did_demo
Qinlu.Hu
2022/11/1
A simulation demo
(1) create simulation data
time.periods <- 4
sp <- reset.sim()
sp$te <- 0 #create blank dataframe# set seed so everything is reproducible
set.seed(1814)
# generate dataset with 4 time periods
time.periods <- 4
# add dynamic effects
sp$te.e <- 1:time.periods
# generate data set with these parameters
# here, we dropped all units who are treated in time period 1 as they do not help us recover ATT(g,t)'s.
dta <- build_sim_dataset(sp)
# How many observations remained after dropping the ``always-treated'' units
nrow(dta)## [1] 15916#This is what the data looks like
dta## # A tibble: 15,916 × 7
## G X id cluster period Y treat
## <dbl> <dbl> <int> <int> <dbl> <dbl> <dbl>
## 1 3 -0.876 1 5 1 5.56 1
## 2 3 -0.876 1 5 2 4.35 1
## 3 3 -0.876 1 5 3 7.13 1
## 4 3 -0.876 1 5 4 6.24 1
## 5 2 -0.874 2 36 1 -3.66 1
## 6 2 -0.874 2 36 2 -1.27 1
## 7 2 -0.874 2 36 3 3.50 1
## 8 2 -0.874 2 36 4 -0.165 1
## 9 2 0.510 4 6 1 1.90 1
## 10 2 0.510 4 6 2 4.10 1
## # … with 15,906 more rows(2) Estimating Group-Time Average Treatment Effects
example_attgt <- att_gt(yname = "Y",
tname = "period",
idname = "id",
gname = "G",
xformla = ~X,
data = dta
)summary(example_attgt)##
## Call:
## att_gt(yname = "Y", tname = "period", idname = "id", gname = "G",
## xformla = ~X, data = dta)
##
## Reference: Callaway, Brantly and Pedro H.C. Sant'Anna. "Difference-in-Differences with Multiple Time Periods." Journal of Econometrics, Vol. 225, No. 2, pp. 200-230, 2021. <https://doi.org/10.1016/j.jeconom.2020.12.001>, <https://arxiv.org/abs/1803.09015>
##
## Group-Time Average Treatment Effects:
## Group Time ATT(g,t) Std. Error [95% Simult. Conf. Band]
## 2 2 0.9209 0.0618 0.7523 1.0896 *
## 2 3 1.9875 0.0676 1.8029 2.1720 *
## 2 4 2.9552 0.0626 2.7845 3.1259 *
## 3 2 -0.0433 0.0634 -0.2163 0.1297
## 3 3 1.1080 0.0675 0.9239 1.2922 *
## 3 4 2.0590 0.0649 1.8820 2.2360 *
## 4 2 0.0023 0.0643 -0.1732 0.1779
## 4 3 0.0615 0.0651 -0.1162 0.2393
## 4 4 0.9523 0.0696 0.7624 1.1423 *
## ---
## Signif. codes: `*' confidence band does not cover 0
##
## P-value for pre-test of parallel trends assumption: 0.60857
## Control Group: Never Treated, Anticipation Periods: 0
## Estimation Method: Doubly Robust(3) plot group-time average treatment effects
ggdid(example_attgt)
(4) Aggregating group-time average treatment effects
Method 1: Simple Aggregation (a weighted average of all group-time
average treatment effects with weights proportional to the group size.
This is available by calling the aggte function with
type = simple)
agg.simple <- aggte(example_attgt, type = "simple")
summary(agg.simple)##
## Call:
## aggte(MP = example_attgt, type = "simple")
##
## Reference: Callaway, Brantly and Pedro H.C. Sant'Anna. "Difference-in-Differences with Multiple Time Periods." Journal of Econometrics, Vol. 225, No. 2, pp. 200-230, 2021. <https://doi.org/10.1016/j.jeconom.2020.12.001>, <https://arxiv.org/abs/1803.09015>
##
##
## ATT Std. Error [ 95% Conf. Int.]
## 1.6583 0.0364 1.5869 1.7297 *
##
##
## ---
## Signif. codes: `*' confidence band does not cover 0
##
## Control Group: Never Treated, Anticipation Periods: 0
## Estimation Method: Doubly RobustMethod 2: Dynamic Effects and Event Studies( aggregate group-time effects into an event study plot)
agg.es <- aggte(example_attgt, type = "dynamic")
summary(agg.es)##
## Call:
## aggte(MP = example_attgt, type = "dynamic")
##
## Reference: Callaway, Brantly and Pedro H.C. Sant'Anna. "Difference-in-Differences with Multiple Time Periods." Journal of Econometrics, Vol. 225, No. 2, pp. 200-230, 2021. <https://doi.org/10.1016/j.jeconom.2020.12.001>, <https://arxiv.org/abs/1803.09015>
##
##
## Overall summary of ATT's based on event-study/dynamic aggregation:
## ATT Std. Error [ 95% Conf. Int.]
## 1.9904 0.0377 1.9165 2.0644 *
##
##
## Dynamic Effects:
## Event time Estimate Std. Error [95% Simult. Conf. Band]
## -2 0.0023 0.0672 -0.1784 0.1830
## -1 0.0105 0.0402 -0.0975 0.1185
## 0 0.9929 0.0302 0.9118 1.0740 *
## 1 2.0231 0.0438 1.9053 2.1410 *
## 2 2.9552 0.0628 2.7863 3.1241 *
## ---
## Signif. codes: `*' confidence band does not cover 0
##
## Control Group: Never Treated, Anticipation Periods: 0
## Estimation Method: Doubly Robustggdid(agg.es)
Method 3:Group-Specific Effects(average effects specific to each group)
agg.gs <- aggte(example_attgt, type = "group")
summary(agg.gs)##
## Call:
## aggte(MP = example_attgt, type = "group")
##
## Reference: Callaway, Brantly and Pedro H.C. Sant'Anna. "Difference-in-Differences with Multiple Time Periods." Journal of Econometrics, Vol. 225, No. 2, pp. 200-230, 2021. <https://doi.org/10.1016/j.jeconom.2020.12.001>, <https://arxiv.org/abs/1803.09015>
##
##
## Overall summary of ATT's based on group/cohort aggregation:
## ATT Std. Error [ 95% Conf. Int.]
## 1.488 0.0305 1.4283 1.5477 *
##
##
## Group Effects:
## Group Estimate Std. Error [95% Simult. Conf. Band]
## 2 1.9545 0.0509 1.8264 2.0827 *
## 3 1.5835 0.0568 1.4405 1.7265 *
## 4 0.9523 0.0656 0.7870 1.1176 *
## ---
## Signif. codes: `*' confidence band does not cover 0
##
## Control Group: Never Treated, Anticipation Periods: 0
## Estimation Method: Doubly Robustggdid(agg.gs)
Method 4: Calendar Time Effects(aggregations across different time periods)
agg.ct <- aggte(example_attgt, type = "calendar")
summary(agg.ct)##
## Call:
## aggte(MP = example_attgt, type = "calendar")
##
## Reference: Callaway, Brantly and Pedro H.C. Sant'Anna. "Difference-in-Differences with Multiple Time Periods." Journal of Econometrics, Vol. 225, No. 2, pp. 200-230, 2021. <https://doi.org/10.1016/j.jeconom.2020.12.001>, <https://arxiv.org/abs/1803.09015>
##
##
## Overall summary of ATT's based on calendar time aggregation:
## ATT Std. Error [ 95% Conf. Int.]
## 1.4808 0.0366 1.409 1.5526 *
##
##
## Time Effects:
## Time Estimate Std. Error [95% Simult. Conf. Band]
## 2 0.9209 0.0662 0.7645 1.0774 *
## 3 1.5491 0.0508 1.4291 1.6692 *
## 4 1.9724 0.0501 1.8538 2.0910 *
## ---
## Signif. codes: `*' confidence band does not cover 0
##
## Control Group: Never Treated, Anticipation Periods: 0
## Estimation Method: Doubly Robustggdid(agg.ct)
## (5) other options
chose control group
# By default, the did package uses the group of units that never participate in the treatment as the control group.
# The other option for the control group is to use the “not yet treated”
example_attgt_altcontrol <- att_gt(yname = "Y",
tname = "period",
idname = "id",
gname = "G",
xformla = ~X,
data = dta,
control_group = "notyettreated"
)
summary(example_attgt_altcontrol)##
## Call:
## att_gt(yname = "Y", tname = "period", idname = "id", gname = "G",
## xformla = ~X, data = dta, control_group = "notyettreated")
##
## Reference: Callaway, Brantly and Pedro H.C. Sant'Anna. "Difference-in-Differences with Multiple Time Periods." Journal of Econometrics, Vol. 225, No. 2, pp. 200-230, 2021. <https://doi.org/10.1016/j.jeconom.2020.12.001>, <https://arxiv.org/abs/1803.09015>
##
## Group-Time Average Treatment Effects:
## Group Time ATT(g,t) Std. Error [95% Simult. Conf. Band]
## 2 2 0.9371 0.0538 0.7852 1.0889 *
## 2 3 1.9612 0.0565 1.8019 2.1205 *
## 2 4 2.9552 0.0630 2.7776 3.1329 *
## 3 2 -0.0393 0.0537 -0.1906 0.1121
## 3 3 1.0747 0.0562 0.9163 1.2331 *
## 3 4 2.0590 0.0636 1.8795 2.2385 *
## 4 2 0.0207 0.0547 -0.1336 0.1751
## 4 3 0.0615 0.0631 -0.1165 0.2395
## 4 4 0.9523 0.0647 0.7698 1.1348 *
## ---
## Signif. codes: `*' confidence band does not cover 0
##
## P-value for pre-test of parallel trends assumption: 0.65089
## Control Group: Not Yet Treated, Anticipation Periods: 0
## Estimation Method: Doubly Robustchoose estimation method
# By default, the did package uses “doubly robust” estimators
# The other options are “ipw” for inverse probability weighting and “reg” # for regression
example_attgt_reg <- att_gt(yname = "Y",
tname = "period",
idname = "id",
gname = "G",
xformla = ~X,
data = dta,
est_method = "reg"
)
summary(example_attgt_reg)##
## Call:
## att_gt(yname = "Y", tname = "period", idname = "id", gname = "G",
## xformla = ~X, data = dta, est_method = "reg")
##
## Reference: Callaway, Brantly and Pedro H.C. Sant'Anna. "Difference-in-Differences with Multiple Time Periods." Journal of Econometrics, Vol. 225, No. 2, pp. 200-230, 2021. <https://doi.org/10.1016/j.jeconom.2020.12.001>, <https://arxiv.org/abs/1803.09015>
##
## Group-Time Average Treatment Effects:
## Group Time ATT(g,t) Std. Error [95% Simult. Conf. Band]
## 2 2 0.9191 0.0703 0.7317 1.1064 *
## 2 3 1.9861 0.0700 1.7998 2.1724 *
## 2 4 2.9537 0.0634 2.7848 3.1226 *
## 3 2 -0.0532 0.0688 -0.2365 0.1301
## 3 3 1.1106 0.0681 0.9291 1.2920 *
## 3 4 2.0609 0.0666 1.8836 2.2382 *
## 4 2 -0.0206 0.0732 -0.2156 0.1744
## 4 3 0.0672 0.0669 -0.1109 0.2452
## 4 4 0.9510 0.0673 0.7718 1.1302 *
## ---
## Signif. codes: `*' confidence band does not cover 0
##
## P-value for pre-test of parallel trends assumption: 0.64383
## Control Group: Never Treated, Anticipation Periods: 0
## Estimation Method: Outcome Regression