Clustered errors and why naive standard errors mislead
Asked at Squarepoint
You pool many observations that arrive in clusters: multiple days per stock, or multiple stocks per sector, where errors within a cluster are correlated (a sector-wide shock hits every name in it).
Does clustering bias your OLS coefficients? If not, what does it break, and how do you fix it?
Show a hint
Independent observations each add fresh information. If observations inside a cluster move together, how much genuinely independent information do you really have?
Your answer
This one is open-ended. Work it through, then check your reasoning against the full solution.