Fixed effects / clustered standard error in multi-level data

Hi everyone :)

I'm analyzing the effect of political corruption on the bid-ask spread of US political entities, namely my data set consists of states, counties and cities. The timeframe of my data is from 2005 to 2015.

For my regression analysis, I use fixed effects on years and and states (e.g. california) and cluster my standard error by government level (i.e. state, county or city) and years.

Is this a sensible approach? Am I missing something? And (not related to linear regression): if I'm t.testing the difference in mean bid-ask spreads, do I have to match the entries based on government level first?

Some expert feedback would be valuable, thanks!