Analysis of Current Issues in Finance
Mohammed H. Khan
(08178887)
University of Hertfordshire
Event Study
Aut vincere aut mori
The impact of certain types of firm-specific events e.g. earnings announcements on the
prices of the affected firms’ securities has been the subject of a number of studies. A major
concern in those ‘event’ studies has been to assess the extent to which security price
performance around the time of the event has been abnormal -- that is, the extent to which
security returns were different from those which would have been appropriate, given the
model determining expected returns. Event studies provide a direct test of market efficiency,
abnormal returns which occur after or before an event are inconsistent with the notion the
security prices adjust quickly to reflect new information. The efficient market hypothesis
states that a market cannot be outperformed because all available information is already built
into all stock prices. (Investopedia: 2011)
This paper begins by explaining what an event study is, it goes on to describe how one is
conducted. The event study here is based on 13th October 2008, where the British
Government injected capital into Royal Bank of Scotland who was on the brink of
insolvency. Penultimately, before concluding results of the event study is analysed.
Investorwords (2011) defines an event study as the analysis of the present or future impact
of a particular news story or significant event related to a firm or financial market. An event
study attempts to determine whether that event already has, or will have a statistically
significant effect on the firm such that it affects its financial performance; more specifically
the company’s share price
The first step to conducting an event study is selecting data according to the event of interest.
The data will then be divided into two sub periods: estimation window and event window.
The estimation window is defined as the periods without the effects of the event, and the
event window as periods that include the effects of the event. (Allen: 2009)
The period of the estimation window is “determined to capture common movements of the
stock or when no relevant event occurs”. The length of the estimation window should be long
so as to represent more data resulting in a more accurate regression, but it should not be too
distant from the event window as it may result “in model parameter instability e.g. because
data with lower explanatory power and a higher probability of confounding events “ (Brown
& Warner: 1985 & Muntermann:2007). Weil [Link]. (2001) state the further the estimation
window stretches from the event window “the less the estimated relation between the stock
price and the market index is likely to represent the underlying relation during the event
window.” Muntermann (2007) suggests the selection of the length of the estimation window
should be shorter if one conducts daily studies instead of monthly studies.
The sample period will be from 2st July 2008 – 20th October 2008 (110 days), with the
estimation window starting on 2th July 2008 – 6th October [Link] event window will start
be 7 days before the actual event and end 7 days after, Hirschheim [Link] (2009) suggests using
an event window consisting of 2 days, however acknowledges that due to insider information
and information leakages and rumours, it is appropriate to have an event window a few days
before the actual event but should not be too many days before to prevent confounding
effects which may bias the analysis.
The share price for Royal Bank of Scotland and the FTSE 100 was collected beginning 2st
July 2008 – 20th October 2008. The FTSE100 is the index selected which is used to control
for outside influences on RBS share price, it was selected based on the criteria that it is based
on listed comparable companies which have similar market capitalisation. (Weil .et al : 2001)
Event Window
Daily data was used in this study, whereas Brown & Warners (1980s) used monthly data.
Shane & Spicer (1983) infer using daily data should not adversely affect the significance of
discovering abnormal return.
The model used for calculating expected returns is the mean adjusted returns. Brown &
Warner (1980) advocate the use of mean adjusted returns as it performs “as well or as better
than more sophisticated” models such as the market adjusted return model and market and
risk adjusted return model. (Shane & Spicer: 1983) However, “for completeness all tests
conducted here were also performed using risk and market adjusted returns”. (Shane &
Spicer: 1983)
The expected return derived from each model will be subtracted from the actual return
giving abnormal return. This is done as we want to assess the extent to which security price
performance around the time of the event has been abnormal i.e. the extent to which security
returns were different from those which would have been appropriate, given the model
determining expected returns (Brown & Warner: 1980).
The three models used to generate the expected return are mean adjusted return, market
adjusted return and market and risk adjusted return.
In the mean adjusted return method, the average return for the return is calculated:
Where n is the number of observations, t is event date and i is firm. The resultant average
return is the expected return which is then used in the calculation of abnormal return. Them
mean adjusted returns model is consistent with the Capital Asset Pricing Model; under the
assumption that a security has constant systematic risk and that the efficient frontier1 is
stationary2, the asset pricing model also predicts that a security’s expected return is constant.
(Brown & Warner:1980).
The second model the market adjusted returns suggests expected return is equal to the
market return for that period. That is:
ARit = Rit – Rmt
Rit for a stock is the percent change in the stock price at time t, , where i denotes the event
(i=1,2,…N), m denotes the market, and t denotes the day of the event (e.g., t = 0 denotes the
day of the vulnerability announcement.). ARit denotes the abnormal return of event i at time t,
Rit denotes the actual return and Rmt denotes the market return at time period t.
(Acquisti .et al : 2006)
The market and risk adjusted model states the return on an asset is determined by a constant
and the return on the market index, that is:
Rit =α +βRmt +ut
Rit is the return to an asset i
α is a constant
β is a slope parameter
Rmt is the return to a market index
ut is an error term
Where α and β are calculated using ordinary least squares.
In this model the expected firm return is a linear function of the market return using an OLS
beta.. (Dykman [Link] : 1984
1
Efficient frontier is the optimal portfolios plotted along the curve have the highest expected return possible for the given amount of risk.
(Investopedia:2011)
2
in Time Series analysis, a stationary series has a constant mean, variance, and autocorrelation through time.(Statistics glossary: 2011)
Once, abnormal return is worked out the standard error is estimated as the standard deviation
of abnormal returns from estimation window. The standard error measures the accuracy with
which a sample represents a population; usually the smaller the standard error, the more
representative the sample will be of the overall population. (Investopedia: 2011)
After calculating standard error, a t statistic3 is calculated for the event window. The t
statistic takes into account any cross-sectional 4 depedence of abnormal returns over the event
window, t tests are performed using crude adjustment method (CAM) sugested by Brown and
Warner (1980). ([Link] al: 2006 and Miles & Rosenfeld: 1983)
The t statistic quantifies the reliability of the calculation i.e. determines if the estimate is
statistically significant or not. Common thresholds are 90%, 95% and 99% confidence levels
requiring a t statistic of 2.66 for 90%, 2 for 5% and 2.66 for 99% or higher for a degree of
freedom of 66. The significance level is measured as a two tailed meaning we are interested
in “whether the computed values lay above or below the true coefficient”. By examining the t
stat it is possible to determine whether the variable i.e. the event of the day had a significant
impact on the share price. (Weil: 2001).
Further, if the event is statistically significant at 1% then the null hypothesis5 is rejected. If
the probability value is below 0.05 but greater than 0.01, then the null hypothesis is typically
not rejected, but is not as strongly significant as at 1% level of confidence. “Probability
values between 5 % and 10% suggest the event was weakly statistical significant. (Hennessy:
1989)
To illustrate:
H0(null hypothesis) : AR=0; If the population AR is indeed zero, then the sample estimate should be
close to zero.
RBS Share price during event
window – Table 1
3
The ratio of the estimated coefficient to its standard error
4
A research design in which events are compared on one or more variables at the same point in time. (Ohio Department of mental
health:2011)
5
The null hypothesis is that the abnormal returns are not significantly different from zero. Under the null hypothesis, the abnormal returns
are independent and identically distributed and normal with a mean of zero and the variance given by the variance of abnormal returns over
the estimation period ([Link] al: 2006)
Date Day RBS (Pence)
06/10/2008 -5 148.1
07/10/2008 -4 90
08/10/2008 -3 90.7
09/10/2008 -2 96
10/10/2008 -1 71.7
Event day 0 65.7
13/10/2008
14/10/2008 1 65
15/10/2008 2 65
16/10/2008 3 65
17/10/2008 4 68.6
20/10/2008 5 84.5
This table shows the actual closing share price of RBS during the event window. The
highlighted box is the share price on event day. This table should be used in conjunction with
the abnormal returns (table 2,3 4) derived Table 3: Market & Risk adjusted model
from the 3 models as it can be used to suggest Date Abnormal return T Stat
whether an event did positively/adversely -5 -1.25728 -0.45776
affect the share price. -4 -50.9462 -18.5488*
-3 14.77014 5.377613*
2 8.725867 3.176974*
-1 -4.64862 -1.6925*
0 -30.1473 -10.9762*
1 -9.75893 -3.5531*
2 19.63338 7.148256*
3 14.47364 5.269664*
Table 2: Mean adjusted return
4 -8.41159 -3.06255*
Day Abnormal return T stat 5 6.570948 2.392396*
-5 -22.755 -3.69478*
-4 -49.6695 -8.06494*
-3 0.9131 0.148262
-2 Table 4: Market adjusted
5.817415 model
0.944586
-1
Date -29.0474
Abnormal return -4.71649*
T stat *** Significant at the 10% level
0 -5 -8.60085
-14.71492303 -1.39654
-3.33771* ** Significant at the 5% level
1 -4 -0.93283
-50.15584338 -0.15147
-11.3766*
2 -3 0.138331
6.092621151 0.022461 * Significant at the 1% level
1.38196
3 -2 0.138331
6.897922324 0.022461
1.564622 Critical values of T at 66 degrees of
4 -1 5.528858
-19.9211892 0.897733 freedom
-4.51863*
5 0 20.98423
-16.67593837 3.407255*
-3.78252* 10% 1.67
1 -4.245594929 -0.96301 5% 1.998
2 7.428654515 1.685006**
3 5.496913237 1% 2.6557
1.246838
4 0.301345985 0.068353
5 15.57969648 3.533867*
Table 5: Mean Market adjusted Market & Risk
AR on event day adjusted return adjusted return
return
AR on day 0 -8.600851 -16.68 -30.1473
T statistic on day 0 -1.39654 -2.03624 -10.9762
Mean adjusted abnormal return for event window
30
20
10
0 Abnorma
-5 -4 -3 -2 -1 0 1 2 3 4 5 l return
-10
-20
-30
-40
-50
-60
The graph above summarises abnormal returns over the event window, it shows on day 0 the
day of the event whereby the government injected capital into RBS. The event t statistic is
statistically insignificant for the event day, however this result does not coincide with the t
statistic derived using the market adjusted and the market and risk adjusted model. The
government had announced the event at 10am RBS stock price on the event day opened at
69p and closed at 65p. A reason for this happening could be that the mean return value is too
close to the event day return. A longer estimation window test was conducted which also
showed a very similar result; i.e abnormal return was statistically insignificant on event day
at all confidence levels.
Another reason could be that the standard deviation of abnormal returns is too high (because
of high volatility in the abnormal returns), making the t-ratio too small. A longer estimation
window was tried which did lower the standard deviation but abnormal return was still
statistically insignificant on event day at all confidence levels.
Given that RBS is a bank and that the date of the event was a volatile period in itself it is
possible events occurring in other industries may have affected the outcome of this event.
B&W (1980) suggest mean adjusted returns does not perform as well as the other two models
if event clustering is present Brown & Warner (1980) state “When there is event clustering,
methodologies which incorporate information about the market’s realized return perform
substantially better than Mean Adjusted Returns.”
From the results above it is apparent that on days -1, -4 and -5 i.e. days before the event,
there happen to be statistically significant negative impact on the firm’s stock value
confirmed by the other two models. This can reflect a possible data breach; however -5 days
before the event, on 6th October (excluding weekend) RBS market value declined by £6
billion following its credit rating being cut by the credit rating company Standard & Poor.
This occurred minutes before the market closed and thus a possible reason for statistically
significant negative impact on the firm’s stock value on day -4 (7th Oct). (Hearald Scotland:
2008)
On day -1 (10 Oct) the Independent (2008) claims RBS share price fell partly because they
asked the government for assistance and partly because the bank has asked shareholders for
another capital investment.
The Share price after the event remained stable and random for 3 days and increased on day
4 (17 Oct). This was due a financial company announcing it would present a rescue plan for
RBS to the government (Bloomberg: 2007) However the market and risk adjusted model only
accounts for this event suggesting abnormal returns was negatively statistically significant;
but this was wrong as the event did not affect the share price adversely.
5 days after the event (20 Oct) reflected by the 3 models which show a strong statistically
significant positive impact on this day. This rise occurred after the announcement from a
company whom agreed to bid for a majority stake in one of RBS’s assets.
In relation to the above events, abnormal returns were justified and days after the event the
share price remained random and stable until an event occurred, but until the event occurred
the share price remained immobile suggesting market efficiency.
In conclusion, the event study shows that markets reacts badly to state capital injections as
reflected in the decrease in RBS share price.
The purpose of the event study was to decipher whether prior or after the event, the stock
price showed abnormal performance which could not be justified inferring the market was
not efficient. However, the event in question was certainly unique; it was arduous to assess
market efficiency due to a series of events occurring during the event window. After studying
the event window abnormal performance on certain days prior and after the event was
justified and there were days during the event window where stock prices remained immobile
until an event actually occurred implying true market efficiency.
The event failed to appear statistically significant using the mean adjusted model but
appeared statically significant at 1% level of significance using the other two models
implying the event had a strong impact on abnormal returns. On the event day there were
many events occurring in other industries which could have impacted RBSs share price; so a
possibility of event clustering which was not explicitly detailed in the news. The other two
models are said to perform better than mean adjusted returns if event clustering is present.
Reference
Ahern,K.R “Sample selection and event study estimation”. Journal of Empirical Finance
Vol.16 (2009) pp.466- 482.
Allen, G.(2009) Computational science: ICCS [Link]: Springer-Verlag
Acquisti, A., Friedman.A, Teland, R. “Is there a cost to privacy breaches? An event study”
Twenty seventh International conference on Informational systems.2006 pp.1-23
Brown, S.J., & J.B. Warner. “Measuring Security Price Performance”. Journal of Financial
Economics, Vol. 8, 1980, pp.205-58.
Bloomberg (2008) Stocks advance in Europe, led by RBS. Available at:
[Link]
Accessed [17 March 2011]
Dyckman,T. Philbrick,D. & Stephan, J. “A Comparison of Event Study Methodologies
Using Daily Stock Returns: A Simulation Approach”. Journal of Accounting Research, Vol.
22, Studies on Current Econometric Issues in Accounting Research (1984) pp. 1-30
Hearald Scotland (2008) Royal Bank of Scotland shares plunge and S&P cuts rating.
Available at: [Link]
1.891395 [ Date accessed: 13 March 2011]
Hennessy, L. (1989), Quantative Analysis. London: Letts
Hirschheim, R, Heinzl, A. & Dibbern, J. (2009) “Information Systems Outsourcing: Enduring
Themes, Global Challenges, and process opportunities” 3rd edn. Berlin: Springer
Investopedia (2011) Define standard error. Available at:
[Link] [Date accessed: 12 March 2011]
Investopedia (2011) Define Abnormal returns. Available at:
[Link] [Date accessed: 28 February 2011]
Investopedia (2011) Define Market efficiency. Available at:
[Link] [Date accessed: 28 February 2011]
Investopedia (2010) Define Default Risk. Available at:
[Link] [ Date accessed : 30 Oct. 10]
Investopedia (2011) Define efficient frontier, Available at:
[Link] [Date Accessed: 1 Mar. 11]
Miles, J.A. & Rosenfield, J.D. “The effect of voluntary spin-off announcements of
shareholder wealth”. The journal of Finance, Vol 9 [5] (1983) pp. 1-10
Munterman, J. (2007) Event-Driven Mobile Financial Information-Services. Frankfurt:
German University Publishers
Ohio : Department of Mental Health (2011) Define cross-sectional. Available at:
[Link]
lab/[Link] [Accessed: 15 March 2011]
Shane, P.B. & Spicer, B. H. “Market response to environmental information produced outside
the firm”. The Accounting Review, Vol. 11 (1983) pp. 521 – 538
Statistics Gloassary (2011) Stationary series. Available at:
[Link]
%20%28in%20Time%20Series%29 [Date Accessed: 1 Mar. 11]
Weil, R. L., Wagner, M. J. & Frank. P. B. (2001) Litigation services handbook: the role of
the financial expert. New York: John Wiley & Sons Inc.