After reading through all of the 51 answers posted here by the time I'm writing this, I had the feeling, that my answer wasn't here yet. Reading the answers brought me to the conclusion that there is a great consent on why to vote for an answer (i.e. promotion of good answers as well as giving credit to the users who provided them). However, there's quite an amount of different opinions on why to vote for questions or whether to vote for them at all.
Before adding my own two cents, I'd like to sum up what I've found to be the essence of most answers (apart from nuances):
People don't vote for questions, because they...
- ...find it more natural to vote for answers than for questions.
- ...see no benefit in voting on questions. Several answers to the same question are comparable amongst each other and voting for answers brings the good ones to the top. However, voting for questions is considered comparing apples and oranges - it just makes no sense.
- ...believe attributing reputation to someone for not knowing
something is inherently wrong or that at least the amount of
reputation gained through an upvoted question should be
substantially smaller than that gained through an upvoted answer.
Regarding 1., there didn't seem to be much reasoning, so I'm not inclined to reason myself in order to refute these opinions.
Regarding 2.: If people vote questions up that they believe are worth asking and that are well written (SSCCE, good English, good formatting), then a questions score would be a measure of quality. This would help to bring high quality questions to the top of the search results and hide less well crafted questions. If that's not a benefit, then what?
Regarding 3., I do understand and to some extent even share this opinion. But I find it important to distinguish reputation and vote count as two completely different concepts that aim to achieve different things.
As already stated, vote count (whether on questions or on answers) is supposed to reflect quality. Reputation on the other hand, is a measure for the merit of an individual. For the individuals the reputation has value, because it is something they can brag about or that they may even be able to use when applying for a new job. Because it has value, it is a good instrument to stimulate participation.
However, the possibility of gaining reputation from asking questions somehow lowers the value of that reputation. Where's the merit in asking a question? Imagine you're looking at an answer posted by a 60k rep user and you think "Whoa", but then you look at his profile and you see that 90% of his rep came from questions. For me that makes a difference. Conversely I'm not too impressed by people who pretend to know everything by never asking any questions. It just shows off their level of narcissism.
Another problem I see with the tight coupling between votes and reputation is that questions that address more common problems will create a higher traffic than those with a very narrow domain. Questions with high traffic and their answers are more likely to get upvoted than those with low traffic. This is not a bad thing by itself, but because upvoting creates reputation I see lots of people accumulating insane amounts of reputation for explaining very basic principles of programming while others who provide sophisticated answers to quite complicated problems receive very little reputation for doing so. I'm not trying to discredit anybody or saying that reputation scores are meaningless, but because of how the system works I believe that they are distorted and not easily comparable.
Coming to an end now: It's too late for that, but if I could make a substantial change to the way SO works, I'd completely decouple votes and reputation. Instead I'd attribute some sort of credits to people depending on their level of participation on this site. These credits wouldn't show up anywhere on their public profile, but they could spend them on other users to increase their reputation.
This is only an idea, which would probably require some more thinking and field testing to actually make it work, but I believe that by decoupling votes and reputation we would see two effects:
It would become clearer to users what they achieve by voting for questions (better search results) and they would be more inclined to do so, because there's no reputation change involved for the person asking. Thus we would see more questions being up- or downvoted.
It would increase the comparability of reputation scores and the scores would better reflect who's actually reputable in the community.
I know that this would be a fundamental change to SO and its siblings and that there's no chance for it to ever become reality, but perhaps someone will read this before launching the next Q&A site.