st: Wooldridge's solution for dynamic probit models

[Sophia Rabe-Hesketh <sophiarh@berkeley.edu>]

In a previous thread on Statalist (see link below),
there was a discussion on "Estimating Wooldridge's CML for
Dynamic Probit model: On the right path?".
http://www.stata.com/statalist/archive/2013-02/msg00433.html

The original email suggested using time averages
of time-varying covariates. Responding to that email,
Jeff Wooldridge stated:

"...simulation evidence seems to suggest that for certain
covariate processes using the time averages can lead to
serious bias"

Such simulation evidence can be found in our recent paper:

Rabe-Hesketh, S. and Skrondal, A. (2013). Avoiding biased versions
of Wooldridge's simple solution to the initial conditions problem.
Economics Letters 20, 346-349.
http://www.sciencedirect.com/science/journal/01651765/120/2
(we're happy to send the paper by email to those who cannot access
the above link)

Briefly, we show that bias can be severe if the time averages
or "within means" are averages over all periods including the
initial period (this is what happens naturally if
-egen mean- is used). The problem can be avoided by either
also including the initial-period covariates or by using
covariates at each period as originally proposed by Wooldridge.

We hope this is useful,

Sophia Rabe-Hesketh and Anders Skrondal
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/