Paying a negative fee immediately from the buyer who pays a positive fee is nothing more than shifting the price. It is a meaningless change.
Taker pays the fee to both maker and the network. It's more than shifting price because it encourages makers thus improving the liquidity.
Sometimes it helps me to imagine an extreme case to get a better feel for what is happening on the smaller scales. So what follows is my own thought experiment.
Lets try the experiment with a 20% taker fee just to be outrageous. Lets have the maker fee be -20%. Under such a market, those who demand liquidity take a 20% loss and those who provide it get a 20% gain. The trader would view such a market similar to a market with a 20% spread. They would be hesitant to buy such an asset because they know they will take an instant 36% loss if they are the taker on both sides. (.8 in and .8 out). This means that someone looking to get in-or-out of such a market with the least loss would have to be a maker for one of the two trades in which case they take a 4% loss (.8*1.2=.96). Those who are willing to "wait" on both sides of the trade can profit by 44% (1.2*1.2). Hence the negative maker fee encourages users to wait (which is the opposite of liquidity). You create "lines" on both side of the order book of people who want to exit their position. I suspect you would see very narrow spreads with steep walls. This market would have the appearance of good liquidity, but the underlying reality is that 'day traders' view it as a market with a 20% spread.
So in this extreme case the takers (aka traders) end up paying for their liquidity TODAY the same way they would pay for their liquidity without the 20/-20 maker/taker rule: via a large spread. Market makers end up making the same profits they would if they had a 20% spread between their buy and sell walls. The only thing the negative maker fee is doing under this model is enforcing a minimal spread that makers can provide, in other words price-fixing the market maker fee. Instead of market makers competing to reduce spread, they are competing to be the first in line of a "virtual" 0 spread. Because no value is moving through time all this price fixing is doing is creating shortages (of takers) and gas lines (those waiting to exit on both sides of the book).
So it is clear that if we set the maker/taker fees to be greater than the natural spreads that things break down. Our real goal is to reduce spreads, not enforce a minimum spread with steep cliffs of liquidity on either side of that minimum spread.
So this means that we want to maximize maker rewards without increasing the cost to the taker. So lets look at another extreme market:
1. Suppose that takers paid a 0.1% fee
2. Suppose that makers earned 20% bonus from someone else (ie: the network).
In this market there would be huge walls of liquidity as people compete to get a 44% return every time they turn over. This 44% return completely eliminates almost all market volatility risk. Traders/takers see an effective spread of just 0.1% which means they feel very comfortable buying the asset because they know they can turn around and sell it instantly with only a 0.2% loss. Assuming there was no limit on the 20% bonus, then people would start trading against themselves. Obviously you would have to mitigate this self trading by making the reward based upon how long the order was on the books before getting filled. This is the situation we really want to create.
So the question becomes how do you compensate makers today without making todays traders (takers) pay for it. My proposal has tomorrows takers pay today's makers by paying for market making today, but not tomorrow.
The proposal here says you give them 0.1% today + a cut of the net present value of all future fees.
The cost of providing liquidity on early on is much more expensive than the cost of providing the same liquidity in a mature market.
Under this model you gain more liquidity from makers early on (when it costs the most) without actually decreasing liquidity available in the future (when it costs less). If you set the decay curve properly you can end up with "constant liquidity" equal to the average liquidity over the entire life of the market. Over a long enough time horizon this means that you should get almost as much liquidity in year 1 as you do in year 30 if market makers believe in the future of a given market.
So when people suggest "simple" rules they are not really getting the result they want.