Why would a researcher use robust standard errors? Why would a researcher use clustered standard errors? If you estimate a model with either robust or clustered standard errors, will the coefficients be different than the version with homoscedastic standard errors? Explicitly address the OLS assumptions in your answer.