Inclusion of a DAI Stability Index, to MKR holder risk profile

Background:
The current MKR holders (~$500M) are directly exposed to:

  1. additional risk of ~$100M in outstanding DAI.
  2. additional incentive of stability fee accrued of ~$5M DAI.
    Of course Indirectly MKR holders face the standard crypto macro economic risks, like any other tokens holders, and have strong vested interest in the overall success of the project as measured by success of the community, stability of DAI, precision of the peg, size of the DAI market, and so on. MKR holders are currently rewarded or punished by these factors indirectly through free market mechanisms.

Problem:
Moral hazard due to low MKR voter turn out.
Assuming 1% voter turn out, at current evaluation levels, less than 0.5% of MKR ($2.5M), can manipulate the DAI market of $100M , for a few percent $1- $3Million, without adversely affecting the MKR owners holdings.
Although I don’t believe such activities have taken place, and the proof or disproof of them will always remain controversial, this is equivalent to a financial system where a small hedge fund runs FOMC and can manipulate LIBOR, as an analogy with respect to the US capital markets. These ratios, (the percent of stake required to manipulation returns) would increase an order of magnitude if DAI is increasingly embedded in DeFi lending products, derivative platforms, and modern structured finance.

Parameterization for an automated quantitative solution:
In order to introduce viable, automated, transparent solutions there is a need for standard quantifications. I like to propose two examples, but these are arbitrary and are only meant as examples:

Example 1) Stability Index (DSI). On a periodic basis (for instance daily, as a short term incentive) we could measure the standard deviation of DAI/USD from a set of reliable exchanges, using Maker’s existing oracle technology, and calculate sigma. We can then set reasonable target of x sigma as the goal then:
a) If sigma = x do nothing
b) If sigma > x penalize
c) If sigma < x reward

Example 2) DAI Growth Index (DGI). On a periodic basis (for instance quarterly, as a long term incentive) we measure the moving average of quantity of DAI in circulation, and penalize/reward based on a set of governance success criteria.

Rewards/Risks mechanism:
Once the appropriate metric(s) are defined, exact mechanisms are required, before implementation. There are two considerations here:

a) Who do we reward/punish?
There are various levels of responsibility with pros/cons for each:
Example 1) All MKR Holders
Example 2) MKR Holders in the voting contract
Example 3) MKR holder responsible for the vote outcome
Example 4) All to different degrees.

b) How do we reward/punish?
Similarly there are various methods, with various scope:
Example 1) Mint/Burn MKR
Example 2) Distribute to/from accrued stability fee, or create a separate pool.

My goal here is not so much to provide the specific details for a specific solution. All these aspects should be polled and voted on. It is more important to get a dialogue started, and to get ideas brewing.

For example we could on the short term, execute large interest rate moves incrementally, rather than abruptly to minimize the threat.

However, I consider the problem of moral hazard as a serious threat to the long term health of the project, if voter participation remains low, and the market size continues to grow.

Happy Halloween!

4 Likes

I really like your post.

Yes! I would argue that this is a power that the fed has now.

I’m skeptical that any mechanism within the system can disciple members. Derivatives outside the system can always invert incentives. I think the ultimate incentive to keep DAI stable is that the code is open-source. A competitor could launch a competitor to DAI if sufficient goodwill is lost due to abuse by MKR holders.

Thanks for your post @PatrickDehkordi

A number of people have been dissatisfied with the MKR voter turnout, but it is actually not really a worry for me.

Here is why:

  1. a lot of MKR is owned by what we could call institutional investors or venture captital firms. These guys and girls will simply never vote. They did their voting when they decided to buy MKR.

  2. other people are voting, including whales, but so far with only a fraction of the MKR they actually own. The reasons are various.

  3. The whole deal with voting is to govern, to decide direction. Right now, and for quite some time into the future, there is no direction except grow. Everybody agrees on that. You could vote for a stability rate change or to fiddle a bit with the debt ceiling but all involved knows these are details. That is why the -4% stability rate change, which on paper is drastic, did not affect the price of DAI, DAI issuance or the MKR price in any big way.

Based on the above I expect governance participation to stay low for quite some time until there are some real issues on the table.

Here are a few issues that could raise the temperature sufficiently to increase voting:

  • Maker finances. Either I have not been paying attention of there have been no news whatsoever about the financial state of Maker. A bit of openess here would improve the risk parameters (pun intended).
  • Vault ethics. At some point questions will be asked about the morality of certain potential vault assets. In theory it is fully doable to use the stock of a Russian land mine supplier as vault material, but I expect it to raise the need for a vault ethics guideline that potentially could generate huge debate.
  • Leftover funds. If there are funds left over after main development is done how will these be spent? Right now this is perhaps less of a worry as Maker is burning through an (quick back of envelope estimate) USD500-600k per month. Once all is done there is possibly not too much left to argue about. But if MKR appreciates there could in theory quickly be a huge pile to divide. Expect heat.

Here is what I am actually worried about: Governance effectiveness. Effectiveness in the sense of evaluating, grading and voting on potential vault assets and the respective risk parameters. From what I have experienced from the processes around USDC, FUND and SNT is that the time necessary to evaluate them and then digest the findings and ultimately decide on risk parameters is simply a factor or two too high. The evaluation process must somehow be more streamlined and more automatic for the Vault to grow. If you ask me - even at a slightly increased risk.

2 Likes

Thanks for the responses.
I know these issues may not be relevant for the immediate term and depend on the success of the project, but the goal should be to make the system as stable and scalable as possible.
If voter turn out is expected to remain low, then I would argue for automation in governance.
Why not consider incorporating something like a simple version of the “Taylor Rule” to monitor DAI stability and adjust rates?

Really like the post @PatrickDehkordi thanks so much for sharing your thoughts on this (and also for using appropriate forum tags).

I’ve been thinking about this on and off for some time now. I’ll share some brief bullet points of my thoughts, hopefully people can use them as a jumping off point.


Problem 1: Voter Turnout

The first problem is voter turnout. We need to increase turnout to secure the system, and in theory having multiple informed parties voting should increase the systems effectiveness. More informed votes = more effective outcome.

Rewards

  • We could provide rewards for voting, any rewards would need to be proportional to the amount of MKR voting due to the sybil problem.
  • There is a risk that providing rewards in this way would cause people to vote in an uninformed manner just to get the voting reward MKR.
  • Adding a lockup period would reduce this problem, and would be beneficial in general, often the effects of a vote are not immediate, MKR Holders are supposed to have responsibility for their actions but this responsibility ends if they can sell their MKR before negative effects manifest.
  • The voting reward would have to come from a portion of the stability fee.
  • To whom do we provide the rewards, all the MKR locked in DSChief, or the MKR voting in executives? What about polls, do we reward participation in those?
  • This could be setup similar to Compound’s ‘c’ tokens. The pool of MKR in the Chief could grow and IOU Tokens would give the holder a right to a share of the MKR locked in the Chief becoming a ‘cMKR’ token.
  • Unlike ‘c’ tokens, ‘cMKR’ tokens shouldn’t be transferable, as this could risk undermining MKR liquidity that is required for the operation of the DCS. (And would break any potential lockup mechanic.

Penalties

  • We could penalise people for not voting. This could work in a number of ways, either by changing the MKR token burn such that those tokens are transferred only to voters, or by transferring MKR directly from non-voters to voters.
  • ‘Direct transfer’ penalties cause major problems with market making and MKR liquidity. I’m pretty sure this type of penalty is flat out unworkable.
  • The alternative is to completely or partially end MKR Burn and instead transfer this value to voting MKR (I’ve pretty much covered this in the Rewards bulletpoints.)
  • We could change how the last resort minting works. For example, we could mint twice the MKR that is required by the debt auction, and distribute half of that MKR proportionally between voting addresses. This has the effect of penalising non-voters to a greater degree than voters in the event of a value shortfall in the DCS.
  • The incentives of the above are questionable. Part of the intention behind the last resort minting is to penalize the MKR Token Holders that have voted in bad parameters.

Problem 2: How close do we try to match the peg?

The second problem is that there isn’t a direct incentive for MKR Holders to have Dai match the peg closely. There are effective incentives to keep it in the right ballpark, but we seem to have landed on a view that ~1% is probably fine. I’m not speaking for or against this, but there is the possiblity, as you said, of directly rewarding or penalising based on the closeness to the peg.

  • Part of the problem with rewarding or penalising based on the ‘DSI’ is that the value has to go to or come from somewhere. If we mint MKR and sell it as a punishment, what do we do with the Dai generated? If we want to burn MKR as a reward, where do we get the Dai to buy the MKR?
  • If we want to punish, we can mint MKR and send the Dai proceeds to the null address, making it inaccessible (better yet, we can store it and vote to distribute it to charity or wider DeFi growth initiatives). There is no way to reward, we can’t create value from nothing.
  • We could setup a sliding scale that punishes based on the ‘DSI’, it could be quadratic based on the time spent sigma spends away from zero. This tuning function could be tweaked. At the very least this would ensure that Dai oscillates around the peg (which I would argue is what we should aim for.)

On a more general note, I think that the idea of having an RSI and DGI as described by you here, are fantastic ideas, and I would love to see these implemented as advisory signals that can inform governance voting.

@Vishesh / @cyrus do you think there is any merit to these metrics, if so, how difficult would it be to record metrics like the proposed DSI and DGI and display them in graph form in the weekly governance and risk meeting? Once we have these metrics, we can propose and vote on targets, try to get some community alignment over what we should be aiming for.

1 Like

Okay, maybe it wouldn’t be a bad idea to making voting mandatory. Check out https://swiftdao.com/blog/automatedexit

Thanks I appreciate the feedback.

I see a getTargetPrice() in dai.js, but not sure if there is a getMarketPrice() or something equivalent in the API. But it shouldn’t be too difficult to implement the indices I mentioned, off chain to track governance performance. I think these would be good addition to the new governance analytics portal.

I think at this point in the project, it’s still better to try to get MKR holders involved in governance rather than automate the functions, cause it is still a relatively small community.