"The Financial Implications of Aviation Disasters"

(1)

Master Thesis

Opleiding: Economie Profiel: Economie & Management Afstudeerrichting: Financiering en Belegging Variant: Waardering & Financieel Management

31 August 2006

"The Financial Implications

of Aviation Disasters"

(2)

Outline:

Outline page 1

1. Introduction page 2

2. Literature review and past empirical research page 4

3. Hypotheses page 8

4. Data and Descriptive statistics page 10

5. Methodology page 14

6. Results page 23

7. Conclusions page 31

8. Recommendations for further research page 32

References page 34

(3)

1. Introduction:

The modern passenger aircraft has made the world smaller by increasing the speed of long distance travelling. For example, from Schiphol Airport one could travel to Australia in just one day or to Los Angeles in just over 11 hours. It also has made long distance travel far more secure than any other form of transport, in terms of accidents for distance covered. As the study of Hinson (1997) points out, the death risk of any passenger of a U.S. commercial airline is reduced to one in eight million over the 1990's. This means that if a passenger chooses one flight at random each day, he or she could, on average, fly around the world for 21,000 years before perishing in a fatal crash. Therefore it makes sense that the number of passengers travelling by airplane is growing rapidly each year. But if the safety is so much better, why do most passengers have at least a small fear of flying? The airlines, aircraft manufacturers and air-safety authorities blame this "irrational" fear of flying on misinformation. Because air crashes are both big and infrequent, they always make the news headlines, while the more frequent but smaller car damages on the roads are largely unreported. Moreover, the airlines state that the passengers know that they are being irrational: if they were really frightened of crashes, they would completely give up flying1_{. As mentioned above, the contrary is happening. What matters even}

more for consumers is not whether air travel is safe but which airlines are safest. Passengers react to the little information that is available. A real life example is that of the company ValuJet. After the crash of a ValuJet airplane in 1996, the passengers stayed away from that airline. Their aircrafts were barely half full against an industry average of 70%2_.

After mentioning these consumer reactions and emotions, it is interesting to see how investors react to aviation disasters. Investors are expected to behave negatively to the event and the stock price is expected to decline. Walker, Thiengtham and Lin (2005) state that although airlines are mostly well insured against claims filed by the surviving relatives, this makes sense because of the legal costs, diversion of management time, rising insurance premiums, loss of consumer confidence and repair and/or replacement costs for the crashed airplane. These are items that will have financial consequences for these companies. Presumably, these financial implications will have an impact on the share price because, as financial theory suggests, the price of a share is the present value of the firm's expected dividends discounted at the risk adjusted rate of return. Following Scott (2003), rationality for investors is assumed, that is, they make decisions so as to

1_{The Economist, (1997a)}

(4)

maximize their expected utility, or satisfaction, from wealth. Because of the calamity, either the risk adjusted discount rate must increase and/or the expected dividends must decrease.

This thesis will examine the financial impact of aviation disasters on a number of variables. First it will examine the effects of the events on the share value of the afflicted airlines and airplane manufacturers (from now on to be referred to as manufacturers). This will be done by an event study in which the performances of the shares of the companies involved are measured on the first day, week and month after the disaster. This will be compared with the market performance. Then there will be examined if the different causes (divided into four categories; human error, mechanical failure, nature and criminal activity) and the number of fatalities (divided into 3 categories; 1-10 people killed, 11-99 people killed and 100 or more people killed) have different influences on the value. Thereafter, the combined impact of both variables will be measured.

The combined problem statement of this thesis will be:

What are the financial implications of aviation disasters?

Within this combined problem statement, the following sub questions will be addressed:

1. What are the effects of an aviation disaster on the financial performance of the afflicted airline?

2. What are the effects of an aviation disaster on the financial performance of the afflicted manufacturer?

3. What is the influence of the different causes on the financial performance of that airline? 4. What is the influence of the different magnitudes on the financial performance of that

airline?

5. What is the influence of the different causes combined with the different magnitudes on the financial performance of that airline?

This thesis also tests to what extend the market is efficient. Fama (1970) states that when the market is efficient and the investors are rational, the market fully and properly reflects all relevant and available information.

(5)

conclusions will be drawn and finally recommendations for further research will be given in section 8.

2. Literature review and past empirical research:

According to FAA's Safety Report of January 20, 1997, aircraft accidents and incidents are events that endanger the safety of aircraft operations and/or the safety of persons involved in those operations. Accidents result in death or serious injury to a person in the aircraft, or in substantial damage to the aircraft itself. Incidents are less serious events "that affect or could affect the safety of operations." In this thesis only accidents will be observed with at least one fatality, this to be sure that the impact is large enough, and the aircraft operations are assumed to start when the aircraft is boarded. In this section, the outcomes of earlier research, where different results were found, will be summed up.

According to Brown (1998), the effects of an aviation disaster will grow due to the fact that the liability limits for passenger deaths and injuries are removed later in 1998. The Warsaw Convention of 1929 settled a liability ceiling of $16,600 per passenger on air carriers. The new convention, expected to be adopted by the International Civil Aviation Organization (ICAO), will accept payouts up to $150,000 requiring only proof of damage. This logically raises the insurance premium charged to the airlines which increases the cost of a disaster. Borenstein and Zimmerman (1988) add that although airlines can insure themselves against most direct costs of an accident, they cannot insure themselves against demand loss.

Chalk performed two studies. In 1986 he examined the effect of a single crash on 25-5-1979 and found that Mc Donnell-Douglas's stockholders suffered a loss of about $200 million. This corresponds with an abnormal return of -0.55% which, according to Chalk, exceeds any reasonable estimate of regulatory or liability cost.

(6)

Walker, Thiengtham and Lin (2005) performed a study with a sample of 138 aviation disasters occurring between July 1962 and December 2003. Only publicly traded U.S. carriers were involved. They observed a stock price drop of 2.8% for the afflicted airline and a drop of 0.8% for the manufacturers within one trading day. For the following week the negative abnormal results were 1.08% and 1% respectively. Additionally they observed that disasters in the U.S. caused by criminal activity (in particular the 9/11 terrorist attacks) cause significantly larger stock price drops in the days following the event. All their results remained significant when the 9/11 crashes were excluded.

Another study about the influence on the share value of the airline and the influence on the share value of the manufacturers was done by Chance and Ferris (1987). They performed a study between 1962 and 1985 on 49 accidents, involving American domestic airlines. Out of these 49, 46 were useful on the abnormal share price returns of airlines and airplane manufacturers. The other 3 had incomplete return series. An additional criterion was that there had to be at least ten fatalities per disaster. They measured an immediate significant negative response of the stock market to the news of the disaster for the afflicted airlines. For the manufacturers they did not find a significant result. They also examined whether the prices of the stocks of those carriers not involved in the accidents reacted. This influence could occur because investors might believe that tighter regulation and higher insurance ratios could be imposed to the industry as a whole. If there is a negative reaction in the performance of the shares of these carriers, investors see aviation disasters as an industry wide problem and not as an isolated event concerning only the carrier involved. If there is a positive reaction, which seems less obvious, investors believe that airlines will benefit by an accident of another airline. This would indicate that investors believe that passenger's confidence in the other companies increases and that their number of bookings increases. They did not find a significant reaction.

Mitchell and Maloney (1989) examined the brand name effect or the potential loss of consumer goodwill of aviation disasters. They used data of 56 crashes in the period of 1965 up to and including 1987 and found that airline crashes cause customers to reduce their demand for the services provided by negligent carriers. Then there is a significant negative stock market reaction to the event.

A number of studies have been performed about the influence of aviation disasters on the share value of airlines alone. On the one hand, some of them found significant negative returns which will be displayed here.

(7)

only the most severe catastrophes had a significant negative effect of -2.42% and -2.89% on the share values of the afflicted airlines.

Bosch, Eckard and Singal (1998) found a significant cumulative negative abnormal return of 2.10% for day 1 and 2.67% for day 2 in their study that observed the stock reactions of 25 airlines from the 4th_{quarter of 1978 until the 4}th_{quarter of 1996.}

Barret, Heuson & Colb (1987) found a significant negative return (-0.62%) on the share values of the airlines as well. Their sample consisted of 78 crashes between January 1962 and December 1985.

On the other hand, Borenstein and Zimmerman (1988) found that demand shows no or only little effect as a result of aviation disasters. Therefore the loss in the airline's equity value following an incident is quite small relative to the social cost. Their research examined 11 airline accidents between 1960 and 1985.

(8)

Year of

publication Author Sample period Number of events manufacturers Airlines/ Abnormal Return Signific-ance

1986 Chalk 25-5-1979 1 Manufacturer -0.55% **

1987 Chalk 1966-1981 76 Manufacturers -$21.3 mln

1987 Chandy & Davidson

Cross 1965-1984 57 Airlines -2.42% / -2.89% ** worst accidents 1987 Heuson Barret, & Colb 1-1-1962 – 31-12-1985 78 Airlines -0.62% **

1987 Chance & _Ferris 1962-1985 49 (46 usable) _{manufacturers}Airlines -1.18% _-0.19% *** _No

1988 Borenstein &

Zimmerman 1960-1985 11 Airlines

Little or no

effect No

1989 Mitchell & _Maloney 1965-1987 56 Airlines -2.31% ***

1998 Eckard Bosch, & Singal 1-10-1978 – 30-9-1996 25 Airlines -2.10% day 1 -2.67% day 2 *** 2005 Thiengtham Walker, & Lin 1962-2003 138 107 manufacturers Airlines -2.8% -0.8% *** *** Table 1: Summary literature review and past empirical research

* = significant at the 10% level, ** = significant at the 5% level and *** = significant at the 1% level. The return column demonstrates what the level of abnormal return is after the disaster. Most abnormal returns are measured in percentages, some in absolute values.

Number of events is the number of crashes involved in the various researches.

(9)

3. Hypotheses:

Based on the outcomes of the performed literature review, the following is expected:

• A significant negative influence in the stock performance of the airlines and airplane manufacturers after an aviation disaster.

Based on the Efficient Market Theory, the share price is expected to decline on the first trading day after the announcement of the disaster. After this decline, the market will find a new equilibrium in which the share price represents the present value of the adjusted future cash flows.

• The various causes of the accidents are not expected to significantly influence the share price in a different way

The various causes are not expected to influence the share value differently, because investors are expected not to distinguish between the various causes. Therefore their reactions are expected to be the same for all causes.

• Accidents of larger maginitude will provoke a significantly larger share price reaction For larger plane accidents, the negative return is expected to be larger as well. So there is expected to be a significant negative relationship between the abnormal return on the share prices and the magnitude. Again, the various declines in the share values are expected on the first trading day after the event. The market is expected to find a new equilibrium in which the share price represents the present value of the adjusted future cash flows.

• The share prices will show a significantly divergent reaction, if a disaster has, both a different cause and different magnitude.

(10)

The tests carried out in this thesis are based on the sub questions raised in the introduction. Based on the expectations mentioned earlier in this section, the following hypotheses can be stated:

For the first two sub questions the following hypotheses are applicable:

H0: After an aviation disaster, there is no significant abnormal return on the share prices of the

afflicted airlines / manufacturers.

Ha: After an aviation disaster, there is a significant negative abnormal return on the share prices

of the afflicted airlines / manufacturers.

For the third sub question the following hypotheses are applicable:

H0: After an aviation disaster, the abnormal returns of the share prices of the afflicted airlines do

not differ significantly as a consequence of different causes.

Ha: After an aviation disaster, the abnormal returns of the share prices of the afflicted airlines

differ significantly as a consequence of different causes.

For the fourth sub question the following hypotheses are applicable:

not differ significantly as a result of different magnitudes.

Ha: After an aviation disaster, the negative abnormal returns of the share prices of the afflicted

airlines are significantly larger when the magnitude is larger.

For the fifth sub question the following hypotheses are applicable:

not differ significantly if the combination of causes and magnitudes of the disaster are different. Ha: After an aviation disaster, the negative abnormal returns of the share prices of the afflicted

(11)

4. Data and descriptive statistics:

This thesis will test the effects of the aviation disasters in a number of countries. The data about the aviation disasters are collected at the website www.aviation-safety.net. On this website of the Flight Safety Foundation a detailed description of almost all aviation disasters can be found. The disasters are stored in several categories such as airline, geographical region / country, cause and date. Because of these categories, all disasters can easily be found and interpreted. Following Walker et al. (2005), additional online databases including www.airdisaster.com and

(12)

the passengers and the crew are taken into account. The airlines that have changed their names within the sample period will be treated as separate companies. Finally, in the two cases that bird ingestion was the cause of the disaster, it is, although not weather-related, counted among the natural causes.

As the impact of 9/11 is unarguable very large, it will influence the results of the research if not handled properly. To overcome this problem, the disasters in the period from August 2001 until January 2003 will be excluded. As there are only 7 events in this period, it is not useful to perform an extra study for these events, especially because 4 out of these 7 have occurred on 9/11. This will influence the study for this period even more. For the crashes that happened after January 1 2003 it is possible to measure an acceptable normal return again. The subdivision of the total sample period is illustrated in Figure 1:

"Normal period" "9/11 period" "Normal period"

1/1/1925 11/8/2001 1/1/2003 31/12/2005 Figure 1, Subdivision for the total sample period

(13)

Airline _AccidentsNr. of _fatalitiesNr. of Airline _AccidentsNr. of _fatalitiesNr. of Airline _AccidentsNr. of _fatalitiesNr. of

Air Florida 1 74 Crossair 1 10 _{Centr. Airl.}Pennsylv. 3 73

Air France 6 124 Delta Air _Lines 5 242 Republic _Airlines 1 1

Air

Midwest 1 21 Eastern Airlines 23 755 Simmons Airlines 1 3

Air

Wiscons. 1 13 Frontier Airlines 2 3 Singapore Airlines 2 87

Alaska

Airlines 3 200 Japan Air Lines 3 578 Skywest Airlines 2 20

Alitalia 1 46 _AirwaysKenya 1 169 Southern _Airways 1 63

All Nippon

Airways 2 2 KLM 1 248 Swissair 1 14

Allegheny 5 143 _{Air Lines}Korean 3 418 International Texas

Airlines 1 11

American

Airlines 22 949 Lauda Air 1 223

Thai Airways

Internat. 2 102

American

Eagle 2 83 Lufthansa 4 66 Caribbean Trans 1 2

Atlantic

Coast 1 5 Malaysia Airlines 1 34 TWA 22 991

Atlantic

Southeast 2 31 Mohawk 5 72 Airlines United 26 956

Braniff 6 204 National _Airlines 8 164 US Air 5 220

Capital

Airlines 6 147 Northeast Airlines 1 32 Valujet 1 110

China

Airlines 3 463 Northwest Airlines 16 622 Western Air Lines 6 113 China

Eastern

Airlines 1 1

Ozark

Airlines 1 38 Wien Air Alaska 1 10

Comair 1 29 Southw. Pac.

Airl. 3 181

Total

Airlines 253 10804

Continent

al Airlines 4 56 Pan Am. World 30 1582

(14)

Manufacturer _AccidentsNr. of _fatalitiesNr. of Boeing 72 4818 Douglas 64 1775 Fokker 3 64 Lockheed 23 1015 Mc Donnel Douglas 24 1373 Total Man. 186 9045

Table 3: Nr. Of Accidents per manufacturer.

Causes _{Accidents Magnitude}Nr. of _AccidentsNr. of _MagnitudeCause & _AccidentsNr. of

1 117 a 82 1a 37 2 58 b 145 2a 16 3 30 c 26 3a 4 4 34 4a 25 ? 14 1b 69 2b 33 3b 23 4b 7 1c 11 2c 9 3c 3 4c 2 ? 14

Total 253 Total 253 Total 253

Table 4: Proportion of different causes, magnitudes and C&M.

(15)

mechanical failure is the cause and there are 1-10 fatalities, 3a is when nature is the cause and there are 1-10 fatalities, 4a is when criminal activity is the cause and there are 1-10 fatalities. 1b is when human error is the cause and there are 11-99 fatalities, 2b is when mechanical failure is the cause and there are 11-99 fatalities, 3b is when nature is the cause and there are 11-99 fatalities, 4b is when criminal activity is the cause and there are 11-99 fatalities. Finally, 1c is when human error is the cause and there are 100 or more fatalities, 2c is when mechanical failure is the cause and there are 100 or more fatalities, 3c is when nature is the cause and there are 100 or more fatalities, 4c is when criminal activity is the cause and there are 100 or more fatalities and a question mark is noted when the cause is unknown again. The number of fatalities is not an issue here.

Table 4 will be used for sub questions 3, 4 and 5. To exclude less reliable outcomes, only the causes combined with magnitudes that have a bigger sample than 10 events will be included in this thesis. This implies that scenario’s 3a, 4b, 2c, 3c and 4c will be eliminated from this test.

In appendix 1, a complete overview of the whole sample and its background is shown.

5. Methodology:

The measurement of the effects of an economic event on the value of firms seems like a difficult task on the surface. But, according to MacKinlay (1997), this can be constructed easily by using an event study. An event study measures the impact of a specific event on the value of a firm by using financial market data.

(16)

5.1 Measuring the effects on the airline and airline manufacturer:

The first test of this thesis is to determine the event effects on the share prices of the airlines and the manufacturers. To measure this, the following tests will be performed. First of all it is important to measure the daily return of a share. Through equation (1) this can be measured:

1 , 1 , , , ) ( − − − = t i t i t i t i P P P R (1)

Ri,t is the return of share i on day t, Pi,tis the price at time t and Pi,t-1 is the price at time t-1.

According to Mac Kinlay (1997), the abnormal return, which is the ex post return of a security over the event window minus the expected return over that event window, can be measured by equation (2):

ARit = Rit – E(Rit|Xt) (2)

ARit is the abnormal return of share i at time t, Rit is the actual return of security i at time t and

E(Rit|Xt) is the expected return on security i at time t given the relevant information available at t.

The expected return is defined as the normal return on that security without the event taking place.

Several economic models can be used to calculate the expected return. Two of these models are the Capital Asset Pricing Model (CAPM) and the Arbitrage Pricing Theory (APT). According to MacKinlay (1997), it was discovered for the CAPM that the results of the studies might be sensitive to the specific CAPM restrictions. For the APT, it was found that the most important factor behaves like a market factor and that additional factors add little explanatory power. Therefore the gains from using an APT motivated model versus the CAPM motivated model are small.

(17)

the empirical fact that the marginal explanatory power of additional factors is small so this results in only little reduction in the variance of the abnormal return.

Therefore there are two common choices for modelling the normal return in a statistical way, the constant mean return model and the market model. In the constant mean return model E(Rit|Xt) is

constant. It assumes that the mean return of a given security is constant over the event window. The market model, which is based on CAPM, assumes that the relation between the market return and the return on a security is stable and linear. To define the estimation window in a normal performance model, it is common to use the period prior to the event. With this parameters the abnormal return can be calculated.

First the two earlier mentioned models for measuring the normal performance will be explained.

5.1.1 The Constant Mean Return Model

As mentioned earlier, the constant mean return model assumes that the expected return on a security is constant over time. Following to Mac Kinlay (1997), the model is defined as done in equation (3): t i i t i R, =

µ

+

ζ

, (3) 0 ) ( i,t = E

ζ

2 , ) var(

ζ

it =

σ

ζ_i

Where Rit is the period-t return on security i, i is the mean return for asset i over the period t = -n

until t = -1 and it is the time period t disturbance term for security i with an expectation of zero

and variance

σ

_ζ2_i.

The mean return on a share can be measured by equation (4):

i = − = − = 1 , t n t t i n R (4)

Where i is the mean (normal / realized average) return of share i over n trading days. This is the

sum of all the Ri,t's divided by the number of trading days n. This realized average return can be

taken as a proxy for the expected return, so i = E(Rit).

(18)

equation (2) is applied in the Constant Mean Return Model it results in the following formula in equation (5):

ARit = Rit – i (5)

ARit is the abnormal return of share i at time t, Rit is the actual return of security i at time t and i

is the expected return on security i at time t.

5.1.2 The market model:

The Market model relates the return of any given security to the return of the market portfolio. The market can be defined in a number of different ways. It can be the 'normal' share market of a country (for instance the AEX Index or NYSE Index) and it can be the PEER group of the airline. More common used indices are the CRSP value weighted Index, the CRSP equal weighted Index, the S&P 500 Index and the Dow Jones Industrial Average Index. Note that airlines and airplane manufacturers are assumed to have a higher level of systematic risk (and therefore a higher beta) compared to other types of companies. Since this is not a problem for this thesis, the Dow Jones Industrial Average Index will be used. According to Mac Kinlay (1997), the model is defined in equation (6): Ri,t = i + iRmt + i,t (6) E( i,t = 0) var( _, ) 2 t t i

σ

ε

=

Rit and Rmt are the returns on security i and the market portfolio over the period t = -n until t = -1,

respectively, and i,t is the zero mean disturbance term. i, i and

σ

_ε2_tare the regression

parameters of the market model. In the market model (

α

ˆ_i +

β

ˆ_iR_m,_t) can be taken as a proxy for

the exprected return, so: (

α

ˆ_i +

β

ˆ_iR_m,_t) = E(Rit).

(19)

ARit = Rit – (

α

ˆi+

β

ˆiRm,t) (7)

ARit is the abnormal return of share i at time t, Rit is the actual return of security i at time t and

(

α

ˆ_i +

β

ˆ_iR_m,_t) is the expected return on security i at time t.

According to Mac Kinlay (1997) the market model has a potential improvement over the constant mean return model. By removing the portion of the return that is related to variation in the return of the market, the variance of the abnormal return is reduced. Therefore the market model will be used in the remainder of this thesis.

After Measuring the abnormal returns in both models, the impact of the disaster can be tested by measuring if the AR is significantly different from zero.

5.1.3 Measuring and analyzing abnormal returns:

After explaining the measurement of the abnormal returns in the previous section, this section accounts for the manner in which the abnormal returns in this thesis are analyzed. Returns will be registered in event time using t. From here on t = 0 is the event date, t = 1/5/20 are respectively one day, one week and one month after the event (in trading days) and t = -n to t = -1 is the estimation window. Important in an event study is that the event itself is not included in the estimation period (therefore the estimation period is from t = -120 until t = -1), because the event can influence the normal performance. For this thesis 120 trading days was taken for the estimation window because it corresponds with approximately 6 months. The measurement of the abnormal returns is done at t = 1, t = 5 and t = 20. These points of time are chosen because they represent one day, one week (in trading days) and approximately one month (in trading days) after the event has taken place. Furthermore, it is typical for an estimation window, an event date and a post event window not to overlap.

(20)

Estimation event post-event Window date window

t = -120 t=-1 t=0 t=1 t=5 t=20

Figure 2, Time line for an event study

After measuring the abnormal return, which is the outcome of equation (7), the abnormal returns should be aggregated. Under the null hypothesis, conditional on the event window mean returns, the abnormal returns will be jointly normal distributed with a zero conditional mean and conditional variance

σ

2(ARi,t)where

− + + = ₂ 2 , 1 2 , 2 ˆ ) ˆ ( 1 1 ) ( m m t m t i R L AR _t

σ

µ

σ

ε (8)

L1 is the length of the estimation window in trading days.

µ

ˆ is the mean return of the market m measured throughout the estimation window and _ˆ2

m

σ

is the squared variance of the market's return. So under the null hypothesis the distribution of the sample abnormal returns of a given observation in the event window is:

ARi,t ~ N(0, 2(ARi,t)). (9)

Walker, Thiengtham and Lin (2005) state that, under the assumption that the returns on each day are independent, the variances are additive. Due to the fact that adding independent normal variables requires adding the variances, the proper variance is the cumulative variance. The following equations describe a firm's cumulative abnormal return (CAR) and the variance of a firm's CAR (var(CAR):

= = = t j t it t i AR CAR 1 , , (10) = = =t j t it t i AR CAR 1 , , ) var( ) var( (11)

j is 1 or 5 or 20 trading days. From these equations the average CAR (ACAR) and its variance

(21)

= = N i it t CAR N ACAR 1 , 1 ₍₁₂₎ = = N i it t _N CAR ACAR 1 , 2 var( ) 1 ) var( (13)

N is the number of firms in both equations.

When using a t-test, the hypothesis that the mean CAR is different from zero on any given day can be tested. The t-test will have the following form:

1) -N df , ( t ~ ) var( = = α t t ACAR ACAR t (14)

Where is the level of significance and df are the degrees of freedom (here: 252 for sub question 1 and 185 for sub question 2). Since it is assumed that the return is negative, a one-tailed t-test is applied. The null hypothesis is, see section 3, that the mean abnormal return during the event window is equal to zero and the alternative hypothesis is that the mean abnormal return is lower than zero. The critical values for the t-test applied here are that:

ACAR is significant at a 10% level if the t-value -1.286 ACAR is significant at a 5% level if the t-value -1.653 ACAR is significant at a 1% level if the t-value -2.347

5.2 Measuring the effects of the different causes and magnitudes on the abnormal returns of the airlines:

For testing if there are significant cause-effects, the ACAR has to be measured for each probable cause. (human error, mechanical failure, nature or criminal activity). Then it is possible to measure the parameters of the different causes. Equation (15) is used for measuring the ACARs of the different causes.

ACARct = = c N i cit c CAR N 1 , , 1 ₍₁₅₎

ACARct is the ACAR at time t=1/5/20 for each cause examined separately, Nc is the number of

(22)

causes (as done in table 4, c,1,t is the ACAR when human error is the cause, c,2,tis the ACAR when mechanical failure is the cause, c,3,t is the ACAR when nature is the cause and c,4,t is the

ACAR when criminal activity is the cause). The null hypothesis of this test is that c1t = c2t = c,3,t = c,4,t, the alternative hypothesis is that c,i,t c,j,t for at least one pair c,i,t, c,j,t. The probabilities will display the significance of the influence of the different causes.

For testing the magnitude of the disaster, the same test will be performed. There are three different groups, defined: (1-10 people killed, 11-99 people killed and 100 or more people killed). Therefore, for the same reasons as mentioned above, equation (16) is applicable:

ACARmt = = m N i t i m m CAR N 1 , , 1 ₍₁₆₎

ACARmt is the ACAR at time t=1/5/20 for each cause examined separately, Nm is the number of

firms in each category of magnitudes and CARm,i,tis the CAR on the share price due to the

different magnitudes (as done in table 4, m,1,tis the ACAR when 1-10 people were killed, m,2,t is the ACAR when 11-99 people were killed and m,3,t is the ACAR when 100 or more people were killed). The null hypothesis of this test is that m,1,t = m,2,t = m,3,t, the alternative hypothesis is that m,i,t m,j,t for at least one pair m,i,t, m,j,t. The probabilities will display the significance of the influence of the different magnitudes.

After testing the different impacts of the different causes and magnitudes these two models will be combined into one, see equation (17). When combining the different magnitudes with the different causes one can see if there is additional explanatory power.

ACARcmt = = cm N i t i cm cm CAR N 1 , , 1 ₍₁₇₎

ACARcmt is the ACAR at time t=1/5/20 for each cause examined separately, Ncm is the number of

firms in each category of causes combined with magnitudes and CARcm,i,tis the CAR on the share

(23)

and 100 or more people were killed). The null hypothesis of this test is that cm,1a,t = cm,2a,t =

cm,4a,t = cm,1b,t = cm,2b,t = cm,3b,t = cm,1c,t, the alternative hypothesis is that cm,i,t cm,j,t

for at least one pair cm,i,t, cm,j,t. The probabilities will display the significance of the influence of the different magnitudes.

To test if the results are significant, an F-test is required. When using an F-test, the significance of the null hypothesis, that the ACAR is the same for the various causes, magnitudes or the combination of the 2, can be tested. Following Newbold, Carlson and Thorne (2003), the F-test will be performed in the following way:

First the sample will be divided into groups of causes, magnitudes and the combination of the two. Thereafter the sum of squares within the groups (SSW) and between the groups (SSG) will be measured. SSW = = = − K i n j i ij i ACAR CAR 1 1 2 ) ( (18) SSG = 2 1 ) (CAR ACAR n K i= i i − (19)

CARij denotes the jth sample observation in the ith group. ACARi, is the sample mean of group i

and ACAR is the overall sample mean. K is the number of groups and ni is the total number of

sample observations. Both variables differ at different tests because of the different samples. From these equations, the within-groups mean square (MSW) and the between groups mean square (MSG) can be denoted.

MSW = K n SSW − (20) MSG = 1 − K SSG ₍₂₁₎

(24)

F = _MSWMSG ~F_(K_-_1,_n_-_K,α₎ (22)

Where is the level of significance. K-1 and n-K are the degrees of freedom in the numerator and the denominator respectively. For this thesis the following degrees of freedom are applicable: 3/235 for sub question 3, 2/250 for sub question 4 and 6/207 for sub question 5.

These tests will be performed for t=1, t=5 and t=20.

6. Results:

(25)

6.1 Results for all airlines:

Graph 1: ACARs All Airlines

For clarity it is chosen to adjust the ACAR of t=0 to zero. This way the ACARs at t=1/5/20 can be directly concluded from the graph.

All Airlines ACAR t-value Sign. t=1 -0.0139 -8.9196 *** t=5 -0.0187 -11.6929 *** t=20 -0.0188 -11.0049 ***

Table 5: ACAR and t-values for all airlines.

The empirical analysis starts by examining the abnormal stock price reaction, of airlines as a result of aviation disasters, over the various time horizons after the disasters. Graph 1 illustrates the ACARs for all airlines graphically. Table 5 gives an overview of the specific results of the aviation disasters on the average cumulative abnormal return on the share value of the afflicted airlines at t=1/5/20. In this table, one can see that the airlines in the sample experience an average abnormal price decline of 1.39% at t=1, 1.87% at t=5 and 1.88% at t=20. All these negative returns are significant at a 1% level. The raise in negative ACAR from t=1 to t=5 implies that the market does not find its new equilibrium immediately after the disaster but adjusts for a couple of

ACARs All Airlines

-3,00%

-2,50%

-2,00%

-1,50%

-1,00%

-0,50%

0,00%

0,50%

1,00%

1,50%

-120

-110

-100

-90

-80

-70

-60 -50

-40

-30

-20

-10

0

10

20 Days before/after the event

(26)

days to find a new equilibrium. This is not in accordance with the hypothesis of efficient markets which implies that all adjustments should be made on the first trading day. A possibility is that this is the result of the release of new information about the disaster, as a result of which a delayed reaction arises. Thereafter the negative ACAR increases slightly. This can indicate that the market moves towards its new equilibrium. Taking into account the outcomes of the first sub question, enough evidence is given to reject H0 that after an aviation disaster there is no

(27)

6.2 Results for all manufacturers:

ACAR's Manufacturers

-1,00%

-0,50%

0,00%

0,50%

1,00%

1,50%

2,00%

2,50%

-12

0 -11

0 -10

0 -90 -80 -70 -60 -50 -40 -30 -20 -10 0 10 20

Days before/after the event

A

C

A

R

Graph 2: ACARs Manufacturers

Although it appears that the ACARs during the estimation window are above zero, no logical explanation can be given. Again, clarity it is chosen to adjust the ACAR of t=0 to zero. This way the ACARs at t=1/5/20 can be directly concluded from the graph.

All

Manufacturers ACAR t-value Sign. t=1 -0.0007 -0.5154 t=5 -0.0038 -2.5801 *** t=20 0.0077 4.8466 ***

Table 6: ACAR and t-values for all manufacturers.

Graph 2 illustrates the ACARs for all manufacturers graphically. Table 6 gives an overview of the impact of the aviation disasters on the abnormal return on the share values of the afflicted

manufacturers at t=1/5/20. In this table, you can see that the manufacturers in the sample

(28)

manufacturers’ shares is positive in the long run. It seems unlikely that investors anticipate on the new demand of an aircraft. Therefore it is a remarkable result considering the nature of the event. When looking at the share prices of Boeing, this result can be explained. Boeings' share price contains a number of large increases in a short period over time. This influences the results strong. Because of these fickle but significant results, it is not possible to draw an unambiguous conclusion for the short-run and long-run. Nevertheless, the H0 of no significant abnormal return

on the share prices of the afflicted manufacturers after an aviation disaster can be rejected.

6.3 Results of the different causes

Causes t=1 t=5 t=20 ACAR t-value Sign. ACAR t-value Sign. ACAR t-value Sign. Human error -0.0132 -5.4886 *** -0.0127 -5.1603 *** -0.0065 -2.4684 *** Mechan. fail. -0.0186 -6.1401 *** -0.0322 -10.2344 *** -0.0474 -14.1098 *** Nature -0.0181 -3.8625 *** -0.0151 -3.1883 *** -0.0338 -6.6679 *** Criminal act. -0.0048 -1.1179 -0.0233 -5.2659 *** -0.0105 -2.2086 ** Mean -0.0139 -0.0193 -0.0204

Table 7: ACAR and t-values for the different causes.

(29)

Causes F-value Sign. t=1 1.347612 t=5 1.040326 t=20 1.451341

Table 8: F- values for the different causes

In table 8 the results of the F-test for the different causes are shown. This table displays if the various ACARs are really different. Since there are no significant outcomes, the table suggests that all causes have the same impact on the negative abnormal return of the airlines. Therefore, the null hypothesis that all causes have the same influence on the share price of the afflicted airlines can not be rejected.

6.4 Results of the different magnitudes:

Magnitudes t=1 t=5 t=20 Nr of fatalities: ACAR t-value Sign. ACAR t-value Sign. ACAR t-value Sign. 1 - 10 -0.0083 -2.9687 *** -0.0040 -1.3766 * -0.0148 -4.7832 *** 11 – 99 -0.0125 -6.3749 *** -0.0191 -9.5814 *** -0.0119 -5.5777 *** 100 or more -0.0391 -6.9702 *** -0.0622 -10.7315 *** -0.0698 -11.2193 *** Mean -0.0139 -0.0187 -0.0189

Table 9: ACAR and t-values for the different magnitudes.

Table 9 shows the influence of the different magnitudes. The results of magnitude a are -0.83% at t=1, significant at a 1% level, -0.40% at t=5, at a 10% significance level and -1.48% at t=20, at a 1% significance level. For magnitude b, t=1 has a result of -1.25%, t=5 has a result of -1.91% and t=20 has a result of -1.19%, all significant at a 1% level. For magnitude c all results are significant at a 1% level as well; -3.91% at t=1, -6.22% at t=5 and -6.98% at t=20. Remarkable in table 8 is that the result of magnitude b is lower than the result of magnitude a at t=20. For the rest of the table, the results are as expected.

(30)

over a longer period than when dealing with smaller disasters. Summarizing this, the results seem to suggest that reactions differ for different magnitudes. To test this, the F-test is applied again.

Magnitudes F-value Sign. t=1 9.0885 *** t=5 7.3032 *** t=20 2.2817

Table 10: F-values for the different magnitudes.

Table 10 illustrates the results of the F-test for the different magnitudes. Again, this table shows if the various ACARs are really different. Since the results are significant at a 1% level for t=1 and t=5, it is safe to say that the ACARs for these different groups differ. At t=20 the results are insignificant. Combining these results it is safe to reject the H0 that after an aviation disaster,

there is no significant different abnormal return on the share prices of the afflicted airlines as a result of different magnitudes.

6.5 Results of the different causes combined with the different magnitudes:

C&M t=1 t=5 t=20 ACAR t-value Sign. ACAR t-value Sign. ACAR t-value Sign. 1a -0.0087 -2.0725 ** 0.0091 2.0658 ** -0.0007 -0.1538 2a -0.0146 -2.3495 *** -0.0194 -3.0829 *** -0.0201 -2.9193 *** 4a -0.0028 -0.5292 -0.0106 -2.0124 ** -0.0318 -5.6727 *** 1b -0.0116 -3.6682 *** -0.0211 -6.5849 *** 0.0006 0.1741 2b -0.0101 -2.8085 *** -0.0165 -4.5045 *** -0.0387 -9.8250 *** 3b -0.0200 -4.1006 *** -0.0146 -2.9661 *** -0.0464 -8.6595 *** 1c -0.0388 -4.7932 *** -0.0340 -4.1966 *** -0.0707 -7.8242 *** Mean -0.0123 -0.0138 -0.0198

Table 11: ACAR and t-values for the different causes combined with the magnitudes.

(31)

is not significant. The result of human error combined with the biggest magnitude is significant at a 1% level at all points in time. The ACAR is -3.88%, -3.40% and -7.07% at t=1/5/20 respectively. The ACAR of mechanical failure is significant at a 1% level at all times and for both magnitudes a and b. For magnitude a, the results are -1.46%, -1.94% and 2.01% at t=1/5/20. For magnitude b, the results are -1.01% at t=1, -1.65% at t=5 and -3.87% at t=20. Nature combined with magnitude b has an ACAR of -2.00% at t=1, -1.46% at t=5 and -4.64 at t=20, all significant at a 1% level. Criminal activity combined with magnitude a has an insignificant result at t=1. At t=2, it has an ACAR of -1.06% significant at a 5% level. At t=20 the ACAR is -3.18% significant at a 1% level. The ACARs of mechanical failure combined with magnitude a and b and of criminal activity combined with magnitude a become more negative over time. When these results are compared with mechanical failure as a cause, the results seem logical since mechanical failure has the same negative slope. For criminal activity combined with magnitude a the result is less obvious. Human error combined with magnitude a does not have a large impact which corresponds with the results for magnitude a. Human error combined with magnitude b corresponds with the results of magnitude b as well. The slope of nature combined with magnitude b corresponds with the slope of nature. Finally, human error's slope combined with magnitude c does not correspond with any of those variables. Again, the results seem to suggest that reactions differ for different magnitudes. To test this, the F-test is applied again.

C&M F-value Sign. t=1 1.918098 * t=5 1.111124 t=20 1.053246

Table 12: F-values for the different causes combined with the magnitudes

Table 12 demonstrates the results of the various causes combined with magnitudes. Once more, this table shows if the various ACARs are really different. The results at t=5 and t=20 are insignificant, but the result at t=1 is significant at a 10% level. Therefore, it is safe to reject the H0

(32)

7. Conclusions:

The results of this thesis suggest that aviation disasters have a negative impact on the share price of the afflicted airlines. The market responds immediately after the announcement by providing the shareholders of the afflicted airline a negative abnormal return. After 5 trading days, the reaction is stronger after which the negative cumulative abnormal return stabilizes at 20 trading days. Because it is unknown if new, important, information about the disaster is disclosed in the following days, it is uncertain if the efficient market hypothesis does not apply. These results are conform the results of most earlier studies. Davidson Chandy & Cross (1987), Barret, Heuson & Colb (1987), Chance & Ferris (1987), Mitchell & Maloney (1989), Bosch, Eckard & Singal (1998) and Walker, Thiengtham & Lin (2005) found a negative cumulative abnormal return as well. Only Borenstein & Zimmerman (1988) did not find a similar result, their outcome was insignificant.

For the share price of manufacturers, it is more difficult to draw an unambiguous conclusion due to the significant but fickle results. Although more research should be done to explain the results, it is safe to say that the stock market reacts negatively to aviation disasters. Only three of the used studies examined the impact on manufacturers as well. Chalk (1986) also found a significantly negative return, but only for one crash. Walker, Thiengtham & Lin (2005) performed a broader study and found negative reactions as well. Chance and Ferris (1987) did not find a negative abnormal return. Generally one can say that the outcomes of this study are equivalent with former studies.

The results of the first two sub questions do indicate that investors relate aviation disasters more to airlines than to the manufacturers. Probably this is because the airline is more illuminated than the manufacturer. Another possibility is that due to the fact that airlines and manufacturers have long lasting contracts, the future impact is larger for airlines than for manufacturers. Where passengers can stay away easily from an airline, it is more difficult to switch manufacturers. Additionally, the included manufacturers are larger companies than the airlines. Therefore theoretically, manufacturers can bear the costs better. Investors anticipate on these reasons, and therefore the consequences can be different for airlines and manufacturers.

(33)

Different magnitudes of disasters do have a different impact on the share price of the airlines. The higher the number of fatalities, the higher the negative abnormal return is the conclusion. This indicates that investors' reactions discriminate between the different magnitudes. Only Davidson Chandy & Cross (1987) examined the influence of bigger magnitudes and they found similar results.

When the different causes are combined with the different magnitudes, there only is a significant different return at trading day 1. This indicates that investors differentiate their reactions on the first trading day and they do not differentiate afterwards.

In general, for the parts of this thesis that overlap earlier studies, this thesis shows outcomes that confirm the outcomes of earlier studies.

8. Recommendations for further research:

First the problems for writing this thesis will be given:

The first problem was in the second sub question. Due to the fact that there are not many manufacturers listed and included in datastream or CRSP, the result is strongly influenced by Boeing and Douglas. They contain more than two-third of the events used in this study.

Another problem is that the announcement time of the disasters is not taken into account in this thesis. Therefore the result a t=1 can be the result over two days because presumably a number of disasters, and their announcement, took place before the stock market was open at t=0. Deliberately t=1 is chosen and not t=0 because when you do not take into account the exact time of the announcement, the result at t=1 is less biased.

More problems came in during the third and fifth sub questions. Most of the times the cause is not immediately released at the announcement time of the disaster as a result of which it is difficult to interpret these short-term outcomes. Normally an investigation is held before releasing the cause. Only in the case of criminal activity and in some very clear other cases, the cause is immediately known.

Finally, although realizing the list of different manners in which this subject can be examined is not complete at all, some other possibilities for further research will be pointed out here.

(34)

The results can get more transparent by analyzing the CARs step by step. If the results are examined day after day without intermittencies, the outcomes will get less biased.

In addition, the following questions could be addressed.

When an airline is afflicted by more than one disaster in a relative short period or has the same cause repeatedly, does that influence the effects on the abnormal return of that security? If the negative effect becomes stronger, it means that investors lose faith in that airline.

Does the abnormal return vary among regional / national or international airlines and among different countries?

Do aviation disasters have an effect on the airline industry as a whole?

(35)

References:

• Barrett, B.W., Heuson, A.J., Kolb, R.W. and Schropp, G.H. (1987), “The adjustment of stock prices to completely unanticipated events,” Financial Review 22, 345-354.

• Binder, J.J. (1985), “On the use of the multivariate regression model in event studies”,

Journal of Accounting Research 23, 370-383.

• Borenstein, S. and Zimmerman, M.B. (1988) "Market incentives for safe commercial airline operation", American Economic Review 78, pp. 913-35.

• Bosch, J.C., Eckard, W. and Singal, V. (1998) "The Competitive Impact of Air Crashes: Stock Market Evidence", Journal of Law and Economics, Vol. 41, No. 2, 503-519

• Brooks, C. (2002), "Introductory econometrics for finance", Cambridge University Press, Cambridge, UK

• Brown, R. (1998), "Rising Costs of Disaster", Reactions 18:1, January, 27-30.

• Brown, S.J. and Warner, J.B. (1980), "Measuring Security Price Performance", Journal

of Financial Economics 8, pp. 205-258, North-Holland Publishing Company

• Brown, S.J and Warner, J.B. (1985), "Using daily stock returns: The case of event studies", Journal of Financial Economics 14, pp. 3-31, North-Holland Publishing Company

• De Bondt, W. F. M. and Thaler, R. (1985), "Does the Stock Market Overreact?", Journal

of Finance, Vol. 40 Issue 3, 793-806

(36)

• Carter, D.E., Simkins, B.J. (2004), " The Market's reaction to unexpected, catastrophic events, the case of airline stock returns and the September 11th attacks", Quarterly

Review of Economics and Finance, 44 (4), 539-558

• Chalk, A. (1986) "Market forces and Aircraft safety: The Case of the DC-10", Economic

Inquiry, 24:1, 43-60.

• Chalk, A. (1987) "Market forces and commercial aircraft safety", Journal of Industrial

Economics, 36:61-81.

• Chance, D.M. and Ferris, S.P. (1987), "The effect of aviation disasters on the air transport industry: A financial market perspective", Journal of Transport Economics and Policy 21

(May), pp. 151-165

• Davidson III, W.N.; P.R. Chandy and M. Cross (1987), Large losses, risk management

and stock returns in the airline industry, The Journal of Risk and Insurance Vol. 54 No.

1, 162-172

• Economist, The. (1997a) "How safe is your airline", January 11, 13,14

• Economist, The. (1997b) "Fasten your safety belt", January 11, 61-63

• FAA, Safety Report: Aviation Safety Data Accessibility function, January 20, 1997

• Fama, E.F., Fisher, F., Jensen, M. and Roll, R. (1969), “The adjustment of stock prices to new information,” International Economic Review 10, 1-21.

• Fama, E.F. (1970), "Efficient capital markets: A review of theory and empirical work",

Journal of Finance 25, pp. 383-417

(37)

• Ito, H. and Lee, D. (2003), "Assessing the Impact of the September 11 Terrorist Attacks on U.S. Airline demand", Brown University working paper, No. 2003-16

• Jaffe, J. and Westerfield, R. (1985), "Patterns In Japanese Common Stock Returns: Day of the Week and Turn of the Year Effects", Journal of Financial and Quantitative

Analysis 20(2), 261-272

• Kallberg, J., Liu, C.H. and Pasquariello, P. (2005), "Updating Expectations: An Analysis of Post-9/11 Returns", AFA 2006 Boston Meetings Paper

• Karafiath, I. (1988), “Using dummy variables in the event study methodology”, Financial

Review 23, 351-357.

• MacKinlay, A.C. (1997), "Event studies in economics and finance", Journal of Economic

Literature 35, pp 13-39

• Mitchell, M.L. and Maloney, M.T. (1989) "Crisis in the cockpit? The role of market forces in promoting air travel safety", Journal of Law and Economics, 32: 329-355.

• Moore, D.S. and McCabe, G.P. (1994), Statistiek in de praktijk, Academic Service, Schoonhoven

• Newbold, P., Carlson, W.L. and Thorne, B.M. (2003), Statistics for Business and

Economics, Prentice Hall, New Jersey

• Park, N.K. (2004), "A guide to using events study methods in multi-country settings",

Strategic Management Journal 25, pp. 655-668

• Ross, S.A., Westerfield, R.W. and Jaffe, J. (2005), Corporate Finance, Mc Graw Hill, Singapore

(38)

• Singal, V. (1997), "Financial Health and Airline Safety", American Economic

Association

• Walker, J.T., Thiengtham, D.J. and Lin M.Y. (2005) "On the performance of airlines and airplane manufacturers following aviation disasters", Canadian Journal of Administrative

(39)

Appendices:

Appendix 1, total sample:

Nr. Airline Date Manufacturer Fatalities POB Cause Mag. C&M Country

1 Air Florida 13-1-1982 Boeing 74 79 1 b 1b USA

2 Air France 18-10-1973 Boeing 1 110 4 a 4a France

3 Air France 24-7-1974 Fokker 3 3 1 a 1a France

4 Air France 28-8-1976 Airbus 1 20 4 a 4a France

7 Air France 25-7-2000 Concorde 109 109 2 c 2c France

8 Air Midwest 1-8-2003 Beech 21 22 2 a 2a USA?

9 Air Wisconsin 12-6-1980 Swearingen 13 15 3 b 3b USA

10 Alaska Airlines 4-9-1971 Boeing 111 111 2 c 2c USA

11 Alaska Airlines 5-4-1976 Boeing 1 57 1 a 1a USA

12 Alaska Airlines 31-1-2000 Mc Donnel Doug. 88 88 2 b 2b USA

13 Alitalia 14-11-1990 Mc Donnel Doug. 46 46 2 b 2b Italy

14 All Nippon 17-3-1977 Boeing 1 ? 4 a 4a Japan

15 All Nippon 23-7-1999 Boeing 1 517 4 a 4a Japan

16 Allegheny 24-12-1968 Convair 20 47 1 b 1b USA

18 Allegheny 9-9-1969 Mc Donnel Doug. 82 82 2 b 2b USA

20 Allegheny 12-2-1979 Mohawk 2 25 1 a 1a USA

21 American Airlines 29-7-1943 Douglas 20 22 3 b 3b USA

23 American Airlines 10-2-1944 Douglas 24 24 ? b ?b USA

27 American Airlines 28-12-1946 Douglas 2 21 2 a 2a USA

29 American Airlines 22-8-1950 Douglas 1 59 2 a 2a USA

30 American Airlines 22-1-1952 Convair 27 27 3 b 3b USA

32 American Airlines 6-7-1954 Douglas 1 ? 4 a 4a USA

35 American Airlines 6-1-1957 Convair 1 10 1 a 1a USA

36 American Airlines 3-2-1959 Lockheed 65 73 1 b 1b USA

37 American Airlines 1-3-1962 Boeing 95 95 2 b 2b USA

(40)

41 American Airlines 20-12-1995 Boeing 160 164 1 c 1c USA

42 American Airlines 1-6-1999 Mc Donnel Doug. 11 145 3 b 3b USA

43 American Eagle 31-10-1994 Saab 68 68 3 b 3b USA

44 American Eagle 13-12-1994 British Aerospace 15 20 1 b 1b USA

45 Atlantic Coast 7-1-1994 Jetstream 5 9 1 a 1a USA

46 Atlantic Southeast 5-4-1991 Embraer 23 23 2 b 2b USA

47 Atlantic Southeast 21-8-1995 Embraer 8 29 2 a 2a USA

48 Braniff 22-8-1954 Douglas 12 19 3 b 3b USA

49 Braniff 17-7-1955 Convair 22 43 3 b 3b USA

50 Braniff 25-3-1958 Douglas 9 24 1 a 1a USA

51 Braniff 29-9-1959 Lockheed 34 34 2 b 2b USA

52 Braniff 6-8-1966 BAC 42 42 1 b 1b USA

53 Braniff 3-3-1968 Lockheed 85 85 3 b 3b USA

54 Capital Airlines 12-12-1949 Douglas 6 23 2 a 2a USA

55 Capital Airlines 6-4-1958 Vickers Viscount 47 47 3 b 3b USA

57 Capital Airlines 12-5-1959 Vickers Viscount 31 31 1 a 1a USA

58 Capital Airlines 12-5-1959 Lockheed 2 44 3 b 3b USA

60 China Air lines 26-4-1994 Airbus 264 271 2 c 2c Taiwan

61 China Air lines 16-2-1998 Airbus 196 196 1 c 1c Taiwan

62 China Air lines 22-8-1999 Mc Donnel Doug. 3 315 1 a 1a Taiwan

63 China Eastern 22-12-1997 Airbus 1 ? 4 a 4a China

64 Comair 9-1-1997 Embraer 29 29 1 b 1b USA

65 Continental 22-5-1962 Vickers Viscount 8 8 4 b 4b USA

66 Continental 29-1-1963 Boeing 45 45 3 a 3a USA

67 Continental 15-9-1975 Boeing 1 5 4 a 4a USA

68 Continental 1-3-1978 Mc Donnel Doug. 2 200 2 a 2a USA

69 Crossair 10-1-2000 Saab 10 10 1 a 1a Switserland

70 Delta Airlines 31-7-1973 Mc Donnel Doug. 89 89 1 b 1b USA

71 Delta Airlines 22-2-1974 Mc Donnel Doug. 3 14 4 a 4a USA

72 Delta Airlines 2-8-1985 Lockheed 134 163 3 c 3c USA

73 Delta Airlines 31-8-1988 Boeing 14 108 1 b 1b USA

74 Delta Airlines 6-7-1996 Mc Donnel Doug. 2 147 2 a 2a USA

75 Eastern Airlines 12-7-1945 Douglas 1 24 1 a 1a USA

76 Eastern Airlines 7-9-1945 Douglas 22 22 2 b 2b USA

82 Eastern Airlines 7-2-1948 Lockheed 1 69 2 a 2a USA

(41)

87 Eastern Airlines 4-10-1960 Lockheed 62 72 3 b 3b USA

92 Eastern Airlines 17-3-1970 Mc Donnel Doug. 1 73 4 a 4a USA

93 Eastern Airlines 29-12-1972 Lockheed 99 176 1 b 1b USA

94 Eastern Airlines 11-9-1974 Mc Donnel Doug. 70 82 1 b 1b USA

95 Eastern Airlines 24-6-1975 Boeing 115 124 3 c 3c USA

96 Eastern Airlines 10-4-1981 Airbus 1 148 4 a 4a USA

97 Eastern Airlines 1-1-1985 Boeing 29 29 1 b 1b USA

98 Frontier 21-12-1967 Douglas 2 2 1 a 1a USA

99 Frontier 20-10-1977 Boeing 1 34 4 a 4a USA

100 Japan Air Lines 27-9-1977 Mc Donnel Doug. 34 79 1 b 1b Japan

101 Japan Air Lines 9-2-1982 Mc Donnel Doug. 24 174 1 b 1b Japan

102 Japan Air Lines 12-8-1985 Boeing 520 524 2 c 2c Japan

103 Kenya 30-1-2000 Airbus 169 179 1 c 1c Kenya

104 KLM 27-3-1977 Boeing 248 248 1 c 1c NL

105 Korean Air 29-11-1987 Boeing 115 115 4 c 4c South Kor.

106 Korean Air 27-7-1989 Mc Donnel Doug. 75 199 3 b 3b South Kor.

107 Korean Air 6-8-1997 Boeing 228 254 1 c 1c South Kor.

108 Lauda Air 26-5-1991 Boeing 223 223 2 c 2c Austria

109 Lufthansa 17-12-1973 Boeing 1 ? 4 a 4a Germany

110 Lufthansa 20-11-1974 Boeing 59 157 1 b 1b Germany

111 Lufthansa 13-10-1977 Boeing 4 91 4 a 4a Germany

112 Lufthansa 14-9-1993 Airbus 2 70 1 a 1a Germany

113 Malaysian Airlines 15-9-1995 Fokker 34 53 1 b 1b Malaysia

114 Mohawk 2-7-1963 Martin 7 43 3 a 3a USA

115 Mohawk 23-6-1967 BAC 34 34 2 b 2b USA

116 Mohawk 19-11-1969 Fairchild 14 14 1 b 1b USA

117 Mohawk 26-1-1972 Fairchild 1 47 4 a 4a USA

118 Mohawk 3-3-1972 Fairchild 16 48 1 b 1b USA

119 National 5-10-1945 Lockheed 2 15 1 a 1a USA

120 National 14-1-1951 Douglas 7 28 1 a 1a USA

121 National 11-2-1952 Douglas 29 63 2 b 2b USA

123 National 16-11-1959 Douglas 42 42 ? b ?b USA

125 National 3-11-1973 Mc Donnel Dou 1 128 2 a 2a USA

126 National 8-5-1978 Boeing 3 58 1 a 1a USA

127 Northeast Airlines 25-10-1968 Fairchild 32 42 2 b 2b USA

128 Northwest Airlines 12-3-1948 Douglas 30 30 1 b 1b USA

129 Northwest Airlines 29-8-1948 Martin 37 37 2 b 2b USA

130 Northwest Airlines 7-3-1950 Martin 13 13 3 b 3b USA

131 Northwest Airlines 23-6-1950 Douglas 58 58 ? b ?b USA