So how long will it take for borrowing to recover? In January, borrowings recovered within a few days. The deleverage in February recovered within 10 days. April took longer again, with 17 days before borrowing levels reached pre-crash levels. The magnitude of the fall in borrowings in May however, is far beyond what we've seen in the last 6 months, both in absolute and relative terms. One month after the peak we are still 43% below peak borrowings The previous falls in borrowing recovered in line with underlying asset values (using ETH as a proxy). The fact that we are still 40% below peak ETH price is instructive, and aligns with the relative position of borrowings to the previous peak. It's reasonable to assume that borrowings will recover to pre-crash levels when we see asset prices make similar gains, unless Compound can attract capital at a faster rate than other players in the market.

The above user behaviour can be seen in the graph of flows below. The net flows into Compound is the sum of inflows to the protocol (deposits + loan repayments) minus the sum of outflows (redemptions and loans), excluding liquidation events. The inflows leading to a rise in the unpegged collateral can be seen on the 13th, 17th-18th and just prior to the crash on the 19th when nearly $1b flowed into Compound in 3 hours.

Given that the positions with stablecoin collateral and stablecoin borrowings are unaffected by the market crash, it's useful to look at the volume of unpegged collateral denominated in native units. For simplicity of analysis, I'll denominate all unpegged coins in eth, with the assumption that all unpegged coins had similar price falls over this period. This helps to see capital levels independent of the rapid price changes in the market

There period from the 16th to the 18th is interesting - crypto prices were falling and loan balances were stable, so users responded by adding more unpegged coin collateral. This reversed for a while on the 19th of May, until around 0100 UTC a rapid fall in the ETH price triggered a rapid build in unpegged collateral. The world fell apart at 1200 UTC on the 19th with a wave of liquidations. More followed later that night and into the 20th of May. Users were stuck - either out of capital, out of gas or out of time. The unpegged collateral levels fell rapidly as positions were liquidated. Another wave of liquidations followed on the 24th of May.

Now we will zoom into the period from May 10th to May 24th and look at the state of the Compound Markets.

You'll note in the graphs below that before the rapid falls on the 19th of May, about half of supplied collateral was denominated in stablecoins (USDC, USDT, DAI). On the loan side, over 90% of loans are denominated in stablecoins. This infers that roughly half of the loan book on Compound suffers an increase in collateralisation factor as a result of the rapid price falls on the 19th of May. Loans approaching the maximum collateralisation factor are in danger of being liquidated.

The crypto market crash of May 2021 occurred after all time highs were struck around May 13th. Large falls in many crypto markets occurred on May 19th, 21st and 24th. The chart below shows the ETH/USD exchange rate, which we will use as a proxy for crypto markets in general.

Looking back over the past 6 months there were significant liquidations in late February 2021. This coincided with a 25% drop in ETH/USD over a few days. Smaller clusters of liquidations occurred in January and April, again coinciding with sharp drops in ETH/USD.