[LINK] Collateral Onboarding Risk Evaluation

Legal Disclaimer: This communication is provided for information purposes only. This communication has been prepared based upon information, including market prices, data and other information, from sources believed to be reliable, but such information has not independently been verified and this communication makes no representations about the enduring accuracy of the information or its appropriateness for a given situation. This content is provided for informational purposes only, and should not be relied upon as legal, business, investment, financial or tax advice. You should consult your own advisers as to those matters. References to any digital assets and the use of finance-related terminology are for illustrative purposes only , and do not constitute any recommendation for any action or an offer to provide investment, financial or other advisory services. This content is not directed at nor intended for use by the MakerDAO community (“MakerDAO”), and may not under any circumstances be relied upon when making a decision to purchase any other digital asset referenced herein. The digital assets referenced herein currently face an uncertain regulatory landscape in not only the United States but also in many foreign jurisdictions, including but not limited to the UK, European Union, Singapore, Korea, Japan and China. The legal and regulatory risks inherent in referenced digital assets are not the subject of this content. For guidance regarding the possibility of said risks, one should consult with his or her own appropriate legal and/or regulatory counsel. Charts and graphs provided within are for informational purposes solely and should not be relied upon when making any decision. The content speaks only as of the date indicated. Any projections, estimates, forecasts, targets, prospects, and/or opinions expressed in these materials are subject to change without notice and may differ or be contrary to opinions expressed by others.

Analysis part 1/2

Summary of Proposed Risk Parameters

Risk Premium: 2%
Liquidation Ratio: 175%
Debt Ceiling: 5 million
Auction Lot Size: 50,000
Minimum Bid Increment: 3%
Bid Duration: 6 hours
Max Auction Duration: 6 hours
Liquidation Penalty: 13%
Dust: 100 Dai


  1. Background
  2. Token data
  3. Exchange activity
  4. 3rd party earning and lending platforms
  5. Risk parameters


Protocol Summary

From the website:

Chainlink is a general-purpose framework for building decentralized oracle networks that give your smart contract access to secure and reliable data inputs and outputs.
Each decentralized oracle network consists of a collection of independent node operators, a method for aggregating data, and pre-made “Chainlinks” (also called external adaptor) that act as middleware to give you access to any API you want to leverage for data and/or external services.

Chainlink currently provides decentralization at both the oracle and data source level. By using multiple independent Chainlink nodes, the user can protect against one oracle being a single point of failure. Similarly, using multiple data sources for sourcing market prices, the user can protect against one data source being a single source of truth. Both of these ensure that the oracle mechanism triggering your important smart contract is as secure and reliable as the underlying blockchain.
You can use Chainlink to connect to data providers, web APIs, enterprise systems, cloud providers, IoT devices, payment systems, other blockchains and much more.

LINK token use

From the whitepaper:

The ChainLink network utilizes the LINK token** to pay ChainLink Node operators for the retrieval of data from off-chain data feeds, formatting of data into blockchain readable formats, off-chain computation, and uptime guarantees they provide as operators. In order for a smart contract on networks like Ethereum to use a ChainLink node, they will need to pay their chosen ChainLink Node Operator using LINK tokens, with prices being set by the node operator based on demand for the off-chain resource their ChainLink provides, and the supply of other similar resources. The LINK token is an ERC20 token, with the additional ERC223 “transfer and call” functionality of transfer(address,uint256,bytes), allowing tokens to be received and processed by contracts within a single transaction.

Regarding the token sale, the whitepaper states:

LINK Tokens are not securities, investments or currency, and are not sold or marketed as such. Also: participating in the sale involves significant technological and systemic risks; the sale is not open to individuals who reside in or are citizens of the United States or Canada. The sale period, duration, pricing, and other provisions may change, as stated in the token sale terms. The LINK Token sale involves known and unknown risks, uncertainties, and other factors that may cause the actual functionality, utility, or levels of use of LINK Tokens to be materially different from any projected future results, use, functionality or utility expressed or implied by SCCL in the terms.

From website

LINK tokens are used to pay node operators for retrieving data for smart contracts and also for deposits placed by node operators as required by contract creators. The LINK token is an ERC677 token that inherits functionality from the ERC20 token standard and allows token transfers to contain a data payload. Read more about the ERC677 transferAndCall token standard.

Token data

On-chain activity

Daily active addresses

The count of active addresses has been continuously rising over the last 12 months. The rise in activity has really sped up since the Black Thursday recovery where it has continuously maintained its level above the two year mean of 1549 daily active addresses. When looking from a distribution point of view, most episodes where the count of active addresses are either above the mean or largely above the mean happened since the second quarter of 2020. The maximum count happened on July 13th where as many as 14253 addresses were active in the network. This rise is likely due also to the increase in token price, where both metrics share a 0.90 two year correlation. However data indicates that the latest rises in price have created more outliers in that correlation between the two variables than historically.

Source: Santiment

New unique addresses

Along with daily active addresses metric, the increase in the count of newly unique addresses has also been very noticeable since the second quarter of 2020. It has mostly moved in tandem with the active addresses over that period. Here the network added as much as 6,508 new unique addresses in a single day on July 13th. As is the case for active addresses, the creation of new unique addresses has been mostly above its two year mean since the Black Thursday events. New unique addresses have also a high degree of correlation with the token price at 0.89. However, it is important to notice here that both new unique addresses and daily active addresses seem to have a high degree of collinearity when it comes to LINK’s price prediction. That means both independent variables would most likely cancel each other’s predictive power in a regression model.

Another important point, according to data, is the growing number of the zero balance addresses in the last couple of months. These accounts transfer or sell out all of their tokens which could indicate a fair amount of speculative activity happening beyond the network fundamentals and demand for the asset.

Sources: Santiment, glassnode

Source: IntoTheBlock

2 year correlation to token price
metric closePriceUsd new_unique_add daily_active_addresses
closePriceUsd 1.00 0.89 0.90
new_unique_add 0.89 1.00 0.97
daily_active_addresses 0.90 0.97 1.00


polynomial coeff: [ 975.38918393 -376.65582042]

polynomial coeff: [ 439.20430438 -203.16809245]

Daily transaction volume (LINK)

The transaction volume shows the aggregate amount of tokens across all transactions that happened on the network for LINK. According to the distribution the bulk of the transaction volume has mostly hovered around its two year mean value of 2,500,150 LINK per day or 0.25% of total supply,despite recent price increases. The last 12 months show however a few spikes in volume activity which seem to indicate these aligned with intervals of significant price decrease. The most recent spikes being Black Thursday and around mid July 2020.

Sources: Santiment

Sources: Santiment

Token allocation

The Chainlink contract has a fixed and capped supply of 1,000,000,000 tokens. According to data from Etherscan, LINK’s token offering launched on September 19, 2017, after a 32 million fundraise hard cap through the ICO. There were two rounds of LINK sale offerings to the market. One pre-sale where LINK was sold at 0.09 USD per token with a 20% bonus. A second public sale sold LINK at 0.11 USD to investors. In total both sales accounted for 350 million LINK sold to the market out of 1 billion tokens issued in the network. The final ICO allocation of tokens was structured as follows:

  • 35% of supply was allocated to token sale investors
  • 35% of supply was reserved to Node Operators and ecosystem rewards that would be distributed to incentivise network participants
  • 30% of supply was allocated to LINK’s parent company, SmartContract.com and its team members to support protocol development

The details of this allocation can be found on the Etherscan transaction here.

The parent company SmartContract.com that is behind the project team has been active since 2014 with the aim to create a bridge product to bring external data sources into public blockchains, through a decentralised oracle network that works along with node operators. Chainlink’s mainnet went live on Ethereum on June 1 2019.

Due to the nature of the ICO token allocation, the actual circulating supply of tokens in the market was initially 35%, i.e. which went to investors. But since the ICO, the addresses associated with the treasury had a few rounds of selling which brings the current circulating supply to an estimated 385,009,556 LINK. It is interesting to note that this increased circulating supply seems to have had an effect in the market price, as the first spikes in market activity happened at the same moment as the team started offloading some of its ICO treasury. Circulating supply calculation is based on coingecko methodology. Finally, LINK tokens held by the team do not seem to have a vesting schedule, according official documentation.

A number of Chainlink node operators provide data feeds to DeFi services against fees determined in LINK by the operator. This link and here display a list of oracle providers, along with their respective transactions, earned LINK from data feeds and other operational metrics. At time of writing, Chainlink’s official staking program for node operators has not been released to the public. For more information on the notion of implicit vs explicit staking, see article. Other FAQs on node operations, can be found here.

In the following sections we will cover asset price returns, token ownership distribution and exchange activity. Token ownership and its concentration amongst top addresses is a particularly important topic in this risk analysis, as it is very noticeable in comparison to other ERC20 projects.

Sources: Etherscan, Messari, Chainlink

Key general information

Variable Value
ID chainlink
TOTAL_SUPPLY 1,000,000,000
MARKET_CAP 4,734,190,898
TOTAL_VOLUME 1,442,365,825
ATH 19.83
ATL 0.15

Source: coingecko

Source: Santiment

Asset price returns

Investors to the LINK token have enjoyed a significant return since the ICO. Those involved in the first presale have seen as much as 18,466% return on their investments. At time of writing, over the last 12 months, investors have seen a 353% yearly return and a two year return of 2,579%. Looking at some technicals, LINK’s price appreciation seems to have been mostly supported by its 200 day moving average over the last year, with only one episode early in 2020 where the 50MA crossed below the 200MA trend. By the end of July both 200 MA and 50 MA were sitting at 4.03 USD and 5.91 USD, respectively. Were the price to drop to its 50 MA levels, that would mean a 31.4% decline. Historically, daily declines below 20% only occurred twice in the last two years: 19 Nov 2018 and Black Thursday. See daily price percentage change distribution below.

Cumulative return 12M

Date Return % Return months closePriceUsd Circulating_supply (est.)
2020-07-31 353.82 12 7.77 382009560
2020-04-30 169.36 9 3.72 379509561
2020-01-31 128.54 6 2.82 377509561
2019-10-31 124.63 3 2.74 377009566

Source: Santiment

Price and moving average

Source: Santiment

Downside risks timeseries

From a daily pricing distribution perspective, LINK has seen negative daily returns 358 times out of a 731 timeseries. 3.1% or 23 of those were sharper than -10% decline in a day. Comparatively, ETH has seen daily price declines a total of 366 times and 20 times below -10% in the same period. The average negative decline for LINK and ETH were -4% and -3% respectively in the two year period.

ETH and LINK have insignificant correlation to each other at 0.33 in this interval.

2 year LINK and ETH correlation

metric ether_closePriceUsd chainlink_closePriceUsd
ether_closePriceUsd 1.00 0.33
chainlink_closePriceUsd 0.33 1.00

source: santiment

source: santiment

Token ownership distribution

Token ownership concentration is one of the topics of greatest concern from a risk perspective in our view. In the following charts we will cover the biggest sticking points when it comes to distribution of assets amongst token holders, the relationship of these to asset pricing and the kind of risks this could bring to the vault portfolio.

Amount held by top addresses: as % of total supply and % of circulating supply

The amount held by top addresses displays the amount of tokens held by the top address holders, which by default includes the top 10 holders. In this metric both exchange and non-exchange top holders are put together. It shows that for the first 12 months of the timeseries the share of ownership as a percentage of total and circulating supplies has remained mostly unchanged at 76% and 35% respectively. Subsequently, those metrics have dropped in the following 12 months.

Source: Santiment

Amount held by top addresses: exchanges vs non-exchange addresses

However, when it comes to splitting out the token amount held by top holders into two groups, top known exchange addresses and top non-exchange addresses, the data provides a diverging trend in the network. Whilst both have reduced their token ownership share in the network in the first year, we can observe a re-accumulation of tokens in the second quarter of 2020 by top non-exchange addresses. That rise in ownership coincides with increased asset price post Black Thursday events. Top 10 non-exchange address ownership represented 63.7% of total supply by July end. Also the reduced amount of tokens in top exchanges could be one of the reasons for the 2020 recent price increase, drawn out of shortage in liquidity.

Source: Santiment

Token amount held by number of addresses

When we look at the timeseries dataset at a more granular level, the following chart shows the amount of tokens held by buckets of address counts over time. As we have seen previously, LINK has attracted a substantial growth in daily active addresses and new unique addresses in the last year. However, a significant portion of that growth in address numbers went to addresses holding 1K tokens or less. At the top percentile of the distribution, the count of addresses has remained mostly unchanged. There is a very small number of addresses holding more than 1M tokens, 6 of which belong to the team treasury and one to node operators. Only 328 addresses hold 100K tokens or more. In other words, only 0.16% of addresses hold the largest shares of the network.

Source: Glassnode

Gini coefficient

The gini index provides an extra insight into the LINK frequency distribution of tokens over addresses. Note that this metric calculation by Glassnode excludes exchange addresses, smart contract addresses, and other special asset-specific addresses e.g. team treasury. This metric is measured on a scale from 0 to 1, where 0 represents perfect distribution equality and 1 perfect inequality of token distribution. As of mid July, LINK has a gini of 0.97. It can be noted that the gini index has increased continuously in the last 12 months, despite the growth in active addresses since the mainnet launch last year and the team treasury offloads into the market. This may be a sign that already existing big buyers are acquiring even more tokens in the market.

This brings a significant ownership risk to the MakerDAO vault portfolio as actions by these holders, whether coordinated or not, may determine potentially significant swings in the asset price and by extension, in case of negative swings, in increased collateral liquidations.

It is important to point out that the use of addresses for the gini index in crypto is purely a proxy for distribution. One entity may own multiple addresses which would bring the gini index even closer to 1.

Note however that this type of high gini index is not particularly uncommon in many crypto projects. This post provides extra insights.

Source: Glassnode

Treasury funds

Team treasury fund addresses have been identified from glassnode and etherscan. Since the ICO, the Chainlink team has offloaded a portion of its original 300 million LINK tokens, mainly in the last 12 months. As per ICO terms, these sales have likely been done to secure funding to support protocol development. In total, 32,009,579 LINK has moved from the team original wallets, bringing their share of total supply from 30% to 27%. The move of tokens out of treasury accounts indicates there is no vesting of funds, which may bring asset price risks should a large portion come into the market in a short interval of time. No official documentation analysed indicates a vesting schedule for these funds. In this analysis treasury tokens have been accounted in the calculation of estimated circulating supply.

In terms of treasury management, in a recent announcement Chainlink appears to have placed a portion of its bitcoin (BTC) and ether (ETH) crypto holdings under Celsius’ treasury management service. Unfortunately, this analysis could not identify the addresses holding ETH or BTC on behalf of the team.

Source: Santiment

Token Velocity & Age consumed

Velocity displays the average number of times that a single LINK token changes addresses daily. Higher token velocity means that a single token is used more often in daily transactions. For the most part of the last 12 months, velocity has hovered around its 2.09 median value, with very little deviance. The biggest outliers in that long term trend happened between February and April 2020, with its highest point on Feb 17, 2020 at 5.67. Since then, velocity has returned to its yearly mean.

On the other end, age consumed may be a valuable metric to observe long term for Risk teams. It tracks the movement of previously idle LINK tokens. The metric shows the amount of LINK changing addresses daily multiplied by the number of days since they last moved. Spikes indicate a significant amount of previously idle LINK tokens moving between addresses. In other words, it gives a sort of view of “network inertia”. This metric can be used to track the movement of big longterm token holders with historically little trading activity. Spikes such as those of July 2019 can be used as preemptive measures for selling pressures. For most of the last 12 months, this metric has not moved much beyond its median 32,054,984 mark. But this recent historical precedent should not be discarded in long term analysis of extreme events.

Source: Santiment


Analysis part 2/2

Exchange activity

Supply on exchanges

Compared to other ERC20 tokens in the market, LINK has a comparatively low percentage of tokens stored in exchange wallets at only 6.95% of the total supply and 18.18% of the circulating supply as of the end of July 2020. It is important to point out that this metric has been continuously deteriorating in the last 12 months without any significant rebounds. For risk, this reinforces the likelihood of liquidity shortages in the event of liquidations.

Also, note a significant negative two year correlation between token supply on exchanges and price change at -0.88. However, that correlation seems to be reducing in the last month or so.

Across different exchanges, Binance has the largest market share by a considerable margin with 87.2% of tokens being stored in their wallets as of July end. This is not surprising considering how high LINK’s daily trading volume is on that exchange.

Sources: Santiment, Glassnode

Summary stats

Stats Coin Supply on Exchanges (as % of total supply)
count 732.00
mean 10.47
std 1.93
min 6.95
25% 8.82
50% 10.48
75% 12.38
max 13.16

Sources: Santiment

2 year correlation of % coins on exchanges and asset price

closePriceUsd Coin Supply on Exchanges (as % of total supply) Coin Supply on Exchanges
closePriceUsd 1.00 -0.88 -0.88
Coin Supply on Exchanges (as % of total supply) -0.88 1.00 1.00
Coin Supply on Exchanges -0.88 1.00 1.00


polynomial coeff: [-0.99961486 12.44690592]

As circulating tokens increased (with treasury selling tokens into the market), supply on exchanges was decreased as % of total. This seems to indicate a purchasing and holding behaviour on the part of large and smaller holders. Gini and other indicators shown above point towards a proportionally large accumulation by larger token holders.

Source: Santiment

Centralised exchanges

List of exchanges

For the remainder of the analysis on centralised exchanges trading activities, we have taken a rather conservative approach by adopting Messari’s methodology for “Real volume” in our assessment. This considers only a subset of significant and regulated exchanges, given the unfortunate common practice of wash-trading across the industry. Click here for further details.

Off that list, it is very visible that Binance and Coinbase are the key exchanges for LINK trading. Both exchanges account for about 75% and 19.25% of trading volume in CEX, respectively, at July 2020 end.

Aggregated exchange volume (snapshot)

Note: See Messari “Real volume”

Sources: Coingecko, Messari

List of exchange pairs

At a more granular level, LINK/USDT and LINK/BTC on Binance and LINK/USD on Coinbase are clearly the dominating trading pairs (86.4% of total) across major centralised exchanges. Again, this reinforces the need for a risk approach to trading volume distribution across a limited number of players and trading pairs.

Exchange pairs traded and volumes (snapshot)

coin_id base target market.name volume volume %
chainlink LINK USDT Binance 17074183 51.96
chainlink LINK BTC Binance 6089910 18.53
chainlink LINK TUSD Binance 65866 0.20
chainlink LINK ETH Binance 483011 1.47
chainlink LINK BUSD Binance 324180 0.99
chainlink LINK USDC Binance 143346 0.44
chainlink LINK USD Binance US 378586 1.15
chainlink LINK USDT Bittrex 21975 0.07
chainlink LINK USD Bittrex 66464 0.20
chainlink LINK ETH Bittrex 8572 0.03
chainlink LINK BTC Bittrex 49187 0.15
chainlink LINK USD Coinbase Pro 5219283 15.88
chainlink LINK ETH Coinbase Pro 530999 1.62
chainlink LINK EUR Coinbase Pro 399065 1.21
chainlink LINK GBP Coinbase Pro 174781 0.53
chainlink LINK BTC Gemini 5892 0.02
chainlink LINK ETH Gemini 1073 0.00
chainlink LINK USD Gemini 151772 0.46
chainlink LINK EUR Kraken 674162 2.05
chainlink LINK USD Kraken 838399 2.55
chainlink LINK ETH Kraken 26636 0.08
chainlink LINK XBT Kraken 107398 0.33
chainlink LINK BTC Poloniex 6924 0.02
chainlink LINK USDT Poloniex 18201 0.06

Sources: Coingecko, Messari

Trading CEX: tokens on exchanges

As mentioned above, the amount of LINK tokens held on known exchange addresses has been declining continuously in the last year, with Binance having the largest share of the pie. There is only 62,316,642 LINK (6.23% of total of supply) readily available in aggregate on exchanges. Given the trend in the last 12 months, in the event of a significant price decline, there is a risk of lack of liquidity for MakerDAO auction keepers. This risk will be proportional to long term debt ceiling levels. Another aspect to take into account is the growing number of leveraging platforms such as Aave including LINK in their listed collaterals. With declining token supply on exchanges, the competition for trading activity tends to accelerate, with likely risky effects for keepers trying to liquidate LINK.

For these reasons this metric should be very closely monitored by Risk teams and debt ceilings as well as risk premiums set accordingly.

Note: unfortunately, in the following chart Coinbase data was not readily available. Therefore there is a significant discount to the actual amount of LINK in wallets.

Source: Glassnode

Trading: CEX volumes

“Real trading volume” has fluctuated significantly across exchanges in the last two years. Although we have a mean daily value of 21,492,798.15 LINK traded in the markets, the variance of that mean is as much as 459053733493354.56, indicating a very high level of oscillation in trading activity. The volume variation goes from a minimum of 1,773,250 LINK to a max of 220,741,335 LINK traded in the two year data. In the second chart below, we smoothed out the real volume metric with a 30 day rolling value to give a view of the oscillation from that mean (in green). Although trading volume association with price was clear in Black Thursday events, there is no significant two year correlation between price and volume (0.23).

Source: Cryptocompare

Trading: DEX & Liquidity pool Volumes

Trading: CEX vs DEX volume

In comparison to centralised exchanges, trading volume in decentralised exchanges and liquidity pools is fairly low for LINK. It has been on the rise from mid-second quarter of 2020, particularly since the launch of Uniswap V2 on May 18. It included LINK/ETH trading pairs. As of the end of July, the biggest share of DEX volume has been traded on Uniswap or 0x liquidity DEXes. Bancor and Kyber also report some increase in volume, but insignificant in comparison.

It is important to note however that timeseries data for decentralised exchange volumes is not only dispersed across different api services, but also not properly maintained. This creates a lot of gaps for longer timeseries’ views or poorly reliable information. Very few data aggregator services (eg. cryptocompare) maintain any timeseries for DEX volumes and often reported volumes conflict with project application dashboard ones. This situation is constantly evolving with new services now starting to fill that gap in the ecosystem. There is an opportunity for monitoring of this dataset based on new data providers coming to the marketplace.

Sources: Cryptocompare, 0xtracker, Bancor

Source: Uniswap.info

Source: 0xtracker

Source: Kyber Tracker

Trading: DEX vs CEX volume distribution

From a distribution perspective, we have a mean value of 4,823 LINK for DEXes and 21,492,798 LINK for centralised exchanges over the same two year period. This share of trading activity may shift over the coming years towards more DEXes with services like Uniswap. The last three months has seen a significant shift in the distribution (31,073 LINK mean value). But for the foreseeable future we expect the large majority of LINK trade still flowing through centralised exchanges.

Sources: Cryptocompare, 0xtracker, Bancor

Sources: Cryptocompare, 0xtracker, Bancor

Source: Cryptocompare

3rd party earning & lending platforms

List of DeFi platforms

A small number of DeFi and CeFi lending platforms have included LINK in their reserves over the last 18 months. At time of writing, Aave is the only one in DeFi proposing both savings and lending services, ever since bzX paused its offering. In CeFi, Nuo has both savings and lending facilities and Celsius and Argent only savings. For comparison with MakerDAO only, this analysis focuses on Aave.

Note: We plan to extend this kind of scope with more data in future onboarding assessments and data monitoring. The purpose is to benchmark some risk policies and strategies applied on these platforms in order to assist in MakerDAO’s own vault portfolio management.

Interest rates

As a snapshot, earning rates on LINK across all platforms stands at a median 0.02%. The outlier in the group is Celsius network running a 1.98% earnings APR. At this point, due to lack of internal data, we cannot provide a clear reason for that rate disparity, other than potentially a special trading relationship with the Chainlink project. However, this assumption has not been verified.

On the borrowing side, Aave has both floating (variable) and fixed (stable) rate tranches for LINK. At time of writing, it runs a 3.25% stable and 0.55% floating, which is close to Nuo’s 2.3% borrowing APR. For the rest of the analysis, we will look mainly at Aave variable reserves, which may have more resemblance with MakerDAO vaults.

Sources: Loanscan.io, aave, alethio

Earn APR

Snapshot view of earn APR

protocol asset metric value change_24h
bzx link earn_apr 0.00 0.00
aave link earn_apr 0.02 -16.03
Nuo link earn_apr 0.10 0.00
Celsius link earn_apr 1.98 0.00
Argent link earn_apr 0.01 0.00

Sources: Loanscan, Alethio

Borrow APR

Snapshot view of borrow APR

protocol asset metric value change_24h
Nuo link borrow_apr 2.30 0.00
aave link borrow_apr 0.55 -8.78
bzx link borrow_apr 8.00 0.00

Sources: Loanscan, Aleth.io

Historical interest rates

Our timeseries runs from 9 Jan 2020, when Aave launched LINK in the protocol. Historically, the mean variable borrow rate has been 0.61% in the platform. However, low variable rates have mainly been sustained in the first two quarters of 2020. From July, with the acceleration of price increases for LINK and growing debt in the platform, Aave has adjusted its risk pricing to a 5.59% max value to rein in some of the downside risks of the growing collateral reserves.

Source: Aleth.io

Outstanding debt

Aave’s LINK debt levels have stayed relatively low for most of the year. Leveraging appetite has picked up from early July which coincided with the pace of price rises in LINK. However this appetite has since diminished to historical levels i.e. closer to the year mean of 359,292 LINK.

Unfortunately, at this stage we have not been able to source LINK liquidation data from aave yet. This may be updated in the near future to consider the extent of liquidations which occured over the last year and during the recent upswing of LINK asset price. Most importantly, it would be valuable to know whether the protocol borrowing rates have picked up not only due to increased leveraging but also in line with a risk-based pricing strategy in reaction to liquidations.

In the future, this sort of benchmarking observation of Aave liquidations may have the advantage of informing MakerDAO’s own ongoing risk policies and parameters. At least, this 3rd party liquidation data may be valuable to assist with management of future LINK vault events, while we may not have enough of our own data observations.

Source: Aleth.io

Dynamic borrow apr vs debt

The aave protocol appears to set a strict linear relationship between its outstanding debt and borrow apr rate for LINK. This reflects in the very dynamic nature of its floating borrowing rates that change in the system to adjust demand and supply for assets. At 0.98 historical positive correlation, both variables move in nearly perfect sync.

polynomial coeff: [1.42119299e-06 1.04068609e-01]

Source: Aleth.io

Summary of LINK stats on Aave

aave_earn_apr aave_borrow_apr aave_link_col_ratio aave_link_outst_debt aave_link_supp_vol
count 222.00 222.00 222.00 222.00
mean 0.12 0.62 5.45 360910.47
std 0.37 1.18 1.11 823042.34
min 0.00 0.00 3.59 0.00
25% 0.00 0.12 4.42 30110.54
50% 0.00 0.22 5.43 107627.86
75% 0.01 0.29 5.99 137784.71
max 2.28 5.59 9.68 4069828.26

Correlation of LINK variables on Aave


Summary of Noticeable Risks:

Below is a brief summary of the key risk factors identified by looking into the data available for this risk analysis:

  • Short term liquidity risk: The available supply of tokens held on major exchanges has been continuously reducing over the last 12 months. In the event of sharp asset price declines and subsequent asset liquidations, it is likely this may cause a short term liquidity problem for the asset while keeper auctions are running. This analysis has indicated the supply on exchanges has deteriorated despite the growing circulating supply throughout the period.
  • Trading volume concentration risk: Binance and Coinbase dominate trading volume for LINK pairs with 75% and 19.25% of all activity, respectively. Within these markets, LINK/USDT, LINK/BTC (Binance) and LINK/USD (Coinbase) account for 86.4% of all trades. In the rather unlikely event that these marketplaces face any technological issue or otherwise, liquidity and trade for LINK pairs would be directly impacted. At this point, no exchange platform, centralised or decentralised, has seriously challenged these two players.
  • Token holding concentration risk: In the LINK network, data indicates only 0.16% of addresses hold 100K tokens or more. This distribution is rather skewed towards already large token holders. Data also indicates that the distribution of token ownership has deteriorated in the last 12 months according to its calculated “proxy” gini index at 0.97, all that despite the increased circulating supply in the market.
  • Low DEX trading volume risk: Trading activity for LINK tokens within decentralised exchanges and liquidity pools has only taken pace in the last couple of months. However, the trading volume in these platforms is insignificant when compared to centralised exchanges such as Binance. Should centralised exchanges face any operational issue preventing trades, it is unlikely decentralised exchanges would be able to offset the trading activity, at least in the short term.
  • Competitive liquidity for collateral risk: This data analysis has indicated that a growing number of DeFi and CeFi lending platforms are adding LINK to their collateral portfolios. As benchmarking analysis has shown, Aave has reached a 4 million debt threshold collateralised in its platform in a relatively short period of time. This analysis has shown already existing LINK liquidity risks across exchanges. Should a few lending platforms have similar volumes as Aave, there is a real risk of “liquidity competition” across DeFi in case of simultaneous liquidations occurring in the industry.
  • Speculative behaviour risk: This study has indicated a continuously growing activity for Chainlink, using daily active addresses and new unique addresses as proxies for ecosystem adoption. However, this study has also shown zero balance addresses are growing in the network over the last few months. In usual settings, the growth of this metric indicates growth in purely speculative behaviour in token ownership.
  • Shared oracle infrastructure risk: There is a growing number of DeFi platforms embedding data sourced from Chainlink’s decentralised oracle network into their own platforms. While in theory this creates efficiencies of scale for the DeFi ecosystem, there is potentially a risk of increased interdependency of big platforms to the services provided by Chainlink oracles. However, this analysis does not have the data to measure that risk quantitatively at this point. Chainlink’s own terms of service covers some of these risk factors.
  • Circulating supply increase risk: The total supply of LINK tokens is capped at 1 billion according to ICO terms. However, this analysis has shown that the initial circulating supply, initially at 350 million LINK, has been increasing since July 2019. Addresses identified to belong to the team have been selling LINK into the market. As the analysis has not identified official documentation covering a vesting schedule for treasury tokens, it indicates a potential risk of sell offs in a short interval of time. In case of such events occurring, this could cause significant downward pressure in the asset price and by extent on vault liquidations.

Proposed Risk Parameters

Risk Premium: 2%
Liquidation Ratio: 175%
Debt Ceiling: 5 million
Auction Lot Size: 50,000
Minimum Bid Increment: 3%
Bid Duration: 6 hours
Max Auction Duration: 6 hours
Liquidation Penalty: 13%
Dust: 100 Dai

For parameter suggestion, the team has used the model described in the governance call here: a credit risk model that estimates the loss distribution of a portfolio of LINK Vaults. Risk parameters are estimated from this hypothetical loss distribution. Inputs to the model are derived from trading volume data along with stressed input parameters. A link to our model specification with inputs and outputs can be found here. Auction parameters have been selected to mirror those for ETH, with the exception of the lot size that has been normalized to a ~$50,000 amount

Beyond the actual model outputs, our analysis of LINK indicates this asset still has a set of noticeable risks (see summary of risks) which point to a cautious and progressive approach to parameter setting. Our proposal here is to collect live vault data which assists in determining whether model expected performance fits the actual performance of vaults. And from there, adjusting risk parameters (e.g debt ceiling, risk premium, liquidation ratio) to actual data.

However, despite simulating for a conservative 3 - 5m debt ceiling, we are conscious that the community may have other priorities such as a higher level of risk appetite for this asset. For this reason, the community input may affect the actual final recommendation for governance voting on LINK.

Lead Researcher: Will Remor


  • Nick Johnson Audit Report: Audit
  • Quantstamp Audit Report: Audit
  • SigmaPrime Audit Report: Audit

MakerDAO forum posts


Below is a list of commonly used primary resources for information and methodology.

Data sources consulted

  • santiment, coingecko, defiexplore, 0xtracker, uniswap, kyber, bancor, alethio, glassnode, coinmetrics, messari, loanscan, aave, intotheblock

Collateral Status Index
[LINK] Collateral Onboarding Application


This is some amount of risk analysis - no wonder LINK took a bit of time. Debt ceiling might be adjusted upwards gradually. Big thumbs up for all involved!


Assuming high LINK liquidity compared to other erc-20 tokens we assessed in the past and lower downside price risks described in this post, it seems that a higher debt ceiling instead of 5m could be proposed in combination with 175% liquidation ratio. However I don’t see any obstacles with the approach Will proposed, which is to test 5m debt ceiling first and proceed increasing it from there.


Shouldn’t we consider the fact that the peg has been horrifically broken for 6 months when setting risk parameters?

1 Like

What does the peg have to do with the underlying risk of the collateral?


That’s what base rate changes are for. Risk premium takes long term risk into account.


I mean the debt ceiling. Link is one of most liquid crypto assets, it should be one of the best tools to try to fix the peg IMO

If we only add 2x5 million debt ceilling per month we can never fix it, we need 100 million per month or more


Debt ceiling can be increased as soon as its hit. Better to start small and grow than start with huge empty ceiling that’s more prone to OSM risk.

1 Like

175% LR is too high
Aave LR is less than 145%

I agree that the initial DC should be higher. Maybe $20 million?

1 Like

This is an important point. Why not set LR to 150% like everything else? There’s no way that there have been a crash in the LINK price where that would make a difference, at that point it would be a jump to 0 like a technical failure, oracle attack or founder exit scam, in which case no matter how high you set the LR it won’t help.

1 Like

From my limited understanding I think CR is related to two factors mainly volatility & liquidity, even though it’s very liquid, I think is quite more volatile than say eth

1 Like

Ref liquidity this point to consider

To briefly comment on some of the questions relating to recommended risk parameters. As mentioned in the analysis, the approach taken for onboarding LINK is based on setting initial conservative parameters that are interactively monitored and adjusted upon intervals of time. The monitoring & adjustment of these parameters (e.g. debt ceiling, liquidation ratio, risk premium…) form the basis for future updated recommendations based on actual data and statistical reasoning.

As mentioned in the analysis, we understand that the community may have other considerations such as a higher risk appetite to take into account in the final governance voting.


Regarding the RP, would you keep it the same if ETH-A had a RP of 2% also?

1 Like

Yeah, though about that too. It would not be consistent with the eth analysis but then again it was made as a stress test analysis.

Perhaps a trade off between a more market fit CR with an increase in the SF although I’m not entirely sure any SF could protect us from a sudden drop in price together with a liquidation competition between protocols. In that sense a low SF with high CR gives us a head start.

1 Like

There are actually at least two possible answers to the above. First, it’s the internal relationship between variables within a same collateral type i.e. LINK. In the stress testing approach, the risk premium has got a certain level of relationship to other input parameters such as debt ceiling, but not uniquely DC. A number of variables influence each other in the simulation, not necessarily in a linear fashion by the way. Second consideration, which is closer to your question, is the external relationship between collateral types and their related risk premium output variables. In this case, the influence of one output (RP) from one collateral type (e.g. ETH) on another output (RP from LINK) requires a wider portfolio view in a sort of “multi-factor” simulation. It is certainly an interesting suggestion which requires a bit of investigation/further modelling to cater for those external relationship between modelled simulations.