Stata Command: qrkd

Stata Command: qrkd.ado

Estimation and robust inference for heterogeneous causal effects in the quantile regression kink designs (Quantile RKD) based on Chiang and Sasaki (2019). Use it when you consider a regression kink design and you are interested in analyzing heterogeneous causal effects of a continuous treatment. The method is robust against large bandwidths and arbitrary functional forms.

Installation:    

    . ssc install qrkd 

Example: 

    . regress outcome covariate

    . predict resid, residuals

    . qrkd resid running_var 

Help:    

    . help qrkd 

Note: this command is effective for continuous treatments. See rkqte.ado for binary treatments.

Reference:  Chiang, H.D. and Y. Sasaki (2019) Causal Inference by Quantile Regression Kink Designs. Journal of Econometrics, 210 (2), pp. 405-433. Paper.

Title
    qrkd -- Executes estimation and robust inference for heterogeneous causal        effects of a continuous treatment in the quantile regression Kink        designs (QRKD).
Syntax
    qrkd y x [if] [in] [, k(real) bpl(real) bpr(real) cover(real) ql(real)                 qh(real) qn(real) bw(real)]
Description
    qrkd executes estimation and robust inference for heterogeneous causal        effects of a continuous treatment in the quantile regression kink        designs (QRKD) based on Chiang and Sasaki (2019).  The command takes        an outcome variable y and a running variable or forcing variable x.        The primary results consist of estimates and a uniform 95% confidence        band of causal effects across multiple quantiles. In addition to        these primary results, the command also conducts tests of: 1. the        null hypothesis that the causal effects are zero for all the        quantiles (i.e., uniformly null causal effects); and 2. the null        hypothesis that the causal effects are constant across all the        quantiles (i.e., homogeneous causal effects) against the alternative        of heterogeneous treatment effects.
    This command works only for a continuous treatment.  For a binary        treatment, refer to rkqte.
Options
    k(real) sets the kink location for the QRKD. The default value is k(0).        (Note: the kink location itself is included as a part of the        observations with negative x.)
    bpl(real) sets the derivative b'(kink-) of the policy function b to the        left of the kink location kink. The default value is bpl(0).
    bpr(real) sets the derivative b'(kink+) of the policy function b to the        right of the kink location kink. The default value is bpr(1).
    cover(real) sets the nominal probability that the uniform confidence band        covers the true causal effects. The default value is cover(.95).
    ql(real) sets the lowest quantile at which the QRKD is estimated. The        default value is ql(.25).
    qh(real) sets the highest quantile at which the QRKD is estimated. The        default value is qh(.75).
    qn(real) sets the number of quantile points at which the QRKD is        estimated. The default value is qn(3).
    bw(real) sets the bandwidth with which to estimate the QRKD. A        non-positive argument, as is the case with the default value bw(-1),        will translate into an optimal rate.
Examples
    (y outcome variable, x running variable)
    Estimation with the policy derivatives b'(k-)=0.04 and b'(k+)=0.00 as in        the analysis of heterogeneous effects of unemployment insurance on        unemployment durations (Chiang and Sasaki, 2019, Sec. 5):
    . qrkd y x, bpl(0.04) bpr(0.00)
    Estimation of the QRKD at 10th, 20th, ..., and 90th percentiles:
    . qrkd y x, bpl(0.04) bpr(0.00) ql(0.1) qh(0.9) qn(9)
    (The default is the inter-quartile range: 25th, 50th & 75th percentiles.)
Stored results
    qrkd stores the following in r():
    Scalars        r(N)           observations        r(h)           bandwidth        r(k)           kink location        r(cover)       coverage probability
    Macros        r(cmd)         qrkd
    matrices        r(q)           quantiles        r(b)           QRKD estimates        r(CBlower)     lower bounds of confidence band        r(CBupper)     upper bounds of confidence band        r(V)           variance matrix
Reference
    Chiang, H.D. and Y. Sasaki. 2019. Causal Inference by Quantile Regression        Kink Designs.  Journal of Econometrics, 210 (2), pp. 405-433.  Link to        Paper.
Authors
    Harold. D. Chiang, Vanderbilt University, Nashville, TN.    Yuya Sasaki, Vanderbilt University, Nashville, TN.