Aave’s loan risk assessment

elenahoo
8 min readDec 1, 2021

Aave is a non-custodial liquidity protocol that provides investors the opportunity to earn interests on deposits from cyrpto assets. It also provides loans for users to borrow given a certain amount of collateral is deposited in Aave.

In order to make sure the borrowers will repay the loan, a series of risk factors are in place to constrain the amount of loan a borrower can get and also to keep track of the healthiness of the loans in the protocol. This article will look into these risk parameters such as collateral, loan-to-value (LTV), liquidation threshold and health factor to see which assets are the most popular ones to borrow and which assets are the best to deposit and to use as collateral; and what risk assessment conclusions can be made about the borrowers.

In Aave’s deposit and borrow markets, everything works very similarly to the traditional bank loans — the lender (depositor of the asset) receives an interest return from the loan, the borrower pays an interest for borrowing. The only thing additional is that the depositor will receive a token called “aToken” that serves as a deposit certificate of the underlying asset and the token also trades on open markets and contains the future potential accrued interests from the loan.

From a borrower’s perspective, the following criteria need to be fulfilled to be able to borrow and keep the loan:

  1. An eligible asset as collateral deposited into Aave. The eligibility of each asset is assessed based on a series of risk factors such as smart contract risk of the token, counterparty risk of the DAO who governs the token, market risks arise from the volatility of the token. The details of the risk factors used to assess the collateral eligibility can be found here.
  2. A loan size less than the LTV threshold set for each asset. Stablecoins usually have a higher LTV threshold than other coins due to its less risky and volatile nature. The LTV threshold for each asset can be found here.
  3. A health factor less than 1 at all time. When the market value of the loan is more than the value of the collateral, the uncollateralised portion will be liquidated at a discount — your collateral will be sold at a lower than market price.

The most popular assets to borrow

Since the launch of Aave protocol v2 at the beginning of 2020, the total amount of loan borrowed in USD has been increasing and reached the peak at the end of April 2021. The number of borrowers although also peaked at the same time as the borrowed amount in April, the recent months have seen some decrease compared to the first quarter of 2021.

Total borrowed amount USD and # of borrowers (original graph)

The most popular assets to borrow are the dollar stablecoins such as USDC, USDT and DAI. The number of borrowers from the stablecoins are more than 3x-5x higher than the other coins. Looking at the daily average number of borrowers using the past 30 days and all time data, the average daily number of borrowers for non-stablecoins is rarely above 10 while the amount borrowed is also mostly at a low level below 2 million USD; whereas for stablecoin, the borrower number could range from 20 to over 100 on a daily basis and the volume is also much higher.

Total # of borrower and amount borrowed (original graph)
Daily average # of borrower and amount — 30d average vs. all-time average (original graph)

In general, stablecoins are mostly used for borrowing while more volatile assets are used as collateral. This is not due to any constraints imposed by Aave for non-stablecoins as collateral. In fact, all listed non-stablecoins can be used as collateral except for RenFIL (see here). One possible reason why borrowers use non-stablecoins as collateral could be that they are long on these coins and expect the values to go up against dollar in the near future, subsequently expecting a lower LTV and more buffer from the liquidation threshold. It could also be that the stablecoins have a wider usage for other purposes across chains so they’re more popular choices.

The best collateral to pledge

Historically, the average LTV across all loans is below 1 most of the time as expected. It is worth noticing that in the early days of the Aave v2 launch, the average LTV is quite volatile and also has more times when the LTV is more than one. In the recent periods, the LTV has been stabilising. Even after the peak in borrowers and borrowing amount in April-May 2021, the spike in LTV to around 1.27 is much lower than the previous peaks back in 2020 (almost 1.5 LTV), showing a safer and healthier environment in recent times than at the beginning of the launch.

Historic LTV (original graph)

The best collateral is the one that holds the most value relative to the borrowed asset and also has enough liquidity in the open market so it can be liquidated easily without a fire sale. To evaluate which collateral holds the most value, one has to look at not only the collateral-debt token pairs but also the pair value at liquidation time. The graphs below show the loan-to-liquidation value (LTV) for each collateral vs. borrowed (debt) token pair at liquidation. The higher the LTV at liquidation, the less value the collateral has relative to the borrowed loan.

Looking at the lowest 10 LTV pairs, WBTC as a collateral holds the most value against a loan in FRAX (44% LTV). This is quite an outlier compared to the other pairs with relatively low LTV in the region of 80%. YFI, XSUSHI and CRV also hold good value as collateral against loans in dollar-pegged stablecoins.

Lowest 10 LTV per collateral-debt token (original graph)

Looking at the average LTV by borrowed asset against all collateral assets for the recent periods, USDT seems to have depreciated against collateral assets more than other borrowed assets, resulting in lower LTVs in May and June in 2021. Vice versa, when looking at the average LTV by collateral asset against all borrowed assets, LINK and UNI hold the best collateral value against borrowed assets at liquidation, with the lowest LTV hence the best collaterals in the recent periods.

Collateral by borrowed asset — bubble size represents the # of borrowers (original graph)
LTV by collateral — bubble size represents the # of borrowers (original graph)
LTV by collateral-debt token pair — bubble size represents the # of borrowers (original graph)

The best asset to deposit

If you are not interested in pledging your deposit as collateral to borrow other assets but only want to deposit it and earn some return, a good way to evaluate the best asset for this purpose is to look at the price of the aTokens.

Since the aToken is a deposit certificate that represents not only the value of the deposit but also the potential future cash flow from the interests earned from the loan, the aToken market value should be evaluated at a higher price than the original deposit asset (see details here). If this is not the case, it usually indicates a poor market expectations of the interests to be earned in the future, either due to lack of demand for borrowing such asset or the interests are too low to matter.

The graph below shows the % of days where the aToken market price is higher than the original asset price i.e. aUSDT’s price is higher than USDT. Most of the stablecoins seem to have higher aToken market value for more than 80% of the days. Stablecoins such as BUSD and DAI show that most of the time the aToken market price (orange line) is higher than the original token price (blue line).

% of days aToken price > reserve token price (original graph)
Examples of aToken price > deposit token price

It is also worth noting that most of the other non-stablecoins have very similar aToken price to their original token. There are two outliers — WBTC and RAI. For WBTC, the aWBTC price deviates below the WBTC price for a period and then goes back to closely tracking the WBTC price; whereas for RAI the aRAI price seems to be fluctuating around the RAI price in a wider range but with no clear tendency to be lower or higher than the RAI price.

Two outliers — WBTC & RAI

Borrower risk indicators

Finally, we look at what the risk indicators tell us about the borrowers. According to Aave’s risk parameters documentation (see here), for each borrower wallet, the following three risk indicators can be used to assess the individual borrower risk:

  • Max LTV — calculated as the weighted average of the LTVs of the collateral assets and their value. The higher the value, the lower the borrower risk.
  • Liquidation Threshold — calculated as the weighted average of the liquidation thresholds of the collateral assets and their value. The higher the value, the lower the borrower risk.
  • Health Factor — calculated as the weighted average of the liquidation thresholds of the collateral assets and the total borrows value. The higher the value, the lower the borrower risk. When the factor is less than 1, the position may be liquidated to maintain solvency.

For all the borrowers, the maximum LTV and liquidation threshold range between 0% to 80%, with most of them concentrated in the range between 70%-80% LTV. All the borrowers have a health factor close to or below 1. The concentration seems to be below 0.2, which indicates quite a healthy borrower environment.

Box plot and dot plot for Max LTV, Health Factor and Liquidation Threshold

The scatter plots also show most of the borrowers concentrate in the high maximum LTV and high health factor region. The liquidation threshold is highly positively correlated with the maximum LTV and is also always higher or equal to the maximum LTV to leave some buffer for the liquidation.

Scatter plot among Max LTV, Health Factor and Liquidation Threshold

Conclusion

Based on the assessment of multiple risk factors in Aave’s loan borrowing, it is shown that most of the borrowers are in the safe and healthy region with enough buffer to repay their loans.

The most popular assets to borrow are:

  • USDC
  • USDT
  • DAI
  • WETH

The best collaterals against stablecoin borrowings are:

  • WBTC
  • YFI
  • xSUSHI
  • CRV

The best assets to deposit according to market value are the stablecoins:

  • GUSD
  • USDC
  • USDT
  • BUSD
  • SUSD

Note: The python code for the plots can be found here.

Sign up to discover human stories that deepen your understanding of the world.

elenahoo
elenahoo

Written by elenahoo

Specialise in AI and DeFi analytics & modelling. You can DM me for questions or discussions on Twitter @elenahoolu

No responses yet

Write a response