Most real world data will probably be heteroskedastic. However, one can still use ordinary least squares without correcting for heteroskedasticity because if the sample size is large enough, the variance of the least squares estimator may still be sufficiently small to obtain precise estimates. (Methods for Detecting and Resolving...)

Heteroskedasticity can also occur if there are subpopulation differences or other interaction effects (e.g. the effect of income on expenditures differs for whites and blacks). (Again, the problem arises from violation of the assumption that no such differences exist or have already been incorporated into the model.) (Richard Williams, Univ. of Notre Dame)

If heteroskedasticity does not cause bias or inconsistency in he OLS estimators, why did we introduce it as one of the Gauss-Markov assumptions? (Hedibert)