Having trouble creating a specific sample

[Aleksandar Filimonov]

Hello everyone,

I am using panel data from Compustat and I need to create a sample of high-growth firms (keep only the gvkey of HGFs). The first definition of an HGF which I need to try is at least a 20% increase in sales for 3 consecutive years and a minimum of 10 employees in the base year. The whole sample period is 7 years (2013-2019) and the 3-year high-growth period can be anywhere within that. After xtset gvkey fyear I was given the following code by a professor of mine (I can't ask him for clarifications as he is currently away):

1. gen sale1 = L.sale
2. gen growth_sales = (sale - sale1) / sale1
3. gen suffgrowth = 1 if growth_sales >= 0.2
4. bysort gvkey: gen baseyear_emp_temp = emp if fyear == (whatever the base year is)
5. bysort gvkey: gen baseyear_emp = max(baseyear_emp_temp)
6. gen hgf = 1 if suffgrowth == 1 & L.suffgrowth == 1 & L2.suffgrowth == 1 & baseyear_emp >= 10
7. gen gvkey_hgf = gvkey if hgf == 1
8. keep gvkey_hgf
9. drop if gvkey_hgf != .

I am not sure how "whatever the base year is" needs to be defined since this should refer to the base year of the 3-year period and not the whole sample period. Even if I try 2013, I get an "invalid synthax" error on line 6.

Also, I am interested to know whether there is any other way of creating a sample with these conditions.

Any help would be appreciated!

[Carlo Lazzaro]

Aleksandar:
welcome to this forum.
1) -emp- variable is missing;
2) as you surmise, "whatever the base year is" should be made numerical.
More details (please see the FAQ) about your dataset can help interested listers helping you out. Thanks.

[Aleksandar Filimonov]

Hi Carlo,

Thank you for your reply!

All variables that should go into this are present in the dataset at the outset:

• gvkey - Global Company Key
• sale - Sales/Turnover
• fyear - Fiscal year
• emp - Number of employees

There was an empty line that seems to have been deleted, so the error I mentioned when using 2013 for the base year actually occurs in line 5. Again, I don't think it should be 2013, this is just something I tried.

Have you got any ideas about how to complete the task?

Many thanks,
Aleksandar

[Carlo Lazzaro]

Aleksandar:
try this (soboptimal, I know) line of code and see it it works:

bysort gvkey: gen baseyear_emp_temp = emp if fyear ==2013

[Aleksandar Filimonov]

Hi Carlo,

As already meantioned, I did try this exact line and I receive an "invalid syntax" error message on the next line. Any idea why that may be?

[Carlo Lazzaro]

Aleksandar:
what you complained about was clear, but in the following toy-example (that you can easy replicate) the code works:

. use "https://www.stata-press.com/data/r17/nlswork.dta"
(National Longitudinal Survey of Young Women, 14-24 years old in 1968)

. bysort idcode: gen baseyear_emp_temp=grade if year==70

. tab baseyear_emp_temp

baseyear_em |
     p_temp |      Freq.     Percent        Cum.
------------+-----------------------------------
          0 |          3        0.18        0.18
          1 |          1        0.06        0.24
          4 |          3        0.18        0.42
          5 |          4        0.24        0.65
          6 |         11        0.65        1.30
          7 |         17        1.01        2.31
          8 |         48        2.85        5.16
          9 |         66        3.91        9.07
         10 |        122        7.24       16.31
         11 |        117        6.94       23.25
         12 |        988       58.60       81.85
         13 |         81        4.80       86.65
         14 |         71        4.21       90.87
         15 |         46        2.73       93.59
         16 |         91        5.40       98.99
         17 |         10        0.59       99.58
         18 |          7        0.42      100.00
------------+-----------------------------------
      Total |      1,686      100.00

.
.

Therefore you have to provide more details and/pr delve into the format of your variables: is there any -string- format variable?
Please help interested listers help you out. Thanks.

[Aleksandar Filimonov]

Hi Carlo,

None of the variables are strings: gvkey - long; fyear - double; emp - double; sale - double; sale1 - float.

This line itself works with my dataset as well. It is the next line where I get an error:

bysort gvkey: gen baseyear_emp = max(baseyear_emp_temp)

Many thanks,
Aleksandar

[Carlo Lazzaro]

Aleksandar:
try:

bysort gvkey: egen baseyear_emp = max(baseyear_emp_temp)

[Aleksandar Filimonov]

Hi Carlo,

Thanks, that did resolve the error! However, I am still not sure that writing 2013 as the base year gives me the data under the conditions that I need. Do you have any suggestion how to make the 'base year' any year within the period? I tried the following:

bysort gvkey: gen baseyear_emp_temp = emp if fyear == 2013 | 2014 | 2015 | 2016 | 2017

But, again, I am not sure whether this is appropriate as there are some firms for which I only get 1 observation.

Thanks,
Aleksandar

[Carlo Lazzaro]

Aleksandar:
you may want to try:

bysort gvkey: gen baseyear_emp_temp = emp if fyear <=2017

You can replace 2017 with 2019 if the latter is the last year in your dataset.