My understanding is that it's cheaper because it is mediated by largely by a single person in a conference room, rather than the heavier process of collecting a jury. And it's less variable because juries are variable. Neither really requires a specific bias to exist for it to be true.

It is definitely possible that arbitrators interact with the company repeatedly and so there is gravity towards deciding in their favor, but I have no idea on the size of that effect.

Have you proven why? To make my point again: we just feel like it won't work so we don't even try