# Re: st: RE: Timevarying IDs in a panel dataset

 From Daniel Höchle To Subject Re: st: RE: Timevarying IDs in a panel dataset Date Tue, 22 Nov 2005 16:31:43 +0100

```Nick,

You are right, the stock's ID is no unambiguous identifier. But for each
single date, ID_now is unique. However, ID_nextperiod is not necessarily
unique as the following might happen (in case of a "merger" or an
"acquisition"):

Date       ID_now       ID_nextperiod         Price_now
1           2               3                    100
1           3               3                    110
2           3               3                    105
3           3               3                    100

In this rare case, the resulting stock returns would be:

ID_now==2, return from Date==1 to Date==2 is 105/100-1=0.05
ID_now==3, return from Date==1 to Date==2 is 105/110-1=-0.0455
ID_now==3, return from Date==2 to Date==3 is 100/105-1=-0.0476

Therefore, what I presume I should do is creating an additional variable
called Price_nextperiod. After generating this variable, the dataset in
my initial example would look as follows:

Date       ID_now       ID_nextperiod         Price_now
Price_nextperiod
1           1               1                    100              110
2           1               1                    110              105
3           1               1                    105              100
4           1               1                    100                .
3           2               2                     22               27
4           2               2                     27                .
1           3               3                     20               25
2           3               2                     25               22
1           4               5                     60               65
2           5               5                     65               67
3           5               5                     67                .
...

Then the stock return may be calculated as

. PriceRet = Price_now / Price_nextperiod - 1

However, I do not know how to generate variable Price_nextperiod. Thanks
again for any suggestions.

Daniel

> It is not clear that these identifiers are
> really identifiers. Or, to put it another way,
> how can this problem be solved at all?
>
> As I understand it, you may need a new identifier
> based on a combination of now and the
> next period. So, any stock that is 3 now
> and 2 next period cannot be the same item
> as one that is 3 and 3, any more than a person
> who is male now and female next week can be the
> same person as someone with more stable gender.
> But does a combination of now and next uniquely
> identify stock?
>
> Nick
> n.j.cox@durham.ac.uk
>
> Daniel Höchle
>
> > I do have a panel of several thousand stocks. Each stock has an id
> > called ID_now. However, there is a problem: The stock's id may
change
> > over time. Therefore, I do have variable ID_nextperiod which
contains
> > the stock's id in the next period. The data looks like this:
> >
> > Date       ID_now       ID_nextperiod         Price_now
> >  1           1               1                    100
> >  2           1               1                    110
> >  3           1               1                    105
> >  4           1               1                    100
> >  3           2               2                     22
> >  4           2               2                     27
> >  1           3               3                     20
> >  2           3               2                     25
> >  1           4               5                     60
> >  2           5               5                     65
> >  3           5               5                     67
> > ...
> >
> > What I would like to do is calculating the discrete stock returns.
If
> > the stock's ID would not change over time this could be
> > achieved (after
> > tssetting the dataset) by simply typing
> >
> > . PriceRet = Price_now / L.Price_now - 1
> >
> > However, I do not have any idea how to calculate the discrete stock
> > returns in case of changing IDs. Any ideas would be appreciated.

*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/
```