Solved – How many observations are enough to perform linear regression with fixed effects

I am new to econometrics. I am studying earnings management in banks during the financial subprime crisis.

I manage to collect data from 23 banks from 2005 to 2010. Few years of them are missing but I end up with 126 observations.

Is that enough to perform a regression analysis with fixed effects (I do not know if that last actually has an impact in the number of minimum required observations).

The question asks about variables (columns, fields), but it seems that it is really about observations (cases, records).

As I understand it, you have 23 banks, would have 6 yearly observations for each, but 12 observations are not available, leaving you with 126.

There is no constraint that bites hard except if you are trying something impossible, in which case the software will refuse to act. Other than that, it is a matter of what you can do well, and that's a matter of degree. In this case, your economic sense should come into play as well as your statistical knowledge. If you have 5 or 6 observations per bank, do you think that is enough information to estimate any of that bank's characteristics? If you also take into account changes within your time period, what then?

Similar Posts:

Rate this post

Leave a Comment