The impact on the airline industry as a result of aviation disasters

The impact on the airline industry as

a result of aviation disasters

Demi Poppe

Student number: S4180704

Student mail: d.a.poppe@student.rug.nl

University of Groningen, Faculty of Economics and Business

Pre-MSc. Finance

Supervisor: J. Offerein

Date: 28-05-2020

ABSTRACT

This paper examines an event study to observe the effect of aviation disasters on the airline industry. The results demonstrate, according to the Cowan generalized sign tests, that the airline industry as a whole (including the operator involved in the crash) is experiencing significant losses reflected by a decline in stock returns. Next to this, this study assumed, through a Mann-Whitney U-test, that the amount of casualties does significantly result in what extent firms are suffering from an aviation disaster.

Keywords: Aviation disasters; Airline industry; Stock price returns; Casualties

1. INTRODUCTION

Altough the aviation history is already more than 100 years old, air crashes continue to occur (Chen, Wu, & Sheu, 2008). Such crashes unfavorably affect the stock prices of the involved airlines since they incur significant financial losses, both during and after tragedies (Ho, Qiu, & Tang, 2013). As a consequence, stock prices tend to drop as a reflection on the decrease in the operators’ expected future cash flows. Sprecher and Pertl (1986) even observed on the day of the event, the abnormal returns presented a negative return of 4 percent. Therefore, there can be assumed that the stock market responds rapidly to

catastrophic events (Change & Ferris, 1987), involving air crashes (Carter & Simkins, 2004). Additionally, several studies show a distinction in the number of fatalities reflected in the impact on the stock prices of the involved airline operators (Kaplanski & Levy, 2010).

Gigerenzer (2004), and Sivak and Flannagan (2004) even observed a decline in the overall air travel demand after an air crash. Hence, aviation disasters should be considered as an

international issue. Therefore, this research will examine the impact on both the involved operator as the airline industry as a result of an aviation disaster.

To elaborate further on what exactly follows in the demand after such a disaster, Tversky and Kahneman (1979) investigated individuals and stated that since people are generally limited in their capability to comprehend and evaluate extreme predictions, greatly unlikely events are either overlooked or ignored. As a consequence, whenever an unexpected event happens, people will perceive this as a shock and involved companies are even more suffering since industrial accidents also cause significant damage to the reputation for the social performance of a business form (Zyglidopoulos, 2001). This underlines the consequences as a result of an air crash and its importance to investigate further. Sequentially, to determine to what extent the airline industry actually suffers after an air crash, the following research question has been determined: “How do aviation crashes affect stock prices in the U.S. airline industry?”.

This paper accounts for an event study that investigates the response on stock prices of 38 companies while encountering aviation disasters. The study consists of 51 events that occurred from 2001 till 2009 in countries considered as Western civilization1_{. The results presenting that} both the carrier and the complete industry suffer from aviation disasters. Also, the number of casualties result in the extent to which the operators are suffering.

2. THEORETICAL BACKGROUND

As aforementioned, aviation disasters seem unavoidable. According to Chen, Wu, and Sheu (2008) these disasters are comparable with explosive bombing and generate various trauma patients. This is corroborated by Kaplanski and Levy (2010), who associated aviation disasters with bad mood and anxiety. Cobb and Primo (2004) states that by nature, individuals are not inclined to relinquish control over their fate. Nevertheless, people do admit this

whenever they decide to fly. As a result, faith in the operator is contingent on the public understanding the reason a plane crashed. Unfortunately, most travelers have little

understanding of such a crash. The reason for this is because they tend to trust the assumption that airline operators are doing everything needed to keep air travel safe and thus not to crash (Siomkos, 2000). Ekeberg, Fauske, and Berg-Hansen (2014) state that 4.5% had cancelled at least one flight because of flight apprehension after a disaster that occurred within the last two years. Hence, airline operators will suffer since their sale is decreasing and as a result, their stock prices will drop (Li et al., 2015). Therefore, the following hypothesis is stated:

H1: As a consequence of an airplane crash in countries considered as Western, U.S. airline operators experience negative stock returns.

H2: The negative market return of the stocks only holds for the involved airline operator and not for the total U.S airline industry.

The impact on firms and stakeholders after a disaster are very diverse per study. Berkman, Jacobsen, and Lee (2011) conclude that the more life-threatening a crisis is, the greater the probability of an intense decline in consumption and, therefore, the huger the effect on stock returns. This is underlined by Kaplanski, and Levy (2010) who found that whenever there is relatively a large amount of casualties concerned at an air crash, the average rates of return will largely decline. This is mainly due to habits to share terrifying stories about certain crashes which will deteriorate individuals’ moods. Nonetheless, Ho, Qiu, and Tang (2013) states that relatively small scale aviation disasters receive less attention from the media and public, and, therefore, the stock prices do barely suffer from its consequences. Hence, the following hypothesis is composed:

3. DATA AND METHODOLOGY 3.1 DATA

The data of this study covers 38 airlines responding to airplane crashes. To get more insight to what extent these disasters have, the historical data of the stock market index of the S&P 500 is included as a comparable framework. The airplane crashes are selected on several criteria. First of all, the crashes itself occurred in countries considered to effectuate a Western civilization since these countries are generally more precise in realizing stock price data. Next to this, the crashes occurred from 2001 till 2009. Flights intended for taxi, ambulance, cargo, military, and private purposes are excluded. The selected disasters incorporate 51 crashes with a total amount of 1.645 casualties. The event list can be seen in the appendices. The primary source of these aviation disasters is The Aviation Safety Network. To ratify the exact casualties and location the dataset data.world has been used.

To provide an answer to hypothesis 1, the impact relative to the total market has been observed. The data necessary to answer this is the S&P 500, which is collected via yahoo.finance and the stock prices of the 38 investigated companies, which are gathered from CRSP. Sequentially, answer the second hypothesis which concerns the financial impact on the involved operator, the author has observed the stock prices of 10 crashes (out of the 51 total) divided over 7 companies. Eventually, to answer hypothesis 3, the author divided the 51 crashes into two groups called the ‘large group casualties’ and the ‘small group casualties’. This distinction is made through the median, which is 6. Less than and including 6 are considered to be a small group and the number of casualties over 6 is considered to be a large group. Table 2 in the appendices includes a small insight into these group divisions.

3.2 METHODOLOGY

To test the null hypothesis and in order to understand how an event affects the stock prices over time, the market model is used. In here, the actual returns of the reference market (S&P 500) are observed with the stock prices of the operators. This model eventually tracks the abnormal returns on the specific days of an event window. The formula for the abnormal return is presented as follows:

𝐴𝑅_$,& = 𝑅_$,&− (𝛼+_$ + 𝛽._$ ∗ 𝑅_0,&) (1)

This formula describes the relationship between the operator’s stock, its reference index (represented by 𝛼+_$ and 𝛽.), and the actual return of the market (𝑅_0,&). 𝐴𝑅_$& represents the abnormal return of a certain stock day 𝑡 on event 𝑖 with 0 as the event date, 𝑅_$& is the stock return on day 𝑡 on event 𝑖, 𝛼+_$ functions as the constant parameter as the OLS estimator, 𝛽._$ is the OLS estimator of the slope parameter and 𝑅0& function as the actual return of the market

(S&P 500) at time 𝑡.

The following procedure is to calculate the average of the total abnormal returns (AAR) to test whether airplane crashes do actually have a negative effect on the U.S. stock market. To perform this, the author made use of the following formula:

𝐴𝐴𝑅_$ = ₅4∑5 𝐴𝑅_$,&

$74 (2)

Here, N is accounted for the total number of events (51 in this study) which is multiplied with all the abnormal returns in order to get the Average Abnormal Returns.

The next step is to measure the total impact of an air crash during the event window. The individual abnormal returns are summed up, presented as CAR (cumulative abnormal return):

𝐶𝐴𝑅$(9:;,:;) = ∑&:&7&4𝐴𝑅$,& (3)

Finally, the CAAR (Cumulative Average Abnormal Return) is calculated through the following formula:

𝐶𝐴𝐴𝑅(𝑡₄, 𝑡_:) =₅4∑5$74𝐶𝐴𝑅_$(𝑡4, 𝑡:) (4)

3.3 DISTRIBUTION

In order to find significant returns, the distribution of the cumulative abnormal returns first have to be determined. Therefore, the Jarcque-Bera test is used within this study. The formula used to obtain the result of this distribution is as follows:

𝐽𝐵 = >_?@𝑠:₊4

B(𝐾 − 3)

:_{E (5)}

Where 𝑁 serves as the number of observations (51), 𝑠 is the sample skewness, and 𝐾 is the kurtosis of the sample.

3.4 SIGNIFICANCE TESTING

Within this study, it turned out that there is a case of a non-normal distribution. Therefore, there are multiple tests used to increase the power of the study. First, the t statistic is provided of the cumulative abnormal returns for each firm. This statistics operates under the null hypothesis of 𝐶𝐴𝑅_$ = 0, and can be defined as:

𝑡_GHI =𝐶𝐴𝑅$ 𝑆_GHI

Where 𝑆_GHI is the standard deviation across firms.

𝑝̂ = 1 NP 1 𝐿₄ 5 $74 P 𝜑_$,& S! &7S" (6)

Where 𝐿4 is the count of non-missing return values in the event window and the last fraction

(𝜑_$,&) counts as 1 if the sign turned out to be positive and -1 otherwise. The formula to obtain the generalized sign test whenever 𝐻_;: 𝐶𝐴𝐴𝑅 = 0, is as follows:

𝒵_XY$X> = (𝑤 − 𝑁𝑝̂) [𝑁𝑝̂(1 − 𝑝) \

(7)

Here, 𝑤 serve as the number of stock with positive CAR during the event period.

Additionally, to test the third hypothesis, the Mann-Whitney U-test is executed. By performing this test, the author observed whether the returns of stock prices are only affected by a relatively large amount of casualties or not. This test is chosen since there is, again, non-normal distribution of the data. The two groups are divided into quantities of 22 and 29. Due to the size of those groups, the value of U approaches a normal distribution and so the null hypothesis can be tested through a Z-score. First, the U scores are calculated according to the following formula:

𝑈_4/: = (𝑛1 ∗ 𝑛2) +𝑛4/:∗ (𝑛4/:+ 1)

2 − 𝑇4/: (8) 𝑇_4/: is the sum of the ranks for one of the two samples. Second, the standard deviation (SD) is calculated as following:

𝑆𝐷 = e(𝑛1 ∗ 𝑛2) ∗ (𝑛1 + 𝑛2 + 1)

12 (9)

And sequentially, the 𝑍 score is computed as following:

𝑍 = 𝑈 − ((𝑛1 + 𝑛2)/2)

4. RESULTS

As aforementioned, to acknowledge significant returns, the distribution of the cumulative abnormal returns has to be determined first. Therefore, the Jarcque-Bera test is used for the first hypothesis as presented underneath:

Table 1 – Jarque-Bera Normality test (JB) H1

CAAR (-20,20) CAAR (-5,5)

Kurtosis -0,951948514 11,07672483

Skewness -0,100131567 -3,274260241

JB Statistic 33,27325506 229,7477845

Conclusion Reject - No normal distribution Reject - No normal distribution

The results of the JB test represents the availability of a non-normal distribution of the data. This result will clarify the tests that the author will examine to give answers to the subsequent hypothesis.

4.1 TESTING THE EFFECT OF AN AVIATION DISASTER ON THE TOTAL AIRLINE INDUSTRY

As can be seen in table 1, there can be concluded that there is no normal distribution identified. Therefore, the author decided to give the results extra power, a t-test is examined as well as the Cowan sign test (1992) which are presented in the following tables.

Table 2 - Significance test H1

CAR (-20,20) CAR (-5,5)

Average -0,04794678 -0,01207935

SD 0,113148028 0,045579046

T-statistic -3,026199401 -1,892620033

Accept/reject Reject* Reject***

Notes. Where *** is significant at 5% level, ** is

Table 3 – Cowan sign test H1 CAR (-20,20) CAR (-5,5) Proportion positive 0 0 Proportion negative 41 11 % Positive returns 0% 0% P-value 9,09495E-13*** 0,000976563***

As table 2 and 3 both present a significant difference, there can be assumed that the airline industry does suffer from an aviation disaster reflected in negative returns (Kaplanski & Levy, 2020). This is also supported by the study of Li et al. (2015) which observed a decrease in sales for airline operators and as a result, a drop in the stock prices (Li et al., 2015).

4.2 TESTING THE EFFECT OF AN AVIATION DISASTER ON THE CARRIER

The sequential hypothesis is whether an airplane crash only has an impact on the involved operator during the crash. Again, the author first executed a JB test to be aware of the distribution. The results (presented in the appendices) also confirm a non-normal distribution. To investigate the mentioned hypothesis, the author used both a t-test as another Cowan (1992) sign test. In the tables underneath, the following results are presented:

Table 4 - Significance test H2

CAR (-20,20) CAR (-5,5)

Average -0,034080855 -0,004796871

SD 0,171827576 0,059896912

T-statistic -1,751720915 -0,70729549

Accept/reject Reject*** Accept

Table 5 – Cowan sign test H2 Notes. Where *** is significant at 5% level, ** is _{significant at 1% level and * is significant at 0.5% level}

Notes. Where *** is significant at 5% level, ** is

Table 5 – Cowan sign test H2 CAR (-20,20) CAR (-5,5) Proportion positive 30 11 Proportion negative 11 0 % Positive returns 73% 100% P-value 0,004324005*** 0,000976563***

From the Cowan generalized sign test there can be assumed that there is a significant difference in both event windows and therefore 𝐻_; can be rejected (p=.004324005). However, the first table which represents the t-statistic gives slightly different results within the event window (-5,5). This can be declared by the fact that the data has a non-normal distribution. The Cowan Generalized Sing Test is accounted for this particular distribution and therefore more reliable than a t-test. Nevertheless, the author tested some additional event windows of the Cowan test which accounts for more robustness and therefore makes sure that the results are in line. The additional event windows (-10,10) and (-2,2) are presented in the appendices. The results represented in the additional tables are in line with the conclusions of table 5.

The results are partially corresponding with the reviewed literature. Several studies, as mentioned at the results of hypothesis 1, found that the complete airline industry suffers from an air crash and so is the involved operator. However, multiple studies did only research the complete airline industry instead of next to the entire industry, the carriers only. Conversely, Chance and Ferris (1987) observed that airplane crashes are considered as isolated events as there are found significant consequences for the carrier only.

After defining the second hypothesis, the different results (AAR & CAAR) between the hypotheses are presented in the following table and the corresponding graph:

Table 6 – Market Model

Market Model

_{Hypotheses 1 testing Hypotheses 2 testing}

Event

day AARs CAAR (-20,20) CAAR (-5,5) AARs CAAR (-20,20) CAAR (-5,5)

Notes. Where *** is significant at 5% level, ** is

Graph 1 – CAARs Market Model

The graph gives a clear indication of the returns of the airline industry (H1) as well as the carriers only (H2). As was also clear from the results and the reviewed literature, H1 represents a significant difference with a decline in returns. The second hypothesis (H2), as discussed in the subchapter above, gave at first somewhat unclear results per test. Nonetheless, the author produced some extra tests for increasing robustness. Therefore, there can be concluded that H2 also represents a significant difference. Nonetheless, as can be seen in the p-value, the numbers are slightly less convincing which is reflected as well in the graph.

4.3 TESTING WHETHER THE NUMBER OF FATALITIES OF AN AVIATION DISASTER NEGATIVELY INFLUENCE THE AIRLINE INDUSTRY

Finally, to determine whether only a relatively large amount of casualties harms the return of stock prices, the Mann-Whitney U-test has been used. Since the absolute value of the obtained Z-score is greater than the critical value; 1,96, 𝐻_; has to be rejected. The results are presented in the following table:

Table 7 – Mann-Whitney U-test

5. CONCLUDING REMARKS

Primarily, the research focused on the effect of airplane crashes on the returns of stock prices. Accordingly, Hypothesis 1 formulated that airline operators are greatly suffering from an airplane crash. From both tests executed, there is clear convince to assume there is a significant difference and, therefore, the complete airline industry is suffering from an aviation disaster.

Hypothesis 2 formulated that the negative return of the stock market will only hold for the carrier and not for other airline operators. Since both tests executed gave slightly different results, the author tested two more event windows to account for robustness. These additional event windows are tested via the Cowan since the data has a non-normal distribution. These results show a significant difference and, therefore, the null hypothesis can be rejected. There can be assumed that the carrier itself suffers from ‘their’ aviation disaster.

Finally, to test whether only a relatively large amount of casualties will negatively affect the stock prices, the data is divided into two groups. Subsequently, the Mann-Whitney U-test is executed which results in rejecting the null hypothesis as well. Accordingly, there can be assumed that the amount of casualties does significantly differs to what extent firms are suffering.

To summarize, the market as well as the involved operator suffer from an aviation disaster. This could be explained because investors believed that increasing insurance rates and tighten regulation might be imposed on the total airline industry (Chance & Ferris, 1987). Next to this, the number of fatalities ‘determines’ to what extent the airline industry is suffering. The more life-threatening a crisis is considered to be, the huger effects observed on stock returns (Berkman, Jacobsen, & Lee, 2011).

6. LIMITATIONS

7. FUTURE RESEARCH

Since this study focused on crashes occurred in Western countries, it would also be interesting to investigate whether there arise the same results of crashes within other

continents. For example, should the carrier only also suffer this large from an air crash in poor countries where the media is not playing such a big role as it does in Western countries? Additionally, it would be interesting to test the type of airplane instead of the operator. What if it is turned out that a certain type of airplane has crashed. Did the airline operators who used this type of airplane recognized in a significant decrease in demand? Or is it the total airline industry that suffers from such a technical fault in the type of airplane? As an example, the Boeing 737 MAX that had after launch, two fatal crashes (Cruz & de Oliveira Dias, 2020). Are these two disasters reflected in the return of the stock prices?

REFERENCES

Barnett, A., LoFaso, A. J., 1983. After the crash: the passenger response to the DC-10 disaster. Management Science, 29(11), 1225-1236.

Berkman, H., Jacobsen, B., Lee, J. B., 2011. Time-varying rare disaster risk and stock returns. Journal of Financial Economics, 101(2), 313-332.

Bibel, G., 2008. Beyond the black box: the forensics of airplane crashes. JHU Press.

Carter, D. A., & Simkins, B. J. (2004). The market’s reaction to unexpected, catastrophic events: the case of airline stock returns and the September 11th attacks. The Quarterly

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

Chance, D. M., Ferris, S. P.,1987. The effect of aviation disasters on the air transport

industry: a financial market perspective. Journal of Transport Economics and Policy, 151-165.

Chen, Y. W., Wu, P. Y., Chen, C. W., & Sheu, J. B. (2008). Empirical study of a multi-objective medical decision system for aviation disaster. WSEAS TRANSACTIONS on

Business and Economics, 5, 118-127.

Cruz, B. S., & de Oliveira Dias, M. (2020). CRASHED BOEING 737-MAX: FATALITIES OR MALPRACTICE?. GSJ, 8(1).

Cobb, R. W., Primo, D. M., 2004. The plane truth: Airline crashes, the media, and transportation policy. Brookings Institution Press.

Davidson, W. N., Chandy, P. R., Cross, M. 1987. Large losses, risk management and stock returns in the airline industry. The journal of Risk and Insurance, 54(1), 162-172. Ekeberg, Ø., Fauske, B., & Berg‐Hansen, B. (2014). Norwegian airline passengers are not

more afraid of flying after the terror act of September 11. The flight anxiety, however, is significantly attributed to acts of terrorism. Scandinavian journal of psychology, 55(5), 464-468.

Fiol, C. M., Kovoor-Misra, S., 1997. Part VI: How Should Reputations be Managed in Good Times and Bad Times?: Two-way mirroring: identity and reputation when things go wrong. Corporate Reputation Review, 1(2), 147-151.

Franke, M., & John, F. (2011). What comes next after recession?–Airline industry scenarios and potential end games. Journal of Air Transport Management, 17(1), 19-26.

Gigerenzer, G. (2004). Dread risk, September 11, and fatal traffic accidents. Psychological

Hirshleifer, D., Shumway, T., 2003. Good day sunshine: Stock returns and the weather. The

Journal of Finance, 58(3), 1009-1032.

Ho, J. C., Qiu, M., Tang, X. (2013). Do airlines always suffer from crashes?. Economics Letters, 118(1), 113-117.

Kaplanski, G., and H. Levy, 2010b, Sentiment and stock prices: the case of aviation disasters,

Journal of Financial Economics 95, 174–201.

Kamstra, M. J., Kramer, L. A., Levi, M. D., 2003. Winter blues: A SAD stock market cycle. American Economic Review, 93(1), 324-343.

Keim, M. E., Noji, E., 2011. Emergent use of social media: a new age of opportunity for disaster resilience. American journal of disaster medicine, 6(1), 47-54.

Larsen, J. (Ed.)., 2013. Responding to catastrophic events: consequence management and policies. Springer.

Li, C. W., Phun, V. K., Suzuki, M., Yai, T., 2015. The effects of aviation accidents on public perception toward an airline. Journal of the Eastern Asia Society for Transportation

Studies, 11, 2347-2362.

Lindsay, B. R., 2011. Social media and disasters: Current uses, future options, and policy considerations.

Siomkos, G. J. (2000). Managing airline disasters:: the role of consumer safety perceptions and sense-making. Journal of Air Transport Management, 6(2), 101-108.

Sivak, M., Flannagan, M.J., 2004. Consequences for road traffic fatalities of the reduction in flying following September 11, 2001. Transportation Research 7 (4–5), 301–305.

Slovic, P., Weber, E., 2002. Perception of risk posed by extreme events. Paper presented at the Conference on Risk Management Strategies in an Uncertain World, New York.

Spielvogel, J. J. (2020). Western civilization. Cengage Learning.

Sprecher, C. R., Pertl, M. A., 1983. Large losses, risk management and stock prices. Journal

of Risk and Insurance, 107-117.

Tversky, A., Kahneman, D., 1979. Prospect theory: An analysis of decision under risk. Econometrica, 47(2), 263-291.

Wenger, D. E., 1985. Mass media and disasters.

Zyglidopoulos, S. C., 2001. The impact of accidents on firms’ reputation for social performance. Business & Society, 40(4), 416-441.

APPENDICES

Table 8 – Overview events

Event Date Operator Fatalities Reason of crash

20-11-00 American Airlines 1 Structural failure

1 15-03-01 Vnukovo Airlines 3 Terrorism

2 04-07-01 Vladivostokavia 145 Pilot error - communication

3 29-08-01 Binter Mediterráneo 4 Engine failed

4 18-09-01 AtlantiAirlines 9 Engine failed

5 10-10-01 Flightline 10 Not known

6 12-11-01 American Airlines 260 Pilot error - experience

7 14-12-01 Eagle Air 6 Shot down

8 04-01-02 Agco Corp 5 Pilot error - experience

9 22-02-02 Military - U.S. Army 10 Not known

10 15-04-02 Air China 128 Pilot error - experience

11 07-05-02 China Northern Airlines 112 Passenger sabotage

12 16-07-02 Bristow Helicopters 11 Structural failure

13 02-09-02 CAAC Air TraffiManagement 3 Not known

14 06-11-02 Luxair 20 Engine failed

15 09-12-02 Raytheon Aircraft 3 Pilot error - experience

16 08-01-03 Turkish Airlines (THY) 75 Weather circumstances

17 27-03-03 PT Air Regional 4 Not known

18 01-06-03 Eurojet Italila 2 Structural failure

19 22-06-03 Brit Air 1 Weather circumstances

20 13-07-03 Air Sunshine 2 Structural failure

21 26-08-03 US Airways Express 2 Structural failure

22 19-09-03 Ameristar Jet Charter 1 Pilot error - experience

23 26-10-03 CATA Linea Aerea 5 Structural failure

24 18-05-04 AZAL Cargo Company 7 Not known

25 28-06-04 United Nations - UTair Charter 24 Not known

26 24-08-04 Sibir Airlines 46 Explosive

27 14-10-04 Pinnacle Airlines/Northwest Airlink 2 Engine failed

28 21-11-04 China Eastern Airlines 53 Weather circumstances

29 16-03-05 Regional Airlines 28 Pilot error - communication

30 07-05-05 Aero-Tropics 15 Pilot error - experience

31 02-08-05 Air France 0 Weather circumstances

32 08-12-05 Southwest Airlines 0 Combination of factors

33 24-03-06 ATESA 5 Not known

34 23-04-06 Yug Avia 4 Lack of Safety instruction

35 02-07-06 Air hamburg 5 Engine failed

36 22-08-06 Pulkovo Airlines 170 Pilot error - experience

38 12-01-07 SunQuest Executive Air Charter 2 Pilot error - experience

39 07-02-07 Metro Aviation 3 Not known

40 17-03-07 UTAir Airlines 6 Pilot error - communication

41 24-07-07 Taquan Air Service 5 Pilot error - experience

42 16-08-07 SeaWind Aviation 6 Pilot error - experience

43 30-11-07 Atlasjet Airlines 57 Pilot error - experience

44 05-01-08 Servant Air 6 Structural failure

45 15-03-08 Trade Wings Aviation Ltd. 3 Not known

46 17-05-08 Chelan Air Service 2 Structural failure

47 31-07-08 East Coast Jets 8 Not known

48 14-09-08 Aeroflot 88 Pilot error - experience

49 15-01-09 US Airways 0 Engine failed

50 12-02-09 Colgan Air 49 Pilot error - experience

51 01-06-09 Air France 228 Weather circumstances

Table 9 - Summary statistics

Number of events 51

Average fatalities 33

Average crashes per year 5,6

Max. Number of fatalities 260

Min. Number of fatalities 0

Median 6

Large group casualties 22

Small group casualties 29

Table 10 – Jarque-Bera Normality test (JB) H2

CAAR (-20,20) CAAR (-5,5)

Kurtosis -0,823574323 10,58260163

Skewness -0,217502257 3,195650647

JB Statistic 31,46901776 208,9822317

Conclusion Reject - No normal distribution Reject - No normal distribution

Table 11 – Cowan sign test H2

CAR (-10,10) CAR (-2,2)

Proportion positive 20 5

Proportion negative 1 0

% Positive returns 95% 100%

P-value 2,09808E-05*** 0,0625** Notes. Where *** is significant at 5% level, ** is