Compare observation to previous ones: Patent Data

[Pascal Meier]

Hello everyone

I have an extensive dataset about individual patents (patnum), belonging to a technology class (class) and subclass (subclass) that a firm (permno) filed. For each patent, I have the exact date (fileddate) the patent was filed and the corresponding year (year_filed). For each patent-firm-year observation, I want to determine whether the technology is new to the firm, i.e. whether it is the first filed application in a specific class (or subclass), or whether it is an application the firm has already knowledge in, i.e. existing patents of the same technology class (or sublass). Thus, for each patent, I aim to have an indicator variable whether the patent is "exploitative" or "explorative" to the firm. In addition, it would be great to have a new variable - proximity - capturing the aggregated number of a firm's patents in class (or subclass) X in relation to the total aggregated number of patents of this firm (up to the year under analysis); in other words, the technological proximity between the patents filed in year t and the existing patent portfolio held by the same firm up to year t−1.

I hope I made it clear what I want to achieve. Any recommendation on how to achieve it is highly appreciated - I really struggle to tackle this issue.

Here is an example of my data

Example generated by -dataex-. To install: ssc install dataex
clear
input long(patnum permno) str3 class str7 subclass float(green fileddate year_filed)
1568124 12503 "056" "006000O" . -14447 1920
1568134 12503 "239" "584000O" 1 -13628 1922
1568142 10401 "333" "168000O" . -14385 1920
1568153 16029 ""    ""        . -13433 1923
1568163 12503 "056" "011500O" . -14487 1920
1568338 14875 "400" "689000O" . -13209 1923
1568578 18227 "192" "003620O" . -12658 1925
1568609 10620 "324" "502000O" . -12838 1924
1568615 12079 "123" "058100O" 1 -14561 1920
1568623 11754 "396" "440000O" . -13955 1921
1568625 11447 "220" "798000O" . -14129 1921
1568648 12079 "072" "085000O" . -14275 1920
1568656 11754 "396" "341000O" . -12713 1925
1568658 11754 "430" "631000O" . -12722 1925
1568667 11754 "546" "101000O" . -13973 1921
1568670 11754 "396" "341000O" . -13045 1924
1568671 11754 "396" "343000O" . -13045 1924
1568682 12060 "310" "219000O" . -13233 1923
1568692 12060 "310" "203000O" . -12852 1924
1568694 12060 "313" "538000O" . -14139 1921
1568702 12060 "346" "035000O" . -13876 1922
1568710 12060 "363" "084000O" . -13426 1923
1568731 12060 "384" "151000O" . -13100 1924
1568735 12060 "318" "741000O" . -13598 1922
1568748 12060 "361" "643000O" . -12590 1925

Best regards and many thanks
Pascal

[Clyde Schechter]

Your data example does not contain any variable identifying the firm. I'll assume you have one, and that it's called firm_id.

I also don't understand everything you want. But I can create for you variables that indicate which patents are first in class and which are first in subclass for the firms, and also how many they already have in that class or subclass at each observation. I couldn't figure out what you meant by proximity, and I don't understand the terms exploratory and exploitive, but I'm guessing that with these variables you can get to those results easily from there.

by firm_id class (fileddate), sort: gen byte first_in_class = (_n == 1)
by firm_id class (fileddate): gen in_class_to_date = _n
by firm_id class subclass (fileddate), sort: gen byte first_in_subclass = (_n == 1)
by firm_id class subclass (fileddate), sort: gen in_subclass_to_date = _n

[Pascal Meier]

Dear Clyde

Many thanks for your prompt and, as always, very helpful answer! The firm identifier is PERMNO. I was able to achieve what I wanted with your code.