The Crypto Loan Economy
Overcollateralized lending is a popular form of loan in crypto. It is a great option for individuals who need exposure to new assets without selling their current position. Here, a question arises: in a world plagued by income disparity, how does overcollateralization fulfill the very purpose of taking a loan? Besides the issue of issue of capital efficiency, overcollateralization prevents one from accessing cheap capital to achieve their financial goals. Meanwhile in a trustless financial system, undercollateralization is only going to make your liquidity providers poor. So what’s the solution? This research article is a collaborative effort of Jack and 0xlol for Polygon, who can be reached on twitter at Jack and 0xlol.
Humans have been involved in the lending-borrowing dynamic from time immemorial. Earliest examples can be traced back to 2000 BCE Mesopotamia, where farmers borrowed seeds and animals to generate agricultural yield. Such social dynamics introduced the complexity of trust and risk, but opened up the horizon for a quantum leap in economic growth. Those with wealth found a new mechanism for earning yield on their assets and those who lacked wealth gained the ability to borrow it and produce their own wealth. Simple, but nothing short of genius!
Leverage is the cornerstone of entrepreneurship and creation, allowing individuals to take on risk and develop new products and services. In fact, excluding the value of outstanding loans, the modern global lending market is currently valued between $7–8 trillion, based on the money made on interest by all private and public institutions.
Moreover, peer-to-peer lending — which DeFi aims to pair with institutional lending in order to kickstart exponential growth — is currently primed for rapid expansion. While institutions now have a firm grip on lending, enabling greater access to individual lenders through DeFi can help provide higher yields on one’s assets by eliminating the need for middle men. In fact, by 2025, the peer-to-peer lending market is expected to reach $1 trillion in value.
This rapidly changing environment is being eyed by many DeFi protocols. They gaze at an extremely profitable market, as our financial system has started rolling over to transparent ledgers i.e. blockchains. While DeFi presents an option for a decentralized and trustless financial future, undercollateralized (UC) lending is the current bottleneck in accelerating DeFi adoption.
TradFi or DeFi?
While traditional finance (TradFi) systems have abundant access to credit scores, KYC documents, and legal protections to recover funds, they suffer from various issues. The issues don’t stop there. Keeping your money with the traditional banks imply that you don’t have custody of your assets. There have been multiple instances where governments and banks have used these assets to bail out ‘friendly frauds.’ Following the 2008 mortgage crisis, major banks were bailed out by the US government, shifting the burden onto taxpayers.
Even with laws preventing race-based discrimination for loans, a 2019 report that examined two million US conventional mortgage applications found that black individuals were 80% more likely to have their applications rejected compared to financially comparable white applicants, largely due to biased algorithms. DeFi presents an opportunity in certain cases to reduce identity based discrimination from lending markets by relying entirely on smart contracts and on-chain identities.
Through governance systems majorly run by the stakeholders DeFi lending markets present a new era of possibilities in which only those who choose to deposit liquidity into the protocols can be on the hook for poor management of funds. It prevents mighty governments from thrusting tax burdens onto those who never participated in bad business practices, such as the sale of subprime loans. In TradFi, if you keep your money in a bank, you earn single-digit annual interest and have no control over whosoever the bank writes loans to. DeFi lets the lenders have a say in its administration, a true form of financial sovereignty!
Most importantly, 31% of adults world-wide are unbanked, while 91% of the world’s population has access to a smartphone. Requiring just an internet connection, DeFi can help unbanked and underserved communities access to previously unreachable financial services. If done correctly, UC loans can give such individuals financial leverage which can change their lives for good.
Beginnings of Lending in DeFi
Decentralized lending first began in its simplest form: overcollateralized lending. Many popular platforms such as AAVE, Compound and Maker pioneered this role, where users deposited their crypto assets as collateral and borrowed an amount that is less than the value of their deposit. Take for example DAI on AAVE, it has a max loan to value (LTV) ratio at 77%, meaning for every $100 deposited in DAI, a user can borrow up to $77 in assets.
On the other hand, UC lending can be either partially collateralized or totally uncollateralized. In a partially collateralized lending system, an individual deposits $x to borrow y($x) worth of assets (y > 1), while in an uncollateralized lending system one could borrow $z with no collateral at all. UC lending is also referred to as unsecured lending while overcollateralized lending is also called secured lending.
We analysed 7 such protocols w.r.t their risk assessment modules, lending mechanisms, total outstanding debt etc. Also, it should be noted that for this article, we aren’t considering flash loans or “one block liquidity” as UC loans. Before diving into the details, let’s first try to understand the different purposes these protocols serve:
Maple Finance, TrueFi, & Clearpool: UC lending protocols for Institutional Borrowers.
DebtDAO: Revolving line of credit for growth- stage DAOs and protocols taken out against future revenue. Offers a new way for DAOs and protocols to raise funds without diluting or releasing a native token.
Goldfinch Finance: Credit lines for real world businesses to exchange for fiat.
Gearbox: 0% funding rate for up to 4x leverage on whitelisted protocols.
Risk Assessment Methods
When dealing with UC loans in a financial sector based on anonymity and trustlessness, risk is the primary concern. Nobody wants to give a loan to a random twitter anon in the hopes of getting repaid while the borrower rides away into the sunset with a fresh lump of cash. So, what actually prevents someone from stealing away all the borrowed funds in real life, but not on-chain? Well, banks and other lending institutes most probably know their whereabouts, can impact their credit score, and cause a lot of legal as well as financial trouble.
Each on-chain protocol tackles this problem with their own custom risk assessment method. The varying methods for risk assessment exist due to the different purposes each protocol serves, such as varying targeted customer bases and levels of control over loaned funds.
Maple Finance, TrueFi, and Clearpool all have similar processes before the loan approval stage. First, each institution gets whitelisted by the individual protocol through an application process where they fill out KYC/KYB (know your customer/business) forms, provide financial documentation, and pass a protocol performed credit score assessment. The credit assessments slightly vary from protocol to protocol, but essentially analyze the business risk and reliability of the applicants. While the initial whitelisting stages are all quite similar in the loan application process, the differences between the institutional lending protocols occur in the specific loan approval process. These are discussed below, alongside the risk assessment methods used by non-institutional lending protocols.
Pool Delegation: Maple Finance, DebtDAO (Yet to Release)
Maple Finance and DebtDAO rely on professional credit analysts that stake the native governance tokens ($MPL and $DEBT) in order to assess and approve loans. However, Maple Finance allows the pool delegates to negotiate the terms of the loan with the borrowers, such as interest rate, date of maturity, and collateral ratios. DebtDAO risk analysts simply approve or deny the loan request that is given to them. It is non-negotiable. These pool delegates are rewarded for their work with portions of the interest earned from borrowers.
Community Vote: TrueFi
Similarly, TrueFi adopts a community voting layer atop the credit risk team. The credit risk team initially sets all the terms for each loan, but the borrower cannot simply accept these terms to receive funding. In order to be funded, the loan must be approved by individuals who stake the native token $TRU for the governance token $stkTRU. Each loan request requires an 80%+ approval rate in order to open the lending pool for deposits. The voting procedure essentially acts as a veto power for the community, given that only two loans have not met the approval threshold.
True risk assessment comes from the professional credit analyst team, but this community engagement allows for greater democratization of the protocol.
Third-party Outside Organization: Clearpool, Goldfinch
Clearpool Finance requires whitelisted institutions to stake at least 500,000 $CPOOL in order to be eligible for a credit assessment. This protocol mentions a different staking requirement. It requires the institution (borrower) to stake governance tokens, rather than the credit risk analyst. The staked amount acts as a sign of good faith from the borrower that outsourcing the credit analysis to its third-party partner, X-Margin, will be worth it. Following approval, the institution’s lending pool is open to receive funds.
Goldfinch also requires businesses to stake its native token, $GFI. Rather than a strict nominal amount, one needs to stake an amount of $GFI equal to 2x the cost of an audit. After which the protocol outsources the credit check of financial statements and identity legitimacy to its third party partners, Persona or Parallel Markets. After a successful audit, each business receives a non-transferable Unique Identity (UID) NFT linked to its specified wallet to avoid sybil attacks. Following the whitelist approval, a 20% threshold must be met in the first-loss capital pool. Then 6 out of 9 randomly selected auditors (individuals who stake $GFI for auditor voting rights) need to approve the loan for the borrower to access capital.
Predetermined Smart Contracts: Gearbox
Gearbox is a totally permissionless protocol that relies on its credit accounts. They are preset smart contracts for each pool type that bind together lenders and borrowers, curated by individuals known as credit managers. The credit accounts hold both the user’s deposited collateral and the borrowed funds. Borrowed funds can only be used in whitelisted protocols and/or swapped for specifically allowed tokens to avoid extreme volatility. Since users deposited funds can be liquidated by smart contracts at a determined risk level, there is no need for a credit assessment.
Though Gearbox might appear to be an interesting protocol, it is not the true form of UC lending that can enable loans for everyday life activities such as car loans and mortgages that TradFi currently has a choke-hold on.
Off-Chain Integration: Teller
Teller gets its own category since it’s not really in the same ballpark as the others. Teller creates standalone products for specific cases that act as marketplaces to connect borrowers and lenders. Such products can be accessed once the proposed borrowers have gone through Teller’s proprietary credit risk algorithms. In fact, USDC.homes on Polygon relies on Teller’s credit assessment before funding a loan from TrueFi. Teller’s algorithmic credit risk protocol computes credit and banking data off-chain privately and securely. It is used to generate individual loan terms for users that are routed on-chain. Inputs involve collateral provided, risk premium, loan-to-bank balance, loan-to-monthly net income, and loan-to-monthly income.
While the risk assessment models provide the initial wall of protection for lenders in UC lending protocols, a ‘first-line of defense’ system offers LPs extra protection from loan defaults by using other stakeholders’ liquidity as the “first-loss capital.”
Stakers as First-loss: Maple Finance, TrueFi, Goldfinch
Maple Finance requires that pool delegates stake 8–12% of the pool’s loan balance as $MPL and $USDC in a 50:50 Balancer pool. In this way pool delegate incentives are aligned with LPs. Also, the funds can be easily liquidated for $USDC in the event of default. Aside from pool delegates, anyone can add to the reserve pool to increase the first line of defense to earn a share of 1/10th of the pool’s interest fees and $MPL rewards. Similarly, Goldfinch relies on backers, who evaluate the borrowers, as the first line of defense. Backers deposit into a “junior pool” for higher nominal interest rates, as they get a 20% share of the interest earned by general LPs in the “senior pool” for taking on the extra risk.
TrueFi relies on the participants who stake $TRU for $stkTRU for voting rights, $TRU emissions and protocol fees as the first line of defense. In the event of a default, up to 10% of $stkTRU may be liquidated to repay the loans. Importantly, there is a cooling down period for staking $stkTRU: it cannot be withdrawn immediately.
Insurance Pool as First Loss: Clearpool
In every block, a percentage of each pool’s interest is diverted to an insurance pool which acts as a claimant pool for LPs if the borrower defaults on the loan. The current governance vote has it set to 5% of a pool’s earned interest.
Spigot Function: DebtDAO
DebtDAO has developed a “spigot function,” which controls the borrower’s revenue generating contracts, and escrows a fixed percentage of the revenue while diverting the rest to the borrower’s treasury. In the case of a default, the spigot can claim the escrowed assets to help pay back interest and principal payments, acting as a debt collector. As arguably the only true form of debt collection in DeFi, this spigot function can be extrapolated for use in other protocols. Its only limitation is in its applicability to protocols and DAOs. If an individual is aware of spigot function claiming any funds sent to their wallet, they can create a new wallet to avoid the penalties.
Tokens Available for Loans and Pool Risk Variety
Aside from risk assessment methods, the types of tokens offered for loans is also an important differentiating factor. While individuals can use UniSwap to exchange the loaned assets for a preferred asset not offered by lending protocols, users may want to avoid facing slippage, transaction fees, and added smart contract risk. The aforementioned protocols offer the popular and dependable USDC, given that it is considered to be the safest stablecoin within the DeFi ecosystem. TrueFi offers UDT, BUSD, and TUSD in addition to USDC, while Maple offers USDC and WETH. Gearbox has the most variety, offering DAI, USDC, WETH, and WBTC for loans since they have the ability to restrict the use of funds to only whitelisted protocols.
Another added functionality of UC lending protocols is the offering of multiple lending pools for the same asset. A great advantage of having multiple pools of the same asset is the ability to have pools with varying levels of risk for liquidity providers. Maple Finance, Clearpool, and Goldfinch offer multiple pools per asset type based on the specified borrowers, while TrueFi, DebtDAO, and Gearbox offer just one single pool per asset type. For example, on Maple Finance users can deposit USDC into either Maven 11 Capital’s pool for 11.3% APY or Alameda Research’s pool for 10.5%, with the difference in APY being due to the delta in estimated risk determined for each institution.
Potential Limits for Undercollateralized Loans in DeFi
While all these on-chain UC lending concepts are exciting, and are pushing the entire DeFi space forward, it is important to discuss their potential drawbacks. First, scalability is a core tenet for the growth potential of such platforms. While the concept of pool managers work well for protocols like Maple Finance, these only succeed as they are limited to institutional borrowers. If such protocols plan to enable the retail’s participation, then the pool-delegation bandwidth required to review and assess borrowers will be enormous. TrueFi has a similar issue: It requires each loan request to pass a vote from $TRUE stakers. What happens when you open the gates to millions of retail borrowers? This is the reason why the loans are only available to institutions. While both protocols lack the design for massive scalability, Maple’s approach to using pool delegates is superior to TrueFi’s community voting mechanism, because it allows for more informed decision making. In reality, community voters with distributed responsibility won’t give the same attention to detail as a single person in charge of one pool.
Moving away from institutional borrowers, on-chain ‘algorithmically determined’ credit scores allow for ease of scalability to the masses, but are relatively weak at the moment due to the limited on-chain data availability. Also, on-chain credit scores, if not connected to a single real-world identity might be susceptible to sybil attacks. In context of an on-chain UC lending, a random individual can slowly build up good credit with a particular account only to borrow a large sum with no plans of paying it back. What happens when someone iterates this process through multiple wallet addresses? Individual’s KYC connected to a single wallet can help avert this disaster-in-the-making. However, for security and personal reasons, individuals often have multiple wallets with assets distributed amongst them. This essentially makes the concept of ‘KYC attached single wallets’ a limiting factor leading to gross inefficiency.
The off-chain credit integration system is relatively scalable and existing institutions seem interested, as Teller has had strong partnerships thus far. Within 1–2 years, one can expect to see people receiving loans on-chain without even realizing the presence of Teller in the back-end.
Additionally, individuals may not want their KYC details kept on-chain for privacy reasons, even if cryptographically encrypted. Along these lines, off-chain credit scores based on the lines of traditional finance not only go against the ethos of DeFi, but also introduce the risk of keeping KYC documents and personal information in a risky off-chain system.
What do the Numbers Say?
Most of the DeFi protocols today are compared based on their TVL. When it comes to lending protocols, it’s surely not the gospel’s truth. Lending protocols, specially those with UC loans usually have their capital’s net-flow outwards rather than inwards. At the same time, this capital should generate a regular income. To understand these aspects better, we take into consideration four different factors (apart from TVL) — total loans originated (TLO), total outstanding debt (TOD), net interest accrued from borrowers (NIA) and treasury or protocol’s revenue (TR).
For a lending protocol, it’s very necessary that capital from your lenders or LPs doesn’t sit idle. It should generate income. So, what do you do? You lend it out to those in the need. All this capital lent to borrowers, cumulatively accounts for protocol’s TLO. This amount is later paid back by borrowers with interest. Usually, a small part of this interest along with a ‘fee’ charged from different parties involved in the process goes to the protocol’s treasury. The rest is paid back to the lenders. The capital which hasn’t been repaid yet accounts for protocol’s TOD.
If we look at the data, it mostly goes in favor of Maple and TrueFi. When compared to the whole UC lending ecosystem through their TLO, NIA and TR, they seem to be establishing themselves as the leaders. They can also be said to have a first mover advantage in the institutional lending space. Goldfinch seems to be picking up the pace with around $120M in TLO, and around $3M and $0.7M in NIA and TR respectively.
Clearpool has effectively differentiated itself within the space by offering variable interest rates rather than the fixed interest rates of Maple and TrueFi. This important yet subtle change primes Clearpool for dominating an untouched sub-sect of undercollateralized lending. As a younger protocol in its growth phase, it is primed to realize its massive potential alongside the growth of crypto UC lending as a whole.
Also, data on all the aforementioned parameters for Teller couldn’t be found as it’s architecture is fundamentally different from other UC lending protocols.
An Optimistic Future
As DeFi continues to proliferate into the mainstream, on-chain credit scores paired with KYC or off-chain credit score integration (for ex. Teller) for retail investors shall push the migration of personal loans from an off-chain manipulated ecosystem to fair on-chain protocols.
More than just credit assessments, legal protections will be pivotal for creating trust between lenders and borrowers. Protocols like TrueFi which have legal agreements with their institutional borrowers shall be able to enforce the loans and recoup any defaults if necessary. Applying this framework to personal loans, on-chain signatures through protocols like EthSign can be implemented to further instill faith in retail UC lending.
While still nascent, on-chain UC lending protocols with their rapid rise and current success have laid solid foundations in this space. It opens up the doors for similar protocols looking to enter DeFi. One can expect further expansion of such concepts into untapped consumer bases. Each and every innovation in this space brings DeFi one step closer to flipping TradFi. Maybe one day, the fortress of Bastille shall fall!