Zou Chuanwei: Reinterpreting DeFi from Eight Key Perspectives and Logics
What direction will the DeFi ecosystem evolve towards? What logic does DeFi evolution follow?Author: Zuo Chuanwei, Chief Economist of Wanxiang Blockchain
Recently, there has been a lot of discussion about DeFi, which can be understood from multiple perspectives. This article proposes understanding DeFi from 8 key perspectives: 1. Financial functions; 2. Discrete time finance; 3. Trustless environment; 4. Oracles and information; 5. Liquidity; 6. Arbitrage; 7. Incentives; 8. Risks and interconnections.
This article attempts to answer two questions:
- What other tools should DeFi developers master besides writing smart contracts?
- From the perspective of DeFi investors, in what direction will the DeFi ecosystem evolve, and what logic does DeFi evolution follow?
I. Financial Functions
Some researchers understand DeFi by referencing mainstream financial institutions, such as discussing what forms banks, securities, and insurance should take in the DeFi space. This analogy is actually not precise enough, as DeFi is built according to financial function modules. This involves the relationship between financial institutions and financial functions.
Financial development has two main intertwined strands, like DNA double helixes: one is financial functions, and the other is financial institutions. Zvi Bodie and Robert Merton proposed six basic financial functions:
Payment and settlement. This is the most fundamental function of financial services to the real economy. Any economic activity relies on payment and settlement to form a value closed loop, aside from barter.
Resource aggregation and equity segmentation. This allows for the aggregation of social resources to undertake larger or more adventurous ventures and share profits and risks among participants. Joint-stock companies are a typical embodiment of this function. Banks gather scattered social funds to support national development, reflecting this function as well.
Transferring resources across time and space. Resources are never evenly distributed across time and space, requiring adjustments in two dimensions to maximize resource utilization efficiency. For example, if Zhang San has idle money that he cannot spend, and Li Si has a project but lacks funds, through the financial system, Zhang San can lend money to Li Si (transfer across space). Later, when Li Si's project generates profits, he can return part of the profits to Zhang San (transfer across time).
Managing risks. The allocation of resources in the financial system occurs in an uncertain environment. Any financial activity, whether centralized or decentralized, essentially involves managing risks to obtain returns.
Providing information. The financial system has an important price discovery function for funds and risks. The information provided by the financial system is significant for the allocation of resources across society.
Solving incentive problems. A large amount of value circulates through the financial system, and the economic incentives generated thereby drive the formation of social division of labor and market order.
Mainstream financial institutions generally perform multiple financial functions simultaneously. For example, banks mainly execute payment and settlement, resource aggregation, transferring resources across time and space, and managing risks.
Bodie and Merton believe that financial functions are more stable than financial institutions; the form of financial institutions depends on the financial functions they perform. For instance, the organizational structure and business forms of banks 100 years ago differ significantly from today, but the financial functions they perform have changed little.
Bodie and Merton's classification of financial functions is also applicable to DeFi (Table 1):
Therefore, understanding DeFi is best approached from the perspective of financial functions. DeFi is constructed according to financial function modules, allowing for good modularity. Multiple DeFi projects assembled together can achieve complex and diverse financial functions, approaching those of mainstream banks, securities, and insurance. However, even a DeFi combination differs significantly from these mainstream financial institutions.
II. Discrete Time Finance
Mainstream finance is continuous time finance. For example, global foreign exchange trading occurs 24/7, and Yu'e Bao generates interest income daily. The time units used in mainstream finance are generally hours, days, weeks, months, quarters, and years.
DeFi is discrete time finance. Any financial activity has cycles and frequencies. The frequency of DeFi depends on the update frequency of the distributed ledger of the public chain. The extension of time in a public chain is reflected in the continuous generation of blocks, with the time unit being the block time. The block time of a public chain is determined by the consensus algorithm run by validating nodes in a distributed network; although there is a statistical average, it is a random variable from a prior perspective.
Discrete time and TPS limitations have comprehensive and profound impacts on DeFi:
They affect the efficiency of DeFi activities; the volume of DeFi activities is inherently constrained by the physical performance of the public chain. For example, during chain congestion, on-chain auctions and on-chain collateral liquidation transactions may not be processed in a timely manner.
They affect the efficiency of information synchronization and arbitrage between on-chain and off-chain. This will be elaborated upon when discussing oracles and arbitrage later.
They affect the efficiency of price discovery and risk clearing. For instance, aside from long-tail crypto assets, most crypto asset price discovery occurs on centralized exchanges. DEXs follow the pricing of centralized exchanges rather than the other way around. Furthermore, during periods of market panic, public chains are prone to congestion, and some transactions that help clear market risks may not be processed by miners or may require higher fees or gas costs to be processed. This not only reduces the efficiency of market risk clearing and returning to equilibrium but also diminishes market participants' confidence in the orderly operation of the market, further amplifying market panic.
What kind of analytical methods should be adopted for discrete time finance? Continuous time finance is easier to analyze than discrete time finance because it can utilize mathematical tools such as calculus. This article proposes the following analytical method for DeFi:
Discrete time finance = Approximation of continuous time finance + Impact of public chain TPS and time lags
In other words, when analyzing DeFi, one should first distill the core financial issues, clarify them in continuous time, and then consider the impacts of public chain TPS and time lags. For example, the analysis of the arbitrage mechanisms in automated market makers and oracles can follow this methodology.
In discrete time finance, the time value of money still applies. For instance, 1 unit of crypto asset after staking for 1 year yields a total of 1.5 units, including principal and interest. Thus, today’s 1 unit of crypto asset is equivalent to 1.5 units of crypto asset one year later. This is the time value of money. The analysis of funding costs and investment returns in DeFi essentially revolves around analyzing the time value of money.
Although the block time is a random variable from a prior perspective, using block time as the time unit allows the introduction of mainstream finance's interest theory into DeFi. Basic concepts and tools such as present value, future value, discount factors, simple interest, compound interest, and no-arbitrage pricing are applicable to DeFi and demonstrate strong application value in many issues.
Yu'e Bao allows real-time querying of investment returns, but due to the limitations of public chain performance, DeFi cannot achieve this. In many DeFi applications, theoretically, each new block generates interest. However, if these newly generated interests are paid out immediately through on-chain transactions, it may incur high gas fees and potentially cause on-chain congestion. Therefore, often it is necessary to appropriately extend the interest payment cycle. For example, accumulating interest first rather than immediately transferring it to the relevant address through on-chain transactions, and then paying out the interest income in a lump sum after some time (e.g., when exiting staking). In such cases, it is even more necessary to introduce precise interest calculation methods.
Interest theory has important applications in staking and DeFi liquidity management. For instance, PoS mining pools need to provide liquidity while offering staking returns to investors. Especially in Ethereum 2.0, the staking lock-up period on the beacon chain is quite long. When investors participate in beacon chain staking, if they have liquidity needs, can they transfer their staking shares? This question is a good application of interest theory.
III. Trustless Environment
Regardless of the roles DeFi participants assume, they are essentially addresses within the public chain. The public chain is a trustless environment where addresses are essentially anonymous, lacking identity and reputation. This is a key difference between DeFi and mainstream finance. In mainstream finance, participants are individuals and institutions. When individuals or businesses apply for loans from banks, banks assess their willingness and ability to repay; when companies issue bonds for financing, rating agencies evaluate their credit ratings; when companies raise equity funds, investors assess their profit prospects. Such assessments occur continuously in mainstream finance but do not exist in DeFi.
The trustless environment is the foundation of DeFi's openness and permissionlessness. However, in a trustless environment, since addresses themselves cannot become credit entities, the fulfillment of financial contracts relies on over-collateralization and staking. For example, suppose address a needs to transfer 1 unit of crypto asset to address b in the future. To ensure the enforceability of this commitment, address a must provide more than 1 unit of crypto asset as collateral. How should we understand over-collateralization and staking?
1. Over-collateralization and staking are important channels through which public chains capture value from DeFi. Without this mechanism, the price interactions between public chains and DeFi could become disconnected.
Over-collateralization locks in liquidity, effectively converting the credit risk of an address into the liquidity risk of the collateral. In both DeFi and mainstream finance, risks do not disappear out of thin air; they often just change form.
Due to over-collateralization, the risk pricing efficiency of DeFi lending is very low. This is reflected in the fact that DeFi lending rates do not include risk premiums for borrowers and are unrelated to the borrower's creditworthiness. In fact, in a trustless environment, it is impossible to define or measure the credit of an address. One could say that DeFi lending is essentially mortgage lending rather than credit lending.
In MakerDAO, over-collateralization ensures that Dai (essentially the liabilities of the CDP) from different collateral debt positions (CDPs) has the same value connotation. Regardless of who initiates the CDP or what collateral is used, as long as the over-collateralization rules are followed, Dai is mutually equivalent.
Staking is a commitment mechanism for stakeholders. This is fully reflected in PoS consensus algorithms to address the Nothing at stake problem.
Financial activities cannot exist without trust. Trust reduces uncertainty about the future and is crucial for lowering transaction costs in financial activities. This holds true for both mainstream finance and DeFi. When we say that blockchain is trustless, it essentially means transforming trust in people and institutions into trust in algorithms and smart contracts, but it is still fundamentally trust. Introducing trust into a trustless environment helps further reduce DeFi transaction costs. There are three ways to introduce trust in DeFi.
Associating addresses with off-chain identities and reputations, such as Gitcoin Grants using GitHub accounts to combat multiple identity attacks and collusion attacks. The controllable anonymity of central bank digital currencies essentially relates anonymous addresses to real identities through KYC.
Repeated games within the public chain can curb opportunistic behavior and form on-chain reputations, such as ChainLink nodes. Repeated games make the long-term losses of opportunistic behavior exceed short-term gains, allowing anonymous addresses to adhere to the rules evolved from the game.
The "invisible hand" ------ arbitrage and economic incentives drive behavior aimed at maximizing interests. This is the basic logic of mechanism design. We do not need to know the values of DeFi participants, such as whether they keep their promises; as long as they are rational economic agents, we can infer their behavioral characteristics through economic analysis. Conversely, we can design arbitrage and economic incentives to encourage DeFi participants to exhibit the behavioral characteristics we desire. This will be discussed in detail later.
IV. Oracles and Information
There are two consensus mechanisms inside and outside the public chain. The first is consensus algorithms such as PoW and PoS, which form consensus on information native to the chain. The second is oracles, which form consensus on information external to the chain. Oracles are the foundation for synchronizing information and arbitrage between on-chain and off-chain. Regardless of the consensus mechanism, both imply a reduction in entropy (i.e., eliminating chaos) and require energy input (or cost consumption). The goal of oracle design is to minimize the ratio of error to cost.
Oracles have various design schemes but can generally be divided into two categories.
The first category of oracles is based on reputation and voting, represented by ChainLink. These oracles rely on multiple quote providers, selecting the average or median of multiple quotes as the oracle quote to control individual quote errors. These oracles also eliminate negligent and malicious quote providers through reputation mechanisms and repeated games.
The second category of oracles is based on trading and arbitrage, using arbitrage mechanisms to converge oracle quotes to market prices. This type of oracle's use of arbitrage mechanisms will be discussed in detail later.
From the perspective of communication engineering, regardless of the form adopted, DeFi oracles are essentially a sampling process with errors and delays. Oracles sample continuous signals from off-chain at discrete time points and then read the discrete signals into the public chain (Figure 1).
Figure 1: Oracles as Sampling Processes
The error of oracles relative to the original continuous signal consists of two parts (Figure 2). First, the error from the signal source. Various oracle schemes essentially aim to minimize this error. Second, the fluctuation of the original signal, which can be amplified by the sampling interval and the time taken to reach consensus. Both parameters are largely influenced by the performance of the public chain.
Figure 2: Decomposition of Oracle Error
Liquidity
Liquidity is a core issue in DeFi. The application and management of liquidity are reflected in many aspects of DeFi. For example, as discussed earlier, over-collateralization transforms the credit risk of an address into the liquidity risk of the collateral.
Liquidity reflects the possibility of selling assets at a reasonable price within a reasonable time. Under otherwise identical conditions, the longer the time, the more likely it is to sell assets at a reasonable price. However, in many situations, it is not possible to sell assets leisurely. Thus, the level of liquidity can significantly impact investors' interests.
Liquidity is a complex economic phenomenon influenced by many factors. For instance, the liquidity of on-chain transactions is limited by public chain TPS. Liquidity is also a product of the interaction between buyers and sellers. The higher the confidence of both parties, the greater the liquidity.
Liquidity is a special type of public good. For most goods, increased demand stimulates supply by pushing up prices. However, because liquidity is related to the confidence of buyers and sellers, supply may be minimal when it is most needed.
For investors, liquidity serves as a commitment and insurance mechanism. Liquidity provides investors with confidence about whether transactions can be completed and at what price.
The commitment mechanisms related to liquidity can be mainly divided into two categories. The first category is reputation-based commitments, such as market makers under order books. These market makers promise to facilitate transactions for buyers and sellers based on their own reputation and strength, profiting from the bid-ask spread. The second category is algorithm-based commitments, such as automated market makers. Automated market makers provide transaction facilitation based on algorithms through liquidity pools but bear the impermanent loss caused by arbitrage, facing challenges in commercial sustainability.
Clearly, under equal conditions, products that can provide liquidity are more attractive to DeFi investors. For instance, the earlier discussed PoS mining pools provide liquidity shares and transfer mechanisms for investors.
Liquidity can be unbounded or bounded. The liquidity of centralized exchanges is unbounded. The liquidity of automated market makers is bounded. For example, regardless of how investors trade with Uniswap, if transaction fees are not considered, the liquidity pool should always satisfy the constant product condition.
PoS mining pools providing liquidity shares to investors also represent bounded liquidity. Regardless of how liquidity shares are transferred among investors, the total amount of liquidity shares they hold remains unchanged ------ the transfer of liquidity shares does not constitute a redemption from the PoS mining pool.
This is similar to how trading in the secondary market of stocks does not affect the number of shares outstanding of a listed company. As Keynes pointed out, there is no liquidity of investment for the entire society.
Liquidity has aggregation effects. For multiple liquidity pools, the liquidity they aggregate together will exceed the sum of their individual liquidity. This, like the risk diversification effect, is a fundamental principle of finance ------ the risk of an asset portfolio is less than the sum of the risks of its parts.
Arbitrage
Arbitrage is not a bad thing. Arbitrage stems from a basic human need. When conditions allow, anyone wants to take advantage of others but does not want others to take advantage of themselves. This is human nature.
Financial development has various driving factors, such as regulation and technology, but the fundamental driving force is arbitrage. Whenever any financial market or product is first introduced, due to imperfect pricing mechanisms, there will always be arbitrage opportunities that attract arbitrageurs. Driven by arbitrageurs, the pricing mechanism is corrected, and financial markets and products tend to improve. This cycle continues, allowing financial development to persist.
Arbitrage leads to price convergence, but convergence requires time and cost. For example, for any oracle based on trading and arbitrage (such as Uniswap), it can be proven through solving optimization problems that there exists a no-arbitrage condition under which no new trades will occur. This no-arbitrage condition can be equivalently defined as a limit on the range of deviation of oracle quotes from market prices.
The economic intuition behind this is very understandable. Once oracle quotes deviate from market prices, it indicates an arbitrage opportunity, but executing an arbitrage strategy incurs costs.
Therefore, arbitrageurs carefully weigh the arbitrage gains against the costs and will only execute the arbitrage strategy when the magnitude of deviation of oracle quotes from market prices is sufficiently large (this is essentially the optimal exercise timing problem of American options). The execution of arbitrage strategies will correct the deviation of oracle quotes from market prices until the arbitrage strategy is no longer economically attractive, and this cycle continues.
It can also be proven that the lower the transaction costs in DeFi, the higher the arbitrage efficiency, and the smaller the deviation of oracle quotes (i.e., reducing the error of the oracle at the information source).
No-arbitrage pricing + interest theory is the basic tool for asset pricing in DeFi. Arbitrage will form the interest rate benchmark curve in DeFi. For instance, for PoS-type crypto assets, staking yields will constitute the "anchor" for DeFi lending rates.
Arbitrage is a universally applicable mechanism design. Arbitrage makes the fewest assumptions about human nature. Arbitrage only requires that people are rational, seek benefits, avoid harm, and maximize their own interests, without needing to know whether they are good or bad people.
In the decentralized environment of DeFi, the incentive and coordination effects of arbitrage become even more significant. For example, the liquidation of collateral in MakerDAO is based on arbitrage design.
The premise for the arbitrage mechanism to function is the existence of an active community of arbitrageurs, so community incentives are crucial. For instance, in oracles based on trading and arbitrage, if there is only one arbitrageur, they will wait until the deviation of oracle quotes from market prices is very large before executing their arbitrage strategy.
If there are multiple arbitrageurs, each will consider the possibility that other arbitrageurs will execute their strategies before them. The competition among arbitrageurs will lead to earlier execution of arbitrage strategies, thereby reducing the deviation of oracle quotes from market prices.
Regardless of the form of the arbitrage mechanism, arbitrage is essentially a zero-sum game, a redistribution of interests, where what one party gains is what another party loses. For example, in automated market makers, the profits of arbitrageurs correspond to the impermanent losses of liquidity providers.
Some automated market makers introduce oracle quotes, which essentially limit the arbitrage space and reduce the impermanent losses for liquidity providers. Clearly, in any arbitrage mechanism, if one party continues to incur losses, the arbitrage game cannot persist, and there will always be a day when it stops.
Incentives
The design of incentive mechanisms should make DeFi an infinite game, rather than a finite game. Community self-organization and self-upgrading are key to the evolution of DeFi. Community members should all be able to derive their own benefits from DeFi. In other words, in the design of DeFi incentive mechanisms, one cannot expect a certain type of participant to always play the role of a "living Lei Feng."
As mentioned earlier, in automated market makers, the profits of arbitrageurs correspond to the impermanent losses of liquidity providers. To compensate for the losses faced by liquidity providers, automated market makers charge transaction fees from investors and transfer the fee income to liquidity providers.
However, due to limited on-chain transaction volumes and low fee standards, the issue of " fee income < impermanent loss" is a common problem, meaning liquidity providers effectively provide this public good at their own expense.
Some automated market makers introduce governance tokens as additional compensation for liquidity providers to alleviate the commercial sustainability issues they face. However, the value-capturing ability of governance tokens is weak. For instance, in the valuation of company stocks, future profits and dividends are generally estimated, followed by estimating the future cash flows of stockholders, which are then discounted to obtain the stock valuation.
If an investor's equity stake does not reach critical points such as 33% or 50%, the value of the corresponding voting rights is generally not considered. Therefore, the effectiveness of governance tokens in automated market makers remains to be further observed. Future directions should focus on improving the fee structures and governance token designs in automated market makers.
These issues are prevalent in the blockchain field. For example, if block rewards are not considered, or if block rewards drop to very low levels, can the transaction fees earned by PoW miners cover their mining costs?
Similarly, can the fees paid by users when calling oracle quotes cover the costs of the oracles? These questions essentially involve the provision and financing of public goods in a decentralized environment. Solutions to these issues should reference economic theories regarding public goods.
V. Risks and Interconnections
The core of DeFi is operational risk, which mainly includes market risk, liquidity risk, technological risk, and credit risk. Market risk arises from the volatility of crypto asset prices. In DeFi, the widespread application of over-collateralization and staking transforms the credit risk of addresses into the liquidity risk of collateral, so credit risk is not as prominent as in mainstream finance (where banks and corporate bond markets primarily deal with credit risk).
The technological risks in DeFi are much more pronounced than in mainstream finance, potentially arising from vulnerabilities in smart contracts or limitations in public chain TPS.
Various DeFi activities essentially involve taking on risks to maximize returns. Risks can be transferred, shared, hedged, converted, and diversified, but they will never disappear.
DeFi projects are constructed around financial function modules, exhibiting modularity. DeFi projects are interconnected and combined through channels such as information, funds, and risks. This helps develop the DeFi ecosystem " from points to surfaces," but it can easily lack overall planning and accumulate risks. Particularly, the more foundational projects in the DeFi ecosystem, while having a stronger "moat" position, are more likely to introduce single point failure risks.
Many researchers categorize the DeFi ecosystem based on different business types, but it is more necessary to depict an overall "risk map" for the DeFi ecosystem. In the future, DeFi projects should conduct audits not only for smart contracts but also for financial risks before going live.
