Thanks to Kit Baum, a new package called xtdhazard is available on SSC.
xtdhazard implements the linear (first) differences instrumental variables estimator, suggested by Farbmacher & Tauchmann (2023) for dealing with unit-level unobserved heterogeneity (possibly correlated with explanatory variables) in the discrete-time hazard model. xtdhazard also implements related, non-linear control-function (cf) estimators. These procedures address the issue that, conventional (linear) fixed-effects panel estimators (within-transformation, first-differences), fail to eliminate unobserved time-invariant heterogeneity and are biased and inconsistent if the dependent variable is a binary dummy indicating an absorbing state. xtdhazard is essentially a wrapper for ivregress 2sls and, depending on which estimator is specified, for community-contributed command cfbinout. xtdhazard temporarily generates first and/or, depending on how the option difference(#) is specified, higher-order own-differences of the explanatory variables, and uses them as instruments for the levels. With estimator 2sls, estimation is by 2sls for which xtdhazard calls ivregress 2sls. With estimators logit, probit, or cloglog, xtdhazard, instead of ivregress, calls the community contributed command cfbinout to run a non-linear control-function (cf, two-stage residuals inclusion) regression; cf. Wooldridge (2015). Unlike conventional fixed effects estimation, these estimators (both, 2sls and cf) rest on the assumption that the unobserved heterogeneity is uncorrelated with the first (or higher-order) own-differences of the explanatory variables. xtdhazard replaces and extends the outdated package xtlhazard that was available on SSC in the past.
References
Farbmacher, H. and Tauchmann, H. (2023). Linear Fixed-Effects Estimation with Non-Repeated Outcomes, Econometric Reviews 42(8), 635-654.
Wooldridge, J.M. (2015). Control Function Methods in Applied Econometrics, The Journal of Human Resources 50(2), 420-445.
Best wishes,
Harald
xtdhazard implements the linear (first) differences instrumental variables estimator, suggested by Farbmacher & Tauchmann (2023) for dealing with unit-level unobserved heterogeneity (possibly correlated with explanatory variables) in the discrete-time hazard model. xtdhazard also implements related, non-linear control-function (cf) estimators. These procedures address the issue that, conventional (linear) fixed-effects panel estimators (within-transformation, first-differences), fail to eliminate unobserved time-invariant heterogeneity and are biased and inconsistent if the dependent variable is a binary dummy indicating an absorbing state. xtdhazard is essentially a wrapper for ivregress 2sls and, depending on which estimator is specified, for community-contributed command cfbinout. xtdhazard temporarily generates first and/or, depending on how the option difference(#) is specified, higher-order own-differences of the explanatory variables, and uses them as instruments for the levels. With estimator 2sls, estimation is by 2sls for which xtdhazard calls ivregress 2sls. With estimators logit, probit, or cloglog, xtdhazard, instead of ivregress, calls the community contributed command cfbinout to run a non-linear control-function (cf, two-stage residuals inclusion) regression; cf. Wooldridge (2015). Unlike conventional fixed effects estimation, these estimators (both, 2sls and cf) rest on the assumption that the unobserved heterogeneity is uncorrelated with the first (or higher-order) own-differences of the explanatory variables. xtdhazard replaces and extends the outdated package xtlhazard that was available on SSC in the past.
References
Farbmacher, H. and Tauchmann, H. (2023). Linear Fixed-Effects Estimation with Non-Repeated Outcomes, Econometric Reviews 42(8), 635-654.
Wooldridge, J.M. (2015). Control Function Methods in Applied Econometrics, The Journal of Human Resources 50(2), 420-445.
Best wishes,
Harald
