To estimate whether exposure to campaign advertising affects political knowledge, a researcher controls for respondents’ knowledge measured after the advertising campaign. Why could this be problematic?