[Limdep Nlogit List] Simulataneous equations models with multi-nominal dependent variables

Christer Berglund Christer.Berglund at ltu.se
Mon Mar 10 23:26:27 EST 2008

Dear subscribers,


I am currently up to the task of assessing the relationship between households’ recycling efforts and a host of socioeconomic variables (using NLOGIT 3.0). I measure recycling efforts by a five-point categorical scale, ranging from “recycle nothing” up to “recycle everything”. The presence of moral norms is a factor that can be expected to be an especially important determinant of household recycling efforts. Since a number of socio-economic variables may be important for both recycling efforts and moral norms, I would like to estimate an equation system that makes it possible to estimate both the direct effect of a variable on recycling efforts and the indirect effect through its impact on moral norms (measured by a seven-point categorical scale). The model thus looks like this (following your notation in the modeling guide):


Prob(Moral norm)=F1(β1’x1) 


Prob(Recycling effort)= F2(β2’x2+γMoral norm) 


I learned from the NLOGIT software package that it is easy to run if the dependent variables are binary (see E17.6 Simultaneous Equations Models in LIMDEP 8.0 Econometric Modeling Guide). However, although it is possible to create a binary variable of the recycling effort (e.g., more than half equal one, else zero), it is much less feasible to do the same with the expressed moral norm.   


So, do anyone have aclue how I should go on about if I were to estimate this with multi-nominal (ordered) dependent variables?


Thanks in advance.


All the best,



Christer Berglund, Ph. D.
Assistant Professor in Economics
Luleå University of Technology
Division of Economics
SE-971 87 Luleå
E-mail: Christer.Berglund at ltu.se
Phone: +46 920 492977
Fax: +46 920 492035
Mobile: +46 70 219 80 67


More information about the Limdep mailing list