Speaking from a survey design point of view, he's more correct here. Especially with topics like this, the self selection effect and subsequent sampling bias can be very strong.

The evidence here is sufficiently poor to provide minimal value. Poorly collected data can often be worse than no data, as it can mislead you.