Robust Standard Errors

Definition

Standard error estimates that remain valid under violations of classical assumptions such as homoscedasticity or independence. White (1980) heteroscedasticity-consistent estimators are the most common variant. For time-series data, Newey-West HAC (heteroscedasticity and autocorrelation consistent) standard errors account for both issues. Clustered standard errors handle within-group correlation in panel or grouped data.

Why It Matters

Classical standard errors assume constant variance and independent observations — conditions rarely met in real-world data. Using ordinary standard errors when heteroscedasticity or autocorrelation is present leads to incorrect p-values and confidence intervals, potentially producing false discoveries or missed effects. Robust standard errors safeguard inference against these violations.

Example

A researcher regresses wages on education and experience using survey data from multiple industries. Wage variance differs markedly across industries (heteroscedasticity). Using White robust standard errors instead of ordinary standard errors widens some confidence intervals and narrows others, yielding more reliable hypothesis tests.

Related Terms

• Heteroscedasticity
• Standard Error
• Ordinary Least Squares (OLS)

Software Notes

• SPSS: Analyze > Regression > Linear; click "Statistics" and check "HC3" or "Robust"
• R: vcovHC() from the sandwich package; e.g., coeftest(model, vcov = vcovHC(model, type = "HC1"))
• Stata: Add , robust or , vce(cluster clustvar) after regression commands