To improve the value of votes on newer answers, votes could be weighted by the amount of time they've been there. I.e. older votes are worth less.
This edit in part consolidates (and perhaps clarifies) my comments on decay, and also proposes an alternative means of accounting for decay.
The problem with votes that come before all the answers are submitted is that people who vote at a particular time (i.e. anytime before all answers have been posted) have temporal parallax, so to speak. Here's an example:
Sets of people A, B respectively view the answer sets 1 & 2, which sets of answers are made in order (i.e. people in set A can see answer set 1, those in B see 1 & 2). There can be as many subsets of A & B as there are answers to a question. Obviously, the best answer in set 1 is not necessarily the best answer out of the union of set 1 & 2. Thus, persons in set A may vote for an answer in set 1 but actually think the best answer was in set 2 (had answer set 2 existed at the time they voted).
Thus A's votes are less informed than B's votes (which are less informed than a hypothetical C's votes that occur after there are more answers after set 2). B therefore has better informed votes (Assuming B votes on the merits - this is notwithstanding that B votes for things that already have votes just because those things already have votes). The value of the votes of people in set B are the only fully informed votes made as they are made with the benefit of all possible selections at that time (notwithstanding any subsequent answer sets).
A single vote may therefore be described as an indication of the perceived value of an answer by the set of users who reviewed the set of answers available at the time they voted. The count of votes on an answer is the perceived value of an answer by the set of users who reviewed the set of answers available at the time they voted. The set of counts of all votes on the answers to the question ranks the answers according to the perceived value of the answer by the set of users who reviewed the set of answers available at the time they voted. (I just find these definitions helpful for clarity and concision of thought).
I therefore posit that votes do not have the same value, and a unit weighting system of vote counting does not accurately reflect the actual value of the answers in many situations. The accuracy of the perceived value of the answers that the votes reflect diminishes where any answer subsequent to the vote would have been considered superior (in the mind of the voter) to the answer they voted upon. In other words, as more answers form, there is increasing probability that votes made prior to those answers (i.e. the perceived value) deviates further from the actual value of the answers.
This foregoing analysis presumes that the 'correct/best answer' actually never changes. In cases where the 'correct/best answer' does actually change (i.e. reality shifts and so the answers on stackoverflow/etc become incorrect), an answer may later appear that is correct and this new answer reflects the fact that the previous answers are invalid.
For example, given A & B and the respective set of answers they know about when they review the question 1 & 2. After, the world changes and the answers in set 1 are now incorrect. If there are subsequent answers (forming answer set 2, which the people in set B know about but A did not) which accurately reflect the new material world change, only people in set B would be able to vote up the correct answer (subject to people in A reviewing their votes).
In cases where A is a large set relative to set B (i.e. lots of people view answer set 1 but not 2), this is highly prejudicial to the Stackoverflow's presumed ambition to value and rank appropriately correct & incorrect answers. In fact, in certain cases this straight vote will actually prevent the Stackoverflow community from being able to highlight and rank at the top the best answers. Those early answers will simply be too far advantaged by existing votes and prevent new answers from percolating to the top.
There is an alternative mechanism for re-evaluating answers that I have thought of: Once a question reaches a certain age (or other appropriate condition), someone can re-ask the question (or engage some mechanism which asks the community whether it's appropriate to re-ask the question), which resets all the votes and thus allows new answers to get a chance to percolate to the top. See e.g. Expiration of answers for questions with novel solutions