Stata: ignore missing values when counting

The system variable _n is fixed in the sense that it denotes the number of the current observation independently of your if condition. What you are asking for with your code is a counter that takes on the value of the number of the observation if it is not missing. This is different from the count of non-missing observations up until a certain point, which is what you really want.

Be aware of the data type of your missings. As you present it, firmage0 seems like a string variable because blanks are missings for strings. Missings for numeric values are represented as . (or .<something>). See help missings.

One solution is doing a cumulative sum conditional on the missings. I assume variable type is numeric:

clear all
set more off

input year firmage0
1975      .  
1975      . 
1975      .      
1975      .  
1975    21
1980    60 
1980    .
1980    89       
1985      . 
1985      .  
1985    23    
1985    45    
1985    47
1985     .   
end 

list, sepby(year)

bysort year: gen counter = sum(!missing(firmage0))
replace counter = . if missing(firmage0) // optional

list, sepby(year)

!missing(firmage0) evaluates to false (0) or true (1) depending on whether the observation for firmage0 is missing or not, respectively. Then sum does the cumulative sum of that.

If the variable is in fact of string type, then replace the corresponding line with:

replace counter = "" if missing(firmage0) // optional

A one-liner (for numeric type) is:

bysort year: gen counter = cond(missing(firmage0), ., sum(!missing(firmage0)))

but I prefer the first way because it is easier to read.