ABSTRACT
This chapter explains that heteroskedasticity literally means different spread, and occurs when the error terms from a regression do not all have the same variance. This is a direct violation of the Classical Linear Regression Model, which states that all the error terms have the same variance. Cross-sectional data are most likely to be heteroskedastic. Heteroskedasticity is not the larger errors associated with higher levels of income. It is the wider frequency distributions of the error terms associated with higher levels of income. The chapter also explains Newey-West technique formulas for the standard errors of the estimators that are less biased in the presence of heteroskedasticity. It describes weighted least-squares technique that remedies heteroskedasticity by dividing the data by some version of the culprit variable. Most econometric software packages apply weighted least-squares on command. Typically, this is the reciprocal of the explanatory variable giving rise to the heteroskedasticity or the reciprocal of the square root of the culprit variable.
