Stata Command: ecic

Stata Command: ecic.ado

The ecic command estimates quantile treatment effects (QTE) at extreme quantiles via changes in changes (CIC) based on Sasaki and Wang (Forthcoming).  The designed setting requires that all the units are untreated in the first period (T=0), all the units in the control group (G=0) remain untreated in the second period (T=1), and all the units in the treatment group (G=1) receive treatments in the second period (T=1). The command assumes repeated cross sections.


To accommodate covariates, one can run preliminary regression of the outcome Y on covariates X.  Replace Y by the residuals in the ecic command.  This residualized procedure is theoretically supported by Sasaki and Wang (Forthcoming; Sec. 6).

Installation:    

    . ssc install ecic

Usage:

    . gen Y_resid = 0

    . foreach g of numlist 0/1 {

    .   foreach t of numlist 0/1 {

    .     regress Y X if G==`g' & T==`t'

    .     predict temp_Y_resid if G==`g' & T==`t', residuals

    .     replace Y_resid = temp_Y_resid if G==`g' & T==`t'

    .     drop temp_Y_resid

    .   }

    . }

    . predict resid_Y, residuals

    . ecic resid_Y G T, q(0.98)

Help:    

    . help ecic

Reference:  Sasaki, Y & Y. Wang (Forthcoming) Extreme Changes in Changes. Journal of Business & Economic Statistics . Paper.

Title
    ecic -- Executes estimation and inference for changes in changes at        extreme quantiles.

Syntax
    ecic Y G T [if] [in] [, q(real)]

Description
    ecic estimates quantile treatment effects (QTE) at extreme quantiles via        changes in changes (CIC) based on Sasaki and Wang (Forthcoming).  The        designed setting requires that all the units are untreated in the        first period (T=0), all the units in the control group (G=0) remain        untreated in the second period (T=1), and all the units in the        treatment group (G=1) receive treatments in the second period (T=1).        The command assumes repeated cross sections.
    To accommodate covariates, one can run preliminary regression of the        outcome Y on covariates X.  Replace Y by the residuals in the ecic        command.  This residualized procedure is theoretically supported by        Sasaki and Wang (Forthcoming; Sec. 6).

Option
    q(real) sets the quantile value. As an extremal quantile, it is natural        to be set either below 0.05 or above 0.95. (A warning message shows        up if q is set between 0.05 and 0.95.) The default value is q(0.99).

Example
    CIC estimation of the QTE at the 98th percentile with an outcome Y, a        covariate X, control/treatment group indicator G = 0, 1, and time        variable T = 0, 1:
    . gen Y_resid = 0    . foreach g of numlist 0/1 {    .   foreach t of numlist 0/1 {    .     regress Y X if G==`g' & T==`t'    .     predict temp_Y_resid if G==`g' & T==`t', residuals    .     replace Y_resid = temp_Y_resid if G==`g' & T==`t'    .     drop temp_Y_resid    .   }    . }    . predict resid_Y, residuals    . ecic resid_Y G T, q(0.98)

Stored results
    ecic stores the following in e():
    Scalars        e(N)           observations        e(n00)         observations with G = 0 and T = 0        e(n01)         observations with G = 0 and T = 1        e(n10)         observations with G = 1 and T = 0        e(n11)         observations with G = 1 and T = 1        e(k00)         order statistics for G = 0 and T = 0        e(k01)         order statistics for G = 0 and T = 1        e(k10)         order statistics for G = 1 and T = 0        e(k11)         order statistics for G = 1 and T = 1        e(q)           quantile value
    Macros        e(cmd)         ecic        e(properties)  b V
    Matrices        e(b)           coefficient vector        e(V)           variance-covariance matrix of the estimator
    Functions        e(sample)      marks estimation sample

Reference
    Sasaki, Y. and Y. Wang. Extreme Changes in Changes. Journal of Business &        Economic Statistics, Forthcoming. Link to Paper.

Authors
    Yuya Sasaki, Vanderbilt University, Nashville, TN.
    Yulong Wang, Syracuse University, Syracuse, NY.