MLwin & multiprocess multilevel modeling

Welcome to the forum for MLwiN users. Feel free to post your question about MLwiN software here. The Centre for Multilevel Modelling take no responsibility for the accuracy of these posts, we are unable to monitor them closely. Do go ahead and post your question and thank you in advance if you find the time to post any answers!

Remember to check out our extensive software FAQs which may answer your question: ... port-faqs/
Post Reply
Posts: 30
Joined: Mon Aug 15, 2011 1:54 pm

MLwin & multiprocess multilevel modeling

Post by julia1633 »

Greetings again,

This time I have a question that is related to the potential of MLwin. Our another current research project uses cmp (Roodman) to estimate jointly the firm decision to invest and sources of firm financing. So, we combine Heckman selection model and SURE framework. However, our data is hierarchical (a mix of firm and country-level data) and ideally we need to use multilevel modeling. So far, we have been controlling for this using country dummies but this does not allow estimating random coefficient effects. Can Mlwin handle multiprocess multilevel modelling. In other words, can we estimate jointly Heckman probit to model a decision to invest and SURE Tobits equations to model sources of finance while also accounting for hierarchical structure of the data? We have a 3-level model as we
also have a time dimension. Altogether we have a panel of 23,000 firms
for the periods 2002, 2005, 2007, 2008, 2009. Initially, we thought we could overcome this by using country-year as one level. However, we have a
crisis dummy for years 2008 and 2009, and it kept dropping once we use
country-year dummies in our SURE Tobits equations. The 'crisis' dummy is
of crucial importance to our research question as we investigate its
effect on credit constraints and sources of financing. Will Mlwin handle estimation of 6 equations jointly? If so, with using Heckman selection model in aML is it essential to have exclusion restrictions (to introduce one identification variable within investment equation, but leave it out in SURE equations)? I would very much appreciate your advice. We use Tobit models to estimate firms' sources of finance within ML SUR framework, but we are concerned that our estimates may be inconsistent due to possible heteroscedasticity. Is there any way to control for this in aML? (4) We also have an observable
endogenous variable (employment growth) which we use in both our Heckman
probit investment equation and Tobits sources of finance equations and
itself as a dependent continuos variable within the ML SUR framework.
Could you please advise if using it as an explanatory variable in probit
and tobits and on its own as a dependent variable still gives consistent
estimates with aML? (5) Alternatively, we can drop it as an explanatory
variable but still keep a continuous model of employment growth, so
effectively estimating three types of equations: Heckman probit for
investment; SUR for 5 sources of finance. So far I only come across with
aML software which we believe could provide a solution to our empirical problem.
However, it has its own limitations. Before moving to explore it in more detail I
would be very grateful if Mlwin allows to estimate multiprocess multilevel model.

Many thanks,

Post Reply