The Fractional Reserve Stablecoin: The Natural Evolution of Modern Banking
Introducing InfiniFi, the first fractional reserve stablecoin to secure greater returns without increased risk by better addressing banking’s duration gap problem.
For the vast majority of history, the world’s financial capital has rested on a flawed foundation. Today, nearly all modern government treasuries and private financial institutions rely on fractional reserve banking. Banks use past behavior to predict how much of their customers’ deposits to keep on hand while lending out the rest to secure greater returns. This can lead to greater capital efficiency and economic expansion in the regions they serve. However, fractional reserve banking also creates duration gaps — a mismatch between the average maturity of a bank’s liquid liabilities and illiquid assets — which can lead to insolvency if enough depositors attempt to withdraw their money at once.
To account for those duration gaps, banks must efficiently manage their balance sheets, effectively model cash outflows, and appropriately handle coordination failures that could lead to bank runs. However, they have imperfect solutions for doing so, relying on reactive lookback models and centralized planning models that misalign incentives between them and their depositors. This leads to increasing threats to the integrity of the global financial system, as witnessed in 2023, the biggest year ever for bank failures, with five banks managing a record $548.7 billion collapsing (led by First Republic Bank, Silicon Valley Bank, and Signature Bank, which represented the second, third, and fourth-largest bank collapses ever).
We propose InfiniFi, a fractional-reserve stablecoin powered by a self-coordinated, duration-matching autonomous balance sheet to directly measure market sentiment and address the problems posed by duration gaps. This depositor-directed system decentralizes how assets are allocated, giving individual depositors choices based on their specific risk, duration, and liquidity preferences. This allows InfiniFi to create more proactive models that better forecast depositor behavior and reduce the risk of coordination failures and bank runs. By better addressing the duration gap, InfiniFi more efficiently deploys capital, allowing it to generate higher returns for any existing asset without increased risk.
As such, InfiniFi represents a generational opportunity to move our shaky financial system to steadier ground, while also providing the first major incentive to tokenize all assets on blockchain technology.
Fractional reserve banking has existed for centuries. It relies on the fact that depositors do not generally ask for all of their money back at once. That means shrewd bankers can lend out a significant portion of those deposits and generate high returns, so long as they keep enough cash on hand to cover expected withdrawals. Rather than deploying all of their cash deposits into liquid assets that can immediately be redeemed for cash, such as treasury bills, banks that use fractional reserve deploy a portion of deposits (typically the majority) into illiquid assets, such as loans, that cannot immediately redeemed but pay higher returns to compensate for the increased risk they represent. This allows banks to increase their returns above what would otherwise be achievable in fully liquid assets.
This increase in capital efficiency comes at a cost. The vast majority of depositor obligations held by banks (their liabilities) are fully liquid (90%) and may be redeemed at any point in time. However, the illiquid assets held by banks have a duration associated with them that must pass before they reach maturity. The mismatch between these zero-duration liabilities and these positive-duration assets is termed the “duration-gap” and serves as the core problem that most banking infrastructure has been built to address. Since banks do not have cash on-hand to cover depositor obligations, they must design and operate systems that manage this shortcoming.
Any entity utilizing a fractional-reserve system must determine how to select assets to ensure that daily cash inflows are sufficient to match projected outflows, how to model and predict what those projected outflows will be, and how to respond when outflows far exceed inflows and a bank run begins. For traditional banks, these problems are addressed with hierarchical balance-sheet management, look-back models trained on past data, and a government-sponsored insurance framework. This traditional approach is as good as the tools of the time have allowed but is neither efficient nor sufficient.
To ensure that the existence of the duration gap does not result in the insolvency of the bank, institutions utilizing fractional reserve systems must determine which assets to select such that daily cash inflows from those assets reaching maturity are roughly equivalent to daily cash outflows. This process of selecting illiquid assets to reach maturity at a certain rate over a given period of time is known as “laddering,” and serves as an important and labor intensive task for any bank.This task of balance-sheet management necessitates an organized approach to address it, and to this end, banks use a top-down hierarchy. Upper management breaks the task of laddering into sub-tasks, then hires employees to perform these sub-tasks, rewarding and promoting the employees whose decisions result in better outcomes for the bank.
This is an efficient and time-tested approach to solving complex problems such as laddering, but unfortunately, the top-down approach that banks utilize to address laddering results in the creation of perverse incentives. As employees are rewarded for producing the best outcomes for their employers, they are encouraged to pursue goals that will provide the best outcomes for the bank, rather than the depositors. This incentive misalignment encourages bankers to pursue balance-sheet management strategies optimized purely for high returns, which directly make the bank more money, rather than optimizing for a balance of risk and reward. With high return strategies comes higher risk, but if the strategies which bankers are incentivized to pursue result in loss of depositor capital and bank failure, the penalty to any individual banker is negligible, as the Global Financial Crisis highlighted (2008). While many individual bankers doubtlessly find the incentives their system creates concerning, those who might otherwise pursue more conservative strategies are put in a situation where they must choose between pursuing higher risk strategies, or hamstringing their own careers. Banks are not fundamentally evil, they simply are the victims of a system design which rewards it.
To ensure that assets match liabilities on a day-to-day basis, banks rely on control systems to measure past depositor behavior and establish margins within which the bank can continue to operate. Initially, this began as simply keeping a sufficient amount of cash-on-hand, or reserves, to buffer bank runs that might occur. As time has progressed, banks have evolved their balance sheet management to measure average inflows and outflows of deposits, laddering illiquid assets to match their projected capital outflows. The process of projecting these cash outflows began with simple averages, but as statistics and finance have evolved, has become an advanced process.
In the present-day, banks utilize algorithms to measure holistic depositor behavior over time, with the most advanced systems utilizing AI, data science, and Machine Learning models to augment the bank’s risk management operations. However, all of these models suffer from a common flaw: they largely rely on look-back models.
Look-back models use historic knowledge to predict what future events will occur. The assumptions that go into these models are that if something has happened, it will likely happen again, and if something has not happened, then it is unlikely to happen. They look to the past to predict the future, but the black-swan events that often precipitate a bank failure are by their nature unpredictable, and fundamentally break the assumptions that go into models of this design. While it is possible for banks to measure what the market is likely to do, current banking models do not permit depositors to tell banks exactly what they will do. As such, banks relying on lookback models are forced to be reactive rather than proactive, limiting their ability to predict cash flows effectively and respond to changing events in real-time.
As a consequence of the duration gap existing, and bank balance-sheet assets not being immediately redeemable 1:1 for cash, depositors cannot all immediately retrieve their money from a fractional reserve system. It is this duration gap that allows for bank runs to occur and that gives rise to the coordination problem of fractional reserve banking.