I'm sure you remember this experiment, which changed the required number of close/reopen votes from 5 to 3. Before getting to the meat of the analysis, I should point out that the change worked almost exactly as intended. Close votes clustered around 3 and the median time to close dropped noticably from November 9 to December 9, 2015:
Here are the statistics I said I'd be looking at:
Month posted Qs Closed Close Rate Closed->Edited Reopened Cl->Ed->Re Deletion ------------ ---- ------ ---------- -------------- -------- ---------- --------- 2014-11-09 1379 572 41.5% 12.1% 2.1% 10.1% 29.8% 2014-12-09 1251 730 58.4% 13.4% 1.5% 7.1% 32.1% 2015-01-09 1490 788 52.9% 11.3% 1.6% 10.1% 32.6% 2015-02-09 1507 816 54.1% 11.0% 1.6% 10.0% 31.7% 2015-03-09 1493 883 59.1% 9.7% 1.1% 8.1% 35.9% 2015-04-09 1391 814 58.5% 12.5% 1.1% 4.9% 34.2% 2015-05-09 1394 726 52.1% 10.2% 2.1% 9.5% 41.9% 2015-06-09 1472 764 51.9% 9.2% 2.1% 14.3% 37.1% 2015-07-09 1496 839 56.1% 10.8% 1.2% 5.5% 36.7% 2015-08-09 1392 785 56.4% 11.6% 0.9% 3.3% 52.3% 2015-09-09 1376 816 59.3% 11.4% 3.3% 18.3% 47.8% 2015-10-09 1463 869 59.4% 8.9% 2.2% 6.5% 43.5% 2015-11-09 1308 955 73.0% 15.4% 4.9% 17.0% 47.0%
The rates remained steady for the most part in the months leading up to the experiment. For the month of the experiment closing, editing of closed questions, and reopening spiked noticeably. (Deletion has a longer lag time so we don't have the final tally.)
The big number that sticks out is the 73% closure rate. We've already gone round and round about what that number should be, so there's no point in rehashing that argument. I think too many questions closed here are salvageable and other people (who have firsthand experience in the close queue) believe that the lines are clear and most closed questions would be bad for the site. If Jeff couldn't grok this site, I don't see why I would either.
The reason I don't understand the goals of this site is, in fact, because I'm not part of this community. Closing is part of how this community, as Clay Shirky puts it, "defends itself so that it can stay on its sophisticated goals and away from its basic instincts." Bikeshed questions represent a considerable portion of those instincts, I suspect. Closing questions that verge toward bad subjective protects the group from junk answers.
Shirky also points out in "A Group Is Its Own Worst Enemy":
You have to find some way to protect your own users from scale. This doesn't mean the scale of the whole system can't grow. But you can't try to make the system large by taking individual conversations and blowing them up like a balloon; human interaction, many to many interaction, doesn't blow up like a balloon. It either dissipates, or turns into broadcast, or collapses. So plan for dealing with scale in advance, because it's going to happen anyway.
Our solution to scaling historically has been to encourage users to trust fellow programmers. While it's not without problems, the philosophy does allow one community to handle 8k questions a day. It's been ages since anyone could read all the questions on the site, but the system works well enough to produce an enormous (and growing) store of knowledge about coding. Most of the questions are crap, but that's more than balanced by the quality of the content that rises to the top of searches.
It's clear that changing the number of votes required to close questions will allow this site to scale without changing the criteria for closing questions. I completely understand why this would be desirable. But I think it would be a short-term hack. It will not fix the root of the problem.
You might as well know what I think the core problem is: distrust. Over and over since I started this experiment, I've heard (whether this is what was intended or not, I don't know) concern that outsiders will take advantage of this community the moment its guard is dropped. And it's not just outsiders: there is a deep division among high-reputation users and former users of the site that is marked by distrust. I have some suggestions about how to work toward a compromise, but I haven't earned any trust here myself.
But in the off chance that it could still be useful, I'm going to toss out my ideas anyway:
Assume that only the sign and not the magnitude of question scores matter. Voting volume tends to follow the number of views. Views are driven largely by how interested the general public is in the question and it's answers. The second most-upvoted question on SO is: How do you undo the last commit? It didn't start off as a particularly good question, but it represents an extremely common problem and so it gets a lot of views. I suppose in a just world, those views would not turn into votes. But in the real world, not everything is tied to merit.
Assume that only the first few answers are ever read. Almost a corollary to the previous point: popular questions will get more answers. If you dig through those answers, it's pretty likely some of them are low-quality. Gallingly, they are likely to have more votes than brilliant answers to obscure questions. But within a question, voting does a good job of surfacing quality. Studies consistently find that people rarely finish reading pages they land on. Either they find the information they are looking for and leave or they give up on the page. Downvote bad answers and trust your fellows to do the same.
There will always be people who misunderstand the nature of this site. Several people pointed me to a comment (on a now-deleted question) that cited this question as a reason for asking something similar. The example has been closed for over a year. Closing questions does not make them less tempting to copy. Don't confuse post hoc justification for motivation. People will ask dumb questions whether there are similar questions around or not.
Be proud of your past. I know this site has gone through a lot of pain. It's also produced some amazing answers. I've seen the accepted answer to Why do game developers prefer Windows? cited several times by people who aren't active on Stack Exchange. The answers to Is there any reason to use C++ instead of C, Perl, Python, etc.? make me wonder if I gave up on C++ templates too soon. While poking around the codebase here at Stack Overflow, I found the answers to What is MVC, really? useful. These bikeshed questions manage to produce exceptional answers in spite of themselves. Or rather, there are some amazing contributors here who have turned leaden questions into answer gold.
The single best defense against worthless opinions is voter diversity. Studies have consistently found that "diverse minds do better, when their decisions are averaged, than expert minds." New users should not be treated as threats, but as potential contributors.
The increase in closure rate makes reducing the number of votes required to close questions impractical. Other effects (particularly reopen rate and median time to close) were encouraging, but did not counter-balance the cost in terms of new questions closed.