Clustering standard errors at year or firm-level.

[Jun Park]

Let's say I am running a regression where Y is stock price and X is natural disaster. Is it appropriate to cluster my standard errors both at the year and firm level?
1. Firm-level because stock prices are auto-correlated.
2. Year-level because all firms at certain years might have had a similar impact because of the natural disaster. But then, I'm not sure if this is related to year fixed effects?.. or standard errors?
Thank you!

[Carlo Lazzaro]

Jun:
you seem to have serial correlation of the epsilon within and across panels.
If you're dealing with a T>N panel dataset, you may want to take a look at -xtpcse-.