Econometrics for Finance

1
Econometrics forFinance

• Session 1 Event Study
• Dr.Arnat Leemakdej

2
Outline of Session

• Overview of Event Study Methodology.
• FFJR (1969)
• Event Study Design
• Assignment1

3
What is an Event Study ?

• An event is the public announcement of a
  (usually voluntary) corporate action, such as a
  merger, a security issue, an earnings
  announcement, a new investment announcement, a
  stock split, a new product launch etc.
• An event study is an econometric procedure for
  isolating the stock price impact of the event or
  its impact on firm value.
• Seminal papers Fama et al.(69), Brown and Warner
  (1980, 85)
• Excellent Review Papers MacKinlay, Journal of
  Economic Literature,1997.

4
Seven Steps to an Event Study

• Event Definition
• Define the event of interest and the period over
  which the impact on security prices will be
  examined -the event window.
• Firm Selection Criteria
• availability of data, characteristics of the data
  sample.
• Measuring Normal and Abnormal Returns
• Actual return minus normal return (estimated
  return as if the event had not occurred).
• Estimation Procedure
• Estimate model parameters over the estimation
  period, usually prior to the event.
• Testing Procedures
• Define Null Hypothesis, aggregation of the ARs,
  statistical tests.
• Empirical Results
• Diagnostics, sample size and any possible
  violations of assumptions.
• Interpretation and Conclusions
• Economic insights into the event and its impact
  on firm value, competing explanations etc.

5
Event Study Methodology- some issues

• Exactly when did news of the event hit the
  market? (at 1115 am, today, this week,
  this month)- event window
• The more precise the answer, the more powerful
  the event study analysis
• Is the announcement of the event partly
  anticipated? (rumors, leakage, prediction)
• The less the event is anticipated, the more
  powerful the event study analysis
• Is the event voluntary? (did managers have a
  choice)
• Voluntary events convey more of managers
  information about the firm
• Does this event trigger future events? (merger
  may trigger antitrust complaint, initial tender
  offer may trigger competing bid, IPO may trigger
  SEO)
• Fewer the triggering events, the easier the
  interpretation of the results

6
Event Study Procedure (1)

• (estimation window (event window (post-event
  window
• Returns indexed in event time. Let 0 be event
  date, event window is T1 1 to T2, (LengthT2-T1)
  
• Estimation window is T0 1 to T1 (length
  L1T1-T0)
• Event window is set around the event day 0, e.g.
  (0,1), (-1,0,1) since you can then study ARs
  around the announcement date.
• Post-event window is sometimes used as well.

T0
T1
0
T2
T3
7
Event Study Procedure (2)

• Abnormal Return is defined as
• Common candidate models for the normal return
  are the constant mean return model and the market
  model
• Mean Return model typically uses nominal daily
  returns but with monthly data nominal and and
  real returns used.
• Could use multifactor models (like Fama French
  three factor model etc.)- but gains are limited.

8
Event Study Procedure (3)

• Next, estimate the market model parameters for
  each firm in the sample during estimation period
  using a suitable market index.
• Next, using the estimated parameters of the
  market model, calculate for each firm period in
  the event window
• H0 Event has no impact, AR jointly normal with
  mean0 and variance

9
Event Study Procedure (4)

• Aggregating Abnormal Returns
• Calculate the CARs over event window in two ways
  
• Across firms and then over time (t1,t2)
• Across event window time (t1,t2) and over firms

10
A pictorial depiction!

• Time?
• Firms?

11
FFJR (1969)

• Original Paper for Event Study
• Impact of Stock Split
• The Information Issues
• Implications on Market Efficiency

12
Event Study Design

• What should be the length of Event Window
• What should be the length of Comparison Period
  (Estimation Period)
• How to formulate Portfolio

13
Assignment1

• Read Brown Warner (1985)
• What is the calendar effect and how can we
  avoid it?
• How to define the Event Date?
• What should be the impact on event study if the
  market already anticipates the news?