We congratulate the authors on their interesting study comparing non-probability Internet samples, low response rate telephone samples, and a high response rate area-probability sample to a benchmark survey. The study differs from others (e.g., Pew, 2016) in exclusively examining cross-classified socio-demographics. The conclusion is in line with previous studies: weights help overcoming errors in non-probability samples, but only to a certain extent. While the general conclusion certainly will be helpful to survey practitioners, there are important practical and theoretical issues that should be examined in light of this paper.