Tobit Panel

[Philipp Leppert]

Hello everyone,

I'm currently struggeling with a tobit panel regression in STATA.

My task is to observe the Impact of Market Timing on the Capital Structure of German firms. The data set consists of 25 firms with data reaching from 2001-2011.
As an example I got a normal tobit regression by a former student dealing with the subject (no panel approach).
E.g.: The Dep. Var. is "Growth in Assets" --- the Indep. Vars. are "Market/Book-Ratio(MB), Proftiability, Tangibility, Size, Book Leverage(DAB)".

I used the xttobit syntax with lower limit zero yielding the same estimates as in the example tobit regression. (so far so good, but no improvement of the estimates)
So my professor told me to try a Fixed Effect Tobit Model instead. As there is only the "Pantob" Syntax from Honoré available I tried to run this model but it does not work out since I am not really understanding how the syntax works (especially the [if, in] part).

Has anyone an idea how to apply the Pantob syntax to the above-mentioned example? Or perhaps a suggestion how I also could conduct the estimation beside the "Pantob" syntax?

Many thanks in advance
Philipp

[Rich Steinberg]

The syntax is pantob 'depvar' 'indepvar' 'name of variable indexing your cross-sectional unit (presumably firms)' if (usual format for subsetting if -- for example, if you wanted to exclude firms where the dummy variable 'whatup' takes the value of zero, including the ones, it would say if whatup ==1). Then a comma, separating options. The default is the estimator Honore developed using a quadratic loss function. If you want his other estimator (using an absolute value loss function) type 'absloss' . The other option is on standard errors -- if you type 'bootstrap' it will calculate by bootstraps. Finally, there is a 'details' option I haven't used. Note it can take a long time to run this, particularly with bootstrap.