My very rough guess is that right now, around 70% of active steem power is redirecting rewards back into their own pockets through either self voting, voting selling to bid bots, spamming micro votes or some other means
This is because the current economic system pays stakers 4x as much to undermine by platform via the above means than to take part in honest curation. We've long passed the point where big stakers can keep each other in check with downvotes because we understand we understand that if any individual staker chose to use their voting power to downvote someone else, not only would it deprive themselves of the opportunity of claiming that reward for themselves, most of the rewards rescued will just be redistributed to others engaging in the exact same actions.
75% of author rewards might sound like a lot, but in reality stakers are claiming both the author and curation rewards of their own votes leaving pitifully little for actual content creators, irrespective of how good their content is. The paper value of the split doesn't matter if none of it goes in the right direction: that is gradually from the stakeholders to the talented content creators via inflation like it is intended. This isn't happening at the moment because our economic system isn't sufficiently incentivizing the stakers to vote honestly, thus the plead for EIP.
I actually want the maximum amount of rewards going from stakers to authors in an honest way at a steady rate. Because the part that goes back into the stakers pockets is redundant; making a inflationary coin just for everyone to reclaim is silly unless it serves a purpose: if it sufficient motivates stakers enough to actually find and reward good content.
Raising curation from 25% to 50% will mean authors will suffer a 33% on paper decrease in rewards. But 50% of an system where votes are mostly honest is vastly more than 75% of next to noting because the money just gets shoved back into stakers own pocket. So in practice, the idea is to get much more money into the hands of authors.
Economic system is not about the naive direct effects of a change, it's about the result at equilibrium. We can make author rewards 100% but if all the large voters just get that money back for themselves, authors see none of it and the system fails. The entire idea behind EIP is to make it more profitable for voting honestly (higher curation) and less profitable to do otherwise (threats of some free downvotes)
Of course if curation was 90%, most of us stakers would vote pretty honestly, but they'll be too little rewards motivating good authors to create content. So you see, it's an optimization problem of finding the right amount to mostly maximize author rewards in practice.
I understand that it is an optimization problem, but I don't understand why you are so fixated on 50/50.
Usually, when you are trying to optimize something you slowly turn it up until you reach a point where you see the change going into the wrong direction again. Then you start testing the range of the breaking point.
Lets say you find the breaking point is 45% curation. Then you try to find the best split in the 38-44% range.
I understand that testing would take months or years, but nobody seems to even consider testing this. Stinc could run a live copy of the steem chain an see how it would affect the distribution if you change the numbers.
But just picking a random number like 50/50 and hoping that it's the best split possible for reaching your proclaimed goals is delusional in my opinion. It might be better than it is now, but is likely not the best split possible.