I'd love to hear what starspirit thinks about the eDollar pegging mechanism, because it's interest rate based, just as his proposal for a peg on bitshares was:
http://forum.makerdao.com/t/how-the-edollar-peg-works/64
paging @starspirit
If I understand it correctly, Rune's proposal appears to use an interest rate to modify supply, while on the long side there is no symmetric mechanism to modify demand. Somebody let me know if this is a wrong interpretation, as I only read Rune's main post.
My personal preference is for a payment system between longs and shorts that modifies both the supply and demand curve. This would lead to greater stability in both total issuance and the interest charged to shorts. For instance, suppose external interest rates rose significantly (e.g. after a Fed hike), and the relative demand for bitUSD fell as a result. Suppose demand at the peg price falls from $10m to $5m. The immediate impact would be a discount as half the current owners try to sell. By charging shorts higher interest as per proposed eDollar, the supply can be contracted to $5m (note this would have happened anyway with settlements, but at least this encourages shorts to act voluntarily). But if instead longs received the interest, a smaller interest rate increase would be required, as this will help lift demand at the same time as reduce supply, meeting somewhere in the middle. [As an aside, somebody might be able to point out to whom Rune intends for the interest to be paid, as I was not clear on this].
Such a transfer scheme is not so easy to implement now in 2.0, because there will no longer be a direct yield mechanism. However, it is possible to do indirectly, which I've been looking at for some time, and it may even be simpler in the end. So for example, suppose we create deposit tokens, where each token represents a notional number of USD that varies over time with regular interest accruals. Longs and shorts maintain a constant exposure to deposit units, but this represents a varying exposure to USD, that is captured in the net asset value of the token. By setting the interest rate with respect to discount or premium to parity (similar to Rune's approach), a mechanism is introduced that tends toward restoring parity over time. For people to be able to spend actual bitUSD in transactions then, one of two approaches are possible. One is to allow conversion of deposit tokens on a 1:1 basis with non-interest paying bitUSD tokens. Parity on the bitUSD is ensured by the 1:1 conversion with deposit tokens. A second approach, which does not require freely held bitUSD, is to allow a mechanism where transactions denominated in bitUSD are always sent and received from deposit reserves, and settled in the appropriate number of deposit tokens. While I expect all this is feasible with Smartcoins, I have to wait and see what practical difficulties there will be with tying together the processes for interest rate formation, feed production, and NAV calculation. I need to speak to the developers more about these things as we get closer to 2.0, but unfortunately am just a bit short of time at the moment.
There are some additional aspects that need to be ironed out though. One problem that Rune identifies is that market shocks can create quite large deviations to parity before the interest rate adjustments have sufficient effect. As a result, he introduces a penalty/reward on settlements. I prefer an approach where the algorithm for adjusting interest rates takes account of the distance from parity, not just the direction. Larger distances can lead to larger rate adjustments, with smaller adjustments as parity is approached. With the algorithm being public and transparent, there ought to be plenty of market-makers willing to step in before such distances get too large, in the knowledge that the rate algorithm will be moving quite responsively to restore parity. In fact, with such an approach, there may not even be a need for settlements at all, adding another great simplification to the structure.
Another problem is the zero-bound on interest rates. Under both Rune's approach and my approach there is likely to an insufficient supply (and persistent premium) if there is insufficient incentive for shorts to participate even at 0% interest. In theory though, it is possible to have negative rates (shorts earn interest from longs), but as a practical issue there is a big question mark around this.
There's more I could add on this wonderful topic, but another time perhaps...
Nice post, it raises some interesting points I will take into consideration.
Regarding the interest on issuance debt, this interest is paid out passively as yield to holders of dai. However dai holders only receive 90% of the interest that dai issuers pay, with the remaining 10% going to Maker in order to pay for undercollateralization insurance. Should a black swan event happen Maker will autonomously buy up all bad collateralized debt positions using the funds it has accumulated in the Maker vault (the account where the insurance payments go to).
In case there aren't enough funds in the Maker vault to cover all the bad debt, makercoin is inflated and sold off to cover the bad debt, at the rate of 1% inflation per week until the system is once again fully collateralized. But when no bad debt exists in the system, every week 1% of all the funds in the Maker vault are instead used to buy up makercoin and permanently burn them, continuously reducing total makercoin supply over time and thus transferring income to the investors of Maker as long as everything is working as it should. Among other things this insurance mechanism incentivises makercoin holders to be active voters and actively ensure that collateral requirements and debt ceilings are set safely for every approved collateral type, since makercoin holders, not dai holders, have to foot the bill if something goes wrong (except in the extreme case where Maker is wiped out, i.e. makercoin price goes to 0. Then the only option is to perform a controlled permanent devalution of the dai).