Comparison between autocorrelation and meteroscedasticity:

Autocorrelation occurs when  the assumption of classical linear  regression that errors corresponding to different observations are uncorrelated breaks down. It occurs in both time-series as well as cross-section analysis. Heteroscedasticity,  on  the other hand, occurs when the assumption of constant error  variance, or homoscedasticity, does not satisfy. The existence of heteroscedasticity often occurs in cross-section data.

In both autocorrelation  and heteroscdasticity,  the least-squares estimators  are  linear and unbiased but  inefficient. While autocorrelation tends to make the variance of the error  term relatively large, heteroscedasticity makes  estimated variances of least-
squares biased. The usual  tests  of statistical significance  such as  t and F  are, therefore, no longer valid in both the cases.