Online advertising platforms host hundreds of thousands of A/B tests, but the platform's delivery algorithm routes each creative to the audience it predicts will engage. Every two-arm test therefore conflates the creative's effect with the algorithm's targeting response, and adjusting for the realized audience is biased because audience is a post-treatment mediator. We propose a three-arm design that adds an arm exposing the algorithm to the treatment metadata while holding the user-facing creative identical to control, point-identifying the natural indirect (algorithmic) and direct (creative) effects without sequential ignorability. In a live Meta campaign with a women-targeted text fragment, the algorithmic channel raises female impression share by +2.07 ppt while the creative channel moves it by -0.68 ppt; roughly three-quarters of the absolute reallocation is algorithmic, and a conventional two-arm test understates the algorithmic channel by a factor of two. The design isolates the contribution of platform's algorithm to the outcome which is separable from creative content.

We study consumer demand in large-scale retail settings with many products, multiple categories and repeated purchase behavior. While inertia and brand loyalty are well documented, existing discrete choice models typically focus on single categories or become computationally infeasible in high-dimensional environments. We propose a dynamic product-level factor model that captures heterogeneity in baseline preferences, price sensitivity and inertia through a shared latent factor structure. By factorizing individual-product coefficients, the model pools information across individuals and categories and allows for correlated heterogeneity. We estimate the model using Bayesian variational inference, enabling scalable estimation with tens of thousands of parameters. In a simulation study calibrated to realistic retail data, we show that the dynamic factor model substantially improves predictive performance relative to static factor models and mixed logit benchmarks, particularly when individual purchase histories are sparse. Accounting for inertia also leads to more elastic demand estimates, underscoring the importance of dynamics for measuring consumer responsiveness. Our results highlight dynamic factor models as a scalable and flexible approach for demand estimation in modern, high-dimensional retail markets.

This paper studies what Bitcoin (BTC) premiums in peer-to-peer (P2P) markets measure. Using transaction-level data from LocalBitcoins, we construct BTC premiums for 80 currencies relative to the U.S. dollar and relate them to blockchain transaction conditions, centralized crypto market (CEX) conditions, cross-border payment frictions, and foreign exchange (FX) markets. We show that these premiums reflect both trading frictions within crypto markets and local frictions in access to cross-border payments. They vary systematically with blockchain conditions and broader crypto market conditions, including BTC returns and volatility, and they are larger in countries facing greater frictions in conventional cross-border payment channels. This pattern is especially pronounced in economies with binding institutional constraints, i.e., tight capital controls and non-floating exchange-rate regimes, consistent with greater reliance on P2P crypto markets as an alternative cross-border payment channel. We further show that rising FX pressure is absorbed mainly through prices rather than trading volumes, and that P2P BTC premiums predict subsequent official exchange rate depreciation. Although premium levels differ across countries, their predictive content remains broadly similar. Overall, P2P BTC premiums reflect limits to arbitrage across crypto trading venues, especially where formal cross-border payment channels are more constrained, and they also embed forward-looking information about currency depreciation.

Decision support methods from operations research are widely used to support complex planning decisions. Within the energy sector, energy system models (ESMs) applying modelling to generate alternatives (MGA) generate large sets of near-optimal, different system configurations. However, they typically generate and analyse alternatives in the model variable space without ensuring stakeholder relevance. Multi-criteria decision analysis (MCDA), in contrast, provides a structured means to account for conflicting objectives and heterogeneous stakeholder interests but often relies on a limited set of pre-defined alternatives that may not appropriately represent the feasible solution space. To address these limitations, this work proposes value-focused modelling to generate alternatives (VF-MGA), a novel methodology that bidirectionally couples MGA and MCDA. Stakeholder objectives elicited within the MCDA inform the MGA-algorithm, enabling a stakeholder-orientated diversification of the alternatives, which are subsequently evaluated within the MCDA based on elicited stakeholder preferences, thereby providing a comprehensive decision basis. Applied to a case study on the decarbonised energy supply of a large university campus, involving eleven stakeholders representing diverse institutional groups, VF-MGA (i) systematically integrates stakeholder objectives into the generation of 691 alternatives reflecting stakeholder-relevant interests, (ii) enables the identification of stakeholder-relevant alternatives from this large set through MCDA-based evaluation, and (iii) provides more differentiated stakeholder preference information by evaluating a large and diverse set of alternatives, thereby revealing acceptable ranges for system options. With this, VF-MGA provides a generalisable methodology for complex planning decision integrating quantitative modelling with participatory decision analysis.

We study how individuals trade off outcome ("what") and process ("how") utility in high-stakes strategic decisions, namely professional tennis. Using optimality conditions and the second-service rule, we derive a sufficient condition for the nonparametric lower bound on the weight of process utility to be positive. Under mild shape restrictions, the high-frequency data indicate that most players likely value process utility positively. We then develop a structural model that recovers player-specific preferences over outcomes and processes. Estimates show that players systematically sacrifice success probabilities to increase process utility, with economically meaningful consequences for match outcomes and expected earnings.

Investment in renewable electricity generation is highly capital intensive and therefore strongly dependent on financing conditions. In Europe, much of this investment has occurred under public support schemes that resemble long-term public contracts such as feed-in tariffs (FiTs) and contracts-for-differences (CfDs). These contracts not only subsidize renewable generation but also stabilize project cash flows by reducing exposure to electricity price volatility, thereby improving debt capacity and lowering financing costs. At the same time, they may distort operational and investment incentives by weakening exposure to wholesale market price signals. This paper studies how alternative public contract designs reduce revenue risk and how this translates into financing outcomes. Using a novel dataset of hourly turbine-level generation covering 63 German onshore wind parks over the period 2013-2024, we simulate project cash flows under two-sided CfDs, one-sided CfDs, and financial CfDs. We then evaluate their implications for cash-flow volatility, debt capacity, and the levelized cost of electricity using a project finance model based on a conservative debt-service coverage ratio (DSCR) constraint. We find that financial CfDs provide hedging performance comparable to conventional two-sided CfDs. The results suggest that the commonly assumed trade-off between revenue stabilization and efficient market integration is not inherent but depends on contract design. More broadly, public contracts can substitute missing long-term hedging markets. These results have direct policy implications for the design of renewable energy support schemes.

Generative artificial intelligence (AI) is expected to transform work, but less is known about how firms reorganize labor demand as the technology diffuses. Existing research has largely focused on which occupations are exposed to AI or whether exposed jobs decline. We extend this debate by examining whether firms adjust by changing where they hire, what jobs contain, or both. Using a nationwide dataset of job postings in the United States, covering all sectors of the economy, we construct a dynamic, posting-level measure of generative AI exposure with a two-stage large language model pipeline. The pipeline identifies the tasks described in each posting and classifies the extent to which generative AI can perform or assist them. We then decompose changes in aggregate exposure into two margins: reallocation of demand across jobs and redesign of tasks within jobs. We document three main findings. First, generative AI exposure is dynamic rather than fixed, changing substantially over time. Second, labor demand adjusts through both margins. Hiring reallocation explains the largest share of the aggregate decline in exposure, accounting for 52% on average, while within-job redesign becomes increasingly important, accounting for 39.5%. A complementary Oaxaca-Blinder decomposition shows that shifts in occupational composition account for about 90% of the exposure change attributable to observable job characteristics. Third, adjustment differs across the job ladder. Senior jobs adjust earlier and mainly through reallocation, whereas junior jobs adjust through a broader mix of reallocation, redesign, and their interaction. These findings suggest that labor-market adjustment to generative AI is a process of organizational reconfiguration, in which firms reshape both hiring demand and the task architecture of work.

Modern economies depend critically on high-voltage power transmission networks. Yet this infrastructure is routinely disrupted by natural hazards ranging from earthquakes and floods to tornadoes and geomagnetic storms. Risk assessments have historically addressed hazards in isolation, leaving no common basis for comparing economic impacts across the full hazard portfolio. This study addresses this gap by developing an integrated framework linking hazard characterization, fragility modeling, and macroeconomic impact propagation. The framework is applied consistently across nine primary hazards and one compound freezing rain and wind gust hazard. Using national hazard datasets and a US high-voltage transmission network of over 13,000 line segments and 10,000 substations, we derive failure probabilities, expected damage, affected population, and downstream economic output losses. Among individual hazards, tropical cyclone wind produces the largest expected daily damage at $137 M/day, followed by lightning at $87 M/day, earthquake at $47 M/day, flood at $46 M/day, tornado at $42 M/day, and landslide at $34 M/day. Downstream economic output losses are largest for tornado at $4.93 B/day, followed by flood at $3.59 B/day and earthquake at $3.02 B/day. A 250-year geomagnetic storm produces $2.07 B/day, placing space weather within the range of major terrestrial hazards. The compound freezing rain and wind gust scenario produces the largest stress-test disruption, affecting 237.4 M people and yielding a modeled downstream output loss of $85.16 B/day. These results should be interpreted as first-order bounding estimates, with the compound scenario representing an upper-bound stress test. Overall, the framework establishes a consistent baseline for prioritizing investments in transmission network resilience.

This paper examines how firm-level determinants of industrial emissions evolve over time as firms adapt to environmental regulation, economic conditions, and organisational constraints. Using a panel of 204 U.S. industrial facilities observed from 1992 to 2023, we link facility-level emissions from the Toxics Release Inventory to firm financial characteristics, managerial attributes, local labour-market conditions, and aggregate macroeconomic indicators. We employ a time-varying mean-group estimator that allows average relationships to change smoothly over time while accommodating persistent heterogeneity across facilities. We find several covariates display episodic associations with emissions growth. The results reveal pronounced stage-like dynamics in emissions determinants, with firm-level characteristics and aggregate conditions dominating in different periods. From an innovation-policy perspective, the findings highlight that firms' responses to environmental regulation are time-dependent and shaped by their adaptive capacity.

Understanding public demand for urban ecosystem services (ES) is crucial for effective green space management, yet the intricate relationships and potential trade-offs among these diverse demands remain poorly understood. Previous studies have yielded inconsistent findings, often limited by small samples or reliance on indirect proxies. Here, we provide the first national-scale, direct assessment of the relationship among demands for nine urban park ES using a survey dataset comprising 20,075 responses across China and a point-allotment experiment that directly quantifies the trade-off patterns among service demands. We found particularly strong preferences among urban residents in China for air purification and recreation services, at the expense of other services. These preferences were further reflected in three distinct demand bundles: air purification-dominated, recreation-dominated, and balanced demands, each delineating a typical group of people with distinct representative characteristics. Socio-economic and environmental factors, such as age, environmental interest, and mean annual precipitation, significantly influence the trade-off intensity among service demands. Our study pioneers the direct, quantitative analysis of relationships among ecosystem service demands, and the results underscore the need for tailored urban park designs that address diverse service demands to sustainably enhance the quality of city life in China and beyond.

Bahar (2025) argues that there is a long-term cointegrating relationship between US job vacancies and southwest border crossings. We show that this conclusion is based on a misspecified Engle-Granger test applied to first differences. Once the Engle-Granger test is correctly applied to levels, evidence for a cointegrating relationship vanishes, invalidating the paper's approach to estimating short- and long-run elasticities. Bahar's approach is therefore uninformative about the relationship between US labor market conditions and migration.

We consider causal inference in dynamic settings where treatment is assigned by thresholding a state variable that can change over time. There is a large literature on regression-discontinuity methods building on the fact that, in the static setting, treatment assignment via threshold crossing induces a quasi-experimental design that enables pragmatic causal inference. But dynamic settings involve challenges not present in the static setting, e.g., past treatments may affect current state and thus future treatments, and so existing regression-discontinuity methods do not apply. Here, we show that dynamic thresholding designs identify a marginal policy effect that nests the classical regression-discontinuity parameter in the static setting; and propose a tailored local linear regression estimator that is consistent for this marginal policy effect. We demonstrate our approach using an experiment that emulates real-world optimization of thresholds for continuous glucose monitoring using data generated from an FDA-approved simulator.

This study investigates the effects of artificial intelligence (AI) adoption in organizations. We ask: First, how should a principal optimally deploy limited AI resources to replace workers in a team? Second, in a sequential workflow, which workers face the highest risk of AI replacement? Third, would the principal always prefer to fully utilize all available AI resources, or are there any benefits to keeping some slack AI capacity? Fourth, what are the effects of optimal AI deployment on the wage level and intra-team wage inequality? We develop a sequential team production model in which a principal can use peer monitoring--where each worker observes a signal of their predecessor's effort--to discipline team members. The principal may replace some workers with AI agents, whose actions are not subject to moral hazard. Our analysis yields four key results. First, the optimal AI strategy stochastically replaces workers rather than fixating on a single position. Second, the principal replaces workers at the beginning and at the end of the workflow, but does not replace the middle worker, since this worker is crucial for sustaining the flow of information obtained by peer monitoring. Third, the principal may optimally underutilize available AI capacity. Fourth, the optimal AI adoption increases average wages and reduces intra-team wage inequality.

Important questions for impact evaluation require knowledge not only of average effects, but of the distribution of treatment effects. The inability to observe individual counterfactuals makes answering these empirical questions challenging. I propose an inference approach for points of the distribution of treatment effects that uses predicted counterfactuals through covariate adjustment. I provide finite-sample valid inference using sample-splitting and asymptotically valid inference using cross-fitting under arguably weak conditions. Revisiting five randomized controlled trials on microcredit that reported null average effects, I find important distributional impacts, with some individuals helped and others harmed by the increased credit access.

This paper develops a set of empirically tractable and flexible sieve estimators for semi-nonparametric multidimensional matching models with transferable utility, focusing on worker-job matching. We generalize the parametric quadratic-Gaussian framework employed by Bojilov and Galichon (2016) and Lindenlaub (2017), which relies on joint normality of observed characteristics. We allow unrestricted distributions of characteristics and show identification of the production technology and the equilibrium wage and matching functions using optimal transport theory. Given identification, we propose efficient, consistent, and asymptotically normal sieve estimators. We revisit Lindenlaub's empirical application and show that, between 1990 and 2010, the U.S. economy experienced much larger technological progress favoring cognitive abilities than the original findings suggest. Furthermore, our flexible model specifications provide a significantly better fit for patterns in the evolution of wage inequality.

This paper examines the sociodemographic and socioeconomic determinants of regional commuter mobility in the Greater Stockholm Area using a heteroscedastic spatial Durbin panel data model estimated via Bayesian Markov Chain Monte Carlo methods. Drawing on mobile phone-derived origin-destination flows from the MIND database, the analysis exploits unusually fine spatial and temporal granularity across a balanced panel of 675 regions over the period 2018-2023. A k-nearest neighbor spatial weight matrix (k = 18), selected via Bayesian model comparison, captures the topological structure of interregional connectivity. By modeling spatial lags in both the dependent and independent variables, the framework enables explicit recovery of direct (own-region) and indirect (spillover) effects from scalar summary measures of the matrix of partial derivatives -- providing robust posterior inference on how sociodemographic and socioeconomic conditions propagate through space. This approach addresses a key limitation of conventional non-spatial methods, which risk producing biased estimates by ignoring spatial interdependence. Empirical results confirm that spatial spillovers predominate over direct effects, with educational attainment and car ownership emerging as the principal determinants of commuter mobility, while age composition plays a comparatively modest role. These findings underscore that evaluating direct effects in isolation systematically underestimates the broader societal returns to mobility-enhancing regional policies.

This paper proposes a computationally efficient mechanism for multi-dimensional matching markets where agents report preferences over object features rather than complete utility assessments. We use Singular Value Decomposition (SVD) to identify the principal direction of variation in feature space and match agents to objects along this dimension, reducing a complex multi-dimensional problem to an effectively one-dimensional problem solvable in $O(N \log N)$ time. We show that when data exhibit low effective dimensionality, our mechanism approximately maximizes Nash Social Welfare, satisfies distributional truthfulness, and achieves symmetry. We establish a novel connection between Nash Social Welfare and Geometric Distributionally Robust Optimization, providing robustness guaranties. Numerical experiments demonstrate that our approach achieves 99\% optimal welfare while running three orders of magnitude faster than direct optimization. The framework applies naturally to school choice, labor markets, and course allocation, where feature-based elicitation reduces the cognitive burden on agents.

The space L of linear value maps on a finite-player cooperative game G^N is finite-dimensional, and admits a canonical inner product induced by the Harsanyi-dividend decomposition of G^N. We show that this inner product is intrinsic: the same value arises from any orthonormal basis of G^N with respect to the Harsanyi inner product. Within this geometry, the subspace L^{ESL} of efficient, symmetric, linear value maps admits a clean structure theorem. The induced orthogonal stratification of L by coalition size yields a canonical linear isomorphism L^{ESL} = R^{n-1}, under which every efficient symmetric linear value map decomposes uniquely into n-1 stratified epsilons, one per coalition size. The classical egalitarian Shapley family of Joosten (1996) is precisely the diagonal slice of this R^{n-1}. The orthogonal projection of any Psi in L^{ESL} onto this diagonal yields an optimal parameter eps*(Psi) equal to the weighted mean of the stratified epsilons under an explicit probability distribution {w_a} over coalition sizes, and the goodness-of-fit R^2(Psi) equals one minus the relative weighted variance of those epsilons. The framework is a literal regression-statistics analogue of the coefficient of determination. At n=4 it produces a clean three-way classification of the standard alternatives to the Shapley value: the Banzhaf value is nearly orthogonal to the egalitarian Shapley axis (R^2 ~ 1%); the equal-surplus-division value is moderately aligned (R^2 ~ 38%); the solidarity value is almost entirely aligned (R^2 ~ 99.6%). Asymptotically R^2(ESD) -> 1, R^2(So) -> 1, and R^2(Bz) -> 1/2, the last reflecting a structural identity between the efficiency defect and the egalitarian-Shapley deviation of the Banzhaf value at every coalition size.

We develop an abstract axiomatic theory of tie-breaking. A tie-breaking input consists of a finite set N of players, a weak order on N representing the standings to be refined, and an auxiliary information item drawn from a set on which the symmetric group Sym(N) acts. Within this minimal framework we prove three theorems. First, no tie-breaking rule producing a strict linear order can be anonymous, provided the input space contains even one intrinsically symmetric situation, a condition met in essentially every realistic application. Second, when we allow the rule to output a partition of N (rather than a strict ranking), there is a unique rule satisfying two natural axioms: it is the partition of N into orbits of the joint stabilizer of the input. Third, every reasonable strict tie-breaking rule decomposes uniquely as the canonical orbit partition followed by an arbitrary completion. The decomposition makes precise the informal observation that real tie-breaking systems are honest until forced to be arbitrary. The framework is broad enough to capture chess tournament tie-breakers, sports league regulations, voting tie-breakers, tie-breaking among symmetric players in cooperative games, and ranking by network centrality measures, all within a single uniform formalism.

What happens when the strongest alliance member pressures a weaker member over territory and strategic control? We examine the Greenland sovereignty crisis as a stress test for LLM geopolitics, centered on the 2019-2026 U.S. push to acquire Greenland from the Kingdom of Denmark. The crisis nests two collective-action problems: Arctic strategic control and whether NATO can enforce alliance norms against the dominant member. We develop three games (asymmetric coercion; a NATO assurance game with a critical-mass tipping point; a triadic extensive-form game with social preferences) and test them with a multi-agent simulation in which eight frontier LLMs play six geopolitical roles (United States, Denmark, Greenland, NATO, Russia, Canada) across 3,604 completed games and 108,120 action observations. Using inverse game theory, we recover each model's structural utility parameters (alpha, beta, gamma, delta, eta) for material self-interest, reciprocity, inequality aversion, norm respect, and commitment consistency. Three findings stand out. First, all eight models become more escalatory under coercion framing (four-action escalation rises from 10.7% to 28.6%). Second, Chinese-origin models show systematically different power-weight profiles from Western-origin models when playing the U.S. role. Third, peaceful US acquisition emerges in only 1.9% of clean games and only 3 of 8 frontier models ever achieve it, most prominently DeepSeek V3.2, which executes a stable five-round playbook through the metropole. Prompts emphasizing jus cogens and self-determination reduce escalation back near baseline in the English-only confirmatory sample; multilingual contrasts are reported as exploratory sensitivity checks. We position this as a structural benchmark for LLM geopolitical behavior, complementing action-frequency benchmarks.

Ambiguity-averse decision makers typically dislike not only the presence of ambiguous events but also their increase, contrary to what standard ambiguity models predict. We axiomatically study such a decision maker. She avoids ex ante randomization over prospects since it only increases the number of relevant ambiguous events without providing a hedge against uncertainty. Our axioms lead to a representation in which the decision maker behaves as if optimizing her ambiguity perception at a cost. We show the uniqueness of the representation, and conduct comparatives of attitudes toward ambiguity and its increase. This identification is not achieved without considering ex ante randomization.