[ARFC] Aave Umbrella - activation

[LlamaRisk]

LlamaRisk’s Umbrella Methodology

As the activation of Umbrella draws near, we aim to provide the Aave DAO with insights into how we will recommend parameters and manage risk in collaboration with other members of the Aave Finance Committee (AFC). At the core of this effort is LlamaRisk’s Umbrella methodology, a quantitative framework specifically designed to determine appropriate Umbrella Fund Caps for assets covered by the mechanism.

This document highlights the key components of LlamaRisk’s Umbrella methodology, including data inputs, analytical processes, outputs, and governance considerations. By leveraging data-driven analysis and scenario-based simulations, our methodology ensures that the protocol is adequately capitalized to withstand predefined market stress scenarios. These recommendations aim to enhance Aave’s risk management, optimize capital efficiency, and support the protocol’s sustainable growth.

Background

Large-scale liquidations pose systemic risks to lending protocols. Insufficient market liquidity can leave undercollateralized debt, leading to bad debt and potential insolvency. To mitigate this, protocols use Umbrella Funds (or Safety Modules) as backstops for shortfalls from failed liquidations. These funds must balance being large enough to handle crises and avoiding inefficient capital allocation.

Aave enhances this approach with Umbrella, a per-asset Umbrella Fund that builds on stkAAVE and stkGHO Safety Modules. The Umbrella methodology ensures Aave can withstand predefined market stress (e.g., bad debt equal to 10% of an asset’s borrow cap) by analyzing liquidity, user positions, and price shocks. It recommends fund caps that ensure safety while minimizing unnecessary capital lock-up.

Methodology Inputs

The accuracy and robustness of the Umbrella methodology depend on the quality and comprehensiveness of input data captured at a specific timestamp. The methodology utilizes two main categories of data:

1. Protocol State Data:

• Detailed parameters for every supported asset (e.g., current market prices, supply and borrow caps, Loan-to-Value (LTV) ratios, Liquidation Thresholds, Liquidation Bonuses, and E-Mode category inclusion).

• A comprehensive snapshot of every user’s account, capturing their collateral and debt balances and E-Mode status.

2. External Market Data:

• Historical price time series for all relevant assets, critical for modeling potential future price shocks using techniques like Value-at-Risk (VaR) and correlation analysis.

• Data on market liquidity and depth for relevant major decentralized exchanges (DEXs) trading pairs.

Methodology Flow and Outputs

The Umbrella simulation follows a multi-stage process:

1. Data Ingestion & Preparation: Collect and consolidate all the necessary Protocol State and Market Data for the chosen timestamp. Filter user positions based on criteria like minimum debt size and potentially filter by E-Mode category if analyzing a specific E-Mode.

Figure: Example state of an asset and the market parameters on Aave

2. Synthetic Price Shock Generation: Based on historical price data analysis, calculating daily returns, identifying stressed regimes, computing VaR99 and inter-asset correlations during drawdowns, generating many distinct, synthetic price shock scenarios. Each scenario represents a potential future market state, assigning a specific negative price change (e.g., -5%, -10%) to each asset within Aave’s Core market, reflecting correlated downturns.

Figure: Example of generated price shock samples

3. Liquidation Simulation (Iterated per Shock Scenario):

• Select one shock scenario and apply the specified price changes to the current market prices of all assets.

• Re-evaluate the Health Factor of all user positions using the shocked asset prices and the protocol’s Liquidation Thresholds (or E-Mode specific thresholds). Identify all positions falling below the liquidation threshold (HF < 1).

• For the specific asset being analyzed, sum up the total debt owed across all identified liquidatable positions. Also, determine the composition and total amount of all collateral assets backing this specific debt.

• Model the process of liquidating the aggregated collateral. This involves simulating swaps through the relevant liquidity pools, accounting for the price impact caused by the liquidation volume. The simulation finds the Optimal Scale factor: the maximum fraction of the total aggregated liquidatable debt (for this asset, in this scenario) that rational liquidators can liquidate profitably, considering the Liquidation Bonus and the simulated price impact. This is achieved using iterative search algorithms to pinpoint the profitability threshold.

• Store the calculated Total Liquidatable Debt and the corresponding Optimal Scale factor for the target asset resulting from each simulated shock scenario.

4. Outlier Filtration: Analyze the distribution of Total Liquidatable Debt and Optimal Scale values across all scenarios. Apply a statistical outlier detection using predefined multipliers. Scenarios falling outside these calculated bounds are flagged as outliers and removed from the subsequent calculation to prevent extreme, potentially unrealistic scenarios from unduly influencing the outcome.

Figure: Example of detected outlier samples

5. Market Liquidation Capacity Calculation: Using only the filtered inlier data points, calculate the mean of the Total Liquidatable Debt values and the mean of the Optimal Scale values. The Market Liquidation Capacity is then computed as:

This value represents the methodology’s estimate of how much debt (in the specific asset) the market can reliably absorb through liquidations under the stress conditions defined by the inlier shock scenarios without causing losses to liquidators.

6. Umbrella Fund Requirement Calculation: Compare the calculated Market Liquidation Capacity against the predefined Safety Target parameter, defining a Safety Threshold over a total Borrow Cap of an asset.

The final Umbrella fund size is then determined by the gap between the Safety Target and Market Liquidation Capacity:

Governance Parameter Management

The Umbrella methodology provides a structured computational framework that needs to align with the protocol’s overall strategy and risk appetite. The Safety Target is the primary parameter for governance interaction, defined as the percentage of an asset’s Debt Cap that the protocol aims to protect through market liquidity and the Umbrella Fund. The percentage of Aave revenue allocated to Umbrella is also a key parameter that will need to be decided by governance, representing another important lever for calibrating the Umbrella mechanism to the protocol’s risk management objectives and financial sustainability goals.

The flexibility of adjusting the Safety Target allows the Aave DAO and the AFC to align Umbrella Fund levels with various considerations. For instance, the inherent risk profile of an asset, its specific borrow concentration patterns, prevailing market conditions, and broader strategic growth initiatives involving risk parameter changes (like LTV adjustments) can all inform the appropriate level for the Safety Target. This adaptability ensures that Umbrella levels reflect both current risks and future ambitions.

LlamaRisk initially recommends starting with moderate Safety Thresholds across major assets. As the DAO proposes and implements strategic market risk adjustments, the Safety Target for affected assets can be reviewed and modified accordingly by the AFC. This creates a direct, dynamic link between risk management and protocol growth objectives. Adjusting the Safety Target becomes a collaborative process, ensuring that the Umbrella methodology’s Umbrella Fund Cap recommendations remain aligned with the evolving strategic direction and risk tolerance of the Aave protocol. Other methodological parameters are generally considered more stable but remain transparent and subject to periodic review.

Initial Scope: Assets and E-Modes

The Umbrella methodology, particularly in its initial implementation, primarily aims to analyze and recommend Umbrella Fund Caps for the most systemically important assets within the protocol. This includes major stablecoins like USDC, USDT, and DAI, constituting a large portion of borrowing activity. It also covers large-cap volatile assets such as WETH (and its staked derivatives like wstETH) and WBTC, representing significant collateral value and borrowing demand.

The methodology is designed with flexibility regarding E-Modes. The structure allows for running the analysis specifically for assets within a designated E-Mode, using the appropriate E-Mode-specific Liquidation Threshold. The goal remains consistent: ensuring sufficient backing (market liquidity + Umbrella) to handle liquidations up to the Safety Target within the specific risk parameters of that asset or E-Mode category.