Hello everyone. I know this subject has been brought up multiple times, but with my limited econometric knowledge and experience I am still unable to solve the issue for my own model.
My model aims to investigate the effects of population density in the place a CEO grew up on corporate risk-taking behavior later in life, using panel data. Additionally, I want to investigate the effect of childhood poverty rates as a moderator variable. Since risk taking is dependent on the industry a firm operates in, I want to use industry fixed effects. I also want to incorporate time fixed effects since my data contains crisis-years.
However, my independent variable, moderator variable, and variable for industry are all time-invariant variables, making the fixed effects model unsuited. To elaborate, population density and poverty rates are time-invariant because they are based on CEO childhood measures, which are constant for each firm/CEO year. What model can I use to deal with this issue? I found somewhere that I can do a two-stage regression, where I first can get a within estimation, and then get the individual mean of these residuals and regress over the time invariant variables and the individual mean of the time variant regressors. However, I have no clue what this means, whether it is appropriate for my model, or what commands to use in Stata.
Furthermore, I am unsure how to define my panel data with the xtset command. I can use xtset gvkey year, which causes it to set the data to panel using a company identifier. I can also use xtset firm_ceo_id year, where firm_ceo_id is a unique identifier for each CEO + company combination, because a number of CEOs was CEO of multiple companies at the same time.
To conclude, I am left with the following questions.
1. What alternative model can I use to deal with the time-invariant variables, and what commands should I use in Stata?
2. How should I set my panel data using the xtset command?
Thanks in advance.
My model aims to investigate the effects of population density in the place a CEO grew up on corporate risk-taking behavior later in life, using panel data. Additionally, I want to investigate the effect of childhood poverty rates as a moderator variable. Since risk taking is dependent on the industry a firm operates in, I want to use industry fixed effects. I also want to incorporate time fixed effects since my data contains crisis-years.
However, my independent variable, moderator variable, and variable for industry are all time-invariant variables, making the fixed effects model unsuited. To elaborate, population density and poverty rates are time-invariant because they are based on CEO childhood measures, which are constant for each firm/CEO year. What model can I use to deal with this issue? I found somewhere that I can do a two-stage regression, where I first can get a within estimation, and then get the individual mean of these residuals and regress over the time invariant variables and the individual mean of the time variant regressors. However, I have no clue what this means, whether it is appropriate for my model, or what commands to use in Stata.
Furthermore, I am unsure how to define my panel data with the xtset command. I can use xtset gvkey year, which causes it to set the data to panel using a company identifier. I can also use xtset firm_ceo_id year, where firm_ceo_id is a unique identifier for each CEO + company combination, because a number of CEOs was CEO of multiple companies at the same time.
To conclude, I am left with the following questions.
1. What alternative model can I use to deal with the time-invariant variables, and what commands should I use in Stata?
2. How should I set my panel data using the xtset command?
Thanks in advance.
