Centrifuge: Onboarding RWA Backed Collateral to MCD

Over the last couple of weeks we have started conversations around how to think of illiquid assets in MCD. We started asking questions and engaging in a dialog on MIP6 proposals, in the Collateral Onboarding Calls and in the Risk & Governance calls. SAI has been live for over two years now and the community has built tools and frameworks for how to manage ETH as a collateral type. With the launch of MCD we only minimally had to change these to extend the process to include one more ERC20 token. So far we’ve could apply more or less the same rules to most collateral types (governance, risk, oracles etc.). Centrifuge is working on not just a single asset but a whole asset class that can be used as collateral in DeFi and specifically in MCD.

I think it’s a good idea to start a forum thread that discusses how “illiquid real world asset backed loans” can be onboarded as collateral to MCD. As we discuss these issues and work out some processes for these new assets, perhaps the result will be the introduction of a new MIP with suggestions on how to treat these assets.

What is a illiquid RWA backed loan pool?

Let’s define this category as broadly as possible: Any loan secured by some asset that legally can be used as collateral. Think of a mortgage backed by a house, a loan to a business secured by an invoice or commodities.

Centrifuge has two MIP6 proposals for assets on the forum. Most of what we will discuss in this thread will be specific to Centrifuge’s set of collateral but there are likely other tokens in DeFi that could fit the bill (RealT comes to mind but likely also DMM though I am not too familiar with their structure).

I think one common trait among these assets are:

  1. they are secured by a legal claim on property, future revenue or something comparable
  2. they have a value in USD or other currency (the amount that a borrower promises to repay).
  3. one ERC20 token represents a basket of collateral of the same type (onboarding individual non-fungible assets such as a single mortgage is possible but doesn’t scale well).
  4. the loan portfolio yields a fixed interest independent of the DSR/base rate (Real estate, factoring and a large part of traditional finance is fixed interest or tied to the overnight risk free rate).

In a recent Risk & Governance call (summary here) I gave a brief introduction on how our assets work and how governance for these assets might look like. If you’re not familiar, this might be a good starting point.

In our recent collateral onboarding calls (recordings and schedule) we started going through how Centrifuge’s assets work. Please also check out the technical documentation on the Tinlake contracts.

TIN & DROP: Junior & Senior Tranches

One particular thing about Centrifuge is that we are creating two tokens that represent a claim on the cash flows from the pool. The TIN token is similar to equity or the junior tranche in a traditional debt fund and the DROP token is the senior tranche. The DROP token offers a fixed yield secured by the TIN tranche that has a higher yield but takes first loss. Look at a some examples on how the tranches work.

Having two tranches greatly reduces the risk for DROP token holders making the DROP an interesting collateral type for MCD. By requiring the asset originators to hold some of their own TIN we can align incentives of the asset originators and give them some skin in the game.

TIN-Adjusted Collateralization Ratio

One interesting way to look at these two tranches its from MCD’s perspective:

In this example the DROP tokens are worth 800k yielding 5%. The DROP token is a claim on that money from the borrowers. But because the TIN token always takes first losses, any loss occurring in the portfolio isn’t proportionally applied to all investors but to TIN investors first. There are a total of $1M worth of loans backing these $800k in debt. The loan portfolio must depreciate 20% in value to be fall below the value of the DROP. You could say that there is a 125% collateralization ratio of the DROP.

When talking about the DROP token there are two important numbers:

  1. The face value of the token: the value of the token that is accumulating interest on an ongoing basis. This is the amount of DAI that you will get back when redeeming the token with Tinlake.
  2. The value of the loan portfolio which if higher than the DROP token face value acts as “over-collateralization”. This could be called the TIN-adjusted collateralization of DROP.

On chain cashflow of loans

An important part in adding DROP to MCD is to make sure that they can be redeemed for DAI as soon as possible. We achieve this by requiring all loan repayments to happen on chain. The Tinlake contracts are used to manage repayments on individual loans. If there are DROP investors that want to redeem their tokens, they get a claim on those revenues first, then TIN and only after can new loans be originated.

Prioritizing the different transactions in this way in Tinlake results in the DROP token holders wanting to redeem always get the best chance at getting their DAI from any loans that are repaid.

Deciding how to add these assets: Primary Issuance or Secondary Lending

Let’s assume there is a hypothetical pool of loans. The loans are secured by a lien on a property and overcollateralized by 120%. Loans pay on average 8%. An interest bearing ERC20 token represents a share in the pool of loans, it’s yield and in the case of default a claim on the underlying collateral. This is a very simplistic description of how Centrifuge’s DROP token works but I believe also fits a lot of other projects. I’m trying to make this as generic as possible.

There are two ways to value this ERC20 token:

  1. based on the value of the collateral used to secure these loans
  2. valuing it at the current outstanding debt of these loans

A rough description of how this could look:

Summarizing this illustration, in the secondary lender scenario Maker has an additional over-collateralization beyond what Tinlake requires with the TIN tranche which brings the effective collateralization ratio when looking at the Tin-adjusted portfolio value up to 139% in this fictional example.

Comparing the two scenarios

1) Primary Issuance
[+] Directly matching the DROP terms: higher yield
[+] Does not need an additional liquidity provider, easier for the Asset Originator (AO)
[+] Directly taking DROP from the AO leads to more stable/predictable borrowing demand
[-] Removes the capability of Maker to control DAI supply with the stability fee, but it could be manipulated by lowering/raising the debt ceiling requiring the Tinlake contracts to remove DROP the vault.

2) Secondary Lender
[+] Adds a “second level” of security/overcollateralization
[+] Can use stability fee to manipulate borrowing demand
[+] More similar to how ETH & other collateral types work in MCD today.
[-] Large part of DROP yield would likely go to the 3rd party providing capital for leverage (because third party would need to manage backup sources of liquidity in case Maker increases stability fee)
[-] A 3rd party arbitraging on the DROP would mean that the second MCD is not the cheapest borrowing market for these assets it would be replaced by something else.

Our current thesis is that going with the second scenario (Secondary Lender Approach) while a bit more complicated for the asset originator will be easier to add to MCD as it matches closer to the risk & gov process for assets we have today.

What are the community’s thoughts on this?

We will be posting more in this thread as we work through these topics together.


  • 2020-06-05: Replaced Graphics with more consistent numbers that serve as an example.

Great post, thank you @spin!

I think the main challenge is establishing valuations for the collateral assets (e.g. DROP) without a liquid trading market. Using the face value of the DROP tokens assumes the fair risk premium for the asset exactly matches the token’s interest rate.

Maker would benefit from an oracle or data source independent of the asset originator or origination platform to judge asset quality, like a credit rating or an auditor/accountant attestation of collateral value. Another possibility would be Maker contracting with an elected paid contributor to handle underwriting, or onboarding additional risk team members specifically for illiquid real world assets.

The drawback is that verification costs money on a per-pool basis, so it would make it less economical to finance smaller pools of loans. But the alternative of taking unverified data from the originator seems risky.


We discussed this thread in the Collateral Onboarding Call yesterday. Here’s the recording for those who couldn’t participate: https://youtu.be/x44aQJEWpEQ

I think the primary issuance scenario is much better.

It should allow Maker to expand much more.

I think the two options can be used together.

I’ve just updated the first thread in this post to add improve the description as well as the numbers used in the example to be more consistent based on feedback by @equivrel in our on Wednesday. I would love to get input on this as well from a few more voices in the community.

@cyrus, what are your two cents on this topic?