| A | B | C | D | E | F | G | H | I | J | K | L | M | N | O | P | Q | R | S | T | U | V | W | X | Y | Z | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
1 | theme | title | authors | organisations | keywords | abstract | ||||||||||||||||||||
2 | fintech innovation | Algorithmic Persuasion - Financial advice in the age of AI | Nielsen, Mads; Zarpala, Labrini | Utrecht University, Netherlands, The | Large Language Models, Bayesian Persuasion, Financial advice | We demonstrate how distortion of content arises endogenously when a content-curating or content-creating algorithm is rewarded for maximizing exposure to particular content regardless of user preferences. We show that when insights from the static model extend to the dynamic setting, characterizing trends in distortion due to fundamentally productive and counter-productive effects. Applying the framework to financial advice by Large Language Models (LLMs), we microfound user utility through portfolio choice. When users suffer from confirmation bias, optimal customization balances productivity enhancements with confirming preferences. Even when they do not, better information quality can paradoxically hurt the precision of users by relaxing their participation constraint, enabling more distortion that offsets the direct gains. | ||||||||||||||||||||
3 | fintech innovation | Building Better FinTech: The Academic Edge | Gargano, Antonio | University of Houston, United States of America | FinTech, Household Finance, Investments | This presentation is not about a single paper but an overview of how FinTechs can benefit from collaborating with academics (drawing from two examples of my own research) | ||||||||||||||||||||
4 | fintech innovation | Robinhood’s Forced Liquidations | Amaya, Diego (1); Garcia Ares, Pedro Angel (2); Pearson, Neil D. (3); Vasquez, Aurelio (4) | 1: Wilfrid Laurier University; 2: ITAM, Mexico; 3: University of Illinois at Urbana-Champaign; 4: ITAM, Mexico | Retail option trading, Robinhood, market microstructure, transaction costs, meme stocks, execution costs, retail investor behavior, option market frictions | Shortly before expiration, Robinhood submits trades to close out the options positions of its customers who do not have the cash or shares to exercise their options or accept assignment. These liquidations result in bursts of customer trades at known times, and allow us to identify the underlying symbols and options positions popular with Robinhood customers. The liquidating trades face adverse execution, as options prices move in unfavorable directions. Underlying equity and ETF prices move in directions consistent with price pressure in the equity and ETF markets as options market makers execute delta hedge trades as they absorb the Robinhood order flow. Our results reveal how brokerage frictions in retail options trading impact options and underlying prices. | ||||||||||||||||||||
5 | fintech innovation | The Rise of Algorithmic Retail Option Traders | Amaya, Diego (1); Garcia Ares, Pedro Angel (2); Pearson, Neil D. (3); Vasquez, Aurelio (4) | 1: Wilfrid Laurier University; 2: ITAM, Mexico; 3: University of Illinois at Urbana-Champaign; 4: ITAM, Mexico | Retail option trading; 0DTE; algorithmic execution; market microstructure; fintech | We show that retail participation is increasingly rule-based rather than discretionary, leaving sharp and recurring intraday footprints in trading activity. Using transaction-level data from SPX zero-days-to-expiration (0DTE) options, we document pronounced volume spikes exactly at the hour and half-hour marks that emerge almost instantaneously and dissipate within seconds. These spikes intensify over time, are concentrated in complex multi-leg trades, and are largely absent in longer-dated contracts. Spike-time trading is dominated by small, standardized, short-premium strategies consistent with template-driven execution and mechanical risk budgeting, and we do not find evidence that these trades are systematically wealth-depleting. Around these deterministic windows, quoted spreads tighten while effective spreads widen, indicating intensified liquidity provision alongside higher execution costs. Overall, fintech-enabled retail automation generates synchronized order flow that reshapes intraday liquidity and market quality. | ||||||||||||||||||||
6 | payments and infrastructure | Almond Fintech: The intelligence layer for global money movement | Swartzbaugh, Adam | Almond FinTech, United States of America | Dynamic Routing Optimization, Non-Stationary Financial Graphs, Liquidity-Aware Execution, Stochastic FX Optimization, Adaptive Multi-Asset Orchestration | Cross-border payment networks are best understood not as static pipelines, but as dynamic, non-stationary graphs in which each routing decision reshapes the system itself - shifting liquidity, altering token prices, and changing the future feasibility and cost of available paths. Traditional routing and max-flow algorithms fail in this setting because they assume fixed capacities and independent edges, while real-world markets exhibit flow-dependent costs, endogenous feedback loops, and rapidly evolving state. This work reframes cross-border FX and stablecoin-based settlement as a real-time control problem under uncertainty, where optimal decisions must account for both current conditions and the impact of execution on future network states. The approach integrates short-horizon forecasting with stochastic optimization to enable adaptive routing across chains, tokens, venues, and time in a continuously changing environment. | ||||||||||||||||||||
7 | payments and infrastructure | Corporate Liquidity Supply from Non-Bank Intermediaries and the Real Effects of Factoring | Zhang, Henry (1); Orestes, Victor (2); Silva, Thiago (3) | 1: CUHK; 2: Wharton School, University of Pennsylvania; 3: Central Bank of Brazil / IMF | Factoring, Receivables, Trade Credit, Working Capital, Liquidity | We show that short-term fluctuations in firms’ ability to convert trade credit receivables into liquidity through factoring have large and persistent real effects, with limited substitution from other financing sources or adjustments in trade credit terms. In Brazil, specialized non-bank intermediaries (FIDCs) securitize receivables and are key providers of working capital financing. Using novel transaction-level data linking factoring, invoices, payments, credit operations, and employment records, we exploit investor inflows to FIDCs in a shift-share design to identify exogenous variation in factoring supply. A one-percentage-point decline in factoring rates increases factoring volumes by 16%, revenues by 6%, and intermediate input expenditure by 4%, with effects persisting for several months. Firms expand permanent employment and demand less temporary labor. A model of corporate liquidity management rationalizes these findings: factoring endogenously transforms production into collateral, tying firms’ real and financial decisions. Model-implied macro-elasticities indicate that lowering economy-wide factoring spreads by 1 percentage point would raise aggregate output and wages by 0.3 to 0.5 percentage points. | ||||||||||||||||||||
8 | payments and infrastructure | Deepening the Secondary Market: Integrating Trade Credit into Market Clearing with the Cycles Protocol | Fleischman, Tomaž; Buchman, Ethan | Cycles Protocol SA, Switzerland | post-trade, trade credit, clearing, setoff, settlement | Current post-trade clearing systems rely almost exclusively on cash or cash-like collateral, leaving vast reserves of short-term liquidity embedded in trade credit outside formal settlement infrastructures. This paper introduces a clearing framework that integrates accounts receivable and payable (AR/AP) into secondary market settlement via the Cycles Protocol—a distributed, multilateral mechanism based on double-entry accounting and atomic cycle execution. The Cycles Protocol acts as a clearing and settlement layer that reduces liquidity needs by using multilateral set-off and cycle removal to maximize balance-sheet compression. It operates without novation and complements, rather than replaces, CCPs’ margining, loss mutualization, and default management. Compared with Liquidity-Saving Mechanisms (LSMs) in Real-Time Gross Settlement (RTGS) systems, this approach extends liquidity optimization beyond interbank payments to real-economy financing networks, reducing systemic reliance on scarce collateral and central intermediaries. | ||||||||||||||||||||
9 | payments and infrastructure | How Fintech Affects Spending Behavior: Evidence from Tap-to-Pay | Voges, Ryan | University of Utah, United States of America | Consumer Spending, Payment Frictions, Tap-to-Pay | I examine how contactless tap-to-pay (TTP) adoption affects consumer spending using item-level transaction data from U.S. convenience stores. Exploiting staggered store rollouts within a triple-difference framework, I show that TTP raises monthly spending primarily by increasing transaction frequency; a phenomenon of consumption fragmentation. Crucially, while monthly totals rise, individual TTP transactions are smaller and contain fewer items. These effects are concentrated in impulse-sensitive categories like beverages, tobacco, and food service products. My findings suggest that reducing interface-level micro-frictions lowers payment salience and weakens self-control, demonstrating that even incremental fintech innovations fundamentally reshape the timing and composition of consumption. | ||||||||||||||||||||
10 | payments and infrastructure | Reclaiming Certainty at the Core of Finance | Brammertz, Willi (1); Mark, Robert (2); Mendelowitz, Allan (3) | 1: Ariadne, Switzerland; 2: Black Diamond; 3: ACTUS Financial Research Foundation | Standards, Algorithmic representation of financial contracts | Financial contracts are promises to pay, specifying who pays whom, how much, and when. These cashflows are the backbone of transaction processing, risk management, accounting, and other systems. The contractual cashflows are determined by a limited set of mathematical algorithms that contain inherent "certainty”. Nevertheless, the legal, academic, and financial communities have failed to recognize this “certainty”, leading to unstandardized representations of the same financial contracts across different systems and uses. This creates inefficiencies, vulnerabilities, and reconciliation burdens. We discuss how operationalizing "certainty" with standardized algorithms is essential for modernizing TradFi and providing the necessary foundation to scale DeFi ecosystems and integrate them with mainstream finance. | ||||||||||||||||||||
11 | blockchain and cryptocurrencies | Buyback Programs for Platform Tokens | Garratt, Rodney (1); van Oordt, Maarten (2,3) | 1: University of California, Santa Barbara; 2: Vrije Universiteit Amsterdam; 3: Tinbergen Institute | Asset pricing, Crypto assets, Exchanges, Platforms, Market manipulation | Pledges to buy back tokens issued by platforms in decentralized finance are becoming increasingly common. We develop a tractable model for the exchange rates of platform tokens that incorporates user demand, investment demand, and buyback pledges. We derive closed-form solutions for the valuation of tokens and the time required to fulfill the pledge. Buyback pledges can increase the value of the tokens, but manipulation of the token supply by market participants may make the pledge more expensive than intended. Sufficiently aggressive buyback pledges induce purely financially motivated investment in platform tokens, which could trigger regulatory classification as securities. | ||||||||||||||||||||
12 | blockchain and cryptocurrencies | Do Cryptocurrency Investors care about Quantum Risks? | Bertucci, Louis (1); Jahanshahloo, Hossein (2); Scharnowski, Stefan (3) | 1: Institut Louis Bachelier, Paris, France; 2: Alliance Manchester Business School, The University of Manchester, UK; 3: University of Mannheim, Germany | Cryptocurrency, Quantum Computing, Cryptography, Security, Bitcoin | Although large-scale quantum computers are not yet available, their future development poses a potential threat to the cryptographic foundations of cryptocurrencies. As an upper bound, we estimate that by the end of 2024, approximately US$ 586 billion in the Bitcoin network alone could be potentially exposed. We then examine whether investors are aware of this risk by analyzing their response to advances in quantum computing. Conventional cryptocurrencies exhibit negative returns and higher trading volume following such news, whereas a quantum-robust cryptocurrency exhibits positive returns. Our findings suggest that some investors are aware of quantum computing risks and respond by shifting toward quantum-resistant cryptocurrencies. | ||||||||||||||||||||
13 | blockchain and cryptocurrencies | Dynamics of exchange trading and Blockchain settlement | Roesch, Dominik | University at Buffalo | Bitcoin, Cryptocurrency, Settlement, Blockchain, Bank-run, Default | I analyze the joint dynamics of Bitcoin trading on exchanges and on the blockchain (“settlement”) from 2009 to 2023. Aggregate trading and settlement exhibit a strong shared trend both on a daily and intra-day basis. Monthly correlations average 40%, though they vary significantly over time and across exchanges. In aggregate and for regulated exchanges, but not for several unregulated exchanges, increased trading ac- tivity leads to higher settlement both contemporaneously and in subsequent periods. Ranking exchanges by how trading influences settlement provides a useful quality met- ric: increased trading driving higher settlement indicates that exchanges or their clients settle trades frequently, making the exchange less vulnerable to bank-runs and default. On the contrary, I identify BTCC, LocalBitcoins, and Yobit as exchanges with elevated default risk. | ||||||||||||||||||||
14 | blockchain and cryptocurrencies | How Do Flash Loans Affect Market Liquidity? | Li, Simeng | Imperial College London | Market Liquidity, Decentralized Finance (DeFi), Flash Loans, Limits to Arbitrage, Automated Market Makers (AMM), Uniswap, Blockchain. | Flash loans are a DeFi primitive that allow users to borrow large amounts of capital without collateral, provided the loan is borrowed and repaid within a single atomic transaction. We ask how this form of on-demand, uncollateralized funding affects market liquidity and execution quality in automated market makers. Exploiting the token-level rollout of flash-loan eligibility on Aave as a natural experiment, we show that flash-loan introduction leads to a pronounced and persistent increase in Uniswap V1 liquidity: executable depth, total value locked, and trading activity rise relative to matched control pools, while slippage and volatility do not deteriorate in a sustained way. Using transaction-level classifications and high-frequency panel regressions, we further show that flash-loan intensity is positively associated with subsequent depth and TVL, with effects concentrated in large debt-restructuring and arbitrage loans. Overall, our evidence suggests that flash loans operate primarily as a liquidity-enhancing funding technology that deepens markets without worsening short-run execution quality, while reallocating risk toward passive liquidity providers. | ||||||||||||||||||||
15 | blockchain and cryptocurrencies | How do Users Gain Influence in Social Networks? | Hui, Yunming (1); Rudinac, Stevan (1); Trimborn, Simon (1,2); Zwetsloot, Inez (1) | 1: University of Amsterdam, Netherlands, The; 2: Tinbergen Institute | Financial Influencers (Finfluencers), Social Networks, Cryptocurrency Markets, Heterogeneous Graph Neural Networks (HGNN) | The growing importance of social media in financial markets has amplified the role of financial influencers (“finfluencers”) in shaping investor attention and market dynamics. Yet little is known about how finfluencers gain and sustain influence in financial social networks. We conduct this study in the context of meme coin markets, a highly social and sentiment-driven segment of cryptocurrency markets. We examine influence formation in Reddit-based meme coin networks during a full boom–bust cycle (September 2024 to March 2025), using data from 103,882 posts, 609,904 comments, and 128,697 users across ten subreddits. We model users, posts, comments, subreddits, and hourly market conditions as a heterogeneous graph and apply a Heterogeneous Graph Neural Network (HGNN) to predict user influence rankings. The HGNN framework allows us to jointly incorporate topological structure, content features, user attributes, and market states without manually specifying interaction patterns. Our results show that influence mechanisms are strongly market-state dependent. During high-volatility periods, network position and engagement timing dominate, while content features lose predictive power. In contrast, content quality becomes important in stable markets. We further identify a highly skewed hierarchy of influence and distinguish sustained finfluencers from occasional ones, with persistence driven by consistent activity and central network positioning. | ||||||||||||||||||||
16 | blockchain and cryptocurrencies | Implied Impermanent Loss for Concentrated Liquidity | Alberici, Luca (1); Papanicolaou, Andrew (2); Schoenleber, Lorenzo (3) | 1: Bayes Business School; 2: North Carolina State University; 3: Collegio Carlo Alberto | Decentralized Exchanges, Decentralized Finance, Impermanent Loss, Derivatives, Risk-Neutral Pricing, Risk Premium, Staking, Yield Farming. | Providing liquidity on decentralized exchanges earns fees but exposes liquidity providers (LPs) to impermanent loss from price movements. With concentrated liquidity, LPs control this risk by choosing how narrowly to deploy capital around the price. Using option prices, we quantify the cost of liquidity provision by developing measures of implied and realized impermanent loss for concentrated liquidity and define the associated impermanent loss risk premium. Empirically, higher expected impermanent loss widens liquidity ranges, while higher risk premia re-center liquidity around the spot price, highlighting opposing effects of risk and compensation. | ||||||||||||||||||||
17 | blockchain and cryptocurrencies | Limits to Arbitrage in Decentralized Finance | Swaminathan, Balasubramaniam | NEOMA Business School, France | Blockchain, Arbitrage, Flash Loans, Decentralized Finance | Flash loans provide uncollateralized, atomic leverage that eliminates traditional capital constraints on arbitrage. We document that flash-loan arbitrage exhibits substantially higher concentration than capital-funded arbitrage, with the top ten participants capturing approximately 50\% of volume compared to 20\% for traditional arbitrageurs. This pattern arises because atomic execution creates option-like payoffs with limited liability but reveals complete strategy information when broadcast publicly. Intense competition induces selection into costly private routing channels requiring non-transferable technological infrastructure. Our findings show that eliminating capital constraints shifts the binding constraint to technological capability, generating concentration through infrastructure requirements rather than financial capacity. | ||||||||||||||||||||
18 | blockchain and cryptocurrencies | Market Efficiency in Prediction Markets - A Comparison with Derivatives | Fabi, Michele (1); Marfe, Roberto (2); Ruffo, Vittorio (3); Schoenleber, Lorenzo (2) | 1: Telecom Paris; 2: Collegio Carlo Alberto, Italy; 3: Frankfurt School of Finance and Management | Prediction Markets, Market Efficiency, Risk-Neutral Pricing, Crypto Derivatives, Behavioral Biases, Decentralized Finance | We study pricing efficiency in decentralized prediction markets by comparing market-implied probabilities from Polymarket with benchmarks derived from option-implied risk-neutral distributions extracted from the derivatives market. We study Bitcoin prediction bets and find that, although Polymarket prices broadly track option-implied benchmarks, they show systematic mispricing driven by complexity, behavioral factors, and market frictions. Mispricing is most pronounced in tail events, during periods of high volatility, major macroeconomic shocks, and reflects the influence of sentiment, attention, and blockchain-specific risks. These results reveal both efficiency and behavioral distortions in prediction markets. | ||||||||||||||||||||
19 | blockchain and cryptocurrencies | Non-native tokens and price discovery | Capponi, Agostino (1); Sokolov, Konstantin (2); Zhang, Jiang (3) | 1: Columbia University; 2: University of Memphis; 3: University of St. Thomas | Blockchain, Non-native tokens, Price discovery, Decentralized exchange, NFT Airdrops | The unrelenting growth in blockchain functionality drives greater adoption and, therefore, ensures the security of transaction settlement. We show that the wider adoption carries its own risk for the blockchain. Specifically, when a large fraction of blockchain transactions involves non-native tokens, the cost of price discovery in the native cryptocurrency increases. This, in turn, leads to lower trade informedness and price efficiency. In a difference-in-differences setup, we find that a shock to non-native transactions typically leads to 1.25 bp of unrealized informed price movement and a 0.60% decrease in price efficiency in the underlying cryptocurrency. Our findings lend support to the theoretical literature arguing that multiple attack vectors on blockchain arise from native cryptocurrency mispricing. | ||||||||||||||||||||
20 | blockchain and cryptocurrencies | On Bubbles in Cryptocurrency Prices | van Oordt, Maarten | Vrije Universiteit Amsterdam | Asset pricing, Bitcoin, crypto assets, exchange rates, rational bubbles | This paper develops a tractable model for the cryptocurrency prices based on the classical framework for rational bubbles. In the baseline equilibrium, investors hold cryptocurrency to sell them to future users. In a bubble equilibrium, investors hold cryptocurrency because they expect its price to appreciate due to future investment inflows. We establish the mathematical relationship between net investment flows and the cryptocurrency's rate of appreciation. The net investment flows required to sustain a bubble equilibrium increase in new coin issuance, the required return and the level of transactional demand, and temporarily decrease when transactional demand expands. The net investment inflows required to sustain a bubble equilibrium are, everything else equal, smaller for cryptocurrencies with a proof-of-stake than for cryptocurrencies with a proof-of-work protocol. | ||||||||||||||||||||
21 | blockchain and cryptocurrencies | Programming money without programmable money | Lee, Michael (1); Martin, Antoine (2) | 1: Federal Reserve Bank of New York; 2: Swiss National Bank | digital money, programmability, payments, monetary systems | Programmability is at the heart of ongoing work on the future of money and payments by central banks around the world. Despite its potential, there is growing concern that programmability conflicts with the provision of “good” money. This paper overviews key principles of “good” money and argues that the discourse on programmability inadequately differentiates between programmable money, which is generally negatively viewed, and programmable payments, which is generally accepted as part of the future. We provide a framework for programmable monetary systems that sharply distinguishes between programmable money and programmable payments. We show that our framework nests a broader set of financial arrangements and revisit the debate on programmability in the design of monetary systems. | ||||||||||||||||||||
22 | blockchain and cryptocurrencies | Stablecoins Activity: Use and Misuse | Gadzinski, Gregory; Maeso, Jean-Michel; Castello, Alessio; Liuzzi, Vito | International University of Monaco, Monaco | Stablecoins, DeFi, CeFi, Manipulations, Concentration | This article develops a taxonomy of stablecoin activity on the Ethereum blockchain. Using all ERC-20 transfers for USDT, USDC, and DAI for the period 2021–2025, we filter noise, Maximal Extractable Value (MEV) signatures and wash-like behaviours, then map the filtered transactions into directed networks and Louvain communities. We show that stablecoin usage cannot be reduced to manipulations and confirm that stablecoins are mainly settlement layers for crypto markets rather than transactional instruments. We also document institutional centralization for exchanges and competitive concentration for DeFi and fintech, highlighting that structural decentralization and economic decentralization are distinct dimensions of stablecoin network dynamics. | ||||||||||||||||||||
23 | blockchain and cryptocurrencies | Statement on the Digital Euro | Grothoff, Christian (1); Kesim, Özgür (2) | 1: Bern University of Applied Sciences; 2: Code Blau GmbH | CBDC, ECB, Digital Euro, security, privacy, corruption | The current implementation of the Digital Euro, as proposed by the European Central Bank, raises serious concerns regarding public spending, competition, privacy, and monetary stability. First, the scale of the proposed budgets is unjustified. For example, the EUR 56 million allocated to the Alias Lookup service (PRO-009485) is far beyond what comparable systems typically require. This is particularly troubling given that standardized and publicly available solutions already exist. At a minimum, this raises questions about procurement discipline and value for money. Second, the tender requirements are structurally exclusionary. By designing the system to function only on the two dominant proprietary mobile platforms, the ECB reinforces an existing duopoly and marginalizes European and Free/Libre open-source alternatives. This approach contradicts stated EU goals of competition, digital sovereignty, and technological independence. Third, the online Digital Euro does not meet citizens' stated demand for payment privacy. Survey data commissioned by the ECB itself shows that Europeans want digital cash. Instead, the proposal delivers a capped, account-like instrument with unclear liability structures. The lack of confidence in its usefulness is reflected in the proposal to mandate merchant acceptance by law. Finally, the offline Digital Euro requirements pose a fundamental and non-negotiable problem. The combination of full anonymity, offline operation, transferability, and zero risk to the recipient and the central bank is not merely an open engineering challenge --- it is mathematically incompatible. This is a formally established result in cryptography: without online reconciliation or a trusted authority, digital assets can always be copied. No consumer hardware can change this fact; hardware can only raise the cost of attack temporarily. As a result, the ECB’s current requirements will result in adilemma: either (1) spend hundreds of millions of euros pursuing an impossible goal, or (2) give up on privacy and discharge unavoidable double-spending risks to citizens and law-enforcement, or (3) deploy a vulnerable system with the ECB remaining fully liable, risking monetary stability from large-scale fraud, potentially by sophisticated or state-level actors. In the current geopolitical and cyber-security environment, this represents a systemic risk to the Euro, not a technical detail. Taken together, these issues risk undermining trust in the Euro and in the institutions responsible for safeguarding it. We therefore urge the ECB to pause the current implementation, reassess its procurement and design assumptions, and switch over to known alternative solutions with a publicly vetted and technically proven design, which do not have the same drawbacks as outlined above. Europe deserves and needs a solution that is competitive, privacy-preserving, technologically sound, and aligned with democratic accountability. | ||||||||||||||||||||
24 | AI in finance | Does Corporate Production of AI Innovation Create Value? | Ahmadi, Ali (1); Kecskes, Ambrus (1); Michaely, Roni (2); Nguyen, Anh (3) | 1: Schulich School of Business at York University; 2: HKU Business School and ECGI; 3: School of Administrative Studies at York University | Artificial intelligence; Innovation; Technology; Labor; Firm value; Corporate finance; Asset pricing; Behavioral finance | Yes, by decreasing firm risk, not by increasing profitability, and with investors taking years to recognize the value created. We start, using novel AI patent data, by documenting significant corporate production of AI innovation as early as 1990. Then, we show that a signification motivation for a firm's AI production is the mutually reinforcing effects of the firm's innovation capacity (exogenous R&D stock) and its labor inputs' AI exposure (both the firm's own and its customers'). We use the interaction of these two effects to instrument for AI production. We find that producing AI creates firm value through a large, permanent decrease in risk (cash flow and stock return, systematic and idiosyncratic). Further evidence suggests that AI lowers physical capital intensity and increases bargaining power for producing firms. The initial market reaction to AI patent announcements is economically small, but abnormal stock returns thereafter are significantly positive (about 5% per year) for (only) roughly three years, suggesting initial undervaluation followed by gradual correction. We find no evidence of investor learning, except during the past five years. We empirically distinguish producing AI innovation versus AI adoption, automation, general technology, and other potential confounds. | ||||||||||||||||||||
25 | AI in finance | GenAI-Based Index of Financial Constraints | Ysmailov, Bektemir | Nazarbayev University | financial constraints, generative AI (GenAI), large language models (LLMs), textual analysis, MD&A disclosures, corporate finance | I construct a new measure of financial constraints by applying a large language model to narrative disclosures in firms' Management’s Discussion and Analysis from Form 10-K filings. The model evaluates each filing as a finance expert and classifies the firm's external financing difficulty on an ordered scale, producing the GenAI FC Index. The index captures contextual signals - such as nuanced liquidity discussions - that traditional accounting-based and prior text-based proxies often miss. It behaves sensibly in both the time series and cross-section and shows only moderate correlations with existing measures, indicating that it contains distinct information. Behavioral tests reveal that firms classified as constrained recycle far less equity and are substantially more likely to omit dividends, and less likely to initiate or increase them. Across these settings, the GenAI FC Index yields stronger and more consistent behavioral separation than benchmark text-based measures. The results demonstrate that generative AI can extract economically meaningful information about firms' financing frictions at scale. | ||||||||||||||||||||
26 | AI in finance | The Evolving Credibility of Stories | Rusonik, Alyssa | HEC Paris, France | Narratives, Stories, Beliefs, Credibility | Despite growing recognition that narratives shape economic outcomes, we lack a formal, empirically tractable definition of what constitutes a “story” or a “narrative.” This project develops a conceptual and computational framework to address that gap. I define a story as a temporally ordered sequence of economically meaningful events and a narrative as a set of stories sharing the same meaning. An axiomatic framework formalizes story similarity and disciplines narrative clustering. Empirically, I implement a three-stage pipeline: large language models extract structured event sequences from unstructured text; stories are compared using a multi-stage similarity measure that combines semantic and structural features; and network-based clustering aggregates similar stories into endogenously emerging narratives. I construct narrative-prominence indices which display systematic co-movement with financial indicators of interest (e.g., prices, returns, volatility, etc.), consistent with markets dynamically reweighting competing narratives over time. Interpreted through a revealed-preference lens, these patterns suggest markets act “as if” particular narratives are believed or attended to at different points in time. Ongoing work develops models of dynamic narrative credibility and evaluates whether narrative-based measures provide incremental forecasting power relative to existing text-based approaches. By providing both a formal definition and an empirically implementable identification strategy, the framework offers a new approach to studying belief formation and aggregate, market-level belief elicitation in financial markets. | ||||||||||||||||||||
27 | investments | (Every) 15 seconds to alpha: Long/short optimization with EVT | Christopoulos, Andreas (1); Barratt, Joshua (2) | 1: University of Cambridge, United Kingdom; 2: Barratt Consulting | CMBS, Extreme value theory, Liquidity, REITs, Risk decomposition | In this paper we model conditional distributions of intraday maximum and minimum REIT prices with extreme value theory (`EVT') techniques. We condition the parameters of these distributions on continuously evolving risk decomposition values derived from the CMBX market. These risk decompositions are interpreted as dynamic state variables, and serve as signals for changing likelihoods of daily REIT extrema. By assessing the model at fifteen-second intervals, intraday, our model generates high-confidence signals of single optimal stopping times for long/short trades for REITs in our study and extraordinary profits with positive and significant alphas in some 90% of our tests. | ||||||||||||||||||||
28 | investments | Asset (and Data) Managers | Zanotti, Marco | Swiss Finance Institute, USI Lugano, Switzerland | Asset Management, DataEconomy, Information, FinTech | This paper studies whether asset management companies use customer data to attract capital. Exploiting information from their websites' codes, I track when fund managers begin collecting and analyzing data on their potential customers using tools like Google Analytics or A/B testing. I show that funds adopting such technologies attract 1.5% higher annual flows and charge higher fees, despite no improvement in performance. These results are concentrated on retail share classes and decline with competition as more rival funds adopt similat tools. At the fund-family level, adopters expand their product offerings, and new fund focus more on retail-oriented themes. Within existing funds, I find evidence of changes in prospectus content and greater sales efforts rather than product differentiation. Overall, data technologies allow managers to raise more capital and charge higher fees. These findings show that technological innovation in asset management extends beyonf portfolio allocation decisions, and it affects how funds attract and retain capital. | ||||||||||||||||||||
29 | investments | Confident Risk Premiums and Investments using Machine Learning Uncertainties | Allena, Rohit | University of Houston, United States of America | Investment Strategies, Stock Return Forecasts, Confidence Intervals, Standard Errors | This paper derives ex-ante confidence intervals of stock risk premium forecasts that are based on a wide range of linear and Machine Learning models. Exploiting the cross-sectional variation in the precision of risk premium forecasts, I provide improved investment strategies. The confident-high-low strategies that take long-short positions exclusively on stocks with precise risk premium forecasts outperform traditional high-low strategies in delivering superior out-of-sample returns and Sharpe ratios across all models. The outperformance increases (decreases) with the model complexity (bias). The confident-high-low strategies are economically interpretable as trading strategies of ambiguity-averse investors who account for confidence intervals around risk premium forecasts. | ||||||||||||||||||||
30 | investments | Fast Flow in Slow-Moving Market: Leveraged Loan Fund Flow and Real Activity | Cheung, Wing Lam (1,2) | 1: University of Lausanne; 2: Swiss Finance Institute | Business cycle, Mutual fund flows, Leveraged loans, Investor demand, Leading indicator | Credit spreads are leading indicators of real activity but only capture pricing information. I show that quantity-based indicators, specifically net flow into open-end funds primarily investing in leveraged loans, provide additional predictive content. Using monthly indicators and conditioning on a broad set of controls, credit spreads reduce out-of-sample forecast errors in the next two years by up to 9%, while loan fund flow further reduces these errors by up to 7%. A dynamic corporate investment model with a slow-moving financing capacity state, which lags behind fast-moving fund flows, explains this price-quantity timing asymmetry in the forecasting power. | ||||||||||||||||||||
31 | investments | Firm-Level Input Price Changes and Their Effects: A Deep Learning Approach | Chava, Sudheer (1); Du, Wendi (2); Mitra, Indrajit (3); Shah, Agam (1); Zeng, Linghang (4) | 1: Scheller College of Business, Georgia Institute of Technology; 2: University of South Carolina; 3: Federal Reserve Bank of Atlanta, United States of America; 4: Babson College | Textual analysis, inflation, input price, cost shocks, passthrough, earnings calls | We develop firm-level measures of input and output price changes using textual analysis of earnings calls. We establish four facts: (1) Input prices increase (decrease) at the median firm once every 7 (30) months. (2) Input price changes contain an equal blend of aggregate and firm-specific components. (3) A firm’s stock price experiences a -1.15% return when our input price change measure is in the top tercile of price increases. Using a structural model, we relate firm markup to this stock price reaction. (4) Firms pass through 70% of their input price changes to output prices in the same quarter. | ||||||||||||||||||||
32 | investments | Hard to Process: Atypical Firms and the Cross-Section of Expected Stock Returns | Weibels, Sebastian | University of Cologne, Germany | atypical firms, processing difficulty, return predictability, mispricing, machine learning | Theories of limited attention predict that investors rely on typical patterns to navigate high-dimensional firm information, making atypical firms hard to process. To quantify this difficulty, we propose a data-driven measure of how atypical a firm's combination of characteristics is using an autoencoder (ATYP). The model learns the typical patterns that describe most firms, and ATYP aggregates the deviations those patterns cannot explain. Unlike measures of disclosure or organizational complexity, ATYP captures the processing difficulty of the underlying information. Empirically, we document that ATYP strongly predicts future returns. A decile portfolio that sells high-ATYP firms and buys low-ATYP firms earns 1.47% per month (equal-weighted) and 0.82% (value-weighted). The effect strengthens precisely where investor attention is low and arbitrage is limited, suggesting mispricing as the explanation. | ||||||||||||||||||||
33 | investments | Pump and Dump: Price Manipulation in Experimental Markets | Kluger, Brian; Saglam, Mehmet | University of Cincinnati, United States of America | price manipulation, social networks | We study how social media messaging affects asset markets using experimental methods. Participants trade in markets with asymmetric information, some markets with and some without the ability to send anonymous public messages. Rather than improving market efficiency through information sharing, we find that messaging facilitates profitable pump-and-dump strategies. Informed traders systematically post misleading messages to manipulate prices. These manipulation schemes are frequently successful, with price manipulators earning substantially more than other informed traders. We also observe deceptive strategies by uninformed traders, though these were generally unprofitable. Both successful and unsuccessful manipulation schemes reduced market efficiency, highlighting an important consequence of investors using social media for financial communication. | ||||||||||||||||||||
34 | investments | The Elusive CAPM: Idiosyncratic News and the Tilt of the Security Market Line | Upenieks, Adam | University of Calgary, Canada | CAPM, Idiosyncratic, News, Beta, Predictability | The capital asset pricing model (CAPM) performs poorly empirically, as market risk (beta) is weakly related to average excess returns. In low news periods, iden tified using idiosyncratic news from the Dow Jones Newswire, market betas have a strong and positive relation with average returns. Higher beta firms earn lower returns to idiosyncratic news, and individual firm betas are consistently lower on days with idiosyncratic news. Consistent with an attention-based mechanism, the beta-return relation is positive when attention to market-wide idiosyncratic news is low relative to macroeconomic attention, and reverses when idiosyncratic attention is high. Hybrid “betting-against-beta” trading strategies exploiting these periods earn high returns. I conclude that waves of high aggregate idiosyncratic news obscure the performance of the CAPM at the firm level and significantly influence asset pricing. | ||||||||||||||||||||
35 | investments | The Fixed Disposition Effect | Ouyang, Qinglin (1); Ouyang, Shumiao (2) | 1: Stockholm University, Sweden; 2: Saïd Business School, University of Oxford | Disposition effect, Investment style, Realization preference, FinTech | We revisit the disposition effect and argue that it is best understood not as a primitive behavioral bias, but as a reduced-form outcome of stable investment styles. Using a unique inter-linked dataset that combines a large-scale experiment with real-world mutual fund transactions, we document strong within-investor persistence in disposition behavior across time and contexts. This persistence is largely driven by a fixed investment style: contrarian investors exhibit a substantially stronger disposition effect, while it is minimal for momentum investors. Investment style explains far more variation in the disposition effect than standard demographic and socioeconomic characteristics. By contrast, realization preference is generally shared. We provide some of the first field evidence that it accounts for roughly 10% of the bias via a sharp jump at the zero-return threshold. Overall, our findings suggest that the disposition effect often emerges as a structural outcome of price-based trading rules, rather than a generic behavioral bias. | ||||||||||||||||||||
36 | investments | The Global Latent Risk Factor in Corporate Debt Distress: Frailty and Spillover Effects | Lee, Yanru | The Hoover Institution, Stanford University, United States of America | Corporate Default Clustering, Text Analytics | This paper employs a dataset containing a comprehensive international coverage of corporate default events. The dataset is primarily constructed using large language models on newspaper articles based on a dictionary of keywords that reflect corporate debt distress. Based on this dataset, I show strong evidence of a common global latent risk (frailty) factor that impacts corporate debt distress risk worldwide. The global latent risk factor identifies substantial common variation among separately estimated dynamic latent risk (frailty) factors of firms at the country level. Estimations of country frailty factors control for observable firm fundamentals capturing systemic risk and omitted macroeconomic factors. Commonalities among country frailty factors highlight global systemic risk. Observable global factors and financing variables can only explain up to 25% of global frailty, indicating the vulnerability of global corporate credit markets to common latent systemic risk. The findings also detect cross-country corporate default risk spillovers, underscoring the international interconnectedness of corporate distress risk. | ||||||||||||||||||||
37 | investments | The Social Risks of Generative AI | Renjie, Rex Wang (1,2); Ceccarelli, Marco (1) | 1: Vrije Universiteit (VU) Amsterdam, Netherlands, The; 2: Tinbergen Institute | Artificial Intelligence, Social Risks, Tail Risks | This paper shows that the equity market prices the novel social risks associated with generative AI. We exploit the release of ChatGPT as an information shock that updated investor beliefs about AI-related tail risks. Using pre-event ESG scores as proxies for firms' social risk management, we show that, controlling for AI productivity exposure, low-ESG firms underperform high-ESG firms by 4 percentage points over the two weeks following the release. The effect is concentrated in the Social pillar, particularly data privacy and security. Increases in option-implied downside risk indicate that changes in discount rates are the channel. A low-minus-high ESG portfolio of AI-exposed firms earns significant alphas during 2023–2024, suggesting that investors demand compensation for bearing AI-related downside risks. Our findings are consistent with a tail-risk model in which social risks of AI are priced through discount rates. | ||||||||||||||||||||
38 | banking | Banking competition and regulation with diverse business models | Eccles, Peter (1); Siciliani, Paolo (1); Grout, Paul (2); Zalewska, Ania (3) | 1: Bank of England. UK; 2: University of Bristol, UK; 3: University of Leicester, UK | Banking regulation, competition, capital requirements, deposit rate ceiling, shadow banking, Open Banking | We develop a model of banking competition where there is a partition between passive depositors who always deposit funds at the same bank and active depositors who can observe and act on all deposit rates in the market. This partition leads to two opposite business models where banks choose either to be (i) a monopolistic bank and serve only passive depositors or (ii) a competitive bank and also serve active depositors. Prudential regulation needs to account for its impact on the relative attractiveness of these two business models. We show that this additional effect, of banks switching between business models, can offset the traditional impact of capital requirements, in that an increase in capital requirements can trigger an intensification of competition, which in turn can increase overall risk-taking. Similarly, the imposition of a deposit-rate ceiling may render the competitive business model comparatively more appealing. We also show that in this situation the introduction of shadow banks has the potential to reduce rather than increase overall risk-taking. | ||||||||||||||||||||
39 | banking | Leader Bias in State Support for Startups | Li, Andy (1,2) | 1: University of Amsterdam, Netherlands, The; 2: Dutch Central Bank, The | Venture Capital, Entrepreneurial Finance, Banking, Loan Guarantee, Public Policy | Governments deploy substantial resources to support innovative startups. Support effectiveness depends critically on which firms receive it, yet little is known about the allocation process. This paper introduces leader bias: public resources disproportionately flow to startups backed by leading investors for reasons unrelated to startup quality. I study leader bias in the context of government loan guarantees, an increasingly important tool for startup finance. My stylized model shows that leader bias arises under application frictions and imperfect observability of startup quality, diverting public resources away from constrained-yet-promising firms. Using confidential credit-registry data matched to venture capital (VC) records, I find that guarantees disproportionately go to startups backed by top-tier VCs. These recipients already enjoy easier credit access without guarantees, do not exhibit different credit risk or innovation output, and are weaker firms within top-tier VC portfolios. To probe mechanisms, I exploit sudden expansions of guarantee programs and find that startups backed by top-tier VCs benefit more from relationship lending and program know-how. A diffusion model provides plausibly causal evidence that program knowledge spreads through VC syndication networks and drives guarantee adoption. These findings have direct implications for program design. Overall, intermediation and informational frictions systematically shape the allocation of public resources. | ||||||||||||||||||||
40 | banking | Regulatory Divergence and Bank Capital Flows | Cisneros, Lucas (1); Gutierrez, Bryan (2) | 1: Superintendencia de Banca, Seguros y AFPs; 2: University of Minnesota | regulatory divergence, bank capital flows, credit allocation, uneven playfield | How does cross-country divergence in banking regulation shape domestic banking systems? We study whether an increase in host-country capital requirements significantly rebalances the competitive landscape toward global banks that are not subject to host regulation. Using novel Peruvian data on global bank lending that relies on representative offices, we uncover that this locally unregulated organizational form accounts for nearly 25% of corporate dollar credit. We find that higher host-country capital requirements significantly reshape credit allocation: banks outside host regulation expand by 7-10pp relative to locally regulated lenders, even for the same borrower. | ||||||||||||||||||||
41 | banking | Technology, Online Banks, and Credit Market Segmentation | Nam, Rachel J. | USI Lugano & Swiss Finance Institute, Switzerland | Banks, Online Banks, FinTech, Credit Market Segmentation, Digital Platforms, Adverse Selection | How does online bank expansion (digital-only depository institutions that originate loans without human intermediation) affect consumer credit market structure? Using loan-level data from Germany, we show that online banks cherry-pick low-risk borrowers, generating adverse selection for traditional banks. We develop and test a model in which online banks have lower costs but weaker screening because they rely solely on hard information. Online banks offer substantially lower rates to low-risk borrowers, but this advantage declines with credit risk, creating a crossover point beyond which traditional banks become more competitive. Using historical branch density as an instrument, we show that supply-driven screening differences contribute to this pattern. Extending the framework to fintech lenders reveals market segmentation: online banks serve the lowest-risk borrowers, traditional banks the medium-risk segment, and fintechs the highest-risk segment. Over time, the traditional online rate gap widens, consistent with deteriorating borrower pools at traditional banks. A shift-share design---combining pre-determined district banking composition with national branch consolidation rates---provides causal support. Our findings highlight that technological development in credit markets can generate important distributional consequences. | ||||||||||||||||||||
42 | behavioral finance | Context-Dependent Memory and Disagreement: Evidence from Household Inflation Expectations | Bao, Yanlin | Singapore Management University, Singapore | Disagreement, Memory, Inflation Expectations, Household Beliefs | This paper documents a novel cognitive source of disagreement: heterogeneous recall of past experiences due to context-dependent memory contributes to divergence in household beliefs. The intuition is illustrated through a stylized model à la Wachter and Kahana (2024). Empirically, using local weather as contextual cues, I construct a measure of memory-based disagreement in U.S. households’ inflation expectations and show that it explains survey-based disagreement across forecast horizons, with stronger effects when disagreement is driven by unusual contexts. The findings align with established memory regularities, generalize to households in other countries, and remain robust across alternative specifications. I further show that heightened disagreement arising from heterogeneous memory weakens monetary policy transmission. These results highlight a previously overlooked cognitive friction shaping disagreement in households’ expectations. | ||||||||||||||||||||
43 | behavioral finance | Neuro-Structured Learning: A Cognitive Architecture for Building Financial Competence | Angelov, Alexander; Atanasov, Vladimir; Carlson, Kurt | William & Mary, United States of America | cognitive depletion, active recovery, EEG, local sleep | We propose a Neuro-Structured Learning framework (NSL), which incorporates the management of cognitive resources in finance education and professional training programs. The framework rests on: 1) a measurement foundation to track real‑time cognitive resource levels and depletion/recovery rates; 2) resource usage optimization, which reduces the cognitive costs of complex learnings tasks by “banking” elemental learning assets until they become low‑cost routines; and 3) resource generation and recovery, which prescribes “effective active recovery” (EAR) tasks and practices to rebuild capacity and optimize performance. The NSL provides a scalable architecture for courses, programs, and corporate training that accelerates the transition from novice to competent practitioner while reducing error rates, training waste, and burnout. | ||||||||||||||||||||
44 | behavioral finance | The Exploring Human | Pirinsky, Christo | University of Central Florida, United States of America | Exploration, evolutionary theory, behavioral biases, firm strategy | This talk presents a novel theory of choice centered around the idea that exploration is at the heart of human decision-making. Through interactions with their increasingly complex environment, people and organizations learn their capabilities, strengths, and limitations. Exploration can be incentivized in two distinct ways. The first one is an innate propensity for exploration (PEX), which makes exploration intrinsically rewarding. Curiosity and boredom, evident in humans as well as animals, are emotions facilitating such rewards. An alternative approach for incentivizing exploration is the adoption of biased perceptions of reality. Unlike PEX, this approach is flexible as people, organizations, and societies can fine-tune their beliefs to achieve desired levels of exploration. Thus, biased perceptions serve the important purpose of fine-tuning exploratory behavior when inherited incentives are sub-optimal. Numerous behavioral patterns that appear irrational at first glance, such as context-dependent choice, self-sabotage and even superstitions, could be rationalized as effective mechanisms guiding individual exploration. Exploration also provides an explanation of economic (hyper) activity. Entrepreneurs start new businesses not only because they believe they have a great idea but also because they would like to explore their abilities to develop and commercialize new ideas. Managers experiment with a new product line not only because they believe it would be profitable, but also because they would like to discover the relative advantages of their enterprise. We are designed to constantly explore an unknown and vastly mysterious world and, in the process, discover and develop who we are. Some of these ideas are developed in the book “A Theory of Dynamic Preferences: Using Economics and Behavioral Sciences to Explore Bias, Irrationality, and Choice” published by Palgrave Macmillan, a part of Springer Nature. | ||||||||||||||||||||
45 | ||||||||||||||||||||||||||
46 | ||||||||||||||||||||||||||
47 | ||||||||||||||||||||||||||
48 | ||||||||||||||||||||||||||
49 | ||||||||||||||||||||||||||
50 | ||||||||||||||||||||||||||
51 | ||||||||||||||||||||||||||
52 | ||||||||||||||||||||||||||
53 | ||||||||||||||||||||||||||
54 | ||||||||||||||||||||||||||
55 | ||||||||||||||||||||||||||
56 | ||||||||||||||||||||||||||
57 | ||||||||||||||||||||||||||
58 | ||||||||||||||||||||||||||
59 | ||||||||||||||||||||||||||
60 | ||||||||||||||||||||||||||
61 | ||||||||||||||||||||||||||
62 | ||||||||||||||||||||||||||
63 | ||||||||||||||||||||||||||
64 | ||||||||||||||||||||||||||
65 | ||||||||||||||||||||||||||
66 | ||||||||||||||||||||||||||
67 | ||||||||||||||||||||||||||
68 | ||||||||||||||||||||||||||
69 | ||||||||||||||||||||||||||
70 | ||||||||||||||||||||||||||
71 | ||||||||||||||||||||||||||
72 | ||||||||||||||||||||||||||
73 | ||||||||||||||||||||||||||
74 | ||||||||||||||||||||||||||
75 | ||||||||||||||||||||||||||
76 | ||||||||||||||||||||||||||
77 | ||||||||||||||||||||||||||
78 | ||||||||||||||||||||||||||
79 | ||||||||||||||||||||||||||
80 | ||||||||||||||||||||||||||
81 | ||||||||||||||||||||||||||
82 | ||||||||||||||||||||||||||
83 | ||||||||||||||||||||||||||
84 | ||||||||||||||||||||||||||
85 | ||||||||||||||||||||||||||
86 | ||||||||||||||||||||||||||
87 | ||||||||||||||||||||||||||
88 | ||||||||||||||||||||||||||
89 | ||||||||||||||||||||||||||
90 | ||||||||||||||||||||||||||
91 | ||||||||||||||||||||||||||
92 | ||||||||||||||||||||||||||
93 | ||||||||||||||||||||||||||
94 | ||||||||||||||||||||||||||
95 | ||||||||||||||||||||||||||
96 | ||||||||||||||||||||||||||
97 | ||||||||||||||||||||||||||
98 | ||||||||||||||||||||||||||
99 | ||||||||||||||||||||||||||
100 |