Skip to main content

Decentralized collateral-based lending and risks

· 6 min read
NEST Protocol

A traditional mortgage is simple. For example, John gives Amy a valuable asset against which Amy can lend John some cash or other assets. At the end of the term, John pays the principal and interest to Amy, and Amy gives John the collateral. If it is not repaid at maturity, Amy has the right to dispose of or auction off the collateral. On the other hand, if the price of the collateral fluctuates dynamically, Amy has the right to ask John to replenish the collateral as the price changes, otherwise it does not have to wait until maturity for Amy to dispose of the collateral.

However, the difficulty of matching, the cost of time for both borrowers and lenders, makes p2p matching very inefficient, which is the reason why most DeFi lending could not be done in the early days. So, people started looking for a way out, one option is to make a pool with high capital utilization, and the other one is to directly let the borrower pledge into some kind of stable coin. As you can see, the former is the model of Compound and the latter is the model of MakerDAO. Both models ensure the stable existence of one of the lending parties before seeking a counterparty to the transaction: Compound ensures that the lender's pool of funds is sufficient, while MakerDAO creates a pool of borrower's assets first; one being investment-oriented and the other borrower-oriented. However, if the collateral operation is on-chain and is done in a decentralized manner, the above process immediately raises the following questions: firstly, who will complete the matching of collateral and lending? Secondly, who will complete the margin call or liquidation once the price changes?

Liquidation risk

In a funding or asset pool model, liquidation risk from mortgage price fluctuations is always important. If the liquidity of the collateral asset is infinite and the price is always valid, then there is never a liquidation risk no matter what the collateral rate is, as long as the value of the collateral asset is higher than the amount borrowed. That is, whether you have a 60% collateralization rate (interest-bearing borrowing size divided by the collateral value) or 90% will always be the same. However, in the real world, liquidity is neither infinite, nor are prices always valid. In some extreme cases, price fluctuations directly cause the value of the collateral asset to fall below the size of the borrowing. This is known as a wearing: since everything is decentralized and you can't demand a wearing payout by locking out the borrower, this loss can only be taken by the lender. In addition, in a pooling model, clearing cannot be done by a designated person because it leads to some kind of trust risk: failure to clear when it should be done results in greater losses. Therefore, the clearing operation must be decentralized. This means that the incentives are designed to ensure that any third party has an incentive to complete the operation, and in the current DeFi project, the clearing residual is designed for this incentive.

Liquidation risk is a typical tail risk. In the Compound and MakerDao, this is addressed by having everyone share the risk. This introduces an increase in lender risk and instability in the intrinsic value of the DAI (no longer equal to $1). The best solution to tail risk like this is to introduce a decentralized insurance mechanism: for the most part, collect insurance premiums and pay out in case of such extreme situations, the whole process is decentralized and anyone can become a participant in the insurance fund or exit it at the right time. The insurance fund is equivalent to the collateral operator selling the tail risk to a third party, which can guarantee the safety of the lender or the stability of the DAI. This design represents that part of the interest on the borrowing, as well as the so-called stabilization fee, should be paid to the insurance fund. Considering that the size of the insurance fund is dynamic, the ability here to determine the risk consideration of the market through the dynamic size of the insurance provides an indicator to view the liquidation risk of the whole pool.

Reinvestment risk

If the lender never has anyone to lend money in the pool, it will cause its own composite return to fall, which is what we call reinvestment risk: after earning interest for a period of time, it needs new opportunities to pay interest rather than a stable long-term return.

Combining the liquidation risk we just discussed with the insurance fund, which in this model will have less ongoing stabilization fees or interest because some of its collateral assets are liquidated, implies a term structure with different liquidation times: the earlier the liquidation, the higher the reinvestment risk, and the later the liquidation, the lower the reinvestment risk. Given the price model, the time to start liquidation is related to the collateral ratio or liquidation line, and given that the two often remain linear, we will measure the composition of the time to start liquidation in terms of the collateral ratio: the higher the collateral ratio (note that it is the borrowed assets divided by the collateral assets), the shorter the time to do liquidation, and the lower the collateral ratio, the longer the time to do liquidation. In this way the mortgage rate constitutes a measure of reinvestment risk, which means that the insurance rate can be set based on the mortgage rate, for example a simple solution is a linear formula. Of course, if strictly speaking, it is perfectly possible to give a more precise formula.

In this way, the insurance fund, the insurance rate associated with the collateral rate (in Compound a part of the interest rate, in MakerDAO a stabilization fee), constitutes the most basic risk management scheme for decentralized collateralized lending or collateralized stable coins. This scheme is fundamental and atomistic. We will see more and more collateral-based programs in the future to make such adjustments.