Blockchain and Financial System Design: Course Summary

This article outlines the course "Blockchain and the Design of Financial Systems" (14.129), which blends computer science and economics to explore the impact of technologies like cryptocurrency, blockchain, tokenization, and computational algorithms on financial systems. The course aims to provide a balanced perspective, moving beyond hype and ideological positions to focus on practical applications and policy design.

Course Overview: Synthesis of Computer Science and Economics

The course integrates concepts from computer science, such as cryptocurrency, blockchain, tokenization, platforms, and computational algorithms, with economic theories including contract theory, mechanism design, general equilibrium theory, and monetary theory. The primary goal is to understand the assumptions and shortcomings of distributed ledgers, smart contracts, and encryption, and their potential impact on existing financial systems, while exploring new financial designs and regulatory implications.

Each lecture or cluster of lectures will feature one or more computer science technologies applied to historical or contemporary financial scenarios, utilizing various economic tools. Examples include community currencies, coordination in financial crashes, tokenized assets, multilateral payment systems, liquidity injections, risk-sharing auctions, and online markets. The course emphasizes unpacking these technologies and their assumptions, sometimes rearranging conventional bundles and labels to foster objective understanding.

The course encourages an open-minded approach to implementation, recognizing that solutions can involve standard packages (e.g., encryption with automated execution on a distributed ledger), computer code without consensus algorithms, or even trusted third parties and escrow accounts. This flexibility allows for a balanced perspective, avoiding hype and ideological stances, and focusing on practical use cases for policy designers.

Lecture Breakdown

Lecture 1: Introduction to Blockchain (Today)

This introductory session sets the stage for the course.

Lecture 2: A Unified View of Distributed Ledgers and Financial Accounts

This lecture treats both blockchains and financial accounts as databases of transactions. While each is associated with a notion of money (e.g., Bitcoin for blockchain, fiat money for financial accounts), the course emphasizes that both are broader concepts not solely defined by money.

Key points: - Databases of Transactions: Both blockchains and financial accounts are fundamentally databases. - Notions of Money: Blockchains are often linked to cryptocurrencies like Bitcoin, while financial accounts relate to fiat money or demand deposits. - Broader Scope: Both distributed ledgers and financial accounts can encompass more than just money, including goods, assets, and certificates that frequently appear in exchange. - Contracts on Blockchain: The concept of a balance and its changes in a transaction can be generalized to state and state changes, allowing for the writing and execution of contracts on the blockchain. - Individual vs. Community Perspective: Financial accounts often focus on individual assets, leading to tensions between storing value and using it in transactions. Historically, double-entry bookkeeping, as conceived by Pacioli, embraced a community perspective where money was seen as a liability, an obligation to spend for community objectives. - CAP Theorem: The course will introduce the CAP theorem (Consistency, Availability, Partition tolerance), an impossibility theorem in computer science, and apply this trilemma to interpret economic systems.

Readings for this lecture include: - Agustin Carstens, "The Future Monetary System—Vision to Reality" (BIS) - "The Baby-Sitting Economy" (reviewed by Paul Krugman) - Goldstein et al., "Payments, Reserves, and Financial Fragility" (on real-time gross settlement) - De Meijer, an article on cryptocurrency use in Kenya for targeting poor households.

Lecture 3: Fragmented Markets, Policy Objectives, and Distributed Ledgers

This lecture explores how distributed ledgers can address problems arising from fragmented markets. It introduces friction into markets to analyze emerging problems and potential remedies, including regulatory solutions and distributed ledgers as a technological solution to information problems.

Key points: - Pareto Optimality: Economists use Pareto optimality as a criterion to assess efficiency, aiming to make some participants better off without harming others. - Market Fragmentation: The US national market system attempts to equate prices across fragmented markets to prevent unfair pricing. The blockchain perspective, however, considers artificially fragmented markets and seeks to achieve competitive outcomes with agents constrained to pairwise trading. - Decentralized Problem Solving: In fragmented markets, even with correct prices, agents generally cannot solve problems in a decentralized way using only their own histories. A common, distributed ledger is often necessary. - Liquidity Requirements: Decentralized monetary systems require significant liquidity for agents to achieve efficient allocations. Real-time gross settlement systems, despite their name, often require liquidity-saving mechanisms. - Warehouse Facilities: Centralized warehouse facilities, while potentially efficient, can lead to market power exploitation by large traders. - Credit vs. Money: Relying solely on credit or overdraft accounts can lead to defaults when agents cannot honor commitments at the end of the day. - Distributed Ledgers as Solution: Distributed ledgers offer a way to address these problems.

Readings for this lecture include: - Chester Spatt, "Regulating Market Microstructure" - Martin and McAndrews, on liquidity-saving mechanisms - Alcazar, on concentrated core banking services and imperfect competition - "Tri-Party" paper, on Federal Reserve liabilities in repo markets.

Lecture 4: Smart Contracts and Coordination Problems

This lecture focuses on smart contracts as a solution to coordination problems, particularly in fragmented markets with privately issued securities or monies.

Key points: - Privately Issued Monies: High-velocity private securities can circulate, but if not coordinated, they can lead to crashes due to incorrect anticipation of asset issuance in other locations. - Historical Context: Walter Bagehot's observations of London's money markets and the periodic crashes associated with bills of exchange provide historical parallels. The first clearinghouse on Lombard Street addressed these issues. - Contemporary Relevance: Similar coordination challenges exist in low and middle-income countries with emerging digitization and in Decentralized Finance (DeFi) markets, where liquidity problems can arise. - Multi-Agent Smart Contracts: Distributed ledgers and smart contracts offer solutions to these coordination problems.

Readings for this lecture include: - Gorton, on private money production without banks (referencing bills of exchange in England) - Bagehot, on money markets - Sargent and Wallace, "Real Bills versus the Quantity Theory" (on inflation and monetary policy debates).

Lecture 5: Tokenized and Programmable Assets

This lecture explores tokenized and programmable assets, focusing on atomic trade and settlement on platforms, and comparing them with legacy systems that experience trade fails.

Key points: - Dynamic Ledgers: The ledger becomes dynamic, allowing for atomic trade and settlement. - Multilateral Smart Contracts: Tokenization and programmable assets are enabled by these contracts, which provide dynamic, explicit representations of future assets and liabilities. - Immediate Settlement: Blockchain-based systems offer immediate settlement, contrasting with legacy systems' end-of-day or T+3 settlement, which are prone to trade fails. - Trade Fails: The magnitude of trade fails in legacy systems (e.g., trillions in US treasuries) highlights the problem. - Challenges of Tokenization: - Information Revelation: If assets can only be programmed by their current holder, trading could reveal private information, affecting pricing. - Interoperability: Emerging systems may be only partially interoperable. - Trust Issues: Tokenization of pre-existing assets requires trust in the issuer and escrow accounts to prevent fraud. - Cross-Border Exchange: Domestic central banks and international agencies are exploring blockchain solutions for cross-border transactions, with the BIS proposing unified ledgers for domestic systems. - Coherence Guarantee: Care must be taken to ensure that new systems do not violate the coherence guarantee, a key feature of multilateral smart contracts.

Readings for this lecture include: - Articles on settlement fails and the risks of tokenization. - Discussions on programmable money and the coherence guarantee. - Work from the Swiss National Bank and other consortia of large banks.

Lecture 6: Algorithmic Flows on Networks for Multilateral Settlement

This lecture examines how algorithms can solve multilateral settlement problems by leveraging network cycles and chains, particularly for trade credit offsets.

Key points: - Network Representation: Obligations between traders can be represented as network diagrams, where edges denote liabilities and assets. - Complexity of Networks: Real-world financial networks are highly complex, with individual agents having limited knowledge of the overall structure. - Ordering of Payments: Even with complete network knowledge, determining the optimal ordering of payments in a multilateral system is a complex combinatorial problem. - Liquidity Injection: If a liquidity pool or overdraft facility exists, deciding where and how much liquidity to inject, and how to recover it, is crucial for system balance. - Trade Credit Overdues: The lecture will focus on using multilateral set-offs (clearing closed cycles) or liquidity injection through chains to address trade credit overdues. - Guest Speaker: Tomaž Fleischman, a computer scientist and collaborator, will present on this topic.

Readings for this lecture include: - Fleischman et al., "Liquidity-Saving Obligation." - Cormen et al., "Introduction to Algorithms." - Optional: Research on the US repo market.

Lecture 7: Dynamic and Stochastic Networks, Risk Sharing, and Contagion

This lecture extends the network concept to dynamic and stochastic settings, focusing on risk-sharing and financial contagion.

Key points: - Federal Funds Market: The federal funds market serves as an example of a dynamic network where traders are interconnected. - General Equilibrium: The course revisits general equilibrium concepts, tracking dates, states of the world, and trading relationships to analyze risk-sharing problems. - Liquidity Injection in Risk Sharing: If new liquidity is available (e.g., from monetary policy on blockchains), the most valued recipients are traders in market clusters experiencing correlated shocks, low average incomes, and high risk aversion. - Contagion vs. Risk Sharing: A contrasting literature views financial markets through the lens of contagion, where adverse shocks spread like disease, arguing for limiting markets. This is opposite to the risk-sharing perspective, which advocates for increasing the number of traders to mitigate risk. - Financially Central Players: The lecture will differentiate between the notions of financially central players in risk-sharing and contagion models.

Readings for this lecture include: - "Contagion! Systemic Risk in..." (a framework used by policymakers). - Martin Sumner's review piece on financial contagion.

Lecture 8: Mechanism Design, Incentives, and Trust

This lecture delves into mechanism design, focusing on incentives to follow protocols and different notions of trust in economics and computer science.

Key points: - Incentives in Mechanism Design: Economics emphasizes incentives to report truthfully and take appropriate actions (e.g., in insurance, mitigating lying). - Implementation without Central Planner: Smart contract code can be validated by agents in advance, with public commitment and escrowed payouts, without needing a central planner. Messages and claims can be recorded in the contract's database. - Pragmatic Approach: Not every piece of blockchain technology is required for effective use; a layered approach (e.g., Layer 1 for code, Layer 2 for agent messages) can be employed. - Computer Science Approach to Trust: Computer science relies on parallel computing and notions of failure. By bounding the probability of failure and replicating computations, truth can be determined even with faulty components or "rogue traders" (Byzantine fault tolerance). - Bitcoin's Proof of Work: This mechanism uses random selection to validate transactions, making it difficult for rogue traders to manipulate the system. - Byzantine Generals Problem: This classic problem illustrates the challenges of communication and coordination among parties when messages can fail. While a protocol exists, it is vulnerable to economic incentives and Bayesian Nash equilibrium, where agents may second-guess each other and fail to act.

Readings for this lecture include: - Articles on computing and mechanism design. - Papers on distributed ledgers and the governance of money.

Lecture 9: Encryption and Keeping Secrets

This lecture introduces encryption as the third leg of new technologies, alongside distributed ledgers and smart contracts.

Key points: - Historical Context: Ancient Mesopotamia's use of clay envelopes with tokens illustrates early methods of securing information. - Modern Encryption: Public and private keys are central to modern encryption, enabling secure communication, authentication of message origin, and verification of message integrity. - Cryptographic Tools: Hashes, cryptographic puzzles, fully homomorphic encryption (allowing computations on encrypted data), and multiparty computation (including zero-knowledge proofs) are key tools. - Zero-Knowledge Proofs: These allow one party to prove they possess certain information without revealing the information itself.

Readings for this lecture include: - A Harvard professor's work on zero-knowledge mechanisms. - Practical implementations like Chainlink, LayerZero, and Dfinity's internet computer.

Lecture 10: Designs Using Encryption

This lecture explores practical applications of encryption, particularly in auctions and insurance.

Key points: - Encrypted Auctions: Bids can be encrypted, yet the winner can be determined without a central, potentially corruptible, auctioneer. - Private Information in Insurance: Encryption can handle situations where agents want insurance for balance sheet shocks but don't want to reveal their private financial state to third parties. - Scrambled Information: Information can be scrambled so no single entity has the complete picture, yet the code can perform desired functions, including randomization at crucial moments. - Fully Homomorphic Encryption and Multiparty Computation: These technologies enable "miracles" like computations on encrypted data and secure multi-party interactions. - Centralized Matching with Encryption: Encrypted supply and demand schedules can be matched to determine equilibrium outcomes in a secure manner. - Real-World Applications: MIT's work with Visa on encryption in blockchain is an example.

Lecture 11: Price-Quantity Strategic Market Games and Decentralized Exchanges

The final lecture juxtaposes classic economic theory with modern decentralized exchanges.

Key points: - Walrasian Outcome: Dubey's "Price-Quantity Strategic Market Games" is a classic paper on achieving the Walrasian outcome through implementation. - SPEEDEX: This decentralized exchange bears similarities to Dubey's work. - Computational Complexity: Finding Nash equilibria in multi-agent systems with complex strategies is computationally challenging. - Algorithmic Guidance: Ongoing research aims to use algorithms to compute ideal equilibrium candidates and signal optimal strategies to agents, reducing the need for endless conjectures. - Daskalakis's Work: This work, by an MIT computer science professor, contributes to the algorithmic approach.

Course Organization and Assessment

The course emphasizes research and exploration. There will be no midterm or final exam. Instead, the focus is on research projects, allowing students to pursue topics of greatest interest to them, with provided pathways and suggested readings.