Event study, is that a mistake?
Posted: 10 October 2013 10:38 PM   [ Ignore ]
Jr. Member
RankRank
Total Posts:  34
Joined  2013-03-31

Dear Joost,

I found even 20 days before the event date, a lot of statistics (from day -20 to 0 ) for daily retun are every significant.

Is it normal? I was thinking that I must make some mistakes….

I am using daily (-266,-30) as the estimation window.

I am using Eventus… Maybe that is the reason for this weird output…

Profile
 
 
Posted: 11 October 2013 01:30 AM   [ Ignore ]   [ # 1 ]
Jr. Member
RankRank
Total Posts:  34
Joined  2013-03-31

This is the output. It seems that something must be wrong. So many stars make me feel scared…

                Market Model, Value Weighted Index                          
                                                               
                  Mean                                            
                Abnormal   Positive:    StdCsect     Generalized            
      Day   N     Return     Negative       Z         Sign Z              
———————————————————————————————————————         
      -30   165       -0.14%  62:103<      -2.629**    -2.497*             
      -29   165       -0.32%  54:111<<<    -3.877***    -3.745***             
      -28   165       0.71%  118:47>>>    5.935***    6.235***             
      -27   165       -0.32%  76:89       0.022       -0.314              
      -26   165       -0.69%  42:123<<<    -5.749***    -5.616***             
      -25   165       0.59%  125:40>>>    6.161***    7.326***             
      -24   165       0.29%  78:87       1.433       -0.002              
      -23   165       -0.50%  49:116<<<    -4.137***    -4.524***             
      -22   165       -0.57%  61:104<<    -3.919***    -2.653**             
      -21   165       0.60%  119:46>>>    5.207***    6.391***             
      -20   165       0.55%  98:67>>      4.056***    3.116**             
      -19   165       -0.28%  71:94       -2.573*      -1.094              
      -18   165       1.12%  121:44>>>    5.080***    6.703***             
      -17   165       0.12%  89:76)      1.769$      1.713$             
      -16   165       -1.06%  29:136<<<    -8.262***    -7.643***             
      -15   165       -1.32%  41:124<<<    -8.687***    -5.772***             
      -14   165       1.52%  132:33>>>    7.413***    8.418***             
      -13   165       0.31%  88:77       1.548       1.557              
      -12   165       -1.45%  31:134<<<    -8.379***    -7.331***             
      -11   165       1.70%  138:27>>>    10.349***    9.353***             
      -10   165       -0.93%  39:126<<<    -7.344***    -6.084***             
        -9   165       -0.61%  63:102<      -3.745***    -2.341*             
        -8   165       -2.27%  25:140<<<    -11.608***    -8.267***             
        -7   165       1.57%  108:57>>>    5.951***    4.675***             
        -6   165       0.75%  95:70>>      3.632***    2.648**             
        -5   165       0.16%  80:85       0.897       0.309              
        -4   165       0.71%  98:67>>      4.333***    3.116**             
        -3   165       -1.11%  37:128<<<    -7.543***    -6.396***             
        -2   165       0.61%  99:66>>      2.143*      3.272**             
        -1   165       0.66%  115:50>>>    4.622***    5.767***             
        0   165       -2.28%  17:148<<<    -12.724***    -9.514***             
        +1   165       1.53%  126:39>>>    8.899***    7.482***             
        +2   165       -2.35%  19:146<<<    -12.890***    -9.202***             
        +3   165       1.75%  129:36>>>    10.247***    7.950***             
        +4   165       -0.41%  56:109<<<    -3.181**    -3.433***             
        +5   165       2.10%  138:27>>>    11.215***    9.353***             
        +6   165       -1.98%  25:140<<<    -10.815***    -8.267***             
        +7   165       2.26%  140:25>>>    12.725***    9.665***             
        +8   165       -1.82%  24:141<<<    -12.108***    -8.423***             
        +9   165       -0.02%  82:83       -1.247       0.621              
      +10   165       -1.62%  32:133<<<    -6.941***    -7.175***             
      +11   165       3.13%  149:16>>>    15.861***    11.069***             
      +12   165       -1.60%  32:133<<<    -9.400***    -7.175***             
      +13   165       -1.66%  30:135<<<    -9.918***    -7.487***             
      +14   165       0.18%  98:67>>      2.629**      3.116**             
      +15   165       0.41%  107:58>>>    4.911***    4.520***             
      +16   165       -1.82%  30:135<<<    -11.073***    -7.487***             
      +17   165       2.11%  133:32>>>    10.702***    8.574***             
      +18   165       0.20%  88:77       1.028       1.557              
      +19   165       -1.63%  29:136<<<    -9.413***    -7.643***             
      +20   165       2.13%  136:29>>>    9.328***    9.041***             
      +21   165       -0.81%  44:121<<<    -4.323***    -5.304***             
      +22   165       -0.16%  66:99(      -0.664       -1.874$             
      +23   165       -0.18%  59:106<<    -2.281*      -2.965**             
      +24   165       -1.20%  40:125<<<    -6.704***    -5.928***

Profile
 
 
Posted: 11 October 2013 02:00 AM   [ Ignore ]   [ # 2 ]
Jr. Member
RankRank
Total Posts:  34
Joined  2013-03-31

Is that because the daily OLS regression is a very bad one?

Profile
 
 
Posted: 11 October 2013 08:49 AM   [ Ignore ]   [ # 3 ]
Administrator
Avatar
RankRankRankRank
Total Posts:  901
Joined  2011-09-19

hi Anna,

What are you testing? If you have an event like an earnings announcement, press release or something else that you can identify precisely, usually a short window (1, 2, 3 or maybe 5 days) around the event is used. So, you would sum up the returns in the window around the event, and then test if that cumulative return is statistically different from zero. If you can’t pinpoint the day and need a wider window, that would introduce ‘confounded’ (unrelated) events and thus noise. In that case, you would still sum it up, for example, the ‘post earnings announcement drift’ in the 30 days a week after the press release, for example. When you output the returns of the individual days, it is not clear if the total over the timeframe is positive/negative. For example, some daily returns are positive and some are negative.

The drawback of Eventus is that it is bit of a ‘black box’. It can be helpful of course but for publication purposes it is probably better to write the code yourself.

best regards,

Joost

 Signature 

To reply/post new questions: Please use the group WRDS/SAS on Google Groups! http://groups.google.com/d/forum/wrdssas

Profile
 
 
Posted: 11 October 2013 01:54 PM   [ Ignore ]   [ # 4 ]
Jr. Member
RankRank
Total Posts:  34
Joined  2013-03-31
Joost Impink - 11 October 2013 08:49 AM

hi Anna,

What are you testing? If you have an event like an earnings announcement, press release or something else that you can identify precisely, usually a short window (1, 2, 3 or maybe 5 days) around the event is used. So, you would sum up the returns in the window around the event, and then test if that cumulative return is statistically different from zero. If you can’t pinpoint the day and need a wider window, that would introduce ‘confounded’ (unrelated) events and thus noise. In that case, you would still sum it up, for example, the ‘post earnings announcement drift’ in the 30 days a week after the press release, for example. When you output the returns of the individual days, it is not clear if the total over the timeframe is positive/negative. For example, some daily returns are positive and some are negative.

The drawback of Eventus is that it is bit of a ‘black box’. It can be helpful of course but for publication purposes it is probably better to write the code yourself.

best regards,

Joost

Thanks so much, Joost!

It is not earnings announcment, it just a single event with much small impact.

It seems that Eventus produced unreliable results, at least for my case. I will use my own code to do it.

Best regards,

Anna

 

Profile
 
 
Posted: 11 October 2013 02:38 PM   [ Ignore ]   [ # 5 ]
Administrator
Avatar
RankRankRankRank
Total Posts:  901
Joined  2011-09-19

ok smile

 Signature 

To reply/post new questions: Please use the group WRDS/SAS on Google Groups! http://groups.google.com/d/forum/wrdssas

Profile