st: mlogit coefs

[Chiara Mussida <cmussida@gmail.com>]

Dear All,
I run a mlogit model for 9 labour market outcomes (transitions between
the three states of employment unemployment and inactivity, therefore
6 transitions and 3 permanences), like:

mlogit transition male_unmarried female_married female_unmarried age
agesq ncomp child northw northe centre Ubenef edu1 edu2 health
qu1nolav qu3nolav qu2nolav nopersincnolav noothersineq qu1ot qu2ot
qu3ot if age>=15 & age<=64, b(3)

the baseline category is the permanence in the state of unemployment.
If I decide to run my mlogit only on the subsample of unemployment,
thereby reducing the number of outcomes to 3:
mlogit unemployedmale_unmarried female_married female_unmarried age
agesq ncomp child northw northe centre Ubenef edu1 edu2 health
qu1nolav qu3nolav qu2nolav nopersincnolav noothersineq qu1ot qu2ot
qu3ot if age>=15 & age<=64, b(3)

and keeping the permanence in the unemployment as baseline, I get
different coefficients signs and significance for a dicrete amount of
covariates, e.g. female_unmarried. My question is: I know that the
first mlogit refer to a larger sample that includes all the labour
force, whilst the second one only refer to the subsample of
unemployed, but this is enough to justify the different behaviour of
the coefs?

The choice to use the "extended" mlogit (9 outcomes) is related to
sample selection issues. SInce I have data to all the labour force it
is better to consider all of them and to avoid an ex ante selection
(and likely a selection bias).

Thanks,
Chiara
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/