Revisiting free downvotes on questions

Question

Disclaimer: I've read Should downvotes on questions be "free"?, but I feel the landscape has changed in the last three years. What I'm looking for here is discussion about the results, looking back. I've included the best stats I could find, but if you have better numbers, please share.

Downvotes on questions are "free", while for answers it "costs" -1 rep. Making the question votes free was done, according to the linked post above, to encourage voting on questions, and as a way to filter the good content from the bad. These are fine ideas, but a few years later, I'm not sure it makes so much sense.

First, some numbers. I'm keeping this focused on Stack Overflow because that's where I'm most active, and I suspect smaller sites may not have the same issues.

Questions 

Total           6.673M
Upvotes        11.735M
Downvotes       1.039M
U/D Ratio      11.291
DV/Question     0.156

Answers 

Total          11.833M
Upvotes        25.929M
Downvotes       0.784M
U/D Ratio      33.081
DV/Answer       0.0663

We can all see that answers are still voted on more than questions, but this (IMO) is natural. What we can also see, though, is that questions are much more heavily downvoted than answers. If you post a question to SO, you have a 1/6 chance of being downvoted, while answers have a rate of 1/15.

Are the questions really that bad? Many would say they are, and this is natural. However, these numbers don't take into account deleted questions. Many "bad" questions are voted down, put on hold, then closed, then deleted. The deletion script in particular specifically looks for downvoted/closed questions to remove them. I feel that if we got the numbers including deletions, it would skew it even more.

Next, look at the Up/Down ratios. Back before question downvotes were free, the ratios were surprisingly even. Both questions and answers had a ratio of about 20:1. Now, questions are at 11:1 and answers are at 33:1. I believe this is direct result of one being free while the other costs rep, and I don't believe this is a change for the better.

In addition to everything so far, we have better tools to get rid of bad questions, and are lacking on answers. You can't put an answer on hold, you can only flag, downvote, delete. Many answers don't fall afoul of the flags, but are still bad. The only recourse here is to downvote, yet it's the only thing on the entire site that costs rep to do. Okay, bounties cost rep, too, but it's the only moderation-like tool that costs rep.

Another relevant topic is "piling on" downvotes, especially for new askers. Often a user wanders in, asks a not-quite-good question, and within ten minutes is staring at a -5 or lower score. Why? Because downvotes on questions are free. Yes, you see answers get downvoted quickly, but almost never as fast or as far. Why? Because they cost rep. Many users feel we should be more "friendly" to new users. I'm not sure where I sit on that, but I feel that in general, the "right" option is to put a question on hold, not downvote it into oblivion. The current system seems to encourage the latter.

Overall, I guess my question boils down to: Does it still make sense for question downvotes to be free?

Update

After restricting this to only the past year, the current ratios for upvotes to downvotes are 3.6:1 for questions and 27:1 for answers. This means the disparity between question/answer voting is wider than I initially thought.

Answer 1

As I mentioned in comments, your data is wrong - and not just because deleted questions aren't included. But I'll get to that in a minute. First, you need to understand the problem this was trying to solve...

This site is full of crap, but there's no way to identify it

This was a huge, huge source of frustration back when free downvotes for questions were rolled out. People complained here on meta, viciously attacked new users on the main site, and flagged so many posts for moderator attention... But they didn't downvote them!

The team was working hard on automated systems to guide or outright block folks asking poor questions, but without any way to identify these "poor quality" questions these efforts were doomed. Something had to be done...

Along with free downvotes, there were several other changes made to assist with this identification:

• Certain close reasons imposed an automatic downvote from the system itself (this was removed fairly recently - see "People just don't downvote enough" below).
• Very Low Quality flags also imposed an automatic downvote.
• The number of votes available per day was increased for folks who voted on a significant number of questions.

These all helped, but I cannot overstate how big a problem this is. And here's problem #1 with the data you collected: you're ignoring questions that don't get any votes at all, which is the real problem these changes were intended to help mitigate. So here:

• 2010-2011: 54% of questions didn't get any votes within 30 days of being asked.
• 2013-2014: 62% of questions didn't get any votes within 30 days of being asked.

Yeah. The problem is actually getting worse. To a certain extent, that's unavoidable: the site is a lot bigger, the tail is a lot longer, and the chances of someone even seeing a given question aren't as good. But that's all the more reason to try and make sure we do as much as possible to encourage folks to vote when they do come across one.

People just don't downvote enough

The number you did get right is the ratio of up-votes to down-votes. It's pretty high, for both questions and answers. It's higher for answers, for various reasons (good and bad) that I won't go into here - but in both cases, folks overwhelmingly prefer up-voting over down-voting even when downvotes are free.

And that's ok... Up to a point.

In addition to everything so far, we have better tools to get rid of bad questions, and are lacking on answers.

We get rid of a lot of answers. But getting rid of things isn't really the point of down-voting - in many cases heavily downvoted posts should be deleted, but this isn't a given. More importantly, voting generates signal - for future readers, for the post's author, and for the system itself - as to what is good and helpful and what isn't.

Those tools for "getting rid" of questions? They're pretty heavy-duty; they require a lot of work from everyone involved. This is why I get so incredibly frustrated when folks talk about closing mediocre questions instead of down-voting them: imagine if your down-vote was held in limbo until 4 other people agreed with it - that's what closing amounts to. It's a way of getting questions that can't or shouldn't be answered off the site, not a "super-downvote" - indeed, a single close vote is very weak.

Which brings me to the data you requested: down-voting on deleted questions. No surprises there: it's higher than down-voting on other questions. Remember, closing is the precursor to deletion, and down-voting on closed questions finally reached the point where we were comfortable disabling those automatic downvotes I mentioned above, something I credit in large part to making down-votes free on questions.

So yes, we do need those. Because they're working. Horray!

Answer 2

I don't think "the numbers" can accurately answer this question, because the cost of down-voting a question has not been the only thing to change on this site in the years it has been in existence.

1. The sheer number of users, and hence the number of questions in general, has skyrocketed.

2. The number of low-quality questions - votes aside, just a subjective observation on my part - has also skyrocketed. I don't know whether the ratio of bad questions : good questions has grown symmetrically, but in the tags I frequent, it certainly seems to have out-paced the growth of new users.

3. The rules of the site itself have changed quite a bit - open-ended and shopping list questions, for example, used to be okay, but now they get down-voted into oblivion (though arguably they should just be voted to close).

4. The number of truly new questions has gone down considerably; the odds of asking a question that has already been asked has hockey-sticked upwards. Yet we still see a ton of questions that are asked without doing prior research, without paying attention to the duplicate suggestions or the related questions to the right, etc. These also tend to get heavily down-voted. And, arguably, rightfully so.

5. Users who have been committed for a long time to the site have learned what types of answers are likely to get up-voted and down-voted. I suspect in certain tags a high percentage of answers are posted by a small percentage of users who know now not to post link-only answers, but may not have known as much a few years ago. (I look back on some of my own answers from years ago, and wonder how they ever earned any up-votes at all.) So this could also explain to some degree why a small percentage of answers get down-voted today in comparison to a few years ago.

6. Flagging has also changed over time, and I think very few answers get down-voted when a flag would suffice. I'm not saying this is a good reason to flag an answer, but it is probably more common than even I want to admit. As a moderator on a low-volume site, I'm appalled at how many people use flags to alert moderators about answers that should be down-voted instead. Can't even imagine what it is like on SO. This definitely does not support any argument that down-voting questions should cost more, but rather that down-voting answers should cost less. (Or maybe even that flagging answers with certain flags should cost rep.)

A lot of these things have changed, and could in and of themselves explain an increase in down-vote ratios even without taking into consideration the change in how much a down-vote costs.

One thing that hasn't changed all that much is the ability to judge a question on its merits alone, without looking at the score or any flags / votes to close. Unfortunately, this is a very tedious and difficult thing to capture with numbers, unless you want to evaluate every single new question as it comes in and determine whether it deserves to be down-voted or not.

Without that information, I think it is impossible to know whether people down-vote answers far too freely now, or whether any change to the cost of down-votes will really have an impact - for the better - on users who post questions or answers that may or may not have been up-voted or down-voted before the change. There is almost certainly some truth to the theory that more down-votes occur on questions than would happen if they cost rep. But I think it is not feasible to determine how much impact that has, and whether making down-votes on questions cost would just lead to a far lower barrier to crap questions. I don't see many cases where questions are down-voted for no good reason, and I don't recall many questions where - if down-voting a question were to cost me 1 or 2 rep - I would not have still clicked the down arrow.

Is it really a bad thing that more questions are down-voted than answers? Maybe that really is a reflection of reality: that the ratio of bad:good questions is that bad. As I alluded to in a comment above, a lot of even the crappiest answers are still useful/helpful to someone. That is more than I can say for the majority of questions below my down-vote threshold.