i am doing a event study thesis right now and are currently using the methodology of McKinlay and Brown & Warner. I have an extremely large sample with over 7000 firms. It is necessary or worthy to use logarithmic returns as actual stock returns? I do not find a clear answer in the literature!
Kind Regards and succes with your website
I am doing an event study on LSAP and effects on selected emerging market countries. I've collected my data and selected emerging markets local currency-denominated bonds (data from Bloomberg or respective central banks) and USD-denominated bonds (JP Morgan EMBIG Index) for countries where data are accessible/available.
I have set up the OLS regression in excel although there are a few things where I have uncertainties.
1. Events (t = 0). Is it ok that I only choose the first time that LSAP was introduce and the first time of tapering? These would be my t = 0. For the former, this corresponds to March 18, 2009. For the latter, this was on December 2013.
However, I note from a paper from the Fed that events could also correspond to announcements. Therefore, with respect to the first LSAP, the announcement occurred on November 25, 2008 and for the tapering, this was on June 2013.
2. Window. Because I am confused about which event to take, I am having trouble setting up/choosing my windows, i.e., estimation period, event and post-event windows.
3. Null hypothesis. I set this up as mean abnormal return (or average residual) = 0. However, I've read papers too where the null is based on the cumulative average return.
I hope you can shed light to my questions. Many thanks in advance for your advice.
Dear Prof. Leemakdej,
I use an event study in my research. Using your research app I've got results of BMP and GRANK test. How can I decide whether I should accept the null hypothesis, that event has no impact on the behaviour of returns, or not. With wich parametrs I should compare CAAR BMP and CAAR GRANK Z
to make conclusions?
Thank you for your answer.
Hi, I am currently working on an event study in excel. I am analyzing 1095 events from different companies. I am trying to calculate the standard deviation we are going to use in our CAAR analyzis, but it seems really low.
What I am doing to find the standard deviation:
Using the formula "STEYX(..:..) to find the standard error for each event.
Then I square all the 1095 standard errors, before i use the formula SUM(..:..) to sum all of them together.
In order to find VAR(AR) (Variance of AR) I use this formula -->
1/N^2 * ( Sum of all the standard errors)
Then, to find the standard deviation I take the square root of the last calculation above.
However, this st. dev seems really low. My question is, is this the right steps to calculate the st. dev?
Dear Prof. Leemakdej
I am a researcher and find the event study tools very useful to do research in my field. Please allow me to say that the eventstudytools.com is the best website I have seen for event study: very comprehensive, detailed, hands-on and very user-friendly. I will introduce the web to my students for research purposes.
For this reason, I am writing to ask three very basic questions:
1. Would it make a huge difference if we use the close pricing rather than justified close price, in your knowledge and experience?
2. Do we have a textbook that systematically explain how the event study is conducted so that I can recommend it to my students?
3. The event study tools offered on this website are completely free?
Thank you for your help. I look forward to hearing from you.
I have only one AR and CAR to test. How to test the AR and CAR for one company and individual event when calculated abnormal returns are not normally distributed?
Can I use t test? If yes how to to justify?
Event window (5 days) Estimation window (30 days)
Or wich non-parametric test is suitable for my event study?
I am doing an event study about market reactions to banks announced to repay their TARP loans. I was planning to employ a 3-day event window (-1,+1) around the event date and use a 143 trading day period(around six month) prior to the event window as the estimation window. Due to several banks paid back the money in several times, my advisor suggested that I should add dummy variables for every sub payment and for every day in event window. I was confused about his advice and could not reach to him due to the break. So can I get some practical advice on how to set different weight on days among event window and how to deal with different amount of event in each bank?
I am trying to conduct event studies for a large sample and I am using your website tools in order to get my abnormal returns, t-tests.
Your template shows estimation window length as:
-2 ; 2 ; -11 ; 120
However, my teacher told me to examine a larger estimation window, of 240 days, with 40 days prior to the event date. How should I put the data in excel in order to extend the window?
I want ask who is the author of t test for CAR. In which book or study I can find a formula for the t test for the cumulative abnormal return in the form shown in the table formulas for testing the significance of returns on this website. I ask you, because I can not find it in references and further readings.
I want to conduct a long term event study (the duration is 4 years) for severel firms. I researched the event dates, the stock market index for that time and the total return index for the companies. To estimate my abnormal return i use the market model like MacKinley:
R(i,t) = a + b*R(m,t) + e,
R(i,t) = observed return for stock i for the time t
R(m,t) = observed return for the market for the time t
Question: how do i get the R(i,t) and R(m,t) to estimate the a and b with the "ordinary least squares" method. Is the equation :
R(i,t) = [K(i,t) - K(i,t-1)]/K(i,t-1)
K(i,t) from the total return index for the stock i for the time t
The right one? Do I get the returns in this way to estimate the normal return?