[Limdep Nlogit List] nested logit of occupational choice

[Tulika Narayan]

Hi,

I have a nested logit model of the choice to not work, and work
(professional, administrative, primary) . One of my branches is degenerate
so I have used the appropropriate model based on the paper by Green et al
(RU2).  I have a bunch of variables on individual and household
characteristics and only one variable that varies by occupations (the
average expected wage in each occupation). I do not have any other variables
in the data that are occupation specific. The objective of the paper is to
see how illnesses affect occupational choice.

The problem is that model is very finicky. I am not able to estimate the
full model (includes all illnesses) and it gives me a divide by zero.  I am
trying to figure out if it is a particular variable that causes the problem.
But here is the model. Is there anything you see that I am doing wrong or
any ideas on what I should look at to find out why the model is so unstable.
Could it be that occupational choice models do not work well if I do not
have adequate information on other choice specific data (occupation specific
variables?) But the multinomial model works just fine.

I would appreciate your inputs!


NLOGIT ; Lhs =choose

; Choices = nw, profman, admin, servpri

; tree = nowork(nw), work(profman, admin, servpri)

; wts =wgttot

; ru2;

; model:

U(nw, profman, admin, servpri) =

<a0, a1, a2, 0> +

<0, bwj1,bwj2,bwj3>*cc_earn+

<0, bh1j1,bh1j2,bh1j3>*h_heart+

?<0, bh2j1,bh2j2,bh2j3>*h_diabet+

<0, bh3j1,bh3j2,bh3j3>*h_cancer/

?<0, bh4j1,bh4j2,bh4j3>*h_asthma+

?<0, bh5j1,bh5j2,bh5j3>*h_migrai+

?<0, bh6j1,bh6j2,bh6j3>*h_lung

?<0, bh7j1,bh7j2,bh7j3>*h_kidney+

?<0, bh8j1,bh8j2,bh8j3>*h_arthri/



U (nowork, work) = <0, bri3_w>*i_educ_n+

<0, bri4_w>*i_educ_b+

<0, bri5_w>*i_educ_g+

<0, bri7_w>*i_white+

<0, bri8_w>*i_black+

<0, bri9_w>*i_asian+

<0, bri10_w>*i_hisp+

<0, bri11_w>*rural+

<0, bri1_w*i_female+

<0,bri2_w> *i_maried+

<0,bri6_w> *i_age+

<0,bri14_w>*totadu+

<0,bri15_w>*i_totchi+

<0, bri13_w>*h_total+

<0, bri12_w>*health+

<0, bri16_w>*i_ageh+

<0, bh1_w*h_bmi+

<0, bh2_w>*h_depres

 ;showtree; start = logit ; maxit = 5000;describe

;all; effects: cc_earn[*] ; prob =pre_pr; Cprob =cond_pr; ivb = iv;
utility=util;

$


-- 


        o
....O-/-`O...to work.
_______________
Tulika A. Narayan
Volunteer for AID

www.aidindia.org