I am estimating a production function as stochastic frontier model with technical inefficiencies modelled as a truncated normal distribution using the frontier command.
Is it sensible to model the mean of the tnormal dist with co-variate X, if X is also an input into the frontier? An example of stata code would be
frontier X, distribution(tnormal) cm(X)
There is a concern that error term depends on X which results in the it been correlated with another co-variate (X) in the frontier model?
This would be akin to stating input X increases the production frontier but for higher values of X inefficiency increases.
Many thanks
Is it sensible to model the mean of the tnormal dist with co-variate X, if X is also an input into the frontier? An example of stata code would be
frontier X, distribution(tnormal) cm(X)
There is a concern that error term depends on X which results in the it been correlated with another co-variate (X) in the frontier model?
This would be akin to stating input X increases the production frontier but for higher values of X inefficiency increases.
Many thanks
