Papers

Intra-household joint home-based tours - trips in which household members depart together, engage in shared activities, and return together - represent a significant share of daily travel, yet are systematically ignored in transport simulations. Conflating joint and solo tours within a single mode choice framework introduces bias in preference parameter estimates. This paper proposes a three-step methodology to integrate joint tours in agent-based transport models: a Random Forest classifier to identify joint tours, a Multinomial Logit model estimating mode choice specific to joint tours, and a Penalized Logistic Regression for driver/passenger assignment. Applied to the Paris region using household travel survey data, the methodology successfully replicates observed joint tour shares and mode distributions in a synthetic population. The proposed framework enables more reliable evaluation of policies whose impacts differ between joint and solo travel, such as HOV lanes or family transit fare discounts.

Most LLM benchmarks rank models on their ability to automate work tasks. In practice, however, models are often used to assist other (human or LLM) agents. The question that drives model selection is therefore not only which model produces the best output, but which model most improves the work of another (weaker) agent. We introduce a unified framework that evaluates the capability of models to automate and augment another agent's performance. Across seven economically grounded real-world tasks, an assistant model writes assistance text for a standardized lower-capacity worker model, which produces the deliverable. In automation mode, the assistant produces the output directly. Outputs are scored through blind pairwise comparisons by an LLM judge panel with task-specific rubrics, replicated across ten runs. Rankings across the two regimes are only modestly correlated, and the automation winner loses augmentation on five of seven tasks. Assistance is not reliably positive. The unaided worker outranks every assisted condition on three tasks, and only one model's guidance beats no guidance on average. These results suggest that automation ability is an incomplete proxy for assistance quality, motivating benchmarks that evaluate models according to the roles they play in human-AI and multi-agent systems.

This paper develops a revolving-door model in which a regulator designs policy signals that affect a regulated firm's capital structure while holding an implicit career option to leave public service for the regulated firm or its advisers. The model has three linked mechanisms. First, regulatory signals alter the firm's pricing, leverage, and risk choices by changing the tradeoff between consumer surplus and firm profit. Second, because the regulator's outside payoff is convex in regulated-firm value, the career option rises with firm value, leverage, and signal-induced volatility. Third, when signal-induced volatility raises the option vega of firm equity, the same mechanism can increase tail risk and reduce distance to default. A proof-of-concept calibration illustrates leverage-vega warning regions and shows how a regulator-induced signal can move a firm from a go-leverage region into an early-warning or crash-risk region. The calibration includes maturity comparisons, sensitivity grids over compensation convexity and initial leverage, a bank-like parameter anchor, stochastic signal simulations, and a stopping-horizon diagnostic. Evidence from federal payroll data on post-employment restrictions supports the career-option primitive by showing that limits on revolving-door opportunities affect post-agency pay. The exercises also organize historical regulatory failures: Lehman Brothers as a regulatory-vacuum case, Silicon Valley Bank as a distorted-signal and weak-supervision case, and Signature Bank as a converted-signal case involving post-public-service incentives, lighter oversight, rapid growth, uninsured-deposit exposure, and tail risk. The model does not predict collapse dates; it identifies conditions under which regulatory forbearance, weak debt penalties, and career-option incentives can delay oversight until firms cross a leverage-vega warning threshold.

Optimal execution (OE) in multi-asset settings involves complex interactions across assets, particularly through shared capital constraints during portfolio rebalancing. While existing models capture cross-impact and portfolio-level dynamics, they largely overlook the role of explicit cash constraints along the execution trajectory. As a result, the feasibility of execution strategies under limited capital remains poorly understood. In this paper, we extend the classical Almgren-Chriss framework to incorporate intertemporal constraints on expected cash consumption, requiring that the expected cumulative cash spent does not exceed a prescribed budget at every trading period. We show that the resulting multi-asset OE problem can be equivalently formulated as a quadratically constrained quadratic program (QCQP), and further establish that it admits a convex representation under mild conditions. This provides a tractable framework for analyzing execution strategies under dynamic capital constraints. Through controlled synthetic experiments, we show that the proposed cash constraints qualitatively alter OE schedules toward cash-feasible sell-first executions as the constraints become tighter. Furthermore, evaluations in an out-of-sample agent-based market simulator demonstrate that our method substantially reduces peak cash drawdown while maintaining implementation shortfall comparable to existing execution strategies. Our results highlight the importance of explicitly modeling financial feasibility in multi-asset execution and provide a foundation for bridging theoretical OE models with practical capital constraints.

Asset-pricing models typically condition on a fixed information set. This paper endogenises the market's conditioning architecture by allowing portfolios to choose representations whose induced exposures affect prices. Capital allocated across representations determines aggregate positions and the clearing premium, while price feedback changes representation value and, through causal certification, admissible representations. A representation equilibrium is the fixed point of this configuration–price–certification loop. The framework separates position crowding through market impact from representation crowding generated by driver-space overlap and priced through a basis-invariant information-capacity cost. Under compactness, resolvent regularity and continuous certification, equilibrium exists at every switching cost; within a stable certification cell, informational congestion yields a concave population game and local uniqueness under a small-gain condition. At a settled configuration, a spectral statistic combining cross-impact, covariance and deployed capacity determines three position-path regimes: below one half the fundamental solution is unique; at the boundary only innovation-free deviations remain; above it, destabilising directions support a continuum of self-confirming conventions. Monotone impact lies inside the uniqueness region, while indefinite impact alone is not sufficient for multiplicity. Intermediary hedging can generate an indefinite response while round trips remain costly. Endogenous risk capacity bounds convention amplitudes, and least-squares learning makes the dominant convention neutral rather than attracting. The threshold is not estimated in data; instead, the empirical exercise documents an out-of-sample, driver-specific signature consistent with representation crowding and states the conditions required for direct measurement.

We introduce a reinforcement learning framework for market making in a limit order book. Our algorithm aims to maximize trading revenue by dynamically submitting market and limit orders of varying sizes across multiple price levels while controlling inventory size. We use multivariate logistic-normal distributions to model order allocations and employ a deep-set encoder to aggregate features from variable-length order sets into a fixed-dimensional latent representation. Additionally, we incorporate potential-based reward shaping to accelerate learning without altering the optimal policy. We illustrate the performance of the method in three simulated market environments consisting of noise traders who submit random trades, tactical traders who respond to instantaneous volume imbalance, and strategic traders who trade in the direction of an exponentially weighted volume imbalance signal.

How level-0 players behave and how they are perceived by higher-level players are central questions in the literature on level-k models of boundedly rational strategic reasoning. To study these twin questions, we apply the intra-team communication method developed by Burchardi and Penczynski (2014) to identify level-0 actions and beliefs in the canonical beauty contest game and in a variant of the 11-20 game of Goeree et al. (2018), in which behavior appears inconsistent with the standard level-k model. In the beauty contest game, we replicate Burchardi and Penczynski (2014)'s finding that elicited level-0 beliefs align with observed level-0 actions. In the variant of the 11-20 game, however, elicited level-0 beliefs and observed level-0 actions diverge, and both depart from the standard level-0 assumption, suggesting a complementary explanation for the behavioral pattern documented by Goeree et al. (2018).

We develop parametric Entropic Value-at-Risk (EVaR) portfolio optimization for tempered stable Lévy returns. We derive portfolio cumulant-generating functions and weight-dependent admissible moment-generating-function domains under two multivariate constructions: a multivariate normal tempered stable approach and an independent component factorization. These expressions allow portfolio EVaR to be evaluated from fitted asset- or component-level parameters without repeated portfolio-level distribution fitting. We construct minimum-EVaR portfolios and two entropic reward–risk portfolios. We test the portfolios in a rolling 2000 to 2026 out-of-sample U.S. sector ETF allocation. In this universe, several entropic portfolios have higher realized Sharpe ratios than their matched CVaR portfolios or standard allocation benchmarks.

We develop simulation-based policy iteration for continuous-time portfolio choice with predictable returns and convex constraints. Each outer step re-evaluates a fixed-latent open-loop backpropagation-through-time (OL-BPTT) adjoint after deployment and solves the constrained update. Shifted-adjoint cancellation controls the adjoint–HJB Hamiltonian-gradient discrepancy by the policy-improvement residual. For CRRA portfolios, exact HJB policy iteration identifies the optimal reduced value factor, while population OL-BPTT iteration converges globally when the adjoint update is directionally improving and approximate stationarity is asymptotically HJB-compatible. A theorem-matched occupation audit yields maximal $95\%$ upper endpoints of $0.066$ for the primitive directional ratio and $0.074$ for a stronger norm-relative ratio, both against the half-step threshold $0.75$. In the high-precision $50$–$50$ occupancy/broad-anchor design of a three-factor, fifty-asset benchmark, current-policy re-evaluation outperforms matched pooled refinement under the on-policy and broad evaluation laws.

Credit decisioning is a high-stakes task in which model outputs must be accurate and explainable to support compliant decisions. Although modern credit risk models such as eXtreme Gradient Boosting (XGBoost) and Graph Neural Networks (GNNs) improve predictive performance, their explanations are often too technical for stakeholders creating communication gaps that can shape approvals, denials, and fairness judgments. We examine whether Large Language Models (LLMs) can serve as explanation layers that translate post-hoc explanation artefacts into stakeholder-appropriate risk narratives. Using Freddie Mac single-family loan-level data, we develop three pipelines: standard tabular (XGBoost + SHAP), and two with alternative data, a pure network-based (GNN + GNNExplainer), and a bimodal one (combining tabular and network data). We generate narratives with three LLM configurations: a small fine-tuned LLM (Gemma 3 4B), a large fine-tuned LLM (DeepSeek R1 70B), and a zero-shot commercial LLM (Gemini 2.5). Explanation quality is evaluated through automated checks across all pipelines and a human study of bimodal explanations comparing credit risk professionals and non-professionals on eight decision-relevant dimensions. We have three main findings. First, the pipeline accounts for higher variance in evidence-grounding scores than the language model, meaning that the binding constraint on explanation quality is the evidence representation, not the model used. Second, the explanation narratives reliably name the influential factors but are less reliable when stating the direction of influence, which may be consequential for adverse-action communication. Finally, professionals apply stricter evidentiary standards than non-professionals. We discuss implications for the governance of risk models, including deployment considerations and the value of domain-aligned LLMs in regulated credit settings.

This paper considers European multi-asset option pricing under Lévy and affine characteristic-function models. The main obstruction is the curse of dimensionality: direct multidimensional COS pricing forms tensor-product coefficient arrays whose size grows exponentially with the number of assets. We study and extend COS-TT-CHF, a low-rank construction that uses TT-cross to compress sampled characteristic-function tensors into tensor-train COS coefficients for arithmetic basket and min/max option pricing. Once built, the compressed representation gives fast post-setup strike-grid and selected component Delta/Vega calculations. The numerical study compares with adaptive-quadrature Fourier benchmarks, direct COS, a tensor-Fourier min-option benchmark, and quasi-Monte Carlo (QMC) references based on randomized Sobol points. The reported timings show a low-dimensional crossover against direct COS as the benchmark moves from $d=2$ to $d=4$, favorable timings against the tensor-Fourier min-option benchmark from $d=3$ onward, and favorable timings against the QMC common-Heston reference already at $d=2$. The reported tests reach $d=30$ for GBM and $d=20$ for VG, NIG, and common-Heston benchmark families, with accuracy, rank, runtime, control-sensitivity, and component Delta/Vega diagnostics reported throughout.

Delegating cognitive operations to generative artificial intelligence redistributes execution and raises a governance problem: where human direction of the task remains. We distinguish two routes. Specified delegation places that direction before execution, through instructions, constraints, or criteria that delimit the task. Iterative coproduction places it during production, through interventions that correct or redirect provisional outputs. To examine both routes, we use aggregate monthly cells from the Anthropic Economic Index for April and May 2026. The AEI distinguishes two modes of use: 1P API, which corresponds to direct traffic through Anthropic's API, and Claude.ai, which combines activity from Chat and Cowork. On this basis, we test whether a stronger work-related orientation of human-AI interaction is associated with more specified delegation within each mode and whether the increase in the iterative profile is greater in Claude.ai than in 1P API. The main analysis uses level-0 O*NET tasks and estimates how both profiles change when an eligible record reallocates ten percentage points from personal use to work-related use. The iterative comparison is restricted to 1,411 node-month pairs observed and eligible in both modes. Specified delegation increases by 2.76 points in 1P API (95% CI: [2.30, 3.22]) and by 1.45 in Claude.ai (95% CI: [0.93, 1.97]). On the common support, iterative coproduction changes by-0.30 points in 1P API and by 0.15 in Claude.ai, yielding a between-mode difference of 0.45 points (95% CI: [0.15, 0.75]). These findings show that work-related orien tation is associated with stronger traces of prior human direction and that the observable iterative response varies across modes of use. The article shifts attention from how much the AI executes to when human direction leaves observable traces.

We present in this article a non-parametric value-at-risk (VaR+CVaR) algorithm that remains accurate for an arbitrarily large number of underlying positions. The algorithm solves the two inherent problems of VaR estimation. First, past history is not directly applicable to the future, but all predictions of the future are based on the past. Second, VaR estimation is equivalent to modeling a single corner of a high-dimensional space (the corner where all bets lose simultaneously). The algorithm only uses mathematical methods that strictly do not degrade in accuracy at high-dimensions. Historical data are then directly incorporated with all high-dimensional relationships present, without manipulation. We test the algorithm with an ensemble of 500 portfolios with random positions across 49 distinct liquid futures of different expiries (VIX, equity indexes, gov. bonds, rates, energy, metals, livestock, agriculture, and softs). All VaR estimations are performed strictly blind to the future. The median portfolio rate of loss exceeding the 99% confidence daily VaR estimate is between $1.0\pm0.1$% depending on algorithm input parameters. 68% of portfolios have a rate of loss exceeding 99% VaR between $1.0\pm0.3$%, and 95% of portfolios between $1.0\pm0.5$%.

Trust in local institutions matters for trade, public goods provision, conflict resolution, and democratic consolidation. Using individual data from rounds 6 and 7 of the Afrobarometer surveys, I document that respondents in former British colonies are substantially more likely to trust traditional leaders than respondents in former French colonies. Cross-country comparisons of this kind are confounded by unobserved heterogeneity in pre- and post-colonial histories. To make progress on identification, I focus on Cameroon, which contains regions colonized by Britain and regions colonized by France within a single modern state, and I exploit the former colonial partition line as a geographic discontinuity, comparing respondents who live close to either side of the anglophone-francophone boundary. The theoretical argument is that the two colonial powers differed systematically in how they treated customary authority: British indirect rule preserved the governance functions of chiefs, while French direct rule subordinated chiefs to the administration and assigned them its most coercive tasks, taxation and forced labor above all, and these divergent experiences shaped the legitimacy of chieftaincy in ways that persist through institutional continuity and socialization. Consistent with this argument, respondents on the formerly British side are 22 to 26 percentage points more likely to trust traditional leaders than their neighbors on the formerly French side, and they are far more likely to have contacted a traditional leader in the previous year.

This paper develops a theoretical framework for signed optimal transport. A global flatness measure induced by the continuum transport equation serves as the regularizer. We examine transport networks from a harmonic analysis perspective, prove the existence and uniqueness of the optimal coupling in the variational problem, and provide an algorithmic scheme that guarantees lossless information transport under bi-marginal constraints. To bridge theory with practice, we design an empirical analysis framework for the resulting optimal estimators, which is sufficiently general to accommodate both time-series and panel-structural analyses of the network, owing to the intrinsic invariance properties of locally compact abelian groups.

Reputation and institutional certification are the two main trust mechanisms under information asymmetry, yet their interaction remains poorly understood. Here we analyze nearly one million eBay listings of Pokemon trading cards. Each listing pairs the seller's reputation with their signal choice, that is, whether they used costly third-party certification, their own costless claim (self-grading), or neither to convey the card's quality. First, we find that certification and self-grading partition the market. Self-grading prevails where reputation is high and value low, whereas certification prevails in the market's opposite corner. Furthermore, seller reputation raises the price premium of self-grading but not of certification. We then show that both empirical patterns follow from a signaling model in which false self-grading damages reputation, whereas certification requires an up-front fee. These results suggest that, even within a single market, trust rests not on a choice between reputation and institutions but on a division of labor.

Financial returns are heavy-tailed, and accurate tail risk estimation is central to portfolio risk management. Modern neural generators sample by pushing a simple base distribution through a learned map, and for training stability that map is built from Lipschitz components. This is the binding constraint: a Lipschitz map of a Gaussian is sub-Gaussian, so heavier-tailed targets admit no exact match at any finite Lipschitz constant. The Monge–Ampère equation ties the Brenier map's local distortion to the density ratio $f/(g\circ T)$, so a deeper trough in the target density requires a higher-gain map and yields a higher-variance estimator. The argument needs only bounded distortion, so it covers normalizing flows, flow matching, GANs, and diffusion samplers alike. Semi-Discrete Optimal Transport (SDOT) relaxes the map's regularity rather than the source's tail class. Its power diagram gives every training observation a cell holding exactly $1/N$ of the source measure, and tail observations are reached by crossing a cell boundary rather than by stretching. Our primary experiment sweeps severity over a calibrated Merton jump-diffusion spanning kurtosis 94 to 1,679. SDOT holds tail ratios at $0.85$–$0.94$ with cross-seed standard deviations below $0.025$, while every learned generator either compresses the tails or inflates them with a variance that grows alongside. Further experiments carry the result to real S&P~500 returns and to a 21-year backtest, where SDOT gives the best risk-adjusted market-neutral strategy under CVaR optimization (Sharpe $0.70$, max drawdown $-2.60\%$, against $0.40$ for the next-best generator).

Platforms summarize providers' past achievements into credentials that buyers use to judge quality. Generative AI can now produce much of the work those achievements certify, raising fears that those quality signals are worthless. We audit how well such credentials stay informative in Kaggle's 2010-2026 archive, where medals are won on predictions scored against withheld answers and two evaluation formats ran side by side. Across 444,698 participations, a medal's power to predict performance sits almost entirely in its first year, in both formats and eras. Fresh medals kept most of their value through the AI transition. About half of the collapse in the informativeness of one format's medals is institutional: the platform had been phasing out that format before AI, and its medal stock aged out on schedule. Old medals look more informative only in isolation. The platform's official lifetime-tier display discards up to a sixth of the medals' predictive power. A recency-weighted index fit before the AI era explains AI-era performance about 13% better than the tiers and selects entrants who perform better on average, though the tiers still identify extreme top performers better. Displaying a recent-performance summary alongside the lifetime tiers would recover the discarded information for buyers.

Decentralized lending lacks a credit bureau: a borrower's capacity to repay must be inferred entirely from public on-chain activity, without income verification or a liability record. This paper presents zLend, a deployed cash-flow underwriting framework that reconstructs a wallet's daily balance history from raw token transfers and derives short-duration repayment-capacity signals from it. The reconstruction is performed twice per wallet, once restricted to a fixed stablecoin basket and once over all fungible transfers, on the premise that a wallet's total token holdings and its liquid, spendable balance are distinct quantities whose conflation misprices risk. From each series we derive liquidity coverage against a fixed loan size, cash-flow volatility and regularity, a drawdown-and-recovery statistic adapted from quantitative finance, and a recurring-counterparty detector that identifies salary-like payment cadence from transfer timing alone. The two views are then compared: a wallet with large aggregate holdings whose stablecoin reserve rarely covers the loan size is flagged as a liquidity mismatch irrespective of total wealth. We specify the pipeline formally, document the golden-master methodology used to verify a cross-language production migration to numerical tolerance 1e-9, and characterize the tier function's parameter sensitivity with an independent reimplementation validated to exact agreement (78 of 78 field assertions) against the deployed system's reference fixtures. Tier assignment is governed predominantly by the reference loan size, with four of six reference wallets changing tier across loan sizes from USD 10 to USD 25,000; the drawdown and coverage criteria bind on disjoint wallets, so neither subsumes the other; and no criterion in the tier rule is inert. zLend is deployed in production, informing real lending decisions via third-party API integrations.

Most digital job-search assistance encourages unemployed workers to broaden their search toward related occupations, targeting one important source of search inefficiency: insufficient occupational diversification. Our analysis suggests that the relevant margin of adjustment depends on the underlying search problem. Building on a detailed analysis of job seekers' beliefs and search behavior, we identify a large group of pessimistic workers for whom the main constraints are low search effort and low aspirations, rather than insufficient occupational diversification. This diagnosis points to an unexpected intervention: rather than encouraging these workers to search in new occupations, we encourage them to search more intensively and apply for better-paying jobs within the occupations they already consider. We evaluate this diagnosis-based intervention, alongside a standard occupational recommendation, in a large-scale randomized experiment conducted with the French Public Employment Service. The motivational intervention increases search effort, raises reservation wages, and improves reemployment outcomes along the predicted margins. Occupational recommendations, by contrast, primarily benefit workers whose search problem lies in the allocation of attention across occupations and operate by activating existing perceptions rather than correcting beliefs. More broadly, our findings show how digital platforms can combine subjective expectations, behavioral data, targeted interventions, and randomized experimentation to diagnose job seekers' needs and iteratively improve intervention design.

We propose a novel valuation framework for contingent convertible (CoCo) bonds based on the issuing bank's Common Equity Tier 1 (CET1) ratio, which is widely acknowledged as an indicator of a bank's solvency. Our approach develops a bivariate jump-diffusion model that captures the dynamic relationship linking the CET1 ratios, share prices, and CoCo bond prices, incorporating both continuous market movements and correlated jump risk. The model advances existing literature through three key innovations: (1) a hybrid mechanism for modeling regulatory discretion in trigger decisions, (2) a class of power conversion schemes that generalizes traditional approaches while maintaining analytical tractability, and (3) a method to overcome the temporal discrepancy between high-frequency market data and low-frequency regulatory reporting. We derive semi-closed form formulas for both write-down and equity-convertible CoCo bonds and validate our model through five case studies spanning from 2009 to 2023, including an in-depth analysis of the 2023 Credit Suisse collapse. The results demonstrate a significant improvement in pricing and hedging performance while highlighting the model's data-adaptive nature that enables short-term predictions.

Job seekers' expectations about reemployment are increasingly used to study job search, but what their biases reveal about underlying beliefs and preferences is ambiguous. We combine new survey data, structural modeling, and machine learning to uncover the informational content of these expectations and show how they can be used to improve the targeting of employment support. Using a new panel of French job seekers' subjective expectations linked to administrative records, we show that reemployment expectation biases are strongly associated with, and summarize, biases in beliefs about the two fundamentals of search, job offer arrival rates and the wage distribution. These underlying biases are heterogeneous but strongly positively correlated, so their effects on search compound. We then estimate a structural job search model with multiple sources of biased beliefs and show that correcting them helps pessimistic job seekers but can demotivate and hurt optimistic ones, providing a rationale for targeting. Finally, we develop a machine-learning stratification that recovers policy-relevant groups, with distinct patterns of biased beliefs and behaviors, from easily elicited reemployment expectations alone. This gives employment services a simple tool to target informational interventions.

We measure volatility roughness across asset classes using a common data infrastructure and pipeline. Our data covers 3,926 United States equities, 34 CME futures roots, rates, FX, and commodities, and options on 44 underlyings over 2010-2025. Realized volatility is rough everywhere. The class-median Hurst estimate ranges from $0.05$ (livestock) through $0.07-0.10$ (rates, FX, agriculture, energy, metals) to $0.13$ (single stocks) and $0.20$ (equity indices). The option-implied measure identifies $H$ only where the leverage effect produces a clean skew term structure. For the equity indices, implied estimates of $0.21-0.28$ are just above realized volatility $H$, while for rates and FX the ATM skew regression fails with an R-squared near zero even though realized volatility remains rough. We also show a mean-reversion contamination formula for the second-moment estimator of the roughness for the stationary fractional Ornstein-Uhlenbeck process. The local slope of the increment second moment deviates from $2H$ by $(1-H)\Gamma(2H+1)(\kappa\Delta)^{2-2H}$ for all $H\in(0,1)$. A correction framework for the second moment, when the log realized volatility measure has additive noise, raised the $H$ estimate slightly but nowhere near the Brownian diffusion framework. Finally, a failure taxonomy discusses where rough-volatility methods apply and where they fail, suggesting alternative paths to further explore the rough-volatility paradigm.

Research activity generates highly localized positive spillovers, yet in the U.S. it has become increasingly spatially misaligned with population and economic activity as people moved away from legacy cities where major research institutions remain anchored. Reallocating researchers toward population centers could broaden local access to knowledge but may reduce knowledge production if legacy institutions or large research clusters raise researcher output. The spatial reallocation of researchers therefore entails a trade-off between gains in knowledge access and losses in knowledge production. To quantify the knowledge production loss from reallocation, we use bibliographic data to separate individual effects from location effects and show that location effects account for substantial differences in research output across locations. Our instrumental-variables estimates provide robust evidence of own-institution agglomeration effects but weaker evidence of external-cluster effects. We then incorporate these estimates into counterfactual reallocations and show that marginally reallocating researchers toward large metropolitan areas with relatively little research activity can plausibly raise aggregate output because the associated knowledge production losses are modest and can be offset by small access gains.

The terminal crisis of capitalism has been predicted regularly since 1848. It has not occurred. We treat this 175-year forecasting record as a stylized fact in need of a model, and we supply one, unifying results from adaptive systems, mechanism design and computational complexity, behavioral decision theory, global games, and the political economy of redistribution. Ten assumptions deliver twelve formal results and one theorem. Recurring shock classes have geometrically decaying collapse hazards: within the banking class alone, the lag from systemic failure to a system-wide deposit backstop fell from 816 days (1930-33) to 18 days (2008) to 2 days (2023). A centralized replacement of the price mechanism fails on at least one of two horns, computation or truthful input, and we are explicit about which horn survives modern algorithms. We derive Davies's J-curve from cumulative prospect theory together with a new result: the revolutionary window closes endogenously, with length inversely proportional to the speed of hedonic adaptation, roughly four to nine years at calibrated parameters, and with its perceptibility threshold derived from the participation constraint rather than assumed. Personalized media shift protest games, in the model, into the unique no-attack equilibrium of a global game. Redistribution is a dial, not an identity: countries with market Ginis of 0.505-0.512 sustain disposable Ginis from 0.273 to 0.395. We state seven testable predictions and six falsification conditions, and map which assumption each rival school must deny.

We design a method for measuring the risk preferences of agents in the deep past. The method combines a structural model of crop choice as a portfolio allocation with machine-learning prediction of expected crop returns, using historic agronomic and climate data. We estimate county-level risk preferences for the United States and farmer-level preferences in Kansas from 1889 to 1929. More risk averse farmers leveraged less, were less likely to purchase novel WWI Liberty Bonds, and were more likely to participate in local risk-sharing institutions. We show that higher risk aversion predicts slower tractor adoption and farm mechanization during the 1920s.

Reinforcement learning has gained increasing attention as a data-driven approach for stock trading. However, learning a policy that is both profitable and stable remains challenging due to non-stationary market behaviour and noisy reward signals. Auxiliary tasks are often used to improve representation learning and stabilize training, yet they are usually designed manually and depend heavily on prior assumptions about targets and prediction horizons. Such fixed designs may not remain suitable across changing market regimes. In this work, we propose a self-supervised framework that automatically discovers auxiliary tasks to support reinforcement learning for stock trading. The auxiliary tasks are formulated as General Value Functions so that their predictions enrich the learned state representation and assist policy optimization. The framework consists of two networks. The main network learns the trading policy along with the auxiliary predictions, while the secondary network generates the definitions of auxiliary tasks through learned cumulants and discount factors. These tasks are updated using a meta gradient mechanism that accounts for their long-term impact on trading performance and improves training stability. We evaluate the proposed approach across four major equity indices: DJI, FTSE, Sensex, and TAIEX. The empirical results demonstrate that automatically discovered auxiliary tasks lead to more robust learning and improved trading performance compared to existing baselines.

We study optimal investment for insurers managing participating (profit-sharing) contracts under probability distortion and probability benchmark (aspiration) constraints. The problem combines three theoretical complexities: (i) nonconcave effective utilities induced by embedded guarantees and surplus-sharing rules, (ii) probability weighting capturing behavioral aspects of long-horizon decisions, and (iii) aspiration-type constraints formalizing solvency requirements. Using quantile formulations and concavification techniques, we derive explicit closed-form solutions for optimal terminal wealth and trading strategies in both complete and incomplete Black-Scholes markets. Our utility class accommodates the piecewise hyperbolic absolute risk aversion (PHARA) family and covers nonconcavities arising naturally in insurance contexts. The framework reveals how probability distortion weakens lock-in behavior and induces time inconsistency: under inverse S-shaped distortions, insurers overestimate upside probabilities and increase risky investment relative to undistorted benchmarks. Asymptotic analysis and numerical illustrations demonstrate regime switches in optimal policies driven by regulatory thresholds and capital constraints. Our results extend the hope-fear-aspirations framework of He and Zhou (2016) and provide practical insights for managing insurance balance sheets under behavioral preferences and solvency constraints.

We study continuous-time multi-asset portfolio choice and consumption under smooth pointwise constraints, including state-dependent feasible sets. The method separates dynamic information acquisition from local constrained recovery. A pointwise-feasible neural actor generates reference rollouts; after training, its realized latent outputs are frozen and first- and second-order adjoints are harvested from a fixed-latent open-loop backpropagation-through-time graph. Feedback therefore generates the reference trajectory without restricting the adjoint formulation to Markov controls. Conditional on the harvested adjoints, deployment solves a local generalized Pontryagin-Hamiltonian problem: quadratic-affine portfolio blocks are recovered exactly by a quadratic program, while a log barrier approximates more general regular KKT branches. We establish local chart representations, an OL-BPTT-to-adjoint correspondence retaining orthogonal martingale residuals, and an end-to-end bound from reference value loss and numerical errors to recovered-policy and local QP-gap errors. Analytical constant- and predictable-opportunity benchmarks validate the adjoints. Common-input experiments show that recovery reduces residual PMP/KKT error left by finite-budget direct policy optimization, including under a state-dependent consumption cap and with up to 100 risky assets. Scalability concerns the constrained action block rather than dimension-free state-space complexity.

We study the blockchain censorship attack introduced in [21], which shows that joining the attack is a dominant strategy. We show that, by introducing certain detectability threshold, joining the attack can lead to strictly less reward for whales, which are defined to be a small number of validators that hold significantly more voting power than the rest (henceforth known as minnows). This leads to a change of the equilibrium: With whales unwilling to participate in the attack, it is difficult for minnows alone to launch the attack. We also perform Monte Carlo simulation to show the existence of reduction for whales' reward in Ethereum and Solana.