I think this is a good rule of thumb in general. When statistics agree with my preconceived notions, I consider them trustworthy, and if not, I assume that reality lines up with what I expect. For example, the referendum in held in the Baltics about leaving the USSR ended in favor of leaving, which I think is a good example of a trustworthy statistic. But the subsequent referendum in the remaining members ended in favor of staying in the USSR, and I think that's a little suspicious, don't you?