help treatrew
-------------------------------------------------------------------------------

Title

    treatrew -   Estimation of Average Treatment Effects by reweighting on
                   propensity score


Syntax

        treatrew outcome treatment [varlist] [if] [in] [weight],
                 model(modeltype) [GRaphic conf(number) vce(robust)]


    fweights, iweights, and pweights are allowed; see weight.



Description

    treatrew estimates Average Treatment Effects by reweighting on propensity
    score as proposed by Rosenbaum and Rubin (1983) in their seminal article.
    Depending on the model specified, treatrew provides consistent estimation
    of Average Treatment Effects under the hypothesis of "selection on
    observables".  Conditional on a pre-specified set of observable exogenous
    variables x - thought of as those driving the non-random assignment to
    treatment - treatrew estimates the Average Treatment Effect (ATE), the
    Average Treatment Effect on Treated (ATET) and the Average Treatment
    Effect on Non-Treated (ATENT), as well as the estimates of these
    parameters conditional on the observable factors x (i.e., ATE(x), ATET(x)
    and ATENT(x)).  Parameters standard errors are provided either
    analytically (following Wooldridge, 2010, p. 920-930) and via
    bootstrapping.  treatrew assumes that the propensity score specification
    is correct.

    According to the syntax:

    outcome: is the target variable over which measuring the impact of the
        treatment

    treatment: is the binary treatment variable taking 1 for treated, and 0
        for untreated units

    varlist: is the set of pre-treatment (or observable confounding)
        variables

     
Options
    
    model(modeltype) specifies the model for estimating the propensity score,
        where modeltype must be one out of these two:  "probit" or "logit".
        It is always required to specify one model.

    graphic allows for a graphical representation of the density
        distributions of ATE(x), ATET(x) and ATENT(x).

    vce(robust) allows for robust regression standard errors in the probit or
        logit estimates.

    conf(number) sets the confidence level of probit or logit estimates equal
        to the specified number.  The default is number=95.


  modeltype_options           Description
  -----------------------------------------------------------------------------
  Model
  probit                      The propensity score is estimated by a probit
                                  regression
  logit                       The propensity score is estimated by a logit
                                  regression
  -----------------------------------------------------------------------------


    treatrew creates a number of variables:


        ATE_x is an estimate of the idiosyncratic Average Treatment Effect.

        ATET_x is an estimate of the idiosyncratic Average Treatment Effect
        on treated.

        ATENT_x is an estimate of the idiosyncratic Average Treatment Effect
        on Non-Treated.



    treatrew returns the following scalars:

        e(N) is the total number of (used) observations.

        e(N1) is the number of (used) treated units.

        e(N0) is the number of (used) untreated units.

        e(ate) is the value of the Average Treatment Effect.

        e(atet) is the value of the Average Treatment Effect on Treated.

        e(atent) is the value of the Average Treatment Effect on Non-treated.


Remarks 

    The treatment has to be a 0/1 binary variable (1 = treated, 0 =
    untreated).

    It is assumed that the probit or logit model is correctly specified.

    Please remember to use the update query command before running this
    program to make sure you have an up-to-date version of Stata installed.


Examples

    *** EXAMPLE ON "JTRAIN2.DTA" ***

   . #delimit ;
   . xi: treatrew re78 train educ black re75 unem74 unem78 lre74 agesq mosinex 
> ,
   . vce(robust) conf(90) model(probit) gr  
   . ;

   . #delimit ;
   . xi: treatrew re78 train educ black re75 unem74 unem78 lre74 agesq mosinex 
> ,
   . model(logit) vce(robust) conf(90)  gr 
   . ;
   

    *** EXAMPLE ON HOW TO BOOTSTRAP STD. ERR. FOR "ATET" AND "ATENT" ***

   . #delimit ;
   . xi: bootstrap ate=e(ate)  atet=e(atet)  atent=e(atent), rep(10):
   . treatrew re78 train educ black re75 unem74 unem78 lre74 agesq mosinex ,
   . model(logit) vce(robust) conf(90)  gr 
   . ;

   
References

    Cameron, A.C., and P.K. Trivedi. 2005. Microeconometrics: Methods and
        Applications.  Chapter 25. Cambridge University Press, New York.

    Cerulli, G. 2012. Ivtreatreg: a new STATA routine for estimating binary
        treatment models with heterogeneous response to treatment under
        observable and unobservable selection, Working Paper Cnr-Ceris, N°
        03/2012.

    Rosenbaum, P., and D.B. Rubin. 1983. The Central Role of the Propensity
        Score in Observational Studies for Causal Effects. Biometrika, 70,
        41-55.

    Wooldridge, J.M. 2002.  Econometric Analysis of Cross Section and Panel
        Data.  Chapter 18. The MIT Press, Cambridge.

    Wooldridge, J.M. 2010.  Econometric Analysis of Cross Section and Panel
        Data, 2nd Edition.  Chapter 21. The MIT Press, Cambridge.


Acknowledgments

    I wish to thank Enrico Viarisio of the Ceris-CNR technical staff for his
    help in formatting the Technical Report accompanying this routine.


Author

    Giovanni Cerulli
    Ceris-CNR
    Institute for Economic Research on Firms and Growth, National Research
        Council of Italy
    E-mail: g.cerulli@ceris.cnr.it



Also see

    Online:  treatreg, ivregress, ivtreatreg, pscore, psmatch2, nnmatch