Fama French 12 industry out of sic codes

[Nini Herzl]

Hey guys,

I would like to create a variable for the Fama French 12 industry classifications (ffinds) with values from 1 to 12 representing the respective industry. I do have SIC codes in my data.
Each industry according to Fama French covers a range of SIC codes. Unfortunately SIC codes are string in my data, so I can't use
gen ffinds = 1 if ((sic<=0299 & sic>=0100) | (sic>=0700 & sic<=0799) | (sic>=0910 & sic<=0919) | sic==2048)

Does anybody have an idea how to solve this problem?

I already tried to recast the variable sic, but this didn't really work. E.g. if the sic code was originally 3089 the value after recasting was 107.

Thanks in advance.

[ben earnhart]

When you say you tried recasting it, what command did you use? The "destring" command seems like it ought to do exactly what you want.

[Sergiy Radyakin]

Unfortunately SIC codes are string in my data.........
I already tried to recast the variable sic, but this didn't really work. E.g. if the sic code was originally 3089 the value after recasting was 107.

Like how? Afaik recast doesn't work across the domain (from numbers to strings or other way around).

What you need is destring, but I am definitely more interested in what you've done to turn "3089" into 107.

For the Sic to FF conversion recode is better:

recode sic (100/299 700/799 910/919 2048 = 1 "Label for first category") (....=2 "Label for second category"), generate(ff)

Best, Sergiy Radyakin

[Nini Herzl]

Thanks Form your anderes. I trief the commands recast and destring. Both revealed that value of 107. As long as you browse the data you can see 3089 for the destringed sic Code. But as soon as you select the cell with 3089 for the sic code, the value 107 ist revealed

[Robert Picard]

Sounds like your sic variable contains data that was encoded from string to numeric using encode. It that's the case, then the simplest is to restore the original string values using decode and then proceed to map sic codes to Fama code.

Note that SIC codes are hierarchical and are usually stored as string. You do not need to convert them to numeric to map them to Fama codes. You can use any relational operators to compare strings, even ranges. Here are some (fake) examples

clear
input str4 sic
0115  
1081  
2399  
2999  
3324  
3679  
4899  
5048  
5181  
6062  
7291  
7374  
7389  
7699  
7941  
7997  
8049  
8093  
8243  
8322 
end

gen ffinds = .
replace ffinds = 1 if inrange(sic,"0100","1999")
replace ffinds = 2 if inrange(sic,"2700","3799")
replace ffinds = 3 if inlist(sic,"7699","7941")
replace ffinds = 4 if sic == "8093"
replace ffinds = 5 if sic > "8093"