ETWFE on Baker.dta

[erinlbritton]

* gen time-varying treatment indicator
gen w = D // time-varying treatment indicator

* gen ever-treated cohort indicators
egen d = max(w), by(i)
sort i t

gen dinf = (d==0) // d-infinity = never-treated cohort
gen d86 = (gvar==1986)
gen d92 = (gvar==1992)
gen d98 = (gvar==1998)
gen d04 = (gvar==2004)

* gen t dummies
foreach t of num 1980/2009 {
	gen t`t' = (t==`t')
}

* leave out last treated cohort; no never treated units
reg Y c.d86#c.t1986 c.d86#c.t1987 c.d86#c.t1988 c.d86#c.t1989 c.d86#c.t1990 c.d86#c.t1991 c.d86#c.t1992 c.d86#c.t1993 c.d86#c.t1994 c.d86#c.t1995 c.d86#c.t1996 c.d86#c.t1997 c.d86#c.t1998 c.d86#c.t1999 c.d86#c.t2000 c.d86#c.t2001 c.d86#c.t2002 c.d86#c.t2003 c.d86#c.t2004 c.d86#c.t2005 c.d86#c.t2006 c.d86#c.t2007 c.d86#c.t2008 c.d86#c.t2009 ///
	c.d92#c.t1992 c.d92#c.t1993 c.d92#c.t1994 c.d92#c.t1995 c.d92#c.t1996 c.d92#c.t1997 c.d92#c.t1998 c.d92#c.t1999 c.d92#c.t2000 c.d92#c.t2001 c.d92#c.t2002 c.d92#c.t2003 c.d92#c.t2004 c.d92#c.t2005 c.d92#c.t2006 c.d92#c.t2007 c.d92#c.t2008 c.d92#c.t2009 ///
	c.d98#c.t1998 c.d98#c.t1999 c.d98#c.t2000 c.d98#c.t2001 c.d98#c.t2002 c.d98#c.t2003 c.d98#c.t2004 c.d98#c.t2005 c.d98#c.t2006 c.d98#c.t2007 c.d98#c.t2008 c.d98#c.t2009 ///
	d86 d92 d98 i.t, cluster(i)
	
*ATTs by cohort
lincom (c.d86#c.t1986+c.d86#c.t1987+c.d86#c.t1988+c.d86#c.t1989+c.d86#c.t1990+c.d86#c.t1991+c.d86#c.t1992+c.d86#c.t1993+c.d86#c.t1994+c.d86#c.t1995+c.d86#c.t1996+c.d86#c.t1997+c.d86#c.t1998+c.d86#c.t1999+c.d86#c.t2000+c.d86#c.t2001+c.d86#c.t2002+c.d86#c.t2003)/18
lincom (c.d92#c.t1992+c.d92#c.t1993+c.d92#c.t1994+c.d92#c.t1995+c.d92#c.t1996+c.d92#c.t1997+c.d92#c.t1998+c.d92#c.t1999+c.d92#c.t2000+c.d92#c.t2001+c.d92#c.t2002+c.d92#c.t2003)/12
lincom (c.d98#c.t1998+c.d98#c.t1999+c.d98#c.t2000+c.d98#c.t2001+c.d98#c.t2002+c.d98#c.t2003)/6

*ATTs by event-time
lincom (c.d86#c.t1986+c.d92#c.t1992+c.d98#c.t1998)/3  // 0
lincom (c.d86#c.t1987+c.d92#c.t1993+c.d98#c.t1999)/3  // 1
lincom (c.d86#c.t1988+c.d92#c.t1994+c.d98#c.t2000)/3  // 2
lincom (c.d86#c.t1989+c.d92#c.t1995+c.d98#c.t2001)/3  // 3
lincom (c.d86#c.t1990+c.d92#c.t1996+c.d98#c.t2002)/3  // 4
lincom (c.d86#c.t1991+c.d92#c.t1997+c.d98#c.t2003)/3  // 5
lincom (c.d86#c.t1992+c.d92#c.t1998)/2  // 6
lincom (c.d86#c.t1993+c.d92#c.t1999)/2  // 7
lincom (c.d86#c.t1994+c.d92#c.t2000)/2  // 8
lincom (c.d86#c.t1995+c.d92#c.t2001)/2  // 9
lincom (c.d86#c.t1996+c.d92#c.t2002)/2  // 10 
lincom (c.d86#c.t1997+c.d92#c.t2003)/2  // 11
lincom (c.d86#c.t1998)  // 12
lincom (c.d86#c.t1999)  // 13 
lincom (c.d86#c.t2000)  // 14
lincom (c.d86#c.t2001)  // 15
lincom (c.d86#c.t2002)  // 16
lincom (c.d86#c.t2003)  // 17

* pretrends; left out first period for each cohort: c.d86#c.t1980 c.d92#c.t1980 c.d98#c.t1980
reg Y c.d86#c.t1986 c.d86#c.t1987 c.d86#c.t1988 c.d86#c.t1989 c.d86#c.t1990 c.d86#c.t1991 c.d86#c.t1992 c.d86#c.t1993 c.d86#c.t1994 c.d86#c.t1995 c.d86#c.t1996 c.d86#c.t1997 c.d86#c.t1998 c.d86#c.t1999 c.d86#c.t2000 c.d86#c.t2001 c.d86#c.t2002 c.d86#c.t2003 c.d86#c.t2004 c.d86#c.t2005 c.d86#c.t2006 c.d86#c.t2007 c.d86#c.t2008 c.d86#c.t2009 ///
	c.d92#c.t1992 c.d92#c.t1993 c.d92#c.t1994 c.d92#c.t1995 c.d92#c.t1996 c.d92#c.t1997 c.d92#c.t1998 c.d92#c.t1999 c.d92#c.t2000 c.d92#c.t2001 c.d92#c.t2002 c.d92#c.t2003 c.d92#c.t2004 c.d92#c.t2005 c.d92#c.t2006 c.d92#c.t2007 c.d92#c.t2008 c.d92#c.t2009 ///
	c.d98#c.t1998 c.d98#c.t1999 c.d98#c.t2000 c.d98#c.t2001 c.d98#c.t2002 c.d98#c.t2003 c.d98#c.t2004 c.d98#c.t2005 c.d98#c.t2006 c.d98#c.t2007 c.d98#c.t2008 c.d98#c.t2009 ///
	c.d86#c.t1981 c.d86#c.t1982 c.d86#c.t1983 c.d86#c.t1984 c.d86#c.t1985 /// // pretrends
	c.d92#c.t1981 c.d92#c.t1982 c.d92#c.t1983 c.d92#c.t1984 c.d92#c.t1985 c.d92#c.t1986 c.d92#c.t1987 c.d92#c.t1988 c.d92#c.t1989 c.d92#c.t1990 c.d92#c.t1991 ///
	c.d98#c.t1981 c.d98#c.t1982 c.d98#c.t1983 c.d98#c.t1984 c.d98#c.t1985 c.d98#c.t1986 c.d98#c.t1987 c.d98#c.t1988 c.d98#c.t1989 c.d98#c.t1990 c.d98#c.t1991 c.d98#c.t1992 c.d98#c.t1993 c.d98#c.t1994 c.d98#c.t1995 c.d98#c.t1996 c.d98#c.t1997 /// 
	d86 d92 d98 i.t, cluster(i)

lincom (c.d86#c.t1985+c.d92#c.t1991+c.d98#c.t1997)/3 // -1
lincom (c.d86#c.t1984+c.d92#c.t1990+c.d98#c.t1996)/3 // -2
lincom (c.d86#c.t1983+c.d92#c.t1989+c.d98#c.t1995)/3 // -3
lincom (c.d86#c.t1982+c.d92#c.t1988+c.d98#c.t1994)/3 // -4
lincom (c.d86#c.t1981+c.d92#c.t1987+c.d98#c.t1993)/3 // -5
lincom (c.d92#c.t1986+c.d98#c.t1992)/2 // -6
lincom (c.d92#c.t1985+c.d98#c.t1991)/2 // -7
lincom (c.d92#c.t1984+c.d98#c.t1990)/2 // -8
lincom (c.d92#c.t1983+c.d98#c.t1989)/2 // -9
lincom (c.d92#c.t1982+c.d98#c.t1988)/2 // -10
lincom (c.d92#c.t1981+c.d98#c.t1987)/2 // -11
lincom (c.d98#c.t1986) // -12
lincom (c.d98#c.t1985) // -13
lincom (c.d98#c.t1984) // -14
lincom (c.d98#c.t1983) // -15
lincom (c.d98#c.t1982) // -16
lincom (c.d98#c.t1981) // -17

I am terrible at plotting in Stata, so I just took the ATTs from my Stata log and plotted them in R (for comparison with Scott Cunningham's substack post: