I have not found a complete discussion of the liquidation backstop for Centrifuge (DROP token) assets, so I wanted to start a new thread. This is an important issue that I hope we can all discuss in a respectful, informative manner. As forewarning, I will raise a couple questions below. Hopefully there are clear answers to them that I just haven’t discovered or have not absorbed yet.
Also, much discussion has surrounded the DROP agreement for Centrifuge assets. This thread is for discussion of the practical mechanism of liquidation and its implications, specifically. An in-depth discussion of the DROP agreements in general is more appropriate elsewhere.
As I understand it, our model for Centrifuge assets is to open a private vault. The SPV (special purpose vehicle) which issues DROP tokens to investors on Centrifuge’s Tinlake platform is the entity able to access that vault. As collateral, they lock DROP tokens in the Maker vault. Note that DROP tokens are the initial collateral. Maker does not own the tokens, just as the ETH or WBTC or YFI locked as collateral is also not in the possession of Maker.
If everything goes smoothly, the SPV will repay its DAI loan to the vault, removing DROP collateral.
If things do not progress smoothly, at the moment liquidations must be initiated manually. This could be done for any number of reasons, and deserve their own thread. But it must be done through a governance vote at the moment.
Should a liquidation be initiated, Maker sends the DROP tokens to Tinlake for redemption. Tinlake pools are timed in blocks called “epochs” which have a minimum but not maximum duration. Whenever the end of an epoch is triggered, a settling occurs of requested redemptions against available DAI reserves for that pool + any new capital contributed by investors waiting for the next epoch to begin. Epochs have different lengths for different pools. I am not a good source for how long each one is – maybe it is a day, maybe it is weeks – and will defer to anyone who knows or can provide a source.
Should there be insufficient capital to meet all redemption requests, there is a Solver Mechanism that kicks in to determine in what order and proportion redemptions are met. I will not pretend to have studied the Solver Mechanism closely enough to guess at its effectiveness or how it operates in real life. To my knowledge – correct me if I am wrong – the Tinlake platform has not yet had to go through this unfortunate liquidation process in real life to provide any guidance from experience. This mechanism could be quite important if Maker liquidates a DROP vault.
While there may be some creative ways for a non-entity like Maker to pursue the limited recourse provided to DROP holders, to me this is the end of our involvement. The backstop is the liquidation process on Tinlake. DROP is our collateral, and whatever you may think does or does not secure DROP tokens, DROP is what Maker has available to liquidate.
My concern is that we are significantly mispricing our risk by accepting DROP collateral. Let me lay it out point by point. I hope someone has a clear, concise answer to each one:
If reserves are less than the value of the DROP Maker liquidates, we suffer a loss. Note that there is no 13% liquidation penalty to provide a buffer. Our involvement in each of these is significant, and I worry that by taking large shares of each DROP pool we onboard, we have exposed ourselves to the possibility of liquidating at steep losses.
We call for liquidation manually. I suspect the main reason we would do this is to get ahead of a potential collapse of the cash flows that collateralize DROP tokens. But to do this requires a governance vote, a public discussion of why we should vote to liquidate, and a reliance upon Maker’s limited ability to monitor the performance of the assets the SPV is managing. Our very discussion here to liquidate seems likely to signal to DROP holders on Centrifuge to also head for the exits, and the DROP we take as collateral is in no way senior to the DROP held by other investors.
Our ability to monitor these assets is currently quite limited. It will largely depend upon the self-reporting of the SPV that issues the DROP to investors and locks DROP to generate DAI. Given that Tinlake’s own representation of a pool’s metrics to prospective investors is bugged, I hesitate to rely upon the platform’s monitoring to supplement our own in a meaningful way. Our own processes and procedures for monitoring are still being refined and have not yet hit execution in a way that is sufficient to protect Maker from significant losses.
These pools are being interpreted by the DAO as being collateralized by the assets under management by the SPV. While these assets are important to consider, as they are the source of cash flows to repay DAI generated in RWA vaults, they are not collateral for Maker or the DROP tokens (See Section 4.G of the DROP agreement for proof of the latter). DROP tokens are what we have available to liquidate, and not what underpins the DROP tokens (which themselves are not secured). The mechanism by which we are able to recover funds is through the Tinlake redemption process. If we believe that DAI can be recovered there in full or large part, then a DROP is sufficient collateral. If we do not believe that DAI can be recovered there in full or large part, then DROP is not valuable collateral.
Maker accepts a lower rate of return than DROP holders on at least some of these pools. The New Silver pool is a good example. We accept a 3.5% stability fee, while DROP holders have a targeted return of 4%. This is actually less of a spread than when the pool was launched, when DROP was targeted at 5% vs our 3.5%. Note that we have less upside than a typical DROP investor, yet share the same exposure to the downside. Most importantly, should Maker and other DROP holders attempt to redeem simultaneously and there are insufficient reserves, Maker enjoys no special preference in repayment, despite accepting a lower yield. (*Edited to add*: Apparently all other investors and the asset originators are subsidized with liquidity rewards as well)
My takeaway is that these loans are under-collateralized, and have been mispriced. The collateralization ratio that is important is that of the value of DROP to DAI debt, not the underlying SPV’s assets under management. Not only are the assets under management not collateral for DROP (Annex B, Page 1 even has a section titled “DROP Tokens are unsecured”; Annex B, Page 2 also has "DROP tokens are non-recourse to the Issuer), Maker has only uncertain methods of pursuing any rights (through arbitration) attached to any DROP tokens it is left holding in the event of an unsuccessful liquidation on Tinlake.
Our approval of DROP vaults should be dependent upon our belief in a robust liquidation process, which will not only depend upon our confidence in Tinlake but also in each pool’s reserves (currently 0 DAI in the case of New Silver). The assets under management should not be our primary concern because while they do influence the ability of the SPV to pay DROP holders, those assets under management are not ultimately how Maker would recover liquidated debt.
We also do not have the benefit of oracles to forewarn us of impending issues, and any liquidation by Maker would provide ample time for other DROP holders to exit before or simultaneous to Maker, depending on the length of an epoch for that pool.
If anyone sees some factual errors in what I have laid out above, or feels I have mischaracterized or misinterpreted anything, please (charitably) explain in the comments. My main request is that you show your work.