This paper introduces a quasi-Bayesian approach for local projection instrumental-variables (LP-IV) estimation. It builds a moment-based quasi-posterior using the generalized method of moments (GMM) objective and applies a roughness-penalty prior to smooth impulse responses over different horizons. The approach maintains the key first-order features of traditional LP-IV methods, while enhancing stability in finite samples and allowing for joint inference through simultaneous bands. Simulations indicate that this regularization decreases root mean squared error compared to standard GMM, especially at medium and longer horizons. An application to Danish electricity markets highlights the method's practical usefulness.

We study a contest-theoretic model of adversarial investment in which an attacker and a defender allocate resources to AI-augmented capabilities across multiple attack surfaces. The attacker's investment operates through two channels: it amplifies offensive potency unconditionally and erodes defensive effectiveness conditionally, generating an adversarial discount that deepens endogenously with the defender's own investment. We derive a closed-form arms race ratio decomposing the relative marginal effectiveness of offensive and defensive investment into six structural primitives and establish equilibrium uniqueness and global convergence under a continuous best-response dynamic. The central result concerns signal cross-correlation, the degree to which threat intelligence on one surface informs detection on another. With full cross-correlation, the arms race ratio is independent of the number of attack surfaces: the attacker's structural advantage from surface proliferation is completely neutralized. Under the benchmark full-dilution case, without cross-correlation, per-surface defense effectiveness vanishes as the attack surface grows. Extending the analysis to heterogeneous defenders facing an attacker who targets by expected value, we argue that the model points to a dual inefficiency: overinvestment in private defense (a zero-sum redirective externality) and underinvestment in shared signal correlation (a public good). These formal results, together with public-good reasoning outside the base model, characterize when collective information aggregation can dominate private capability investment as the decisive margin in adversarial contests.

Restricting individuals' access to some opportunities may steer their desire toward their substitutes, a phenomenon known as the forbidden fruit effect. We axiomatize a choice model named restriction-sensitive choice (RSC), which rationalizes the forbidden fruit effect and is compatible with the prominent psychological explanations: reactance theory and commodity theory. The model is identifiable from choice data, specifically from the observation of choice reversals caused by the removal of options. We conduct a normative analysis both in terms of the agent's freedom and welfare. We apply our model to shed light on two phenomena: the backfire effect of beliefs and the backlash of integration policies targeted towards minorities.

This paper introduces the Generality-Accuracy-Simplicity (GAS) framework to analyze how large language models (LLMs) are reshaping organizations and competitive strategy. We argue that viewing AI as a simple reduction in input costs overlooks two critical dynamics: (a) the inherent trade-offs among generality, accuracy, and simplicity, and (b) the redistribution of complexity across stakeholders. While LLMs appear to defy the traditional trade-off by offering high generality and accuracy through simple interfaces, this user-facing simplicity masks a significant shift of complexity to infrastructure, compliance, and specialized personnel. The GAS trade-off, therefore, does not disappear but is relocated from the user to the organization, creating new managerial challenges, particularly around accuracy in high-stakes applications. We contend that competitive advantage no longer stems from mere AI adoption, but from mastering this redistributed complexity through the design of abstraction layers, workflow alignment, and complementary expertise. This study advances AI strategy by clarifying how scalable cognition relocates complexity and redefines the conditions for technology integration.

This paper offers a commentary on the use of notions of statistical significance in choice modelling. We review the reasons for uncertainty in parameter estimates, provide a precise discussion on the computation of measures of uncertainty and confidence intervals, and discuss the use of statistical tests. We argue that, as in many other areas of science, there is an over-reliance on 95\% confidence levels, and misunderstandings of the meaning of significance. We also observe a lack of precision in the reporting of measures of uncertainty in many studies, especially when using $p$-values and even more so with \emph{star} measures. The paper also stresses the importance of considering behavioural or policy significance in addition to statistical significance. Finally, we stress a number of points that are specific to choice modelling and which require special attention, notably in relation to derived measures such as willingness-to-pay, the treatment of random heterogeneity, and the use of repeated choice data.

We establish the existence and uniqueness of the equilibrium for a stochastic mean-field game of optimal investment. The analysis covers both finite and infinite time horizons, and the mean-field interaction of the representative company with a mass of identical and indistinguishable firms is modeled through the time-dependent price at which the produced good is sold. At equilibrium, this price is given in terms of a nonlinear function of the expected (optimally controlled) production capacity of the representative company at each time. The proof of the existence and uniqueness of the mean-field equilibrium relies on a priori estimates and the study of nonlinear integral equations, but employs different techniques for the finite and infinite horizon cases. Additionally, we investigate the deterministic counterpart of the mean-field game under study.

Macroeconomic dynamics is typically modeled under the assumption that the economy evolves according to a single invariant law of motion. This paper shows that this assumption imposes a structural restriction. We develop Dynamic Macroeconomics with Multiple Regimes (DMR), a framework in which economic evolution is governed by multiple regime-specific propagation operators. As a result, trajectories arise from ordered compositions of heterogeneous operators rather than from the iteration of a single mapping. We establish three structural results. First, invariant-law and regime-dependent systems are not topologically equivalent. Second, regime dependence is dynamically irreducible: it cannot be eliminated through any injective transformation of the state space. Third, whenever regime operators fail to commute, there exists no map $F:\mathbb{R}^n\to\mathbb{R}^n$ whose iterates reproduce all regime-admissible trajectories. These results establish a structural separation between invariant-law macroeconomics and regime-dependent dynamics, implying that stability, policy evaluation, and structural characterization must be conducted at the level of interacting propagation operators rather than within a single invariant mapping.

Many tractable TANK models treat redistribution as a contemporaneous wedge. I show that this view is incomplete once wage contracts are type-specific. In a tractable Two-Agent New Keynesian model, each household type adjusts its nominal wage relative to its own previous wage. This own-lag contract makes the cross-type wage gap a payoff-relevant distributional state variable. The wage gap follows a second-order expectational law of motion and feeds back into aggregate demand through consumption dispersion. Inflation stabilization or contemporaneous profit-wedge neutralization therefore generally fails to restore the corresponding representative-agent allocation. Under the maintained commitment benchmark, RANK-equivalent stabilization from period $t=1$ onward requires history-dependent transfers that respond to inherited wage dispersion, not only current profits. In the benchmark calibration, wage rigidity raises the peak output response to a transfer shock by a factor of 3.27, from $2.49\times 10^{-4}$ to $8.14\times 10^{-4}$.

We develop a model of choice over social norms that allows for complementarities along two dimensions: \textit{technological}, analogous to complementarities between consumption goods, and social, capturing returns from conformity. Together, these determine whether two norms are complements, substitutes, or independent, as defined by how the equilibrium prevalence of one norm responds to a marginal shift in the utility of another. We estimate the model using repeated cross-sections from Sierra Leone and Nigeria, focusing on female genital cutting, polygyny, and child marriage. Social returns are significant across all specifications. For female genital cutting and child marriage, we find evidence of complementarities, especially strong in Sierra Leone. For polygyny and child marriage, we find evidence of social substitutability, particularly in Nigeria. We interpret these differences using insights from anthropology. Finally, we iterate the model forward to study policy counterfactuals, assessing the potential effects of legal reforms and social interventions.

We study double descent and benign overfitting in macroeconomic forecasting. We document that double-descent risk curves arise in standard macroeconomic datasets that are driven by a small number of latent factors, and we characterize when the underlying benign-overfitting mechanism holds. The conditions of Bartlett et al. (2020) are satisfied under the exact factor model and can also hold under the more realistic approximate factor model, provided idiosyncratic variances are not too dispersed across series. Because macroeconomic panels have only moderate dimensions, the overparameterization ratio N/T required by the theory is not naturally available. Our solution is to augment the data with synthetic copies from an estimated factor model and we prove that this strategy converges to a kernel ridge regression with a factor-structured kernel. Using monthly (FRED-MD) and quarterly (FRED-QD) US data, the resulting estimator consistently outperforms the Stock-Watson factor model for point forecasting across all series and horizons, with gains that are pervasive, statistically significant, and increasing with the forecast horizon. Our results suggest that benign overfitting, when it works, succeeds because overparameterization implicitly constructs a well-behaved kernel, not because overparameterization is intrinsically desirable.

Selecting the right electricity market region for a hyperscale AI datacenter requires reasoning across live electricity prices, grid carbon intensity, technology cost trajectories, and causal grid dynamics -- a multi-step, multi-source analytical task that static knowledge benchmarks cannot evaluate. We introduce EnergyAgentBench, the first agentic benchmark grounded in live electricity market data for this problem class. The benchmark comprises 70 task variants across five families: datacenter siting under cost-carbon trade-offs (F1), long-horizon portfolio siting (F1-LH), lifetime LCOE ranking over multi-decade cost trajectories (F2), 30-year portfolio optimization (F2-LH), and causal grid diagnosis (F3). Tasks require 3 to 48 sequential tool calls against live endpoints from the QuarluxAI infrastructure platform, the U.S. Energy Information Administration (EIA), and the National Renewable Energy Laboratory (NREL) with ground truth derived from trained XGBoost cost-surface models (R^2 0.967--0.995) and the NREL Annual Technology Baseline 2024. We evaluate nine models across Anthropic, OpenAI, and HuggingFace over 1,414 runs at three random seeds. Claude Sonnet 4.6 achieves the highest overall score (0.900) at one-quarter the cost of Claude Opus 4.7 (0.889). Claude Haiku 4.5 leads on long-horizon procedural siting (0.986), outperforming all frontier models including those costing 16x more per run. F3 Causal is the most discriminating family, with a 30.7-point spread between Sonnet (0.793) and Llama 3.3 70B (0.486), versus a 6.6-point spread on F1 Siting. A failure taxonomy of 135 coded failures identifies null-value integration in NREL ATB trajectories as the dominant failure mode (70%), followed by premature commitment on causal tasks (20%) and adversarial injection blindness (6%). Benchmark code, run trajectories, and the failure taxonomy dataset are publicly released.

Instruction-tuned language models exhibit behavioural fairness in high-stakes decisions while retaining biased associations in their internal representations. However, whether these suppressed representations can affect model outputs - and whether such causal potency is symmetric across demographic groups - remains unknown. We investigate the use of open-weight models for mortgage underwriting using matched applications that differ only in racially-associated names and reveal a critical disconnect: models show no output-level bias, yet retain and amplify demographic representations across model layers. Through activation steering and novel cross-layer interventions, we demonstrate that this suppressed information is decision-relevant: when reinjected at critical layers, it produces near-complete decision reversals. Critically, this latent bias is asymmetric - steering interventions affect decisions in one demographic direction, while producing minimal effects in reverse - and susceptible to adversarial prompt engineering and parameter-efficient fine-tuning. These findings demonstrate that behavioural audits focused on outputs are insufficient: fair outputs can mask exploitable internal biases. They also motivate dual-layer testing frameworks combining output evaluation with representational analysis for AI governance in high-stakes decisions.

Financial markets such as bond, derivatives, and repo markets form networks of interdependent obligations. Existing multilateral netting methods typically trade off the extent of netting against preservation of counterparty exposure: central clearing reallocates exposure to a central counterparty, while trade compression may alter bilateral counterparty relationships. TradeMech is a mechanism for markets in which one or two homogeneous fungible objects are traded. The mechanism transforms a network of initial bilateral contracts into chains and cycles, nets the designated object multilaterally on those chains and cycles, and replaces initial contracts with multiparty contracts whose assigned trades remain fractions of the original bilateral trades. The construction achieves maximal multilateral netting of the designated object while preserving each agent's contractual profit and preserving the location of counterparty risk. When a party fails to pre-commit a required object, the affected assigned trade is recovered as a bilateral contract between the same original counterparties and the remaining assigned trades are re-netted on residual chains, so no new counterparty exposure is created.

In educational settings, AI can be used as a learning aid, but can also be used to avoid schoolwork, thereby passing classes while learning little. Many existing studies on the impact of AI on education focus on AI use in controlled settings or with specialized tools. In this paper, the dropoff in ChatGPT activity during non-school summer months in 2023 and 2024 is used to identify areas with heavy educational AI use and thus estimate the educational impact of AI as it is actually used. I find no meaningful impact of AI usage on high school test score averages in either direction. These results imply that, to the extent that high school students use AI to avoid learning, it either does not matter much for their test performance or is cancelled out by positive uses of AI in the aggregate.

In the context of aggregating von Neumann-Morgenstern utilities, we show that bounded violations of the Pareto conditions characterize aggregation rules that are approximately utilitarian. When a single utility function is intended to represent the preference judgments of a group of individuals and the Pareto principles are nearly satisfied, we prove that its distance from a weighted sum of individual cardinal utilities does not exceed half of the positive parameter that differentiates our weaker versions of the Pareto conditions from their conventional forms. This result suggests the stability of Harsanyi's (1955) aggregation theorem, in that small deviations from the Pareto principles lead to aggregation rules that remain close to utilitarian aggregation.

It has long been established that, if a panel dataset suffers from attrition, auxiliary (refreshment) sampling restores full identification under additional assumptions that still allow for nontrivial attrition mechanisms. Such identification results rely on implausible assumptions about the attrition process or lead to theoretically and computationally challenging estimation procedures. We propose an alternative identifying assumption that, despite its nonparametric nature, suggests a simple estimation algorithm based on a transformation of the empirical cumulative distribution function of the data. This estimation procedure requires neither tuning parameters nor optimization in the first step, i.e., it has a closed form. We prove that our estimator is consistent and asymptotically normal and demonstrate its good performance in simulations. We provide an empirical illustration with income data from the Understanding America Study.

We study stochastic choice across decision problems, each represented as a menu of action labels paired with observable outcome vectors. We propose a consistency condition for behavior in decision problems composed of two separable components: choice probabilities must agree with those obtained when each component is considered in isolation. Together with monotonicity and continuity, this separability requirement characterizes the family of random coefficients logit rules.