Why you should not use mean imputation for missing data

I encountered the question today of what to do with missing values when conducting null hypothesis testing or regression? I have seen many suggest doing mean imputation. That is, simply replace any missing values with the mean of the variable cal...