Dear all,
Anybody familiar with how to generate IV test statistic (weak identification and underidentification tests: Kleibergen-Paap) with cluster-robust errors, and particularly with bootstrapped errors (boottest) ? David Roodman
I'm running the IV panel regression below. I obtain the regression coefficients using 'ivreghdfe' and obtain Wild-bootstrapped t-statistics and confidence intervals using 'boottest'.
I noticed that the IV test statistics may differ substantially, based on the specification of the error clustering (see the output below).
Based on this difference, I wonder how the IV test statistics can be obtained based on Wild-bootstrapped errors (WRE) from 'boottest'.
Many thanks,
Arjan
Anybody familiar with how to generate IV test statistic (weak identification and underidentification tests: Kleibergen-Paap) with cluster-robust errors, and particularly with bootstrapped errors (boottest) ? David Roodman
I'm running the IV panel regression below. I obtain the regression coefficients using 'ivreghdfe' and obtain Wild-bootstrapped t-statistics and confidence intervals using 'boottest'.
I noticed that the IV test statistics may differ substantially, based on the specification of the error clustering (see the output below).
Based on this difference, I wonder how the IV test statistics can be obtained based on Wild-bootstrapped errors (WRE) from 'boottest'.
Code:
global i firm_id global t i.year * Results using CRVE clustering by country // there are 32 countries ivreghdfe $deparvar $indepvars_exog $t ($indepvar_endog = $indepvar_instrument), absorb($i) cluster(country) *** Output: * Underid. test statistic (Kleibergen-Paap): 8.6621*** (p-value 0.0086) * Weak id. test statistic (Kleibergen-Paap): 69.7332*** * Results using CRVE clustering by country and sector // there are 32 countries and 15 sectors ivreghdfe $deparvar $indepvars_exog $t ($indepvar_endog $indepvar_endog_lag1 $indepvar_endog_lag2 = $indepvar_instrument $indepvar_instrument_lag1 $indepvar_instrument_lag2), absorb($i) cluster(country sector) *** Output: * Underid. test statistic (Kleibergen-Paap): 3.1035* (p-value 0.0781) * Weak id. test statistic (Kleibergen-Paap): 106.7399*** * Results using Wild bootstrap errors (WRE), clustering country and sector // this is my baseline specification for computing the errors and t-statistics boottest $indepvar_endog = 0, cluster(country sector) bootcluster(sector) nograph seed(999) reps(999)
Arjan
Comment