PS5
Grace Gao
2025-04-04
library(pacman)
p_load(wooldridge, multiwayvcov, causalweight, lmtest, did, usmap,
sandwich, AER, ivmodel, haven, estimatr, tidyverse, lubridate,
gridExtra, stringr, readxl, reshape2, scales, broom,
ggplot2, fixest, lfe, stargazer, foreign, ggthemes, ggforce,
ggridges, latex2exp, viridis, extrafont, kableExtra, snakecase,
janitor, tableone, causaldata, multcomp, etable)Problem 1
data(kielmc)
attach(kielmc)
as.data.frame(table(nearinc, y81))## nearinc y81 Freq
## 1 0 0 123
## 2 1 0 56
## 3 0 1 102
## 4 1 1 40summary(rprice)## Min. 1st Qu. Median Mean 3rd Qu. Max.
## 26000 59000 82000 83721 100230 300000length(unique(cbd))## [1] 34?kielmci. DID without covariates
kielmc <- kielmc %>%
mutate(post = ifelse(y81 == 1, 1, 0),
treat = ifelse(nearinc == 1, 1, 0),
interaction = post * treat)
model1 <- lm(rprice ~ post + treat + interaction, data = kielmc)
cluster_se1 <- cluster.vcov(model1, kielmc$cbd)
stargazer(model1, type = "text", se = list(sqrt(diag(cluster_se1))),
title = "Difference-in-Differences Estimation (With Covariates)",
dep.var.labels = "House Prices",
covariate.labels = c("Post", "Treat", "Interaction"),
out = "table1.txt")##
## Difference-in-Differences Estimation (With Covariates)
## ===============================================
## Dependent variable:
## ---------------------------
## House Prices
## -----------------------------------------------
## Post 18,790.290***
## (5,154.864)
##
## Treat -18,824.370**
## (7,796.294)
##
## Interaction -11,863.900*
## (6,621.818)
##
## Constant 82,517.230***
## (3,221.924)
##
## -----------------------------------------------
## Observations 321
## R2 0.174
## Adjusted R2 0.166
## Residual Std. Error 30,242.900 (df = 317)
## F Statistic 22.251*** (df = 3; 317)
## ===============================================
## Note: *p<0.1; **p<0.05; ***p<0.01The intercept tells us that during the pre-treatment period, houses that located farther away from the incinerators have a 82517.2 higher real housing price. The coefficient of post implies that the real housing price increase by 18790.3 after the installation of incinerator for all observations on average. The coefficient of treat says that houses that located near the incinerators have a 18824.4 lower real prices on average. All of these coefficients are statistically significant at 5%. The coefficient of interaction term shows that after the incinerators were installed, houses that located near the incinerators have real prices that are 11863.9 lower than those located farther away on average. However, this estimate is only statistically significant at 10%.
ii. DID with covariates
library(ipw)
# DID
model2 <- lm(rprice ~ post + treat + interaction + rooms + baths + age + agesq + larea + lland, data = kielmc)
cluster_se2 <- cluster.vcov(model2, kielmc$cbd)
stargazer(model2, type = "text", se = list(sqrt(diag(cluster_se2))),
title = "Difference-in-Differences Estimation (With Covariates)",
dep.var.labels = "House Prices",
covariate.labels = c("Post", "Treat", "Interaction", "Rooms", "Baths", "Age", "Age Squared", "Log Area", "Log Land"),
out = "table2.txt")##
## Difference-in-Differences Estimation (With Covariates)
## ===============================================
## Dependent variable:
## ---------------------------
## House Prices
## -----------------------------------------------
## Post 16,672.710***
## (3,714.861)
##
## Treat 15,554.710
## (10,224.750)
##
## Interaction -17,545.700***
## (5,821.574)
##
## Rooms 2,978.963**
## (1,401.381)
##
## Baths 7,458.220***
## (2,595.273)
##
## Age -616.245***
## (128.447)
##
## Age Squared 3.053***
## (0.776)
##
## Log Area 32,815.340***
## (8,487.841)
##
## Log Land 7,194.257
## (4,660.224)
##
## Constant -279,726.300***
## (93,485.450)
##
## -----------------------------------------------
## Observations 321
## R2 0.637
## Adjusted R2 0.627
## Residual Std. Error 20,235.080 (df = 311)
## F Statistic 60.690*** (df = 9; 311)
## ===============================================
## Note: *p<0.1; **p<0.05; ***p<0.01model3 <- lm(rprice ~ post + treat + interaction + rooms + baths + age + agesq + area + land, data = kielmc)
cluster_se3 <- cluster.vcov(model3, kielmc$cbd)
stargazer(model3, type = "text", se = list(sqrt(diag(cluster_se3))),
title = "Difference-in-Differences Estimation (With Covariates)",
dep.var.labels = "House Prices",
covariate.labels = c("Post", "Treat", "Interaction", "Rooms", "Baths", "Age", "Age Squared", "Area", "Land"),
out = "table3.txt")##
## Difference-in-Differences Estimation (With Covariates)
## ===============================================
## Dependent variable:
## ---------------------------
## House Prices
## -----------------------------------------------
## Post 15,210.440***
## (3,859.677)
##
## Treat 9,964.965
## (7,484.951)
##
## Interaction -15,020.680***
## (5,297.257)
##
## Rooms 3,280.511**
## (1,440.462)
##
## Baths 7,489.358***
## (2,725.948)
##
## Age -657.159***
## (136.595)
##
## Age Squared 3.083***
## (0.725)
##
## Area 17.830***
## (5.325)
##
## Land 0.116
## (0.105)
##
## Constant 2,189.075
## (9,333.944)
##
## -----------------------------------------------
## Observations 321
## R2 0.652
## Adjusted R2 0.642
## Residual Std. Error 19,823.800 (df = 311)
## F Statistic 64.683*** (df = 9; 311)
## ===============================================
## Note: *p<0.1; **p<0.05; ***p<0.01# DID using IPW with cluster bootstrap
covariates_model4 <- c("rooms", "baths", "age", "agesq", "larea", "lland")
model4 <- didweight(kielmc$rprice,
kielmc$nearinc,
kielmc$y81,
x = as.matrix(kielmc[, covariates_model4]),
boot = 1000,
cluster = kielmc$cbd)
treatment_effect_model4 <- as.numeric(model4$effect)
standard_error_model4 <- as.numeric(model4$se)
p_value_model4 <- as.numeric(model4$pvalue)
results_model4 <- data.frame(
Coefficient = c("Treatment Effect", "Standard Error", "P-value"),
Value = c(treatment_effect_model4, standard_error_model4, p_value_model4)
)
stargazer(results_model4, type = "text", summary = FALSE,
title = "Difference-in-Differences Estimation with IPW and Cluster Bootstrap",
out = "table4")##
## Difference-in-Differences Estimation with IPW and Cluster Bootstrap
## ==============================
## Coefficient Value
## ------------------------------
## 1 Treatment Effect -18,380.370
## 2 Standard Error 23,264.120
## 3 P-value 0.429
## ------------------------------covariates_model5 <- c("rooms", "baths", "age", "agesq", "area", "land")
model5 <- didweight(kielmc$rprice,
kielmc$nearinc,
kielmc$y81,
x = as.matrix(kielmc[, covariates_model5]),
boot = 1000,
cluster = kielmc$cbd)
treatment_effect_model5 <- as.numeric(model5$effect)
standard_error_model5 <- as.numeric(model5$se)
p_value_model5 <- as.numeric(model5$pvalue)
results_model5 <- data.frame(
Coefficient = c("Treatment Effect", "Standard Error", "P-value"),
Value = c(treatment_effect_model5, standard_error_model5, p_value_model5)
)
stargazer(results_model5, type = "text", summary = FALSE,
title = "Difference-in-Differences Estimation with IPW and Cluster Bootstrap",
out = "table5")##
## Difference-in-Differences Estimation with IPW and Cluster Bootstrap
## =============================
## Coefficient Value
## -----------------------------
## 1 Treatment Effect -5,896.836
## 2 Standard Error 10,622.170
## 3 P-value 0.579
## -----------------------------The sign of the treatment effect is consistent among all models and the negative sign intuitively makes sense. But the magnitudes and statistical significance of the treatment effects vary across models. Including covariates greatly increases the explanatory power of the model, but using larea&lland instead of area&land does not help much with capturing variations in rprice. Using IPW with cluster bootstrapping, the model is more robust against bias caused by comparability of treatment groups and gives more precise standard errors. So technically, I prefer the IPW model with area&land even though with a loss of statistical significance. But I’m not sure why the treatment effect in the model is much smaller compared to others.
iii. TWFE
As we see heterogeneous treatment effects across treated observations, approach 2 suits better for this analysis because approach assumes constant treatment effects within treated units.
library(modelsummary)
# Friend's data
i <- c(1, 1, 2, 2, 3, 3, 4, 4)
t <- c(1, 2, 1, 2, 1, 2, 1, 2)
P_t <- c(0, 1, 0, 1, 0, 1, 0, 1)
D_it <- c(0, 0, 0, 0, 0, 1, 0, 1)
D_i <- c(0, 0, 0, 0, 1, 1, 1, 1)
Y_it <- c(6, 7, 10, 9, 8, 14, 8, 12)
df <- data.frame(i=i, t=t, P_t=P_t, D_it=D_it, D_i=D_i, Y_it=Y_it)
rm(i, t, P_t, D_it, D_i, Y_it)
df <- df %>%
mutate(interaction = D_i * P_t)
# approach 1
model6 <- feols(Y_it ~ D_it + i + t, data = df)
summary(model6)## OLS estimation, Dep. Var.: Y_it
## Observations: 8
## Standard-errors: IID
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 6.833333 3.167763 2.157148 0.097199 .
## D_it 4.333333 2.173067 1.994109 0.116900
## i 0.333333 0.687184 0.485071 0.652994
## t 0.333333 1.611590 0.206835 0.846241
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## RMSE: 1.19024 Adj. R2: 0.599327# approach 2
model7 <- feols(Y_it ~ D_i + P_t + interaction, data = df)
summary(model7)## OLS estimation, Dep. Var.: Y_it
## Observations: 8
## Standard-errors: IID
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 8.00e+00 1.22474 6.531973e+00 0.0028378 **
## D_i 3.55e-15 1.73205 2.050000e-15 1.0000000
## P_t 3.55e-15 1.73205 2.050000e-15 1.0000000
## interaction 5.00e+00 2.44949 2.041241e+00 0.1107872
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## RMSE: 1.22474 Adj. R2: 0.575758# approach 3
model8 <- feols(Y_it ~ interaction + i + t, data = df)
summary(model8)## OLS estimation, Dep. Var.: Y_it
## Observations: 8
## Standard-errors: IID
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 6.833333 3.167763 2.157148 0.097199 .
## interaction 4.333333 2.173067 1.994109 0.116900
## i 0.333333 0.687184 0.485071 0.652994
## t 0.333333 1.611590 0.206835 0.846241
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## RMSE: 1.19024 Adj. R2: 0.599327models <- list("Approach 1" = model6, "Approach 2" = model7, "Approach 3" = model8)
modelsummary(models, stars = TRUE)| Approach 1 | Approach 2 | Approach 3 | |
|---|---|---|---|
| + p < 0.1, * p < 0.05, ** p < 0.01, *** p < 0.001 | |||
| (Intercept) | 6.833+ | 8.000** | 6.833+ |
| (3.168) | (1.225) | (3.168) | |
| D_it | 4.333 | ||
| (2.173) | |||
| i | 0.333 | 0.333 | |
| (0.687) | (0.687) | ||
| t | 0.333 | 0.333 | |
| (1.612) | (1.612) | ||
| D_i | 0.000 | ||
| (1.732) | |||
| P_t | 0.000 | ||
| (1.732) | |||
| interaction | 5.000 | 4.333 | |
| (2.449) | (2.173) | ||
| Num.Obs. | 8 | 8 | 8 |
| R2 | 0.771 | 0.758 | 0.771 |
| R2 Adj. | 0.599 | 0.576 | 0.599 |
| AIC | 33.5 | 33.9 | 33.5 |
| BIC | 33.8 | 34.3 | 33.8 |
| RMSE | 1.19 | 1.22 | 1.19 |
| Std.Errors | IID | IID | IID |
Approach 3 is preferred for this analysis because approach 3 has the same fit as approach 1 while including an interaction term that captures the heterogeneity in treatment effects.
Problem 2
library(did)
data(mpdta)i. Descriptive county-level maps
library(dplyr)
library(viridis)
mpdta <- mpdta %>% rename(fips = countyreal)
mpdta <- mpdta %>% mutate(treat = factor(treat, levels = c(0, 1), labels = c("Untreated", "Treated")))
mpdta <- mpdta %>% mutate(first.treat = factor(first.treat, levels = c(2004, 2006, 2007)))
plot_treat <- plot_usmap(data = mpdta, values = "treat", regions = "counties") +
scale_fill_manual(values = c("pink", "purple"), name = "Treatment") +
labs(title = "Minimum Wage Treatment Status by County") +
theme_minimal()
plot_treat
plot_first_treat <- plot_usmap(data = mpdta, values = "first.treat", regions = "counties") +
scale_fill_manual(values = c("orange", "purple", "pink"), name = "First Treatment Year") +
labs(title = "Year of Minimum Wage Implementation") +
theme_minimal()
plot_first_treat
lemp_summary <- mpdta %>%
filter(!is.na(first.treat)) %>%
group_by(fips, year, first.treat) %>%
summarise(mean_lemp = mean(lemp, na.rm = TRUE), .groups = "drop") %>%
mutate(fips = as.character(fips))
plot_lemp <- function(treat_year) {
cohort_data <- lemp_summary %>% filter(first.treat == treat_year)
cohort_data <- cohort_data %>% mutate(year = as.factor(year))
plot_usmap(data = cohort_data, values = "mean_lemp", regions = "counties", include = cohort_data$fips) +
scale_fill_viridis_c(option = "magma", name = "Avg. log(Teen Employment)", na.value = "gray80") +
facet_wrap(~"year", nrow = 2) +
labs(title = paste("Treatment Group:", treat_year)) +
theme_minimal()
}
map_0 <- plot_lemp(0)
map_2004 <- plot_lemp(2004)
map_2006 <- plot_lemp(2006)
map_2007 <- plot_lemp(2007)
map_0
map_2004
map_2006
map_2007
The treated counties distribute all over the US while the most
concentrated part is in the midwest. Among those treated, most counties
are treated in 2007 while the very few were treated in 2004.
From the map, I can’t really observe any changes of lemp over time for each group. It’s either too subtle or no changes.
ii. Cohort-year-specific average treatment effects on the treated
mpdta$first.treat <- as.numeric(as.character(mpdta$first.treat))
mpdta$first.treat[is.na(mpdta$first.treat)] <- 0
att1 <- att_gt(
yname = "lemp",
tname = "year",
idname = "fips",
gname = "first.treat",
xformla = ~lpop,
control_group = "nevertreated",
clustervars = "fips",
data = mpdta,
est_method = "reg"
)
summary(att1)##
## Call:
## att_gt(yname = "lemp", tname = "year", idname = "fips", gname = "first.treat",
## xformla = ~lpop, data = mpdta, control_group = "nevertreated",
## clustervars = "fips", 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]
## 2004 2004 -0.0149 0.0224 -0.0742 0.0443
## 2004 2005 -0.0770 0.0309 -0.1587 0.0047
## 2004 2006 -0.1411 0.0360 -0.2361 -0.0460 *
## 2004 2007 -0.1075 0.0345 -0.1987 -0.0164 *
## 2006 2004 -0.0021 0.0224 -0.0612 0.0571
## 2006 2005 -0.0070 0.0196 -0.0587 0.0447
## 2006 2006 0.0008 0.0192 -0.0499 0.0514
## 2006 2007 -0.0415 0.0202 -0.0949 0.0118
## 2007 2004 0.0264 0.0145 -0.0120 0.0647
## 2007 2005 -0.0048 0.0153 -0.0453 0.0358
## 2007 2006 -0.0285 0.0179 -0.0759 0.0189
## 2007 2007 -0.0288 0.0168 -0.0733 0.0157
## ---
## Signif. codes: `*' confidence band does not cover 0
##
## P-value for pre-test of parallel trends assumption: 0.23116
## Control Group: Never Treated, Anticipation Periods: 0
## Estimation Method: Outcome RegressionFor the 2006 and 2007 groups, the att are not statistically significant in all time periods. For the 2004 group, the att is negative and statistically significant in 2005, 2006, and 2007.
att2 <- att_gt(
yname = "lemp",
tname = "year",
idname = "fips",
gname = "first.treat",
xformla = ~lpop,
control_group = "nevertreated",
clustervars = "fips",
data = mpdta,
est_method = "ipw"
)
summary(att2)##
## Call:
## att_gt(yname = "lemp", tname = "year", idname = "fips", gname = "first.treat",
## xformla = ~lpop, data = mpdta, control_group = "nevertreated",
## clustervars = "fips", est_method = "ipw")
##
## 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]
## 2004 2004 -0.0145 0.0240 -0.0775 0.0484
## 2004 2005 -0.0764 0.0305 -0.1563 0.0034
## 2004 2006 -0.1405 0.0388 -0.2420 -0.0389 *
## 2004 2007 -0.1069 0.0337 -0.1953 -0.0186 *
## 2006 2004 -0.0009 0.0216 -0.0574 0.0557
## 2006 2005 -0.0064 0.0188 -0.0557 0.0429
## 2006 2006 0.0012 0.0192 -0.0492 0.0516
## 2006 2007 -0.0413 0.0192 -0.0916 0.0090
## 2007 2004 0.0266 0.0141 -0.0105 0.0636
## 2007 2005 -0.0047 0.0168 -0.0488 0.0394
## 2007 2006 -0.0283 0.0193 -0.0789 0.0223
## 2007 2007 -0.0289 0.0178 -0.0755 0.0177
## ---
## Signif. codes: `*' confidence band does not cover 0
##
## P-value for pre-test of parallel trends assumption: 0.23604
## Control Group: Never Treated, Anticipation Periods: 0
## Estimation Method: Inverse Probability WeightingHere the 2004 group is not getting significant att in 2005. Everything else stays pretty much similar to the last estimation.
att3 <- att_gt(
yname = "lemp",
tname = "year",
idname = "fips",
gname = "first.treat",
xformla = ~lpop,
control_group = "nevertreated",
clustervars = "fips",
data = mpdta,
est_method = "dr"
)
summary(att3)##
## Call:
## att_gt(yname = "lemp", tname = "year", idname = "fips", gname = "first.treat",
## xformla = ~lpop, data = mpdta, control_group = "nevertreated",
## clustervars = "fips", est_method = "dr")
##
## 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]
## 2004 2004 -0.0145 0.0228 -0.0744 0.0453
## 2004 2005 -0.0764 0.0291 -0.1527 -0.0001 *
## 2004 2006 -0.1404 0.0376 -0.2390 -0.0419 *
## 2004 2007 -0.1069 0.0361 -0.2015 -0.0123 *
## 2006 2004 -0.0005 0.0236 -0.0623 0.0614
## 2006 2005 -0.0062 0.0198 -0.0582 0.0458
## 2006 2006 0.0010 0.0198 -0.0509 0.0528
## 2006 2007 -0.0413 0.0202 -0.0944 0.0118
## 2007 2004 0.0267 0.0143 -0.0109 0.0643
## 2007 2005 -0.0046 0.0158 -0.0459 0.0367
## 2007 2006 -0.0284 0.0190 -0.0784 0.0215
## 2007 2007 -0.0288 0.0165 -0.0721 0.0146
## ---
## Signif. codes: `*' confidence band does not cover 0
##
## P-value for pre-test of parallel trends assumption: 0.23267
## Control Group: Never Treated, Anticipation Periods: 0
## Estimation Method: Doubly RobustThis one still produces similar results.
ggdid(att3)
I think the results are consistent with the maps. The effects are mostly
insignificant.
iii. Cohort-specific and aggregate average treatment effects on the treated
aggte <- aggte(att3, type = "group")
summary(aggte)##
## Call:
## aggte(MP = att3, 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.]
## -0.0328 0.0124 -0.0572 -0.0085 *
##
##
## Group Effects:
## Group Estimate Std. Error [95% Simult. Conf. Band]
## 2004 -0.0846 0.0266 -0.1415 -0.0276 *
## 2006 -0.0202 0.0178 -0.0581 0.0178
## 2007 -0.0288 0.0167 -0.0645 0.0069
## ---
## Signif. codes: `*' confidence band does not cover 0
##
## Control Group: Never Treated, Anticipation Periods: 0
## Estimation Method: Doubly RobustHere we are still seeing that only the 2004 group experienced significant effects over time after the treatment. Having the minimum wage policy decreased the employment rate of the 2004 group by about 8.5%.
# 1
library(twang)
mpdta$treat <- ifelse(mpdta$treat == 2, 1, mpdta$treat)
mpdta$treat <- ifelse(mpdta$treat == 1, 0, mpdta$treat)
mpdta$treat <- ifelse(mpdta$treat == 2, 1, mpdta$treat)
mpdta$treat <- as.numeric(mpdta$treat)
ps <- ps(treat ~ year + lpop + treat, data = mpdta, stop.method = "es.mean")## Fitting boosted model
## Iter TrainDeviance ValidDeviance StepSize Improve
## 1 1.3102 nan 0.0100 nan
## 2 1.2907 nan 0.0100 nan
## 3 1.2717 nan 0.0100 nan
## 4 1.2530 nan 0.0100 nan
## 5 1.2348 nan 0.0100 nan
## 6 1.2170 nan 0.0100 nan
## 7 1.1995 nan 0.0100 nan
## 8 1.1824 nan 0.0100 nan
## 9 1.1656 nan 0.0100 nan
## 10 1.1491 nan 0.0100 nan
## 20 1.0010 nan 0.0100 nan
## 40 0.7727 nan 0.0100 nan
## 60 0.6058 nan 0.0100 nan
## 80 0.4800 nan 0.0100 nan
## 100 0.3831 nan 0.0100 nan
## 120 0.3075 nan 0.0100 nan
## 140 0.2478 nan 0.0100 nan
## 160 0.2004 nan 0.0100 nan
## 180 0.1624 nan 0.0100 nan
## 200 0.1319 nan 0.0100 nan
## 220 0.1073 nan 0.0100 nan
## 240 0.0874 nan 0.0100 nan
## 260 0.0712 nan 0.0100 nan
## 280 0.0581 nan 0.0100 nan
## 300 0.0474 nan 0.0100 nan
## 320 0.0388 nan 0.0100 nan
## 340 0.0317 nan 0.0100 nan
## 360 0.0259 nan 0.0100 nan
## 380 0.0212 nan 0.0100 nan
## 400 0.0173 nan 0.0100 nan
## 420 0.0142 nan 0.0100 nan
## 440 0.0116 nan 0.0100 nan
## 460 0.0095 nan 0.0100 nan
## 480 0.0078 nan 0.0100 nan
## 500 0.0064 nan 0.0100 nan
## 520 0.0052 nan 0.0100 nan
## 540 0.0043 nan 0.0100 nan
## 560 0.0035 nan 0.0100 nan
## 580 0.0029 nan 0.0100 nan
## 600 0.0023 nan 0.0100 nan
## 620 0.0019 nan 0.0100 nan
## 640 0.0016 nan 0.0100 nan
## 660 0.0013 nan 0.0100 nan
## 680 0.0010 nan 0.0100 nan
## 700 0.0009 nan 0.0100 nan
## 720 0.0007 nan 0.0100 nan
## 740 0.0006 nan 0.0100 nan
## 760 0.0005 nan 0.0100 nan
## 780 0.0004 nan 0.0100 nan
## 800 0.0003 nan 0.0100 nan
## 820 0.0003 nan 0.0100 nan
## 840 0.0002 nan 0.0100 nan
## 860 0.0002 nan 0.0100 nan
## 880 0.0001 nan 0.0100 nan
## 900 0.0001 nan 0.0100 nan
## 920 0.0001 nan 0.0100 nan
## 940 0.0001 nan 0.0100 nan
## 960 0.0001 nan 0.0100 nan
## 980 0.0001 nan 0.0100 nan
## 1000 0.0000 nan 0.0100 nan
## 1020 0.0000 nan 0.0100 nan
## 1040 0.0000 nan 0.0100 nan
## 1060 0.0000 nan 0.0100 nan
## 1080 0.0000 nan 0.0100 nan
## 1100 0.0000 nan 0.0100 nan
## 1120 0.0000 nan 0.0100 nan
## 1140 0.0000 nan 0.0100 nan
## 1160 0.0000 nan 0.0100 nan
## 1180 0.0000 nan 0.0100 nan
## 1200 0.0000 nan 0.0100 nan
## 1220 0.0000 nan 0.0100 nan
## 1240 0.0000 nan 0.0100 nan
## 1260 0.0000 nan 0.0100 nan
## 1280 0.0000 nan 0.0100 nan
## 1300 0.0000 nan 0.0100 nan
## 1320 0.0000 nan 0.0100 nan
## 1340 0.0000 nan 0.0100 nan
## 1360 0.0000 nan 0.0100 nan
## 1380 0.0000 nan 0.0100 nan
## 1400 0.0000 nan 0.0100 nan
## 1420 0.0000 nan 0.0100 nan
## 1440 0.0000 nan 0.0100 nan
## 1460 0.0000 nan 0.0100 nan
## 1480 0.0000 nan 0.0100 nan
## 1500 0.0000 nan 0.0100 nan
## 1520 0.0000 nan 0.0100 nan
## 1540 0.0000 nan 0.0100 nan
## 1560 0.0000 nan 0.0100 nan
## 1580 0.0000 nan 0.0100 nan
## 1600 0.0000 nan 0.0100 nan
## 1620 0.0000 nan 0.0100 nan
## 1640 0.0000 nan 0.0100 nan
## 1660 0.0000 nan 0.0100 nan
## 1680 0.0000 nan 0.0100 nan
## 1700 0.0000 nan 0.0100 nan
## 1720 0.0000 nan 0.0100 nan
## 1740 0.0000 nan 0.0100 nan
## 1760 0.0000 nan 0.0100 nan
## 1780 0.0000 nan 0.0100 nan
## 1800 0.0000 nan 0.0100 nan
## 1820 0.0000 nan 0.0100 nan
## 1840 0.0000 nan 0.0100 nan
## 1860 0.0000 nan 0.0100 nan
## 1880 0.0000 nan 0.0100 nan
## 1900 0.0000 nan 0.0100 nan
## 1920 0.0000 nan 0.0100 nan
## 1940 0.0000 nan 0.0100 nan
## 1960 0.0000 nan 0.0100 nan
## 1980 0.0000 nan 0.0100 nan
## 2000 0.0000 nan 0.0100 nan
## 2020 0.0000 nan 0.0100 nan
## 2040 0.0000 nan 0.0100 nan
## 2060 0.0000 nan 0.0100 nan
## 2080 0.0000 nan 0.0100 nan
## 2100 0.0000 nan 0.0100 nan
## 2120 0.0000 nan 0.0100 nan
## 2140 0.0000 nan 0.0100 nan
## 2160 0.0000 nan 0.0100 nan
## 2180 0.0000 nan 0.0100 nan
## 2200 0.0000 nan 0.0100 nan
## 2220 0.0000 nan 0.0100 nan
## 2240 0.0000 nan 0.0100 nan
## 2260 0.0000 nan 0.0100 nan
## 2280 0.0000 nan 0.0100 nan
## 2300 0.0000 nan 0.0100 nan
## 2320 0.0000 nan 0.0100 nan
## 2340 0.0000 nan 0.0100 nan
## 2360 0.0000 nan 0.0100 nan
## 2380 0.0000 nan 0.0100 nan
## 2400 0.0000 nan 0.0100 nan
## 2420 0.0000 nan 0.0100 nan
## 2440 0.0000 nan 0.0100 nan
## 2460 0.0000 nan 0.0100 nan
## 2480 0.0000 nan 0.0100 nan
## 2500 0.0000 nan 0.0100 nan
## 2520 0.0000 nan 0.0100 nan
## 2540 0.0000 nan 0.0100 nan
## 2560 0.0000 nan 0.0100 nan
## 2580 0.0000 nan 0.0100 nan
## 2600 0.0000 nan 0.0100 nan
## 2620 0.0000 nan 0.0100 nan
## 2640 0.0000 nan 0.0100 nan
## 2660 0.0000 nan 0.0100 nan
## 2680 0.0000 nan 0.0100 nan
## 2700 0.0000 nan 0.0100 nan
## 2720 0.0000 nan 0.0100 nan
## 2740 0.0000 nan 0.0100 nan
## 2760 0.0000 nan 0.0100 nan
## 2780 0.0000 nan 0.0100 nan
## 2800 0.0000 nan 0.0100 nan
## 2820 0.0000 nan 0.0100 nan
## 2840 0.0000 nan 0.0100 nan
## 2860 0.0000 nan 0.0100 nan
## 2880 0.0000 nan 0.0100 nan
## 2900 0.0000 nan 0.0100 nan
## 2920 0.0000 nan 0.0100 nan
## 2940 0.0000 nan 0.0100 nan
## 2960 0.0000 nan 0.0100 nan
## 2980 0.0000 nan 0.0100 nan
## 3000 0.0000 nan 0.0100 nan
## 3020 0.0000 nan 0.0100 nan
## 3040 0.0000 nan 0.0100 nan
## 3060 0.0000 nan 0.0100 nan
## 3080 0.0000 nan 0.0100 nan
## 3100 0.0000 nan 0.0100 nan
## 3120 0.0000 nan 0.0100 nan
## 3140 0.0000 nan 0.0100 nan
## 3160 0.0000 nan 0.0100 nan
## 3180 0.0000 nan 0.0100 nan
## 3200 0.0000 nan 0.0100 nan
## 3220 0.0000 nan 0.0100 nan
## 3240 0.0000 nan 0.0100 nan
## 3260 0.0000 nan 0.0100 nan
## 3280 0.0000 nan 0.0100 nan
## 3300 0.0000 nan 0.0100 nan
## 3320 0.0000 nan 0.0100 nan
## 3340 0.0000 nan 0.0100 nan
## 3360 0.0000 nan 0.0100 nan
## 3380 0.0000 nan 0.0100 nan
## 3400 0.0000 nan 0.0100 nan
## 3420 0.0000 nan 0.0100 nan
## 3440 0.0000 nan 0.0100 nan
## 3460 0.0000 nan 0.0100 nan
## 3480 0.0000 nan 0.0100 nan
## 3500 0.0000 nan 0.0100 nan
## 3520 0.0000 nan 0.0100 nan
## 3540 0.0000 nan 0.0100 nan
## 3560 0.0000 nan 0.0100 nan
## 3580 0.0000 nan 0.0100 nan
## 3600 0.0000 nan 0.0100 nan
## 3620 0.0000 nan 0.0100 nan
## 3640 0.0000 nan 0.0100 nan
## 3660 0.0000 nan 0.0100 nan
## 3680 0.0000 nan 0.0100 nan
## 3700 0.0000 nan 0.0100 nan
## 3720 0.0000 nan 0.0100 nan
## 3740 0.0000 nan 0.0100 nan
## 3760 0.0000 nan 0.0100 nan
## 3780 0.0000 nan 0.0100 nan
## 3800 0.0000 nan 0.0100 nan
## 3820 0.0000 nan 0.0100 nan
## 3840 0.0000 nan 0.0100 nan
## 3860 0.0000 nan 0.0100 nan
## 3880 0.0000 nan 0.0100 nan
## 3900 0.0000 nan 0.0100 nan
## 3920 0.0000 nan 0.0100 nan
## 3940 0.0000 nan 0.0100 nan
## 3960 0.0000 nan 0.0100 nan
## 3980 0.0000 nan 0.0100 nan
## 4000 0.0000 nan 0.0100 nan
## 4020 0.0000 nan 0.0100 nan
## 4040 0.0000 nan 0.0100 nan
## 4060 0.0000 nan 0.0100 nan
## 4080 0.0000 nan 0.0100 nan
## 4100 0.0000 nan 0.0100 nan
## 4120 0.0000 nan 0.0100 nan
## 4140 0.0000 nan 0.0100 nan
## 4160 0.0000 nan 0.0100 nan
## 4180 0.0000 nan 0.0100 nan
## 4200 0.0000 nan 0.0100 nan
## 4220 0.0000 nan 0.0100 nan
## 4240 0.0000 nan 0.0100 nan
## 4260 0.0000 nan 0.0100 nan
## 4280 0.0000 nan 0.0100 nan
## 4300 0.0000 nan 0.0100 nan
## 4320 0.0000 nan 0.0100 nan
## 4340 0.0000 nan 0.0100 nan
## 4360 0.0000 nan 0.0100 nan
## 4380 0.0000 nan 0.0100 nan
## 4400 0.0000 nan 0.0100 nan
## 4420 0.0000 nan 0.0100 nan
## 4440 0.0000 nan 0.0100 nan
## 4460 0.0000 nan 0.0100 nan
## 4480 0.0000 nan 0.0100 nan
## 4500 0.0000 nan 0.0100 nan
## 4520 0.0000 nan 0.0100 nan
## 4540 0.0000 nan 0.0100 nan
## 4560 0.0000 nan 0.0100 nan
## 4580 0.0000 nan 0.0100 nan
## 4600 0.0000 nan 0.0100 nan
## 4620 0.0000 nan 0.0100 nan
## 4640 0.0000 nan 0.0100 nan
## 4660 0.0000 nan 0.0100 nan
## 4680 0.0000 nan 0.0100 nan
## 4700 0.0000 nan 0.0100 nan
## 4720 0.0000 nan 0.0100 nan
## 4740 0.0000 nan 0.0100 nan
## 4760 0.0000 nan 0.0100 nan
## 4780 0.0000 nan 0.0100 nan
## 4800 0.0000 nan 0.0100 nan
## 4820 0.0000 nan 0.0100 nan
## 4840 0.0000 nan 0.0100 nan
## 4860 0.0000 nan 0.0100 nan
## 4880 0.0000 nan 0.0100 nan
## 4900 0.0000 nan 0.0100 nan
## 4920 0.0000 nan 0.0100 nan
## 4940 0.0000 nan 0.0100 nan
## 4960 0.0000 nan 0.0100 nan
## 4980 0.0000 nan 0.0100 nan
## 5000 0.0000 nan 0.0100 nan
## 5020 0.0000 nan 0.0100 nan
## 5040 0.0000 nan 0.0100 nan
## 5060 0.0000 nan 0.0100 nan
## 5080 0.0000 nan 0.0100 nan
## 5100 0.0000 nan 0.0100 nan
## 5120 0.0000 nan 0.0100 nan
## 5140 0.0000 nan 0.0100 nan
## 5160 0.0000 nan 0.0100 nan
## 5180 0.0000 nan 0.0100 nan
## 5200 0.0000 nan 0.0100 nan
## 5220 0.0000 nan 0.0100 nan
## 5240 0.0000 nan 0.0100 nan
## 5260 0.0000 nan 0.0100 nan
## 5280 0.0000 nan 0.0100 nan
## 5300 0.0000 nan 0.0100 nan
## 5320 0.0000 nan 0.0100 nan
## 5340 0.0000 nan 0.0100 nan
## 5360 0.0000 nan 0.0100 nan
## 5380 0.0000 nan 0.0100 nan
## 5400 0.0000 nan 0.0100 nan
## 5420 0.0000 nan 0.0100 nan
## 5440 0.0000 nan 0.0100 nan
## 5460 0.0000 nan 0.0100 nan
## 5480 0.0000 nan 0.0100 nan
## 5500 0.0000 nan 0.0100 nan
## 5520 0.0000 nan 0.0100 nan
## 5540 0.0000 nan 0.0100 nan
## 5560 0.0000 nan 0.0100 nan
## 5580 0.0000 nan 0.0100 nan
## 5600 0.0000 nan 0.0100 nan
## 5620 0.0000 nan 0.0100 nan
## 5640 0.0000 nan 0.0100 nan
## 5660 0.0000 nan 0.0100 nan
## 5680 0.0000 nan 0.0100 nan
## 5700 0.0000 nan 0.0100 nan
## 5720 0.0000 nan 0.0100 nan
## 5740 0.0000 nan 0.0100 nan
## 5760 0.0000 nan 0.0100 nan
## 5780 0.0000 nan 0.0100 nan
## 5800 0.0000 nan 0.0100 nan
## 5820 0.0000 nan 0.0100 nan
## 5840 0.0000 nan 0.0100 nan
## 5860 0.0000 nan 0.0100 nan
## 5880 0.0000 nan 0.0100 nan
## 5900 0.0000 nan 0.0100 nan
## 5920 0.0000 nan 0.0100 nan
## 5940 0.0000 nan 0.0100 nan
## 5960 0.0000 nan 0.0100 nan
## 5980 0.0000 nan 0.0100 nan
## 6000 0.0000 nan 0.0100 nan
## 6020 0.0000 nan 0.0100 nan
## 6040 0.0000 nan 0.0100 nan
## 6060 0.0000 nan 0.0100 nan
## 6080 0.0000 nan 0.0100 nan
## 6100 0.0000 nan 0.0100 nan
## 6120 0.0000 nan 0.0100 nan
## 6140 0.0000 nan 0.0100 nan
## 6160 0.0000 nan 0.0100 nan
## 6180 0.0000 nan 0.0100 nan
## 6200 0.0000 nan 0.0100 nan
## 6220 0.0000 nan 0.0100 nan
## 6240 0.0000 nan 0.0100 nan
## 6260 0.0000 nan 0.0100 nan
## 6280 0.0000 nan 0.0100 nan
## 6300 0.0000 nan 0.0100 nan
## 6320 0.0000 nan 0.0100 nan
## 6340 0.0000 nan 0.0100 nan
## 6360 0.0000 nan 0.0100 nan
## 6380 0.0000 nan 0.0100 nan
## 6400 0.0000 nan 0.0100 nan
## 6420 0.0000 nan 0.0100 nan
## 6440 0.0000 nan 0.0100 nan
## 6460 0.0000 nan 0.0100 nan
## 6480 0.0000 nan 0.0100 nan
## 6500 0.0000 nan 0.0100 nan
## 6520 0.0000 nan 0.0100 nan
## 6540 0.0000 nan 0.0100 nan
## 6560 0.0000 nan 0.0100 nan
## 6580 0.0000 nan 0.0100 nan
## 6600 0.0000 nan 0.0100 nan
## 6620 0.0000 nan 0.0100 nan
## 6640 0.0000 nan 0.0100 nan
## 6660 0.0000 nan 0.0100 nan
## 6680 0.0000 nan 0.0100 nan
## 6700 0.0000 nan 0.0100 nan
## 6720 0.0000 nan 0.0100 nan
## 6740 0.0000 nan 0.0100 nan
## 6760 0.0000 nan 0.0100 nan
## 6780 0.0000 nan 0.0100 nan
## 6800 0.0000 nan 0.0100 nan
## 6820 0.0000 nan 0.0100 nan
## 6840 0.0000 nan 0.0100 nan
## 6860 0.0000 nan 0.0100 nan
## 6880 0.0000 nan 0.0100 nan
## 6900 0.0000 nan 0.0100 nan
## 6920 0.0000 nan 0.0100 nan
## 6940 0.0000 nan 0.0100 nan
## 6960 0.0000 nan 0.0100 nan
## 6980 0.0000 nan 0.0100 nan
## 7000 0.0000 nan 0.0100 nan
## 7020 0.0000 nan 0.0100 nan
## 7040 0.0000 nan 0.0100 nan
## 7060 0.0000 nan 0.0100 nan
## 7080 0.0000 nan 0.0100 nan
## 7100 0.0000 nan 0.0100 nan
## 7120 0.0000 nan 0.0100 nan
## 7140 0.0000 nan 0.0100 nan
## 7160 0.0000 nan 0.0100 nan
## 7180 0.0000 nan 0.0100 nan
## 7200 0.0000 nan 0.0100 nan
## 7220 0.0000 nan 0.0100 nan
## 7240 0.0000 nan 0.0100 nan
## 7260 0.0000 nan 0.0100 nan
## 7280 0.0000 nan 0.0100 nan
## 7300 0.0000 nan 0.0100 nan
## 7320 0.0000 nan 0.0100 nan
## 7340 0.0000 nan 0.0100 nan
## 7360 0.0000 nan 0.0100 nan
## 7380 0.0000 nan 0.0100 nan
## 7400 0.0000 nan 0.0100 nan
## 7420 0.0000 nan 0.0100 nan
## 7440 0.0000 nan 0.0100 nan
## 7460 0.0000 nan 0.0100 nan
## 7480 0.0000 nan 0.0100 nan
## 7500 0.0000 nan 0.0100 nan
## 7520 0.0000 nan 0.0100 nan
## 7540 0.0000 nan 0.0100 nan
## 7560 0.0000 nan 0.0100 nan
## 7580 0.0000 nan 0.0100 nan
## 7600 0.0000 nan 0.0100 nan
## 7620 0.0000 nan 0.0100 nan
## 7640 0.0000 nan 0.0100 nan
## 7660 0.0000 nan 0.0100 nan
## 7680 0.0000 nan 0.0100 nan
## 7700 0.0000 nan 0.0100 nan
## 7720 0.0000 nan 0.0100 nan
## 7740 0.0000 nan 0.0100 nan
## 7760 0.0000 nan 0.0100 nan
## 7780 0.0000 nan 0.0100 nan
## 7800 0.0000 nan 0.0100 nan
## 7820 0.0000 nan 0.0100 nan
## 7840 0.0000 nan 0.0100 nan
## 7860 0.0000 nan 0.0100 nan
## 7880 0.0000 nan 0.0100 nan
## 7900 0.0000 nan 0.0100 nan
## 7920 0.0000 nan 0.0100 nan
## 7940 0.0000 nan 0.0100 nan
## 7960 0.0000 nan 0.0100 nan
## 7980 0.0000 nan 0.0100 nan
## 8000 0.0000 nan 0.0100 nan
## 8020 0.0000 nan 0.0100 nan
## 8040 0.0000 nan 0.0100 nan
## 8060 0.0000 nan 0.0100 nan
## 8080 0.0000 nan 0.0100 nan
## 8100 0.0000 nan 0.0100 nan
## 8120 0.0000 nan 0.0100 nan
## 8140 0.0000 nan 0.0100 nan
## 8160 0.0000 nan 0.0100 nan
## 8180 0.0000 nan 0.0100 nan
## 8200 0.0000 nan 0.0100 nan
## 8220 0.0000 nan 0.0100 nan
## 8240 0.0000 nan 0.0100 nan
## 8260 0.0000 nan 0.0100 nan
## 8280 0.0000 nan 0.0100 nan
## 8300 0.0000 nan 0.0100 nan
## 8320 0.0000 nan 0.0100 nan
## 8340 0.0000 nan 0.0100 nan
## 8360 0.0000 nan 0.0100 nan
## 8380 0.0000 nan 0.0100 nan
## 8400 0.0000 nan 0.0100 nan
## 8420 0.0000 nan 0.0100 nan
## 8440 0.0000 nan 0.0100 nan
## 8460 0.0000 nan 0.0100 nan
## 8480 0.0000 nan 0.0100 nan
## 8500 0.0000 nan 0.0100 nan
## 8520 0.0000 nan 0.0100 nan
## 8540 0.0000 nan 0.0100 nan
## 8560 0.0000 nan 0.0100 nan
## 8580 0.0000 nan 0.0100 nan
## 8600 0.0000 nan 0.0100 nan
## 8620 0.0000 nan 0.0100 nan
## 8640 0.0000 nan 0.0100 nan
## 8660 0.0000 nan 0.0100 nan
## 8680 0.0000 nan 0.0100 nan
## 8700 0.0000 nan 0.0100 nan
## 8720 0.0000 nan 0.0100 nan
## 8740 0.0000 nan 0.0100 nan
## 8760 0.0000 nan 0.0100 nan
## 8780 0.0000 nan 0.0100 nan
## 8800 0.0000 nan 0.0100 nan
## 8820 0.0000 nan 0.0100 nan
## 8840 0.0000 nan 0.0100 nan
## 8860 0.0000 nan 0.0100 nan
## 8880 0.0000 nan 0.0100 nan
## 8900 0.0000 nan 0.0100 nan
## 8920 0.0000 nan 0.0100 nan
## 8940 0.0000 nan 0.0100 nan
## 8960 0.0000 nan 0.0100 nan
## 8980 0.0000 nan 0.0100 nan
## 9000 0.0000 nan 0.0100 nan
## 9020 0.0000 nan 0.0100 nan
## 9040 0.0000 nan 0.0100 nan
## 9060 0.0000 nan 0.0100 nan
## 9080 0.0000 nan 0.0100 nan
## 9100 0.0000 nan 0.0100 nan
## 9120 0.0000 nan 0.0100 nan
## 9140 0.0000 nan 0.0100 nan
## 9160 0.0000 nan 0.0100 nan
## 9180 0.0000 nan 0.0100 nan
## 9200 0.0000 nan 0.0100 nan
## 9220 0.0000 nan 0.0100 nan
## 9240 0.0000 nan 0.0100 nan
## 9260 0.0000 nan 0.0100 nan
## 9280 0.0000 nan 0.0100 nan
## 9300 0.0000 nan 0.0100 nan
## 9320 0.0000 nan 0.0100 nan
## 9340 0.0000 nan 0.0100 nan
## 9360 0.0000 nan 0.0100 nan
## 9380 0.0000 nan 0.0100 nan
## 9400 0.0000 nan 0.0100 nan
## 9420 0.0000 nan 0.0100 nan
## 9440 0.0000 nan 0.0100 nan
## 9460 0.0000 nan 0.0100 nan
## 9480 0.0000 nan 0.0100 nan
## 9500 0.0000 nan 0.0100 nan
## 9520 0.0000 nan 0.0100 nan
## 9540 0.0000 nan 0.0100 nan
## 9560 0.0000 nan 0.0100 nan
## 9580 0.0000 nan 0.0100 nan
## 9600 0.0000 nan 0.0100 nan
## 9620 0.0000 nan 0.0100 nan
## 9640 0.0000 nan 0.0100 nan
## 9660 0.0000 nan 0.0100 nan
## 9680 0.0000 nan 0.0100 nan
## 9700 0.0000 nan 0.0100 nan
## 9720 0.0000 nan 0.0100 nan
## 9740 0.0000 nan 0.0100 nan
## 9760 0.0000 nan 0.0100 nan
## 9780 0.0000 nan 0.0100 nan
## 9800 0.0000 nan 0.0100 nan
## 9820 0.0000 nan 0.0100 nan
## 9840 0.0000 nan 0.0100 nan
## 9860 0.0000 nan 0.0100 nan
## 9880 0.0000 nan 0.0100 nan
## 9900 0.0000 nan 0.0100 nan
## 9920 0.0000 nan 0.0100 nan
## 9940 0.0000 nan 0.0100 nan
## 9960 0.0000 nan 0.0100 nan
## 9980 0.0000 nan 0.0100 nan
## 10000 0.0000 nan 0.0100 nan
##
## Diagnosis of unweighted analysis
## Calculating standardized differences
## Optimizing with es.mean.ATE stopping rule
## Optimized at 2999
## Diagnosis of es.mean.ATE weightsatt4 <- att_gt(yname = "lemp", tname = "year", idname = "fips",
gname = "first.treat", xformla = ~lpop,
control_group = "nevertreated", clustervars = "fips",
data = mpdta, est_method = "dr",
weights = ps$weights)
summary(att4)##
## Call:
## att_gt(yname = "lemp", tname = "year", idname = "fips", gname = "first.treat",
## xformla = ~lpop, data = mpdta, control_group = "nevertreated",
## weightsname = ps$weights, clustervars = "fips", est_method = "dr")
##
## Reference: Callaway, Brantly and Pedro H.C. Sant'Anna. "Difference-in-Differences with Multiple Time Periods." Journal of Econometrics, Vol. 225, No. 2, pp. 200-230, 2021. <https://doi.org/10.1016/j.jeconom.2020.12.001>, <https://arxiv.org/abs/1803.09015>
##
## Group-Time Average Treatment Effects:
## Group Time ATT(g,t) Std. Error [95% Simult. Conf. Band]
## 2004 2004 -0.0145 0.0223 -0.0730 0.0439
## 2004 2005 -0.0764 0.0285 -0.1514 -0.0015 *
## 2004 2006 -0.1404 0.0366 -0.2365 -0.0444 *
## 2004 2007 -0.1069 0.0324 -0.1919 -0.0219 *
## 2006 2004 -0.0005 0.0233 -0.0618 0.0609
## 2006 2005 -0.0062 0.0188 -0.0557 0.0433
## 2006 2006 0.0010 0.0188 -0.0485 0.0504
## 2006 2007 -0.0413 0.0204 -0.0948 0.0122
## 2007 2004 0.0267 0.0148 -0.0122 0.0657
## 2007 2005 -0.0046 0.0161 -0.0470 0.0378
## 2007 2006 -0.0284 0.0191 -0.0785 0.0216
## 2007 2007 -0.0288 0.0171 -0.0736 0.0160
## ---
## Signif. codes: `*' confidence band does not cover 0
##
## P-value for pre-test of parallel trends assumption: 0.23267
## Control Group: Never Treated, Anticipation Periods: 0
## Estimation Method: Doubly Robust# 2
library(fixest)
mpdta$treat_year <- ifelse(mpdta$first.treat == 0, NA, mpdta$first.treat)
att5 <- feols(lemp ~ i(year, treat_year, ref = 2003) + lpop | fips + year,
data = mpdta)
summary(att5)## OLS estimation, Dep. Var.: lemp
## Observations: 955
## Fixed-effects: fips: 191, year: 5
## Standard-errors: Clustered (fips)
## Estimate Std. Error t value Pr(>|t|)
## year::2004:treat_year 0.014843 0.008215 1.80685 0.0723672 .
## year::2005:treat_year 0.031743 0.011551 2.74813 0.0065711 **
## year::2006:treat_year 0.039293 0.013366 2.93982 0.0036915 **
## year::2007:treat_year 0.022030 0.013474 1.63494 0.1037170
## ... 1 variable was removed because of collinearity (lpop)
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## RMSE: 0.111216 Adj. R2: 0.993005
## Within R2: 0.013004# 3
library(did2s)
att6 <- did2s(
data = mpdta,
yname = "lemp",
first_stage = ~ lpop + fips + year,
second_stage = "treat",
treatment = "treat",
cluster_var = "fips"
)
summary(att6)## OLS estimation, Dep. Var.: lemp
## Observations: 2,500
## Standard-errors: Custom
## Estimate Std. Error t value Pr(>|t|)
## treat -0.005309 0.045442 -0.116838 0.907
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## RMSE: 0.545008 Adj. R2: -3.778e-4