I am working with longitudinal bond data. The bonds have some fixed characteristics such as maturity, size of issue, etc., and some time-variant variables such as profitability, leverage, etc. This implies that if I run a panel fixed effects OLS regression using plm command in R, my fixed (time-invariant) variables drop and I have coefficients for the time-varying variables. I use the following command for panel fixed effects/within regression:
library(plm)
result <- plm(SPREAD ~ X + lnISSUESIZE + ROA + LEV + Maturity + factor(t), index = "id", model = "within", data = df)
summary(result)
I perform panel quantile regression with individual penalized fixed effects using the rqpd command to estimate coefficients across quantiles:
library(quantreg) library(rqdp)
rqpd_model <- rqpd::rqpd(SPREAD ~ X + lnISSUESIZE + ROA + LEV + Maturity + as.factor(t)| as.factor(id), panel(method = "pfe", taus=c(0.1, 0.2, 0.3, 0.4, 0.50, 0.6, 0.70, 0.8, 0.9), tauw=rep(1/9, 9)) , data = df, control = quantreg::sfn.control(tmpmax = 50000))
summary.rqpd(rqpd_model)
In panel quantile regression, I get single set of estimates for the bond fixed effects (for all quantiles) and the fixed variables do not get omitted (like in OLS panel FE regression above) and I get coefficient estimates for fixed coefficients too.
Could anyone please help me understand why that happens? Any help would be appreciated!
Thank you!
I tried panel quantile regression with penalized fixed effects using rqpd command and I was expecting that the time-invariant categorical variables would drop.