Cross Validated
2023-03-23 21:50 UTC
By narrativera
AI-113-20230323-social-media-16c48da3
Levels of clustered standard errors and fixed effects
A while ago I read (probably on stata forums) that FE can be on a lower level than SE clusters, but not vice versa. For example, firm/household FE with industry/state SE are good, but doing state FE with household SE would be wrong. Intuitively it seems to make sense. If treatment assignment is on a household level, shouldn't household FE be the first choice? But it seems that there are papers in different fields that use higher level FE with lower level SE. What am I missing?
A while ago I read (probably on stata forums) that FE can be on a lower level than SE clusters, but not vice versa. For example, firm/household FE with industry/state SE are good, but doing state FE with household SE would be wrong. Intuitively it seems to make sense. If treatment assignment is on a household level, shouldn't household FE be the first choice? But it seems that there are papers in different fields that use higher level FE with lower level SE. What am I missing?
Full article content could not be extracted automatically. Read the original below.
Source:
Cross Validated
· stats.stackexchange.com