Matching through different SIC codes

[Boy van Reede]

Hello you all,

I am currently working on an assignment were I want to match a group of companies doing an IPO to a control group by industry and market cap.

For the industry I need to match on SIC codes, the problem now is that not every treated company has a control company with the same 3-digit SIC code.
I am trying to make an algorithm (a loop) which will first look at the 3-digit SIC code, but when there is no match, it should look at the 2-digit SIC code. And when
there is no match on the 2-digit SIC code, looks on the 1-digit SIC code. But i haven't found the right algorithm at this moment.

I know that I need to merge the two groups of companies and use:

merge 1:M 3SICCODE using control group
keep if _merge ==3

But before or after this there needs to be a line where it will search for another match of a 2-digit SIC code or even a 1-digit SIC code. Because it could also be possible that
the _merge is 1(treated group) or 2(control group), so there is no match.

Maybe someone of you can help me?

Many thanks

[Carlo Lazzaro]

Boy:
welcome to the list.
I'm not clear whether your data are sparsed in different datasets (as it would seem from -merge-) or not.
If you have access to a unique dataset (or you can create somehow), you can create different groups of companies for further analyses,as in the following toy-example (I created a variable named -IPO- which takes on 0 if the company was not involved in an IPO and 1 otherwise):

. set obs 10
number of observations (_N) was 0, now 10

. g SIC_CODE=10001 in 1/5
(5 missing values generated)

. replace SIC_CODE=10002 if SIC_CODE==.
(5 real changes made)

. g IPO=1 in 1/2
(8 missing values generated)

. replace IPO=0 in 3/7
(5 real changes made)

. replace IPO=1 if IPO==.
(3 real changes made)

. egen flag=group( SIC_CODE IPO)