Papers

Knowledge spillovers have largely been studied as unintended externalities diffusing through geographic proximity, while deliberate interregional networks have been examined mostly through co-inventor ties rather than the scientific layer carried by co-authored publications. This study asks whether spillovers from interregional AI-related co-publications increase regional AI patenting, and whether the ownership composition of regional R&D shapes the capacity to capitalize on them. Using a network-econometric approach across 26 Chinese provincial-level regions from 2006 to 2020, we find that co-publication spillovers foster AI patenting, but the returns concentrate almost entirely in regions where privately owned enterprises dominate R&D, reinforcing regional divergence.

Marketers now deploy generative AI agents as synthetic consumers to pretest visual assets such as logos, packaging, and advertising at a fraction of human-panel cost. However, this procedure assumes that a model seeing a visual cue can also perceive its consumer meaning, which is largely untested. We stress-test the assumption using six canonical visual marketing experiments, varying the two levers managers control: model generation (GPT-4o-mini vs. GPT-5.4-mini) and input format (plain text vs. JSON). Every resulting configuration passed the manipulation checks; however, none of the configurations reproduced more than two of the six human effects, and the remainder were nonsignificant. The one exception was a significant reversal of the human pattern. Providing conceptual or empirical evidence through in-context learning steers average responses toward the human effect. Yet steering has a limit: even when it succeeds, a configuration reproduces less than half of the natural spread of human responses and so understates consumer heterogeneity. We integrate these results into an AI governance protocol (Calibrate, Intervene, Deploy) that delineates when synthetic consumers can responsibly screen creatives and when human panels remain necessary.

Standard valuation methods, including discounted cash flow, the income approach standard IDW S 1 of the Institute of Public Auditors in Germany, and market multiples, compress milestone probabilities, continuation options, and risk shifts into opaque aggregate parameters; none provides a structured protocol for decomposing AI integration into auditable option-level assumptions. We propose an industry-agnostic taxonomy separating AI Integrators from AI Providers. AI Integrators are further classified by their Integration Depth Level, ranging from no integration to AI at the core of the product or process. A milestone-gated real-options overlay decomposes milestone state value into five components, and an Analytic Hierarchy Process-based Success Readiness Index derives per-option probabilities from structured pairwise comparisons for scenario analysis. Applied to an AI-native energy software-as-a-service firm, the framework yields a coherent valuation band traceable to identifiable option-level assumptions. Risk concentrates in later-stage continuation options, matching the structural prediction for AI Providers. The protocol applies across the firm lifecycle, including mergers and acquisitions due diligence. The case is a single-firm demonstration of protocol coherence, not empirical validation; multi-case testing against realised post-exit valuations is left to future research.

Organizations invest heavily in internal mechanisms such as incentive systems, governance structures, performance measurement, analytics, and repeated reorganizations. Yet realized performance gains are often weak or short-lived, particularly in large and mature organizations. Standard explanations emphasize allocative inefficiency arising from incentive misalignment, information problems, or bounded rationality. While relevant, these explanations struggle to account for environments in which effort is high, optimization capabilities are sophisticated, and aggregate performance remains far below apparent potential. This paper develops a theory of structural inefficiency inside organizations. The central mechanism is the presence of internal externalities: unpriced cross-effects of organizational mechanisms acting simultaneously on multiple performance objectives. When such effects are mixed in sign, organizational effort is partially canceled internally, creating a structural ceiling on realized performance. The paper introduces a simple representation of organizational primitives and key performance indicators and derives a realized-performance law in which internal conflict enters as a multiplicative discount factor. Under weak and realistic conditions, structural inefficiency dominates allocative inefficiency, explaining why optimization often fails before organizational coherence is restored.

We study the design of priority pricing systems with heterogeneous agents in environments in which improving quality for some agents reduces the average quality that can be provided. Contrary to the equity-efficiency tradeoff emphasized in public debates, we show that under economically natural conditions priority pricing can Pareto-improve on an equal-allocation benchmark. Three priority tiers suffice for such an improvement, combining higher quality for a fee, lower quality with compensation, and an intermediate tier at the benchmark quality; two tiers are never enough. Our results provide a framework for overcoming equity-efficiency tensions in applications such as lane pricing, waiting-line design, public provision, and insurance.

People often face environments where multiple models compete to explain the same observations. This paper examines how people update beliefs in such settings and how preferences over payoff-relevant states shape model selection and belief updating. This paper first develops a framework where preference-driven bias distorts the perceived model, affecting Bayesian and best-fit updating differently. In a laboratory experiment, most participants are classified as Bayesian updaters, who average across models, while a substantial minority are classified as best-fit updaters, who select the model that best fits the observed signal. Within-participant comparisons between the symmetric payoff and asymmetric payoff conditions indicate that asymmetric payoffs shift reported beliefs toward the preferred state, particularly among participants classified as best-fit updaters. Relative to symmetric payoffs, asymmetric payoffs increase the reported belief of the preferred state by about 8 percentage points among best-fit updaters, while the estimated effect among Bayesian updaters is close to zero. These findings help us better understand model-based learning and have implications for domains such as political polarization and financial investment, where competing narratives and strong preferences often coexist.

AI systems are becoming participants in the evaluation and use of scientific research. They encounter citation counts, download statistics and lists of popular articles developed around human readers, but the collective consequences of these signals for artificial readers remain uncertain. This paper adapts the Music Lab design to a market for academic attention. In the first experiment, 1,000 AI agents choose papers from the titles and abstracts of all 114 regular research articles published in the American Economic Review in 2025. The experiment has five independent-choice communities and five social-influence communities, each with 100 sequential agents. Only agents in the social-influence condition observe earlier selections within their community. Agents may select any number of papers. Social-information communities select 17.2 percent fewer papers per agent, concentrate their choices more heavily, and collectively cover 73 papers, compared with 90 independently. Between-community variation is greater under social information. In a second experiment with 200 agents across twenty social communities, randomly assigning papers five initial selections raises their subsequent selection rate by 45.55 percentage points (95% CI: 41.20 to 49.90). Choices have modest correspondence with external citations and little correspondence with download counts. The results show how a simple information rule shapes the volume, breadth and distribution of scientific attention in an artificial population.

We estimate housing externalities in a high-density city, exploiting the staggered rollout of Singapore's nationwide Main Upgrading Programme for public housing. Controlling for nonrandom neighborhood exposure, we find that upgrading raises treated buildings' prices by 11.5% upon completion and neighboring buildings' resale prices by about 2% within 500 meters, decaying to zero beyond. A model with distance-decaying externalities shows that in dense settings spillovers justify the distortions of in-kind provision; this advantage diminishes and reverses at lower densities. Administrative data on over 2 million residents show that upgrading disproportionately retains older incumbents, suggesting age-specific amenities as an underexplored externality channel.

People increasingly compete against AI agents rather than other human opponents. We distinguish two channels: an opponent effect and an information effect. These are different elements with different consequences: the opponent effect is specific to a given computational system, the information effect a property of the information environment that an organisation or policymaker can control. We separate them in a preregistered experiment (N = 1,395) using a dynamic all-pay auction, a repeated contest in which escalation of commitment arises from the incentives. What participants are told about the opponent (human, an AI trained to imitate people, or an AI trained to compete well) is varied and crossed with who they actually face, in a deception-free design. What people are told influences escalation: the median price rises by 6.7 points when a human might be the opponent and falls by 8.8 when an optimising machine might be, a spread of about 15% of the prize value of the competition, produced by information alone. Competing against the AI agents lowers prices, yet reduces the chance that both sides finish with positive earnings, showing distinct effects of the opponent channel. The information effect is not explained by articulated strategy, or individual differences, and is consistent with a competitive response engaged when a human is a live possibility. This shows that describing an AI competitor is not behaviourally neutral.

We examine whether post-origin training information inflates the measured accuracy and economic value of financial forecasts. We evaluate five sets of financial time-series foundation models, each comprising independently trained annual vintages under U.S., global, and factor-augmented training environments, across 14 equity markets and four forecast horizons. Rolling comparisons vary the annual vintage for a fixed forecast; fixed-vintage comparisons hold the vintage fixed as target windows move across its training cutoff. Each alternative forecast is paired with an origin-aligned point-in-time (PIT) benchmark using identical numerical histories and inference protocols. In the U.S.-trained reference environment, post-origin vintages materially revise informative PIT forecasts but generally reduce accuracy in both designs. Pooled rolling comparisons yield higher mean squared forecast errors in 18 of 20 U.S. model-set-horizon combinations. The origin-crossing update also performs worse on average than an equally long pre-origin update. Under a common constrained allocation rule using one-month forecasts, median exposed-minus-PIT differences in annualized certainty-equivalent returns are -1.77 percentage points in the United States and -2.14 points internationally. Global and factor-augmented training produce more mixed predictive effects. An exact squared-error decomposition shows that revisions improve accuracy when their error-correcting benefit exceeds their mean squared magnitude; under U.S. training, alignment with PIT errors generally falls short of this requirement. Temporal exposure therefore establishes an information-set violation, not sufficient evidence of inflated predictive accuracy or investor value.

Organizations often invest substantial effort in improving internal mechanisms such as processes, governance, and technology, yet realize only modest performance gains. This paper proposes a structural explanation for this pattern. We introduce the concept of internal externalities: unpriced cross-effects generated by internal organizational mechanisms that act simultaneously on multiple objectives. When these effects oppose one another, effort cancels internally, limiting attainable performance even without incentive misalignment or informational constraints. We formalize this mechanism by modeling internal mechanisms as primitives that exert signed effects across performance objectives. This structure gives rise to internal conflict, a measurable property that captures the fraction of organizational activity lost to cancellation. We show that conflict constitutes a form of structural inefficiency that constrains the returns to effort and renders allocative optimization ineffective unless addressed first. The framework implies a natural sequencing of interventions, in which structural cleanup precedes effort reallocation and optimization. The analysis yields testable predictions and governance implications for how organizations should scale, constrain, redesign, or deprioritize internal mechanisms.

I consider a neoclassical growth model with a constant absolute risk aversion (CARA) utility function and derive a global closed form approximation that is arbitrarily precise as the discount rate $\rho$ is close to the population growth rate $n$. I use it to show that the consumption function is strictly concave and that countries can have two different paths converging to the steady-state: front-loading and back-loading.

Recent crises have made supply security central to trade policy. We show how tradeable import certificates (TIC) implement targets for domestic production and reliable foreign supply while preserving gains from trade. A single certificate market per country decentralizes the welfare-maximizing allocation under heterogeneous targets, with certificate prices adjusting endogenously as conditions change. TIC robustly protect these targets against a range of deviations from trade agreements. A model of economic coercion microfounds the targets, linking them to shortage salience, supply reliability, bilateral trade dependence, and norms against yielding to coercion. Tariff-subsidy agreements require more information and are more vulnerable to hidden deviations. While a common carbon price provides a natural focal point for climate agreements, quantity-based security targets could be a more natural focal point for trade agreements than tariff and subsidy rates.

Prediction-market shares differ from traditional financial products in that, with no information or outside utility, classical delta-neutral Central Limit Order Book~(CLOB) market making cannot be financed by payoff-uninformative noise flow. Transaction-level evidence from a major prediction-market CLOB platform shows makers profiting not from spread but from carrying an under-priced side to settlement — the empirical signature of behavioral tail demand rather than classical, randomized noise. We build this tail demand directly into the model and study an LMSR and a CLOB on the same event. A pre-shock CLOB quote inside the common-signal band is picked off; competitive quotes therefore screen informed traders out of the book. CLOB makers earn screening rent by carrying the under-priced side to resolution, while informed flow routes to the LMSR. The venues coexist: the CLOB supplies the tail-demand rent margin that lets the LMSR recover part of its loss to informed flow, and AMM depth moves the CLOB premium with a sign set by maker-side contestability—widening it where standing quotes can be undercut, compressing it where a committed maker carries the book. With three or more outcomes, binary-book CLOBs pin switch prices but leave implied beliefs indeterminate, whereas the LMSR prices the outcome simplex coherently and uses collateral more efficiently.

LLM-based GUI agents increasingly act on behalf of users in digital environments that were designed with human users in mind. These graphical user interfaces were designed to support, but also deliberately steer, the behaviour and decisions of users. While behavioural biases in the textual outputs of LLMs are well-documented, far less is known about how such influence operates when models act as agents that perceive interfaces and execute decisions—and, in particular, whether the reasoning capabilities increasingly built into these agents make them more robust to it. Drawing on Dual-Process Theory, we empirically investigate whether LLM-based GUI agents are susceptible to automatic (Type 1) and reflective (Type 2) digital nudges, and how their reasoning configuration moderates this susceptibility. In a randomized online shopping experiment with 3,600 agents and a total of 21,600 simulations across six frontier models from three providers, we found that agents were vulnerable to both nudge types. Crucially, the reasoning configuration moderated these effects in opposing directions, reducing susceptibility to automatic default nudges while heightening it to reflective social influence nudges. Extensive reasoning therefore did not make agents more robust but redirected the route through which choice architecture takes effect. Exploratory analysis further showed this redirection to be systematically structured by model scale. Beyond establishing nudge susceptibility as a behavioural property of agentic AI, the study positions interface design as a governance concern for organizations that delegate decisions to autonomous agents.

We examine the effectiveness of work requirements as a screening device in the Supplemental Nutrition Assistance Program (SNAP). Work requirements for "able-bodied adults without dependents" were suspended after the Great Recession and gradually reinstated across counties and states in the 2010s. Using linked administrative SNAP and employment data from five states and a triple-differences design, we find that work requirements reduce SNAP participation by seven percent without increasing labor supply and disproportionately screen out low-income individuals. We develop a welfare framework to interpret these results and find that the social costs of work requirements exceed budget savings.

This paper documents how inventory dynamics vary across levels of development and how they respond to real and financial shocks. Using a balanced panel of 72 small open economies over 1993-2022, spanning advanced to low-income economies, we show that inventories are strongly procyclical in advanced economies but become acyclical at lower income levels, while volatility and persistence rise as development declines. Using local projections, we estimate inventory responses to a long-run productivity shock and to a financial shock identified from the spread between U.S. corporate bond yields and Treasury yields. Productivity shocks generate temporary inventory accumulation across all groups, with smoother adjustment in advanced economies and more irregular cycles in emerging and developing economies. Financial shocks trigger decumulation followed by a rebound in higher-income groups, but a delayed, unreversed decline in low-income economies.

The Poisson pseudo-maximum likelihood (PPML) estimator is widely used for estimating bilateral gravity equations. Its consistency requires only a correctly specified conditional mean. Conventional inference, however, also requires finite-variance scores and Gaussian limits. We show that these conditions fail: bilateral flows are Pareto-tailed, PPML scores have a stable limit under a structural gravity data-generating process, and sandwich confidence intervals are too narrow. We retain PPML for point estimation but replace sandwich inference with an m-out-of-n bootstrap robust to heavy tails. Across three bilateral data settings, the correction is large and overturns conventionally significant gravity coefficients.

In-context learning (ICL) enables large language model (LLM) agents to improve decisions using interaction history, yet it remains unclear whether such improvement reflects refined internal reasoning or mere extrapolation of statistical patterns. To disentangle these mechanisms, we study LLM agents in multi-agent incomplete-information games that require recursive belief reasoning. By constructing a public goods game and manipulating the statistical structure of historical feedback, we evaluate decision quality against a history-independent rational expectations equilibrium (REE) benchmark. Our experiments reveal that when historical statistical patterns are disrupted, the benefits of longer context largely vanish, degrading decision quality to the no-context baseline in a way sharply amplified by stronger strategic interdependence. These results suggest that, in such strategic environments, ICL behavior is more consistent with statistical extrapolation than with strategic reasoning. Our work extends the mechanistic study of ICL to strategic multi-agent settings, introduces REE as a diagnostic tool for distinguishing reasoning from extrapolation, and provides a reusable framework for probing the boundaries of LLM reasoning in recursive belief tasks.

Discrete choice model specification is a time-consuming task in which modellers often specify and estimate multiple models while balancing goodness-of-fit, parsimony, and behavioural plausibility. We present Delphos, a multitask reinforcement learning framework that learns transferable specification strategies across transport choice datasets. Delphos frames model specification as a sequential decision-making problem in which it applies a sequence of modelling actions and receives feedback from an estimation environment based on model performance and convergence. To transfer modelling decisions across datasets with different sets of variables, Delphos represents utility specifications as sets of modelling terms using a DeepSet-Q architecture, allowing a shared specification policy to learn across multiple datasets. Trained on nine transport choice datasets, Delphos consistently outperforms independently trained single-task agents, indicating that sharing modelling experience improves learning efficiency and helps identify promising sequences of modelling decisions with fewer unsuccessful estimation attempts. When applied without further training to the unseen Swissmetro and Decisions datasets, the same agent identifies competitive specifications in less than 20 minutes on a standard CPU. It achieves a higher log-likelihood per observation than the VNS metaheuristic on Swissmetro and performance comparable to a published MNL specification developed by expert modellers on Decisions. These findings show that accumulating and reusing modelling experience enables Delphos to function as an intelligent assistant for discrete choice model specification. It reduces manual trial-and-error while allowing modellers to retain control over model diagnosis, refinement, and final selection.

We perform a detailed study of a principal–agent problem in a continuous time version of the celebrated Holmström–Milgrom model (Econometrica 55 (2), 1987) where we add limited liability for the Agent. We develop a probabilistic methodology to prove that the Principal's value function is the unique bounded classical solution to a fully nonlinear and fully degenerate partial differential equation (PDE) with Cauchy-Dirichlet boundary conditions on $[0,T]\times[0,\infty)$. Indeed, we also prove infinite continuous differentiability of the solution in the interior of the domain. The strength of our regularity result is such that we can ensure existence of optimal controls in strong form—a rare occurrence in dynamic contracting—and we obtain fine properties of the optimal control map, including a characterisation via a further nonlinear degenerate PDE.

Qualcomm's No License, No Chips policy let it levy a royalty surcharge on every handset, whether or not it used a Qualcomm modem chip. In FTC v. Qualcomm, the Ninth Circuit reversed the district court after accepting Qualcomm's argument that, because the surcharge did not vary with the chip, it was "chip neutral" and left handset makers' choices undistorted. I develop an equilibrium model of the modem chip market and show the defense to be wrong: the surcharge's facial neutrality does not imply economic neutrality. For per handset surcharges, a surcharge and an equal government tax affect the rival's pricing identically, but not Qualcomm's: a tax is remitted to the Treasury, whereas Qualcomm collects the surcharge – including on handsets using a rival's chip. Raising its own price therefore yields Qualcomm a smaller gain under the surcharge (the surcharge it collects on the demand diverted to the rival) than under the tax (the tax it avoids on its own lost sales), because the diversion ratio is less than one. Under the very conditions that would make a tax chip neutral, the surcharge raises the rival's all in price by strictly more than Qualcomm's – and, under symmetric demand, lowers its output by more as well – tilting handset makers toward Qualcomm. For ad valorem surcharges, the defense fails for a different reason: even a non discriminatory tax is generically chip neutral only if the FRAND royalty rate is zero, so the argument's premise itself does not hold. I also analyze discriminatory surcharges.

Large language models (LLMs) now serve as conversational shopping assistants on platforms that also sell advertising. These AI agents face a conflict of duty. They advise consumers who rely on their judgment, yet are deployed by platforms that benefit when sponsored listings are chosen. Sponsorship disclosures, designed to allow consumers to penalize paid placements, now reach the AI agent rather than the consumer, and the agent's evaluation of them is hidden from the consumer. Drawing on the fiduciary concept of conflict of duty, we argue that an agent's evaluation of a sponsored listing should not depend on which party deployed it. In controlled choice experiments, we manipulate assigned roles in the system prompt to name either a traveler or a booking platform as the agent's principal. Platform delegation significantly attenuates the penalty that agents apply to sponsored listings and weakens the skepticism that disclosure triggers in their reasoning traces. We replicate out findings across LLMs and reasoning depths. A second study decomposes the disclosure label and shows that the divergence between the two delegates widens significantly when the paid placement is attributed to the platform. Stricter terminology ("Sponsored" instead of "Promoted") lowers choice of paid listings but does not close this gap when the platform is named. The findings show that disclosure mandates designed for human consumers cannot by themselves protect consumers in AI-mediated commerce.

Scientific progress is a key driver of economic growth and prosperity. There is great excitement - but also concerns - about the impacts of AI on science, but so far little data. We provide early insights on this from three data sources: a sample of 15 million Gemini interactions, an inventory of over 2,600 specialized AI models across disciplines, and a survey of over 600 scientists. We map these data to a new taxonomy of scientific tasks to study how scientists are using AI. Four main findings emerge. First, we find broad adoption and coverage: scientists use AI more than most other occupations. Specialized AI models have broad disciplinary coverage and are highly cited. Nearly half of the scientists surveyed report using some form of AI every day. Second, we document evidence that LLMs (proxied through Gemini usage) and specialized models act as complements– LLMs are used for general analysis, coding, and manuscript preparation, while specialized models provide domain-specific predictions, data generation and classification. Third, scientists report large productivity gains from using AI: a saving of nearly 7 hours per week, time which is primarily re-invested in more research. Finally, we show that AI is already changing the scientific process. As some stages of scientific research become easier, bottlenecks shift downstream. Scientists report an increased backlog of untested hypotheses and substantial demand for output verification. Our findings suggest that AI holds significant potential to increase scientific productivity. However, as with other sectors, its ultimate impact will be governed by complex task interdependencies and investment into the elimination of emerging bottlenecks.

Telework has expanded rapidly, and understanding the behaviors employees enact under it has become correspondingly important. This study develops an integrative conceptual framework specifying the multidimensional nature of telework behavior. A systematic literature review following PRISMA identified 114 review articles, which were analyzed using constructivist grounded theory. Six behavioral dimensions were identified, covering performance, communication, environmental, task, policy, and well-being conduct. Antecedents were grouped into individual factors, job characteristics, organizational norms, technological factors, and work environment factors. Outcomes were grouped into job satisfaction, productivity, turnover, health and well-being, work-life balance, and social isolation. Three contextual moderators were identified, namely telework modality, telework preference, and cultural and national context. Fourteen propositions link these categories and are advanced as testable claims rather than established findings, since they are derived from published review evidence and have not been empirically tested. The framework proposes telework behavior as the construct through which the conditions of remote work are translated into employee and organizational outcomes, and it specifies a content domain from which measures of the construct can be developed. It also indicates where organizations can direct policy, training, and intervention.

Green economic complexity provides a generalizable framework for examining countries' productive capabilities in a defined product set. We apply this framework to AI-enabling goods within the full product space, linking current specialization with adjacent diversification opportunities. Using BACI exports for 2007-2023 and 103 AI-enabling goods, we measure complexity-weighted specialization (AECI), product-level adjacent opportunities (AIAP), and average complexity-weighted relatedness of remaining candidates (AECP). In 2023, Japan leads AECI, while China leads AECP; portfolio breadth accounts for much of the variation in raw AECI. Initial raw potential is positively associated with subsequent changes in the AI-enabling export share, but its associations with changes in AECI and specialization counts are not statistically significant at the 5% level. Our contribution is a trade-based assessment of AI-enabling productive capabilities and related opportunities. The results and public dashboard provide a preliminary complement to publication and patent indicators, not a comprehensive measure of national AI performance or a validated forecast of diversification.

We study coordination games in which every action requires planning and preparation. Before players act, they can revise their plans based on partially revealing information about their adversary's preparations. Precise information enables agents to screen for cooperation, selecting the payoff-dominant equilibrium either via small exogenous trembles or via payoff uncertainty. Across scenarios, we emphasize that decomposing an action into (i) preparation and (ii) execution allows us to analyze simultaneous-move games where players partially observe each other's contemporaneous actions.

As artificial intelligence's capabilities improve, it is increasingly viewed as a general scientific method. But how true are these claims? Does AI outperform all techniques, or only some, and how is this changing? To assess the claims, we assemble a corpus of 2,507 head-to-head comparisons between AI and other scientific analysis techniques across 27 scientific disciplines from papers published between 2000 and early 2025. We find a profound dichotomy. Relative to traditional statistics, AI often outperforms, but at a significantly higher computational cost. But there are also nearly a quarter of cases where AI is both more expensive and performs worse than traditional statistical techniques and this fraction has been stable for a decade. Relative to scientific computing, AI often underperforms, but at lower computational cost. This has begun to change: since 2020, AI's performance against scientific computing has notably strengthened and it now outperforms on more than half of comparisons. These patterns suggest that AI is therefore not a universal replacement for existing methods, but rather a valuable – and improving – part of a new AI-enabled scientific frontier.

Numbers this large invite a defensive reflex: reach for the reassuring figure and move on. By the most widely cited measure, the World Bank's extreme-poverty line of \$3.00 per day (2021 PPP), approximately 847 million people, or 10.4% of the world's population, lived in poverty in 2024. That figure is accurate as measured. It is also, by design, a floor: a threshold built to mark bare survival in the poorest economies, not to describe what it takes for a family anywhere to live with basic security. The central argument of this report is that the reassurance offered by that single line is largely an artifact of how we chose to measure. Held to the \$3.00 floor, roughly 847 million people are poor; held to the standards of their own societies, the number is several times larger, and even the most cautious, honest count exceeds a billion. Measured against income standards appropriate to each country's level of development, some 3.7 billion people fall below the World Bank's \$8.30/day upper-middle-income reference line, roughly 1.1 billion are multidimensionally poor, and around 2.3 billion are food insecure. These are not competing errors but answers to different questions. This report does not dispute the official data. It takes those figures as accurate and asks what they do and do not capture. It disaggregates by geography, age, and gender and supplements the monetary count with multidimensional measures of food security, working poverty, social protection, and relative poverty. It also documents a data-quality problem that runs in one direction: survey coverage is weaker in regions with the worst poverty, so official figures are far likelier to understate deprivation than to overstate it.

We model the economics of recursive self-improvement (RSI) and assess its plausibility and impacts. First, we build a sequence of increasingly rich models of AI progress to highlight the feedback loops behind RSI. We represent our models as directed graphs and show that net acceleration in AI capabilities depends on the product of elasticities across each feedback loop. Second, we distinguish between "narrow" and "broad" AI capabilities, capturing the possibility that AI systems improve narrowly at optimizing AI R&D benchmarks without improving at broader economically valuable tasks. Third, we document existing estimates of key parameters and provide a wish list of empirical objects that AI companies can measure and feasibly share publicly. Finally, we calibrate the model with existing data. A back-of-the-envelope calculation suggests that feedback loops are not currently strong enough to generate a self-sustaining acceleration, though they appear to be strengthening. We conclude by assessing the plausibility and implications of such an acceleration.