Impact of an average change on dependent variable over time

[geoffrey croes]

Hello! For my thesis I wanted to test a hypothesis which requires a model that I wouldn't know how to design myself. For a panel data (12 years) sample of 500 companies, I have a dummy variable that equals 1 if a company discloses a certain information variable, and another dummy variable that equals 1 if a company is loss-making (not profitable). I should test whether companies stop disclosing the information when they become profitable.

example: company A is loss-making from 2007-2013 and discloses the information, and is profitable from 2014 and onwards and stops disclosing the information
--> I want to test if this, on average, holds for every company

My Stata knowledge is kind of limited, I only got a minor course about it last year, and I really have no idea how I would be able to test this

I hope there is someone here who could help me!

[Carlo Lazzaro]

Geoffrey:
welcome to this forum.
You should declare you dependent variable to help interested listers helping you out in turn.

[geoffrey croes]

Sure!

My dependent variable: DISCLOSURE (= 1 if a certain company discloses the information in a specific year; 0 if else)

I also have information for each year if a firm made a loss: LOSS_MAKING (= 1 if a certain company made a loss in a specific year; 0 if else)

(+ some control variables of course)

So really my first problem is how to set up the independent variable for it to show "whether a firm is no longer loss making", and than the potential impact of it "no longer being loss-making" on the chance that a firm discloses the information. So I guess I need some kind of LPM or probit model, but I wouldn't know how to design the model (and the indep. variable)

[geoffrey croes]

Sure!


I also have information for each year if a firm made a loss: LOSS_MAKING (= 1 if a certain company made a loss in a specific year; 0 if else)

This would of course be the (basis for the) independent variable, just to clarify

[Oscar Ozfidan]

here is a starting point

logit LOSS_MAKING i.DISCLOSURE
if coefficient of DISCLOSURE is significant that would indicate it matters. but that conclusion could be driven by the effect of time (good economic environment this year vs next) and/or company (a company struggling vs prospering for reasons other than disclosure effect)
so
logit LOSS_MAKING i.DISCLOSURE i.year i.company should be a good starting point for further investigations etc there are 500 companies which means 500 coefficients for each company so you will see a pretty long coefficient list

[geoffrey croes]

Okay, thank you very much!