Hello guys, I am new here, many thanks for your time in considering my problem.
I have a question regarding whether I should be making use of the panel data in testing.
I am interested in employee contract, and my dataset has a structure like this:
I’m seeing (a) what determines Contract Duration (#years the employee is signed for) and Wage (US$), and (b) how Contract Duration and Wage affect firm profits.
In answering (a), I treat my data as cross-sectional, because Contract Duration and wage are determined by the firm-level and employee-level variables at the time of signing the contract.
In answering (b), I treat my data as panel, because I wanted to see how Tenure (# years the employee has stayed in the firm) and Wage affect firm profits.
Now, the reviewer is asking if I have a mediation model, where Contract Duration and Wage mediate the antecedents to and outcome of signing the contract.
My question is, how do I deal with the mediation in my problem?
I have a question regarding whether I should be making use of the panel data in testing.
I am interested in employee contract, and my dataset has a structure like this:
| FirmID | EmployeeID | Year | ContractDuration | Tenure | Wage | FirmProfit | Firm-level vars | Employee-level vars |
| 1 | 1 | 2001 | 3 | 1 | ||||
| 1 | 1 | 2002 | 3 | 2 | ||||
| 1 | 1 | 2003 | 3 | 3 | ||||
| 2 | 2 | 2002 | 1 | 1 |
In answering (a), I treat my data as cross-sectional, because Contract Duration and wage are determined by the firm-level and employee-level variables at the time of signing the contract.
In answering (b), I treat my data as panel, because I wanted to see how Tenure (# years the employee has stayed in the firm) and Wage affect firm profits.
Now, the reviewer is asking if I have a mediation model, where Contract Duration and Wage mediate the antecedents to and outcome of signing the contract.
My question is, how do I deal with the mediation in my problem?
- Since Contract Duration and Wage are determined in the year of signing the contract, is it appropriate to treat it as a panel? I get that I can take the benefits of using a panel to control for unobserved effects. But, the control variables (lagged) when Tenure >1 do not precede Contract Duration which is determined at Tenure=1.
- If I do a panel analysis, how do I study the mediation effect? On the one hand, Contract Duration is time-invariant, and it is difficult to analyze it in a fixed-effect model (my data should not be a random-effect model). On the other, in studying a mediator effect, should I use Contract Duration or Tenure as my mediator?
