I am interested in examining the effect of government funds on the start-up patenting activity. I observe a set of startups and their patenting activity over time. My dataset covers 20 years, and the startups contained in the dataset are launched in different years. In the data, I observe patenting activities of start ups to decay over time i.e. if they generate 10 patents in their first year, then they generate 8 in the next year, 5 after 2 years later and so on. Given this information:
My first question is: If they get the funds on their second years, most likely I would be underestimating the effect of government funds due to natural decay of number of patents over time. How to deal with this?
My second question is: How to deal with multiple treatments i.e. if start up receives more than one government fund?
My third question is: Since year fixed effects basically demeans the average patents for the particular year and start-ups are more likely to come up with more patents in their first couple of years, fixed effects are likely to de-maen more if there are more start ups entering in a given year. How could I deal with this?
Example data is here:
Code:
* Example generated by -dataex-. To install: ssc install dataex
clear
input double company_id float(postfund n_patents year)
 57593 0  5 1996
 57593 0  0 1997
 57593 0  0 1998
 57593 0  0 1999
 57593 0  0 2000
 57593 0  1 2001
 57593 0  0 2002
 85598 0  2 1988
 85598 0  0 1989
 85598 0  1 1990
 85598 0  0 1991
 85598 0  0 1992
 85598 0  1 1993
 85598 0  0 1994
 85598 0  0 1995
 85598 0  1 1996
 85598 0  0 1997
 85598 0  0 1998
 85598 0  0 1999
 85598 0  0 2000
 85598 0  0 2001
 85598 0  0 2002
164211 0  3 1994
164211 0  0 1995
164211 0  0 1996
164211 0  0 1997
164211 0  0 1998
164211 0  0 1999
164211 0  0 2000
164211 0  0 2001
164211 0  0 2002
219942 0  4 1991
219942 0  8 1992
219942 0  0 1993
219942 0  1 1994
219942 0  0 1995
219942 0  1 1996
219942 0  0 1997
219942 0  0 1998
219942 0  0 1999
219942 0  0 2000
219942 0  0 2001
219942 0  0 2002
259008 0  2 1984
259008 0 15 1985
259008 0 15 1986
259008 0  1 1987
259008 0  1 1988
259008 0  1 1989
259008 0  0 1990
259008 0  1 1991
259008 0  1 1992
259008 0  1 1993
259008 0  0 1994
259008 0  1 1995
259008 0  0 1996
259008 0  1 1997
259008 0  0 1998
259008 0  0 1999
259008 0  0 2000
259008 0  0 2001
259008 0  0 2002
259613 0 17 1983
259613 0  1 1984
259613 0  1 1985
259613 1  1 1986
259613 2  2 1987
259613 2  1 1988
259613 2  1 1989
259613 2  0 1990
259613 2  1 1991
259613 3  2 1992
259613 3  0 1993
259613 3  0 1994
259613 3  0 1995
259613 4  0 1996
259613 4  0 1997
259613 4  0 1998
259613 4  1 1999
259613 4  0 2000
259613 4  1 2001
259613 4  0 2002
264590 0  6 1994
264590 0  0 1995
264590 0  0 1996
264590 0  0 1997
264590 0  0 1998
264590 0  0 1999
264590 0  0 2000
264590 0  0 2001
264590 0  0 2002
266164 0  1 1989
266164 0  0 1990
266164 0  0 1991
266164 0  0 1992
266164 0  0 1993
266164 0  0 1994
266164 0  0 1995
266164 0  0 1996
266164 0  0 1997
end
I am currently using following command:
Code:
xtreg log_patents postfund i.year,fe vce(cluster company_id)