HAC Standard Errors

HAC standard errors (heteroskedasticity and autocorrelation consistent) — most commonly Newey-West — are the time-series analogue of robust standard errors. They correct OLS standard errors for both Heteroskedasticity and Serial Correlation in the residuals, using a kernel that down-weights more distant lags. HAC SEs are almost always larger than naive OLS SEs, so t-stats fall and p-values rise to honest levels.

When to use

Use HAC SEs as the default for any time-series regression. R: coeftest(model, vcov = NeweyWest(model)) from the sandwich / lmtest packages. The bandwidth (number of lags) is chosen by Newey-West's automatic rule (NeweyWest()) or set manually for the autocorrelation horizon you expect. PS_3 illustrates the contrast: under OLS the HNC coefficient looks significant, but under HAC the t-stat drops and significance can disappear.