Quant- Conditional Heteroskedasticity vs Serial Correlation

A little help here: struggling to differentiate between conditional heteroskedasticity and serial correlation. It appears to me that in both violations of regression assumptions, the regression errors are correlated with the independent variables. By the variance of the regression errors being conditioned on the independent variable under conditional heteroskedasticity, doesn’t that also effectively mean those errors are correlated with their independent variable?

I think I’ve got it. In serial correlation, it’s the error terms across different observations that are correlated with each other (therefore an error in y1 affects the error in y2). While in conditional heteroskedasticity, the regression error terms are correlated with the independent variable.

Yup, there is an acceptable level of serial correlation (see durbiin-watson statistic). On the other hand, conditional heteroskedasticity of errors are unacceptable.

It is not generally true that serial correlation means the errors are related to the independent variables. Serial correlation means values within a series are related to one another. Heteroscedasticity does not mean the unequal variance is related to the independent variable unless it is said to be “conditional” heteroscedasticity.

In practice, these are on a spectrum, so serial correlation and conditional heteroscedasticity may be present but to a degree that doesn’t materially impact the project.