Liquidation auctions for MKR are unnecessary, leak value, and increases bad debt accumulation
Summary
The current MKR liquidation design in the LockstakeEngine (LSE) and LockstakeClipper results in significant value leakage and bad debt accumulation.
The process follows the established pattern for external collateral assets. However, MKR is an internally owned (burnable and mintable) collateral. Thus, the cycle of first market selling MKR during liquidations to arbitrageurs and paying for liquidations incentives, and later possibly buying it back through the flapper mechanism if a surplus is generated, is unnecessary and leaks a large amount of value (mostly during the liquidation auctions). Over time this leakage can be estimated to easily be above 0.5% or even 5% of protocol reserves.
Selling it in liquidation auctions to arbitrageurs, when the price is low, leaks value due to the necessity of arbitrage profits, gas prices, and keeper incentives. By replacing this process with immediate MKR burning during liquidations, and only selling through flopper auctions if and when necessary, in larger lots, and more controlled pace, the protocol could significantly reduce value leakage, and either reduce bad debt accumulation or increase profits.
Root Cause
The root cause of this issue lies in the design of the liquidation and surplus auction processes which is designed with immutable assets in mind and is unsuitable for an internally owned asset:
MKR is sold in liquidation auctions when positions become unsafe, at unfavorable rates due to the market conditions that triggered the liquidations.
Surplus DAI accumulates in the vow, which is then possibly used to buy MKR through the flapper (Splitter) mechanism, again from the open market.
This cycle of selling low (and maybe buying later) leads to value leakage through multiple inefficiencies:
Liquidation auctions, via many separate lots, incur gas costs, external profits, and keeper incentives (chips and tips).
MKR is often sold when its price is low during market stress and high gas prices.
Liquidations put additional, correlated, and uncontrollable selling pressure on the asset price, exacerbating the problem further, and compounding the problem of "selling-low". The additional pressure even risks liquidation cascades and additional bad debt.
Smaller in total size due to cancelling out with protocol surplus and various thresholds.
Smaller in total size since exact debt (and not chop multiplied equivalent debt) needs to be covered.
Controllable and more gradual pace.
Decoupled in time from short term market wide price action.
Decoupled in time from gas prices spikes.
Decoupled in time from market wide liquidity reduction (high slippage), and high arbitrageur opportunity costs.
Reduced risk of incentive farming (since chip and tip can be smaller, and collateral cannot be bought back during liquidation)
Internal pre-conditions
n/a
External pre-conditions
Market volatility leading over time to significant cumulative liquidation volume.
Attack Path
Market wide price decline reduces MKR price making a large amount of liquidations necessary.
Liquidation auctions are triggered at the depressed, low price, and high gas prices.
Protocol pays incentives in the form of chip and tip to the keepers.
The collateral is than sold to takers, but the price is not only low due to market conditions, but also the final take discount needs to generate arbitrage profits, cover high gas costs, reduced liquidity (slippage), and opportunity costs for the bots, so is considerably lower than market price.
The MKR auctions cause further price decline, triggering additional liquidations, and further protocol losses.
Impact
Because the chop (liquidation penalty) is commonly set to 113%, it's reasonable to conclude that liquidation process inefficiency is up to 13% loss. I'll assume a more conservative 5% leakage below. The exact number is likely to vary, but regardless of the exact figure, the inefficiencies accumulate over time.
The impact of this inefficiency is substantial:
Value Leakage: Assuming the conservative 5% value leakage on MKR sold in liquidation auctions, when cumulative liquidation volume reaches 10M USD, approximately 500K USD will be lost to inefficiencies. This is equivalent to a [0.5% loss vs. the assumed 100M protocol reserves figure](https://github.com/makerdao/sherlock-contest/blob/master/README.md#severity-definitions.
Long-term Losses: Over a prolonged period (years), when cumulative liquidations reach 100M USD, the total value leaked could be as high as 5M USD, or 5% of assumed protocol reserves.
Increased Bad Debt: The inefficient process may lead to increased bad debt accumulation, as the protocol consistently sells MKR at unfavorable prices, and risk bad debt generation due to exacerbating liquidation cascades.
Reduced Protocol Sustainability: Continuous value leakage reduces the protocol's ability to maintain a healthy balance sheet and manage risk effectively.
PoC
n/a
Mitigation
Simplify the current liquidation auction process by replacing it with MKR burning:
When a position becomes unsafe, on LockstakeClipper's kick can avoid initiating any auctions.
Instead onKick will burn the full chopped amount of MKR.
The chip and tip for this clipper may be much smaller.
The callback process can be simplified to a single step (onKick).
Conversely, the clipper's take and redo can be removed as well.
mrhudson890
Medium
Liquidation auctions for MKR are unnecessary, leak value, and increases bad debt accumulation
Summary
The current MKR liquidation design in the LockstakeEngine (LSE) and LockstakeClipper results in significant value leakage and bad debt accumulation.
The process follows the established pattern for external collateral assets. However, MKR is an internally owned (burnable and mintable) collateral. Thus, the cycle of first market selling MKR during liquidations to arbitrageurs and paying for liquidations incentives, and later possibly buying it back through the flapper mechanism if a surplus is generated, is unnecessary and leaks a large amount of value (mostly during the liquidation auctions). Over time this leakage can be estimated to easily be above 0.5% or even 5% of protocol reserves.
Selling it in liquidation auctions to arbitrageurs, when the price is low, leaks value due to the necessity of arbitrage profits, gas prices, and keeper incentives. By replacing this process with immediate MKR burning during liquidations, and only selling through flopper auctions if and when necessary, in larger lots, and more controlled pace, the protocol could significantly reduce value leakage, and either reduce bad debt accumulation or increase profits.
Root Cause
The root cause of this issue lies in the design of the liquidation and surplus auction processes which is designed with immutable assets in mind and is unsuitable for an internally owned asset:
In comparison, using flopper auctions has the following advantages:
chop
multiplied equivalent debt) needs to be covered.Internal pre-conditions
n/a
External pre-conditions
Attack Path
Impact
Because the chop (liquidation penalty) is commonly set to 113%, it's reasonable to conclude that liquidation process inefficiency is up to 13% loss. I'll assume a more conservative 5% leakage below. The exact number is likely to vary, but regardless of the exact figure, the inefficiencies accumulate over time.
The impact of this inefficiency is substantial:
PoC
n/a
Mitigation
Simplify the current liquidation auction process by replacing it with MKR burning:
kick
can avoid initiating any auctions.onKick
will burn the full chopped amount of MKR.chip
andtip
for this clipper may be much smaller.onKick
).take
andredo
can be removed as well.