|
on Computational Economics |
| By: | Kirill Zernikov (New Economic School) |
| Abstract: | This paper studies empirical deep hedging for S&P 500 index options under a local downside-shortfall reward. It moves beyond performance comparison by asking what the learned hedge does, when it fails, and whether it can be made auditable. TD3 agents are compared with a daily-updated Black-Scholes delta hedge on the same option episodes. In walk-forward tests from 2015 to 2023, the agents usually learn a systematic delta haircut relative to Black-Scholes. The correction is explained by spot-implied-volatility co-movement and often improves accumulated reward and terminal downside variance, but it is regime-fragile: 2022 exposes losses in adverse daily states, while 2023 shows that underhedging can raise ordinary variance when option P&L is spot-dominated and the volatility channel is unusually weak. Symbolic regression distills the neural policies into compact formulas that can be traded out of sample; these formulas preserve much of the reward, downside-variance, and CVaR advantage over Black-Scholes, and sometimes sharpen it, but inherit the same fragility in difficult regimes. |
| Date: | 2026–05 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2605.21696 |
| By: | Zhe Sage Chen; Quanyan Zhu |
| Abstract: | We propose a game-theoretic framework for adaptive multi-agent intelligent systems. Unlike classical game theory, which often treats strategies as primitive objects chosen by perfectly rational agents, the proposed framework provides a mathematical foundation for studying equilibrium in NeuroAI and can be viewed as an extension of game theory under relaxed assumptions, including partial observability, bounded computation, and uncertainty. At its core, Multilevel Interactive Equilibrium (MIE) generalizes the classical Nash equilibrium to intelligent systems with internal computation. Rather than being defined solely at the level of observable behavior, equilibrium emerges when neural learning dynamics, cognitive representations, and behavioral strategies mutually stabilize between interacting agents. This framework applies uniformly to interactions between two biological brains, two artificial agents, or hybrid human-AI systems. We discuss applications of multilevel game theory to human-autonomous vehicle driving, human-machine interaction, human-large language model (LLM) interaction, and computational psychiatry. We also outline experimental strategies and computational methods for estimating MIE and discuss challenges and prospects for future research. |
| Date: | 2026–05 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2605.10505 |
| By: | Santiago Picasso |
| Abstract: | A stylized factin modern economies is that the more developed a country is, the greater the weight of the service sector.The economics of complexity has provided a new perspective that explains this growth in modern economies.However, thestudy of economic complexity through the standard measure of thecomplexity index presents an increasingly relevant omission in understanding the economic process and its growth.Ingeneral, the data used to measure the EconomicComplexity Index(ECI) are based on information about goods;however, there is a lack of informationon services.This paper proposes an ew methodology to retrieve information on the economic complexity in services.Forthis purpose, the US input-output matrix is used.This work is novel because, thanks to the structure of the data as a network, it is possible to infer them is sing information on the complexity of services. Using a machinelearning method, it ispossible to impute the complexity index for 146services, a level of disaggregation, that is strikingly higher than in other works.The index recovered by this method is consistent with previous results that found service sectors to be more complex than goods.The second result shows that the more restricted the core is in the center of the network, the greater the centrality of services and their complexity.Finally, the results confirm the relevance of the economic complexity index. However, the ECI forservices is better than the ECI for goods for predicting growth;aone-unit increase in the ECI of services increases GDP growth by more than 1 percentage point. |
| Keywords: | Economic Complexity; Services Sector; Input–Output Networks; Machine Learning; k-Nearest Neighbors; Structural Transformation; Economic Growth; Spatial Econometrics |
| JEL: | C45 C55 O11 O14 O47 L80 |
| Date: | 2026–02 |
| URL: | https://d.repec.org/n?u=RePEc:ude:wpaper:0126 |
| By: | Jagdish Tripathy; Marcus Buckmann |
| Abstract: | Instruction-tuned language models exhibit behavioural fairness in high-stakes decisions while retaining biased associations in their internal representations. However, whether these suppressed representations can affect model outputs - and whether such causal potency is symmetric across demographic groups - remains unknown. We investigate the use of open-weight models for mortgage underwriting using matched applications that differ only in racially-associated names and reveal a critical disconnect: models show no output-level bias, yet retain and amplify demographic representations across model layers. Through activation steering and novel cross-layer interventions, we demonstrate that this suppressed information is decision-relevant: when reinjected at critical layers, it produces near-complete decision reversals. Critically, this latent bias is asymmetric - steering interventions affect decisions in one demographic direction, while producing minimal effects in reverse - and susceptible to adversarial prompt engineering and parameter-efficient fine-tuning. These findings demonstrate that behavioural audits focused on outputs are insufficient: fair outputs can mask exploitable internal biases. They also motivate dual-layer testing frameworks combining output evaluation with representational analysis for AI governance in high-stakes decisions. |
| Date: | 2026–05 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2605.15217 |
| By: | Marco Gregnanin; Johannes De Smedt; Giorgio Gnecco; Maurizio Parton |
| Abstract: | Generating synthetic data for financial time series poses challenges, especially considering their non-stationary nature. Traditional statistical time series models normally assume weak stationarity. However, this assumption can constrain their effectiveness. Deep learning models, particularly Generative Adversarial Networks (GANs), have exhibited considerable potential in emulating complex probability distributions. GANs employ a generator-discriminator framework, where the generator creates data samples, while the discriminator distinguishes real from generated data. In this research, we introduce the Sig-Graph GAN model, which integrates the time-series signature, offering a structured summary of its temporal evolution; the Long Short-Term Memory network, capturing its inherent autoregressive structure; and Graph Neural Networks (GNNs), leveraging geometric patterns within the time-series data. To employ GNNs optimally, we use the visibility graph algorithm to derive a graph-based representation of the underlying time series. Numerical evaluations demonstrate that the Sig-Graph GAN model outperforms baseline methods in replicating the distribution of logarithmic returns across different stock exchanges. The integration of the graph structure with the autoregressive component effectively captures both geometric and temporal patterns embedded in time-series data. This research advances the field of GAN models for time series by introducing a model capable of leveraging both autoregressive properties and geometric structures for synthetic data generation. |
| Date: | 2026–05 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2605.22215 |
| By: | Mohammad Jalili Torkamani; Pedro Gomes; Amirmohammad Sadeghnejad; Jason Le |
| Abstract: | The film industry is characterized by significant financial uncertainty, where large production investments do not always guarantee commercial success. This study analyzes the relationship between release season, production budget, and movie financial performance using the Full TMDB Movies Dataset 2024. A data mining framework incorporating association rule mining, clustering, machine learning, and SHAP analysis was applied to identify key drivers of revenue and profitability. The results show that release season has limited predictive influence on revenue and return on investment (ROI). In contrast, production budget, popularity, and audience ratings are significantly more influential. Association rule mining revealed that high-budget films with poor ratings are strongly associated with negative ROI outcomes. Random Forest regression achieved substantially stronger predictive performance than Decision Tree regression, with an $R^2$ value of 0.652. SHAP analysis further confirmed that budget and popularity are the dominant predictors of box office revenue, while timing-related variables contribute minimally. These findings suggest that financial success in the film industry is driven more by production investment and market attention than by seasonal release strategies, providing practical insights for budgeting, release planning, and financial risk management. |
| Date: | 2026–05 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2605.12551 |
| By: | Kausar, Shafiya (INSEAD) |
| Abstract: | We evaluate LLM-augmented reinforcement learning for stock trading on Nasdaq- 100 (2019–2023) and report a previously unmeasured experimental phenomenon: the relationship between LLM signal coverage density and trading performance is non-monotonic, with a clearly identifiable harmful regime. In a controlled coverage sweep over {0%, 5%, 20%, 50%, 80%, 100%}, signal injection at 5% and 20% coverage degrades performance below the no-signal baseline, becoming net-positive only at ≥ 50% coverage. The FNSPID dataset’s 9.7% non-neutral coverage sits inside this harmful regime—meaning that for typical research configurations available today, adding LLM signals to the RL pipeline reduces returns. Beyond this density finding, we report three further negative results that the LLMRL trading literature has not adequately addressed. First, our LLM-augmented RL agent (158.11% cumulative return as a 3-seed ensemble) is outperformed by three standard non-RL baselines that prior work in this thread does not report: momentum top-10 (250.45%), equal-weight buy-and-hold (235.00%), and equal-weight monthly rebalanced (214.06%), all of which also exceed the Nasdaq- 100 buy-and-hold benchmark (164.52%). Second, we control for the daily-vs.- monthly rebalancing-frequency confound by deploying the same trained agents under matched-frequency monthly execution; the monthly variant underperforms its daily counterpart by 47pp (111.01% vs. 158.11%), confirming that the baseline gap is not driven by transaction-cost differences. Third, a v3-matched ablation finds that removing the CVaR tail-risk constraint produces a difference within the seedto- seed variability of the experiment. Across two independent runs, the sign of this difference flipped, providing direct empirical evidence that the algorithmic risk-tail machinery contributes no detectable return benefit in this setting. A regime decomposition reveals one clear win for the agent: in the 2023 recovery period, the 3-seed ensemble (52.6%) outperforms all non-RL baselines, suggesting the learned policy may have regime-specific advantages that single-window evaluation obscures. We argue that LLM-RL trading research should adopt non-RL baselines as standard practice, report signal coverage density as a first-class experimental variable, and decompose results by regime. Code and trained models are available at https: //anonymous.4open.science/r/signal-densi ty-llm-trading-9966/. |
| Date: | 2026–05–14 |
| URL: | https://d.repec.org/n?u=RePEc:osf:socarx:nxvdp_v1 |
| By: | Kamil Kashif; Robert \'Slepaczuk |
| Abstract: | This study develops and evaluates a deep reinforcement learning framework for dynamic portfolio allocation across global equity markets. The Soft Actor-Critic algorithm is used to learn continuous portfolio weights within a Markov Decision Process, incorporating transaction costs, turnover penalties, and diversification constraints into the reward function. Five model configurations are compared, varying in reward formulation, policy structure (flat versus hierarchical Dirichlet), portfolio constraints, and temporal encoder (LSTM versus Transformer), and evaluated via walk-forward optimization across sixteen out-of-sample folds spanning 2003-2026 on the Nasdaq-100, Nikkei 225, and Euro Stoxx 50. Results show that RL strategies achieve competitive risk-adjusted performance primarily in the Euro Stoxx 50, where statistically significant abnormal returns are observed, but the central hypothesis is only partially confirmed: no strategy achieves statistically significant excess returns relative to Buy and Hold under HAC-robust inference across all markets. Regime analysis reveals that RL adds the most value during periods of elevated uncertainty, while ensemble aggregation across markets improves risk-adjusted performance and confirms the benefits of geographic diversification. |
| Date: | 2026–05 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2605.17307 |
| By: | Marjan Petreski |
| Abstract: | This paper combines traditional panel econometrics with random forest machine learning to revisit the relationship between exchange rate regimes and economic growth for 27 transition economies over 1991-2019. Exploiting the Couharde-Grekou (2024) probabilistic synthesis classification, the random forest approach non-parametrically confirms and sharpens what fixed-effects and system GMM estimation establish parametrically intermediate exchange rate regimes consistently underperform fixed arrangements, with growth penalties ranging from -1.0 to -10.4 percentage points, while floating regimes show negative but largely insignificant differentials. Beyond regime effects, the machine learning analysis reveals that the intermediate regime penalty is sharpest precisely where institutions are weakest - non-parametric validation that institutional capacity, not regime label alone, determines whether exchange rate anchoring pays off. The regime-growth relationship is further concentrated in the pre-2003 stabilization era and is absent among EU member economies, suggesting the growth dividend from exchange rate anchoring eroded as institutional convergence advanced. Together, these findings demonstrate how machine learning variable importance metrics can corroborate and enrich causal inference from panel methods, while supporting the view that exchange rate anchoring carried a meaningful credibility dividend during the formative phase of transition. |
| Date: | 2026–05 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2605.17391 |
| By: | Werner, Sven; Trotter, Philipp |
| Abstract: | Development finance increasingly funds entrepreneurship in developing countries, but evidence of its impact on entrepreneurship is mixed. Existing studies analyze total development finance flows as entrepreneurship-specific development finance data did not previously exist. By training and validating a machine-learning classifier on development finance project descriptions (2000-2022; 5 million projects; 97% accuracy), we introduce a scalable, replicable measure of specific entrepreneurship-support development finance (ESDF). Crucially, this measure allows us to assess which entrepreneurship margins respond to development finance. In a 19-year panel of 50 developing countries, two-way fixed-effects regressions show that higher ESDF is associated with higher entrepreneurial intentions, while total development finance is not. ESDF is not significantly linked to early-stage entrepreneurial activity, however, suggesting conversion bottlenecks in current entrepreneurial processes. |
| Abstract: | Entwicklungsfinanzierung richtet sich zunehmend auf die Förderung von Entrepreneurship im Globalen Süden. Die makroökonomische Evidenz zur Wirksamkeit dieser Förderung ist bislang jedoch uneinheitlich. Bisherige Studien greifen auf aggregierte Daten zur Entwicklungsfinanzierung zurück, da spezifische Daten zur Entrepreneurship-Förderung bislang nicht verfügbar waren. In diesem Papier entwickeln wir ein skalierbares und replizierbares globales Maß für die Förderung von Entrepreneurship durch Entwicklungsfinanzierung (entrepreneurship-support development finance, ESDF). Dazu trainieren und validieren wir ein Machine-Learning-Klassifikationsmodell auf Basis der Beschreibungen von 5 Millionen Entwicklungshilfeprojekten aus den Jahren 2000 bis 2022 (Genauigkeit des Modells: 97 %). Mit diesem Maß kann die Wirkung von ESDF auf verschiedene Stufen des Gründungsprozesses untersucht werden. Auf Basis eines Panels von 50 Ländern über 19 Jahre zeigen Regressionen mit Länder- und Jahreseffekten, dass ein höheres ESDF-Volumen mit stärkeren Gründungsabsichten einhergeht, während sich für aggregierte Entwicklungsfinanzierung kein entsprechender Zusammenhang zeigt. Zugleich ergibt sich kein signifikanter Zusammenhang zwischen ESDF und der Gründungsaktivität. Dies spricht dafür, dass zusätzliche Förderung zwar die Gründungsneigung erhöht, sich aber nicht automatisch in tatsächliche Gründungen übersetzt. |
| Keywords: | Entrepreneurship-support development finance, international assistance, entrepreneurial intentions, early-stage entrepreneurship, machine learning classification |
| JEL: | F35 O19 L26 C23 C45 |
| Date: | 2026 |
| URL: | https://d.repec.org/n?u=RePEc:zbw:rwirep:341094 |
| By: | Christos Spyridon Koulouris; Carlo Campajola |
| Abstract: | In this paper, we investigate whether deep reinforcement-learning agents interacting in a shared optimal-execution environment can sustain supra-competitive outcomes, in the sense of achieving lower implementation shortfalls than the relevant game-theoretical competitive benchmark. We study a two-agent Almgren-Chriss liquidation game and examine how learned behavior depends on intra-episode environment feedback, the ability to interpret the mid-price and the agent's knoledge of the past. We first use ex-ante schedule-learning agents to remove intra-episode feedback and isolate what can arise when agents commit to complete liquidation trajectories before execution begins. We then allow agents to condition on the evolving state using a variety of DDQN architectures. We find that, when agents are given access to intra-episode history, especially recent prices and own past actions, supra-competitive outcomes become substantially more frequent and more persistent. These findings indicate that supra-competitive behavior in this execution game is driven not by multi-agent learning or by current price observation alone, but by feedback, memory, and state-contingent interaction along the realized execution path. |
| Date: | 2026–05 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2605.20348 |
| By: | Francesco A. Fabozzi; Dasol Kim; William N. Goetzmann |
| Abstract: | We introduce a novel approach to emotion modeling that shifts the focus from identification to evaluation, addressing the limitations of discrete classification in applied domains such as finance. By constructing a dataset of emotional intensity scores and fine-tuning open-weight generative language models to output continuous values from 0-100, we demonstrate a more expressive, generalizable framework for sentiment and emotion analysis. Our findings not only outperform classification baselines but also reveal surprising generalization capabilities and transfer effects to related constructs such as sentiment and arousal. This work contributes to the interdisciplinary recontextualization of NLP by introducing emotion intensity evaluation as an alternative to classification, arguing that this shift better aligns with the needs of domains--such as finance--where the degree of emotional content is central to interpretation and decision-making. |
| Date: | 2026–05 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2605.16613 |
| By: | Qinwen Zhu; Wen Chen; Nicolas Langren\'e |
| Abstract: | This paper develops a deep learning-based framework for pricing convertible bonds with path-dependent contractual features, namely downward conversion price reset and issuer call clauses under rolling-window trigger rules, which are widespread in the convertible bond market. We formulate the valuation problem as a path-dependent partial differential equation (PPDE), which explicitly captures the dependence of the convertible bond value on the historical path of the underlying asset and the dynamic evolution of the conversion price. We derive consistent PPDE formulations for three canonical underlying dynamics: geometric Brownian motion (GBM), constant elasticity of variance (CEV) and Heston stochastic volatility. We then construct a discrete-time dynamic programming scheme in which conditional expectations are approximated by neural networks, which remains tractable in such high-dimensional path-dependent setting. Empirical tests on China CITIC Bank Convertible Bond show that our framework produces stable and accurate prices and sensitivity patterns across all model specifications. Three key economic insights emerge: 1. Contractual features dominate underlying dynamics in determining convertible bond values. 2. The call provision decreases convertible bonds prices by truncating upside gains. 3. Counterintuitively, despite improving conversion terms, the downward reset provision further decreases the price of convertible bonds by lowering the effective call threshold and making early redemption more likely. The proposed PPDE-deep learning approach provides an efficient, flexible tool for pricing convertible bonds with complex path-dependent structures. |
| Date: | 2026–05 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2605.12189 |
| By: | Jesus Cañas; Emily Kerr |
| Abstract: | Learning how businesses use artificial intelligence (AI) helps policymakers understand changing economic conditions, particularly involving employment and productivity. |
| Keywords: | artificial intelligence (AI); labor; manufacturing; economic surveys |
| Date: | 2024–06–25 |
| URL: | https://d.repec.org/n?u=RePEc:fip:d00001:98445 |
| By: | Mathias Mesfin |
| Abstract: | This paper compares gradient boosting and long short-term memory (LSTM) architectures for intraday directional prediction in Micro E-Mini Nasdaq 100 futures (MNQ). Motivated by recent foundation-model research on financial candlestick data, including the Kronos architecture, we test whether five-minute OHLCV bar sequences contain exploitable sequential predictive structure at the scale of a single instrument dataset. Using 944 trading days from 2021-2025, four model configurations are evaluated under strict expanding-window walk-forward validation across three out-of-sample periods. The target variable is whether the session close exceeds the 10:30 AM open by more than ten points. No configuration produces statistically significant out-of-sample accuracy above the 51.8% base rate. Combined OOS accuracies range from 50.00% to 50.89% across gradient boosting variants, while the LSTM achieves 50.59%. Permutation tests yield p-values of 0.135 for the best gradient boosting model and 0.515 for the LSTM, indicating no statistically significant predictive edge. Feature importance instability across walk-forward folds suggests noise fitting rather than stable structural signal capture. The results indicate that four years of single-instrument five-minute OHLCV data are insufficient for reliable sequential ML-based intraday forecasting. The primary contribution is a documented evaluation of a Kronos-inspired architecture on a constrained real-world dataset, providing an empirical lower bound on data scale requirements for sequential financial ML. |
| Date: | 2026–05 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2605.17724 |
| By: | Lee, Kamwoo; Blankespoor, Brian; Newhouse, David |
| Abstract: | Fine-grained spatial data are critical for informed decision-making in domains ranging from economic planning to environmental management. However, many statistics are only available for coarse administrative units, necessitating techniques for fine-scale spatial disaggregation. This paper introduces a graph neural network (GNN) based framework for disaggregating aggregated indicators to a finer spatial resolution. The GNN approach leverages graph representations of spatial units to incorporate both feature information and spatial relationships, addressing challenges of heterogeneity and data sparsity. The approach also adopts the H3 hierarchical hexagonal indexing system to define fine-resolution cells, providing a globally consistent, multi-resolution spatial grid well suited to graph-based modeling. The paper demonstrates the framework using gross domestic product (GDP) as a representative example, disaggregating national or regional GDP to fine-resolution cells. The proposed methodology is applicable to a broad class of aggregate indicators, offering a flexible and scalable tool for spatial analysis of economic, social, and environmental statistics. The results show that the framework produces high-resolution estimates that are consistent with known aggregates and aligned with ancillary covariate patterns. This general-purpose approach to spatial disaggregation enables more detailed mapping of indicators like GDP and beyond, unlocking finer insights from coarse data. |
| Date: | 2026–04–23 |
| URL: | https://d.repec.org/n?u=RePEc:wbk:wbrwps:11360 |
| By: | Stefano Blando; Giorgio Fagiolo; Mauro Napoletano; Tania Treibich; Andrea Vandin |
| Abstract: | Agent-based models (ABMs) are increasingly used in macroeconomics, but their analysis still often relies on ad hoc Monte Carlo campaigns with heterogeneous statistical effort across parameter settings. We show how statistical model checking (SMC), implemented through MultiVeStA, can provide a principled analysis layer for a realistic macroeconomic ABM without rewriting the simulator in a dedicated formalism. Our case study is the heuristic-switching Keynes+Schumpeter(K+S) model, analysed hrough a transient sensitivity campaign over one-parameter sweeps, two macro observables (unemployment and GDP growth), and one auxiliary micro-level probe (market share) on the post-warmup phase of a 600-step horizon. The analysis is driven by reusable temporal queries, observable-specific precision targets, and confidence-based stopping rules that automatically determine the simulation effort required by each configuration. Results show a clear contrast across parameter families: macro-financial and structural sweeps produce the strongest transient effects, whereas several heuristic-rule sweeps remain much weaker under the same precision policy. More broadly, the paper shows that SMC can support reproducible and informative quantitative analysis of substantively rich economic ABMs, while making uncertainty estimates and simulation cost explicit parts of the reported results. |
| Date: | 2026–05 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2605.10447 |
| By: | Bj\"orn L\"ofdahl Grelsson |
| Abstract: | Historical Simulation (HS) and its extensions form a popular class of methods for estimating Value-at-Risk for portfolios of financial assets based on historical data. In this note, we seek to unify several ideas and models from throughout the literature into a single modeling framework. By explicitly defining a parametric model form for the asset returns and extracting the realized increments of the driving innovation process from historical data, we are able to reproduce the Historical Simulation, filtered Historical Simulation, and displaced Historical Simulation methods. This shows beyond a doubt that these methods need more underlying assumptions than what is often alluded to. |
| Date: | 2026–05 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2605.10066 |
| By: | Marco Gregnanin; Johannes De Smedt; Giorgio Gnecco; Maurizio Parton |
| Abstract: | Forecasting univariate time series in the financial market is a challenging endeavor. While numerous statistical and machine learning models have been introduced to address this challenge, they typically concentrate solely on analyzing temporal patterns within the time series data. In this research, we study the statistical significance of the inclusion of geometric patterns in enhancing forecasting accuracy within the context of time series analysis. We introduce the Time-Geometric model, a combination of models designed to exploit both geometric and temporal patterns. The contribution of this research lies in advancing the domain of univariate time series prediction, as demonstrated through extensive empirical evaluations. Our findings underscore that leveraging geometric patterns, captured through Graph Neural Networks, yields statistically significant improvements in forecasting accuracy. |
| Date: | 2026–05 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2605.21192 |
| By: | Emmanouil Sofianos (BETA - Bureau d'Économie Théorique et Appliquée - AgroParisTech - UNISTRA - Université de Strasbourg - Université de Haute-Alsace (UHA) - Université de Haute-Alsace (UHA) Mulhouse - Colmar - UL - Université de Lorraine - CNRS - Centre National de la Recherche Scientifique - INRAE - Institut National de Recherche pour l’Agriculture, l’Alimentation et l’Environnement); Thierry Betti (BETA - Bureau d'Économie Théorique et Appliquée - AgroParisTech - UNISTRA - Université de Strasbourg - Université de Haute-Alsace (UHA) - Université de Haute-Alsace (UHA) Mulhouse - Colmar - UL - Université de Lorraine - CNRS - Centre National de la Recherche Scientifique - INRAE - Institut National de Recherche pour l’Agriculture, l’Alimentation et l’Environnement); Theophilos Papadimitriou (DUTH - Democritus University of Thrace); Amélie Barbier-Gauchard (BETA - Bureau d'Économie Théorique et Appliquée - AgroParisTech - UNISTRA - Université de Strasbourg - Université de Haute-Alsace (UHA) - Université de Haute-Alsace (UHA) Mulhouse - Colmar - UL - Université de Lorraine - CNRS - Centre National de la Recherche Scientifique - INRAE - Institut National de Recherche pour l’Agriculture, l’Alimentation et l’Environnement); Periklis Gogas (DUTH - Democritus University of Thrace) |
| Abstract: | Forecasting public debt is essential for effective policymaking and economic stability, yet traditional approaches face challenges due to data scarcity. While machine learning (ML) has demonstrated success in financial forecasting, its application to macroeconomic forecasting remains underexplored, hindered by short historical time series and low-frequency (e.g., quarterly/annual) data availability. This study proposes a novel hybrid framework integrating dynamic stochastic general equilibrium (DSGE) modeling with ML techniques to address these limitations, focusing on the evolution of France's public debt. We first generate a large artificial macroeconomic dataset using an estimated DSGE model for France, which allows for efficient training of ML algorithms. These trained models are then applied to actual historical data for directional debt forecasting. The results show that the best machine learning model is an XGBoost achieving 90% accuracy, outperforming an elastic net model, used as benchmark. Our results highlight the viability of combining structural economic models with data-driven techniques to improve macroeconomic forecasting. |
| Keywords: | public debt, machine learning, France, forecasting, DSGE, DSGE forecasting France machine learning public debt |
| Date: | 2026–03–05 |
| URL: | https://d.repec.org/n?u=RePEc:hal:journl:hal-05620169 |
| By: | Aditya Retnanto (Asian Development Bank); Yohan Iddawela (Asian Development Bank); Elaine Tan (Asian Development Bank) |
| Abstract: | Systematic reviews and meta-analyses (SRMAs) are important tools for evidence synthesis but have historically required substantial manual effort, particularly during the data extraction phase. To address this bottleneck, we developed and evaluated an automated pipeline that utilizes large language models (LLMs) to ingest full text scientific articles and extract structured metadata. We benchmarked the performance of leading models, including Gemini 2.5 Pro, GPT-5, and Sonnet 4.0, across two distinct domains: mobile health interventions and education. Our results indicate that Gemini 2.5 Pro achieved the strongest performance in qualitative metadata extraction and outcome identification. However, quantitative metadata extraction remained a significant challenge. Models struggled to interpret complex data across multiple tables and failed to calculate effect sizes when only raw figures were reported. Crucially, we find that human annotators often applied implicit filtering criteria not documented in the coding manual, which made benchmarking the results challenging. We discuss the implications of these findings, emphasizing that while LLMs can accelerate the coding process, reliable automation requires significantly more prescriptive coding manuals to strictly steer model behavior and ensure fair benchmarking. |
| Keywords: | evidence synthesis automation;large language models (LLMs);data extraction benchmarking;systematic reviews and meta-analyses (SRMA) |
| JEL: | C88 |
| Date: | 2026–05–15 |
| URL: | https://d.repec.org/n?u=RePEc:ris:adbewp:022484 |
| By: | Lake Yang; Junwei Su; Jingfeng Zeng; Wenhao Lu; Xingzhi Qian; Weitong Zhang; Chuan Wu; Dunhong Jin |
| Abstract: | Herding -- where agents align their behaviors and act collectively -- is a central driver of market fragility and systemic risk. Existing approaches to quantify herding rely on price-correlation statistics, which inherently lag because they only detect coordination after it has already moved realised returns. We propose GeomHerd, a forward-looking geometric framework that bypasses this observability lag by quantifying coordination directly on upstream agent-interaction graphs. To generate these graphs, we treat a heterogeneous LLM-driven multi-agent simulator -- each financial trader instantiated by a persona-conditioned LLM call -- as a forecastable world, and evaluate the geometric pipeline on the Cividino--Sornette continuous-spin agent-based substrate as our headline financial testbed. By tracking the discrete Ollivier--Ricci curvature of these action graphs, GeomHerd captures the structural topology of emerging coordination. Theoretically, we establish a mean-field bridge mapping our graph-theoretic metric to CSAD, the classical macroscopic herding statistic, linking GeomHerd to downstream price-dispersion measurement. Empirically, GeomHerd anticipates herding long before aggregate market baselines: on the continuous-spin substrate, our primary detector fires a median of 272 steps before order-parameter onset; a contagion detector ($\beta_{-}$) recalls 65% of critical trajectories 318 steps early; and on co-firing trajectories the agent-graph signal precedes price-correlation-graph baselines by 40 steps. As a complementary indicator, the effective vocabulary of agent actions contracts during cascades. The geometric signature transfers out-of-domain to the Vicsek self-driven-particle model, and a curvature-conditioned forecasting head reduces cascade-window log-return MAE over detector-conditioned and price-only baselines. |
| Date: | 2026–05 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2605.11645 |
| By: | Eliseo Curcio |
| Abstract: | Selecting the right electricity market region for a hyperscale AI datacenter requires reasoning across live electricity prices, grid carbon intensity, technology cost trajectories, and causal grid dynamics -- a multi-step, multi-source analytical task that static knowledge benchmarks cannot evaluate. We introduce EnergyAgentBench, the first agentic benchmark grounded in live electricity market data for this problem class. The benchmark comprises 70 task variants across five families: datacenter siting under cost-carbon trade-offs (F1), long-horizon portfolio siting (F1-LH), lifetime LCOE ranking over multi-decade cost trajectories (F2), 30-year portfolio optimization (F2-LH), and causal grid diagnosis (F3). Tasks require 3 to 48 sequential tool calls against live endpoints from the QuarluxAI infrastructure platform, the U.S. Energy Information Administration (EIA), and the National Renewable Energy Laboratory (NREL) with ground truth derived from trained XGBoost cost-surface models (R^2 0.967--0.995) and the NREL Annual Technology Baseline 2024. We evaluate nine models across Anthropic, OpenAI, and HuggingFace over 1, 414 runs at three random seeds. Claude Sonnet 4.6 achieves the highest overall score (0.900) at one-quarter the cost of Claude Opus 4.7 (0.889). Claude Haiku 4.5 leads on long-horizon procedural siting (0.986), outperforming all frontier models including those costing 16x more per run. F3 Causal is the most discriminating family, with a 30.7-point spread between Sonnet (0.793) and Llama 3.3 70B (0.486), versus a 6.6-point spread on F1 Siting. A failure taxonomy of 135 coded failures identifies null-value integration in NREL ATB trajectories as the dominant failure mode (70%), followed by premature commitment on causal tasks (20%) and adversarial injection blindness (6%). Benchmark code, run trajectories, and the failure taxonomy dataset are publicly released. |
| Date: | 2026–05 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2605.15230 |
| By: | Dmitry Dagaev; Egor Ivanov; Petr Parshakov; Alexey Savvateev; Gleb Vasiliev |
| Abstract: | The emergence of large language models (LLMs) has spurred economists to study how humans and LLMs behave in strategic settings. We organized a series of round-robin tournaments in the Colonel Blotto game. This game attracts game theorists' attention due to high-dimensional action space and the absence of pure strategy Nash equilibria. In the first tournament, more than 200 human participants competed against one another. In the second tournament, several popular LLMs were invited to submit strategies. In the third tournament, we matched the number of LLM strategies to the number submitted by humans. We find that humans more often employ better-calibrated intermediate-level allocation heuristics and outperform the simpler, more stereotyped strategies submitted by LLMs. Strategic sophistication is key to success if and only if the necessary level of reasoning depth is reached, while lower and higher levels of reasoning offer no clear advantage over the primitive strategies. Among humans, field of study weakly predicts success: participants with STEM backgrounds perform better in the first tournament. Surprisingly, humans almost do not adjust their strategies across tournaments with different sets of opponents. This result suggests that humans base their choices primarily on the game's rules rather than on the identity of their opponents, treating LLMs much like human competitors. |
| Date: | 2026–05 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2605.22095 |
| By: | Firmin Ayivodji (International Monetary Fund); Etienne Briand (University of Quebec in Montreal); Kevin Moran (Laval University); Dalibor Stevanovic (University of Quebec in Montreal) |
| Abstract: | News media coverage of monetary policy is not a passive transcript of central-bank communication: it filters announcements, macroeconomic news, and editorial choices into narratives that move expectations and policy decisions. We embed media sentiment into a behavioral New-Keynesian model in which the central bank reacts to sentiment and sentiment follows an explicit law of motion. We construct monetary-policy sentiment indicators from more than 50, 000 Canadian newspaper articles using dictionary methods, transformer models, and a generative-AI framework. Media sentiment shifts household inflation and wage expectations, improves out-of-sample forecasts of GDP growth and inflation, and loads positively on the Bank of Canada's estimated Taylor rule once treated as endogenous. A Bayesian SVAR identifies anticipated and unanticipated monetary-policy shocks together with a narrative shock; the narrative shock contributes a non-trivial share of medium-horizon macroeconomic variance, and a counterfactual that shuts down the dynamic feedback from media sentiment attenuates the propagation of monetary policy to output and prices. |
| Keywords: | Monetary policy, text analysis, news media, machine learning, forecasting |
| JEL: | E52 E58 E71 D84 C32 C55 |
| Date: | 2026–05 |
| URL: | https://d.repec.org/n?u=RePEc:bbh:wpaper:26-03 |
| By: | Runyao Yu; Julia Lin; Derek W. Bunn; Jochen Stiasny; Wentao Wang; Yujie Chen; Tara Esterl; Peter Palensky; Jochen L. Cremer |
| Abstract: | Accurate and efficient imbalance electricity price forecasting is critical for industrial energy trading systems, especially as battery assets and automated bidding pipelines increasingly participate in balancing markets. However, real-time forecasting is complicated by nonlinear market-rule-based price formation, heterogeneous input signals, and incomplete data availability caused by communication delays, publication lags, and measurement outages. This paper proposes a market-rule-informed neural forecasting framework that embeds imbalance price formation rules into the latent space of an expressive neural network. The proposed framework preserves raw signal information while exploiting transparent market-rule priors. We further analyze operational robustness by removing price-component information and characterize how forecasting performance scales with input length and forecasting horizon. Experimental results show that the proposed model achieves competitive forecasting performance with substantially fewer trainable parameters and shorter training time than generic deep learning baselines. Experimental results show that the proposed model achieves competitive forecasting performance with substantially fewer trainable parameters and shorter training time than generic deep learning baselines, demonstrating that market-rule priors and expressive neural networks should be jointly used for accurate and computationally sustainable forecasting in industrial energy trading applications. The implementation is publicly available at https://runyao-yu.github.io/MRINN/. |
| Date: | 2026–05 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2605.09061 |
| By: | Lin William Cong; Ke Tang; Jingyuan Wang |
| Abstract: | We adapt attention-based neural networks and reinforcement learning to direct portfolio construction, allowing broader portfolio-management objectives (including non-time-additively separable ones) and in a data-driven way, searching over a much richer policy/strategy space than low-dimensional parametric rules or human-specified strategies. As arguably the first non-text-based, “large” GenAI model in Finance, AlphaPortfolio accommodates long- and short-range path dependence in firm and market states (e.g., using Transformer encoder), cross-asset information, flexible (path-dependent) objectives (incl. Sharpe ratio, which is non-additively separable across periods) for end-to-end (rather than step-by-step) optimizations. In U.S. equities, AlphaPortfolio yields superior out-of-sample performance (e.g., Sharpe ratio above two and risk-adjusted alpha over 13% with monthly rebalancing) robust under various market conditions and economic restrictions (e.g., exclusion of small/illiquid stocks) and over time. The gains come from the direct construction, effective sequence modeling, and cross-asset attention network. We further demonstrate AlphaPortfolio's flexibility to incorporate transaction costs, state interactions, and alternative objectives, before developing a polynomial-feature-sensitivity analysis to uncover key drivers of performance, including their rotation and nonlinearity. |
| JEL: | C14 C58 G11 G12 |
| Date: | 2026–05 |
| URL: | https://d.repec.org/n?u=RePEc:nbr:nberwo:35195 |
| By: | Bektemir Ysmailov (Nazarbayev University, Graduate School of Business) |
| Abstract: | 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. |
| Keywords: | financial constraints, generative AI (GenAI), large language models (LLMs), textual analysis, MD&A disclosures, corporate finance |
| JEL: | G30 G32 M41 C81 |
| Date: | 2026–01 |
| URL: | https://d.repec.org/n?u=RePEc:asx:nugsbw:2026-01 |
| By: | Michele Zampa (Geneva Graduate Institute) |
| Abstract: | Machine-learning models often perform poorly when asked to generalize beyond the support of the training distribution. This paper asks whether the same limitation shapes the research capabilities of autonomous large language model (LLM) agents: do they perform better when generating papers that follow research paradigms already well represented in the literature? I study this question using evidence from the Autonomous Policy Evaluation (APE) project, an open platform developed by the Social Catalyst Lab at the University of Zurich in which LLM agents generate empirical economic policy papers and compete in a tournament-style evaluation against human-written benchmarks. I construct a measure of literature support by locating each paper abstract in the semantic space of economics using a comprehensive corpus of English-language economics abstracts from OpenAlex. This measure captures whether a paper lies in a crowded or sparse region of the discipline’s existing research landscape. I then test whether literature support predicts tournament performance. I find that literature support shows a statistically significant positive association with performance for AI-generated papers, but not for human-written papers. Because outcomes are assigned by an LLM judge, this relationship could partly reflect evaluation bias toward more familiar topics. However, the absence of a comparable pattern among human papers suggests that the result is not purely judge-side. The evidence is more consistent with a production-side interpretation: autonomous research agents perform better when operating in areas that are more densely represented in the existing literature and, plausibly, in model training data. The findings shed some light on both the promise and the limits of agentic LLM systems as producers of scientific research. |
| Keywords: | LLM agents; AI-generated research; economics of science; semantic embeddings; scientific novelty |
| JEL: | O33 B41 A14 D83 |
| Date: | 2026–05–18 |
| URL: | https://d.repec.org/n?u=RePEc:gii:giihei:heidwp14-2026 |
| By: | Massimo Giannini |
| Abstract: | This paper assesses whether NASA Black Marble nightlight intensity can serve as an early indicator of annual taxable income at the Italian municipal level, where official data are released with a 12--18 month lag. Using a panel of 7{, }631 municipalities over 2012--2021, we compare four recurrent neural network architectures (LSTM, BiLSTM, GRU, Transformer) against six benchmarks: simple persistence, panel fixed effects, autoregressive distributed lag, and two spatial econometric specifications (SAR, Spatial Durbin) on a queen-contiguity matrix. Models are trained on 2012--2019 and evaluated out-of-sample on 2020--2021 with a cross-sectional Diebold--Mariano test. A single-layer GRU achieves a median forecast error of 1.07 million euros across the cross-section of municipalities -- approximately $4\%$ of the median municipal IRPEF income of 29 million euros -- statistically dominating every benchmark (DM $>4$ against persistence, $>40$ against spatial linear models, all $p |
| Date: | 2026–05 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2605.08782 |
| By: | Omar Abdel Haq; Amitabh Chandra; Tomáš Jagelka; Erzo F.P. Luttmer; Joshua Schwartzstein |
| Abstract: | Large Language Models (LLMs) are trained on a prodigious corpus of human writing and may reveal human preferences over characteristics of life courses, such as income, longevity, and working conditions. We present OpenAI's GPT-5.4 and a broadly representative sample of Americans with pairs of life stories and ask them to choose the life they would prefer for themselves. A person's choice is better predicted by the LLM's choice than by another person’s choice over the same stories, and LLM valuations of several life attributes are similar to those derived from human responses. Our results suggest that LLM responses offer a scalable and cost-effective complement to existing methods for studying human preferences. |
| JEL: | D0 H0 I0 |
| Date: | 2026–05 |
| URL: | https://d.repec.org/n?u=RePEc:nbr:nberwo:35185 |
| By: | Philippe Goulet Coulombe |
| Abstract: | Average forecast accuracy is not the same as forecast reliability. I treat forecast loss differentials relative to a benchmark as a return series. I then evaluate these returns using risk-adjusted performance measures from finance, including the Sharpe ratio, Sortino ratio, Omega ratio, and drawdown-based metrics. I also introduce the Edge Ratio capturing a model's propensity to deliver uniquely informative predictions relative to the forecasting frontier. I apply this framework to U.S. macroeconomic forecasting, comparing econometric benchmarks, machine learning models, a foundation model (TabPFN), and the Survey of Professional Forecasters. While it is often feasible to beat professional forecasters in terms of average accuracy, it is much harder to beat them on a risk-adjusted basis. They rarely exhibit catastrophic failures and often achieve high Edge Ratios, plausibly reflecting the value of contextual judgment. Nonetheless, selected machine learning methods deliver attractive risk profiles for specific targets. The framework naturally extends to meta-analyses across targets, horizons, and samples, illustrated with a density forecast evaluation and the M4 competition. |
| Date: | 2026–05 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2605.09712 |
| By: | Sebastian Bell; Ali Kakhbod; Martin Lettau; Abdolreza Nazemi |
| Abstract: | We propose AlphaGlass, an inherently interpretable machine-learning framework for constructing portfolios that directly optimize investment objectives. AlphaGlass maps stock characteristics into additive signals with sparse interactions and converts these signals into long-short portfolios through a differentiable rank-and-mask layer. This end-to-end design allows the model to optimize objectives such as the Sharpe ratio or mean-variance utility while keeping portfolio weights interpretable and traceable to specific characteristics and interactions. We show theoretically that in-sample objective maximization consistently estimates the population objective and that the differentiable rank-and-mask layer is a faithful smooth proxy for the corresponding conventional long-short quantile portfolio. In U.S. equities, AlphaGlass delivers strong out-of-sample performance and reveals economically interpretable drivers of long and short positions. |
| JEL: | C14 C45 G10 G11 G12 |
| Date: | 2026–05 |
| URL: | https://d.repec.org/n?u=RePEc:nbr:nberwo:35186 |