[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]
st: Which discrete choice model should I use?
I have an unbalanced panel of firms with a maximum of 6 (and a minimum 3) observations on each firm.
I wish to model the characterisitcs of firms that restructure (change) via an LBO (levergaed buyout).
I have a sample of LBOs and have constructed a stratified random control sample. There are 3 levels of decision: (1) do not change or change via an LBO, (2) If a firm decides to undertake an LBO it can either do it via a management buyout (MBO) or management buyin (MBI), (3) if a firm has been through an MBO or MBI it can stay as an MBO or MBI or it can exit from an MBO or MBI. Thus, (I think) my decision tree looks like this:
Corporate change decision
Do not change (control sample) LBO
| / \
non-LBO (control) MBO MBI
| / \ / \
non-LBO (control) Stay Exit Stay Exit
MBO MBO MBI MBI
To me it looks like I should be using a nested logit, but I am stuck constructing the depvar (though the problem might be to do with how the data are organizaed) in that given each firm at any moment in time is either: non-LBO, MBO, MBI, stay MBO, exit MBO stay MBI, or exit MBI and so my depvar is simply a column of 1s (ones). Note I am trying to model the charactersitics of each "type" of firm (type being non-LBO, stay MBO, exit MBO, etc.) rather than the decision itself. The latter would result in a variable where one indicates a decison is made (zero otherwise) while modelling the characteristics means that, say, a variable takes the value one to indicate the MBO governance structure is in place (zero otherwise).
Not that I have modelled the MBO and MBI decision using a multinomial logit and the IIA condition is satisfied. My problem comes when I also want to model 'exit' firms characteristics as it is not possible to be an exit firm without having first been either an MBO or MBI.
I'm sorry my email is convoluted. Sample data are attached so you can see how I have constructed my binary variables and to give some indication of the nature of the data
Any help is very much appreciated.
This message has been checked for viruses but the contents of an attachment
may still contain software viruses, which could damage your computer system:
you are advised to perform your own checks. Email communications with the
University of Nottingham may be monitored as permitted by UK legislation.
Description: Binary data