OLS Regression and Fixed Effects Model

[Derick Coleman]

Hi everyone,

I was just wondering if anyone could explain the difference between using an OLS regression and using a fixed effects model. I believe that the fixed effects model is used for Panel data, but you can also run an OLS regression on Panel data. Sorry if I am incorrect but it would be greatly appreciated if someone can clear this up.
The two codes I am using to compare models are:

/////OLS
reg lsp esgscore qr ltdp ebit roe

////////Fixed-effects model
xtreg lsp esgscore qr ltdp ebit roe

Where:
lsp = Log of the stock price
esgscore = esgscore
qr = quick ratio
ltdp = Long term debt percentage of capital
ebit = earnings before interest and taxes
roe = return on equity

Also what is the difference of using the random effects model and a fixed effects model for panel data?
Any help would be greatly appreciated!

[Carlo Lazzaro]

Derick:
1) while it's true that you can appy OLS to fixed effect panel data regression, there's no gain with using -regress- instead of -xtreg,fe-;
2) your first code can be applied to cross-sectional data. To make it suitable for -fe- panel data regression, you should tweak it as follows:

regress lsp esgscore qr ltdp ebit roe i.panelvar i.year, vce(cluster panelid)

3) there are tons of literature on going -fe- vs -re-. Just take a look at -xtreg- entry, Stata .pdf manual and skim through the references.