[Limdep Nlogit List] Abowd model

[Sarah Rice]

Hi,
I am trying to run a partial observability model, specifically the Abowd model.  One of my dummy variables (included in both equations) generates a large standard error in the biprobit stage of the model (the OLS output before that looks fine). Here is my code:

NAMELIST ; X1 = ONE,PREV_RES,NEG_IB,W_LNMKTV,SCL_AUD0,
SCL_NONA,DOM_AUD, DYEAR2,DYEAR3,DYEAR4,DYEAR5
; X2 = ONE,W_LNMKTV,NEG_IB,W_XFIN_A,
LIT_IND_,AUDCHANG,SCL_NONA,CFO_CHAN,CEO_CHAN,DYEAR2,DYEAR3,DYEAR4,DYEAR5$ 
PROBIT ; LHS = ICW1 ; RH1 = X1 ; RH2 = X2 
; SELECTION 
; TLF = 0.0001 ; TLB = 0.0001 ; TLG = 0.0001 
;MAXIT=1000 
; MARGIN$

Here is the output, the problem is with the MW302 variable in the first equation:

Index function for probability
 Constant -.4818966208      .51698754        -.932   .3513
 MW302     4.608961997      3399.4395         .001   .9989  .75819672E-01
 PREV_RES  .4675092916      .16033951        2.916   .0035  .42418033
 NEG_IB    .5123620086      .18698872        2.740   .0061  .31147541
 W_LNMKTV -.5989551894E-01  .70235377E-01    -.853   .3938  6.4948502
 SCL_AUD0  .9713727510      .21995908        4.416   .0000  .63548487
 SCL_NONA  .1220528478      .68791035         .177   .8592  .12731751
 DOM_AUD  -.5566853243      .31217491       -1.783   .0745  .90778689
 DYEAR2    .2316147399      .21451992        1.080   .2803  .29508197
 DYEAR3    .3282624853      .26865556        1.222   .2218  .16598361
 DYEAR4   -.2061929855      .35686753        -.578   .5634  .10655738
 DYEAR5    1.021448995      3.1479586         .324   .7456  .45081967E-01
 
Constant  4.429475262      1.7964627        2.466   .0137
 MW302    -1.232412343      .76963506       -1.601   .1093  .75819672E-01
 W_LNMKTV -.3211294293      .17924421       -1.792   .0732  6.4948502
 NEG_IB    .9791616539      .56448514        1.735   .0828  .31147541
 W_XFIN_A -8.596653693      3.6545031       -2.352   .0187  .19497226E-01
 LIT_IND_ -.7386809943      .59401485       -1.244   .2137  .34631148
 AUDCHANG  4.071501323      2.2204806        1.834   .0667  .11475410
 SCL_NONA -2.806828918      1.4636268       -1.918   .0551  .12731751
 CFO_CHAN  .2565828958E-01  .41137382         .062   .9503  .33401639
 CEO_CHAN  1.936762588      .97871717        1.979   .0478  .21516393
 DYEAR2   -.7450245427      .57612396       -1.293   .1960  .29508197
 DYEAR3   -1.669746440      .83050800       -2.011   .0444  .16598361
 DYEAR4   -2.184862798      1.2794867       -1.708   .0877  .10655738
 DYEAR5   -5.015709500      1.9902950       -2.520   .0117  .45081967E-01







 Since it is not a critical variable I omitted it from first one and then both equations, and in both cases when I re-run the model I get this error in red.

Models - estimated variance matrix of estimates is singular
Current estimated covariance matrix for slopes is singular.

I have been trying to figure out why this is the case when I omit the variable.  If collinearity were a problem I would think omitting it would help.  Also, I thought maybe the model was too cumbersome, but again I would think omission of a variable would help that. I am also finding the model is sensitive to the addition/subtraction of dummy variables in general and I get this error often when I play with the inclusion/exclusion of some dummy variables.  I have made sure I have sufficient variance between these variables and the LHS but so far am at a loss.  Also, my continuous variables are scaled so they are not vastly different from the rest of the data.  Does anyone have insights as to what is going on here?  I've been wracking my brain for days on this.  Thanks much in advance!


Sarah Rice
Assistant Professor of Accounting
2100 Hillside Road, Unit 1041
The University of Connecticut School of Business
Storrs, CT 06269-2

Office Phone: 860-486-1928
Dept Fax:  (860) 486-4838