-xthtaylor- endogenous versus exogenous

[Victoria Rogers]

Normally the model is Alpha= beta MRP + beta SMB + beta HML + beta MOM + error term

I want to test the effect of gender (variable called Male), so the model becomes Alpha= beta Male+ beta MRP + beta SMB + beta HML + beta MOM + error term

My control variables include a time-invariant called Family and time-varying variables Volatility Size

Now I'm trying to do -xthtaylor- model with the following code

 
 xthtaylor alpha MRP SMB HML MOM Male Family Volatility Size, endog(Male MRP SMB HML MOM) constant(Male)

Is that the correct code in this case? I want to use -xthtaylor- to control for all fixed effects while testing the effect of the time-invariant Male

[Sarah Edgington]

Can you explain why you're treating gender as endogenous?

Without explanations of what your variables are it's hard to figure out how you're thinking about this model. It would help if you told us about the variables in your model including which variables you think are endogenous (and why) and which variables you think are exogenous (and why). You seem to be treating gender as an time-invariant endogenous variable. I'm not extremely familiar with this type of model, but a quick skimming of the manual suggests that in that case you would need to argue that one of your other variables serves as an instrument for gender. This strikes me as a very difficult argument to make.

[Victoria Rogers]

Based on the things I read on the Internet I thought that all variables of the original code can be seen as endogenous because they have a direct effect on the independent variable.

MRP SMB HML MOM are called the 4 factors of Fama French (e,g, MRP= market risk premium) which tend to explain a lot of the independent variable alpha (excess return of an individual manager compared to the stock market)......Family can be seen as a classification, volatility is the standard deviation of stock returns, size is the size of the stock portfolio

I'm not familiar at all with this model, but I've read everything about the model on the Internet including the manual and other examples.( a few weeks ago I checked everything, now I refreshed)

[Sarah Edgington]

I would spend some time reviewing the concept of endogeneity before pursuing this model any further.
In this model you are saying that Male, MRP, SMB, HML, and MOM are correlated with the individual-level error terms and that Family, Volatility, and Size are not. I'm skeptical but I don't know enough about finance and the particular variables you're using to say one way or the other whether that makes sense. If you're going to use this model you need to be able to make a coherent argument for why it's plausible.

[Victoria Rogers]

So, it's endogenous if the variables are directly related with my panelvar and exogenous if they are indirectly related with my panelvar?

It's difficult for me to make a distinction between endogenous and exogenous because on the Internet I only saw relatively easy examples like demand and price being endogenous and weather exogenous in case of products like corn, carrots etc.

[Sarah Edgington]

When you describe a variable as endogenous or exogenous you're describing the relationship of that variable with the error term(s) in the model.
I really would not recommend using this model without a strong understanding of the concepts of endogeneity and instrumental variables.

[Victoria Rogers]

I'm also too afraid yet to use this model, therefore I'm now focussing on -xtmixed- which is a lot more intuitive. However, could you please check if it's correct or not:

http://www.statalist.org/forums/foru...r-mixed-models

[Phil Bromiley]

I don't think you're hearing the suggestions of the other commentators (both here and in your xtmixed discussion). You need to understand what endogenous really means before you try to specify anything regardless of the estimation technique. You sound like you're messing with techniques without understanding some of the basics. If sex is endogenous and you include it as an explanatory variable, then both xthtaylor and xtmixed are unlikely to give you consistent estimates. @Sarah's comment questioning sex as endogenous is that generally (but not always) sex is predetermined. It might not be predetermined if you were somehow selecting individuals based on sex.