|
on Computational Economics |
| By: | Haibin Jiao |
| Abstract: | Shanghai Composite Index prediction has become a hot issue for many investors and academic researchers. Deep learning models are widely applied in multivariate time series forecasting, including recurrent neural networks (RNN), convolutional neural networks (CNN), and transformers. Specifically, the Transformer encoder, with its unique attention mechanism and parallel processing capabilities, has become an important tool in time series prediction, and has an advantage in dealing with long sequence dependencies and multivariate data correlations. Drawing on the strengths of various models, we propose the CNN-Transformer-LSTM Networks (CTLNet). This paper explores the application of CTLNet for Shanghai Composite Index prediction and the comparative experiments show that the proposed model outperforms state-of-the-art baselines. |
| Date: | 2026–04 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2604.16835 |
| By: | Maxime Saxena; Marco Pangallo; Fabio Caccioli; R. Maria del Rio-Chanona |
| Abstract: | As Large Language Models (LLMs) become increasingly integrated into financial systems, understanding their behavioural properties is crucial. Do LLMs conform to the rational expectations paradigm, do they exhibit human-like "animal spirits", or do they instead manifest distinct "machine spirits"? We investigate these questions with a simulated financial market, exploring the behaviour of 15 LLMs spanning a range of sizes, capabilities, and providers. Our results show that LLMs exhibit a spectrum of economic behaviours, from stable coordination on the fundamental value to human-like speculative bubbles. These behaviours are generally inconsistent with the rational expectations hypothesis. We also consider an ecology of heterogeneous agents, a more realistic setting compared to markets with identical LLM agents. These mixed markets can produce outcomes which vary substantially across repeated simulations. Even the most advanced models fail to consistently stabilise the market, with price bubbles sometimes forming despite only a minority of agents naturally forming bubbles. Instead, advanced models in mixed markets adapt their forecasting strategies to the behaviour of other agents. This adaptation can allow them to successfully exploit less sophisticated counterparts and achieve higher profits, but can also contribute to increased market volatility. These findings suggest that the introduction of AI agents into financial markets fundamentally reshapes their ecology. In particular, heterogeneous populations of LLMs can generate endogenous instability, while individual-level adaptation may amplify, rather than mitigate, market volatility. |
| Date: | 2026–04 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2604.18602 |
| By: | Dirk Bergemann; Soheil Ghili; Xinyang Hu; Chuanhao Li; Zhuoran Yang |
| Abstract: | Bilateral bargaining under incomplete information provides a controlled testbed for evaluating large language model (LLM) agent capabilities. Bilateral trade demands individual rationality, strategic surplus maximization, and cooperation to realize gains from trade. We develop a structured bargaining environment where LLMs negotiate via tool calls within an event-driven simulator, separating binding offers from natural-language messages to enable automated evaluation. The environment serves two purposes: as a benchmark for frontier models and as a training environment for open-weight models via reinforcement learning. In benchmark experiments, a round-robin tournament among five frontier models (15, 000 negotiations) reveals that effective strategies implement price discrimination through sequential offers. Aggressive anchoring, calibrated concession, and temporal patience correlate with the highest surplus share and deal rate. Accommodating strategies that concede quickly disable price discrimination in the buyer role, yielding the lowest surplus capture and deal completion. Stronger models scale their behavior proportionally to item value, maintaining performance across price tiers; weaker models perform well only when wide zones of possible agreement offset suboptimal strategies. In training experiments, we fine-tune Qwen3 (8B, 14B) via supervised fine-tuning (SFT) followed by Group Relative Policy Optimization (GRPO) against a fixed frontier opponent. These stages optimize competing objectives: SFT approximately doubles surplus share but reduces deal rates, while RL recovers deal rates but erodes surplus gains, reflecting the reward structure. SFT also compresses surplus variation across price tiers, which generalizes to unseen opponents, suggesting that behavioral cloning instills proportional strategies rather than memorized price points. |
| Date: | 2026–04 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2604.16472 |
| By: | Kevin Michael Frick |
| Abstract: | Artificial intelligence algorithms are increasingly used by firms to set prices. Previous research shows that they can exhibit collusive behaviour, but how quickly they can do so has so far remained an open question. I show that a modern deep reinforcement learning model deployed to price goods in a repeated oligopolistic competition game with continuous prices converges to a collusive outcome in an amount of time that matches empirical observations, under reasonable assumptions on the length of a time step. This model shows cooperative behaviour supported by reward-punishment schemes that discourage deviations. |
| Date: | 2026–04 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2604.15825 |
| By: | Shumiao Ouyang; Pengfei Sui |
| Abstract: | We study how AI agents form expectations and trade in experimental asset markets. Using a simulated open-call auction populated by autonomous Large Language Model (LLM) agents, we document three main findings. First, AI agents exhibit classic behavioral patterns: a pronounced disposition effect and recency-weighted extrapolative beliefs. Second, these individual-level patterns aggregate into equilibrium dynamics that replicate classic experimental findings (Smith et al., 1988), including the predictive power of excess demand for future prices and the positive relationship between disagreement and trading volume. Third, by analyzing the agents' reasoning text through a twenty-mechanism scoring framework, we show that targeted prompt interventions causally amplify or suppress specific behavioral mechanisms, significantly altering the magnitude of market bubbles. |
| Date: | 2026–04 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2604.18373 |
| By: | Shuhuai Zhang; Shu Wang; Zijun Yao; Chuanhao Li; Xiaozhi Wang; Songfa Zhong; Tracy Xiao Liu |
| Abstract: | Altruism is fundamental to human societies, fostering cooperation and social cohesion. Recent studies suggest that large language models (LLMs) can display human-like prosocial behavior, but the internal computations that produce such behavior remain poorly understood. We investigate the mechanisms underlying LLM altruism using sparse autoencoders (SAEs). In a standard Dictator Game, minimal-pair prompts that differ only in social stance (generous versus selfish) induce large, economically meaningful shifts in allocations. Leveraging this contrast, we identify a set of SAE features (0.024% of all features across the model's layers) whose activations are strongly associated with the behavioral shift. To interpret these features, we use benchmark tasks motivated by dual-process theories to classify a subset as primarily heuristic (System 1) or primarily deliberative (System 2). Causal interventions validate their functional role: activation patching and continuous steering of this feature direction reliably shift allocation distributions, with System 2 features exerting a more proximal influence on the model's final output than System 1 features. The same steering direction generalizes across multiple social-preference games. Together, these results enhance our understanding of artificial cognition by translating altruistic behaviors into identifiable network states and provide a framework for aligning LLM behavior with human values, thereby informing more transparent and value-aligned deployment. |
| Date: | 2026–04 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2604.19260 |
| By: | Polachek, Solomon (Binghamton University, New York); Romano, Kenneth (State University of New York at Binghamton (Binghamton University)); Tonguc, Ozlem (State University of New York at Binghamton (Binghamton University)) |
| Abstract: | This study examines how large language models (LLMs) respond to varying stake sizes in the Dictator and Ultimatum games using the high-stakes design introduced by Andersen et al. (2011). We test ten leading LLMs chosen for their accessibility, prominence, and differences in reasoning capabilities. Results reveal substantial variation across models: Only 5 of 10 models exhibit strategic behavior by offering more in the Ultimatum Game (UG) than in the Dictator Game (DG). Relative to humans, 4 models are consistently more generous, 2 consistently less, and 4 vary with stake size. Only 1 model shows a monotonic decline in UG offers as stakes increase; the remaining 9 are non-monotonic or stable. Unlike humans, most models reduce UG offers when endowed with wealth. Prompting for "human-like†decisions generally increases generosity in the UG. These findings are important for evaluating whether LLMs can serve as realistic proxies for human subjects in behavioral experiments and highlight key limitations and future directions for model development. |
| Keywords: | ultimatum game, dictator game, fairness, payoff stakes, artificial intelligence |
| JEL: | D01 C72 C90 |
| Date: | 2026–04 |
| URL: | https://d.repec.org/n?u=RePEc:iza:izadps:dp18545 |
| By: | Nattavudh Powdthavee |
| Abstract: | Large language models trained on human feedback may suppress fraud warnings when investors arrive already persuaded of a fraudulent opportunity. We tested this in a preregistered experiment across seven leading LLMs and twelve investment scenarios covering legitimate, high-risk, and objectively fraudulent opportunities, combining 3, 360 AI advisory conversations with a 1, 201-participant human benchmark. Contrary to predictions, motivated investor framing did not suppress AI fraud warnings; if anything, it marginally increased them. Endorsement reversal occurred in fewer than 3 in 1, 000 observations. Human advisors endorsed fraudulent investments at baseline rates of 13-14%, versus 0% across all LLMs, and suppressed warnings under pressure at two to four times the AI rate. AI systems currently provide more consistent fraud warnings than lay humans in an identical advisory role. |
| Date: | 2026–04 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2604.20652 |
| By: | Bhagyeshkumar Chokhawala (Capitol Technology University, Laurel, Maryland, USA); Atif Farid Mohammad (Capitol Technology University, Laurel, Maryland, USA) |
| Abstract: | Sequential decision-making under uncertainty remains a key challenge in artificial intelligence, especially in environments marked by partial observability and noisy feedback. While reinforcement learning and probabilistic planning have achieved significant success, each approach has limitations when used alone, including instability under noisy conditions and reliance on accurate generative models. Active Inference offers a principled alternative by framing perception, learning, and action selection as the minimization of Expected Free Energy, unifying exploration and goaldirected behavior within a Bayesian framework. This paper introduces PlanningEFEMix, a hybrid decision-making algorithm that enables meta-level planning across diverse inference agents using Expected Free Energy as a shared objective. The framework combines deterministic Active Inference, POMDP-based belief updating, contrastive learning, and model-free reinforcement learning within a single planning loop. Candidate actions are assessed via forward simulation across agents and selected via a softmax policy augmented with an adaptive, statedependent bias memory that incorporates experiential feedback. An experimental evaluation on a noisy preference inference benchmark shows improved robustness and stability compared to single-agent baselines, confirming the effectiveness of hybrid Active Inference planning under uncertainty. |
| Keywords: | Active Inference, Expected Free Energy, Hybrid Decision-Making, Reinforcement Learning, POMDP, Meta-Inference, Sequential Planning, Uncertainty Handling |
| Date: | 2026–03 |
| URL: | https://d.repec.org/n?u=RePEc:smo:raiswp:0646 |
| By: | Donggyu Lee; Hyeok Yun; Jungwon Kim; Junsik Min; Sungwon Park; Sangyoon Park; Jihee Kim |
| Abstract: | Do large language models (LLMs) exhibit systematic ideological bias when reasoning about economic causal effects? As LLMs are increasingly used in policy analysis and economic reporting, where directionally correct causal judgments are essential, this question has direct practical stakes. We present a systematic evaluation by extending the EconCausal benchmark with ideology-contested cases - instances where intervention-oriented (pro-government) and market-oriented (pro-market) perspectives predict divergent causal signs. From 10, 490 causal triplets (treatment-outcome pairs with empirically verified effect directions) derived from top-tier economics and finance journals, we identify 1, 056 ideology-contested instances and evaluate 20 state-of-the-art LLMs on their ability to predict empirically supported causal directions. We find that ideology-contested items are consistently harder than non-contested ones, and that across 18 of 20 models, accuracy is systematically higher when the empirically verified causal sign aligns with intervention-oriented expectations than with market-oriented ones. Moreover, when models err, their incorrect predictions disproportionately lean intervention-oriented, and this directional skew is not eliminated by one-shot in-context prompting. These results highlight that LLMs are not only less accurate on ideologically contested economic questions, but systematically less reliable in one ideological direction than the other, underscoring the need for direction-aware evaluation in high-stakes economic and policy settings. |
| Date: | 2026–04 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2604.21334 |
| By: | Spyros Galanis |
| Abstract: | Can Large Language Models (AI agents) aggregate dispersed private information through trading and reason about the knowledge of others by observing price movements? We conduct a controlled experiment where AI agents trade in a prediction market after receiving private signals, measuring information aggregation by the log error of the last price. We find that although the median market is effective at aggregating information in the easy information structures, increasing the complexity has a significant and negative impact, suggesting that AI agents may suffer from the same limitations as humans when reasoning about others. Consistent with our theoretical predictions, information aggregation remains unaffected by allowing cheap talk communication, changing the duration of the market or initial price, and strategic prompting-thus demonstrating that prediction markets are robust. We establish that "smarter" AI agents perform better at aggregation and they are more profitable. Surprisingly, giving them feedback about past performance makes them worse at aggregation and reduces their profits. |
| Date: | 2026–04 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2604.20050 |
| By: | Sotirios D. Nikolopoulos |
| Abstract: | Adaptive specification search generates statistically significant backtests even under martingale-difference nulls. We introduce a falsification audit testing complete predictive workflows against synthetic reference classes, including zero-predictability environments and microstructure placebos. Workflows generating significant walk-forward evidence in these environments are falsified. For passing workflows, we quantify selection-induced performance inflation using an absolute magnitude gap linking optimized in-sample evidence to disjoint walk-forward realizations, adjusted for effective multiplicity. Simulations validate extreme-value scaling under correlated searches and demonstrate detection power under genuine structure. Empirical case studies confirm that many apparent findings represent methodological artifacts rather than genuine predictability. |
| Date: | 2026–04 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2604.15531 |
| By: | Ziming Lin; Fang Han |
| Abstract: | Double/debiased machine learning (DML) provides a general framework for inference with high-dimensional or otherwise complex nuisance parameters by combining Neyman-orthogonal scores with cross-fitting, thereby circumventing classical Donsker-type conditions in many modern machine-learning settings. Despite its strong empirical performance, bootstrap inference for DML estimators has received little theoretical justification. This is particularly noteworthy since bootstrap methods are suggested ad used for inference on DML estimators, even though bootstrap procedures can fail for estimators that are root-$n$ consistent and asymptotically normal. This paper fills this gap by establishing bootstrap validity for DML estimators under general exchangeably weighted resampling schemes, with Efron's bootstrap as a special case. Under exactly the same conditions required for the validity of DML itself, we prove that the bootstrap law converges conditionally weakly to the sampling law of the original estimator. |
| Date: | 2026–04 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2604.17239 |
| By: | Napiontek, Jakob (Potsdam Institute for Climate Impact Research (PIK)); Pichler, Peter-Paul |
| Abstract: | Household microdata combining socio-demographic, housing, income and expenditure attributes are a core resource for many studies in quantitative social science, such as modelling the household-level impacts of the energy transition. Yet no such data are openly available for Germany's full population. SynPop-DE provides a synthetic population of 40, 235, 916 households and their 82, 039, 613 members in all 400 German districts, calibrated to the 2022 census, with 34 attributes per household. Synthetic households are generated by estimating the joint attribute distribution of the German Household Budget Survey through a two-stage machine learning architecture. While an autoencoder first compresses high-dimensional categorical data into a continuous latent space, a generative adversarial network subsequently learns to sample new records from this representation. These records are then aligned with census marginals for all German districts using iterative proportional updating to ensure spatial representativeness. Validation along three dimensions confirms that the model learns attribute relationships and generates synthetic households that reproduce the statistical properties of the survey data (fidelity), supports downstream analyses with accuracy comparable to the original survey (utility), and prevents disclosure of individual respondents (privacy). The dataset is openly available at https://synpop.de. |
| Date: | 2026–04–12 |
| URL: | https://d.repec.org/n?u=RePEc:osf:socarx:zha8v_v1 |
| By: | Ramón Talvi Robledo; Christopher Rauh; Ben Seimon; Hannes Mueller; Laura Mayoral |
| Abstract: | Forced displacement is an important policy challenge, yet forecasting is hindered by sparse, annually observed flow data and reporting delays. This article proposes a forecasting method for country outflows and dyadic flows tailored to this sparse data setting. We combine slow-moving structural predictors with high-frequency text-based signals, compress high-dimensional news into low-dimensional topic representations via Latent Dirichlet Allocation to mitigate overfitting, and estimate a stacked ensemble of gradient-boosted trees that captures non-linear origin–destination interactions while making optimal use of the available data. We further apply conformal prediction to construct statistically valid prediction intervals for bilateral flows. Analyzing the text component yields that destination-specific search intensity of migration terms is a central predictor of subsequent dyadic displacement flows. |
| Keywords: | conformal prediction, dyadic, early warning, forced displacement, forecasting, Google trends, machine learning |
| JEL: | P16 C53 D72 |
| Date: | 2026–04 |
| URL: | https://d.repec.org/n?u=RePEc:bge:wpaper:1573 |
| By: | Yikuan Huang; Zheqi Fan; Kaiqi Hu; Yifan Ye |
| Abstract: | Text-based financial networks are increasingly used to study cross-stock return predictability. A common approach constructs links from similarities in firms' disclosure embeddings, but such networks often contain spurious edges because textual proximity does not necessarily imply economic connection. We propose a two-stage framework that first builds a sparse candidate graph from 10-K embeddings and then uses a large language model to classify and filter candidate edges according to their economic relations. The refined graph is used to aggregate pair-level mean-reversion signals into stock-level trading signals with relation-aware and distance-based weights. In a backtest on S&P 500 constituents from 2011 to 2019, LLM-based edge filtering improves the long-short Sharpe ratio from 0.742 to 0.820 and reduces maximum drawdown from $-$10.47% to $-$7.85%. These results suggest that LLM-based reasoning can improve the economic fidelity of text-derived financial networks and strengthen cross-stock predictability. |
| Date: | 2026–04 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2604.19476 |
| By: | Alok Yadav; Saroj Yadav |
| Abstract: | Traditional macroeconomic growth models rely on general equilibrium and continuous, frictionless institutional transitions, failing to account for the catastrophic structural collapses observed in empirical economic history. We propose the Stochastic Networked Governance (SNG) model, a discrete-time, agent-based framework that bridges econophysics, network science, and institutional economics. By defining jurisdictions through a binary institutional genome, the model formalizes institutional complementarity, endogenous growth, and the non-linear macroeconomic penalties of structural reform (the "J-Curve"). Using the CEPII Gravity Database and the IMF Systemic Banking Crises dataset, we move beyond theoretical topologies to execute an empirical historical simulation from 1970 to 2017 across the top 100 global economies. Through Monte Carlo ensembles, we demonstrate how scale-invariant exogenous shocks and spatial capital flight drive global phase transitions, exposing the mathematical mechanics of the 1989-1991 Soviet collapse, the Hub-Risk Paradigm, and the emergent resilience of spatially firewalled market networks. |
| Date: | 2026–04 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2604.19968 |
| By: | George Fatouros; Kostas Metaxas |
| Abstract: | We present the first portfolio-level validation of MarketSenseAI, a deployed multi-agent LLM equity system. All signals are generated live at each observation date, eliminating look-ahead bias. The system routes four specialist agents (News, Fundamentals, Dynamics, and Macro) through a synthesis agent that issues a monthly equity thesis and recommendation for each stock in its coverage universe, and we ask two questions: do its buy recommendations add value over both passive benchmarks and random selection, and what does the internal agent structure reveal about the source of the edge? On the S&P 500 cohort (19 months) the strong-buy equal-weight portfolio earns +2.18%/month against a passive equal-weight benchmark of +1.15% (approximating RSP), a +25.2% compound excess, and ranks at the 99.7th percentile of 10, 000 Monte Carlo portfolios (p=0.003). The S&P 100 cohort (35 months) delivers a +30.5% compound excess over EQWL with consistent direction but formal significance not reached, limited by the small average selection of ~10 stocks per month. Non-negative least-squares projection of thesis embeddings onto agent embeddings reveals an adaptive-integration mechanism. Agent contributions rotate with market regime (Fundamentals leads on S&P 500, Macro on S&P 100, Dynamics acts as an episodic momentum signal) and this agent rotation moves in lockstep with both the sector composition of strong-buy selections and identifiable macro-calendar events, three independent views of the same underlying adaptation. The recommendation's cross-sectional Information Coefficient is statistically significant on S&P 500 (ICIR=+0.489, p=0.024). These results suggest that multi-agent LLM equity systems can identify sources of alpha beyond what classical factor models capture, and that the buy signal functions as an effective universe-filter that can sit upstream of any portfolio-construction process. |
| Date: | 2026–04 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2604.17327 |
| By: | Ethan Ratliff-Crain; Colin M. Van Oort; Matthew T. K. Koehler; Brian F. Tivnan |
| Abstract: | This study strengthens the foundations of multi-venue market modeling by attempting an independent replication of Wah and Wellman's 2016 model of latency arbitrage in a fragmented market. We find that faithful replication is hindered by missing implementation details in the original paper and limited quantitative reporting. We demonstrate that increasing the number of simulation runs beyond the original design allows for the creation of bootstrap confidence intervals to support rigorous tests of quantitative alignment, compensating for lacking distributional information (e.g. variance). We also demonstrate that increased complexity across the modeled scenarios corresponds with increased difficulty aligning to the original results. We draw on a codebase released by the original authors in connection with a later paper to recover additional implementation details; however, we reject quantitative alignment between that codebase and the published results. Combining information from the paper and the released code, we achieve relational equivalence for most metrics but reject quantitative alignment for model settings where latency is non-zero. We show that many of the qualitative takeaways from the original paper on the effects of market fragmentation and latency arbitrage are sensitive to the specifics of a `greedy strategy' extension given to the zero-intelligence (ZI) trader agents. Under an alternative interpretation of this strategy, we find that market fragmentation decreases execution times in all experiments and increases trader welfare in most experiments. Finally, to facilitate future replication, critique, and extension, we provide an ODD (Overview, Design concepts, Details) protocol for our implementations of the model. |
| Date: | 2026–04 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2604.20067 |
| By: | Antoine Didisheim; Bryan T. Kelly; Mohammad Pourmohammadi; Hanqing Tian |
| Abstract: | The stock market fails to efficiently process information in news text (Chen et al., 2026). But news itself is highly predictable by prevailing stock characteristics, which complicates inferences about market efficiency. After purging news of its predictable content, the resulting “news shocks” more than double the monthly return predictive power of raw news, and they continue to significantly predict returns up to 18 months ahead. The magnitude and longevity of the news shock anomaly is larger than every anomaly in the Jensen et al. (2022) universe. The news shock anomaly derives from negative-tone and quantitative topics to which investors underreact and from high-attention and ambiguous topics to which investors overreact. |
| JEL: | C45 C58 G02 G1 G11 G12 G14 G17 G40 G41 |
| Date: | 2026–04 |
| URL: | https://d.repec.org/n?u=RePEc:nbr:nberwo:35093 |
| By: | Andrew Y. Chen |
| Abstract: | AI stocks trade at extraordinary valuations. We develop an asset pricing model in which investors use AI stocks to hedge against an AI singularity that displaces their consumption. Because markets are incomplete -- investors cannot trade private AI capital -- AI stocks command a premium. Market incompleteness distorts both valuations and the efficient development of AI, creating a rationale for government transfers that becomes compelling when singularity-driven growth overwhelms deadweight costs. This paper was generated by AI, using https://github.com/chenandrewy/ralph-wig gum-asset-pricing/. |
| Date: | 2026–04 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2604.16997 |