This type of rating system is always disappointing. Candidates and interviewers each have different ideas of what expertise means -- I guess because there is no real context around the question. Maybe it's a trope to say, but it seems like competence and likelihood to rate oneself an expert are inversely correlated (barring hubris or real expertise!).
I used to ask candidates to rate themselves from 1-10, e.g., in database performance analysis, and often would get people rating themselves a 9 or a 10 without being able to articulate anything about the topic. It just seems very meaningless, since we were going to have a discussion anyway -- and the conversational part of the interview is more revealing, in any case.
I usually find that (within a broad range of capability) someone's actual knowledge on a given topic is inversely correlated with their self-rating on a 1-10 scale.
I’ve found some companies use the rating systems to tailor interviews to your strengths by providing descriptions at each level (e.g. “10 means you wrote the book about it”). It’s not always useless.
I personally never experienced it myself. But I can imagine it being useful in a relativistic sense.
For example if the candidate puts 9 for PHP and 5 for Haskell, you can get the rough idea what language they prefer or are more comfortable in using. You are not comparing these ratings against other candidate's ratings.
I used to ask candidates to rate themselves from 1-10, e.g., in database performance analysis, and often would get people rating themselves a 9 or a 10 without being able to articulate anything about the topic. It just seems very meaningless, since we were going to have a discussion anyway -- and the conversational part of the interview is more revealing, in any case.