I think purely from a fundamental attribution error point of view we should expect the average "stupid" person we encounter to be less stupid than they seem.

(which is not to say stupidity doesn't exist of course, just that we might tend to overestimate its prevalence)

I guess the other question would be, are there any biases that might lead us to underestimate someone's stupidity? Illusion of transparency, perhaps, or the halo effect? I still think we're on net biased against thinking other people are as smart as us.