Propensity*score matching or heckman self selection bias.

[andy macrobarty]

I have a question concerning if I need to use a propensity score matching or heckman self selection bias.

These are 'my dependent categories 'TSR'', ''EPS'' or ''TSR and EPS category). TSR stands for plans with total shareholder return (share price measure relative to peer group)

I look at the impact of the market beta on the selection of ''TSR'', ''EPS'' or ''TSR and EPS category) by utilizing a multinomial logit model.

In latter part of my paper, I only look at sample of plans with TSR conditions thus removing or ignoring condition of plans with EPS CONDITION ( taking TSR plans only and taking TSR plans from the joint category) sample and look at the effect on market beta on the characteristics of relative performance plans (threshold targets, or peer group selection) in TSR plans. Some of the characteristics are peer group choices so used as 0 or 1 (logit model). Else, some of these are threshold targets, this I employed, normal regression model.


I run these model separately, now I am thinking whether there is need to employ propensity or heck man selection because in second part I am only consider plans with TSR condition.If there is any bias if any

[Phil Bromiley]

You didn't get a quick answer. Your question is a bit complex and hard to understand. Naturally, following the FAQ on asking questions (Stata code in code delimiters, readable Stata output, and sample data using dataex) will also help.

I have trouble understanding "at the effect on market beta on the characteristics of relative performance plans". So, you're doing a outcome model with the categories as iv's? Generally, you can select on the rhs without problems.

I'd consider whether you may be creating a endogeneity problem when beta is on both sides of equations (but I'm not sure that is what you're proposing).

[andy macrobarty]

I had some independent variables in both separate models. In one model, I run a multinomial logit model with three categories...... (TSR, EPS, and EPS and TSR) categories.I look at effect of market beta on these three difference choices employed

In second model, I only pick observations from TSR and also TSR from (EPS and TSR joint category) which eventually means the same, But, I look at other aspects of these categories. What are further characteristics in these plans???????????????

Since, I am selecting TSR only is there the selection bias or heckman selection best?
Then, I use different instruments at the first stage....

[andy macrobarty]

Is it best to use heckman selection or propensity score matching? I am looking at TSR only category and also EPS and TSR joint from a perspective of TSR only

[Priyesh VP]

I think you should refine your question to get attention from the forum members. I suppose you need clarification on when/not to use PSM and Heckamn correction