3 Mexican Flu Infections ... 70

III.4 Safety of an airline ... 70

III.5 Member of an Alliance ... 71

A

PPENDIX

IV

F

INANCIAL

V

ARIABLES

... 72

IV.1 Financial figures ... 72

IV.2 Financial ratios ... 73

IV.3 Financial Risk Rating ... 74

A

PPENDIX

V

N

ON

-F

INANCIAL

V

ARIABLES

... 76

V.1 Fleet Information ... 77

A

PPENDIX

VI

AMS-S

PECIFIC

V

ARIABLES

... 78

A

PPENDIX

VII

T

HE

A

IRLINE

R

ATING

M

ODEL

... 79

A

PPENDIX

VIII

V

ALIDATION

... 85

VIII.1 SkyEurope ... 85

VIII.2 Delta Airlines ... 86

VIII.3 United Airlines ... 87

VIII.4 Sterling Airlines ... 88

1. Introduction

Airports are exposed to various risks as part of their business activities. These risks can be of strategic, operational or financial nature, or may be related to compliance with statutory rules and regulations. One of the most important operational risks, besides insufficient safety and security, is the dependence on third parties. To a large extent, the smooth operation of an airport depends on the efforts of third parties, such as air traffic control, baggage handlers, Customs, authorities, governmental institutions, the Military Police and of course, the airlines. This research focuses on the latter and aims at aiding an airport in forecasting airlines’ performance in general and with relevance to a specific airport, in this case Amsterdam Airport Schiphol.

1.1 Problem Statement

The airline industry is one of the most vulnerable and volatile industries in the world. After a period of growth and expansion, the industry’s landscape changed during the turbulent years of the terrorist attacks of September 11, 2001, the Iraq war and the SARS epidemic. The current economic crisis has decreased passenger growth to an all-time low; in the first quarter of 2009, passenger decline reached a post World War II level2_.

The lower number of passengers obviously leads to a decreased performance for airlines. Operating losses may lead to the decision to cut cost and eliminate destinations from the network. Other factors influencing this decision are for instance external factors such as the rapidly spreading Mexican flu, or operational factors such as the average load factor on a route.

Anticipating on this potential loss is crucial for an airport’s competitive position. It facilitates the short term forecasting process and thus the preciseness of the budgets. It could also prevent airlines from actually canceling the destination from its network. And although predicting a further decline or a modest growth in passenger flows is an everyday business for airports, the current turbulent times are making this increasingly difficult3_.

To facilitate this forecasting process, this research aims at developing a tool, an ‘Airline Rating Model’, with which an indication about the expected short term performance of an airline can be made. Besides general performance, performance at a specific airport is also taken into account. The case example this research is based on is Amsterdam Airport Schiphol. Hence, the main research question of this study is:

Is it possible to define and predict an airline’s performance and the impact of its airport-specific performance on Amsterdam Airport Schiphol in the short term?

2 _{Source: Airline Value Analysis Model, Schiphol Group}

3 _{Long term forecasts in the air transport industry are especially hard to make. Even the two main aircraft}

In order to answer this question, the following sub questions need to be answered:

• To what extent is an airline’s general performance impacted by external, financial and non-financial factors?

• What variables determine an airline’s external, financial and non-financial performance? • What variables determine an airline’s performance at Amsterdam Airport Schiphol?

• Is there a relationship between an airline’s general performance, and its performance at Amsterdam Airport Schiphol?

• Is there a relationship with an airline’s performance at Amsterdam Airport Schiphol and its decision to cease operations and cancel Amsterdam from its network?

1.2 Amsterdam Airport Schiphol

To assess the impact of an airline’s performance at a specific airport, Amsterdam Airport Schiphol (AMS), or Schiphol Airport, is used as a case example. Schiphol Airport is Europe’s fifth largest passenger airport4_{, and third largest in terms of cargo}5_{. With its 47.4 million passengers in 2008, it}

ranks fourteenth worldwide. Schiphol is the only airport still residing in the same place as it was once established, 3.4 meters below sea-level in the Haarlemmermeerpolder. It is located 20 minutes outside Amsterdam and one hour away from the harbor of Rotterdam. Schiphol’s core function is that of ‘connecting’. Its network of 262 destinations is the main reason foreign and domestic businesses choose to establish operations in the Randstad area and, as such, forms an important driver for the Dutch economy. Moreover, 43% of Schiphol’s passengers are transfer passengers, flying to Schiphol mainly for its ideal geographical location, its connectivity and its network. This network is one of Schiphol’s main competitive advantages. It fully relies on the performance of the airlines.

Knowing what performance to expect from these airlines and which airlines are likely to cease operations at Schiphol can make an important difference for the airport. It creates the possibility for account managers to reinforce client relations before the actual decision of ceasing local operations is made. Moreover, it creates the possibility for marketing managers to timely find another carrier that might fill the gap in the network. Hence, knowing what to expect might prevent the (negative consequences of) actual cancellations. Furthermore, by using the Airline Rating Model, forecasts about expected passenger traffic become more accurate. This leads to accurate budgeting and precise availability of staff and facilities, reducing costs of over- or under capacity.

The Airline Rating Model is created by using multiple regressions and 2008 data of 45 large airlines currently operating at Schiphol Airport.

1.3 Introduction to the theoretical framework

The Airline Rating Model consists of five different models; the External, Financial and Non-Financial Model; a summary of those three is the General Model; and the fifth model is the AMS-Specific Model6_{. While the latter is new in this field of research and has not been investigated before, the}

impact of non-financial and especially financial variables on airline’s performance has often been written about.

Edward Altman (1968) was the first to develop a failure forecasting model by combining several variables in a multiple discriminant regression. His model is adjusted many times and applied to various industries. Pilarski and Dinh (1999) used Altman’s Z as a base for their model which identifies the risk for bankruptcy specifically for airlines. Their model consists of five ratios, and does not require an individual to invest significant amounts of time or expense in credit research. This latter aspect is important for a model’s manageability, which is an essential condition for the success of the ‘Airline Rating Model’. Besides this financial model which is of course of utmost importance for an airline’s operations and payment behavior at an airport, a second model is integrated into the Airline Rating Model.

In 2002, Gudmundsson developed a non-financial prediction model that could be applied worldwide, taking into account differences in the political and economic environment of airlines. As opposed to Pilarski and Dinh, Gudmundsson´s focus was on predicting distress preceding bankruptcy rather than bankruptcy per se, and on operational ratios instead of financial ratios. His model was distinctive in the sense that his conceptual foundation was more robust than any prior research in this field, making it very interesting to include in the Airline Rating Model. Gudmundsson’s focus on variables such as political and economic influence also led to the inclusion of an external factor in this model.

While the aforementioned models both describe only part of an airline’s health and efficiency, this research aims at covering multiple aspects of an airline’s performance and summarizing it in one single overall score. Moreover, this model is the first to assess the relationship between an airline’s performance at a specific airport and its decision to continue its local operations or to cancel the destination from its network.

In the next section the factors that are of influence on an airline’s general and airport-specific performance are introduced; the external, financial, non-financial and AMS-specific factors. These factors are conceptualized by defining several variables. While the financial variables and most of the non-financial variables are derived from respectively the Pilarski and Dinh (1999) and the Gudmundsson (2002) research, all of the external and some of the non-financial variables are based on different theories7_{. Section 3 starts by describing the sample used in this research. Subsequently}

the data collection process and the applied methodologies are reflected upon8_{. The multiple}

regressions result in an external model, a financial model, a non-financial model, a summary of

6 _{Because the possibility to apply the AMS-Specific Model to other airports as well (which will be elaborated on}

in paragraph 7.2), this model might sometimes be referred to as the airport-specific model.

these three; the general model, and the AMS-specific model. The weight and significance of each variable in the models are all presented in the section 49_{. The findings of the model for the 45}

sample-airlines in 2009 are presented in the Airline Rating Report in Section 5.

2. Conceptual and Theoretical Background

The aim of this research is to create a model that can indicate an airline’s expected short-term performance in general and in Schiphol-specific terms. As stated before in the previous chapter, the model will take into account external factors that are affecting airlines’ firm performance, general indicators to measure an airline’s financial health and operational efficiency, and evidently some specific Schiphol-related operational factors (the AMS-Specific-Factor). This section gives an overview of the conceptual and theoretical background of all the factors included in the model. Figure 1 shows the relationship of those factors.

Figure 1: Building blocks of the Airline Rating Model

2.1 External Variables

Travel and tourism is the world’s largest industry and also represents the top three industries in many countries (Goeldner, Ritchie, and McIntosh, 2000). The airline sector is largely dependent on travel and tourism and it is thus important to include a measure that takes factors of influence on this industry into account. In this section those external factors - economic and political situation, national health, perception of safety and being a member of an airline- are conceptualized and their theoretical background is emphasized.

2.1.1 Economic situation

The economic situation of a country is inextricably bound up with the level of commercial activity, in business as well as leisure. An appropriate estimate for a nation’s economic well-being is its inflation and GDP growth percentage. Experts at Schiphol researched the in- and outbound passenger flows and its relation to the GDP of the origin as well as the destination. The aggregate outcome is a model that forecasts future passenger developments by means of using the expected GDP growth of the nations it concerns. An average multiplier of 2 was found, meaning that 1%

change in the GDP growth on average results in 2% change in passenger volume. The outcomes of this model are assumed to be relevant for the short term performance of airlines serving Schiphol and are integrated into the Airline Rating Model.

Another variable that indicates the economic situation of a country is its inflation rate, the year-on-year percentage change in Consumer Price index. Both figures are summed up in Appendix III.1.

2.1.2 Political situation

Travel surveys consistently find that safety and security are important concerns among tourists (Poon and Adams, 2000). The risks associated with political instability and potential violence or terrorism have been identified as particularly influential in changing travel intentions, even among experienced travellers (Sömez and Graefe, 1998a). Therefore, an indicator to measure the political (in-)stability of a country, derived from the Worldwide Governance Indicators (WGI) project, is included in this model. The WGI reports governance indicators for 212 countries and territories over the period 1996–2007, for six dimensions of governance: ‘Voice and Accountability’, ‘Political Stability and Absence of Violence/Terrorism’, ‘Government Effectiveness’, ‘Regulatory Quality’, ‘Rule of Law’ and ‘Control of Corruption’. The indicators combine the views of a large number of enterprise, citizen and expert survey respondents in industrial and developing countries. The individual data sources underlying the aggregate indicators are drawn from a diverse variety of survey institutes, think tanks, non-governmental organizations, and international organizations10_.

‘Political Stability and Absence of Violence/Terrorism’ is integrated in this model, the others are irrelevant to airline performance. This indicator measures the perceptions of the likelihood that the government will be destabilized or overthrown by unconstitutional or violent means, including domestic violence and terrorism. Appendix III.2a shows a world map that ranks all countries in terms of political stability.

Since this variable includes political stability as well as absence of violence/terrorism in one score, there is a loss of information concerning what factor impacts airline performance the most; political stability or a safe and violence-free environment. To avoid bias or a loss of information the impact of both factors is double checked by including two risk assessments of AON11_{, Political Risk and}

Terrorism Threat. Political Risk per country is shown in Appendix III.2b and includes risks of exchange transfer, strike, riot and civil commotion, war, terrorism, sovereign non-payment, legal and regulatory risk, the risk of political interference and supply chain vulnerability. The aviation industry, among others, can be largely impacted by these risks. Terrorism threat per country is shown in Appendix III.2cand takes several terrorist groups into account (i.e. far right/reactionary, far left/revolutionary, kidnap, single interest, nationalist/separatist, national Islamist, global Islamist and other religious/extremist cultist). Both risk maps were created in 2009 by polling AON analysts and insurance underwriters in the UK, the US and other countries.

10 _{Source: http://info.worldbank.org/governance/wgi/index.asp. The aggregate indicators do not reflect the}

official views of the World Bank, its Executive Directors, or the countries they represent. The WGI are not used by the World Bank Group to allocate resources or for any other official purpose.

11 _{Aon Corporation is the leading global provider of risk management services, insurance}

2.1.3 National health

The air traffic industry is one of the most vulnerable sectors to outbreaks of worldwide infectious diseases. Traveling by air (especially to infected areas) becomes less attractive, particularly if the disease is transferable via face-to-face contact12_.

Between November 2002 and July 2003 the outbreak of SARS (Severe Acute Respiratory Syndrome) caused a huge shock in worldwide air travel4_{. Within a matter of weeks SARS spread from the}

Guangdong province of China to rapidly infect individuals in some 37 countries around the world. Infected cases mounted up to 8096 with 774 fatalities, the vast majority of which were in China, Hong Kong and Taiwan. As a result, 25% of the total scheduled flights on Hong Kong International Airport were cancelled13_.

Currently the world is facing the rapidly spreading influenza A (H1N1) also called the Mexican flu. First infections were identified in Mexico City at March 18 2009 and at June 11 2009 the World Health Organization declared the Mexican flu to be a pandemic14_{. This is the first pandemic since}

the Hong-Kong flu in 1968-1969, which killed an estimate of one million people worldwide15_.

Worldwide, pharmaceutical organizations started the mass production of the antiviral drug oseltamivir (known as Tamiflu)16_{, but the first cases of H1N1 viruses which are resistant to the anti}

drug have already appeared17_.

In order to include the risk of less passenger travel due to infectious diseases, national health18 _is

included as a third external variable in the Airline Rating Model.

2.1.4 Perception of safety

The risk of being involved in a fatal commercial aircraft accident is approximately one in 12.5 million19_{. In the United States, flying is 22 times safer than traveling by car and contrary to current}

perceptions about the mortality rate of airline accidents, data collected for the seven largest U.S. domestic carriers reveal that only 14% resulted in fatalities, 84% resulted in serious injuries and 71% resulted in minor injuries (Squalli, 2006). However, despite this impressive safety record, when accidents occur, people appear to form distorted perceptions about an accident’s survival rate. Professor Jay Squalli is the first to measure the impact of different accident severity levels on enplanement. He finds that safety perceptions about accidents with minor injuries have no statistically significant impact on enplanement, while perceptions about accidents with serious injuries and fatalities lead to cumulative decreases in enplanement. His results indicate that, perhaps due to the extensive media coverage that large disaster carriers face, passengers do not necessarily feel safer with larger airlines over time. In 2009 he investigates the impact of accidents

12 _{http://www.who.int/csr/sarsarchive/2003_03_27/en/} 13 _{That is, in April 2003.}

14 _{http://www.reuters.com/article/topNews/idUSTRE55A1U720090611} 15 _{http://www.jsonline.com/news/usandworld/43705182.html}

16

http://www.asia-manufacturing.com/news-314-shanghaipharmaceuticalgroup-roche-tamiflu-glaxosmithkline-news6.html, http://www.silobreaker.com/roche-steps-up-production-of-tamiflu-5_2262312416660422656

17 _{http://www.who.int/csr/disease/swineflu/newsbriefs/h1n1_antiviral_resistance_20090708/en/index.html} 18 _{as becomes clear in this sub paragraph, the term ‘national health’ in this research reflects ‘suddenly changing}

on advertising expenditures. He finds that airlines reduce their advertising efforts after a fatal accident because media coverage can distort and influence consumer perceptions. This may indicate the carriers’ fear of signaling desperate attempts at selling a service of low quality, particularly in the presence of the constant ‘memory refreshing’ role that the media plays following accidents. In fact, because flying is an experience good, advertising cannot be informative. Claims that the offered service is of high quality could thus be perceived as misleading and potentially rejected by consumers (Squalli, 2009). The lower marketing expenditures might result in an even lower enplanement level and thus lead to a negative impact of an airline’s operations.

2.1.5 Member of an alliance

An airline alliance is an agreement between multiple airlines to cooperate on a substantial level. It provides a network of connectivity and convenience for international passengers and international packages. The first large alliance started in 1989, when Northwest and KLM agreed to code sharing20

on a large scale. This agreement developed into SkyTeam, one of the three large alliances, which carried over 462 million passengers to 905 destinations worldwide in 2008. The other two large alliances are known as Star Alliance and oneworld.

There are several advantages to being a member of an airline alliance. The most important one is an extended and optimized network (Chen and Chen, 2003). It also reduces costs by sharing sales offices, maintenance- and operational staff or facilities (e.g. catering or computer systems). Being a member of an alliance can also be a form of back up. If an airline faces decreasing sales, its codeshare-partners can set this off by their sales and the operating airline does not necessarily have to cancel flights. Hence, alliances can serve as a kind of safety net for shocks in demand.

Airlines cannot just join an alliance; they have to be allowed into it. An airline becomes eligible when it adds value to the network, and when it might contribute to the load factors of the other airlines. A first condition that must be met is a stable political and economic environment in the home country of the airline (Chen and Chen, 2003). This implies that being an alliance member -even a regional or partner member- reflects a common trust in the airline’s home country and can thus serve as a good measure for the environmental stability.

Factors representing General Performance

After having discussed the environmental stability on which an airline cannot exert any influence, the following building blocks of the model describe two types of general airline performance; financial and operational performance. Two existing models have been used, the financial Pilarski Score model (1999) and Gudmundsson’s non-financial model (2002). These logistic regression models are very suitable for predictions of airline distress in the short term. Gudmundsson (1999) performed a comparison of the prediction accuracy of various models and found that non-financial models

19 _{Based on the 539 worldwide fatalities in 2008 reported by the British aviation consulting firm ‘Ascend’, and the}

6.74 billion world population of 2008 reported by the U.S. Census Bureau.

20_{Code sharing enables multiple airlines to sell seats on the same flight. It refers to a practice where a flight}

perform better than financial models two and three years prior to distress, whereas financial models perform better than non-financial models one year prior to distress. Hence, by using both models the overall prediction accuracy in the short term (one to three years prior) is more reliable. The models are not suitable for longer term predictions. This is partly due to the static nature of the models, not taking cycles of demand into account (Clarke et al. 2005).

2.2 Financial Variables

The most obvious reason for an airline to leave an airport is financial distress, or even worse, bankruptcy. Failure prediction models have been used extensively to evaluate the financial performance of a company and as early warnings systems of potential business failure (Theodossiou, 1991; Gudmundsson, 2002). Four main approaches have been used in the development of prediction models; Univariate Analysis, Multivariate (Discriminant) Analysis (MDA), Conditional Probability and Neural Networks21_{. The Univariate Analysis was first performed by Beaver (1966) and resulted in one}

ratio (cash flow to total debt) that was able to classify a firm as likely to fail or not. Although his predictor performed fairly well, the main difficulty with this approach is that the classification could take place for only one ratio at a time, while other ratios might impact this classification as well. Altman (1968) argued that a model with multiple variables was needed to accurately predict the risk of bankruptcy, and he developed the Multivariate Analysis. Nowadays this method is the most widely applied method in failure prediction literature (Gudmundsson, 2002).

Altman combined several financial ratios to assess four aspects of a firm’s financial well-being. These four types of ratios measure liquidity (the ability of a firm to pay its obligations as they come due), leverage (the extent to which a firm uses debt as a method of finance), turnover (the efficiency of asset usage), and profitability (Gritta et al., 2006). This resulted into a model that produced an overall score, the Z-score that measures the extent to which a firm fits a bankruptcy profile. In 1983 Altman suggested the use of a variant of the original Z-Score, the Z’’ Score model, to measure financial health specifically in the service industry22_{. Altman’s Z-score model is the most}

commonly used MDA model to predict bankruptcy (Clarke et al., 2005).

Based on Altman’s model, Pilarski and Dinh (1999) computed an airline-specific bankruptcy prediction model. They used logistic regression analysis to estimate the probability of bankruptcy and to rank airlines in terms of financial strength. Rather than producing a score that must be compared to a scale, as is the case with the previous models, this model produces the probability of bankruptcy. P is that probability. The higher the P value, the greater the financial stress and the more likely is the chance of failure.

The Pilarski Score Model was based on quarterly data for 36 bankruptcies and 280 non-bankruptcies. Although this sample is overwhelmingly composed of non-bankrupt observations it did not bias the results. Over fifty-five variables relating to profitability, efficiency, asset utilization, liquidity,

21 _{For a description about the latter two methods and models developed using those methods see Gudmundsson,}

2002.

22 _{The Z’’ Score model was applied to the airline industry and found to be reliable (Altman and Gritta, 1984). It}

leverage, and capital adequacy were considered. The final model consists of the five most significant variables (three of which –F1, F2, and F3- were also represented in Altman’s Model), which cover all of

the earlier mentioned aspects of a firm’s financial well-being;

F1 = operating revenues/total assets (turnover)

F2 = retained earnings/total assets (earnings)

F3 = equity/total debt obligations (leverage)23

F4 = liquid assets/current maturities of total debt obligations (liquidity)

F5 = earnings before interest and taxes/operating revenues (margin)

P is then calculated by z e P + = 1 1 Where Z = -1.98F1 –4.95F2 –1.96F3 –0.14F4 –2.38F5

And e = 2,718 (the base for natural logarithms)

A P-score above 38% indicates a significant bankruptcy risk in the short term. The accuracy rate of the Pilarski Score Model was 98.5% and it predicted insolvency on average 4.1 quarters before the carrier filed for bankruptcy. Because all variables are ratios, the model withstands differences in currencies and is thus internationally applicable. The five indicators that predict airline bankruptcy according to Pilarski and Dinh are described below.

2.2.1 Operating revenues/total assets

This ratio represents the turnover, the productive capacity of the airline’s assets. In this industry, the turnover ratio is particularly important because an aircraft is a highly expensive piece of equipment in terms of maintenance, operational and depreciation costs. Moreover, aircraft seats are perishable and revenues are not maximized when aircraft fly with empty seats. F1 measures an airline’s ability

to maximize revenues on all its flights.

2.2.2 Retained earnings/total assets

Pilarski and Dinh considered both this variable and total equity / total assets, but because these ratios have the same informational content only one was used. Retained earnings reflect the cumulative earnings of a firm. In case of significantly negative earnings the retained losses might result in a negative total equity. This would imply that total liabilities would exceed total assets. Although this situation does not necessarily cause bankruptcy, it substantially increases its likelihood. Pilarski and Dinh decided to include retained earnings / total assets in their model

23 _F

because besides being an indicator of net worth, this variable measures earnings accrued over time more directly then total equity/ total assets.

2.2.3 Equity / debt

Leverage is a good indicator of the bankruptcy risk of a firm. When operating conditions are poor, this ratio can decrease by either eroding equity due to high losses, or by increasing debt as the firm seeks more financing. As the equity/ debt ratio becomes more negative, the likelihood of bankruptcy increases.

2.2.4 Current assets / current liabilities

Liquidity ratios are good identifiers of impending bankruptcy. When firms do not have the wherewithal to meet all debt obligations, they seek bankruptcy protection to obtain a reprieve from creditors and to reorganize their operations. Hence, the immediate cause for bankruptcy arises when a firm is unable to meet its current debt payment. This ratio represents a firm’s ability to meet this short term obligations by quickly converting liquid assets into cash.

2.2.5 EBIT / operating revenues

The last financial variable measures profitability and efficiency. It represents the part of the operating revenues that results in gross earning (i.e. after expenses -aside from interests, tax, accounting, and extraordinary expense- have been accounted for). By using income before interest, this ratio demonstrates the earning power of a firm’s core operations. Because of the leverage typically used, the EBIT/ operating revenue ratio should be high enough so that the firm can cover interest expenses.

2.3 Non-Financial Factors

Instead of focusing on financial data like most of the prediction models, Gudmundsson (2002) computed a prediction model that uses non-financial data and proxy variables for governmental influence and quality of the economic environment. Another large difference in this model is that rather than focusing on bankruptcy, Gudmundsson computed a model that focuses on distress preceding bankruptcy. Another difference in this research compared to other prediction models is his variable selection method. He specified a conceptual relationship a priori24 _{and eliminated}

unnecessary variables based on correlation analysis (Gudmundsson, 2002). He states that through this methodology his model is conceptually more robust than if a traditional approach was used, that is selecting variables based on prediction ability alone. However, most of the variables used in his research were statistically insignificant25_{. Therefore, his model is not entirely integrated into the}

Airline Rating Model, but his conceptual foundation is used and the variables he specified are individually tested for significance.

24 _{Unfortunately he only shortly elaborates on the relationship and characteristics of the variables.}

25 _{Only average fleet age and employees per aircraft was significant at the respectively 10% and 5% level. The}

Gudmundsson classified an airline as distressed when it had operating losses in 1997 and 1998 or in three (or more) out of five years in the period 1994 until 1998.

As shown in figure 2, Gudmundsson’s framework assumes that airline performance is related to input resources, political influence and economic factors. Input resources in airlines cluster around two main elements, labor and aircraft equipment. Poor management of equipment (fleet acquisition, utilization and composition) and low labor productivity26 _{is assumed to be related to}

poor airline performance.

Figure 2: Gudmundsson’s model to predict airline distress

Just as Pilarski ad Dinh, Gudmundsson also used a logistic regression analysis to find nine variables to significantly correlate with airline distress:

N1 = Average fleet age

N2 = Number of different aircraft brands in fleet

N3 = Departures per aircraft

N4 = Load factor

N5 = Passengers carried per departure

N6 = Hours flown per pilot

N7 = Employees per aircraft

N8 = Political influence

N9 = Economic factors

P, the probability of financial distress, is then calculated by

And e = 2,718 (the base for natural logarithms)

The overall prediction accuracy rate of this model was 90.3%. The nine indicators that predict financial distress according to Gudmundsson are described below.

2.3.1 Fleet age

Newer aircraft are generally more efficient to run than older aircraft due to the higher maintenance costs and fuel consumption. A high average fleet age should thus be a characteristic of poorer performing airlines. There can be several reasons for this. First, the financial situation of the airline does not allow the acquisition or leasing of new aircraft. Second, fleet acquisition and planning is poorly managed due to inexperience or political influence. The latter can play a role when political procedures supersede an airline’s operating interests in a market of substantial government influence, i.e. at government regulated airlines or in monopolistic market.

2.3.2Variety of aircraft brands operated

Aircraft purchases often take two to three years, which means that economic forecasting is very important for airlines to manage the introduction of a new aircraft (capacity increase) in harmony with industry cycles. If this forecasting, the fleet planning or the aircraft acquisition policy is not managed well, airlines might be forced to turn to alternative short-term solutions. This often leads to other aircraft brands that poorly fit the current fleet composition. This raises costs due to increased crew costs and maintenance burden. Thus, it is assumed that airlines operating excessive number of aircraft brands will be poor performers, partly due to poor forecasting management but mainly due to increased crew costs.

2.3.3 Departures per aircraft

The utilization of aircraft, given the large associated capital outlay and costs, is very important in airline management. Non-distressed carriers are expected to have higher number of departures per aircraft as a consequence of better overall management (in terms of schedules, marketing and distribution).27

2.3.4 Load factor

The cost efficiency of an airline is mainly dependent on the average load factor. The higher the percentage of seats sold, the lower the unit costs.

2.3.5 Passengers carried per departure

26 _{For most airlines remuneration of flying personnel is the key cost driver (typically 25-40% of total operating}

costs) (Williams, 2008, Sorestö 1997).

27 _{An expected intervening factor is average ‘stage length’. A good performing carrier operating mostly}

Economies of scale in the airline industry exist in terms of aircraft size; the larger the aircraft the lower the operating cost per seat (Holloway (2008, page 179), Seristö, (1997), Gritta (2006)). Hence, the greater the average number of passengers carried per flight, the better the airline’s operating performance28_.

2.3.6 Hours flown per pilot

Pilots are usually the most expensive labor resource. Their labor productivity, a high number of flight hours per pilot, is thus assumed to be related to better performing carriers.

2.3.7 Employees per aircraft

Another measure for labor productivity is the number of employees per aircraft; the fewer the number of employees per aircraft the higher the assumed productivity. Gudmundsson took notice of aircraft size as an intervening factor, but found no significant relation between the average fleet size and the number of employees per aircraft.

2.3.8 Governmental influence

In a bankruptcy prediction model one would expect proportionally high government ownership to work as a deterrent to bankruptcy that is to be linked to non-failure. Alitalia is a good example of an airline that would not have existed without its government’s back up. In this distress prediction model another assumption was made: the higher the proportional government ownership the less incentive there is for an airline to pursue competitive cost structures and other efficiency measures. Thus, government ownership of a majority of the airline’s equity is linked with poorer performance and higher likelihood of distress status.

2.3.9 Economic factors

To include the impact of the quality of the domestic economy the inflation percentage was selected as a proxy. Gudmundsson assumed that high inflation rates indicate poor unstable economic management having negative impact on airlines’ operating results.

Besides Gudmundsson’s variables, three extra operational factors of influence base on other topic related literature were defined. The three variables, aircraft size, aircraft ownership & lease contract flexibility, and fuel hedging strategy are described below.

2.3.10 Aircraft size

Although Gudmundsson (2002) and some other authors claim that operational costs decrease as size of an airplane increases (see variable O5), Wu and Caves (2000) find that a heavier fleet structure in

terms of aircraft size is related to higher operational costs.

In their article on aircraft operational costs and turnaround efficiency at airports authors Wu and Caves (2000) found aircraft operational costs to significantly differ among airlines, one of the reasons being the structure of the fleet in terms of airplane size. Airlines with a higher percentage of heavy aircraft (jumbo jets with an average seat capacity of 400) deal with higher operational costs than airlines with a higher percentage of large aircraft (wide-body jets with an average seat capacity of 250) or medium aircraft (narrow-body jets with an average seat capacity of 150). To cover the operational costs, airlines with heavier fleet structure need higher load factors and are thus more vulnerable to negative shocks in demand (Wu and Caves, 2000).

A third opinion on aircraft size versus operational costs is presented by Morell (2007). Figure 3 shows that aircraft size according to Morrel is non-linearly correlated with fuel efficiency -and thus operational costs (25% of operational costs is fuel-related (Williams, 2008). He found that wide-body jets, the large aircraft with an average seat capacity of 250, are the most fuel efficient. This model includes ‘aircraft size’ to research relationship between an airline’s fleet size and its performance.

Fuel Efficiency per aircraft type

0 5 10 15 20 25 0 10 20 30 40 50 60 70 80

Passengers and cargo payload (x1,000 kg)

T o n n e -k m a v a il a b le p e r U S g a ll o n

Figure 3: Fuel efficiency versus aircraft size by aircraft type, P. Morrell (2007)

2.3.11 Aircraft ownership and lease contract flexibility

Every airline leases at least a part of the aircraft in its fleet. Leasing offers an airline the opportunity to operate aircraft without the financial burden of buying them, and to provide a temporary increase in capacity. Another advantage of leasing could be the flexibility to trade up to more technologically advanced aircraft if the lease is structured that way. There are two main types of leasing agreements; wet leases and dry leases.

was not significantly correlated with the load factor and thus a separate measurement in Gudmundsson’s model.

A dry lease is a leasing arrangement whereby an aircraft financing entity29 _{(lessor) provides an}

aircraft without insurance, crew, ground staff, supporting equipment and maintenance to an airline (lessee). Dry lease is typically utilized by leasing companies and banks that profit from the tax reduction resulting from the amortization and depreciation write offs of the aircraft (this advantage is worthless to an airline making little or no money). A typical dry lease starts from two years onwards and bears certain conditions with respect to e.g. depreciation, maintenance and insurances, depending also on the geographical location, political circumstances, etc.

A wet lease is a leasing arrangement whereby one airline (lessor) provides an aircraft, complete crew, maintenance, and insurance to another airline (lessee). The lessee provides fuel and covers airport fees, other duties and taxes. Wet leases are paid for by the hour, and generally last one month to two years. It is typically utilized during peak traffic seasons or when the lessee’s own fleet is undergoing a heavy maintenance check. It could also serve to initiate new routes, or to serve routes into countries where the lessee, for whatever reason, is banned from operating.

The structure of fleet ownership and lease contracts can make a significant difference in an airline’s ability to adjust its capacity to changing market demands. Unlike carriers owning or dry leasing the majority of their aircraft, carriers with mainly wet lease contracts are very flexible and can easily cut off routes and frequencies when revenue decreases. Hence, this latter group forms a larger threat to Schiphol Airport.

2.3.12 Hedging strategy

Because jet fuel constitutes a large percentage30 _{of airline operating costs and oil prices are highly}

volatile, airlines face an incentive to hedge fuel price risk (Carter et al., 2006). A lot of research has been done in the field of fuel hedging and its impact on firm value and firm performance. In this section two of the most recent works are discussed and integrated in the Airline Rating Model. Morrell and Swan (2006) state that the main purpose of hedging is not to increase profit, but to reduce the volatility of profits. A policy of permanent hedging of fuel costs should leave expected long-run profits unchanged. This is underlined by the economic fundamentals of hedging such as the expected value of a fuel hedge of zero. When shocks in oil prices are caused by political and consumer uncertainty leading to slower economic growth, hedging will indeed decrease the volatility in profit. However, increased oil prices could also be a result of strong economic growth and oil supply constraints. In this case hedging gains will increase volatility of profits.

Other benefits of hedging are the accounting role in moving profits from one period to another and the insurance against bankruptcy. Perhaps the most compelling argument for airline hedging is the zero-cost-signal to investors that management is technically alert and competent.

Carter, Rogers and Simkins (2006) look at the pros and cons of hedging from another perspective and are convinced of its positive impact on firm value. They define the principal benefit of airlines’ fuel hedging as the reduction of underinvestment costs. Most of the 5%-10% hedging premium is attributable to the interaction of hedging with investment opportunities. When fuel prices rise

many small or financially unhealthy airlines might reach a level of near bankruptcy and are sometimes forced to sell parts of their assets below book-value. These are profitable investments for other airlines, and the ones that hedged their fuel price are not negatively affected by the increased oil price and thus in the position to exploit this investment opportunity. This is highly valuated by shareholders and significantly increases firm value.

Although both articles look at fuel hedging from another perspective, they both underline its -either direct or indirect- profitability.

2.4 AMS-Specific Variables

Not the least important part of this research focuses on indicators that are related to an airline’s operations at Amsterdam Airport Schiphol, the AMS-specific variables. In the end the goal is to create a model that can predict the risk and probability that an airline crosses out its route to, from or through Schiphol. A set of four AMS-specific factors are defined and included in the model; the load factor on routes to Amsterdam Airport Schiphol Group, the number of frequencies per route, the average punctuality of an airline, the market share per route.

2.4.1 Load factor

Low load factors are the main reason for airlines to cancel certain routes (Dennis, 2002). Many of the route cancellations that were published in newspaper articles from the end of 2008 onwards were a consequence of decreased load factors (VirginBlue, Mexican Airlines, WestJet, RyanAir, Thomsonfly, Flyblobespan, Brisith Airways, Delta Airlines, Northwest Airlines). Each flight has a minimum load factor under which operating the flight becomes unprofitable. If this level is reached the risk of an airline to cancel the route increases.

2.4.2 Frequencies per route

Figure 4: Changes of market share ratio with capacity share ratio based on increase of service frequency versus increase of aircraft size (Wei and Hansen, 2005).

Other authors (Clark, 2007; Dennis, 2002) confirm this finding but add to it that there is a point at which increasing the number of frequencies does not proportionally increase an airline’s market share.

Figure 5: S-curve of market share versus frequencies

This is underpinned by the S-curve of market share and frequency share (Clark (2007), Dennis (2002)). There is a non linear relation between the number of flights an airline operates and the market share it has on that route, see figure 5. Simply put, where two or more carriers compete, the airline that offers less than half of the frequencies can be expected to gather a proportionally smaller proportion of the market. Under and above a certain number of frequencies market share does not notably increase. Conversely, when frequency share predominates, then that airline can expect to gain a proportionally greater share of the market (Clark, 2007). This might imply that the

tendency to cancel routes increases when frequency share on one route is extra low, because the market share fitting that number of frequencies is disproportionally low.

It is important to note that the S-cure does not apply to all markets. For example long-haul routes and leisure markets are less like to respond to frequencies than short-haul routes and business markets (Clark, 2007).

2.4.3 Punctuality

People evaluate the quality of a product or service relative to a certain reference point and response asymmetrically to gains and losses. Consumers give heavier weights to the losses (negative deviations from the reference point) than to the equivalent-sized gains (positive deviations). This is known as loss aversion (Tversky and Kahneman, 1991). Suzuki is the first to apply loss aversion to the transportation literature (Suzuki and Tyworth, 1998) and relates on time performance of an airline it to market share (Suzuki 2000).

His findings were threefold. First, passengers are likely to be loss averse with respect to punctuality; the switching rate of passengers with delay experience is higher than that of passengers without delay experience. Second, when passengers switch airlines, their choices may be affected by the airline’s airport dominance, but not by other carrier specific characteristics. This implies that airlines that possess greater dominance at origin and/or destination airports may attract more switchers. Third, on-time performance affects an airline’s market share through the passengers’ experience, but not through the ‘advertisement' of performance (Suzuki, 2000).

2.4.4 Market share

A high market share on an airport (its general presence related to other airlines) is expected to be positively correlated to a stable flight and route schedule. Furthermore it is hypothesized that market leaders on routes (airlines that serve 70% or more of the passengers on a route) will be reluctant to cancelling that route.

Yield per passenger

Another factor that might influence an airline’s performance and one would thus expect to find in this chapter is the relationship of business versus leisure passengers i.e. ‘yield per passenger’. However, because of the paradoxical and economically dependent nature of this variable it is not incorporated in the model.

high-cost carriers seem to be more vulnerable than low-high-cost carriers in times of strongly fluctuating demand (Seristö, 1997). Due to this vulnerability towards the economic situation makes it hard to include ‘yield per passenger’ in this model.

All variables described in this theoretical framework result in the theoretical Airline Rating Model (see figure 6), and will be the bases for the data gathering which is explained in the next chapter.

General

Performance

Financial Score

AMS-Specific

Performance

External Score External Variables X1 Economic situation X2 Political stability X3 National health X4 Perception of safety X5 Member of an alliance Non-financial Variables N1 Fleet age

N2 Different brands in fleet

N3 Departures per aircraft

N4 Load factor

N5 Pax per departure

N6 Hours flown per pilot

N7 Employees per aircraft

N8 Governmental influence

N9 Inflation

N10 Aircraft size

N11 Aircraft ownership and lease flexibility

N12 Hedging Strategy

Financial Variables

F1 operating rev/ total assets

F2 retained earnings/total assets

F3 equity/total debt obligations

F4 liquid assets/current debt obligations

F5 ebit/operating revenues

AMS-Specific Variables

AMS1 Load factor

AMS2 Frequencies

AMS3 Punctuality

AMS4 Market share

3. Research Methodology

In this section an overview of the methodological issues is presented. First the sample is discussed, subsequently the process of data collecting is outlined and the methodology is elaborated on and afterwards.

3.1 Sample

In 2008, 161 different airlines offered services to 47.4 million passengers to and from Amsterdam Airport Schiphol. 60 of those airlines carried 97.9% of all passengers, of which KLM was the largest, flying 51% of total passengers31_.

The sample for this research contains 45carriers. First, all charter airlines (Onur Airlines, Corendon, Inter Airlines and Sun Express) have been removed from the sample. Those carriers operate on a charter basis (flights that take place outside normal schedules) and they are thus irrelevant for this research. Secondly, all direct KLM-related airlines (KLM, Air France, Transavia, Martinair and VLM Airlines) were excluded from the sample. These airlines are inextricably bound up with the airport, and the relationship among the two interdependent parties is very close. This relation is maintained by means of consultative bodies32 _{that discuss market developments, strategic plans, network}

adjustments, potential new markets, etc. These meetings facilitate the exchange of important knowledge from KLM and its subsidiaries to Amsterdam Airport Schiphol and vice versa. Because of the intense relationship between the two parties, it is unnecessary to include them in this airline rating model.

The top 45 remaining airlines represent 32.8% of all passengers at Schiphol in 2008.

Appendix 1 gives an overview of the top 104 carriers and the sample for this research, arranged in order of passengers carried from to and through Schiphol Airport.

3.2 Data

Because the ‘Airline Rating Model’ needs an annual update it is important that input for the model is easily accessible and data collecting does not require an individual to invest a significant amount of time. For that reason several data sources have been tapped and reviewed for each individual variable. All the data sets are discussed in this paragraph and summarized in Appendices III to VI.

3.2.1 External Factors

Inflation and GDP figures per country were derived from globalinsight.com, ‘the global leader in economic and financial analysis, forecasting and market intelligence’. It provides comprehensive

31 _{Source: Schiphol’s Airline Value Analysis Model}

32 _{i.e. Werkgroep Markt, Werkgroep Air, Werkgroep Ground, Werkgroep Operationeel Plan, Operationeel}

information on countries, regions and industries and was useful as a quick access to national economic figures such as the GDP growth rate and the inflation percentage33 _{of 2008. The GDP}

growth-rate was used as input in Schiphol’s GDP-model that links GDP growth to passenger growth at Schiphol Airport. Inflation figures and GDP impact on Schiphol’s passenger traffic is presented in appendix III.1.

Figures reflecting a nation’s political situation were retrieved from the Worldbank as is shortly discussed in paragraph 2.1-X2. For the 35 countries in this sample the combined view of an average

of 11 experts, survey institutes, think tanks, non-governmental organizations, and international organizations is represented. The average standard error of the ‘Political Stability and Absence of Violence Indicator’ was 0.22. The second dataset, that separates the political situation and the threat of terrorism, is derived from AON. Every country is categorized in one of the respectively six and five risk categories AON has distinguished. World maps are shown in appendices III.2a to III.2c. Due to the development of the Mexican Flu, the third variable, national health, had to be excluded from this research. The H1N1 virus has become incomparable to any previous disease. The impact on Mexican Airports was huge and at first seemed similar to the effects of SARS on Hong Kong International Airport34_{. However, at July 6 2009, the Mexican flu had reached 135 countries and}

caused 429 deaths, accounting for 0.45% of all infected35_{. It is, of course, very hard to say how this}

disease will expand, but the WHO stated that further spread of the pandemic, within affected countries and to new countries, is inevitable. This assumption is fully backed by experience. The Mexican Flu pandemic has spread internationally with unprecedented speed. In past pandemics, influenza viruses have needed more than six months to spread as widely as the new H1N1 virus has spread in less than six weeks. As this pandemic evolves, the data needed for risk assessment, both within affected countries and at the global level, are also changing. Priorities are shifting and the WHO decided to change its reporting methods of infections and fatalities. The report in appendix III.3 is the last published global table showing the number of confirmed cases for all countries. All by all, the conclusion had to be drawn that ‘national health’ could not be included in this model. However, ignoring national health as a factor impacting the willingness of people to travel would harm the reliability of the airline rating report, and it is thus mentioned as a separate ‘’incident factor (see chapter 5)

During this research process the world faced as much as three severe accidents within six weeks, killing a total of 439 people36_{. As stated in paragraph 2.1-X}

4, the perception of an airline’s safety can

be largely influenced by fatal accidents in the short term past. This results in a decrease in passengers and marketing expenditures, influencing an airline’s profitability. Airfleets.net presents

33 _{Consumer Price Index, Year-on-Year percentage change}

34_{http://www.moodiereport.com/document.php?c_id=1178&doc_id=20652,}

http://www.tradingmarkets.com/.site/news/Stock%20News/2299394/,http://www.nowpublic.com/health/swine-flu-cancels-flights-mexico-list-travel-companies

35 _{These figures show a death rate of more than 20 times as low compared to SARS, implying a lower severity and}

less danger. However, the number of infections is almost 12 times higher in four months than SARS was in nine months.

36 _{On June 1, an Airbus A330 from Air France crashed into the Atlantic Ocean on its way from Rio de Janeiro to}

an easy-reference database on airline accidents per airline, which can be double checked by airdisaster.com in chronological order. Appendix III.4 gives an overview of fatal accidents since 2006 of the sample airlines. Because there are too few data points meeting the criteria it was decided to exclude ‘the perception of an airline’s safety’ from this research. It will, however be included in the ‘overall incident factor’ (see chapter 5).

The list of members of each of the three alliances (SkyTeam, oneworld and Star Alliance, see appendix III.5) was derived from the Airline Industry Guide (2008/2009). The alliance dummy is formed by assigning a 1 to ‘Full’, ‘Regional’ and ‘Associate’ member airlines, and a 0 to independent airlines. It should be taken into account that Continental Airlines is currently a SkyTeam member but will switch to Star Alliance as of October 24th_{, 2009.}

3.2.2 Financial Factors

For the financial figures the annual reports of the individual airlines of the financial year of 2008 have been used. To complete the Pilarski Score Formula the operating revenue, the liquid and total assets, the liquid and total liabilities, the retained earnings/reserves, the shareholders’ equity and the earnings before interest and tax (EBIT) were retrieved from the consolidated balance sheet, income statement and in some cases from the statement of changes in shareholders’ equity. Unfortunately it was not possible to retrieve financial data of every airline in the sample. Several wholly owned subsidiaries (e.g. Arke Fly, Swiss Airlines, BMI (-Baby), Jet2, Atlas Blue), government owned airlines (e.g. Surinam Airways, Royal Air Maroc, Olympic Airlines, Alitalia and Ukraine International Airlines) do not publish any annual report. Others (e.g. Vueling, TAP Portugal, Malev, LOT Polish Airlines, EL AL, Flybe) did not have their ’08 report published online in the data gathering phase of this research (July 2009)37_{. For nine of these airlines only the EBIT margin (F}

5)

could be calculated by using the latest release of the World Airline Report38_{. The remaining 29}

airlines form the sample for the financial part of this research. All financial figures and ratio’s can be found in appendix IV.

3.2.3 Non-Financial Factors

Traffic figures were also primarily retrieved from the annual reports. Although some authorities and organizations such as ICAO39_{, IATA}40 _{and AEA}41 _{and Airline Business offer interesting databases,}

after reviewing trial versions the conclusion was drawn that most of the databases were incomplete and not meeting the criteria for this research. The number of passengers and the load factor of each carrier are outlined in every annual report. Because some of the airlines in the sample do not publish annual reports, the World Airline Report (produced in July 2009 by Air Transport World) was

37 _{Efforts were made to retrieve the financial information by e-mail or telephone but this turned out to be}

unrewarding.

38 _{In July 2009, Air Transport World published ‘The World Airline report’ containing this information for every}

airline in American dollars, except for Arke Fly, Alitalia, TAP Portugal, Malev, Surinam Airways, Royal Air Maroc and Olympic Airlines.