Aave should create a new loan type where a portion of the interest paid by the borrower contributes to a reduction in their collateralization ratio. As borrowers build credit they become able to take undercollateralized loans. Lenders sacrifice a portion of their yield by diverting it to collateral reduction, but enable novel borrowing behavior.
In the case of a borrower successfully building a credit history but then defaulting on an undercollateralized loan, the lender would still ensure net profit over the lifetime of the lender-borrower relationship.
I agree this can become the primary source of on-chain consumer credit if it can be paired with a social graph.
Currently in Aave, borrowers that want to take a bigger loan than their last one have to add collateral. This is in addition to the transaction they make when paying back the loan+interest to the lender. This proposal saves the borrower the step of adding collateral and lets them contribute their interest to it as another means of taking a bigger loan.
@seccode the issue with reducing the collateralization based on repayments would be that you can easily game the system (borrow, repay constantly). However, undercollateralized lending is something that we are researching actively at the moment and I think one interesting idea is to utilize the Credit Delegation feature that is build into the Aave Protocol.
Perhaps even create permissionless Credit Delegation Vaults that anyone can spin off but would appear in the Aave UI based on Uniswap-style token lists.
You could set different parameters to these vaults for example how much interest you would pay to the delegators etc.
@stani This system of collateralization is actually not one that can be gamed by a borrower. For a borrower to borrow+repay constantly in order to reduce their collateralization ratio, the lender will accumulate a net profit over the course of these efforts.
The collateral reduction is explicitly tied to a fraction of the lender’s profit, so to get an uncollateralized loan immediately via gaming mechanisms, the borrower will have to pay the lender directly. A lender would not be against a borrower doing this.
The idea of credit delegation is super interesting for this idea because a borrower with a credit history that allows them to take uncollateralized loans can co-sign a loan for someone they know. The new borrower can now take out a loan against the credit history of the original borrower.
This proposal features a clear value add to the Aave protocol - not sure why this hasn’t gotten more attention. Can someone from the community gives this another look?
I don’t see any system non “gameable” on which the borrower doesn’t pay more in advance to depositors than the difference between over and under collateralization.
If for example, a borrower borrows a value of $1’000 with a value of $1’500 collateral, first, the interest rate he pays should still follow exactly the same dynamics as currently, so the proposed redirection of interest to reduce over-collateralization requirements should be on top. Second, the amount of under collateralization allowed would be exactly the same as that interest paid on top by the borrower, so it becomes a net 0 dynamic.
In general, under-collateralization implemented on the protocol layer is not achievable without risk added to the system, a risk that can be of multiple natures but boiling down to a mechanism outside the protocol that should enforce liquidation.
Credit Delegation solves this by moving the risk to the depositor delegating “collateral power”, and the game-changing aspect is that a depositor can be an autonomous entity (smart contract) coding any type of logic managing the risk.
Using your example, if the borrower profits from the loan and successfully pays it back, they may want to take a larger loan in the future. To do this, they not only have to pay back the interest from the previous loan, but also add more collateral. Consolidating these two transactions into one improves UX and saves gas costs. Aave should definitely implement this because it is a clear value add.