Having different variances. A set, or a vector, of observations is heteroscedastic if the variance of the random error is different for different observations. Heteroscedasticity observed in cross-sectional data is typically related to the scale effect: often, larger cross-sectional units are subject to larger values of the random component. In time-series data it may take the form of serial correlation in the variance (autoregressive conditional heteroscedasticity). It may also be introduced by model misspecification. In the presence of heteroscedasticity the ordinary least squares estimators of the coefficients are consistent but inefficient; those of the standard errors are inconsistent, and hence the standard inference based on the estimated standard errors is invalid. Among the popular tests for heteroscedasticity are the Breusch–Pagan test, the Glejser test, and White’s test. Two approaches to estimation with heteroscedastic data are generalized least squares (both the coefficients and the standard errors are re-estimated) and heteroscedasticity-consistent standard errors (only the estimated standard errors are corrected).