分享一篇使用合成控制法SCM进行实证研究的文章的代码

凡是搞计量经济的，都关注这个号了

邮箱：[e-mail]

所有计量经济圈方法论丛的code程序, 宏微观数据库和各种软件都放在社群里.欢迎到计量经济圈社群交流访问.

下面是一篇使用合成控制法SCM进行实证研究的文章的代码。对SCM方法感兴趣的学者，可以将其复制到Stata中对照着文章进行理解，以便将来开展自己的相关研究。针对各种方法的代码，公众号里已经有很多推荐了，有需要的学者可以自行检索【计量经济圈公众号搜索功能及操作流程演示】。

合成控制法SCM1.合成控制法创始人如何用SCM做实证呢？这些规定动作一个都不能少！2.关于合成控制法SCM的33篇精选Articles专辑！小组惊动了阿里巴巴！3.再谈合成控制法SCM, 帮你寻找因果推断控制组，4.最新: 运用机器学习和合成控制法研究武汉封城对空气污染和健康的影响!5.合成控制法与HCW方法, 谁能够走得更远？6.广义合成控制法gsynth, Stata运行程序release，7.广义合成控制法gsynth, 基于交互固定效应的因果推断，8.DID, 合成控制, 匹配, RDD四种方法比较, 适用范围和特征，9.合成控制法什么鬼? 因果推断的前沿方法指南，10.用回归控制法HCW(附程序和数据)评估各国经济刺激计划的影响, 真是一个好主意！11.计量院士为你讲解回归控制法HCW的操作, 并配上自己写的一篇范文！12.HCW面板数据政策评估方法, panel数据构造对照组，13.断点回归RD和合成控制法SCM免费课程, 文章, 数据和代码全在这里, 有必要认真研究学习！14.中文刊上用断点回归RDD和合成控制法SCM的实证文章有哪些？不看至少需要收藏一下！15.用合成控制法, 机器学习和面板数据模型开展政策评估的论文！

*社群也上传的有，群友可以自行下载。

**下面的代码复制一篇文章“Carl Bonander，2017，Compared ith what? Estimating the effects

of injury prevention policies using the synthetic control method，Injury Prevention”

____________________________________________________________

Open the dataset

______________________________________________________________________

*/

*Change ... to the directory where you unzipped your data file.

cd "..."

use rep_opioids.dta, clear

*Define the incidence rate variable (per 100.000 population)

gen rate = deaths/population*100000

**Identify and drop states with missing outcome values (Synth doesn't like

**missing values)

gen missing = missing(rate)

tab state if missing==1

by statecode (missing), sort: drop if missing[_N]==1

*Tell Stata that the data is a panel dataset

xtset statecode year

/*

_________________________________________________________________________

Notes on the Synth code

_________________________________________________________________________

The code works as follows:

synth outcomevar covariate(s), options

where outcomevar is the outcome variable (rate in this case) and covariate(s) 

is a list of variables to match on. We can specify options for each variable

using () to modify the period matched on (see help file for further details).

The options trunit(), trperiod() specifiy the treated state (id for Florida=12)

and the start of the intervention period (2010).

keep() tells synth to store the results in a separate file, which is useful

for e.g. plotting in external programs, since it stores the outcomes in the

synthetic control unit. We will also exploit this feature later on to produce

inferential statistics using permutation tests.

unitnames() makes the output easier to interpet by displaying state names,

as stored in the variable state, instead of id numbers.

_________________________________________________________________________

Section 1: Model selection.

Run a set of different matching specifications to select the best model

_________________________________________________________________________

We begin by selecting the best fitting matching specifiation,

using the models suggested by Ferman et al. (2016, referenced in the paper).

The results are stored in separate files (see the keep() option), which

can be used for plotting them together.

We can use this test for model selection by choosing the model that minimizes

the pre-intervention RMSPE (which is what Synth reports in the output), and present

this model as the main result (as is done in the paper). We can also plot the

stored results to identify if the results holds (approximately) for all or any

alternative specification (which lends more confidence in the results).

As discussed in the paper, we choose the best fitting model of the options

that do not include the full vector of pre-intervention outcome years, since

other covariates are included in the model (see notes below).

*/

*All years (note that covariate weights are set to zero in models with all outcome year)[1]

synth rate rate(1999) rate(2000) rate(2001) rate(2002) rate(2003) rate(2004) rate(2005) rate(2006)

rate(2007) rate(2008) rate(2009) gdp inc unemp_rate, ///

trunit(12) trperiod(2010) fig keep(spec_1.dta, replace) unitnames(state)

*Odd years

synth rate rate(1999) rate(2001) rate(2003) rate(2005) rate(2007) rate(2009) gdp inc unemp_rate, ///

trunit(12) trperiod(2010) fig keep(spec_2.dta, replace) unitnames(state)

*Even years (best fit)

synth rate rate(2000) rate(2002) rate(2004) rate(2006) rate(2008) gdp inc unemp_rate, ///

trunit(12) trperiod(2010) fig keep(spec_3.dta, replace) unitnames(state)

*Average

synth rate rate gdp inc unemp_rate, ///

trunit(12) trperiod(2010) fig keep(spec_4.dta, replace) unitnames(state)

*First half

synth rate rate(1999) rate(2001) rate(2002) rate(2003) rate(2004) rate(2005) gdp inc unemp_rate, ///

trunit(12) trperiod(2010) fig keep(spec_5.dta, replace) unitnames(state)

*First three fourths

synth rate rate(1999) rate(2001) rate(2002) rate(2003) gdp inc unemp_rate, ///

trunit(12) trperiod(2010) fig keep(spec_6.dta, replace) unitnames(state)

/*NOTE [1]

We can extract the information regarding variable weights to probe how much

importance is given to each predictor in the model (this is based on their

predictive power). In the model using the entire range of outcome years,

the covariate weights for gdp inc and unemp_rate will be set to 0 (and are

hence redundant). Choosing one of the other model options

is likely preferable in cases when other covariates are included.

The main output tells us the unit weights and predictor balance,

but omits the variable weights. However, the variable weight matrix

can still be extracted after estimation using:

*/

matlist e(V_matrix)

/*

Which gives the results for the lastest model estimated. Now check

to compare against the model that matches on all years:

*/

qui synth rate rate(1999) rate(2000) rate(2001) rate(2002) rate(2003) rate(2004) rate(2005) rate(2006)

rate(2007) rate(2008) rate(2009) gdp inc unemp_rate, ///

trunit(12) trperiod(2010)

matlist e(V_matrix) //the variable weights for gdp, inc and unemp_rate are now = 0

/*

_________________________________________________________________________

Section 2: Main analysis.

Run synth to perform the main analysis as presented in the paper.

__________________________________________________________________________

Here we match on the outcome itself (every even year in the pre-intervention

period) and the pre-intervention averages of gdp, inc and unemp_rate. We explore

other options below, but for now: this is the model with the best pre-intervention

fit, as measured by the RMSPE (lower=better).

*/

synth rate rate(2000) rate(2002) rate(2004) rate(2006) rate(2008) gdp inc unemp_rate, ///

trunit(12) trperiod(2010) fig keep(main_res.dta, replace) unitnames(state)

/*

________________________________________________________________________________

Section 2: Inferential statistics as suggested by Abadie, Diamond and Heinmuller

We will now perform two placebo permutation tests (one graphical and one

"formal" test).

The graphical test gives us a hint regarding the distribution of placebo results from untreated states, but is sometimes hard to read.

The other produces a pseudo p-value by ranking the states (placebo and Florida) according to their post/pre-intervention RMSPE ratio (larger ratio=bigger effect estimate irrespective of direction).

________________________________________________________________________________

*/

******************

**SPAGHETTI PLOT**

******************

/*

This code runs through all units in the dataset and estimate a synthetic control unit for each one of them. It then stores and merges the results from all of these iterations (46 total), producing a new variable containing the difference between the observed values of outcome Y to the synthetic values

of Y for every year in the time series. We define the variable as the estimated effect for unit i at time t.

In the end, it plots the placebo results in gray against the main results for the treated unit (Florida) in black.

NOTE: Make sure to run the entire loop at once.

*/

set more off

egen id = group(statecode)

xtset id year

forvalues i = 1/46 { //change last number to number of units in the data

tempfile synth_`i' //tempfile to store the results from model `i'

qui synth rate rate(2000) rate(2002) rate(2004) rate(2006) rate(2008) gdp inc unemp_rate, ///

trunit(`i') trperiod(2010) keep(`synth_`i'', replace)

}

forvalues i = 1/46 { //change last number to number of units in the data

use `synth_`i'', clear

rename _time years

gen tr_effect_`i' = _Y_treated - _Y_synthetic

keep years tr_effect_`i'

drop if missing(years)

save `synth_`i'', replace

}

use `synth_1', clear

forvalues i = 2/46 { //change last number to number of units in the data

qui merge 1:1 years using `synth_`i'', nogenerate

}

local lp

forvalues i = 1/46{ //change last number to number of units in the data

 local lp `lp' line tr_effect_`i' years, lcolor(gs12) lpattern(solid) ||

}

twoway `lp' || line tr_effect_9 years, ///

lcolor(black) lpattern(solid) legend(off) xline(2010, lpattern(dash) lcolor(black)) ylab(,nogrid)

/* NOTES:

(1) Change "line tr_effect_9" to tr_effect_[treatunit] to highlight other treatment

units in the plot. Also, change xline([year],...) to the intervention year.

In the paper, I use the resulting dataset (now in use) to generate Fig. 4, which

also highlights the placebos with the same or greater post/pre-RMSPE ratio. I use an

external plotting program for this, although code can easily be added to

the twoway plot above to highlight specific placebos. 

(2) The forvalues loop is only valid if id codes are in consecutive order.

If not, generate a new id var using "egen newid = group(oldid)" and identify

value of the treatment unit on the new id variable and the change input values

indicated in the code above. (This is not a problem in this dataset)

*/

***********************************************

**"PSEUDO P-VALUES" AND POST/PRE RMSPE RATIOS**

***********************************************

/*

Next, let's calculate a pseudo p-value based on the rank of the treatment unit's

post/pre-RMSPE ratio compared to the untreated placebo units post/pre-RMSPE

ratios (as in Abadie, Diamond and Heinmuller 2010, referenced in the paper).

We use the placebo effect dataset genereted by the code above for this,

so the code for the spaghetti plot above must be run first.

*/

preserve

gen post = 0

replace post = 1 if years>=2010 //change intervention year if needed

forvalues i = 1/46 { //change last number to to number of units/panels in the data

gen error_squared_`i' = (tr_effect_`i')^2

}

collapse (mean) error_squared_*, by(post)

reshape long error_squared_, i(post) j(id)

reshape wide error_squared_, i(id) j(post)

gen rootpre = sqrt(error_squared_0)

gen rootpost = sqrt(error_squared_1)

gen Ratio = rootpost/rootpre

summarize Ratio if id==9, meanonly //change id to treatunit

scalar treatRMSPE = r(mean)

gen prob = 0

replace prob = 1 if Ratio>=treatRMSPE

summarize prob, meanonly

scalar pval = r(mean)

tempname resmat //list post/pre ratios for all units

 forvalues i = 1/46 { //change last number to to number of units/panels in the data

summarize Ratio if id==`i', meanonly

matrix `resmat' = nullmat(`resmat') \ r(mean)

 local names `"`names' `"`i'"'"'

 }

 mat colnames `resmat' = "Post/Pre Ratio"

 mat rownames `resmat' = `names'

 matlist `resmat' , row("Temp ID")

matrix pvalue = pval

mat colnames pvalue = "p-value"

mat rownames pvalue = "Ratio trunit"

matlist pvalue //probability of finding post/pre ratio => ratio in treatment unit

restore

/*

_______________________________________________________________________________

Section 3: Sensitivity analysis (leave-one-out)

_______________________________________________________________________________

In this section, we run the leave-one-out sensitivity analysis as presented

in the paper. We begin by opening the analysis dataset again.

*/

use opioids.dta, clear

*Define the incidence rate variable (per 100.000 population)

gen rate = deaths/population*100000

*Add covariates

merge 1:1 state year using "Covariates\unemp_panel.dta", nogenerate

merge 1:1 state year using "Covariates\capita_gdp_panel.dta", nogenerate

merge 1:1 state year using "Covariates\personal_capita_income_panel.dta", nogenerate

**Identify and drop states with missing outcome values (Synth doesn't like

**missing values)

gen missing = missing(rate)

tab state if missing==1

by statecode (missing), sort: drop if missing[_N]==1

*Tell Stata that the data is a panel dataset

xtset statecode year

******************************

*********LEAVE ONE OUT********

******************************

/*

This procedure can get a little tedious as we will do it by hand,

checking the pre-intervention RMSPE for each model, and iteratively

removing the state that recieved the largest weight in the previous

iteration. As we saw in the main analysis, Maine was the largest

contributor to the synthetic control unit. As such, we begin by

dropping Maine from the donor pool. In iteration 2, we will notice

that Hawaii is the largest contributor, so iteration 3 will drop

both Maine and Hawaii, etc, until we obtain a model with twice

the prediction errors as in the main analysis (RMSPE for the main analysis

is 0.55, so we will keep going until it is >1.10. Turns out

this results in 35 iterations before we stop. (In an ideal case,

we would probably be better off by writing a algorithm that

automates this process.)

*/

**Drop Maine (Iter 1)

preserve

drop if state=="Maine"

synth rate rate(2000) rate(2002) rate(2004) rate(2006) rate(2008) gdp inc unemp_rate, trunit(12)

trperiod(2010) unitnames(state) fig keep(loo_1.dta, replace)

restore

**Drop Hawaii & Maine (Iter 2)

preserve

drop if state=="Hawaii"|state=="Maine"

synth rate rate(2000) rate(2002) rate(2004) rate(2006) rate(2008) gdp inc unemp_rate, trunit(12)

trperiod(2010) unitnames(state) fig keep(loo_2.dta, replace)

restore

**Drop Pennsylvania, Hawaii & Maine (Iter 3)

preserve

drop if state=="Pennsylvania"|state=="Hawaii"|state=="Maine"

synth rate rate(2000) rate(2002) rate(2004) rate(2006) rate(2008) gdp inc unemp_rate, trunit(12)

trperiod(2010) unitnames(state) fig keep(loo_3.dta, replace)

restore

**Drop Arkansas,Pennsylvania, Hawaii & Maine (Iter 4)

preserve

drop if state=="Arkansas"|state=="Pennsylvania"|state=="Hawaii"|state=="Maine"

synth rate rate(2000) rate(2002) rate(2004) rate(2006) rate(2008) gdp inc unemp_rate, trunit(12)

trperiod(2010) unitnames(state) fig keep(loo_4.dta, replace)

restore

**Drop Cali, Arkansas,Pennsylvania, Hawaii & Maine (Iter 5)

preserve

drop if state=="California"|state=="Arkansas"|state=="Pennsylvania"|state=="Hawaii"|state=="Maine"

synth rate rate(2000) rate(2002) rate(2004) rate(2006) rate(2008) gdp inc unemp_rate, trunit(12)

trperiod(2010) unitnames(state) fig keep(loo_5.dta, replace)

restore

**Drop Kansas, Cali, Arkansas,Pennsylvania, Hawaii & Maine (Iter 6)

preserve

drop if

state=="Kansas"|state=="California"|state=="Arkansas"|state=="Pennsylvania"|state=="Hawaii"|state=

="Maine"

synth rate rate(2000) rate(2002) rate(2004) rate(2006) rate(2008) gdp inc unemp_rate, trunit(12)

trperiod(2010) unitnames(state) fig keep(loo_6.dta, replace)

restore

**Drop NJ, Kansas, Cali, Arkansas,Pennsylvania, Hawaii & Maine (Iter 7)

preserve

drop if state=="New

Jersey"|state=="Kansas"|state=="California"|state=="Arkansas"|state=="Pennsylvania"|state=="Hawaii

"|state=="Maine"

synth rate rate(2000) rate(2002) rate(2004) rate(2006) rate(2008) gdp inc unemp_rate, trunit(12)

trperiod(2010) unitnames(state) fig keep(loo_7.dta, replace)

restore

**Drop Virginia, NJ, Kansas, Cali, Arkansas,Pennsylvania, Hawaii & Maine (Iter 8)

preserve

drop if state=="Virginia"|state=="New

Jersey"|state=="Kansas"|state=="California"|state=="Arkansas"|state=="Pennsylvania"|state=="Hawaii

"|state=="Maine"

synth rate rate(2000) rate(2002) rate(2004) rate(2006) rate(2008) gdp inc unemp_rate, trunit(12)

trperiod(2010) unitnames(state) fig keep(loo_8.dta, replace)

restore

**Drop Maryland, Virginia, NJ, Kansas, Cali, Arkansas,Pennsylvania, Hawaii & Maine (Iter 9)

preserve

drop if state=="Maryland"|state=="Virginia"|state=="New

Jersey"|state=="Kansas"|state=="California"|state=="Arkansas"|state=="Pennsylvania"|state=="Hawaii

"|state=="Maine"

synth rate rate(2000) rate(2002) rate(2004) rate(2006) rate(2008) gdp inc unemp_rate, trunit(12)

trperiod(2010) unitnames(state) fig keep(loo_9.dta, replace)

restore

**Drop New Hamp., Maryland, Virginia, NJ, Kansas, Cali, Arkansas,Pennsylvania, Hawaii & Maine (Iter 10)

preserve

drop if state=="New Hampshire"|state=="Maryland"|state=="Virginia"|state=="New

Jersey"|state=="Kansas"|state=="California"|state=="Arkansas"|state=="Pennsylvania"|state=="Hawaii

"|state=="Maine"

synth rate rate(2000) rate(2002) rate(2004) rate(2006) rate(2008) gdp inc unemp_rate, trunit(12)

trperiod(2010) unitnames(state) fig keep(loo_10.dta, replace)

restore

**Drop Washing, New Hamp., Maryland, Virginia, NJ, Kansas, Cali, Arkansas,Pennsylvania, Hawaii &

Maine (Iter 11)

preserve

drop if state=="Washington"|state=="New

Hampshire"|state=="Maryland"|state=="Virginia"|state=="New 

Jersey"|state=="Kansas"|state=="California"|state=="Arkansas"|state=="Pennsylvania"|state=="Hawaii

"|state=="Maine"

synth rate rate(2000) rate(2002) rate(2004) rate(2006) rate(2008) gdp inc unemp_rate, trunit(12)

trperiod(2010) unitnames(state) fig keep(loo_11.dta, replace)

restore

**Drop Ariz., Washing, New Hamp., Maryland, Virginia, NJ, Kansas, Cali, Arkansas,Pennsylvania, Hawaii &

Maine (Iter 12)

preserve

drop if state=="Arizona"|state=="Washington"|state=="New

Hampshire"|state=="Maryland"|state=="Virginia"|state=="New

Jersey"|state=="Kansas"|state=="California"|state=="Arkansas"|state=="Pennsylvania"|state=="Hawaii

"|state=="Maine"

synth rate rate(2000) rate(2002) rate(2004) rate(2006) rate(2008) gdp inc unemp_rate, trunit(12)

trperiod(2010) unitnames(state) fig keep(loo_12.dta, replace)

restore

**Drop Texas, Ariz., Washing, New Hamp., Maryland, Virginia, NJ, Kansas, Cali, Arkansas,Pennsylvania,

Hawaii & Maine (Iter 13)

preserve

drop if state=="Texas"|state=="Arizona"|state=="Washington"|state=="New

Hampshire"|state=="Maryland"|state=="Virginia"|state=="New

Jersey"|state=="Kansas"|state=="California"|state=="Arkansas"|state=="Pennsylvania"|state=="Hawaii

"|state=="Maine"

synth rate rate(2000) rate(2002) rate(2004) rate(2006) rate(2008) gdp inc unemp_rate, trunit(12)

trperiod(2010) unitnames(state) fig keep(loo_13.dta, replace)

restore

**Drop Idaho, Texas, Ariz., Washing, New Hamp., Maryland, Virginia, NJ, Kansas, Cali,

Arkansas,Pennsylvania, Hawaii & Maine (Iter 14)

preserve

drop if state=="Idaho"|state=="Texas"|state=="Arizona"|state=="Washington"|state=="New

Hampshire"|state=="Maryland"|state=="Virginia"|state=="New

Jersey"|state=="Kansas"|state=="California"|state=="Arkansas"|state=="Pennsylvania"|state=="Hawaii

"|state=="Maine"

synth rate rate(2000) rate(2002) rate(2004) rate(2006) rate(2008) gdp inc unemp_rate, trunit(12)

trperiod(2010) unitnames(state) fig keep(loo_14.dta, replace)

restore

**Drop Georgia, Idaho, Texas, Ariz., Washing, New Hamp., Maryland, Virginia, NJ, Kansas, Cali,

Arkansas,Pennsylvania, Hawaii & Maine (Iter 15)

preserve

drop if

state=="Georgia"|state=="Idaho"|state=="Texas"|state=="Arizona"|state=="Washington"|state=="New

Hampshire"|state=="Maryland"|state=="Virginia"|state=="New

Jersey"|state=="Kansas"|state=="California"|state=="Arkansas"|state=="Pennsylvania"|state=="Hawaii

"|state=="Maine"

synth rate rate(2000) rate(2002) rate(2004) rate(2006) rate(2008) gdp inc unemp_rate, trunit(12)

trperiod(2010) unitnames(state) fig keep(loo_15.dta, replace)

restore

**Drop Utah, Georgia, Idaho, Texas, Ariz., Washing, New Hamp., Maryland, Virginia, NJ, Kansas, Cali,

Arkansas,Pennsylvania, Hawaii & Maine (Iter 16)

preserve

drop if

state=="Utah"|state=="Georgia"|state=="Idaho"|state=="Texas"|state=="Arizona"|state=="Washington

"|state=="New Hampshire"|state=="Maryland"|state=="Virginia"|state=="New

Jersey"|state=="Kansas"|state=="California"|state=="Arkansas"|state=="Pennsylvania"|state=="Hawaii

"|state=="Maine"

synth rate rate(2000) rate(2002) rate(2004) rate(2006) rate(2008) gdp inc unemp_rate, trunit(12)

trperiod(2010) unitnames(state) fig keep(loo_16.dta, replace)

restore

**Drop Louisiana, Utah, Georgia, Idaho, Texas, Ariz., Washing, New Hamp., Maryland, Virginia, NJ,

Kansas, Cali, Arkansas,Pennsylvania, Hawaii & Maine (Iter 17)

preserve

drop if

state=="Louisiana"|state=="Utah"|state=="Georgia"|state=="Idaho"|state=="Texas"|state=="Arizona"|

state=="Washington"|state=="New Hampshire"|state=="Maryland"|state=="Virginia"|state=="New

Jersey"|state=="Kansas"|state=="California"|state=="Arkansas"|state=="Pennsylvania"|state=="Hawaii

"|state=="Maine"

synth rate rate(2000) rate(2002) rate(2004) rate(2006) rate(2008) gdp inc unemp_rate, trunit(12)

trperiod(2010) unitnames(state) fig keep(loo_17.dta, replace)

restore

**Drop Alabama, Utah, Georgia, Idaho, Texas, Ariz., Washing, New Hamp., Maryland, Virginia, NJ,

Kansas, Cali, Arkansas,Pennsylvania, Hawaii & Maine (Iter 18)

preserve

drop if

state=="Alabama"|state=="Louisiana"|state=="Utah"|state=="Georgia"|state=="Idaho"|state=="Texas"

|state=="Arizona"|state=="Washington"|state=="New

Hampshire"|state=="Maryland"|state=="Virginia"|state=="New

Jersey"|state=="Kansas"|state=="California"|state=="Arkansas"|state=="Pennsylvania"|state=="Hawaii

"|state=="Maine"

synth rate rate(2000) rate(2002) rate(2004) rate(2006) rate(2008) gdp inc unemp_rate, trunit(12)

trperiod(2010) unitnames(state) fig keep(loo_18.dta, replace)

restore

**Drop Montana, Alabama, Utah, Georgia, Idaho, Texas, Ariz., Washing, New Hamp., Maryland, Virginia,

NJ, Kansas, Cali, Arkansas,Pennsylvania, Hawaii & Maine (Iter 19)

preserve

drop if

state=="Montana"|state=="Alabama"|state=="Louisiana"|state=="Utah"|state=="Georgia"|state=="Ida

ho"|state=="Texas"|state=="Arizona"|state=="Washington"|state=="New

Hampshire"|state=="Maryland"|state=="Virginia"|state=="New 

Jersey"|state=="Kansas"|state=="California"|state=="Arkansas"|state=="Pennsylvania"|state=="Hawaii

"|state=="Maine"

synth rate rate(2000) rate(2002) rate(2004) rate(2006) rate(2008) gdp inc unemp_rate, trunit(12)

trperiod(2010) unitnames(state) fig keep(loo_19.dta, replace)

restore

**Drop Ohio, Montana, Alabama, Utah, Georgia, Idaho, Texas, Ariz., Washing, New Hamp., Maryland,

Virginia, NJ, Kansas, Cali, Arkansas,Pennsylvania, Hawaii & Maine (Iter 20)

preserve

drop if

state=="Ohio"|state=="Montana"|state=="Alabama"|state=="Louisiana"|state=="Utah"|state=="Georgi

a"|state=="Idaho"|state=="Texas"|state=="Arizona"|state=="Washington"|state=="New

Hampshire"|state=="Maryland"|state=="Virginia"|state=="New

Jersey"|state=="Kansas"|state=="California"|state=="Arkansas"|state=="Pennsylvania"|state=="Hawaii

"|state=="Maine"

synth rate rate(2000) rate(2002) rate(2004) rate(2006) rate(2008) gdp inc unemp_rate, trunit(12)

trperiod(2010) unitnames(state) fig keep(loo_20.dta, replace)

restore

**Drop NC, Ohio, Montana, Alabama, Utah, Georgia, Idaho, Texas, Ariz., Washing, New Hamp.,

Maryland, Virginia, NJ, Kansas, Cali, Arkansas,Pennsylvania, Hawaii & Maine (Iter 21)

preserve

drop if state=="North

Carolina"|state=="Ohio"|state=="Montana"|state=="Alabama"|state=="Louisiana"|state=="Utah"|stat

e=="Georgia"|state=="Idaho"|state=="Texas"|state=="Arizona"|state=="Washington"|state=="New

Hampshire"|state=="Maryland"|state=="Virginia"|state=="New

Jersey"|state=="Kansas"|state=="California"|state=="Arkansas"|state=="Pennsylvania"|state=="Hawaii

"|state=="Maine"

synth rate rate(2000) rate(2002) rate(2004) rate(2006) rate(2008) gdp inc unemp_rate, trunit(12)

trperiod(2010) unitnames(state) fig keep(loo_21.dta, replace)

restore

**Drop Nevada, NC, Ohio, Montana, Alabama, Utah, Georgia, Idaho, Texas, Ariz., Washing, New Hamp.,

Maryland, Virginia, NJ, Kansas, Cali, Arkansas,Pennsylvania, Hawaii & Maine (Iter 22)

preserve

drop if state=="Nevada"|state=="North

Carolina"|state=="Ohio"|state=="Montana"|state=="Alabama"|state=="Louisiana"|state=="Utah"|stat

e=="Georgia"|state=="Idaho"|state=="Texas"|state=="Arizona"|state=="Washington"|state=="New

Hampshire"|state=="Maryland"|state=="Virginia"|state=="New

Jersey"|state=="Kansas"|state=="California"|state=="Arkansas"|state=="Pennsylvania"|state=="Hawaii

"|state=="Maine"

synth rate rate(2000) rate(2002) rate(2004) rate(2006) rate(2008) gdp inc unemp_rate, trunit(12)

trperiod(2010) unitnames(state) fig keep(loo_22.dta, replace)

restore

**Drop Tennessee, Nevada, NC, Ohio, Montana, Alabama, Utah, Georgia, Idaho, Texas, Ariz., Washing,

New Hamp., Maryland, Virginia, NJ, Kansas, Cali, Arkansas,Pennsylvania, Hawaii & Maine (Iter 23)

preserve

drop if state=="Tennessee"|state=="Nevada"|state=="North

Carolina"|state=="Ohio"|state=="Montana"|state=="Alabama"|state=="Louisiana"|state=="Utah"|stat

e=="Georgia"|state=="Idaho"|state=="Texas"|state=="Arizona"|state=="Washington"|state=="New

Hampshire"|state=="Maryland"|state=="Virginia"|state=="New

Jersey"|state=="Kansas"|state=="California"|state=="Arkansas"|state=="Pennsylvania"|state=="Hawaii

"|state=="Maine"

synth rate rate(2000) rate(2002) rate(2004) rate(2006) rate(2008) gdp inc unemp_rate, trunit(12)

trperiod(2010) unitnames(state) fig keep(loo_23.dta, replace)

restore

**Drop Connecticut, Tennessee, Nevada, NC, Ohio, Montana, Alabama, Utah, Georgia, Idaho, Texas,

Ariz., Washing, New Hamp., Maryland, Virginia, NJ, Kansas, Cali, Arkansas,Pennsylvania, Hawaii & Maine

(Iter 24)

preserve

drop if state=="Connecticut"|state=="Tennessee"|state=="Nevada"|state=="North

Carolina"|state=="Ohio"|state=="Montana"|state=="Alabama"|state=="Louisiana"|state=="Utah"|stat

e=="Georgia"|state=="Idaho"|state=="Texas"|state=="Arizona"|state=="Washington"|state=="New

Hampshire"|state=="Maryland"|state=="Virginia"|state=="New

Jersey"|state=="Kansas"|state=="California"|state=="Arkansas"|state=="Pennsylvania"|state=="Hawaii

"|state=="Maine"

synth rate rate(2000) rate(2002) rate(2004) rate(2006) rate(2008) gdp inc unemp_rate, trunit(12)

trperiod(2010) unitnames(state) fig keep(loo_24.dta, replace)

restore

**Drop Minnesota, Connecticut, Tennessee, Nevada, NC, Ohio, Montana, Alabama, Utah, Georgia,

Idaho, Texas, Ariz., Washing, New Hamp., Maryland, Virginia, NJ, Kansas, Cali, Arkansas,Pennsylvania,

Hawaii & Maine (Iter 25)

preserve

drop if state=="Minnesota"|state=="Connecticut"|state=="Tennessee"|state=="Nevada"|state=="North

Carolina"|state=="Ohio"|state=="Montana"|state=="Alabama"|state=="Louisiana"|state=="Utah"|stat

e=="Georgia"|state=="Idaho"|state=="Texas"|state=="Arizona"|state=="Washington"|state=="New

Hampshire"|state=="Maryland"|state=="Virginia"|state=="New

Jersey"|state=="Kansas"|state=="California"|state=="Arkansas"|state=="Pennsylvania"|state=="Hawaii

"|state=="Maine"

synth rate rate(2000) rate(2002) rate(2004) rate(2006) rate(2008) gdp inc unemp_rate, trunit(12)

trperiod(2010) unitnames(state) fig keep(loo_25.dta, replace)

restore

**Drop Delaware, Minnesota, Connecticut, Tennessee, Nevada, NC, Ohio, Montana, Alabama, Utah,

Georgia, Idaho, Texas, Ariz., Washing, New Hamp., Maryland, Virginia, NJ, Kansas, Cali,

Arkansas,Pennsylvania, Hawaii & Maine (Iter 26)

preserve

drop if

state=="Delaware"|state=="Minnesota"|state=="Connecticut"|state=="Tennessee"|state=="Nevada"|st

ate=="North

Carolina"|state=="Ohio"|state=="Montana"|state=="Alabama"|state=="Louisiana"|state=="Utah"|stat

e=="Georgia"|state=="Idaho"|state=="Texas"|state=="Arizona"|state=="Washington"|state=="New

Hampshire"|state=="Maryland"|state=="Virginia"|state=="New 

Jersey"|state=="Kansas"|state=="California"|state=="Arkansas"|state=="Pennsylvania"|state=="Hawaii

"|state=="Maine"

synth rate rate(2000) rate(2002) rate(2004) rate(2006) rate(2008) gdp inc unemp_rate, trunit(12)

trperiod(2010) unitnames(state) fig keep(loo_26.dta, replace)

restore

**Drop Massachusetts, Delaware, Minnesota, Connecticut, Tennessee, Nevada, NC, Ohio, Montana,

Alabama, Utah, Georgia, Idaho, Texas, Ariz., Washing, New Hamp., Maryland, Virginia, NJ, Kansas, Cali,

Arkansas,Pennsylvania, Hawaii & Maine (Iter 27)

preserve

drop if

state=="Massachusetts"|state=="Delaware"|state=="Minnesota"|state=="Connecticut"|state=="Tennes

see"|state=="Nevada"|state=="North

Carolina"|state=="Ohio"|state=="Montana"|state=="Alabama"|state=="Louisiana"|state=="Utah"|stat

e=="Georgia"|state=="Idaho"|state=="Texas"|state=="Arizona"|state=="Washington"|state=="New

Hampshire"|state=="Maryland"|state=="Virginia"|state=="New

Jersey"|state=="Kansas"|state=="California"|state=="Arkansas"|state=="Pennsylvania"|state=="Hawaii

"|state=="Maine"

synth rate rate(2000) rate(2002) rate(2004) rate(2006) rate(2008) gdp inc unemp_rate, trunit(12)

trperiod(2010) unitnames(state) fig keep(loo_27.dta, replace)

restore

**Drop Colorado, Massachusetts, Delaware, Minnesota, Connecticut, Tennessee, Nevada, NC, Ohio,

Montana, Alabama, Utah, Georgia, Idaho, Texas, Ariz., Washing, New Hamp., Maryland, Virginia, NJ,

Kansas, Cali, Arkansas,Pennsylvania, Hawaii & Maine (Iter 28)

preserve

drop if

state=="Colorado"|state=="Massachusetts"|state=="Delaware"|state=="Minnesota"|state=="Connectic

ut"|state=="Tennessee"|state=="Nevada"|state=="North

Carolina"|state=="Ohio"|state=="Montana"|state=="Alabama"|state=="Louisiana"|state=="Utah"|stat

e=="Georgia"|state=="Idaho"|state=="Texas"|state=="Arizona"|state=="Washington"|state=="New

Hampshire"|state=="Maryland"|state=="Virginia"|state=="New

Jersey"|state=="Kansas"|state=="California"|state=="Arkansas"|state=="Pennsylvania"|state=="Hawaii

"|state=="Maine"

synth rate rate(2000) rate(2002) rate(2004) rate(2006) rate(2008) gdp inc unemp_rate, trunit(12)

trperiod(2010) unitnames(state) fig keep(loo_28.dta, replace)

restore

**Drop Iowa, Colorado, Massachusetts, Delaware, Minnesota, Connecticut, Tennessee, Nevada, NC,

Ohio, Montana, Alabama, Utah, Georgia, Idaho, Texas, Ariz., Washing, New Hamp., Maryland, Virginia,

NJ, Kansas, Cali, Arkansas,Pennsylvania, Hawaii & Maine (Iter 29)

preserve

drop if

state=="Iowa"|state=="Colorado"|state=="Massachusetts"|state=="Delaware"|state=="Minnesota"|sta

te=="Connecticut"|state=="Tennessee"|state=="Nevada"|state=="North

Carolina"|state=="Ohio"|state=="Montana"|state=="Alabama"|state=="Louisiana"|state=="Utah"|stat

e=="Georgia"|state=="Idaho"|state=="Texas"|state=="Arizona"|state=="Washington"|state=="New

Hampshire"|state=="Maryland"|state=="Virginia"|state=="New

Jersey"|state=="Kansas"|state=="California"|state=="Arkansas"|state=="Pennsylvania"|state=="Hawaii

"|state=="Maine"

synth rate rate(2000) rate(2002) rate(2004) rate(2006) rate(2008) gdp inc unemp_rate, trunit(12)

trperiod(2010) unitnames(state) fig keep(loo_29.dta, replace)

restore

**Drop Indiana, Iowa, Colorado, Massachusetts, Delaware, Minnesota, Connecticut, Tennessee, Nevada,

NC, Ohio, Montana, Alabama, Utah, Georgia, Idaho, Texas, Ariz., Washing, New Hamp., Maryland,

Virginia, NJ, Kansas, Cali, Arkansas,Pennsylvania, Hawaii & Maine (Iter 30)

preserve

drop if

state=="Indiana"|state=="Iowa"|state=="Colorado"|state=="Massachusetts"|state=="Delaware"|state=

="Minnesota"|state=="Connecticut"|state=="Tennessee"|state=="Nevada"|state=="North

Carolina"|state=="Ohio"|state=="Montana"|state=="Alabama"|state=="Louisiana"|state=="Utah"|stat

e=="Georgia"|state=="Idaho"|state=="Texas"|state=="Arizona"|state=="Washington"|state=="New

Hampshire"|state=="Maryland"|state=="Virginia"|state=="New

Jersey"|state=="Kansas"|state=="California"|state=="Arkansas"|state=="Pennsylvania"|state=="Hawaii

"|state=="Maine"

synth rate rate(2000) rate(2002) rate(2004) rate(2006) rate(2008) gdp inc unemp_rate, trunit(12)

trperiod(2010) unitnames(state) fig keep(loo_30.dta, replace)

restore

**Drop Wisconsin, Indiana, Iowa, Colorado, Massachusetts, Delaware, Minnesota, Connecticut,

Tennessee, Nevada, NC, Ohio, Montana, Alabama, Utah, Georgia, Idaho, Texas, Ariz., Washing, New

Hamp., Maryland, Virginia, NJ, Kansas, Cali, Arkansas,Pennsylvania, Hawaii & Maine (Iter 31)

preserve

drop if

state=="Wisconsin"|state=="Indiana"|state=="Iowa"|state=="Colorado"|state=="Massachusetts"|state

=="Delaware"|state=="Minnesota"|state=="Connecticut"|state=="Tennessee"|state=="Nevada"|state=

="North

Carolina"|state=="Ohio"|state=="Montana"|state=="Alabama"|state=="Louisiana"|state=="Utah"|stat

e=="Georgia"|state=="Idaho"|state=="Texas"|state=="Arizona"|state=="Washington"|state=="New

Hampshire"|state=="Maryland"|state=="Virginia"|state=="New

Jersey"|state=="Kansas"|state=="California"|state=="Arkansas"|state=="Pennsylvania"|state=="Hawaii

"|state=="Maine"

synth rate rate(2000) rate(2002) rate(2004) rate(2006) rate(2008) gdp inc unemp_rate, trunit(12)

trperiod(2010) unitnames(state) fig keep(loo_31.dta, replace)

restore

**Drop Illinois, Wisconsin, Indiana, Iowa, Colorado, Massachusetts, Delaware, Minnesota, Connecticut,

Tennessee, Nevada, NC, Ohio, Montana, Alabama, Utah, Georgia, Idaho, Texas, Ariz., Washing, New

Hamp., Maryland, Virginia, NJ, Kansas, Cali, Arkansas,Pennsylvania, Hawaii & Maine (Iter 32)

preserve

drop if

state=="Illinois"|state=="Wisconsin"|state=="Indiana"|state=="Iowa"|state=="Colorado"|state=="Mass

achusetts"|state=="Delaware"|state=="Minnesota"|state=="Connecticut"|state=="Tennessee"|state==

"Nevada"|state=="North

Carolina"|state=="Ohio"|state=="Montana"|state=="Alabama"|state=="Louisiana"|state=="Utah"|stat

e=="Georgia"|state=="Idaho"|state=="Texas"|state=="Arizona"|state=="Washington"|state=="New

Hampshire"|state=="Maryland"|state=="Virginia"|state=="New

Jersey"|state=="Kansas"|state=="California"|state=="Arkansas"|state=="Pennsylvania"|state=="Hawaii

"|state=="Maine"

synth rate rate(2000) rate(2002) rate(2004) rate(2006) rate(2008) gdp inc unemp_rate, trunit(12)

trperiod(2010) unitnames(state) fig keep(loo_32.dta, replace)

restore

**Drop Vermont, Illinois, Wisconsin, Indiana, Iowa, Colorado, Massachusetts, Delaware, Minnesota,

Connecticut, Tennessee, Nevada, NC, Ohio, Montana, Alabama, Utah, Georgia, Idaho, Texas, Ariz.,

Washing, New Hamp., Maryland, Virginia, NJ, Kansas, Cali, Arkansas,Pennsylvania, Hawaii & Maine (Iter

33)

preserve

drop if

state=="Vermont"|state=="Illinois"|state=="Wisconsin"|state=="Indiana"|state=="Iowa"|state=="Color

ado"|state=="Massachusetts"|state=="Delaware"|state=="Minnesota"|state=="Connecticut"|state=="T

ennessee"|state=="Nevada"|state=="North

Carolina"|state=="Ohio"|state=="Montana"|state=="Alabama"|state=="Louisiana"|state=="Utah"|stat

e=="Georgia"|state=="Idaho"|state=="Texas"|state=="Arizona"|state=="Washington"|state=="New

Hampshire"|state=="Maryland"|state=="Virginia"|state=="New

Jersey"|state=="Kansas"|state=="California"|state=="Arkansas"|state=="Pennsylvania"|state=="Hawaii

"|state=="Maine"

synth rate rate(2000) rate(2002) rate(2004) rate(2006) rate(2008) gdp inc unemp_rate, trunit(12)

trperiod(2010) unitnames(state) fig keep(loo_33.dta, replace)

restore

**Drop Missi., Vermont, Illinois, Wisconsin, Indiana, Iowa, Colorado, Massachusetts, Delaware,

Minnesota, Connecticut, Tennessee, Nevada, NC, Ohio, Montana, Alabama, Utah, Georgia, Idaho, Texas,

Ariz., Washing, New Hamp., Maryland, Virginia, NJ, Kansas, Cali, Arkansas,Pennsylvania, Hawaii & Maine

(Iter 34)

preserve

drop if

state=="Mississippi"|state=="Vermont"|state=="Illinois"|state=="Wisconsin"|state=="Indiana"|state=="

Iowa"|state=="Colorado"|state=="Massachusetts"|state=="Delaware"|state=="Minnesota"|state=="Co

nnecticut"|state=="Tennessee"|state=="Nevada"|state=="North

Carolina"|state=="Ohio"|state=="Montana"|state=="Alabama"|state=="Louisiana"|state=="Utah"|stat

e=="Georgia"|state=="Idaho"|state=="Texas"|state=="Arizona"|state=="Washington"|state=="New

Hampshire"|state=="Maryland"|state=="Virginia"|state=="New

Jersey"|state=="Kansas"|state=="California"|state=="Arkansas"|state=="Pennsylvania"|state=="Hawaii

"|state=="Maine"

synth rate rate(2000) rate(2002) rate(2004) rate(2006) rate(2008) gdp inc unemp_rate, trunit(12)

trperiod(2010) unitnames(state) fig keep(loo_34.dta, replace)

restore

**Drop Missouri, Missi., Vermont, Illinois, Wisconsin, Indiana, Iowa, Colorado, Massachusetts, Delaware,

Minnesota, Connecticut, Tennessee, Nevada, NC, Ohio, Montana, Alabama, Utah, Georgia, Idaho, Texas,

Ariz., Washing, New Hamp., Maryland, Virginia, NJ, Kansas, Cali, Arkansas,Pennsylvania, Hawaii & Maine

(Iter 35)

preserve

drop if

state=="Missouri"|state=="Mississippi"|state=="Vermont"|state=="Illinois"|state=="Wisconsin"|state==

"Indiana"|state=="Iowa"|state=="Colorado"|state=="Massachusetts"|state=="Delaware"|state=="Minn

esota"|state=="Connecticut"|state=="Tennessee"|state=="Nevada"|state=="North

Carolina"|state=="Ohio"|state=="Montana"|state=="Alabama"|state=="Louisiana"|state=="Utah"|stat

e=="Georgia"|state=="Idaho"|state=="Texas"|state=="Arizona"|state=="Washington"|state=="New

Hampshire"|state=="Maryland"|state=="Virginia"|state=="New

Jersey"|state=="Kansas"|state=="California"|state=="Arkansas"|state=="Pennsylvania"|state=="Hawaii

"|state=="Maine"

synth rate rate(2000) rate(2002) rate(2004) rate(2006) rate(2008) gdp inc unemp_rate, trunit(12)

trperiod(2010) unitnames(state) fig keep(loo_35.dta, replace)

restore

/*

_________________

Plot the results

_________________

*/

use loo_1, clear

rename _time years

keep years _Y_treated _Y_synthetic

rename _Y_treated observed

rename _Y_synthetic loo_1

drop if missing(years)

save loo_1, replace

forvalues i = 2/35 { //change last number to number of leave one out-iterations

use loo_`i', clear

rename _time years

keep years _Y_synthetic

rename _Y_synthetic loo_`i'

drop if missing(years)

save loo_`i', replace

}

use loo_1, clear

forvalues i = 2/35 { //change last number to number of leave one out-iterations

qui merge 1:1 years using loo_`i', nogenerate

}

local lp

forvalues i = 1/35{ //change last number to number of leave one out-iterations

 local lp `lp' line loo_`i' years, lcolor(gs12) lpattern(solid) ||

}

twoway `lp' || line observed years, ///

lcolor(black) lpattern(solid) legend(off) xline(2010, lpattern(dash) lcolor(black)) ylab(,nogrid)

下面这些短链接文章属于合集，可以收藏起来阅读，不然以后都找不到了。

2.5年，计量经济圈近1000篇不重类计量文章，

可直接在公众号菜单栏搜索任何计量相关问题,

Econometrics Circle

数据系列：空间矩阵 | 工企数据 | PM2.5 | 市场化指数 | CO2数据 |  夜间灯光 | 官员方言  | 微观数据 | 内部数据计量系列：匹配方法 | 内生性 | 工具变量 | DID | 面板数据 | 常用TOOL | 中介调节 | 时间序列 | RDD断点 | 合成控制 | 200篇合辑 | 因果识别 | 社会网络 | 空间DID数据处理：Stata | R | Python | 缺失值 | CHIP/ CHNS/CHARLS/CFPS/CGSS等 |干货系列：能源环境 | 效率研究 | 空间计量 | 国际经贸 | 计量软件 | 商科研究 | 机器学习 | SSCI | CSSCI | SSCI查询 | 名家经验计量经济圈组织了一个计量社群，有如下特征：热情互助最多、前沿趋势最多、社科资料最多、社科数据最多、科研牛人最多、海外名校最多。因此，建议积极进取和有强烈研习激情的中青年学者到社群交流探讨，始终坚信优秀是通过感染优秀而互相成就彼此的。