On page 373 of the CFAI 2019 books, it says:“Because we assume that the error term is serially correlated, by definition the error term is correlated with the dependent variable.” Would the error term be correlated with the dependent variable even if no serial correlation was present? I thought the error term included everything that affects Y that is not captured in the Xs. I guess my confusion stems from the definition of serial correlation. Serial correlation is when error terms are correlated across observation, not correlated with the dependent variable, right? So why would a regression model having serial correlation exhibit a correlation between the error terms and the dependent variable? I would expect models with no serial correlation to also have the error terms be correlated with the dependent variable. I think I am missing something.