diff-in-diff with multiple treatment periods but questions asked in different waves

[Karen Arulsamy]

Hi all,

I am working on a project to understand the effect of the Covid-19 pandemic on health behaviours using Understanding Society with a diff-in-diff set up comparing male and female outcomes. Health behaviours (hindex_cutoff) is an index based on six variables (dfruit_simp dvege_simp fvamt_cutoff alchfreq alchbinge metminutesmv_cutoff) where three are asked in Jul 2020 & Jan 2021 and the rest in Sep 2020 & Jan 2021 i.e, there are two treatment periods depending on the variable. I have two questions:

1) I tried to create a variable where 1 = Pre, 2 = July/Sep 2020 and 3 = Jan 2021 but when running the estimations using this index, 2 is omitted. Is there a way around this?

2) Alternatively, I could run the regressions separately for the six variables for each time period but the sample size is changing quite a bit (I have tried experimenting with saving the estimation samples from the smallest available sample per outcome variable and using that but this doesn't seem to work). How can I ensure that the sample size is constant across the outcome variables and time periods?


**Treatment periods

gen July2020 = .
replace July2020 = 1 if wave == 15
replace July2020 = 0 if wave == 9

gen Sep2020 = .
replace Sep2020 = 1 if wave == 16
replace Sep2020 = 0 if wave == 9

gen Jan2021 = .
replace Jan2021 = 1 if wave == 18
replace Jan2021 = 0 if wave == 9


**Treatment vs. "control" groups

gen treatfem = .
replace treatfem = 1 if sex_main == 0
replace treatfem = 0 if sex_main == 1


keep if inlist(wave,9,15,16,17,18,19)
bys pidp: keep if _N == 6


xtreg hindex_cutoff i.treatfem##i.covidindicator i.dimonth_final2 i.dimonth_final3 i.dimonth_final4 i.dimonth_final5 i.dimonth_final6 i.dimonth_final7 i.dimonth_final8 i.dimonth_final9 i.dimonth_final10 ///
i.dimonth_final11 i.dimonth_final12 i.gor_main, fe vce(cluster pidp)

* Example generated by -dataex-. To install: ssc install dataex
clear
input float(hindex_cutoff dfruit_simp dvege_simp alchfreq metminutesmv_cutoff alchbinge treatfem covidindicator imonth_final gor_main) long pidp byte wave float(July2020 Sep2020 Jan2021)
. . . . . . 1 . 11  5    76165 17 . . .
. . . . . . 1 .  3  5    76165 19 . . .
. . . 1 . 1 1 2  9  5    76165 16 . 1 .
. 1 0 . . . 1 2  7  5    76165 15 1 . .
1 1 1 1 1 1 1 1  4  5    76165  9 0 0 0
1 1 0 1 1 1 1 3  1  5    76165 18 . . 1
. . . 0 1 1 1 2  9  1  1587125 16 . 1 .
. . . . . . 1 . 11  1  1587125 17 . . .
1 1 1 1 1 1 1 3  1  1  1587125 18 . . 1
. 0 0 . . . 1 2  7  1  1587125 15 1 . .
1 1 1 1 0 1 1 1  9  1  1587125  9 0 0 0
. . . . . . 1 .  3  1  1587125 19 . . .
. . . . . . 0 .  3 11  4849085 19 . . .
. 1 1 . . . 0 2  7 11  4849085 15 1 . .
. 1 1 0 . 1 0 3  1 11  4849085 18 . . 1
. . . 0 . 0 0 2  9 11  4849085 16 . 1 .
. . . . . . 0 . 11 11  4849085 17 . . .
0 0 0 0 0 1 0 1  4 11  4849085  9 0 0 0
1 0 0 1 1 1 1 3  1  7 68002725 18 . . 1
. . . . . . 1 .  3  7 68002725 19 . . .
0 0 1 1 0 1 1 1  3  7 68002725  9 0 0 0
. . . 0 1 1 1 2  9  7 68002725 16 . 1 .
. 0 0 . . . 1 2  7  7 68002725 15 1 . .
. . . . . . 1 . 11  7 68002725 17 . . .
. 1 1 . . . 1 2  7  1 68008847 15 1 . .
. . . . . . 1 .  3  1 68008847 19 . . .
1 1 0 1 0 1 1 1  3  1 68008847  9 0 0 0
. . . . . . 1 . 11  1 68008847 17 . . .
. 0 0 1 . 1 1 3  1  1 68008847 18 . . 1
. . . 1 0 1 1 2  9  1 68008847 16 . 1 .
. . . . . . 1 . 11  1 68010887 17 . . .
. . . 1 . 1 1 2  9  1 68010887 16 . 1 .
. . . . . . 1 .  3  1 68010887 19 . . .
1 1 1 1 1 1 1 1  3  1 68010887  9 0 0 0
. 0 1 1 . 1 1 3  1  1 68010887 18 . . 1
. 1 1 . . . 1 2  7  1 68010887 15 1 . .
. . . . . . 1 . 11  5 68031967 17 . . .
0 0 1 0 1 1 1 1  3  5 68031967  9 0 0 0
. 0 0 . . . 1 2  7  5 68031967 15 1 . .
. . . . . . 1 .  3  5 68031967 19 . . .
. . . 0 1 1 1 2  9  5 68031967 16 . 1 .
0 1 0 0 0 1 1 3  1  5 68031967 18 . . 1
. . . 1 1 1 0 2  9  7 68035365 16 . 1 .
1 0 0 1 1 1 0 3  1  7 68035365 18 . . 1
. . . . . . 0 .  3  7 68035365 19 . . .
. 1 0 . . . 0 2  7  7 68035365 15 1 . .
1 1 1 1 0 1 0 1  2  7 68035365  9 0 0 0
. . . . . . 0 . 11  7 68035365 17 . . .
. . . . . . 0 .  3  1 68035367 19 . . .
. 1 1 . . . 0 2  7  1 68035367 15 1 . .
end

Many thanks
Karen

[Jared Greathouse]

I'm not understanding the premise. What is your treatment here?

[Karen Arulsamy]

The treatment period is July 2020, September 2020 and January 2021 (compared against Wave 9: 2017-2019) and we are comparing females against males (the control group in this case). It is a little unusual in this case as everyone was treated i.e., they were all affected by the pandemic but this male female comparison has been done before in published papers.