fixed effect - areg versus xtreg

[Matt Oldy]

Hi everyone,

I've been using the answers given in this forum for a while. Thank you to all the experts who contribute.

I have a question that I couldn't find an answer to and believe you can help me with:

I have panel data with the following information about private companies:

1. Company Name
2. Year
3. Quarter
4. Revenue
5. Number of Employees

I want to test the connection between the number of employees and the company's revenue.
At the same time, for some companies, I have 4 observations each year, and for some, I have only 1, 2, or 3 observations per year.
My hypothesis is that the year can have an exogenous effect, so I want to control for it; and - most importantly - the companies are from different sectors so I need to FE by the company.

All in all, I was thinking about two options for my analysis:

Option 1 - using arge:
gen companyCluster=companyName+string(year,"%02.0f") //to be able to cluster for multiple observations by some companies in a given year
areg revenue number_of_employees i.year, absorb(company_name) cluster(companyCluster)

Option 2 - using xtreg:
xtset company_name
xtreg annual_reported_revenue number_of_employees i.year, fe


Are these options valid? If option 2 is valid, do I need to add clustering?

Thank you all in advance, Matt

[Carlo Lazzaro]

Matt:
welcome to this forum.
You should get the same sanple estimates with both codes.
That said, I would go -xtreg,fe- with clustered standard errors.

[Matt Oldy]

Thank you Carlo!

[George Ford]

if you are uninterested in the i.year coefficients, then you could use reghdfe.