UPDATE Feb 15, 2018: The diagnosis of the problem in this post was correct, but the solution wasn't. So, I published another post to propose a solution to this problem: How to Solve the Reward Pool Abuse Problem Once and For All
The reason for the reward pool abuse problem is that 100% of the total stake isn't voting actively at least 10 times every day. If they did that, reward pool abuse wouldn't be profitable.
I will quote my previous post to explain this issue.
Let's make the following assumptions for simplicity's sake.
Each year 9.5% of the existing amount of Steems are created out of thin air and added to the Steem supply. That means each Steem token will lose approximately 8.68% of its value each year in fiat currency, e.g. USD.
That means everybody who invests in Steem starts with a disadvantage. There are several ways to offset that disadvantage. The first and obvious one is that more investors invest their fiat money in Steem and increase its price. Since that benefits everybody the same, we ignore its effect with our first assumption above.
The second way to offset the dilution is by earning rewards from those 9.5% extra Steems. The reward pool is distributed using the following formula:
- 75% to the reward pool
- 15% awarded to holders of Steem Power
- 10% witnesses to power block chain
For the following calculation, let's assume that everybody upvotes themselves to maximize their share of reward pool and no one else.
A regular user can only receive 90% of the extra Steems, which means they can only recuperate 90% of the dilution maximum. That means they will still lose 0.868% at the end of the year.
A witness seems to have 100% recuperation of the dilution on the surface, but don't forget that they have to pay the electricity bills.
This means, if someone wants to abuse the system, the best they can make is around 1% loss on their capital every year or break-even on their capital but electricity bills to be paid and equipment depreciation.
So, taking into account the assumptions above, Steemit reward pool abuse is not a profitable business in theory.
The real world is not ideal. Not everybody votes ten times every day. The reward share of the votes that aren't cast are distributed to the content and curation that are voted upon. That makes the reward pool abuse business profitable.
UPDATE, Feb. 15, 2018: After running the numbers, I came to the conclusion that this system won't work either. As long as there's reward money paid, there will be abuse. That's why I came up with another solution, which you can find in this post: How to Solve the Reward Pool Abuse Problem Once and For All.
In order to deal with this problem, we need to issue new Steem tokens proportional to the percentage of the total stake that is cast in a period, instead of a fixed 9.5% per year.
If only 60% of the total stake is used in a given year, then we create a new inflation rate of 0.6 * 9.5 %, which is 5.7%. That inflation rate would ensure that the reward pool abuse business doesn't make any profits, similar to the ideal scenario above.
I want to thank 
If you have any questions or feedback or if you see any errors in this reasoning, please let me know in the comments.
If not, please resteem this proposal so that we can get it implemented in the next hard fork, and overcome the reward pool abuse problem.
