fixed effect in this case

[Kelly Ha]

I have a measure that varies within the industry. Specifically, it's a rank that varies from 0 to 9. The measure is quite sticky, meaning if a firm is assigned to the 8th bucket, it's likely that that firm will end up in 8th again in the following year. Can this be a reason NOT to use firm fixed effect? The reason being that I am utilizing variation within the industry, so having firm fixed effect would absorb the necessary information? Thank you for the help in advance!

[Carlo Lazzaro]

Kelly:
as you are surely aware of, -fe- estimators wipes out time-invariant predictors.
If -fe- expected beahviour is or not a good reason for switching to different estimators (provided they're justified), is difficult to say (for me, at least).