Personally I up vote questions I think are useful but not necessarily important to me (I only up vote answers I understand and agree with) and I ALWAYS up vote a question I answer.
But it obvious that people are very miserly with their question up-votes. I think this is because of the reputation rewarded for an up-vote.
I propose a reduction in reputation for question up-votes to 3.
This will encourage people to vote more and thus the amount of reputation earned would be similar.
I think people are pretty free with the up votes. My own analyis has show an up to down vote ratio starting at 10 last fall and growing to about 12 now.
Like Brent Ozar, I also investigated the relationship between votes and other user data (like reputation and up:down vote ratios, etc.) in a couple of blog posts, like these:
Here are some of my favorite graphs:
Down (x-axis) vs. Up (y-axis) vs. Reputation (bubble size)
Up Votes as a percentage by User Join Date and Average User Reputation by Join Date
Larger images and other stats on the blog.
This was one of the questions I explored at SPWho2.com:
What kinds of questioners do the most upvoting?
Or in other words, what tags are the friendliest? Are people who ask questions with a particular tag more likely to do upvotes? These are NOT the upvotes performed on their question - these are the upvotes done by the questioner themselves.
As it turns out, yes, there's a huge variation. Some examples:
You can see the full report on SPWho2.com, along with a report on average questioner downvotes. I find it hilarious that people who ask vim and emacs questions are much more likely to perform downvotes than other users - a little grumpy, eh?
Next month I'll add one for average answerer upvotes too. For example, are people who answer "PHP" questions more likely to upvote others? What are the friendliest tags, and what are the least friendly?
I've pondered this, but then you end up with questions about how people use scale. For instance I have an acquaintance who has said that on a scale of 1 to 10 he never ever rates anything 10 because there is always room for improvement. So even on a small scale is anything less than 3 reflect poor quality? Would people start up voting with 1s poor-mediocre answers?
The current system is clear an unambiguous. -1 if it was poor. +1 if it was good. 0 if it was unremarkable. Simple is better.
Generally speaking, the number of upvotes an answer can expect is a function of its potential audience. Fun/CW questions have the widest audiences and receive the most votes. The most popular tags (like C#) are next. Most questions and probably most upvotes per answer of the "real" questions. And it trails away after that. JPA questions, for example, aren't often easy but will rarely garner more than 1-2 votes.
