Is there any evidence that the size of a bounty affects the quality or quantity of responses? Anecdotally it seems that it does not, and that the existence of a bounty is sufficient. However, is there any data behind such assertions? I can buy those arguments, but I can also buy the argument that large bounties attract a better audience. Has this been studied at all?
2 Answers
I've tried to approximate this with this query:
SELECT BountyAmount, AVG(Score) AS AverageScore, COUNT(*) AS NumPosts
FROM Posts
INNER JOIN Votes
ON Posts.Id = Votes.PostId
WHERE BountyAmount > 0
GROUP BY BountyAmount
HAVING COUNT(*) >= 20
ORDER BY BountyAmount
It doesn't account for whether or not the post was made during or before the bounty period, though.
The answer appears to be yes.
-
+1 Nice! If this question had a (large) bounty. I'm sure you would get it ;)– LixFeb 6, 2012 at 21:55
-
2"It doesn't account for whether or not the post was made during or before the bounty period, though." I greatly appreciate the effort (+1), but I feel this could be problematic. A designed use-case of bounties is to reward good answers (which presumably have high vote tallies already), so there's already a built-in correlation if not properly accounting for existing votes. Feb 6, 2012 at 22:00
-
1+1. Interesting, but a 500 bounty may work like advertising to the voters, in addition to the effect on the actual answer texts. It would be interesting to get the Average Score per page view on the Y axis and the Bounty score on the X axis. Feb 6, 2012 at 22:04
I wasn't really happy with the query in the other answer; it seems to include the score of the question as well as the answers. So I made a new one, which looks at the following figures:
- the total (current) score of all answers posting during the bounty
- the maximum (current) score of all answers posting during the bounty
- the average (current) score of all answers posting during the bounty
- the number of answers posting during the bounty
and takes the average of those. On Stack Overflow, there's definitely an increasing trend (i.e. higher bounties attract better answers). 350 and 450 point bounties are relatively rare and figures are skewed; it's not logical that a 450 point bounty should perform better than a 500 one. This makes sense, given that there are (currently) 478 active bounties and users are likely to sort by amount.
On smaller sites, like Meta Stack Exchange, where the bountied questions always fit on one page, the difference isn't really noticeable:
(again, results may be skewed because of low occurrence of certain bounty amounts; if you're curious about that 550 figure, check this)
-
It's also worth noting that the order in which the list of bountied questions was sorted was different at the time. Back then, bounties expiring within 24 hours would be listed first in order of closer expiration time, and then remaining bounties would be listed in descending order of reputation value without regard to expiration time; today, they're simply listed in expiration order without regard to reputation value. Jun 27, 2020 at 19:29