|
on Big Data |
| By: | Xavier Gabaix; Ralph S. J. Koijen; Robert J. Richmond; Motohiro Yogo |
| Abstract: | Firm characteristics, based on accounting and financial market data, are commonly used to represent firms in economics and finance. However, investors collectively use a much richer information set beyond firm characteristics, including sources of information that are not readily available to researchers. We show theoretically that portfolio holdings contain all relevant information for asset pricing, which can be recovered under empirically realistic conditions. Such guarantees do not exist for other data sources, such as accounting or text data. We build on recent advances in artificial intelligence (AI) and machine learning (ML) that represent unstructured data (e.g., text, audio, and images) by high-dimensional latent vectors called embeddings. Just as word embeddings leverage the document structure to represent words, asset embeddings leverage portfolio holdings to represent firms. Thus, this paper is a bridge from recent advances in AI and ML to economics and finance. We explore various methods to estimate asset embeddings, including recommender systems, shallow neural network models such as Word2Vec, and transformer models such as BERT. We evaluate the performance of these models on three benchmarks that can be evaluated using a single quarter of data: predicting relative valuations, explaining the comovement of stock returns, and predicting institutional portfolio decisions. We also estimate investor embeddings (i.e., representations of investors and their strategies), which are useful for investor classification, performance evaluation, and detecting crowded trades. We discuss other applications of asset embeddings, including generative portfolios, risk management, and stress testing. Finally, we develop a framework to give an economic narrative to a group of similar firms, by applying large language models to firm-level text data. |
| JEL: | C53 G12 G23 |
| Date: | 2025–04 |
| URL: | https://d.repec.org/n?u=RePEc:nbr:nberwo:33651 |
| By: | Dimitrios Kanelis; Lars H. Kranzmann; Pierre L. Siklos |
| Abstract: | We analyze how financial stability concerns discussed during Federal Open Market Committee (FOMC) meetings influence the Federal Reserve’s monetary policy implementation and communication. Utilizing large language models (LLMs) to analyze FOMC minutes from 1993 to 2022, we measure both mandate-related and financial stability-related sentiment within a unified framework, enabling a nuanced examination of potential links between these two objectives. Our results indicate an increase in financial stability concerns following the Great Financial Crisis, particularly during periods of monetary tightening and the COVID-19 pandemic. Outside the zero lower bound (ZLB), heightened financial stability concerns are associated with a reduction in the federal funds rate, while within the ZLB, they correlate with a tightening of unconventional measures. Methodologically, we introduce a novel labeled dataset that supports a contextualized LLM interpretation of FOMC documents and apply explainable AI techniques to elucidate the model’s reasoning. |
| Keywords: | explainable artificial intelligence, financial stability, FOMC deliberations, monetary policy communication, natural language processing |
| JEL: | E44 E52 E58 |
| Date: | 2025–04 |
| URL: | https://d.repec.org/n?u=RePEc:een:camaaa:2025-26 |
| By: | Wang, Yuyan (Stanford U); Li, Pan (Georgia Institute of Technology); Chen, Minmin (Google, Inc) |
| Abstract: | Modern recommender systems use machine learning (ML) models to predict consumer preferences based on consumption history. Although these “black-box†models achieve impressive predictive performance, they often suffer from a lack of transparency and explainability. While explainable AI research suggests a tradeoff between the two, we demonstrate that combining large language models (LLMs) with deep neural networks (DNNs) can improve both. We propose LR-Recsys, which augments state-of-the-art DNN-based recommender systems with LLMs’ reasoning capabilities. LR-Recsys introduces a contrastive-explanation generator that leverages LLMs to produce human-readable positive explanations (why a consumer might like a product) and negative explanations (why they might not). These explanations are embedded via a fine-tuned AutoEncoder and combined with consumer and product features as inputs to the DNN to produce the final predictions. Beyond offering explainability, LR-Recsys also improves learning efficiency and predictive accuracy. To understand why, we provide insights using high-dimensional multi-environment learning theory. Statistically, we show that LLMs are equipped with better knowledge of the important variables driving consumer decision-making, and that incorporating such knowledge can improve the learning efficiency of ML models. Extensive experiments on three real-world recommendation datasets demonstrate that the proposed LR-Recsys framework consistently outperforms state-of-the-art black-box and explainable recommender systems, achieving a 3–14\% improvement in predictive performance. This performance gain could translate into millions of dollars in annual revenue if deployed on leading content recommendation platforms today. Our additional analysis confirms that these gains mainly come from LLMs’ strong reasoning capabilities, rather than their external domain knowledge or summarization skills. LR-RecSys presents an effective approach to combine LLMs with traditional DNNs, two of the most widely used ML models today. Specifically, we show that LLMs can improve both the explainability and predictive performance of traditional DNNs through their reasoning capability. Beyond improving recommender systems, our findings emphasize the value of combining contrastive explanations for understanding consumer preferences and guiding managerial strategies for online platforms. These explanations provide actionable insights for consumers, sellers, and platforms, helping to build trust, optimize product offerings, and inform targeting strategies. |
| Date: | 2025–02 |
| URL: | https://d.repec.org/n?u=RePEc:ecl:stabus:4234 |
| By: | Guy Stephane Waffo Dzuyo (Forvis Mazars, LORIA - Laboratoire Lorrain de Recherche en Informatique et ses Applications - Inria - Institut National de Recherche en Informatique et en Automatique - UL - Université de Lorraine - CNRS - Centre National de la Recherche Scientifique, SYNALP - Natural Language Processing : representations, inference and semantics - LORIA - NLPKD - Department of Natural Language Processing & Knowledge Discovery - LORIA - Laboratoire Lorrain de Recherche en Informatique et ses Applications - Inria - Institut National de Recherche en Informatique et en Automatique - UL - Université de Lorraine - CNRS - Centre National de la Recherche Scientifique); Gaël Guibon (LORIA - Laboratoire Lorrain de Recherche en Informatique et ses Applications - Inria - Institut National de Recherche en Informatique et en Automatique - UL - Université de Lorraine - CNRS - Centre National de la Recherche Scientifique, LIPN - Laboratoire d'Informatique de Paris-Nord - CNRS - Centre National de la Recherche Scientifique - Université Sorbonne Paris Nord, SYNALP - Natural Language Processing : representations, inference and semantics - LORIA - NLPKD - Department of Natural Language Processing & Knowledge Discovery - LORIA - Laboratoire Lorrain de Recherche en Informatique et ses Applications - Inria - Institut National de Recherche en Informatique et en Automatique - UL - Université de Lorraine - CNRS - Centre National de la Recherche Scientifique); Christophe Cerisara (SYNALP - Natural Language Processing : representations, inference and semantics - LORIA - NLPKD - Department of Natural Language Processing & Knowledge Discovery - LORIA - Laboratoire Lorrain de Recherche en Informatique et ses Applications - Inria - Institut National de Recherche en Informatique et en Automatique - UL - Université de Lorraine - CNRS - Centre National de la Recherche Scientifique); Luis Belmar-Letelier (Forvis Mazars) |
| Abstract: | The identification of the financial characteristics of industry sectors has a large importance in accounting audit, allowing auditors to prioritize the most important area during audit. Existing company classification standards such as the Standard Industry Classification (SIC) code allow to map a company to a category based on its activity and products. In this paper, we explore the potential of machine learning algorithms and language models to analyze the relationship between those categories and companies' financial statements. We propose a supervised company classification methodology and analyze several types of representations for financial statements. Existing works address this task using solely numerical information in financial records. Our findings show that beyond numbers, textual information occurring in financial records can be leveraged by language models to match the performance of dedicated decision tree-based classifiers, while providing better explainability and more generic accounting representations. We think this work can serve as a preliminary work towards semi-automatic auditing. Models, code, and a preprocessed dataset are publicly available for further research at https://github.com/WaguyMz/hybrid company classification |
| Keywords: | Machine Learning, Industry Sectors, Large Language Models, LLM Applications, Audit, Financial Statement |
| Date: | 2025–02–25 |
| URL: | https://d.repec.org/n?u=RePEc:hal:journl:hal-05031499 |
| By: | Lee, Chaewon; Gates, Kathleen |
| Abstract: | Machine learning (ML) has extended the scope of psychological research by enabling data-driven discovery of patterns in complex datasets, complementing traditional hypothesis-driven approaches and enriching individual-level prediction. As a principal subfield, supervised ML has advanced mental health diagnostics and behavior prediction through classification and regression tasks. However, the complexity of ML methodologies and the absence of established norms and standardized pipelines often limit its adoption among psychologists. Furthermore, the black-box nature of advanced ML algorithms obscures how decisions are made, making it difficult to identify the most influential variables. Automated ML (AutoML) addresses these challenges by automating key steps such as model selection and hyperparameter optimization, while enhancing interpretability through explainable AI. By streamlining workflows and improving efficiency, AutoML empowers users of all technical levels to implement advanced ML methods effectively. Despite its transformative potential, AutoML remains underutilized in psychological research, with no dedicated educational material available. This tutorial aims to bridge the gap by introducing AutoML to psychologists. We cover advanced AutoML methods, including combined algorithm selection and hyperparameter optimization (CASH), stacked ensemble generalization, and explainable AI. The utility of AutoML is demonstrated using the ‘H2O AutoML’ R package with publicly available psychological datasets, performing regression on multi-individual cross-sectional data and classification on single-individual time-series data. We also provide practical workarounds for ML methods currently unavailable in the package, allowing researchers to use alternative approaches when needed. These examples illustrate how AutoML democratizes ML, making it more accessible while providing advanced methodologies for psychological research. |
| Date: | 2025–04–18 |
| URL: | https://d.repec.org/n?u=RePEc:osf:osfxxx:j4xuq_v1 |
| By: | Matthew Smith; Francisco Alvarez |
| Abstract: | Machine learning (ML) is becoming an essential tool in economics, offering powerful methods for prediction, classification, and decision-making. This paper provides an intuitive introduction to two widely used families of ML models: tree-based methods (decision trees, Random Forests, boosting techniques) and neural networks. The goal is to equip practitioners with a clear understanding of how these models work, their strengths and limitations, and their applications in economics. Additionally, we briefly discuss some other methods, as support vector machines (SVMs) and Shapley values, highlighting their relevance in economic research. Rather than providing an exhaustive survey, this paper focuses on practical insights to help economists effectively apply ML in their work. |
| Date: | 2025–04 |
| URL: | https://d.repec.org/n?u=RePEc:fda:fdaddt:2025-03 |
| By: | Xian Wu |
| Abstract: | This paper proposes a new algorithm -- Trading Graph Neural Network (TGNN) that can structurally estimate the impact of asset features, dealer features and relationship features on asset prices in trading networks. It combines the strength of the traditional simulated method of moments (SMM) and recent machine learning techniques -- Graph Neural Network (GNN). It outperforms existing reduced-form methods with network centrality measures in prediction accuracy. The method can be used on networks with any structure, allowing for heterogeneity among both traders and assets. |
| Date: | 2025–04 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2504.07923 |
| By: | Fredy Pokou (CRIStAL, INOCS); Jules Sadefo Kamdem (MRE); Fran\c{c}ois Benhmad (MRE) |
| Abstract: | In an environment of increasingly volatile financial markets, the accurate estimation of risk remains a major challenge. Traditional econometric models, such as GARCH and its variants, are based on assumptions that are often too rigid to adapt to the complexity of the current market dynamics. To overcome these limitations, we propose a hybrid framework for Value-at-Risk (VaR) estimation, combining GARCH volatility models with deep reinforcement learning. Our approach incorporates directional market forecasting using the Double Deep Q-Network (DDQN) model, treating the task as an imbalanced classification problem. This architecture enables the dynamic adjustment of risk-level forecasts according to market conditions. Empirical validation on daily Eurostoxx 50 data covering periods of crisis and high volatility shows a significant improvement in the accuracy of VaR estimates, as well as a reduction in the number of breaches and also in capital requirements, while respecting regulatory risk thresholds. The ability of the model to adjust risk levels in real time reinforces its relevance to modern and proactive risk management. |
| Date: | 2025–04 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2504.16635 |
| By: | Sohom Ghosh; Sudip Kumar Naskar |
| Abstract: | Government fiscal policies, particularly annual union budgets, exert significant influence on financial markets. However, real-time analysis of budgetary impacts on sector-specific equity performance remains methodologically challenging and largely unexplored. This study proposes a framework to systematically identify and rank sectors poised to benefit from India's Union Budget announcements. The framework addresses two core tasks: (1) multi-label classification of excerpts from budget transcripts into 81 predefined economic sectors, and (2) performance ranking of these sectors. Leveraging a comprehensive corpus of Indian Union Budget transcripts from 1947 to 2025, we introduce BASIR (Budget-Assisted Sectoral Impact Ranking), an annotated dataset mapping excerpts from budgetary transcripts to sectoral impacts. Our architecture incorporates fine-tuned embeddings for sector identification, coupled with language models that rank sectors based on their predicted performances. Our results demonstrate 0.605 F1-score in sector classification, and 0.997 NDCG score in predicting ranks of sectors based on post-budget performances. The methodology enables investors and policymakers to quantify fiscal policy impacts through structured, data-driven insights, addressing critical gaps in manual analysis. The annotated dataset has been released under CC-BY-NC-SA-4.0 license to advance computational economics research. |
| Date: | 2025–04 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2504.13189 |
| By: | Wei Jiao |
| Abstract: | We introduce a new machine learning approach to detect value-relevant foreign information for both domestic and multinational companies. Candidate foreign signals include lagged returns of stock markets and individual stocks across 47 foreign markets. By training over 100, 000 models, we capture stock-specific, time-varying relationships between foreign signals and U.S. stock returns. Foreign signals exhibit out-of-sample return predictability for a subset of U.S. stocks across domestic and multinational companies. Valuable foreign signals are not concentrated in those largest foreign markets nor foreign firms in the same industry as U.S. firms. Signal importance analysis reveals the price discovery of foreign information is significantly slower for information from emerging and low-media-coverage markets and among stocks with lower foreign institutional ownership but is accelerated during the COVID-19 crisis. Our study suggests that machine learning-based investment strategies leveraging foreign signals can emerge as important mechanisms to improve the market efficiency of foreign information. |
| Date: | 2025–04 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2504.07855 |
| By: | Simon Jehnen; Joaqu\'in Ordieres-Mer\'e; Javier Villalba-D\'iez |
| Abstract: | Recent advancements in information availability and computational capabilities have transformed the analysis of annual reports, integrating traditional financial metrics with insights from textual data. To extract valuable insights from this wealth of textual data, automated review processes, such as topic modeling, are crucial. This study examines the effectiveness of BERTopic, a state-of-the-art topic model relying on contextual embeddings, for analyzing Item 7 and Item 7A of 10-K filings from S&P 500 companies (2016-2022). Moreover, we introduce FinTextSim, a finetuned sentence-transformer model optimized for clustering and semantic search in financial contexts. Compared to all-MiniLM-L6-v2, the most widely used sentence-transformer, FinTextSim increases intratopic similarity by 81% and reduces intertopic similarity by 100%, significantly enhancing organizational clarity. We assess BERTopic's performance using embeddings from both FinTextSim and all-MiniLM-L6-v2. Our findings reveal that BERTopic only forms clear and distinct economic topic clusters when paired with FinTextSim's embeddings. Without FinTextSim, BERTopic struggles with misclassification and overlapping topics. Thus, FinTextSim is pivotal for advancing financial text analysis. FinTextSim's enhanced contextual embeddings, tailored for the financial domain, elevate the quality of future research and financial information. This improved quality of financial information will enable stakeholders to gain a competitive advantage, streamlining resource allocation and decision-making processes. Moreover, the improved insights have the potential to leverage business valuation and stock price prediction models. |
| Date: | 2025–04 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2504.15683 |
| By: | Lidia Ceriani; Chiara Gigliarano; Paolo Verme |
| Abstract: | Multidimensional indexes are ubiquitous, and popular, but present non-negligible normative choices when it comes to attributing weights to their dimensions. This paper provides a more rigorous approach to the choice of weights by defining a set of desirable properties that weighting models should meet. It shows that Bayesian Networks is the only model across statistical, econometric, and machine learning computational models that meets these properties. An example with EU-SILC data illustrates this new approach highlighting its potential for policies. |
| Date: | 2025–04 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2504.06012 |
| By: | Biewen, Martin (University of Tuebingen); Glaisner, Stefan (University of Tübingen) |
| Abstract: | This paper utilises distributional random forests as a flexible machine learning method for analysing income distributions. Distributional random forests avoid parametric assumptions, capture complex interactions among covariates, and, once trained, provide full estimates of conditional income distributions. From these, any type of distributional index such as measures of location, inequality and poverty risk can be readily computed. They can also efficiently process grouped income data and be used as inputs for distributional decomposition methods. We consider four types of applications: (i) estimating income distributions for granular population subgroups, (ii) analysing distributional change over time, (iii) spatial smoothing of income distributions, and (iv) purging spatial income distributions of differences in spatial characteristics. Our application based on the German Microcensus provides new results on the socio-economic and spatial structure of the German income distribution. |
| Keywords: | small area estimation, poverty, inequality, grouped income data |
| JEL: | D31 C55 I3 |
| Date: | 2025–03 |
| URL: | https://d.repec.org/n?u=RePEc:iza:izadps:dp17774 |
| By: | Yunhua Pei; John Cartlidge; Anandadeep Mandal; Daniel Gold; Enrique Marcilio; Riccardo Mazzon |
| Abstract: | Accurate financial market forecasting requires diverse data sources, including historical price trends, macroeconomic indicators, and financial news, each contributing unique predictive signals. However, existing methods often process these modalities independently or fail to effectively model their interactions. In this paper, we introduce Cross-Modal Temporal Fusion (CMTF), a novel transformer-based framework that integrates heterogeneous financial data to improve predictive accuracy. Our approach employs attention mechanisms to dynamically weight the contribution of different modalities, along with a specialized tensor interpretation module for feature extraction. To facilitate rapid model iteration in industry applications, we incorporate a mature auto-training scheme that streamlines optimization. When applied to real-world financial datasets, CMTF demonstrates improvements over baseline models in forecasting stock price movements and provides a scalable and effective solution for cross-modal integration in financial market prediction. |
| Date: | 2025–04 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2504.13522 |
| By: | Stella Jia; Abhishek Nagaraj |
| Abstract: | Large Language Models (LLMs) have demonstrated remarkable capabilities in text generation, but their performance may be influenced by the datasets on which they are trained, including potentially unauthorized or pirated content. We investigate the extent to which data access through pirated books influences LLM responses. We test the performance of leading foundation models (GPT, Claude, Llama, and Gemini) on a set of books that were and were not included in the Books3 dataset, which contains full-text pirated books and could be used for LLM training. We assess book-level performance using the “name cloze” word-prediction task. To examine the causal effect of Books3 inclusion we employ an instrumental variables strategy that exploits the pattern of book publication years in the Books3 dataset. In our sample of 12, 916 books, we find significant improvements in LLM name cloze accuracy on books available within the Books3 dataset compared to those not present in these data. These effects are more pronounced for less popular books as compared to more popular books and vary across leading models. These findings have crucial implications for the economics of digitization, copyright policy, and the design and training of AI systems. |
| JEL: | K24 O36 |
| Date: | 2025–03 |
| URL: | https://d.repec.org/n?u=RePEc:nbr:nberwo:33598 |
| By: | Ben Weidmann; Yixian Xu; David J. Deming |
| Abstract: | We show that leadership skill with artificially intelligent (AI) agents predicts leadership skill with human groups. In a large pre-registered lab experiment, human leaders worked with AI agents to solve problems. Their performance on this “AI leadership test” was strongly correlated (ρ=0.81) with their causal impact as leaders of human teams, which we estimate by repeatedly randomly assigning leaders to groups of human followers and measuring team performance. Successful leaders of both humans and AI agents ask more questions and engage in more conversational turn-taking; they score higher on measures of social intelligence, fluid intelligence, and decision-making skill, but do not differ in gender, age, ethnicity or education. Our findings indicate that AI agents can be effective proxies for human participants in social experiments, which greatly simplifies the measurement of leadership and teamwork skills. |
| JEL: | J24 M54 O30 |
| Date: | 2025–04 |
| URL: | https://d.repec.org/n?u=RePEc:nbr:nberwo:33662 |
| By: | Dimitrios Kanelis; Pierre L. Siklos |
| Abstract: | We combine modern methods from Speech Emotion Recognition and Natural Language Processing with high-frequency financial data to precisely analyze how the vocal emotions and language of ECB President Mario Draghi affect the yields and yield spreads of major euro area economies. This novel approach to central bank communication reveals that vocal and verbal emotions significantly impact the yield curve, with effects varying in magnitude and direction. Our results reveal an important asymmetry in yield changes with positive signals raising German, French, and Spanish yields, while negative cues increase Italian yields. Our analysis of bond spreads and equity markets indicates that positive communication influences the risk-free yield component, whereas negative communication affects the risk premium. Additionally, our study contributes by constructing a synchronized dataset for voice and language analysis. |
| Keywords: | artificial intelligence, asset prices, communication, ECB, high-frequency data, speech emotion recognition |
| JEL: | E50 E58 G12 G14 |
| Date: | 2025–04 |
| URL: | https://d.repec.org/n?u=RePEc:een:camaaa:2025-25 |
| By: | Ma, Ji (The University of Texas at Austin) |
| Abstract: | Large language models (LLMs) increasingly serve as human-like decision-making agents in social science and applied settings. These LLM-agents are typically assigned human-like characters and placed in real-life contexts. However, how these characters and contexts shape an LLM's behavior remains underexplored. This study proposes and tests methods for probing, quantifying, and modifying an LLM's internal representations in a Dictator Game -- a classic behavioral experiment on fairness and prosocial behavior. We extract ``vectors of variable variations'' (e.g., ``male'' to ``female'') from the LLM's internal state. Manipulating these vectors during the model's inference can substantially alter how those variables relate to the model's decision-making. This approach offers a principled way to study and regulate how social concepts can be encoded and engineered within transformer-based models, with implications for alignment, debiasing, and designing AI agents for social simulations in both academic and commercial applications. |
| Date: | 2025–04–18 |
| URL: | https://d.repec.org/n?u=RePEc:osf:osfxxx:8p7wg_v1 |
| By: | Darin Christensen; Tamma Carleton; Esther Rolf; Cullen Molitor; Shopnavo Biswas; Karena Yan; Graeme Blair |
| Abstract: | Artisanal and small-scale mining (ASM) supplies livelihoods and critical minerals but has been linked to conflict and environmental degradation. We enable monitoring of this largely informal sector by creating high-resolution maps of ASM's footprint in Africa using machine learning models that integrate geographic features and satellite imagery. We find ASM is more extensive than documented: in five countries with on-the-ground surveys, we predict over 231, 000 1-km2 grid cells [±2 standard errors: 170, 153-297, 710] contain ASM activity – over 40 times that recorded by surveyors. Adapting methods for spatial domain adaptation, we map ASM across 20 total countries, estimating that 4% [2-8%] of territory and 17% [10-30%] of the population are impacted by ASM, which encroaches on a larger share of settlements and ecosystems than previously understood. |
| JEL: | Q32 Q49 |
| Date: | 2025–04 |
| URL: | https://d.repec.org/n?u=RePEc:nbr:nberwo:33646 |
| By: | Alejandro Lopez-Lira; Yuehua Tang; Mingyin Zhu |
| Abstract: | Large language models (LLMs) cannot be trusted for economic forecasts during periods covered by their training data. We provide the first systematic evaluation of LLMs' memorization of economic and financial data, including major economic indicators, news headlines, stock returns, and conference calls. Our findings show that LLMs can perfectly recall the exact numerical values of key economic variables from before their knowledge cutoff dates. This recall appears to be randomly distributed across different dates and data types. This selective perfect memory creates a fundamental issue -- when testing forecasting capabilities before their knowledge cutoff dates, we cannot distinguish whether LLMs are forecasting or simply accessing memorized data. Explicit instructions to respect historical data boundaries fail to prevent LLMs from achieving recall-level accuracy in forecasting tasks. Further, LLMs seem exceptional at reconstructing masked entities from minimal contextual clues, suggesting that masking provides inadequate protection against motivated reasoning. Our findings raise concerns about using LLMs to forecast historical data or backtest trading strategies, as their apparent predictive success may merely reflect memorization rather than genuine economic insight. Any application where future knowledge would change LLMs' outputs can be affected by memorization. In contrast, consistent with the absence of data contamination, LLMs cannot recall data after their knowledge cutoff date. |
| Date: | 2025–04 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2504.14765 |
| By: | Jinhui Li; Wenjia Xie; Luis Seco |
| Abstract: | This study introduces a dynamic investment framework to enhance portfolio management in volatile markets, offering clear advantages over traditional static strategies. Evaluates four conventional approaches : equal weighted, minimum variance, maximum diversification, and equal risk contribution under dynamic conditions. Using K means clustering, the market is segmented into ten volatility-based states, with transitions forecasted by a Bayesian Markov switching model employing Dirichlet priors and Gibbs sampling. This enables real-time asset allocation adjustments. Tested across two asset sets, the dynamic portfolio consistently achieves significantly higher risk-adjusted returns and substantially higher total returns, outperforming most static methods. By integrating classical optimization with machine learning and Bayesian techniques, this research provides a robust strategy for optimizing investment outcomes in unpredictable market environments. |
| Date: | 2025–03 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2504.02841 |
| By: | Boris Nikolov (University of Lausanne; Swiss Finance Institute; European Corporate Governance Institute (ECGI)); Norman Schuerhoff (Swiss Finance Institute - HEC Lausanne); Sam Wagner (University of Lausanne) |
| Abstract: | A key question in automating governance is whether machines can recover the corporate objective. We develop a corporate recovery theorem that establishes when this is possible and provide a practical framework for its application. Training a machine on a large dataset of firms' investment and financial decisions, we find that most neoclassical models fail to explain the data since the machine learns from managers to underestimate the shadow cost of capital. This bias persists even after accounting for financial frictions, intangible intensity, behavioral factors, and ESG. We develop an alignment measure that shows why managerial alignment with shareholder-value remains imperfect and how to debias managerial decisions. |
| Keywords: | Corporate Purpose, Inverse Reinforcement Learning |
| JEL: | D22 G30 L21 |
| Date: | 2025–03 |
| URL: | https://d.repec.org/n?u=RePEc:chf:rpseri:rp2523 |
| By: | Mengjie (Magie); Cheng; Elie Ofek; Hema Yoganarasimhan |
| Abstract: | We study how media firms can use LLMs to generate news content that aligns with multiple objectives -- making content more engaging while maintaining a preferred level of polarization/slant consistent with the firm's editorial policy. Using news articles from The New York Times, we first show that more engaging human-written content tends to be more polarizing. Further, naively employing LLMs (with prompts or standard Direct Preference Optimization approaches) to generate more engaging content can also increase polarization. This has an important managerial and policy implication: using LLMs without building in controls for limiting slant can exacerbate news media polarization. We present a constructive solution to this problem based on the Multi-Objective Direct Preference Optimization (MODPO) algorithm, a novel approach that integrates Direct Preference Optimization with multi-objective optimization techniques. We build on open-source LLMs and develop a new language model that simultaneously makes content more engaging while maintaining a preferred editorial stance. Our model achieves this by modifying content characteristics strongly associated with polarization but that have a relatively smaller impact on engagement. Our approach and findings apply to other settings where firms seek to use LLMs for content creation to achieve multiple objectives, e.g., advertising and social media. |
| Date: | 2025–04 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2504.13444 |
| By: | David Jaggi (Zurich University of Applied Sciences; University of Zurich - Department of Finance); Markus Leippold (University of Zurich; Swiss Finance Institute); Tingyu Yu (University of Zurich - Department Finance) |
| Abstract: | This study presents ClimateControversyBERT, a novel open-source language model for real-time detection and classification of corporate climate controversies (i.e., brown projects, misinformation, ambiguous actions) from financial news. Validated using RepRisk and Refinitiv metrics, the model effectively identifies inconsistencies between corporate climate commitments and actions as they emerge. We document significant negative market reactions to these controversies: firms experience an immediate average stock price drop of 0.68%, with further declines over subsequent weeks. The impact is intensified by high media visibility and is notably stronger for firms with existing emission reduction commitments, underscoring the market's penalty for perceived environmental failures. |
| Keywords: | Climate controversy, corporate greenwashing, natural language processing |
| JEL: | G14 |
| Date: | 2025–04 |
| URL: | https://d.repec.org/n?u=RePEc:chf:rpseri:rp2545 |
| By: | Yanampally Abhiram Reddy; Siddhi Agarwal; Vikram Parashar; Arshiya Arora |
| Abstract: | In the age of social media, understanding public sentiment toward major corporations is crucial for investors, policymakers, and researchers. This paper presents a comprehensive sentiment analysis system tailored for corporate reputation monitoring, combining Natural Language Processing (NLP) and machine learning techniques to accurately interpret public opinion in real time. The methodology integrates a hybrid sentiment detection framework leveraging both rule-based models (VADER) and transformer-based deep learning models (DistilBERT), applied to social media data from multiple platforms. The system begins with robust preprocessing involving noise removal and text normalization, followed by sentiment classification using an ensemble approach to ensure both interpretability and contextual accuracy. Results are visualized through sentiment distribution plots, comparative analyses, and temporal sentiment trends for enhanced interpretability. Our analysis reveals significant disparities in public sentiment across major corporations, with companies like Amazon (81.2) and Samsung (45.8) receiving excellent sentiment scores, while Microsoft (21.7) and Walmart (21.9) exhibit poor sentiment profiles. These findings demonstrate the utility of our multi-source sentiment framework in providing actionable insights regarding corporate public perception, enabling stakeholders to make informed strategic decisions based on comprehensive sentiment analysis. |
| Date: | 2025–04 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2504.15448 |
| By: | Geiger, Felix; Kanelis, Dimitrios; Lieberknecht, Philipp; Sola, Diana |
| Abstract: | Central bank communication has become a crucial tool for steering the monetary policy stance and shaping the outlook of market participants. Traditionally, analyzing central bank communication required substantial human effort, expertise, and resources, making the process time-consuming. The recent introduction of artificial intelligence (AI) methods has streamlined and enhanced this analysis. While fine-tuned language models show promise, their reliance on large annotated datasets is a limitation that the use of large language models (LLMs) combined with prompt engineering overcomes. This paper introduces the Monetary-Intelligent Language Agent (MILA), a novel framework that leverages advanced prompt engineering techniques and LLMs to analyze and measure different semantic dimensions of monetary policy communication. MILA performs granular classifications of central bank statements conditional on the macroeconomic context. This approach enhances transparency, integrates expert knowledge, and ensures rigorous statistical calculations. For illustration, we apply MILA to the European Central Bank's (ECB) monetary policy statements to derive sentiment and hawkometer indicators. Our findings reveal changes in the ECB's communication tone over time, reflecting economic conditions and policy adaptions, and demonstrate MILA's effectiveness in providing nuanced insights into central bank communication. A model evaluation of MILA shows high accuracy, flexibility, and strong consistency of the results despite the stochastic nature of language models. |
| Keywords: | Central bank communication, monetary policy, sentiment analysis, artificial intelligence, large language models |
| JEL: | C45 E31 E44 E52 E58 |
| Date: | 2025 |
| URL: | https://d.repec.org/n?u=RePEc:zbw:bubtps:316448 |