We study calibeating, the problem of post-processing external forecasts online to minimize cumulative losses and match an informativeness-based benchmark. Unlike prior work, which analyzed calibeating for specific losses with specific arguments, we reduce calibeating to existing online learning techniques and obtain results for general proper losses. More concretely, we first show that calibeating is minimax-equivalent to regret minimization. This recovers the $O(\log T)$ calibeating rate of Foster and Hart [FH23] for the Brier and log losses and its optimality, and yields new optimal calibeating rates for mixable losses and general bounded losses. Second, we prove that multi-calibeating is minimax-equivalent to the combination of calibeating and the classical expert problem. This yields new optimal multi-calibeating rates for mixable losses, including Brier and log losses, and general bounded losses. Finally, we obtain new bounds for achieving calibeating and calibration simultaneously for the Brier loss. For binary predictions, our result gives the first calibrated algorithm that at the same time also achieves the optimal $O(\log T)$ calibeating rate.

This paper addresses the challenge of time-inconsistent stochastic control within a continuous-time framework. Its primary focus lies in uncovering a probabilistic representation, specifically in the shape of a system of backward stochastic differential equations (BSDEs). These equations encapsulate the equilibrium value function essential for resolving cases where the present state affecting the target functional triggers the inconsistency. Additionally, the paper offers an application exemplifying this theory through the time-inconsistent linear--quadratic regulator.

This paper models firm-to-firm trade in a production network as a set of double auctions. Firms have multilateral market power, namely, can affect prices in both input and output markets. The size and division of surplus are endogenous and depend only on technology, network position, and consumer preferences. The standard simplifying assumption of price-taking on input markets (unilateral market power) has systematic effects: it underestimates the final price and overestimates the surplus going upstream. These phenomena affect the model predictions for the welfare impact of mergers.

A number of recent contributions have put forward the topological structure of production networks as a key determinant of macro-economic dynamics. However, firm-to-firm production networks data is generally not available. Against this background, reconstruction method based on firms' size have been developed. This paper enriches this set of reconstruction methods by integrating input-output sectoral flows in the reconstruction process. We derive analytical expressions for the maximum entropy solutions to the firm network reconstruction problem with sectoral input-output constraints, first for binary networks and then for weight reconstruction. We perform a numerical analysis comparing standard and input-output based reconstruction methods using Hungarian production network data. Our results show that adding input-output constraints substantially reduces deviations from the input-output structure compared with standard methods. Our augmented method provides an almost perfect fit to input-output data, though all methods have difficulties reproducing other structural characteristics.

We study the associations between everyday economic decision-making quality and people's emotional states. Using high-frequency, highly disaggregated consumer "scanner" data, we show that the cost of poor decision-making is substantial, on average equal to around half of day-to-day consumption budgets. While material circumstances help explain decision-making quality, how people feel about those circumstances is equally important. Contrary to evidence that stress and worry impair performance in settings where distraction is costly, we find these same feelings are associated with improved decision-making for frequently made consumption choices. This is consistent with worry increasing attentiveness to decisions within households' locus of control.

We study an information acquisition problem in which an informed trader acquires costly information prior to trading in the Kyle equilibrium. The cost of information acquisition is represented by an entropy cost. Regardless of the prior distribution of the asset payoff, continuous signals are optimal. Moreover, any continuously distributed signal, together with an associated logit type posterior distribution of the payoff, yields the same ex-ante value for the informed trader, the same distribution of posterior expected payoff, and the same unconditional distribution of the informed trader's trading strategy. Consequently, a normally distributed signal can be adopted without loss of generality. We further show that when the information acquisition cost increases or the volatility of noise trades decreases, the variance of the posterior expected payoff declines, the profit potential from trading diminishes, meanwhile the posterior expected payoff increasingly resembles a normal distribution, and the information leakage cost from trading decreases.

Recommendation systems (RSs) are increasingly used to guide job seekers on online platforms, yet the algorithms currently deployed are typically optimized for predictive objectives such as clicks, applications, or hires, rather than job seekers' welfare. We develop a job-search model with an application stage in which the value of a vacancy depends on two dimensions: the utility it delivers to the worker and the probability that an application succeeds. The model implies that welfare-optimal RSs rank vacancies by an expected-surplus index combining both, and shows why rankings based solely on utility, hiring probabilities, or observed application behavior are generically suboptimal, an instance of the inversion problem between behavior and welfare. We test these predictions and quantify their practical importance through two randomized field experiments conducted with the French public employment service. The first experiment, comparing existing algorithms and their combinations, provides behavioral evidence that both dimensions shape application decisions. Guided by the model and these results, the second experiment extends the comparison to an RS designed to approximate the welfare-optimal ranking. The experiments generate exogenous variation in the vacancies shown to job seekers, allowing us to estimate the model, validate its behavioral predictions, and construct a welfare metric. Algorithms informed by the model-implied optimal ranking substantially outperform existing approaches and perform close to the welfare-optimal benchmark. Our results show that embedding predictive tools within a simple job-search framework and combining it with experimental evidence yields recommendation rules with substantial welfare gains in practice.

As large language models (LLMs) and vision-language-action models (VLAs) become widely deployed, the tokens consumed by AI inference are evolving into a new type of commodity. This paper systematically analyzes the commodity attributes of tokens, arguing for their transition from intelligent service outputs to compute infrastructure raw materials, and draws comparisons with established commodities such as electricity, carbon emission allowances, and bandwidth. Building on the historical experience of electricity futures markets and the theory of commodity financialization, we propose a complete design for standardized token futures contracts, including the definition of a Standard Inference Token (SIT), contract specifications, settlement mechanisms, margin systems, and market-maker regimes. By constructing a mean-reverting jump-diffusion stochastic process model and conducting Monte Carlo simulations, we evaluate the hedging efficiency of the proposed futures contracts for application-layer enterprises. Simulation results show that, under an application-layer demand explosion scenario, token futures can reduce enterprise compute cost volatility by 62%-78%. We also explore the feasibility of GPU compute futures and discuss the regulatory framework for token futures markets, providing a theoretical foundation and practical roadmap for the financialization of compute resources.

We establish concentration rates for estimation of treatment effects in experiments that incorporate prior sources of information -- such as past pilots, related studies, or expert assessments -- whose external validity is uncertain. Each source is modeled as a Gaussian prior with its own mean and precision, and sources are combined using Bayesian model averaging (BMA), allowing data from the new experiment to update posterior weights. To capture empirically relevant settings in which prior studies may be as informative as the current experiment, we introduce a nonstandard asymptotic framework in which prior precisions grow with the experiment's sample size. In this regime, posterior weights are governed by an external-validity index that depends jointly on a source's bias and information content: biased sources are exponentially downweighted, while unbiased sources dominate. When at least one source is unbiased, our procedure concentrates on the unbiased set and achieves faster convergence than relying on new data alone. When all sources are biased, including a deliberately conservative (diffuse) prior guarantees robustness and recovers the standard convergence rate.

This paper incorporates fixed capital into a multi-sectoral input-output model to reassess the Okishio Theorem. We establish the existence of a critical wage elasticity strictly less than unity, beyond which cost-reducing technical progress leads to a declining equilibrium rate of profit. This implies that profit rates may fall even under Kaldor's Stylized Facts or a moderately declining labour share, significantly extending the theorem's domain of validity. Game-theoretic analysis reveals a strict Prisoner's Dilemma structure underlying technical adoption. Empirical evidence from Chinese industrial data confirms that fixed capital intensity exerts a significant dampening effect on the profit-enhancing impact of productivity growth.

This paper develops a novel multi-agent reinforcement learning (MARL) framework for reinsurance treaty bidding, addressing long-standing inefficiencies in traditional broker-mediated placement processes. We pose the core research question: Can autonomous, learning-based bidding systems improve risk transfer efficiency and outperform conventional pricing approaches in reinsurance markets? In our model, each reinsurer is represented by an adaptive agent that iteratively refines its bidding strategy within a competitive, partially observable environment. The simulation explicitly incorporates institutional frictions including broker intermediation, incumbent advantages, last-look privileges, and asymmetric access to underwriting information. Empirical analysis demonstrates that MARL agents achieve up to 15% higher underwriting profit, 20% lower tail risk (CVaR), and over 25% improvement in Sharpe ratios relative to actuarial and heuristic baselines. Sensitivity tests confirm robustness across hyperparameter settings, and stress testing reveals strong resilience under simulated catastrophe shocks and capital constraints. These findings suggest that MARL offers a viable path toward more transparent, adaptive, and risk-sensitive reinsurance markets. The proposed framework contributes to emerging literature at the intersection of algorithmic market design, strategic bidding, and AI-enabled financial decision-making.

Reinsurance optimization is a cornerstone of solvency and capital management, yet traditional approaches often rely on restrictive distributional assumptions and static program designs. We propose a hybrid framework that combines Variational Autoencoders (VAEs) to learn joint distributions of multi-line and multi-year claims data with Proximal Policy Optimization (PPO) reinforcement learning to adapt treaty parameters dynamically. The framework explicitly targets expected surplus under capital and ruin-probability constraints, bridging statistical modeling with sequential decision-making. Using simulated and stress-test scenarios, including pandemic-type and catastrophe-type shocks, we show that the hybrid method produces more resilient outcomes than classical proportional and stop-loss benchmarks, delivering higher surpluses and lower tail risk. Our findings highlight the usefulness of generative models for capturing cross-line dependencies and demonstrate the feasibility of RL-based dynamic structuring in practical reinsurance settings. Contributions include (i) clarifying optimization goals in reinsurance RL, (ii) defending generative modeling relative to parametric fits, and (iii) benchmarking against established methods. This work illustrates how hybrid AI techniques can address modern challenges of portfolio diversification, catastrophe risk, and adaptive capital allocation.

Using only a binary intervention and outcome and the design of the randomization within an experiment, we construct a design-based likelihood of the joint distribution of potential outcomes in the sample -- the numbers of always takers, compliers, defiers, and never takers. We develop a visualization to show that samples with defiers can sometimes generate the data in more ways than samples without, yielding a higher likelihood. This likelihood can vary within the Frechet bounds, even though the traditional likelihood does not. Evidence is weak, but it exists, as we illustrate with health applications and our dbmle package.

Algorithms increasingly automate bidding in online auctions, raising concerns about tacit bid suppression and revenue shortfalls. Prior work identifies individual mechanisms behind algorithmic bid suppression, but it remains unclear which factors matter most and how they interact, and policy conclusions rest on algorithms unlike those deployed in practice. This paper develops a computational laboratory framework, based on factorial experimental designs and large-scale Monte Carlo simulation, that addresses bid suppression across multiple algorithm classes within a common methodology. Each simulation is treated as a black-box input-output observation; the framework varies inputs and ranks factors by association with outcomes, without explaining algorithms' internal mechanisms. Across six sub-experiments spanning Q-learning, contextual bandits, and budget-constrained pacing, the framework ranks the relative importance of auction format, competitive pressure, learning parameters, and budget constraints on seller revenue. The central finding is that structural market parameters dominate algorithmic design choices. In unconstrained settings, competitive pressure is the strongest predictor of revenue; under budget constraints, budget tightness takes over. The auction-format effect is context-dependent, favouring second-price under learning algorithms but reversing to favour first-price under budget-constrained pacing. Because the optimal format depends on the prevailing bidding technology, no single auction format is universally superior when bidders are algorithms, and applying format recommendations from one algorithm class to another leads to counterproductive design interventions.

This paper investigates a novel behavioral feature of recursive preferences: aversion to risks that persist over time, or simply \textit{correlation aversion}. Greater persistence provides information about future consumption but reduces opportunities to hedge consumption risk. I show that, for recursive preferences that exhibit a preference for early resolution of uncertainty, correlation aversion is equivalent to increasing relative risk aversion. To quantify correlation aversion, I develop the concept of the persistence premium, which measures how much an individual is willing to pay to eliminate persistence in consumption. I provide an approximation of the persistence premium in the spirit of Arrow--Pratt, which provides a quantitative representation of the trade-off between information and hedging. I show that correlation-averse preferences have a variational representation, linking correlation aversion to concerns about model misspecification. I present several applications. I first illustrate how correlation aversion shapes portfolio choices, and then show how the persistence premium can improve the calibration of macro-finance models. In an optimal taxation model, I show that recursive preferences -- unlike standard preferences -- lead to redistributive tax policies that increase social mobility.

This paper develops numerical and causal interpretations of two-way fixed effects (TWFE) regressions in settings with nonbinary, nonstaggered treatments and time-varying covariates. Using the equivalence between TWFE and pooled first-difference regressions, I express the TWFE coefficient as a weighted average of first-difference coefficients across all horizons, clarifying how short- and long-run changes contribute to the estimate. Causal interpretation relies on common-trends assumptions across all horizons and conditioning on covariate changes rather than levels. I propose diagnostic procedures to assess these assumptions across horizons and illustrate them by reexamining TWFE estimates of minimum-wage effects on employment.

When organisations adopt commercial AI systems for decision support, they inherit value judgements embedded by vendors that are neither transparent nor renegotiable. The governance puzzle is not whether AI can support decisions but which recommendations the system can actually produce given how its vendor has configured it. I formalise this as a behavioural feasible set, the range of recommendations reachable under vendor-imposed alignment constraints, and characterise diagnostic thresholds for when organisational requirements exceed the system's flexibility. In scenario-based experiments using binary decision scenarios and multi-stakeholder ranking tasks, I show that alignment materially compresses this set. Comparing pre- and post-alignment variants of an open-weight model isolates the mechanism: alignment makes the system substantially less able to shift its recommendation even under legitimate contextual pressure. Leading commercial models exhibit comparable or greater rigidity. In multi-stakeholder tasks, alignment shifts implied stakeholder priorities rather than neutralising them, meaning organisations adopt embedded value orientations set upstream by the vendor. Organisations thus face a governance problem that better prompting cannot resolve: selecting a vendor partially determines which trade-offs remain negotiable and which stakeholder priorities are structurally embedded.

Extreme economic outcomes are not shaped by tails alone. They are also shaped by unequal access to opportunities. This paper develops a theory of heterogeneous extremes by taking the distribution of opportunity access as the object of study. In a mixed Poisson search setting, normalized maxima admit a Laplace mixture representation that yields order comparisons and a clean benchmark against the homogeneous economy. The main contribution is geometric: a canonical coupling turns differences in heterogeneity into optimal transport bounds for the whole induced law of extremes, the full schedule of top quantiles, and structured counterfactual paths between economies. The paper also derives a second order expansion that separates classical extreme value approximation error from heterogeneity effects. As a complementary normative exercise, it studies an entropy regularized design problem for reallocating opportunities under a mean constraint. A stylized labor market network application interprets heterogeneity as unequal access to job opportunities and shows how the framework can be used for tail counterfactuals and robustness analysis of top wage distributions.

We characterize optimal monetary policy when policy endogenously moves risk premia through redistribution across agents who differ in their willingness to bear risk. The analytical core is Marginal Risk Capacity, the covariance of monetary policy exposures with marginal propensities to take risk. This sufficient statistic governs this channel as MPCs govern the consumption channel. MRC enters the Ramsey criterion as a risk premium wedge that breaks divine coincidence, vanishes if and only if macroprudential tools are available, and generates a new inflation bias under discretion. Solving the Ramsey problem globally reveals a risk capacity trap where transmission collapses, and optimal policy preemptively prevents it.

We characterize the maximal attainable power-size gap in overidentified instrumental variables models with heteroskedastic or autocorrelated (HAC) errors. Using total variation distance and Kraft's theorem, we define the decision theoretic frontier of the testing problem. We show that Lagrange multiplier and conditional quasi likelihood ratio tests can have power arbitrarily close to size even when the null and alternative are well separated, because they do not fully exploit the reduced-form likelihood. In contrast, the conditional likelihood ratio (CLR) test uses the full reduced-form likelihood. We prove that the power-size gap of CLR converges to one if and only if the testing problem becomes trivial in total variation distance, so that CLR attains the decision theoretic frontier whenever any test can. An empirical illustration based on Yogo (2004) shows that these failures arise in empirically relevant configurations.

E-commerce is shifting from search-based shopping to agentic purchasing. Rather than relying on keywords, AI shopping agents learn customer preferences through targeted multi-round conversations and then recommend a tailored set of products. We develop a solicit-then-suggest framework to study this setting. In a d-dimensional preference space, an agent conducts m rounds of solicitation to refine its belief about the customer's ideal product, then recommends k products from which the customer chooses. Our analysis identifies the key economic tradeoff. Under a Gaussian prior, we establish an uncertainty decomposition: solicitation depth and assortment breadth are substitutes, with total prior uncertainty split between what solicitation resolves and what assortment breadth hedges. The two instruments improve match quality at very different rates. Expected loss decreases on the order of 1/m with solicitation depth, but only on the order of k^(-2/d) with assortment breadth, reflecting a curse of dimensionality. Thus, a few well-designed questions can achieve what would otherwise require far more recommendations. We also characterize the optimal policy. The optimal assortment forms a Voronoi partition, assigning each product to the posterior region it best serves. With a single recommended product, the optimal solicitation follows a water-filling rule that equalizes posterior uncertainty across dimensions. With multiple products, the optimum may allocate less precision to dimensions that the assortment can hedge. This single-product water-filling rule also yields a general approximation guarantee for larger assortments, and the gap vanishes as dimension grows. Beyond the Gaussian case, the uncertainty decomposition and substitutability between solicitation depth and assortment breadth continue to hold for non-Gaussian priors.

The Riesz representer is a central object in semiparametric statistics and debiased/doubly-robust estimation. Two literatures in econometrics have highlighted the role for directly estimating Riesz representers: the automatic debiased machine learning literature (as in Chernozhukov et al., 2022b), and an independent literature on sieve methods for conditional moment models (as in Chen et al., 2014). These two literatures solve distinct optimization problems that in the population both have the Riesz representer as their solution. We show that with unregularized or ridge-regularized linear, sieve, or RKHS models, the two resulting estimators are numerically equivalent. However, for other regularization schemes such as the Lasso, or more general machine learning function classes including neural networks, the estimators are not necessarily equivalent. In the latter case, the Chen et al. (2014) formulation yields a novel constrained optimization problem for directly estimating Riesz representers with machine learning. Drawing on results from Birrell et al. (2022), we conjecture that this approach may offer statistical advantages at the cost of greater computational complexity.

This paper investigates the barriers to gender convergence using Japan as a salient environment to explore the interactive effects of labor market structures and social norms. I develop a quantitative model of household labor supply where couples jointly decide their occupations and working hours. The model features a labor market with inflexible "regular" jobs with convex pay schedules and flexible "non-regular" jobs, interacting with social norms regarding spousal earnings. The calibrated model successfully reproduces observed gender gaps in participation, occupation, and working hours, and explains 48% of the gender wage gap. The model also accounts for cross-regional differences in gender gaps solely through variation in social norms. Counterfactual simulations show that while increasing job flexibility substantially reduces wage and occupational gaps, the working hours gap persists due to the unequal burden of domestic work. Closing this remaining gap requires policies such as affordable household services. Furthermore, the model suggests that the effects of structural reforms can depend on the strength of gender norms, with larger reductions in gender gaps in more conservative environments.

We study the employment effects of the 22% increase in the Spanish minimum wage in 2019, focusing on young workers. Using census-grade administrative tax data covering the universe of formal wage bills and employment (Models 190/390 linked to personal income tax records), we construct several measures of treatment intensity, including two structurally grounded bite indicators based on the incidence of young minimum-wage workers and the implied increase in the wage bill obtained via Exponential Tilting. Difference-in-differences estimates with two-way fixed effects, dynamic event-study specifications, and robust confidence intervals from the HonestDiD framework all point to the same conclusion: the reform did not generate net disemployment effects for young workers. Point estimates of the elasticity are small and often positive, and confidence internals comfortably include zero even with sizable deviations from parallel trends. A triple-difference design exploiting pre-existing tourism dependence further shows that the sharp employment collapse of 2020 is primarily explained by the COVID-19 shock operating through tourism-intensive sectors, rather than by the minimum-wage hike itself. Our results suggest that, in the macroeconomic and institutional environment prevailing in Spain in 2019, with the minimum wage rising to around 60% of the average wage in a recovering economy, the labour market absorbed a large discrete increase in the wage floor without destroying aggregate youth employment. More broadly, the paper highlights how the choice of treatment definition, the use of census-grade data, robust DiD inference, and explicit modelling of concurrent shocks can shape conclusions about the effects of minimum-wage policies.

The Nobel Memorial Prize in Economics has been awarded annually since 1969. Who wins the prize is a topic of much interest and tracks the whole course of the academic discipline over the last 57 years. Explaining who wins the prize in any given year is a complex process, which involves the subtle endogeneity of the choice of the field and the individual(s) who should be honoured. Citations, track records, networks of past winners, institutional factors along with field rotation and Economic Prize Committee composition may all play a role. A dynamic sample involving a changing stock of would-be candidates along with a moving flow -- both into and out of the sample -- add complexities to the modelling. We find robust evidence that the Nobel Prize rotates in a semi-regular way between the fields of economics. Earlier awards were for a single paper, later ones for a body of work. Networks do not matter, but having a Nobel student or co-author does. There is some evidence that the personal preferences of Committee members had an effect on either field or individual winner. The Committee's decisions changed after Lindbeck retired.

Contemporary societies face a severe crisis of demographic reproduction. Global fertility rates continue to decline precipitously, with East Asian nations exhibiting the most dramatic trends -- China's total fertility rate (TFR) fell to approximately 1.0 in 2023, while South Korea's dropped below 0.72. Simultaneously, the institution of marriage is undergoing structural disintegration: educated women rationally reject unions lacking both emotional fulfillment and economic security, while a growing proportion of men at the lower end of the socioeconomic spectrum experience chronic sexual deprivation, anxiety, and learned helplessness. This paper proposes a computational framework for modeling and evaluating a Stratified Polyamory System (SPS) using techniques from agent-based modeling (ABM), multi-agent reinforcement learning (MARL), and large language model (LLM)-empowered social simulation. The SPS permits individuals to maintain a limited number of legally recognized secondary partners in addition to one primary spouse, combined with socialized child-rearing and inheritance reform. We formalize the A/B/C stratification as heterogeneous agent types in a multi-agent system and model the matching process as a MARL problem amenable to Proximal Policy Optimization (PPO). The mating network is analyzed using graph neural network (GNN) representations. Drawing on evolutionary psychology, behavioral ecology, social stratification theory, computational social science, algorithmic fairness, and institutional economics, we argue that SPS can improve aggregate social welfare in the Pareto sense. Preliminary computational results demonstrate the framework's viability in addressing the dual crisis of female motherhood penalties and male sexlessness, while offering a non-violent mechanism for wealth dispersion analogous to the historical Chinese Grace Decree (Tui'en Ling).

We study optimal dynamic persuasion in a bandit experimentation model where a principal, unlike in standard settings, has a single-peaked preference over the agent's stopping time. This non-monotonic preference arises because maximizing the agent's effort is not always in the principal's best interest, as it may lead to a dead end. The principal privately observes the agent's payoff upon success and uses the information as the instrument of incentives. We show that the optimal dynamic information policy involves at most two one-shot disclosures: an accelerator before the principal's optimal stopping time, persuading the agent to be optimistic, and a brake after the principal's optimal stopping time, persuading the agent to be pessimistic. A key insight of our analysis is that the optimal disclosure pattern -- whether gradual or one-shot -- depends on how the principal resolves a trade-off between the mean of stopping times and its riskiness. We identify the Arrow-Pratt coefficient of absolute risk aversion as a sufficient statistic for determining the optimal disclosure structure.

We develop a theoretical framework for sample splitting in A/B testing environments, where data for each test are partitioned into two splits to measure methodological performance when the true impacts of tests are unobserved. We show that sample-split estimators are generally biased for full-sample performance but consistently estimate sample-split analogues of it. We derive their asymptotic distributions, construct valid confidence intervals, and characterize the bias-variance trade-offs underlying sample-split design choices. We validate our theoretical results through simulations and provide implementation guidance for A/B testing products seeking to evaluate new estimators and decision rules.

Economic decision-making depends not only on structured signals such as prices and taxes, but also on unstructured language, including peer dialogue and media narratives. While multi-agent reinforcement learning (MARL) has shown promise in optimizing economic decisions, it struggles with the semantic ambiguity and contextual richness of language. We propose LAMP (Language-Augmented Multi-Agent Policy), a framework that integrates language into economic decision-making and narrows the gap to real-world settings. LAMP follows a Think-Speak-Decide pipeline: (1) Think interprets numerical observations to extract short-term shocks and long-term trends, caching high-value reasoning trajectories; (2) Speak crafts and exchanges strategic messages based on reasoning, updating beliefs by parsing peer communications; and (3) Decide fuses numerical data, reasoning, and reflections into a MARL policy to optimize language-augmented decision-making. Experiments in economic simulation show that LAMP outperforms both MARL and LLM-only baselines in cumulative return (+63.5%, +34.0%), robustness (+18.8%, +59.4%), and interpretability. These results demonstrate the potential of language-augmented policies to deliver more effective and robust economic strategies.

We introduce Agentic Economic Modeling (AEM), a framework that aligns synthetic LLM choices with small-sample human evidence for reliable econometric inference. AEM first generates task-conditioned synthetic choices via LLMs, then learns a bias-correction mapping from task features and raw LLM choices to human-aligned choices, upon which standard econometric estimators perform inference to recover demand elasticities and treatment effects.We validate AEM in two experiments. In a large scale conjoint study with millions of observations, using only 10% of the original data to fit the correction model lowers the error of the demand-parameter estimates, while uncorrected LLM choices even increase the errors. In a regional field experiment, a mixture model calibrated on 10% of geographic regions estimates an out-of-domain treatment effect of -65\pm10 bps, closely matching the full human experiment (-60\pm8 bps).Under time-wise extrapolation, training with only day-one human data yields -24 bps (95% CI: [-26, -22], p<1e-5),improving over the human-only day-one baseline (-17 bps, 95% CI: [-43, +9], p=0.2049).These results demonstrate AEM's potential to improve RCT efficiency and establish a foundation method for LLM-based counterfactual generation.