In this post, user 
For the TLDR answer, take a look at my favorite webcomic:
First, votes are soft rate-limited -- you have voting power that regenerates at a constant rate, so if you vote too fast, each vote ends up being worth less. You can't get around the rate limit with multiple accounts, because you have to fund them all with SP.
Upvotes on popular articles give more rewards to the author than upvotes on less popular articles. So if you're colluding with the post author, you might think you can wait until your author/partner's post gets traction, then upvote the post, adding mad $$$ to the payout and getting a kickback from the author.
This "wait-and-upvote" strategy is actually dominated by a simpler "upvote-immediately" strategy [1]: Just upvote your partner/author's post from the beginning, your partnership will then end up with more curation rewards and a better chance of getting eyeballs because it has your combined voting power pushing it for a larger percentage of its lifetime. At this point it's basically equivalent to the author and you pooling your SP.
The annualized rate of return for posting at the maximum rate [2] and upvoting your own posts is currently about 4-6% and should decline somewhat over time [3]. Upvoting yourself for large payouts requires an enormous up-front purchase of SP, and the rate of return on the purchase price is modest while the non-liquid SP gives you huge exposure to Steem. Basically the only way to make staggering amounts of money as a poster is to convince other posters to upvote you.
Getting kickbacks for upvoting is relatively harmless; it's economically similar to loaning out your SP for a very modest APR. The rewards earned per voting SP provided are simply too small. And then there are the hazards of being downvoted or having your collusion called out...
The financial gains of selling votes are nonexistent or very, very small if you consider the capital deployed and consider the base level to be what you get by voting honestly. Add in the risks and added logistical steps, and there's simply no sane economic reason to sell your vote to milk the curation/posting rewards.
Please read my disclaimer. This post is my own personal opinion.
[1] After older voting mechanics were discovered to encourage sniping bots that would monitor the blockchain and upvote a post as soon as it was posted, we added a gradually decreasing curation reward penalty for upvotes in the first 30 minutes; the penalty is given to the author, so there is no reason for the author (or anyone in a partnership that shares the author's payout) not to vote for it immediately.
[2] We're thinking the next blockchain update should introduce a per-account soft limit of about 4 posts per day, much less than the voting soft limit, which means that self-voting with maximum effect will require non-trivial logistics involving multiple accounts -- a fair amount of effort regardless of whether you do it manually or build some automated tools to help you.
[3] Like many cryptocurrencies, Steem started out extremely "hot" with new STEEM being mined very quickly, and then "cools" at a pre-programmed rate. Unlike many cryptocurrencies, Steem has a nonzero long-term equilibrium annualized rate to ensure incentives are aligned for continued growth, adoption and retention. See the whitepaper for more.