Yeesh, this study is...not great. The survey design is odd at best—respondents were placed in different segments based on survey questions, undermining any ability to discern correlation and causation. The study sets up as a pretty obvious case of potential type 2 error as well.