Your idea is based on a false premise. The key is that the metric of reputation conflates, by design, answer-writing ability and technical knowledge. By extension, your statement that "we can expect higher-quality answers from experienced users" is problematic: what do you mean by higher-quality?
Answer-writing ability consists of:
- How well the user understands the site's rules and is thus able to write answers which do not violate any guidelines or the Terms of Service;
- How skilled the user is at commanding the language being spoken (in most cases, English) and writing effective prose;
- The user's ability to trade off length and thoroughness of their answer vs. conciseness.
Technical knowledge consists of:
- What subject areas the user has personal experience with;
- What research (Google searches, reading academic papers, etc.) the user has done in order to produce their answer;
- The rigor with which the user has checked any facts they claim to be true, vs. which of the user's statements have they qualified with opinion-making statements such as "I think that" (an opinion is less valuable than a fact, when a fact is readily available to be found and cited.)
As users become more experienced with the site, there is going to be a fairly positive trend between answer-writing ability and reputation. However, the correlation between reputation and technical knowledge is not there, and cannot ever be there. The primary reason is that reputation is a conflation of these two topics. If you get a downvote for your answer, what is the reason that the downvoter provided for that downvote? Is it technical-related, or is it answer-writing-related?
Consider that, compared to a well-written answer with technical inaccuracies, it is much more difficult to edit a poorly-written or off-topic answer to continue to be relevant to the site and worth viewing. If the answer is well-organized and explains itself well, it's not that hard to edit anything that is inaccurate. If the answer is inscrutable because it's riddled with terrible grammar, spelling and doesn't attempt to justify the statements it makes, it's more likely to be deleted after being downvoted to a negative score.
If your definition of "higher-quality" means answer-writing ability, then your proposition is well-intentioned, but it won't work out in the end. The problem is that people use the downvote mechanism for any reason at all; the motive behind the downvote is completely invisible to the person receiving the downvote. Were they just in a bad mood? Did they not like that the answer -- 99% of it basically perfect -- contains a single typo? Do they have a false idea in their head and therefore downvote it because they think it's wrong, even when it isn't?
If your definition of "higher-quality" means "more technically knowledgable", I'm afraid you're barking up the wrong tree. Reputation does not, and cannot, measure technical knowledge. It never will. It's far too easy to game the system (bounty hunting, for instance) to gain reputation far disproportionate with the user's technical knowledge. I believe it is impossible to design a pure numeric value that accurately and fairly describes a user's technical knowledge relative to their peers. To even attempt that, you would have to design a suite of the world's fairest academic tests that have ever existed, on every conceivable subject, and then force every user of StackExchange to take them all and be graded on each.
Lastly, the biggest problem I have with your proposal is that it opens up a huge avenue for abuse, unless you also take a proportionally higher amount of reputation from the people casting the downvotes. We as a community have no way to police bad behavior with respect to downvotes; we can't review downvotes like we can review answers or questions. We can't even see (by design) who downvoted anything, and we don't require anyone to provide reasons why they downvoted. The whole system is just there to encourage participation.
Remember that, if users are getting many more downvotes on the whole than they get upvotes, their question/answer scores will be negative. If you have too many questions/answers with a negative score, you will actually become question or answer-banned on the site, indefinitely, until you fix your old contributions. This auto-ban mechanism is typically the way we handle users who are making bad contributions.
I think you are putting too much emphasis on the reputation of users as a means of judging their worth or merit or knowledge. The purpose of voting on questions and answers is to show what the community collectively thinks are the most meritorious answers and questions. The reputation is secondary. Upvotes and downvotes are important due to the effect they have on the net score of the questions and answers, NOT due to the effect they have on the user's reputation.
To put it into real life terms: reputation can be thought of as how many years' experience you have on a job. Being experienced means you probably won't make any rookie mistakes, and probably know the rules, but it doesn't mean you're very good at your job -- just that you're not so terrible as to be fired. Trying to shoe-horn any additional meaning into reputation other than that is going to be impossible due to the many ways it can be gamed (both in the positive and negative directions), and your proposal only magnifies the problem with gaming it.
