Good afternoon,
I am currently working on a research paper in which I look at the effect of various human capital factors of start-up owners on the chance that they receive investment by a venture capitalist (VC). My sample is a single cohort panel of new businesses which are tracked for a period of 8 years, every year they answer the same questionnaire. My dependent variable is if they receive investment by a VC in a specific year, it is therefore a dummy variable with an outcome of 0 or 1. I therefore use xtlogit but it seems to me that i still have to decide whether I use a fixed effects (fe) model or random effects (re) model, for this i run a Hausman test. Here comes the rub, when i run a Hausman test stata informs me that in the fe model 8881 observations are omitted because these groups (new ventures) only have positive or negative outcomes, this makes sense as most new ventures in my dataset (~98%) will never receive investment from a VC, however it leaves me with only 172 observations over 36 companies (groups) in the fe model. I imagine this creates a very heavy bias in the fe model, it also makes all of my independent variables very insignificant (which is quite surprising as there is quite a bit of research on the topic which points the other direction). The re model does encompass all 9053 observations over 2207 companies, however, the Hausman test then proceeds to indicate that i should run a fixed effects regression.
My first question would then be, is there any validity to his hausman test as it excludes such a large amount of relevant observations?
Secondly, could I under these circumstances just decide to use an re model? And if not, is there a different approach I can take (I understand this last question will be quite difficult to answer without more specific knowledge but I am curious nonetheless).
Below I have added the code I used for the hausman test
In the f3_eq_invest_vent_cap = if the company receives investment by a VC
The other variables are indicators of the amount of human capital present in the venture.
Kind regards,
Floris
I am currently working on a research paper in which I look at the effect of various human capital factors of start-up owners on the chance that they receive investment by a venture capitalist (VC). My sample is a single cohort panel of new businesses which are tracked for a period of 8 years, every year they answer the same questionnaire. My dependent variable is if they receive investment by a VC in a specific year, it is therefore a dummy variable with an outcome of 0 or 1. I therefore use xtlogit but it seems to me that i still have to decide whether I use a fixed effects (fe) model or random effects (re) model, for this i run a Hausman test. Here comes the rub, when i run a Hausman test stata informs me that in the fe model 8881 observations are omitted because these groups (new ventures) only have positive or negative outcomes, this makes sense as most new ventures in my dataset (~98%) will never receive investment from a VC, however it leaves me with only 172 observations over 36 companies (groups) in the fe model. I imagine this creates a very heavy bias in the fe model, it also makes all of my independent variables very insignificant (which is quite surprising as there is quite a bit of research on the topic which points the other direction). The re model does encompass all 9053 observations over 2207 companies, however, the Hausman test then proceeds to indicate that i should run a fixed effects regression.
My first question would then be, is there any validity to his hausman test as it excludes such a large amount of relevant observations?
Secondly, could I under these circumstances just decide to use an re model? And if not, is there a different approach I can take (I understand this last question will be quite difficult to answer without more specific knowledge but I am curious nonetheless).
Below I have added the code I used for the hausman test
In the f3_eq_invest_vent_cap = if the company receives investment by a VC
The other variables are indicators of the amount of human capital present in the venture.
Code:
xtlogit f3_eq_invest_vent_cap Total_Age Total_other_bus_started Total_edu_owner c3a_owner_operators, fe est store fe xtlogit f3_eq_invest_vent_cap Total_Age Total_other_bus_started Total_edu_owner c3a_owner_operators, re est store re hausman fe re
Kind regards,
Floris
Comment