Collateralization ratio

The ratio, and the same number written backwards

Divide collateral value by debt value. Post $15,000 of collateral against $10,000 of borrowings and the ratio is 150%. Loan to value is the same relationship inverted, debt over collateral, which is 66.7% on that position, and lending protocols set their limits in those terms. Aave's published risk parameters state it directly: at an LTV of 75%, for every 1 ETH worth of collateral a borrower can draw 0.75 ETH worth of another asset.¹

Convert that maximum into a ratio and it is 133%. The number founders should carry out of this page is not the ratio at opening, though. It is the distance between where the position sits now and where it gets closed, measured in price rather than in percentage points of ratio.

Three thresholds, and only one of them closes the position

The opening limit is loan to value, which caps what can be borrowed on day one. The maintenance limit is the liquidation threshold, which Aave defines as the percentage at which a position is treated as undercollateralised: at a threshold of 80%, once borrowings pass 80% of collateral value the position may be liquidated.¹ The summary statistic is the health factor, collateral value weighted by each asset's liquidation threshold divided by total borrowings, and below 1 the position may be liquidated to keep the protocol solvent.¹

Compound expresses the same idea with different vocabulary, mapping user balances to prices through a price oracle and then to risk weights it calls collateral factors, with a comptroller contract making the determination.² The naming differs between protocols and the arithmetic does not, which is worth knowing before comparing two positions across two venues.

Now run the numbers on that pair. A borrower at 75% LTV with an 80% liquidation threshold is liquidated once collateral falls to 93.75% of its value at opening. That is a 6.25% move, and it does not care whether the borrower intended to be conservative.

Why this is not the coverage ratio

The two get used as synonyms in conversation and they measure different things on different clocks. Coverage ratio, as used in economic simulation, is cumulative revenue divided by cumulative cost over a modelled horizon, read across percentiles to see how often a design goes underwater. It describes the solvency of a business over time and the consequence of a breach is treasury drawdown.

Collateralization ratio is a point in time balance sheet measure on one position, priced by an oracle, and the consequence of a breach is automated: somebody else repays your debt and takes your collateral at a discount. One is a question about whether a model survives. The other is a question about whether a position survives this afternoon. A design review needs both, and confusing them produces a treasury plan that reads healthy while individual positions are being closed.

The numerator is a quote and the thresholds are somebody's decision

Both halves of this ratio are softer than they look. The collateral value is a price the protocol did not observe, delivered by an oracle, and a 150% ratio against an asset that gaps 40% in one block was never a 150% ratio. The number is only as good as the depth behind the price feed, which is why the honest version of this test is run at the price where liquidations start cascading rather than at spot.

The thresholds move too. Aave's access control contract defines named on chain roles including pool admin, emergency admin, risk admin and asset listing admin, each grantable and revocable through governance.³ A risk admin can change the parameter that decides when your position is closed. That is a governance permission working as designed, not a flaw, but it means a collateralization ratio is a policy setting rather than a law of the protocol, and it belongs in the admin capability register alongside every other privileged action.

The stablecoin version measures something else entirely

The same phrase gets applied to reserves, and the design target is different. A fiat backed stablecoin aims at 1:1, so the interesting question is not the size of a buffer above 100% but the quality and liquidity of what sits inside the 100%. Circle disclosed on 11 March 2023 that the USDC reserve was 77% short dated US Treasury bills worth $32.4 billion and 23% cash worth $9.7 billion, with $3.3 billion of that cash at Silicon Valley Bank, about 8% of the total.⁴

Overcollateralised designs are the ones where a buffer is the product. A position backed at 150% is carrying a 50% cushion for volatility, and the cushion is what lets an automated liquidation still recover the debt after a price fall. Comparing a 150% CDP ratio to a 100% fiat reserve as if the higher number is safer misreads both. One is absorbing price risk on a volatile asset, the other is holding an instrument that is not supposed to move.

What to settle before your token is accepted as collateral

Getting listed as collateral looks like a milestone and behaves like a transfer of risk onto your own liquidity. Four questions belong in the analysis first. What liquidation threshold applies, and what price move it implies given real depth rather than historical volatility. Which oracle prices your asset and how expensive it is to move that oracle relative to the profit from doing so. What the liquidation discount is, since that is the price your token gets sold at during a cascade. And where the liquidators route, because it is usually the pool you seeded.

In our view the parameter to argue about is the threshold, not the LTV. The LTV decides how much leverage a user can take. The threshold decides how far the market has to move before your token becomes forced supply, and that number was set by a governance process you do not control. None of this is advice about any asset or any position; it is the arithmetic a design team should run before the listing, rather than after the first cascade.