Working Papers
2025
- The Price of Processing: Information Frictions and Market Efficiency in DeFiPablo D. Azar, Sergio Olivas, and Nish D. SinhaApr 2025
This paper investigates the speed of price discovery when information becomes publicly available but requires costly processing to become common knowledge. We exploit the unique institutional setting of hacks on decentralized finance (DeFi) protocols. Public blockchain data provides the precise time a hack’s transactions are recorded—becoming public information—while subsequent social media disclosures mark the transition to common knowledge. This empirical design allows us to isolate the price impact occurring during the interval characterized by information asymmetry driven purely by differential processing capabilities. Our central empirical finding is that substantial price discovery precedes common knowledge: approximately 36 percent of the total 24-hour price decline (∼27 percent) materializes before the public announcement. This evidence suggests sophisticated traders rapidly exploit their ability to process complex, publicly available on-chain data, capturing informational rents. We develop a theoretical model of informed trading under processing costs which predicts strategic, slow information revelation, consistent with our empirical findings. Our results quantify the limits imposed by information processing costs on market efficiency, demonstrating that transparency alone does not guarantee immediate information incorporation into prices.
- The Financial Stability Implications of Tokenized Money Market FundsPablo D. Azar, Francesca Carapella, and Alexandros VardoulakisApr 2025
The market for repurchase agreements (repo) is central to short-term funding and monetary policy transmission, relying critically on high-quality collateral, typically sovereign debt. While the exchanged security serves as the primary collateral, cash is commonly posted to cover haircuts or variation margin, ensuring the transaction remains adequately collateralized against market fluctuations. This paper examines the financial stability implications of substituting tokenized Money Market Fund (MMF) shares for cash in this crucial margining function. We argue that while technologically feasible, this substitution introduces novel systemic risks. Tokenized MMFs inherit the run propensity of traditional MMFs and add unique technological vulnerabilities (blockchain finality, smart contract risk). Embedding these instruments within the repo margining process creates a new transmission channel for shocks from both the MMF sector and the digital asset ecosystem into the core of the traditional financial system, potentially amplifying procyclicality and systemic fragility. We draw parallels to the information sensitivity of private debt used as collateral, suggesting tokenized MMFs may be insufficiently robust to serve the role typically played by cash margin in mitigating counterparty risk during stress.
- Banking on the BlockchainPablo D. Azar, and Will F. DiamondApr 2025
The ability of banks to transform maturities creates economic value but also inherent fragility, potentially leading to inefficient runs and liquidations, as shown by Diamond and Dybvig (1983). Traditional solutions like deposit insurance or suspension of convertibility mitigate runs but introduce other distortions. This paper asks whether banks can achieve efficient liquidity transformation without runs, relying solely on market mechanisms augmented by modern cryptographic tools, thereby potentially reaching a constrained first-best outcome. We develop a model where banks face idiosyncratic depositor liquidity shocks and where revealing liquidity stress publicly incurs stigma, triggering runs. We argue that cryptographic primitives require verifiable on-chain representations of bank state, achievable via specific blockchain architectures (e.g., Bowe, Gabizon, and Miers 2020) or trusted auditors. Our core analysis shows that anonymous funding mechanisms, enabled by cryptography and verifiable state, allow solvent-but-stressed banks to access liquidity without incurring stigma, thus eliminating inefficient runs. Under conditions where such funding is sufficiently available, the equilibrium allocation approaches the constrained first-best, characterized by minimal precautionary liquidity hoarding and the absence of belief-driven runs. We also analyze the complementary role of Zero-Knowledge Proofs for state verification and, in an extension, Verifiable Random Functions for mitigating adverse selection in liquidations. The paper provides a micro-founded analysis of how specific cryptographic tools, underpinned by verifiable state information, can address fundamental informational frictions in banking.
- Computation, Misallocation and GrowthPablo D. AzarApr 2025
This paper develops a novel economic growth model where computational capacity serves as a fundamental constraint on allocation efficiency. We model an economy facing an infinite sequence of resource allocation problems with varying dimensionality, where finding optimal solutions requires exponentially increasing computational resources as problem complexity grows. Using results from computational complexity theory—specifically the hardness of the Closest Vector Problem—we derive the optimal allocation of computational resources across problems and characterize the steady state. Our model demonstrates that computational investment represents a critical channel for economic growth, particularly as economies become more complex. The framework provides new insights into the diminishing returns to computational advancement observed in mature sectors and establishes conditions for sustainable growth paths that balance computational capacity with increasing economic complexity.
- Blockchain Breakdown: Causal Evidence that Proof of Work is not StrategyproofPablo D. AzarApr 2025
Digital assets, increasingly integrated into finance via ETFs and tokenized products, rely on consensus protocols like Proof-of-Work (PoW) for security. However, whether these PoW protocols truly incentivize honest participation or allow strategic deviation (selfish mining) remains debated. Exploiting high-frequency Ethereum data and a novel instrumental variables strategy, we causally demonstrate PoW’s manipulability. We find that higher mining rewards—-driven by exogenous shocks such as crypto hacks and crises—-significantly increase stolen and uncle blocks—proxies for consensus instability linked to selfish mining. Quantitatively, a one percent increase in average hourly transaction fee revenue leads to approximately 1.23 additional stolen blocks per hour. We explain these empirical findings with a model where selfish miners can adjust their hashing rate at the intensive margin by renting computational power from mining pools. This reveals a fundamental tension: the very incentives designed for blockchain security can degrade stability, posing systemic risks and highlighting critical market design challenges for decentralized financial infrastructure as these assets become more interconnected with traditional financial institutions.
2021
- Moore’s Law and Economic GrowthPablo D. AzarMay 2021Revised October 2022
Over the past sixty years, semiconductor sizes have decreased by 50 percent every eighteen months, a trend known as Moore’s Law. Moore’s Law has increased productivity in virtually every industry, both by increasing the computational and storage power of electronic devices, and by allowing the incorporation of electronics into existing products such as vehicles and industrial machinery. In this paper, I examine the physical channel through which Moore’s Law affects GDP growth. A new model incorporates physical constraints on firms’ production functions and allows for new types of spillovers from the physical characteristics of products. I use the model, and a new data set of product weights, to estimate the effect of the electronic miniaturization channel on productivity growth. The results show that between 11.74 and 18.63 percent of productivity growth during 1960 to 2019 can be attributed to physical changes in the size of electronic components. This effect is highest during the 1990s and early 2000s.