|
on Computational Economics |
| By: | Wei Ouyang |
| Abstract: | This technical note describes the artificial intelligence (AI) machine learning (ML) algorithm used in the Canadian Exporter Business Register (LFTTD). It examines the challenges associated with matching business names in Canadian export data with those in the Business Register. The note also outlines the considerations underlying the algorithm and explains how it addresses these data-matching challenges. |
| Keywords: | LFTTD |
| Date: | 2026–09 |
| URL: | https://d.repec.org/n?u=RePEc:cen:tnotes:26-44 |
| By: | Zongwu Cai (Department of Economics, The University of Kansas, Lawrence, KS 66045, USA); Ying Fang (The Wang Yanan Institute for Studies in Economics, Xiamen University, Xiamen, Fujian 361005, China and Department of Statistics & Data Science, School of Economics, Xiamen University, Xiamen, Fujian 361005, China); Ming Lin (The Wang Yanan Institute for Studies in Economics, Xiamen University, Xiamen, Fujian 361005, China and Department of Statistics & Data Science, School of Economics, Xiamen University, Xiamen, Fujian 361005, China); Zeqin Liu (School of Statistics, Shanxi University of Finance and Economics, Taiyuan, Shanxi 030006, China) |
| Abstract: | We review applications of large language models (LLMs) in economic and financial research, focusing on text analysis and variable construction, forecasting and decision support, simulation of economic agents, and research assistance and automation. We examine what researchers use LLMs to do, how the reported results compare with existing methods, and what limits their use. Contextual interpretation improves several classification tasks, but variable construction and measurement error affect subsequent analysis. Some forecasts improve on benchmarks, although gains vary and predictive signals need not yield trading profits. Human observations help calibrate agents and improve predictions of survey responses. Automated systems generate economic mechanisms and connect empirical analysis with drafting, while researchers still judge questions and explanations. We then discuss future directions for improving research quality and making better use of researchers' expertise. |
| Keywords: | Large language models, text analysis, economic forecasting, economic agent simulation, research assistance and automation |
| Date: | 2026–09 |
| URL: | https://d.repec.org/n?u=RePEc:kan:wpaper:202618 |
| By: | Filipović, Slađana; Kozarević, Safet; Mujić, Haris |
| Abstract: | Digitalization has expanded insurers' databases, making machine learning (ML) a valuable tool for identifying key rsik factors in casco insurance pricing. While Generalized Linear Models (GLM) remain the standard, modelling claims frequency and severity separately is essential for accurately assessing the impact of each feature. Extreme gradient boosting (XGBoost), an ensemble ML method, can complement traditional GLM by improving predictive accuracy through feature selection. This study investigates a proposed hybrid XGBoost-GLM approach using a real-world dataset of 116.205 casco policies from a Bosnian insurer covering the period 2011-2023. The results show that XGBoost-based feature selection did not outperform the baseline GLM, which included all initial features without any selection. Deviance analysis confirmed that the baseline GLM with the full initial set of features achieved lower deviance and represented the optimal specification for the frequency-severity models. The modelling process was conducted in the R software environment (R Core Team, 2025). |
| Keywords: | casco insurance pricing, GLM, feature selection, machine learning, XGBoost |
| Date: | 2025 |
| URL: | https://d.repec.org/n?u=RePEc:zbw:esconf:343875 |
| By: | Hiroyuki Kasahara; Allen Peters; Oliver Xu |
| Abstract: | Machine learning predicts outcomes well, but predictive accuracy does not ensure reliable counterfactual responses. We examine how to combine machine learning and theory for measurement and counterfactual analysis, using maritime CO2 emissions where physics provides a benchmark speed response. Matching hourly tracking data for dry bulk and container ships to annual fuel consumption reported under EU regulations, we compare engineering calculations, structural regressions, hybrid models, and machine learning. Out of sample, all estimated models predict within a few percent of reported totals, outperforming standard engineering calculations. Yet pure machine learning and unrestricted structural regression imply attenuated speed responses. Hybrids preserve the structural component's speed response by excluding speed-related inputs from the machine learning component. A cost-benefit analysis of speed reductions illustrates the policy stakes: an attenuated speed response can flip the sign of net benefits. Accurate aggregate predictions therefore cannot substitute for scrutiny of the restrictions determining counterfactual responses. |
| Date: | 2026–10 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2610.01008 |
| By: | Hamid Boustanifar; Sasan Mansouri |
| Abstract: | Researchers increasingly use generative large language models (LLMs) to convert corporate text into empirical variables. We examine the extent to which LLM-based textual measures are invariant to model choice using thirteen measures, including sentiment, management clarity, uncertainty, answer specificity, and climate and political risk. Seven LLMs from different providers score earnings call transcripts of S&P 500 companies on these constructs. Cross-model rank correlations average only 0.52, and transcript-level differences common across providers account for only 34% of total score variation. Cross-model disagreement does not predict subsequent analyst or market disagreement, consistent with a substantial model-specific component rather than common ambiguity in the underlying disclosure. Model choice significantly affects downstream inference, with coefficient magnitudes, signs, and statistical significance varying substantially across models. Averaging across providers makes transcript rankings more stable for most constructs, but score levels remain sensitive to the models included in the ensemble. LLM-generated variables should therefore be treated as model-contingent measurements and validated across providers. |
| Date: | 2026–09 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2609.31013 |
| By: | Georgios Chionas; Charalampos Kleitsikas; Stefanos Leonardos; Leandro S\'anchez-Betancourt; Carmine Ventre |
| Abstract: | We present SAiFE_gym, a Python module that provides a collection of simulation environments for studying trading problems in Constant Product Markets (CPMs) with Concentrated Liquidity (CL). These markets give Liquidity Providers (LPs) granular control over how their capital is allocated and enable them to adjust their range of liquidity provision dynamically based on market conditions, which in turn, dictates how they earn fees. We decompose the microstructure of CPMs with CL in interactive components that allow researchers and practitioners to capture various economic settings. We employ a vectorized approach to optimize our environments, making them scalable for high dimensional Reinforcement Learning (RL) workflows that best describe sequential decision problems. We demonstrate the benefits of our environments by evaluating the performance of RL agents in CPMs with CL under uncertainty in market parameters. |
| Date: | 2026–09 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2609.17788 |
| By: | Aleksandar Arandjelovic; Uwe Schmock |
| Abstract: | We study simple predictable processes whose coefficients are represented by neural networks. On finite measure spaces, we establish density results for neural networks in Orlicz spaces. For filtrations generated by a stochastic process, measurable random variables, including at stopping times, can be approximated by neural networks depending on finitely many observations. Every stochastic integral with respect to a semimartingale can then be approximated, in the semimartingale topology, by integrals of such simple predictable processes. We show that restricting trading strategies to this class leaves the minimal mean-variance hedging error under partial information unchanged and obtain a no-free-lunch characterization in terms of equivalent martingale measures. Finally, the Bichteler-Dellacherie characterization of semimartingales remains valid even upon restricting the predictable integrands to those whose coefficients are represented by neural networks. |
| Date: | 2026–09 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2609.31578 |
| By: | Jeffrey D. Michler; Kieran Douglas; Anna Josephson |
| Abstract: | Researchers increasingly use generative AI, particularly large language models (LLMs), to automate tasks across the research pipeline. We study the reliability of these tools at the reading, classification, and synthesis of large bodies of academic literature. We frame LLM-assisted literature reviews as a measurement problem, treating models as measurement systems and tracing how their errors affect downstream conclusions. As a test case, we use three different implementations of ChatGPT to identify and extract metadata from economics papers that use rainfall as an instrumental variable. We benchmark each implementation against a subset of human-labeled evaluation data, and then deploy those implementations to extract metadata from the full corpus. The LLMs perform well on binary classification, but performance deteriorates as tasks demand greater contextual interpretation. More importantly, how much researchers can rely on model outputs depends not only on the complexity of the reading task but also on the type of claims the data is asked to support. The same amount of measurement error substantially affects paper-level claims while having little effect on broader claims about the literature. Measurement error in LLM-generated data is thus most consequential at precisely the level of detail that constitutes an LLM's principal value added over human reviewers. We conclude that standard model performance metrics are informative about the quality of generated data but do not by themselves establish the credibility of downstream inference. Researchers must also evaluate whether substantive claims are robust to the measurement system used to generate the underlying data. |
| Date: | 2026–09 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2609.27686 |
| By: | Zhenhao Fu; Ruipeng Xu; Qibing Ren |
| Abstract: | Individually protective decisions can produce avoidable collective failures. As large language model (LLM) agents take on greater roles in financial decision-making, financial AI safety must therefore be considered not only at the level of individual agents, but also at the level of the systems they jointly create. We study this problem with FRAIL, a controlled experimental framework that places LLM agents in three dynamic financial environments---bank runs, debt rollover, and reward crowdfunding---where agents' decisions reshape the financial conditions faced by others. Across seven leading LLMs, we find widespread collective fragility even when no agent is instructed to destabilize the system: 77\% of baseline bank-run episodes and 83\% of debt-rollover episodes end in failure. We then compare three interaction mechanisms based on compensated commitments, centralized commitment agreements, and participant-led coalitions. All three improve aggregate outcomes, but no single mechanism performs best across all financial structures. Across mechanisms, successful stabilization shares a common temporal pattern: broad commitment forms early, before defensive behavior becomes self-reinforcing. Our findings show that individually capable agents do not automatically form safe financial systems, highlighting system-level evaluation and interaction design as central problems for financial AI safety. Code is available at https://anonymous.4open.science/r/FinFra il-CF26. |
| Date: | 2026–09 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2609.30940 |
| By: | Alexey Chernysh; Orkhan Ekhtibarov; Dmitry Zmitrovich |
| Abstract: | Large language models are strong general-purpose code generators, but executable algorithmic trading remains a demanding specialization target: a model must translate a natural-language strategy specification into correct program logic for a specialized trading framework, execute on historical data, produce trades, and remain semantically faithful to the request. We study two complementary mechanisms for specializing language models for this setting: continued pretraining on algorithmic-trading framework code and supervised fine-tuning (SFT) on agent-validated request-to-code pairs. Evaluation is centered on QuantCode-Bench, our 400-task benchmark for Backtrader strategy generation, together with a repository-level SWE-bench-like track. Continued pretraining improves single-turn Judge Pass from 41.5% to 47.5% for Qwen3.5-397B-A17B and from 27.8% to 33.0% for Qwen3.6-35B-A3B. SFT applied after continued pretraining yields a larger gain for Qwen3.6-35B-A3B, reaching 58.2% Judge Pass and 83.5% successful backtests; in agentic evaluation it raises first-turn success from 22.3% to 58.3% and final success after up to 10 turns from 47.5% to 79.5%. Continued pretraining alone improves first-turn agentic success but lowers final success after repair from 47.5% to 32.5%, consistent with degraded instruction following, whereas SFT improves both. We also identify a capability-retention failure: domain specialization degrades parser-conformant structured tool calling, and targeted recovery SFT restores tool-call formatting but not the base checkpoint's repository-level agent performance. The results show that framework-oriented pretraining, validated SFT, and explicit capability-retention evaluation address distinct failure modes in domain-specific executable code generation. |
| Date: | 2026–09 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2609.39420 |
| By: | Christos Spyridon Koulouris; Carlo Campajola |
| Abstract: | In this paper, we extend earlier findings of supra-competitive outcomes in optimal-execution games by identifying a learned punitive mechanism that deters deviations and provides behavioural evidence of collusion. We investigate this mechanism in a two-player, finite-horizon Almgren-Chriss liquidation game. Independent proximal policy optimisation agents with access to within-episode price and action histories achieve costs below the Nash benchmark. We identify a profitable deviation by training against the mean learned liquidation schedule, then impose its first trade on one of the original agents. The opponent responds by accelerating liquidation. This response more than offsets the deviator's gain in every run and both player roles, while leaving the punisher's average payoff materially unchanged relative to not punishing under the same deviation. The punisher imposes greater losses on the deviator while preserving its own average payoff, despite the availability of more profitable, less punitive liquidation plans. Matching deviations and subsequent additional selling rise and later decline during training, while final policies retain an effective punitive response. We formalise two checks: whether punishment outweighs the gain from deviating, and whether the change in trading behaviour is large enough to account for the loss imposed. Both checks hold for the tested deviation. Together, these findings provide behavioural and economic evidence supporting a collusive interpretation of the learned supra-competitive outcomes. |
| Date: | 2026–09 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2610.00619 |
| By: | Hammed A. Olayinka; Saheed O. Olayemi |
| Abstract: | We develop a Dynamic Spatial Panel Bayesian Additive Regression Trees model with Horseshoe shrinkage (DSP-BART-HS) for high-dimensional spatio-temporal panel data. We jointly evaluate the model against a comprehensive suite of structural spatial econometrics, non-parametric machine learning methods, and small-area estimators across nine data-generating scenarios spanning 200 replicates each, including irregular spatial topologies, dense policy effects, and non-linear individual-level interactions. DSP-BART-HS is the best or statistically indistinguishable from the best estimator in every scenario. Conventional region-time-aggregate comparators suffer severe performance degradation -- trailing by a factor of three or more -- whenever individual-level non-linearity drives outcome variance, a limitation this framework's tree-ensemble design directly addresses. The model also maintains strong predictive accuracy under a zero-training-region spatial holdout via its spatial diffusion mechanism. We demonstrate practical utility across two U.S. county-level panel applications -- intergenerational economic mobility and geographic income inequality -- under genuine unseen-region, random, and temporal holdouts, with results validated through repeated-split uncertainty quantification and paired significance testing against every comparator. One clear limitation emerges: under temporal extrapolation specifically, a lighter per-region autoregressive specification (MTS-CAR-X) achieves a robust, and consistent advantage. DSP-BART-HS nonetheless establishes strong predictive performance, automated variable selection, and flexible inference for hierarchical spatio-temporal panel data. |
| Date: | 2026–09 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2610.00072 |
| By: | Hengyi Yang; Sida Lin; Yiyan Qi; Yankai Chen; Haohan Zhang; Xianhua Peng; Jian Guo |
| Abstract: | Stock price forecasting is a long-standing challenge in computational finance, driven by the inherent randomness of markets and complex temporal patterns. While recent deep-learning models have raised forecasting accuracy by jointly modeling inter-stock and temporal price dynamics, they conflate inter-stock relationships with intra-stock temporal dependencies and focus solely on the univariate objective of price movement. To address these limitations, we propose LiMT, a Hierarchical Multi-Task Learning framework that integrates liquidity-aware signals for stock price forecasting. LiMT employs a Market Regime Encoder (MRE) module that first extracts contemporaneous cross-stock dependencies, then models each stock's temporal dynamics, yielding a unified latent state. Building on this latent state, we introduce a Liquidity-Driven Learning (LDL) module, a mixture-of-experts architecture that features cross-task gating mechanisms to jointly predict price movement, volatility, and trading volume. We further design an Adaptive Portfolio Optimization (APO) mechanism that converts multi-task forecasts into executable portfolio weights under transaction-cost and liquidity constraints. Extensive experiments on the CSI300 and CSI500 benchmarks show that LiMT performs best among strong neural and tree-based baselines across the reported metrics. In realistic CSI300 backtests, APO improves annualized return from 3.99% to 10.01% and Sharpe ratio from 1.22 to 1.86 over equal weighting, showing that the multi-task forecasts translate into deployable portfolio gains. |
| Date: | 2026–09 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2609.25617 |
| By: | Duong Hien Chi Kien; Thanh Trung Huynh |
| Abstract: | Reinforcement learning for trading often struggles to balance upside participation with drawdown control. Profit-only policies can collapse toward passive long exposure on upward-drifting assets, while aggressively risk-penalized rewards can become too defensive during volatile periods. This paper proposes PPO-HRAP, a hybrid regime-aware policy that combines Proximal Policy Optimization with an interpretable regime prior. The agent observes both market features and portfolio-state variables, receives a reward combining portfolio log return, VIX-conditioned drawdown-increase penalty, target-exposure deviation, and turnover cost, and executes a blended action between the PPO actor output and a regime-derived target exposure. On the held-out 2020-2022 SPY test window, PPO-HRAP achieves 27.62% total return, 8.48% annualized return, 0.6447 Sharpe ratio, 0.8588 Sortino ratio, and 0.4592 Calmar ratio, while reducing maximum drawdown from 34.10% for Buy and Hold to 18.47%. Across five SPY seeds, PPO-HRAP remains stable with mean total return $0.2725 \pm 0.0109$ and mean Sharpe ratio $0.6219 \pm 0.0565$. Single-run cross-asset tests on QQQ and DIA further show that the proposed method ranks first on total return and Sharpe ratio for all three reported assets. These results suggest that blending learned actions with a volatility-aware regime prior is a practical way to improve risk-adjusted trading behavior, although the current policy still incurs high turnover and cross-asset robustness beyond SPY remains limited to single-run evidence. |
| Date: | 2026–10 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2610.01325 |
| By: | Seunghan Lee; Jun Seo; Jaehoon Lee; Junhyeok Kang; Sangjun Han; Sungdong Yoo; Minjae Kim; Tae Yoon Lim; Dongwan Kang; Hwanil Choi; Soonyoung Lee; Wonbin Ahn |
| Abstract: | Language models used in financial backtests suffer from look-ahead bias, as a model trained on text published after the study period has already observed the outcomes it is asked to predict. To handle this issue, point-in-time (PIT) language models are pretrained on chronologically filtered corpora and released as one checkpoint per calendar year, each with a documented cutoff. However, each additional year costs a full pretraining run, and whether that run is necessary has never been tested. In this paper, we show that the annual pretraining run is not necessary. We instead compare each checkpoint against the newer one that replaced it, and find that the newer checkpoint scores no better on the same evaluation window. Motivated by this observation, we propose PALM (Point-in-time Adaptation for financial Language Models), a simple yet effective alternative to annual pretraining that fits a low-rank adapter on text published before the decision date without modifying any pretrained weight. We further find that a small adapter is enough to add a new period to the knowledge an old checkpoint already encodes, and that this outperforms continued pretraining. We validate PALM on a decade of financial news and on various families of PIT models, whose cutoffs span two decades and whose sizes range from 1.3 to 4.2B. Code is available at: https://github.com/seunghan96/palm. |
| Date: | 2026–09 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2609.30316 |
| By: | Satoshi Nakano; Kazuhiko Nishimura |
| Abstract: | Understanding how supply chains endogenously transform requires a parametric model of processing networks with non-neutral substitution elasticities. While the Cascaded CES (CCES) production function provides a rigorous framework for these multi-layered linkages, dynamically calibrating its structural parameters from time-series data constitutes a highly non-convex inverse optimization problem. Enforcing the strict microeconomic concavity constraint causes standard monolithic approach to fail due to extreme ill-conditioning and the curse of dimensionality. To overcome this computational bottleneck, we propose a novel structure-exploiting algorithm. By leveraging the physical upstreamness topology of the network, our hybrid heuristic alternates between a vertical cascade-sequential descent and a horizontal block coordinate descent. Our framework successfully calibrates the fundamental elasticities of a 10-sector marcoeconomic model of the United States, providing a tractable computational engine to fully endogenize and predict complex supply-chain transformations. |
| Date: | 2026–09 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2609.15452 |
| By: | Vassilis Polimenis |
| Abstract: | Lepto-regression is defined as the machine learning process of constructing a Regression Tree of a target feature on itself. It is a novel, model-free method potentially revealing information on important sample structure properties. But it is yet not clear what the informational content of lepto-variance is and how it is related to other well-known statistics of a sample. One significant finding is that 58% of the historical US stock return variability is 1-bit lepto-variance that can not be explained by any financial factor. The central question investigated in this paper is to use small normal N(0, 1) drawn samples to explore how the 1-bit sample lepto-variance and lepto-ratio relate to sample variance, skewness and excess kurtosis. Using a large sample simulation, the lepto ratio of a normal is found to converge to 36.3%. For smaller normally distributed simulated N(0, 1) samples, while lepto-variance itself is highly correlated to sample variance, lepto-variance as a fraction of total variance is highly correlated to excess kurtosis. Both lepto-variance and lepto-ratio are orthogonal to sample skew. Another finding is that while lepto-ratio is strongly correlated to lepto-variance it remains orthogonal to sample variance. |
| Date: | 2026–09 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2609.25144 |
| By: | Jawad Addoum; Vitaly Meursault; Justin Murfin |
| Abstract: | Stock and bond investors treat different aspects of firm information as value-relevant, generating distinct mappings of text to returns (“dictionaries”). Comparing stock and bond responses to earnings calls, bondholders emphasize operations and downside risk (relative to growth and innovation). While consistent with asset payoff differences, dictionary differences follow the investor cohort and not the payoff structure. Among bonds that are de facto equities but priced by bondholders—junior debt for which firm value is less than senior debt face value—bond investors continue to interpret information through a creditor’s lens. Information ignored by one dictionary but not the other generates price underreaction. |
| Keywords: | Cross-asset returns; Natural Language Processing; Machine Learning; Market Segmentation |
| Date: | 2026–09–21 |
| URL: | https://d.repec.org/n?u=RePEc:fip:fedpwp:103784 |
| By: | Abhinav Tomar; Lakshya Nagpal; Vikas Chauhan; S. R. Hassan |
| Abstract: | Interconnected financial systems are vulnerable to cascading failures arising from cross-holdings and nonlinear contagion, making the analysis and mitigation of systemic risk a challenging computational problem. In this work, we develop a unified optimization framework for financial network analysis based on Ising models and Quadratic Unconstrained Binary Optimization (QUBO). Starting from the Elliott Golub Jackson financial network model, we extend equilibrium valuation to incorporate threshold-induced failures, formulate the Maximum Cascade Failure Problem, and derive an equivalent QUBO representation. We then formulate the Optimal Bailout Allocation Problem as a controlled Ising model and transform the resulting bi-level optimization into a single joint QUBO that simultaneously determines equilibrium failures and optimal interventions under budget constraints. To characterize the influence of individual institutions, we introduce bailout susceptibility as a response-based measure of systemic importance and develop a susceptibility-driven greedy intervention strategy. Numerical simulations demonstrate equilibrium valuation, worst-case cascade identification, optimal bailout allocation, and susceptibility analysis on financial networks of varying sizes. Beyond optimization, the Ising representation provides a general statistical-mechanical framework for analyzing financial contagion, enabling the application of response theory, Monte Carlo methods, and other techniques developed for interacting spin systems. The proposed framework establishes a unified approach for systemic risk analysis that is compatible with classical annealing, quantum-inspired optimization, and emerging quantum annealing technologies |
| Date: | 2026–07 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2609.17415 |
| By: | Aojie Yuan; Haiyue Zhang; Zijian Su |
| Abstract: | Decision-focused learning (DFL) trains predictors through downstream objectives, but a different loss need not provide an independent parameter-update direction. We characterize this restriction through the predictor Jacobian, using sparse index tracking to distinguish the covariance entries read by the optimizer from the parameter directions available to learning. Rank-one Jacobians make nonzero per-example gradients collinear; a conditional spectral bound describes near-collinearity. A batch-subspace characterization and counterexamples show why these local statements imply neither common minimizers nor collinear batch updates. Experiments examine when geometry translates into decision quality. Across 38 one-parameter equity configurations, DFL gains over MSE remain below 1.8%; a 385-parameter conditional predictor also has pointwise rank one. In validation-tuned shortest-path and knapsack experiments, full-capacity SPO+ reduces mean regret by 11.6% and 10.6%, respectively; only knapsack survives correction across eight comparisons. The capacity contrast persists on fresh datasets across batch orders and training budgets. Holding expressivity fixed, invertible coordinate scaling lowers spectral effective rank and ordinary SGD gains; compensating for the scaling restores the original trajectories. Financial forward-target controls separate forecast accuracy from decision quality; a matched neural comparison finds no aggregate DFL advantage in the tested architecture. These findings distinguish local rank restrictions, coordinate-dependent optimization and predictive accuracy. Predictor geometry helps explain available learning directions, while held-out decision quality remains the test of practical benefit. |
| Date: | 2026–09 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2609.39261 |
| By: | Gabriel Manso; Emma Fu; Neil Thompson |
| Abstract: | As artificial intelligence's capabilities improve, it is increasingly viewed as a general scientific method. But how true are these claims? Does AI outperform all techniques, or only some, and how is this changing? To assess the claims, we assemble a corpus of 2, 507 head-to-head comparisons between AI and other scientific analysis techniques across 27 scientific disciplines from papers published between 2000 and early 2025. We find a profound dichotomy. Relative to traditional statistics, AI often outperforms, but at a significantly higher computational cost. But there are also nearly a quarter of cases where AI is both more expensive and performs worse than traditional statistical techniques and this fraction has been stable for a decade. Relative to scientific computing, AI often underperforms, but at lower computational cost. This has begun to change: since 2020, AI's performance against scientific computing has notably strengthened and it now outperforms on more than half of comparisons. These patterns suggest that AI is therefore not a universal replacement for existing methods, but rather a valuable -- and improving -- part of a new AI-enabled scientific frontier. |
| Date: | 2026–09 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2609.16258 |
| By: | Hoyoung Lee; Suyeol Yun; Jack Haverty; Yunju Cho; Meesong Kim; Daekyung Park; Sumin Kim; Jihoon Kwon; Jasmine Jia Geng; Andrew Chin; Yin Luo; Edward Tong; Yu Yu; Zach Golkhou; Minkyu Kim; Igor Halperin; Young Cha; Alejandro Lopez-Lira; Chanyeol Choi; Yongjae Lee |
| Abstract: | Evaluating finance research agents requires rubrics that reflect expert standards and fix the values correct as of an information cutoff. Expert-reviewed finance benchmarks rely on fixed, per-item rubrics, which are costly to extend and cannot encode each institution's own standard. In FinAutoRubric, experts specify reusable evaluation guidance, while agents and code carry out query-specific rubric generation, review, and validation. This expert guidance governs every agent, as prompts and as rules that code enforces, and a Task Bank of reusable criteria carries it across tasks. In long-horizon loops that follow the expert guidance, a writer agent researches every expected value and a reviewer agent verifies it, and failures escalate to a human. On three expert-authored finance benchmarks, its rubrics track expert scoring as closely as the strongest evaluated generator while stating the expert rubric's expected value for more criteria, their scores agree with human grading, and in-house analysts prefer them in a blind review. The released 100-query FinAutoRubric Benchmark, built from in-house analysts' key questions across 78 tasks and eight asset classes, shows that rubrics from an earlier model generation still leave headroom for a later one. |
| Date: | 2026–09 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2609.35744 |
| By: | Lucas Parmentier (CEMOI - Centre d'Économie et de Management de l'Océan Indien - UR - Université de La Réunion) |
| Abstract: | I study the optimal taxation of artificial intelligence (AI). My analysis highlights a wage-compression effect that provides a rationale for subsidizing AI, as AI adoption may reduce rather than increase wage inequality. When this effect outweighs AI's social costs, AI should be subsidized instead of taxed. I treat AI as a productive input distinct from robots, as the two technologies perform different ranges of tasks. Using numerical simulations, I demonstrate that the optimal policy for AI may differ sharply from that for robots. Under the benchmark calibration, AI is subsidized, whereas robots are taxed. Therefore, the case for taxing AI is weaker than is commonly assumed. When AI lowers the skill premium, fiscal policy may need to encourage, rather than slow, adoption. |
| Abstract: | Nous étudions la taxation optimale de l'intelligence artificielle (IA). Notre analyse révèle un effet de compression des salaires constituant un motif de subvention à l'IA, dans la mesure où l'IA pourrait réduire les inégalités salariales au lieu de les accroître. Lorsque cet effet domine les coûts sociaux de l'IA, cette dernière devrait être subventionnée au lieu d'être taxée. Nous traitons l'IA comme un intrant distinct des robots, car ces deux technologies effectuent des tâches différentes. À travers différentes simulations numériques, nous montrons que la politique optimale concernant l'IA peut être très différente de celle concernant les robots. Avec le calibrage de base, l'IA est subventionnée, tandis que les robots sont taxés. Ainsi, l'argument en faveur d'une taxation de l'IA est plus faible qu'il n'y paraît. Lorsque l'IA réduit la prime de compétence, les pouvoirs publics pourraient devoir encourager son adoption, plutôt que la ralentir. |
| Keywords: | Artificial intelligence, Optimal taxation, Redistribution, Wage inequality, Technological change, Inégalités salariales, Progrès technique, Taxation optimale, Intelligence artificielle |
| Date: | 2026–11 |
| URL: | https://d.repec.org/n?u=RePEc:hal:journl:hal-05748060 |
| By: | Dawn C Parker (University of Waterloo [Waterloo]); Corrine Ida (LISA - Laboratoire « Lieux, Identités, eSpaces, Activités » (UMR CNRS 6240 LISA) - CNRS - Centre National de la Recherche Scientifique - Università di Corsica Pasquale Paoli [Université de Corse Pascal Paoli]); Claudio Detotto (LISA - Laboratoire « Lieux, Identités, eSpaces, Activités » (UMR CNRS 6240 LISA) - CNRS - Centre National de la Recherche Scientifique - Università di Corsica Pasquale Paoli [Université de Corse Pascal Paoli], CRENOS - UniCa - Università degli Studi di Cagliari = University of Cagliari = Université de Cagliari); Tatiana Filatova; Jiaqi Ge (TU Delft - Delft University of Technology); Yu Huang (OUC - Ocean University of China); Eric Innocenti (LISA - Laboratoire « Lieux, Identités, eSpaces, Activités » (UMR CNRS 6240 LISA) - CNRS - Centre National de la Recherche Scientifique - Università di Corsica Pasquale Paoli [Université de Corse Pascal Paoli]); Nicholas Magliocca (UA - University of Alabama [Tuscaloosa]); J. Gareth Polhill (The James Hutton Institute); Dominique Prunetti (LISA - Laboratoire « Lieux, Identités, eSpaces, Activités » (UMR CNRS 6240 LISA) - CNRS - Centre National de la Recherche Scientifique - Università di Corsica Pasquale Paoli [Université de Corse Pascal Paoli]); Shahab Valaei Sharif (University of Waterloo [Waterloo]) |
| Abstract: | In recent years, a wide range of agent-based real-estate market models (ABM-REMMs) have been developed to incorporate actor and environmental heterogeneity and feedbacks between scales, features not possible in traditional real-estate market models. Despite close communication and collaboration between scholars, independent research groups have developed customized model codes, often in different programming languages. To synthesize this knowledge, we present a new meta-model template "MR POTATOHEAD: Real Estate Market Edition" (MP-REME), developed collaboratively by the authors, to describe and compare six independently developed ABM-REMMs. We demonstrate that the six models are special cases of a generalized meta-model, which is suitable for a wide range of agent-based exploration of real-estate market dynamics. The MP-REME templates can serve as a design pattern to catalyze development of a community ABM-REMM code base, facilitating faster model development, model transparency and replicability, and model comparison. Residential real estate (land and housing) markets impact a wide variety of critical socio-ecological outcomes at local, national, and international scales, and scientifically robust and transparent real-estate market models are needed to explore the impacts of these markets on socio-ecological outcomes. Given the range of differences between models, development of such a standard model is critical to increase confidence in the application of ABM-REMMs to policy analysis of issues such as urban flooding, carbon sequestration, heat island mitigation, and biodiversity preservation. |
| Keywords: | reuseable building blocks, model replication, design pattern, real-estate markets, agent-based modeling |
| Date: | 2026–09–11 |
| URL: | https://d.repec.org/n?u=RePEc:hal:journl:hal-05747447 |
| By: | Esteve Almirall; Steve Willmott; Ulises Cort\'es |
| Abstract: | Why does a superior technology sometimes spread and sometimes stall, even when its expected returns are much higher? We develop an agent-based model in which firms adopt a digital technology by imitating successful neighbours through a fast-and-frugal heuristic. We compare frequency-based imitation, where firms follow the local majority, with performance-conditioned imitation, where they follow better-performing neighbours. We define a cascade as high final adoption reached through rapid, accelerating takeoff rather than slow accumulation. In the calibrated regime, four results emerge. First, dynamic cascades arise under performance-conditioned imitation, but not under an exogenous adoption hazard or frequency-based imitation. Second, on degree-controlled networks, adoption tips once the network provides sufficient reach, at a small-world rewiring threshold near 0.41, with a 95 percent bootstrap confidence interval of 0.39 to 0.43. This reflects short paths, low clustering, and high-reach nodes rather than one graph statistic. Third, at a fixed mean threshold, concentrating receptiveness in the lower tail can generate cascades that homogeneous thresholds cannot. Fourth, cascade probability depends strongly on the positions of initial adopters and low-threshold pioneers. At human-scale observation degrees, the mechanism is genuine complex contagion: adoption requires reinforcement from multiple successful exemplars, survives a requirement of at least two successful digital neighbours when seeded by a clustered critical mass, and weakens beyond Dunbar-scale neighbourhoods. The model identifies where interventions may have leverage rather than forecasting industries. Broad cultural change is not required for system-wide adoption: a small minority of receptive firms, positioned where the network can transmit success, can be enough. In short, a few misfits can change the world. |
| Date: | 2026–07 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2609.30269 |