Count different portfolios

[Juan Gonzalex]

Hi,

My dataset is as following:

* Example generated by -dataex-. For more info, type help dataex
clear
input long gvkey float(time change5 count)
101340 422 2 20
101538 422 1 21
208224 422 1 21
100131 422 1 21
100499 422 4 14
101317 422 4 14
102283 422 1 21
221020 422 2 20
102902 422 5 16
 19591 422 2 20
 12384 422 4 14
 19565 422 5 16
100760 422 4 14
100862 422 1 21
100095 422 1 21
101048 422 1 21
 11749 422 2 20
100149 422 3 19
100980 422 1 21
 18636 422 3 19
100619 422 3 19
100913 422 4 14
100045 422 2 20
102569 423 2 20
100816 423 4 14
101310 423 1 21
 19904 423 5 16
101207 423 4 14
100728 423 5 16
101343 423 3 19
222305 423 5 16
100774 423 4 14
200189 423 3 19
 14620 423 3 19
220392 423 2 20
 19565 423 5 16
100951 423 3 19
 19579 423 5 16
  4439 423 1 21
 12368 423 5 16
100060 423 1 21
end
format %tm time

Change5 are portfolios which are formed each month. I would like to count how many portfolios consist of less than 30 stocks (count).In my head I have something like the following formula: count if time, change5 and count is the same. So for example:

input long gvkey float(time change5 count)
101538 422 1 21
208224 422 1 21
100131 422 1 21
102283 422 1 21
100862 422 1 21
100095 422 1 21
101048 422 1 21
100980 422 1 21

All these observations from the sample above are in the same portfolio, so this would count as 1 portfolio.

Is there an easy formula for this?

Thanks

[Carlo Lazzaro]

Juan:
are you looking for something along the following lines?

. bysort gvkey (time) change5: gen wanted=1 if count<30

. total wanted, over( change5)

Total estimation                                Number of obs = 41

------------------------------------------------------------------
                 |      Total   Std. err.     [95% conf. interval]
-----------------+------------------------------------------------
c.wanted@change5 |
              1  |         11          0             .           .
              2  |          7          0             .           .
              3  |          7          0             .           .
              4  |          8          0             .           .
              5  |          8          0             .           .
------------------------------------------------------------------

[Juan Gonzalex]

Hi Carlo,

Thanks for your response, but with the code you provided I get the number of observations per change5, but what I want is the number of portfolios. So if the time and change5 are the same, it is a portfolio. So I want to count how many different portfolios are there, counting if time and change5 are the same. In the example above (#1), you can see that the time (422) and the change5 (1) are the same, so all those observations are in 1 portfolio.

[Carlo Lazzaro]

Juan:
see -collapse-.

[Juan Gonzalex]

Thanks Carlo, it worked!