Papers

We extend the clustered connectedness framework of Buchwalter, Diebold and Yilmaz (2026) in two complementary directions that improve the robustness and interpretability of cross-cluster connectedness. First, we develop a diagnostic for residual ordering sensitivity by characterizing the distribution of cluster-level net connectedness across all admissible identification orderings and, in particular, by pairing first- and last-position orderings while holding fixed the relative ordering of all other clusters. Second, we introduce a dedicated cluster of control variables to absorb variation associated with observed common macro-financial factors while preserving the computational scalability of the clustered framework. The control cluster is fixed first, and bank innovations are residualized with respect to it before the remaining bank clusters are permuted and orthogonalized as usual, leaving the number of admissible bank-cluster identification strategies unchanged. Under the maintained recursive assumption that control-cluster innovations are contemporaneously exogenous to bank-cluster innovations, the remaining cross-cluster connectedness among the bank clusters can be interpreted as bank-to-bank transmission net of those observed common-factor shocks. We apply the methodology to seventy-one global banks grouped into seven regional clusters over 2003–2024. The treatment of common macro-financial factors materially affects both system-wide cross-group connectedness and cluster-level net positions. Placing the controls in a dedicated first cluster also substantially reduces paired first-versus-last ordering sensitivity across all seven bank clusters, with especially large reductions for the United States and the European clusters.

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.

This paper studies the effects of p-hacking on the bias of published estimates when papers with statistically significant results are selectively published. We show that fast p-hacking—actions that lead to large changes in p-values—always exacerbates the bias from selective publication. On the other hand, slow p-hacking—actions that lead to small changes in p-values—exacerbates bias when selection is weak, but mitigates it when selection is strong. In a model featuring both types of p-hacking, we show that a normality assumption identifies the true distribution of effects as well as the counterfactual mean that would obtain under selective publication without p-hacking. Applying the model to meta-analyses on the effects of behavioral nudges and development aid, we find suggestive evidence that both mitigation and exacerbation can arise in practice.

Building on the pioneering paper of Kearns, Roth, and Ryu (SODA'26), we study information aggregation in a networked learning model. The model captures a central pattern in agentic AI: each agent sees only part of the data and passes on only its own conclusion. Their model considers a linear regression problem with the mean squared error (MSE) loss. Agents sit in a DAG and each sees only a subset of the features and its parents' predictions, fits a linear predictor, and passes only its prediction forward. The benchmark is the full-feature learner that sees all raw features. A path of depth $D$ is $M$-covered if every block of $M$ consecutive agents collectively sees all raw features. Kearns, Roth, and Ryu proved that the excess mean squared error of the last agent on such a path is $O(M/\sqrt D)$, and gave a cyclic instance with excess error $\Omega(M/D)$ for $D<M^2$. We close this gap: the correct rate is constant up to depth $M^2$, and $\Theta(M^2/D)$ beyond it. We first give a sharper analysis of the cyclic instance and improve its lower bound to $\Omega(\sqrt{M/D})$ for $D<M^2$. We then construct, for every depth $D\ge M^2$, an $M$-covered path of depth $D$ with excess error $\Omega(M^2/D)$. The same instance gives the constant lower bound for all $D < M^2$. We also show that for any fixed distribution the excess error contracts geometrically along the path, ruling out any single instance that witnesses any polynomial lower bound at every depth. Finally, we prove the same optimal rate for logistic classification in the logit-passing model of Bateni et al., which considers the binary cross-entropy (BCE) loss. The same improved upper bound of $O(M^2/D)$ holds, and we transfer all the regression lower bounds by showing that on those examples the logistic path follows the least-squares path up to rescaling.

Creativity researchers often distinguish between two stages of the creative process: generation versus selection. While much is known about the psychology of idea generation (e.g., the factors that lead to a greater number of novel and useful ideas), less is understood about the nature of selection, or how generation and selection interact. Here we investigate how the act of generating ideas may potentially distort the selection process. Using an incentive-compatible paradigm in which pairs of participants reviewed the same ideas and were rewarded for submitting only high-quality ideas, we find that people submit a greater number of lower-quality ideas when selecting among their own ideas than when selecting among another person's ideas (the Creative Endowment Effect). This effect generalizes across three tasks in two domains and is resistant to an informational intervention (i.e., explicitly telling people about the effect). However, having participants revisit their ideas several months later increases their selectivity. The broader implications for individuals and organizations are discussed.

Measuring the intergenerational transmission of lifetime economic status is complicated by researchers often only observing snapshots of income at specific ages. Consequently, standard practice estimates intergenerational mobility using income averages, introducing life-cycle bias that compromises reliability and comparability across studies, time, and place. I develop a missing data framework that exploits available income data and observable characteristics to eliminate life-cycle bias. This method combines nonparametric identification with Neyman-orthogonal moments to construct debiased machine learning estimators for intergenerational income mobility measures under plausible missing-at-random and testable independence assumptions. I apply this framework to estimate the intergenerational elasticity for the U.S. using the Panel Study of Income Dynamics across birth cohorts from 1954 to 1977 with rolling 10-year windows. While existing approaches estimate values between 0.41 and 0.54, the proposed method yields substantially higher estimates ranging from 0.6 to 0.7, averaging 0.64. These results align closely with recent evidence using long time averages over mid-career periods, reinforcing high U.S. intergenerational persistence.

We represent a non-Bayesian agent as one who does not completely trust the information they receive. The behavioral expression of complete trust lies in a homogeneity property of Bayesian updating: posterior beliefs do not change if a signal is made arbitrarily rare by scaling down its likelihood vector. We show that simply dropping this property and retaining all other Bayesian behavioral properties yields a unique representation where the agent is still Bayesian but has subjective uncertainty over the information structure generating the signal. The representation result is proved using the Fundamental Theorem of Projective Geometry. We analyze how various updating biases may be rationalized by a lack of trust.

This paper analyzes the implementation of blockchain-based integrity mechanisms in Greek Fiscal Electronic Mechanisms (FEMs) and the central tax information system eSEND. The study examines the cryptographic architecture of fiscal devices, including Electronic Cash Registers, Fiscal Printers, Fiscal Signing Machines, and FEMAS devices, which implement double or triple hash-chain structures to ensure transaction immutability. The transmission protocol between fiscal devices and the central database is also evaluated with respect to encryption, sequential validation, and blockchain verification. In contrast, the architecture of Electronic Invoicing Provider Services and the myDATA central platform is analyzed, highlighting the absence of blockchain-based integrity guarantees. The comparison demonstrates that hardware-based fiscal mechanisms provide stronger guarantees for transaction completeness and tamper resistance than purely software-based invoicing infrastructures. The findings highlight architectural weaknesses in the current e-invoicing framework and propose improvements for ensuring transaction integrity in digital tax ecosystems.

When natural disasters strike, individuals, communities, and even entire countries can suffer. Researchers have studied the impacts of disasters on various factors of interest, from mental health, to poverty, to economic activity. However, the impact of disasters on the nonprofit sector is understudied despite the nonprofit sector's perhaps surprising role in local or national economies as well as its role in disaster response and recovery. Thus, we study the effect of natural disaster damage on different county-level nonprofit outcomes using a panel dataset spanning 1991 to 2021 and causal inference methods tailored to panel data. Contrary to prior work, which found small but positive associations between disaster damage and nonprofit revenue or assets, we find no evidence of a causal effect.

We study tournaments with managerial discretion in hiring. A manager selects a coworker from a pool of candidates and then competes against him in a Lazear–Rosen–style tournament with a prize equal to a share of total output. A profit-maximizing principal sets the prize share together with a head start (or handicap)—an advantage (or disadvantage) in the output comparison—granted to the manager. The head start affects output through three channels: (i) encouraging the manager, (ii) discouraging the new hire, and (iii) inducing the manager to hire a stronger candidate. The hiring effect dominates the discouragement effect until the strongest candidate is hired; beyond that point, any further head start discourages the new hire more than it encourages the manager. The optimal contract therefore grants a head start just large enough to induce the manager to hire the strongest candidate.

I provide the first analytic construction of a canonical Aiyagari economy exhibiting at least three steady states. Along the way, I provide new upper and lower bounds on the stationary capital supply for the case where the net return on saving is negative. I also give parameter restrictions that guarantee the existence of a steady state with a rental rate in the neighborhood of a remarkably simple number: the capital share times the depreciation rate.

We study robust mechanisms when the designer possesses a Bayesian belief over some components of agents' private information but faces ambiguity over others. The designer evaluates mechanisms by their worst-case performance over all joint distributions consistent with her belief over the Bayesian components. The framework encompasses settings such as multidimensional delegation in which a principal knows the distribution of the state but not the agent's preferences (e.g., his tradeoffs across dimensions), screening in which a seller only has misspecified estimates of buyer preferences, and auction and voting design when agents' beliefs about each other are ambiguous to the designer. We provide conditions under which a knowledge-based mechanism—one that conditions only on the Bayesian components but not the ambiguous ones—is robustly optimal. Our results unify earlier work across distinct economic environments and uncover new applications.

We give an exact randomization-based confidence set for the average treatment effect (ATE) in matched-pair studies with a binary outcome, requiring neither monotonicity nor any distributional assumption beyond the within-pair coin flip. At its core is an analytic solution to the worst-case allocation of attributable effects: two binomial-symmetry lemmas identify the pattern hardest to reject as a single boundary corner, so testing null hypotheses needs no integer program and no numerical search. Inverting the test via binary search yields a prediction set for the attributable effect in O(log S) Binomial tail calculations; the Bonferroni proposition of Rigdon and Hudgens (2015) produces the ATE confidence set at the same computational cost. The same corner extends without further machinery to a sensitivity analysis for matched observational studies under Rosenbaum's $\Gamma$-model. A simple formula for the design sensitivity illuminates when an observational study can hope to provide evidence for an effect.

People increasingly use language models to support life decisions. Many such decisions involve a probabilistic forecast: How likely is a major life event, a natural disaster, or an economic outcome? Users of language models may implicitly trust that these forecasts fall out of a coherent world model. In this paper, we evaluate the coherence of language model probabilistic forecasts through a procedure that builds on a theorem due to de Finetti. We elicit forecasts from language models across events generated from stock returns data. We then use linear programs to compute the largest Dutch-book profit - the profit an arbitrageur could guarantee by betting against model-generated probabilities - which we use as a measure of incoherence. Our procedure does not require outcome labels, so we can evaluate coherence even in settings where outcomes are not observed or have not yet resolved. We find substantial evidence of incoherence in language model forecasts. Such incoherence increases when there are richer logical relationships between events, and irrelevant contextual details can increase incoherence by an order of magnitude. We conclude by discussing how alternative training strategies may improve probabilistic coherence.

Extending the Price-and-Choose (P&C) mechanism of Echenique and Nunez (2025), we propose the Price-Accept-and-Choose (PA&C) mechanism, which preserves efficiency while eliminating P&C's first-mover advantage. We then analyze randomized and bidding variants and show that the resulting equilibrium payoffs correspond to standard solutions in transferable utility games: the Center of the Imputation Set value for P&C and the Shapley value for PA&C. In the randomized variants, these solutions arise in expectation; in the bidding variants, they are implemented on every equilibrium path. We further relate other efficient designs, such as balanced VCG payments, to additional solution concepts, forging an interpretable bridge between implementation theory and cooperative game theory.

In the presence of interference, where the treatment assigned to one unit can affect the outcomes of others, many causal estimands depend on the treatment-assignment policy under which the experiment is conducted. This policy dependence creates a fundamental challenge for off-policy estimation, where the goal is to estimate causal quantities under a hypothetical intervention policy different from the one used to collect data. We study this problem of off-policy estimation of causal effects for heterogeneous Bernoulli policies. By representing exposure-weighted potential outcomes in the biased Fourier basis of the experimental design, we construct, for any prespecified Fourier subspace encoding the assumed interference structure, the unique minimum-$L^2$ weight that transports every function in that subspace. Global and local inverse-probability weights, linear-interference weights, and no-interference weights are special cases. The weight variance is a structured chi-square distance between the experiment and target policies. When the assumed interference structure is misspecified, the introduced bias couples the omitted outcome spectrum with the corresponding policy-shift coefficients, yielding a sharp robustness bound and a bias-variance trade-off. A Fourier-neighborhood-overlap condition gives consistency under structured interference, and we state a Doob-martingale central limit theorem for off-policy estimators. As the variance is not identified, we derive identifiable bounds and associated conservative estimators of the variance. Simulations illustrate these theoretical results for the design and analysis of experiments under network interference and design mismatch.

We establish the consistency and asymptotic normality of a two-step estimator of conditional expectiles in the context of conditional scale models. We first estimate the conditional variance parameters by quasi-maximum likelihood and then compute the unconditional expectile of the innovations using the empirical distribution of the standardized residuals. We show how replacing true innovations with standardized residuals affects the asymptotic variances of both conditional and unconditional expectile estimators. Finally, our empirical analysis reveals that conditional expectiles assess tail risk in cryptomarkets in a more robust manner than traditional quantile-based risk measures, such as value at risk and expected shortfall.

Large language models (LLMs) are increasingly deployed as economic agents, yet there is little evidence whether LLM agents are suited for participating in market mechanisms designed for humans, and whether these mechanisms deliver desired outcomes when faced with LLM agents. We address this question by replicating seminal economic experiments, replacing human subjects with LLM agents. We place agents in a double auction environment, which is a widely-used market mechanism. We check whether such a market is able to deliver an efficient allocation of resources, thereby testing a novel dimension of alignment of LLM agents – their compatibility with a fundamental market mechanism. We find that markets populated by LLM agents exhibit slower or no convergence towards market equilibrium, thus providing less efficient allocations than markets populated by humans. We then analyze agents' individual trading decisions and find substantial heterogeneity both across model families and market roles. We also run a lexical analysis of Chain-of-Thought (CoT) traces generated by the agents. We find that the decision to execute a trade rather than continue incrementally adjusting prices is associated with a shift from strategic considerations toward urgency. We publicly release our testing framework, which can be used for future evaluations.

We study a network formation game in which agents sponsor links at a linear cost in order to maximize centrality, defined as a weighted sum of walk counts with positive and weakly decreasing weights. This class includes Katz Bonacich centrality and total communicability and captures environments in which access decays with distance. Our main result is a sharp equilibrium characterization: every Nash equilibrium network is core periphery, meaning there exists a set $C$ such that every node is linked to every node in $C$, and there are no other edges. The driving force is that linking to better connected agents generates many additional short connections at once, so incentives concentrate links on a dense subset. Finally, we show that the welfare maximizing network is always either empty or complete.

Two contestants with possibly different marginal costs compete across identical battlefields governed by a Tullock technology with discriminatory power at most one. A symmetric schedule divides a fixed prize according to the number of victories. Allowing for inactivity, unequal efforts across battlefields, and arbitrary mixed strategies, we prove the existence and uniformity of equilibrium. Equilibrium may be pure, semi-pure (one contestant mixes), or two-sided mixed; in a two-sided mixed equilibrium, each contestant uses at most countably many positive effort levels. Multiple equilibria with different structures can coexist while generating the same expected effort, prize share, cost, and payoff. For every admissible schedule and cost ratio, equilibrium is unique with six or fewer battlefields, whereas seven first permits multiplicity or two-sided mixing. We also characterize all equilibria under majority rule.

We study poverty minimization via direct transfers, framing this as a statistical learning problem while retaining the information constraints faced by real-world programs. Using nationally representative household consumption surveys from 34 countries that together account for 76% of the world's poor, we estimate that reducing the poverty rate to 1% (from a baseline of 13%) would cost $211 B nominal per year. This is 4.0 times the corresponding reduction in the aggregate poverty gap, but only 19% of the cost of universal basic income. Extrapolated globally, the results imply a cost of 0.28% of global GDP to (approximately) end extreme poverty.

This paper provides a clear philosophy on codes of human conduct within economic and social institutions like markets and government. It categorizes these codes into ethics and morals: ethics involve nonmaleficence (avoiding harm), while morals involve beneficence (doing good). Ethics, such as honesty, respecting property, and keeping promises, are universally accepted minimal norms essential for the functioning of social institutions. In contrast, morals, including altruism, self-sacrifice, and generosity, are subjective and context-dependent, linked to individual virtues and influenced by character, culture, and religion. The paper argues that ethical behavior is a necessary condition for an act to be considered moral. Accordingly, human acts can be classified into four categories: unethical-immoral, solely ethical, perceived moral, and truly moral, with examples provided for each type. It posits that the free market system, unlike socialism or communism, requires only ethical agents to succeed, as they ensure the market's mission of improving human well-being. These agents do not necessarily need to be moral for the market to function effectively. Furthermore, the paper discusses that voluntary morality, in the sense of helping others out of compassion, can greatly benefit any human-based institution, including markets, by enhancing social capital alongside ethics and the justice system. However, the forced incorporation of personal moral principles into the economic system can be detrimental to the virtuousness of intended moral acts, economic efficiency, individual liberty, responsibility, and social cohesion.

For an exogenous source of information and a prespecified number of messages k, I characterize the set of maximal–that is to say, undominated–summaries of the information that use at most k messages. When the exogenous information produces a regular distribution over posteriors (if it has a density on its affine hull, for instance), the maximal summaries take a precise geometric form. They are hierarchical power diagrams: convex subdivisions of the space of beliefs constructed by recursively subdividing cells according to weighted proximity criteria.

I show that the equivalence in finite social-choice environments of Bayesian incentive compatibility and dominant-strategy incentive compatibility holds universally if and only if every type-dependent difference in the valuation of prizes operates through a single scalar type index. Beyond binary domains this universal robustifiability is nongeneric.

To study a scalar parameter, a researcher may consider multiple research designs. Based on the evidence across designs, the researcher may wish to formulate a headline estimate of the parameter. I examine how to choose this headline when it is unclear which design is most appropriate for studying the parameter. I model this setting by assuming that (i) exactly one of the designs is valid for the parameter and (ii) the researcher has ambiguity about which design is valid, represented by a class of priors over the candidate designs. To account for ambiguity, I propose reporting the headline estimate that minimizes the worst-case posterior risk over the class of priors. In three applications, I show cases where accounting for ambiguity materially affects the quantitative conclusion and cases where an existing headline is already close to optimal.

Inference in linear regression commonly treats OLS residuals as proxies for unobserved errors. This approximation can fail when the regression projection is nonlocal relative to the error-dependence structure. Residualization then shifts covariance information across observations and clusters, while conventional heteroskedasticity-consistent (HC) and cluster-robust variance estimators (CRVE) retain only diagonal or within-cluster residual moments and may therefore understate sampling uncertainty. This paper develops a projection-geometry framework for robust variance estimation. The variance of the OLS estimator is represented exactly as a Riesz functional of latent covariance blocks, and observable residual moments are linked to the target through a linear operator determined by the full regression projection. This formulation reduces variance estimation to a linear inverse problem. I propose a Riesz variance estimator that combines within- and cross-cluster residual moments. Conventional HC and CRVE emerge as restricted approximations whose validity depends on negligible projection spillovers. The estimator remains well defined when cluster-specific leverage matrices are singular and is computed by an iterative algorithm that avoids explicit matrix inversion. Simulations show substantial undercoverage by conventional methods under projection spillovers, whereas the proposed estimator restores near-nominal coverage. In an application to colonial governor promotions, the correction changes the significance of four of five reported coefficients.

We develop a framework for mechanism design with AI agents whose alignment (preferences) and capabilities (feasible actions and information) are unknown. We want such agents to act on our behalf so mechanisms must incentivize both honesty and obedience. A one-sided imitation structure—capabilities can be concealed but not counterfeited—yields a revelation principle, a characterization of implementable policies via nested cyclical monotonicity, and conditions under which eliciting higher-order beliefs can discipline multiple agents. We apply our framework to stylized examples of (i) sandbagging in which a more capable agent pretends to be less capable; (ii) an alignment–interpretability trade-off, where the two are substitutes in the instrument but complements in value; (iii) discipline via peer scoring; (iv) coupling rewards to induce competition among multiple agents; and (v) scalable oversight and reward shaping.

We characterise the division rules for claims problems that satisfy equal treatment of equals, bilateral consistency, composition down, and composition up. The rules are precisely the members of a one-parameter log-exponential family $\{r^\theta\}_{\theta\in[-\infty,+\infty]}$, with constrained equal awards (CEA) and constrained equal losses (CEL) as its endpoints. For finite $\theta$, $r^\theta$ is the equal-sacrifice rule in awards for $u_\theta=\log\varphi_\theta$, where \[ \varphi_\theta(x):=\frac{e^{\theta x}-1}{\theta}\quad(\theta\ne0), \qquad \varphi_0(x):=x, \] and simultaneously the equal-sacrifice rule in losses for the dual utility $u_{-\theta}$. The proportional rule is the midpoint, $\theta=0$. Continuity of the rules are not assumed but a consequence of the other axioms.

The paper explores a theoretical freemium model for the sale of information, drawing on mathematical tools used in the study of repeated zero-sum games and Bayesian persuasion. Unlike standard Bayesian persuasion models, the information seller (IS) is indifferent to the actions taken by the information buyer (IB) and is concerned solely with maximizing the revenue from selling information. Offering some information for free may increase the IB's willingness to pay for additional information. The information that the IB seeks is about the state of the world. Initially, the IB only knows the prior distribution over possible states. The IS supplies both free and paid information through signals whose state-dependent distributions determine the IB's posterior via Bayes' rule. The IB's utility is a function of the posterior. An optimal free signal is one that maximizes the IS's expected revenue from the subsequent paid signal. That revenue is equal to the IB's expected utility gain when moving from the posterior induced by the free signal to that induced by the paid signal. The paper characterizes the optimal free and paid signals and derives a formula for the maximal revenue in terms of the IB's utility function. It shows that a revenue gain for the IS from the provision of free information is accompanied by a loss to the IB. Whether free information can increase the IS's revenue depends on the form of the IB's utility function. In the two-state case, that dependence is fully characterized. In the general case, only necessary conditions are obtained. In particular, if the IB's utility function is convex, the IS can never profit from providing free information. This occurs, in particular, when the IB uses the information to solve a decision problem. By contrast, when the IB is engaged in a strategic interaction with a third party, the IS may benefit from providing free information.

Policies with a common objective and implementation date may differ in details or context. We distinguish the aggregate average treatment effect on the treated (ATT) from sub-aggregate ATTs defined by implementation cohort, jurisdiction, period, or policy type. UN-DID and DID-INT, two DiD estimators that construct jurisdiction-by-time effects, estimate these ATTs under parallel-trends conditions matched to the aggregation. In CPS placebo-law simulations, randomization inference is generally well-sized, though some jurisdiction-specific tests are conservative. The jackknife can be undefined for sub-aggregate ATTs; when defined, it over-rejects with few treated or comparison jurisdictions. Estimands and inference methods should match the policy question and implementation setting.