I teach physics at a community college and have been on a number of hiring committees for tenure-track jobs. Usually we interview about 6 people for such a position, and of those, usually something like 3 of them show a lack of knowledge of the subject in the interview. When I say a lack of knowledge, I mean that they can't do easy freshman stuff. We will ask them a question that is posed as a request for them to pretend we're a class of students and give a short explanation of something, so it's nominally a test of their ability to teach it, but in reality we find candidates who simply don't know it. These are people who have a PhD from an accredited school. (We hardly ever interview people who only have a master's.)
[EDIT] (To clarify, I'm talking about very basic knowledge, such as understanding what Newton's laws mean. We're not asking candidates to recall obscure trivia. Personally, I don't really care whether they know which is Newton's 1st law, 2nd, or 3rd. I'm talking about candidates who actually demonstrate elementary misunderstandings of Newton's laws. Since a lot of academia.SE users are in math, I think a good analogy would be if someone applied to teach math, and that person didn't know the chain rule -- which I have heard from colleagues in math departments at community colleges is also common. Continuing this analogy, the issue is not that they fail to remember the term "chain rule." The issue is that if they're asked to differentiate sin cos x, they can't do it. They do silly things like attempting to use the product rule, as if the sine was multiplied by the cosine. Or they throw up their hands and won't try, even if the committee tries to help them out.)
Are there good strategies for screening these people out at an earlier stage, without needlessly losing too many good candidates from our pool? Here's what we're already doing:
We require applicants to submit undergraduate transcripts, and we are less likely to interview people who have poor undergraduate grades (such as Ds and Fs in math and physics).
We prefer applicants who have graduate degrees from more prestigious schools.
We prefer a candidate who has taught a wide variety of courses to one who has only, e.g., taught mechanics.
Undergraduate grades do seem to correlate with what we see in the interview, but it's also possible that someone started out their undergraduate education with a weak high school background and then overcame that disadvantage. There is also the difficulty of comparing different countries' grading standards. We require graduate transcripts, but I find those hard to extract useful information from.
We are currently only asking them to give the names of references, but not to supply letters of reference along with their applications. Would it help to make them send letters?
You would think that someone with really weak competence would never get into a good graduate program, and therefore we could just not interview people who have degrees from low-prestige programs. However, our pool doesn't usually include a ton of people who have degrees from the best graduate programs, and we have also seen people in the past with degrees from renowned universities who nevertheless displayed major gaps in their knowledge, as well as highly competent people who got their degrees from no-name schools.