The impact on stock returns of an airline following an aviation
disaster
Thesis supervisor: A. Pua
Veronique Ackerman 10329153
Bachelor’s Thesis Economics July 15, 2015
Statement of Originality
This document is written by Student Veronique Ackerman who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
Abstract
This study examines the financial effect on returns of airlines after an aviation disaster. The focus of this paper will solely be on the financial impact on the returns of the respective airlines following an aviation disaster. This is tested between March 1977 and March 2015 with a sample of 30 aviation disasters involving airplanes of commercial airlines which are, at the time of the incident, publicly traded. To measure the abnormal returns of airlines after these disasters, an event study methodology is used. The tests performed were based on the guidelines of the models of Campbell, Lo and Mackinlay(1997). In addition, two different sources are used as reference for the calculation of the abnormal returns. These are the broad-based stock market index and the aviation industry of the country where the airline is located. The results obtained suggest that airlines experience a stock return decline of 2,0% on the day of the disaster. This abnormally return does not continue beyond the day of the crash.
Table of contents
1. Introduction 4
2. Literature review 5
3. Data collection and description 6
4. Methodology 7
4.1 Hypothesis
4.2 Event study methodology
5. Results and discussion 13
5.1 Broad-based stock market 15
5.2 Aviation industry 16 5.3 Discussion 16 6. Conclusion 17 6.1 Conclusion 6.2 Discussion Bibliography 19 Literature Websites Appendix A 21 Appendix B 22 Appendix C 23
1. Introduction
Nowadays, flying proves to be the safest way to travel according to the National Safety Council. However, if an aviation accident occurs, it can cost the lives of humans. In addition, it might affect the financial performance of the airplane manufacturers, the air industry and the airplanes involved. The focus of this study will solely be on financial performance of the airlines involved. This study uses ‘event study’ methodology to measure this possible negative financial impact on the returns. This kind of methodology is widely used in studies that evaluate the effects of unanticipated events on the stock performance of companies.
An accurate measure of the financial consequences of an aviation disaster is to examine how the prices of the company’s common stock reacts. The stock prices plunge to reflect the decreases in the airlines’ expected future cash flows. The main requirement of event studies is that the stock market is efficient. That is, the stock prices react quickly and accurately to new information. Previous research by Fama(1970) concluded that the stock market is quite efficient if the event is relevant to the wealth of shareholders. This means that the effects of an event will immediately be reflected in stock prices.
In addition, this study also includes the re-investment of dividends. Thus by looking at the total return index, the price index performance is augmented by the performance impact due to dividend. In the research of Agrrawal and Bordman (2010) the authors suggest that the total return gives more accurate information for the investors instead of looking at only the relative price changes of stocks.
In order to answer the research question, this will be tested by conducting event studies with various time intervals within an event window and performing several t-tests to see if the negative abnormal returns are significantly different from zero.
This paper proceeds as follows. It continues with prior related literature in section 2. Section 3 explains the method of collection of the data and also describes the collected data. Section 4 explains the event study methodology. In addition, it will describe the hypothesis of the paper. Subsequently, section 5 reports the event study results and discusses them. Finally, section 6 concludes this paper.
2. Literature review
Several previous studies examine the stock market reaction to commercial air crashes. Chance and Ferris(1987) conducted a study about the financial impact on stock prices to U.S. commercial airline crashes, the airline-industry and airplane manufacturers. For the period of Feb 1963 until July 1985, they investigated a sample of 49 events. They found that the stock market for the carrier involved reacts significant negatively to the news of an airplane crash within one trading day. This does not continue following the day of the crash. Contrary to the involved carriers, the authors did not found a negatively significant stock price reaction for the airline-industry and airplane manufacturers.
Another research that indicates the effect of a negative reaction on stock prices to air crashes is the study of Davidson, Chandy and Cross(1987). The sample includes 57 crashes belonging to 22 firms and the time period they selected for the study was 1965 to 1984. Subsequently, they partitioned the crashes into two subsamples. The first subsample included 30 crashes which had fatalities of ten or more people for 11 airlines. The second subsample included only the worst crash of each of the 11 airlines in the first subsample. They only found statistically significant evidence for the second subsample on the day that the crash occurred. Hence, the most severe crashes with the largest number of fatalities had a
significant negative reaction on the return of the airlines on the day of the crash. Similarly to Chance and Ferris(1987), these returns reversed itself within the days following the crash.
Barett, Heuson, Kolb and Schropp(1987) also attempted to examine if crashes resulting in the highest number of fatalities, as well as less severe crashes had an financial impact on the return of the involved carrier. In contrast with the research of Davidson et al.(1987), they found for both subsamples an average market decline for carries stock prices on the day of the crash, which is statistically significant. The results are more potent for the most severe crashes.
In the research of Kaplanski and Levy(2010), the authors found that an airplane crash is followed by significantly negative return in the stock market accompanied by a reversal effect two days later.
Thiengtham, Yi Lin and Walker(2005) provide results for the financial impact of airline crashes on short- and long-term performance of airlines, as well as airplane
manufacturers. Their research consist of 138 aviation disasters involving airplanes of publicly traded U.S. carries in the time period 1962 to 2003. Walker et al. observe a significant price reaction for both airlines involved and airplane manufacturers following the announcement of
an aviation disaster. The cumulative abnormal returns remain significant negative until six months after the crash for the airline carrier. This is not in line with the study of Davidson et al.(1987) and Baret et al.(1987) and Chance and Ferris(1987) who found a significantly negative return only the day of the aviation disaster. Walker et al. (2005) also found that the magnitude for airplane manufacturers is relatively smaller during the first week following the disaster. Contrary to the airline carrier, the airplane manufacturer experiences a significant positive stock price reaction within six months of a disaster.
It is clear that there are similar studies who examine the financial impact on stock prices following an aviation disaster. In sum, scientific literature finds consensus on the fact that, in general, the stock market imposes value losses on firms on the day of the aviation
disaster((Davidson et al.,1987), (Chance and Ferris,1987) and (Thiengtham et al, 2005)). However, scientific literature differs when it comes to evaluating the duration of the negative airline’s financial performance after accidents. Corresponding with the above-mentioned studies, this study provides complementary evidence for a possible confirmation or denial of the statistically negative financial impact of the stock return of the involved airline carrier following an aviation disaster. The focus of this study will solely be on the airlines involved. In addition, in comparison with other studies this study will use more recent data. Hence, this study is more up-to-date. Moreover, relatively little research of the same nature has been conducted on worldwide markets. Accordingly, this paper seeks to fill this gap by focusing on the worldwide market.
3. Data collection and description
For the event study conducted in this paper datasets were collected using ‘Datastream’. This is a financial database which contains company data, equities and macro-economic data. The market data and company data is updated daily and was directly installed to excel by using Datastream through Microsoft Excel. Because data before 1975 are not available on Datastream, airplane disasters that occurred prior to 1975 are excluded. Additionally, the stock markets are closed on national holidays and weekends. Hence, the datasets includes only information on the remaining trading days per week.
Detailed information regarding the specific aviation crash has been gathered from two websites. The website published by the National Transportation Safety Board
(www.NTSB.gov) provides detailed information on airline crashes dating back to 1935. To complete the dataset for countries outside the United States www.aviation-safety.net was
used. For each aviation disaster are recorded the date of the crash, the name of the airline, the country of the airline involved and the number of fatalities in the ground and in the air.
Only the events that caused at least one fatality, in the air or on the ground, were included. This arbitrarily chosen amount is to ensure that the study would contain only events of possible significant economic impact. In addition, only commercial airlines were included which are, at the time of the incident, publicly traded.
The dataset includes information on 30 airplane disasters owned by publicly traded
commercial airlines between March 27, 1977 and March 24, 2015. All incidents are listed in the table A in the Appendix A.The analysis has been carried out in Excel, at a significance level of 1%, 5% and 10%.
Because there are two event windows which overlap in calendar time, three examinations has to be done to indicate if the two event windows which overlap show too many multicollinearity. According to Carter Hill et al.(2012), multicollinearity arises when events are too strongly correlated with each other. Hence, the results will affect each other. This will result in an undefined least squares estimator. To prevent multicollinearity, the two companies (the eleventh and twenty-ninth company listed on the table A in Appendix A) have to be dropped in order for the estimation to be correct. Hence, the first examination will consist of companies excluding the two companies of which the event windows overlap. The second examination will consist all companies excluding the eleventh ‘overlapping’
company. Finally, the last examination will consist all companies excluding the twenty-ninth ‘overlapping’ company. Test 2 and test 3 are done to see if they might show multicollinearity. Each test will be carried out of the same methodology.
4. Methodology
In this chapter, the methodology used for testing abnormal returns following an aviation disaster is explained. Hypothesis will be tested in this paper to arrive at conclusions of the financial performance of involved airline companies following an aviation disaster. The hypothesis is listed in the following part of this section before the methodology is further explained.
4.1 Hypothesis
Before analyzing the results of this study, the following hypothesis will aid in answering the research question of this paper:
The null hypothesis, H0, assumes that an aviation disaster has no financial impact on the mean or variance of the returns.
The alternative hypothesis, H1, assumes that there is a negative financial impact on the mean or variance of the returns after an aviation disaster.
Hence, the financial impact of the accidents on returns will be tested. Testing whether the results obtained are statistically significant form zero for the three tests is done using a Student’s t-test.
4.2 Event study methodology
To analyze the returns occurring after the aviation disaster an event study will be conducted. This event study methodology measures whether abnormal returns are statistical significant for the airline involved after an airplane crash for specific time intervals. To test this, this study will follow the guidelines given by Campbell et al.(1997).
The abnormal return is calculated by subtracting the normal return of the airline over the event window from its actual ex post return of the airline over the event window. Event study methodology possibly demonstrates that the drop in the stock return is related to the
announcement of the aviation disaster instead of an unrelated market or industry factor or due to random fluctuation in stock prices. Hence, the change in the value of the firm is caused by the (exogenous) event.
For each security s and event date τ:
Єst* = Rst – E[Rst | Xt] (1)
where
єst*= abnormal return for time period t
Rst = actual return for time period t
Rs,t is the is the return on investment, including interest payments, as well as appreciation or
depreciation in the price of the stock s at time t.
The formula used to calculate the return index is as follows:
RIt = RIt-1 * . (2) where: RI = return index P = clean price A = accrued interest
NC = next coupon. Adjustment made when a stock goes ex-dividend. CP = value of any coupon received on t or since t-1
t = time
In Datastream this datatype is called ‘RI’. Subsequently the total return has to be calculated as follows:
Rt = (RIt – RIt-1) / RIt-1 (3)
where:
R = total return.
Abnormal returns are measured in the period 20 trading days prior to the event and ends 20 trading days after the event to find out how long it takes for the market to fully incorporate the new information. Hence, this study uses a short-, as well as mediocre length intervals within the event window. The event window is common in similar event studies; it gives sufficient data from which causal patterns can be revealed that might come from the disaster.
First, the normal return, E(Rst), has to be estimated. The normal return is the return of
a stock as if the event had not happened. This can be estimated using several benchmark models. Here, the market model is used.
4.2.1 Market model
The market model assumes a stable relation between the company stock return and the market return. According Campbell et al.(1997) represents the market model improvement over other statistical models because the variance of the abnormal return in reduced. For any security with a specific stock s we have:
Rs,t = αi + βi Rm,t + us,t (4)
E[us,t] = 0 var[us,t]= σ2єi
where the subscripts s, t and m indicate respectively a specific stock, time and market. Additionaly, us,t is the random error term for stock s at time t and αi ,βi and σ2єi are the
parameters to be estimated by the market model. The beta is the measure of correlation between the industry index and the stock.
According to Campbell et al.(1997), it is conventional to assume that the returns are ‘jointly multivariate normal’ and ‘independently and identically distributed’ through time. This assumption is sufficient for a market model. The alpha-and beta-values are then
calculated by regressing an airline return on the return of the market portfolio using ordinary least squares (OLS).
To estimate the parameters of the market model a subset of data has to be used known as the ‘estimation window’. According to Campbell et al.(1997), it is conventional using daily data and the market model to calculate the parameters for the normal returns for the event window using data from more than 100 trading days prior to the event. This is a long enough event window, because the crash cannot be predicted by the market. To prevent an influence on the parameters of the normal return model, the event period itself is not
included, resulting in an estimation window’ of 119- until 21 trading days prior to the event for the different intervals within the event window of (-20, +20).
For this study, there are two different market indices collected to proxy for the
market. The first one is the broad-based stock index for the country of the involved company. The second one is the equity index of the whole airline-industry in the country of the
involved company.
The definition of some notation is as follows. In this study the event date is the date on which the aviation crash took place. The τ = 0 represents the event date. This is under the
assumption that the announcement of the crash is at the same day as the crash (τ = 0). If the event date was absent in the collected data, the date prior to the event date was picked as the ‘event date’. T1 represents 20 trading days before the event and T2 = 20 trading days after the
event. T0 represents 119 trading days before the event. Hence, T1 to T2 represents the event
window and T0 to (T1 -1) As in similar studies, this study does not make use of a post-event
window in which they investigate the longer term company performance following the event. The length of the estimation window(L1) is represented by L1 = (T1 -1) – (T0 -1) and the
length of the event window by L2 = (T2 + 1)– T1. This results in L1 = 99 and L2 = 41. The
time line is illustrated in figure 1. The intervals has to be within the event window. Hence, T1
≤ t1 < t2 ≤ T2.
Figure 1. The time-line for this event study
4.2.2. Abnormal returns
To estimate the abnormal returns Єst* in the event window, an implementation of equation (4)
in equation (1) yields:
Єst* = Rst – E[αi + βi Rm,t + єs,t | Xt] ,
Est* = rst – ( i + i Rm,t) (5)
As mentioned before, ordinary least squares (OLS) is under general conditions an unbiased and efficient estimation procedure for the market model parameters. The parameters and i are then estimates of the true parameters of the estimation window. A condition for
the market return over the event window is that the abnormal returns will be jointly normally distributed with an expectation of zero and a covariance matrix(Vi) which consists of two
parts. The first parts consists of the variance due to the future disturbances. The last part is the additional variance due to the sampling error. Because of the fact that the estimation window is large the second term will approach zero. Hence, we have the distribution for any abnormal return observation within the event window:
Est* N(0,Vs) (6)
The abnormal returns have to be aggregated over time and across securities. To get the Cumulative Abnormal Returns(CAR), they have to be aggregated over time. This is the measure of the total abnormal returns during the specific interval of the event period. Under the assumption that the returns on each day are independent, the standard errors are
cumulative. For the corresponding number of trading days in the event window, the following equations describe respectively the firms CAR and the variance of the cumulative abnormal returns with t1 being the first time variable of the interval and t2 being the second within the
event period:
s(t1,t2) = st* (7)
Var( s(t1,t2)) = σs2(t1,t2) = γVsγ
where:
γ = a vector (L2 x 1) with ones in positions t1 to t2 and zeros elsewhere in the event period.
According to equation (6) we have:
s(t1,t2) N(0, s2(t1,t2))
Subsequently, the returns have to be aggregated across firms.
* = st* (8)
var( *)= V = s
To get one test statistic for hypothesis, this average abnormal returns vector have to be aggregated over time by using the same approach as for the individual airline’s vector.
Hence:
(t1,t2)) = * = ( s(t1,t2) ) (9)
var( (t1,t2)) = 2(t1,t2) = γVγ ( = s2(t1,t2) )
Assuming that the covariance tends to zero:
A Student t-test has to be done to examine if the CARS are significantly different from zero on any given day during the event window. This is to note if there is significantly market response to the stock prices of airline companies. This is on the form:
t =
~
t(α, df = N – 1) (11)
A one-tailed t-test is appropriate, because the alternative hypothesis, H1, assumes that the
return is negative.
5. Results and discussion
This section will represent the main results of the event study methodology by Campbell et al.(1997) depicted in the previous section. The empirical tests should be able to assess the validity of the hypothesis presented in section three. As described in section two, similar earlier studies have concluded a significant abnormal return the day of the announcement of the aviation disaster, but are inconsistent the days following the announcement. In line with similar studies, the event study approach in section four will test if such abnormal return exists for particular intervals within the event window in the sample used in this study.
As described earlier, two out of thirty events of this sample overlap in their event window. These events are illustrated in Appendix A. To not immediately exclude these ‘overlapping’ firms and see if these ‘overlapping’ companies will show multicollinearity in their results, three test have to be done. The first test includes firms whose event window do not overlap (N=28). The second test will include all firms except the eleventh event of the table A shown in Appendix A (N=29). The last test will include all firms except the twenty-ninth event of the table A shown in the Appendix A (N=29). In addition, to test if the event has significant negative performance on the day of the crash, the interval [-1,0] is used. However, for some firms (nine out of thirty, including the eleventh ‘overlapping’ company) the stock market is closed on the day of the disaster. These firms are excluded for the event window [-1,0], because these samples do not reveal any causal patterns on the day of the disaster. Hence, test 1 consists of N=20, test 2 consists of N=21 and test 3 consists of N=20 for the time period [-1,0]. For the other intervals whose stock market is closed for the firm on the day of the announcement, one day before the disaster is used as the event date. For the event window [-1;0], the critical values for test 1, test 2 and test 3 are 1.729, 1.725 and 1.729
respectively for a significance level of 5%. For the other symmetric intervals [-1;+1], [-4;+4] and [-19;+19] the critical values for test 1, test 2 and test 3 are 1.703, 1.701 and 1.701
respectively for a significance level of 5%. In sum, table 1 presents the number of observations with their critical t-value for the time period [-1;0].
table 1: the number of observations along with their critical t- values for an one-tailed test for the event window
[-1;0]
#observartions critical t- value critical t- value critical t- value
1% level 5% level 10% level
Test 1 20 2.539 1.729 1.328
Test 2 21 2.528 1.725 1.325
Test 3 20 2.539 1.729 1.328
Additionaly,table 2presents the number of observations tested for the intervals [-1;1],[-4;+4] and [-20;+20].
Table 2: the number of observations along with their critical t- values for an one-tailed test for the event window
[-1;1], ],[-4;+4] and [-20;20]
#observartions critical t- value critical t- value critical t- value
1% level 5% level 10% level
Test 1 28 2.473 1.703 1.314
Test 2 29 2.467 1.701 1.313
Test 3 29 2.467 1.701 1.313
The start of the analysis will consist of the three tests described in section three for four different intervals within the event window, by which the normal return is calculated using the broad-based stock index for the country of the involved company as the market return for calculating the alpha- and betaparameters for the normal return in the estimation period by using the market model described earlier. The second part of this section will consist of the three tests for four different intervals by which the normal return is calculated using the aviation industry as a whole of the specific country as reference. The last part of this section will give a discussion of the results of two different approaches. In addition, the three tests will be considered.
5.1 Broad-based stock index
In appendix B the graphics of the abnormal returns are showns for each of the three tests. The graphs show that the abnormal returns immediately fall on the day of the crash. Thereafter, the abnormal returns seem to reverse.
In addition, table 3 provides the results for the four different intervals within the event window. The first interval is one day before the event till the day of the event. Then, the days are increased to symmetric intervals using the theory of Campbell et al.(1997) to show possible impacts over the days. For the different intervals, cumulative abnormal returns (CARs) are shown for commercial airlines whose airplanes were involved in a disaster during the sample period. For calculating the normal return, the alphas and betas were calculated using the broad-based stock index as the market return. The second column of each test provides test statistics for a one-tailed t-test.
Table 3: Abnormal performance for the airline companies following an aviation accident.
Abn. returns Compared to the broad-based stock index used as Rmkt
Test 1 (N=28) Test 2 (N=29) Test 3 (N=29)
interval CAR t-test CAR t-test CAR t-test
-1;0 -0,0203772 -2,522** -0,0220262 -2,679*** -0,02038 -2,522** -1;+1 -0,007261 -0,925 -0,0094499 -1,177 -0,00735 -0,961 -4;+4 -0,0032707 -0,234 -0,0048976 -0,341 -0,00329 -0,241 -20;+20 -0,0010604 -0,032 -0,0017919 -0,052 -0,00111 -0,034
Table 3. Using a one-sided test : *** = statistical significance at the 1% ** = statistical significance at the 5% * = statistical significance at the 10%
As shown in table 3, the airlines in the sample experience an average return decline of 2,038% during the trading day of the event for test 1. For test 2 and test 3 these declines are 2,202% and 2,038% respectively.
In order to answer the research question of this paper, it is necessary to compare the found t-test values with the critical values at the 1%, 5% and 10% levels in. The significance are shown by the stars. Only the day of the crash shows a significance impact on the returns for a 5% significance level. In addition, test 2 is also significant at the 1% level at this interval. For the other intervals, the return reactions are economically and statistically insignificant.
5.2 Aviation industry
In appendix C the graphics of the abnormal returns are showns for each of the three tests. Table 4 depicts the same approach for the four different intervals within the event window as part 1(5.1) of this section, but now the aviation industry for the country of the airline involved is used (instead of the broad-based stock market index) to calculate the parameters for the normal returns.
Table 4: Abnormal performance for the airline companies following an aviation accident.
Abn. returns Compared to aviation industry of the specific country Test 1 Test 2 Test 3
interval CAR t-test CAR t-test CAR t-test
-1;0 -0,0062148 -1,150 -0,0093076 -1,594* -0,00621 -1,150 -1;+1 -0,0043461 -0,801 -0,0065546 -1,133 -0,00435 -0,829 -4;+4 -0,0019307 -0,199 -0,0034265 -0,331 -0,00193 -0,206 -20;+20 -0,0008571 -0,037 -0,0014624 -0,059 -0,00086 -0,038
Table 4. Using a one-sided test : *** = statistical significance at the 1% ** = statistical significance at the 5% * = statistical significance at the 10%
As can be seen from table 4, the airlines in the sample experience an average return decline of 0,621%, 0.931% and 0,621% for test 1, test 2 and test 3 respectively on the day of the
aviation disaster. Only the t-value of test 2 experiences a significant negative impact on the returns of the day of the crash at the 10% significance level.
5.3 Discussion
In this part, the results will be discussed.
Test 2 results in both parts (in the ‘broad-based stock index’-part, as well as in the ‘aviation industry’-part) in a higher significance level compared to test 3. Test 2 excludes the nineteenth ‘overlapping’ company and includes the twenty-ninth ‘overlapping’ company. For test 3 it is vice versa. This follows the intuition that the crash with the highest death rate will have a higher significant impact on the performance of the airline involved. The nineteenth crash, as listed in Appendix A, is a crash that resulted in the death of 2 out of 142 people. Hence, 1.4% of the persons on the airplane lost their lives. The twenty-ninth crash was a more severe crash, which resulted in the death of 230 out of 230 people. This results in a
death rate of 100%. Hence, in line with the intuition, test 2 gives a more significant t-test value compared with test 3.
The results reflect that the ‘aviation industry’-part is less significant for all tests on the intervals [-1;0], [-1;+1] and [-4;+4] compared with the ‘broad-based stock market’-part. A possible reason for this is that the aviation industry for the specific country is also hit by a crash. There might be negative spillover effects to the airline industry. Thus, this is a industry wide effect. Hence, the effect is less if you take the airline industry as reference. Therefore, the based stock market index is a better reference. Because of the fact that the broad-based stock market index is a better reference, only part 1 is chosen to draw the conclusion.
6 Conclusion and discussion
6.1 Conclusion
The majority of previous work established that the abnormal returns immediately fall only on the day of the crash.
This paper examines the financial impact on the days following an aviation disaster. This study affirms that an aviation disaster leads to a significant negative return effect on the day of the disaster. These findings are also in agreement with the hypothesis of Fama(1999) who states that the stock market is efficient. This abnormally return does not continue beyond the day of the crash.
6.2 Discussion
Future studies can differentiate between the crashes and the airlines. To detail the effect of a disaster, different characters can be added. To not shift away from the actual research
question, these details are excluded. But including these details might offer policy makers to provide proper safety management policies.
This study gives a strong believe that the severity of the crash, thus the death-rate, plays a significant role in the impact of the negative returns following a crash. This is an interesting area for future research.
This study has several benefits compared to earlier studies. As mentioned, this study is more up-to-date and uses a world-wide focus. In addition, previous research used the price index change to calculate returns, this thesis used the datatype ‘return index’ to calculate
returns. This has the benefit that it includes the re-investment of dividends, which is usually considered as a more accurate measure of perfomance.
However, this study has several limitations. The first limitation is the sample size of the thesis. For example, Chance and Ferris (1987) used a sample of 49 events and the
Davidson, et al.(1987) used a sample 57 events in their research. Thus compared with earlier studies, the sample size of 30 observations is below average.
The second limitation is that Datastream contains only data of quite large international companies. Hence, the effect is still unknown for small regional airlines. If data can be
obtained from smaller regional airlines, this could be further examined in future studies. Last, following intuition is not enough. Hence, the negative spillover has to be examined in the aviation industry.
Bibliography 1. Literature
Agrrawal, P., & Borgman, R. (2010). What is wrong with this picture? A problem with comparative return plots on finance websites and a bias against income generating assets. The Journal of Behavioral Finance, 11(4), 195-210 .
Barrett, W.B., Heuson, A.J., Kolb, R.W., Schropp, G.H., (1987). The adjustment of stock prices to completely unanticipated events. Financ. Rev. 22, 345-354.
Campbell, J., Lo, A., MacKinlay, A., & Whitelaw, R. (1998). The econometrics of financial markets. Macroeconomic Dynamics, 2(4), 559-562.
Carter Hill,R., Griffiths,William E., Lim, Guay C. (2010), Principles of Econometrics, 4th edition, John Wiley & Sons.
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, 151-165.
Davidson ΙΙΙ, W.N., Chandy, P.R., and Cross, M. (1987). Large losses, risk management and stock returns in the airline industry. The Journal of Risk and Insurance, 54, 162-172.
Fama, E.F. (1970). Efficient capital markets: A review of theory and empirical work, Journal of Finance, 25, 383-417.
Flouris, T., and Walker, T.J. (2005). The financial performance of low-cost and fullservice Airlines in times of crisis. Canadian Journal of Administrative Sciences, 22, (1), 3-20.
Lin, M.Y., Thiengtham, D.J., and Walker, T.J. (2005). On the performance of airlines and airplane manufacturers following aviation disasters. Canadian Journal of Administrative Sciences, 22, (1), 21-34.
Kaplanski, G., & Levy, H. (2010). Sentiment and stock prices: The case of aviation disasters. Journal of Financial Economics, 95(2), 174-201.
2. Websites
Aviation safety network: www.aviation-safety.net
National Safety Council: http://www.nsc.org/pages/home.aspx
Appendix A:
Table A:Overview of the sample
* The event windows of these two air crashes overlap. Number Date accident (DD/MM/YYY Y) Airline Country Fatalities/ (passengers+crew) (on ground) 1 14/09/2008 Aeroflot Russia 88/88(0)
2 20/04/1998 Air france-KLM France 53/53(0)
3 25/07/2000 Air france-KLM France 109/109(4)
4 26/06/1988 Air france-KLM France 3/136(0)
5 01/06/2009 Air france-KLM France 228/228(0)
6 16/02/1998 China airlines China 196/196(7)
7 25/05/2002 China airlines China 225/225(0)
8 21/11/2004 China Eastern Airlines China 53/53(0)
9 15/11/1987 Continental Airlines United States 28/82(0)
10 02/08/1985 Delta Airlines United States 134/163(1)
11* 06/07/1996 Delta Airlines United States 2/142(0)
12 31/08/1988 Delta Airlines United States 14/108(0)
13 29/09/2006 Gol Linhas Aereas Intelligentes Brazil 154/161(0)
14 04/04/1994 KLM Netherlands 3/24(0)
15 27/03/1977 KLM Netherlands 583/644(0)
16 06/08/1997 Korean Air Lines Korea 228/254(0)
17 24/03/2015 Lufthansa Germany 150/150(0)
18 17/07/2014 Malaysia airlines Malaysia 298/298(0)
19 15/09/1995 Malaysia airlines Malaysia 34/53(0)
20 08/03/2014 Malaysia airlines Malaysia 239/239(0)
21 10/07/2006 Pakistan International Airlines Pakistan 45/45(0)
22 08/10/2001 Scandinavian Airlines Sweden 114/114(4)
23 13/03/1986 Simmons airlines United States 3/9(0)
24 31/10/2000 Singapore Airlines Singapore 83/179(0)
25 08/12/2005 Southwest Airlines United States 0/103(1)
26 02/09/1998 Swissair Swiss 229/229(0)
27 03/03/2001 Thai Airways Thailand 1/5(0)
28 11/12/1998 Thai Airways Thailand 102/146(0)
29* 17/07/1996 Trans World air United States 230/230(0)
30 25/02/2009 Turk Hava Yollari (Turkish
Airlines)
Appendix B
Part 1, section 5: used the Broad-based stock market index for the calculation of the normal return. Graph 1. Test 1 Graph 2. Test 2 -0,04 -0,03 -0,02 -0,01 0 0,01 0,02 -30 -20 -10 0 10 20 30
av. Abnormal returns
av. Abnormal returns
-0,04 -0,03 -0,02 -0,01 0 0,01 0,02 -30 -20 -10 0 10 20 30
av. Abnormal returns
Graph 3. Test 3
Appendix C
Part 2, section 5: used the aviation industry for the calculation of the normal return.
Graph 4. Test 1 -0,04 -0,03 -0,02 -0,01 0 0,01 0,02 -30 -20 -10 0 10 20 30
av abnormal returns
av abnormal returns -0,015000 -0,010000 -0,005000 0,000000 0,005000 0,010000 -30 -20 -10 0 10 20 30av. Abnormal returns
Graph 5. Test 2 Graph 6. Test 3 -0,020000 -0,015000 -0,010000 -0,005000 0,000000 0,005000 0,010000 -30 -20 -10 0 10 20 30
av. Abnormal returns
av. Abnormal returns
-0,015000 -0,010000 -0,005000 0,000000 0,005000 0,010000 -30 -20 -10 0 10 20 30
