# The Efficacy of Mathematics Stack Exchange

I have two questions.

First, how "effective" is asking a question on the math stack exchange? I find 3252 pages of questions in the "unanswered" section - these questions have received zero responses. This is almost 20 percent of the 17465 total pages asked, meaning 1 in 5 queries never get a single answer.

Second, is it known what percent of questions on the exchange are asked by "established" users, and what percent are asked by random anonymous users (like me!) who basically sign on for one day to get unstuck on a homework problem and then never come back?

I don't ask these questions because I'm critical of the site, but I wonder if there would be some way to make it more effective (assuming its goal is to be conducive of the best discussion in mathematics).

• One correction. Many questions in the "unanswered" bin do indeed have answers, but the questioner has failed to accept an answer. I very frequently have to remind the questioners please to do so once they're satisfied. Apr 7 '14 at 1:30
• @TedShifrin Thank you, I did not notice that. It is true that a sizable proportion have some answers. From skimming a few of the pages, I get the sense that only a small amount (maybe 10%?) of those forgot to accept an answer - as in, the "answers" provided were not really answers, but general "advice", links to other pages, partial solutions, etc. But maybe I should revise my statistics... Apr 7 '14 at 1:32
• @TedShifrin I was under the impression that questions don't appear in the unanswered bin unless they have no upvoted answers, i.e. the answers that do exist aren't good enough. Apr 7 '14 at 1:33
• @MarioCarneiro: Yes, you're right, but my comment stands. Plenty of users don't know the protocol to upvote or accept. I've had some questioners accept with no upvote. I've had some who were satisfied and did neither. Apr 7 '14 at 1:38
• The irony of this is that it's not getting an answer... Apr 7 '14 at 2:17
• @MarioCarneiro I second Ted's objection to the inference "i.e., ... aren't good enough". There are thousands of answers that resolve the question completely, but remain with zero score. Voting system isn't a perfect judge of answer quality. Apr 7 '14 at 3:14
• @TedShifrin But unlike the protocol for accepting a question, anyone at all (including you) can upvote a question, so if you see an "unanswered" question that in fact has a good answer that has merely not been upvoted, you can just upvote it yourself and get it off the queue; there is no reason why the non-participation of other users should stop these questions from getting off the queue. Apr 7 '14 at 3:15
• @MarioCarneiro Yes, and there was (is?) a somewhat concerted effort to get those question out of the Unanswered list: Crusade of Answers. It's not as easy as it sounds... the site is huge, and the number of people involved in the Crusade is tiny. Also, it's very hard or impossible to judge quality of a technical answer in a field out of one's area of expertise. So, answers in more narrow areas, where few experts are present, are mostly out of luck. Apr 7 '14 at 3:25
• @cheap Actually, in my opinion the worst offenders for incorrectly categorized "unanswered" questions are those questions where the problem is completely resolved in the comments. In this case your only recourse is to post your own answer (and be accused of stealing the "glory") or prompt the comment answerer to post a real answer. Apr 7 '14 at 3:25
• @MarioCarneiro I haven't seen a single accusation of stealing in such cases, directed either at me or at others. I did post a few such answers; I even shared my workflow for this. Apr 7 '14 at 3:27
• @cheap Also, I don't see any problem with simply dealing with these cases as they arise in the natural course of things, rather than a concerted effort to clean it up, unless you are thinking like the OP and want to take aggregate statistics on the total number of unanswered questions. Apr 7 '14 at 3:29
• @cheap I guess I didn't mean the "stealing" literally; but I gather that this behaviour is generally discouraged in preference to the person who actually answered the question posting an answer, so that they can receive their justly earned rep. Apr 7 '14 at 3:31
• @cheap I'm not very active on meta, so I don't know where to refer you to; but it seems consistent with the general philosophy of the rep system. Do you not agree that it would be better if the person who answered the question posted that answer so that they could be rewarded for their hard work? Apr 7 '14 at 3:44
• @MarioCarneiro Your second sentence makes it sound as if someone prevents those users from posting an answer. If they chose to post a comment but not an answer, it's their choice and I respect that. My choice is to post an answer, possibly using their content (with attribution), which is exactly in the SE spirit. I haven't seen any sign of discouragement when I did that. Apr 7 '14 at 3:47
• @MarioCarneiro Questions answered in comments have been discussed on meta several times, posting a CW-answer summarizing the comments is the recommended approach, if the commenter does not post the answer: meta.math.stackexchange.com/questions/1148/… Apr 7 '14 at 5:23

The efficacy is not accurately measured by the size of the Unanswered category. Yes, there are about 48K questions there, but only 34K actually have no answers. The rest have at least one answer, which in many cases fully deserves an upvote, but does not have one yet. And among the questions with no answers at all many received helpful comments; unfortunately there is no way to obtain a count of those where the comments helped.

Let's count the questions that truly sank without a sound: no answers and no comments. Exclude those that were somehow rejected (closed or downvoted). We are left with ... 9461 questions, out of the total of 262K questions currently on the site. Not that bad.

To your second question: I took the reputation of the askers of the most recent 10000 questions; focusing on recent questions allows to reduce the effect of postfactum reputation changes, and also gives a better idea of the current state of events. Out of $10000$ questions:

• $82$ were asked by users with reputation $10000$ or more
• $183$ were asked by users with reputation $5000$ or more
• $955$ were asked by users with reputation $1000$ or more (called "established users" by SE)

The median was $61.5$.

So, less than $10\%$ of askers were "established" by the SE standard. The reputation of the rest is presented in the histogram below:

Then I did the same for $10000$ most recent answers. Out of these:

• $700$ were posted by users with reputation $100000$ or more
• $2995$ were posted by users with reputation $10000$ or more
• $7140$ were posted by users with reputation $1000$ or more (called "established users" by SE)

The median was $3734$. (The author of a randomly chosen answer is more likely than not to be able to vote to close.) Established users, who ask less than $10\%$ of questions, answer more than $70\%$ of them.

This time, the histogram is based on users with rep under 100K:

• The main statistic I get out of this is that 24.5K - just over half - of all unanswered questions are in the "commented, no answers" category, a rather worrysome statistic for the issue of answers in the comments. Apr 7 '14 at 4:08
• It might be useful to break this group down by question vote total - most of the highly upvoted questions with no answers are simply good open problems. Apr 7 '14 at 4:11
• @MarioCarneiro There are comparatively few of those however. Apr 7 '14 at 6:28
• Also towards the second question, there are 57755 questions by users with more than 1000 rep, and 7038 by users with more than 10000 rep. Apr 7 '14 at 6:29
• Some statistics on the number of questions asked by high-rep users are given here. They are from May 2013. I thought it was worth mentioning if someone wants to compare how these numbers have changed. Apr 7 '14 at 7:16
• While browsing through your remaining 9461 questions, I noticed that a) very fresh questions are doomed to be counted there; and b) Questions like this are counted, which is merged with a question having an upvoted and accepted answer Apr 8 '14 at 17:07
• @HagenvonEitzen Yes, so the precision in "9461" is misleading; it should be "nine thousands or so". There is no easy way to filter out merged (locked) posts, as far as I can tell. Apr 8 '14 at 23:37