每周一14：00自动抓取最新NBER工作论文，中文内容由Kimi翻译。欢迎关注微信公众号：芾椿YA
(Updated on: 2026-01-12)
Nuclear Operations with a High Penetration of Renewables: The Case of France
Authors: Nicolas Astier, Frank A. Wolak
Abstract: Nuclear and intermittent renewables (wind and solar) are generally regarded as the only scalable technologies producing low-carbon electricity. However, the extent to which these technologies can co-exist in a reliable power system depends on whether nuclear units can adjust their operations to renewable output fluctuations. Using hourly data from the French power system, we find that nuclear units are operated quite flexibly, and that the foregone energy production due to “load following” actions (relative to the counterfactual of operating at full capacity during load following events) is currently limited. However, we find that an additional load following event is associated with a slightly higher likelihood of a unit failure. We also find that unit-level minimum output constraints are binding more frequently as system-wide renewable generation increases, especially so for units most exposed to solar generation. In 2024, hours during which available nuclear flexibility was exhausted are associated with non-positive hourly day-ahead prices.
摘要: 核能和间歇性可再生能源（风能和太阳能）通常被认为是唯一能够产生低碳电力的可扩展技术。然而，这些技术在可靠电力系统中共存的程度取决于核能装置是否能够调整其运行以适应可再生能源输出的波动。利用法国电力系统的小时数据，我们发现核能装置的运行相当灵活，并且由于“负荷跟踪”行动而放弃的能源生产（相对于在负荷跟踪事件期间以全容量运行的反事实情况）目前是有限的。然而，我们发现额外的负荷跟踪事件与装置故障的可能性略高有关。我们还发现，随着全系统可再生能源发电的增加，装置级别的最小输出限制更频繁地生效，尤其是对于那些最暴露于太阳能发电的装置。在2024年，可用核灵活性耗尽的小时与非正的小时日前价格相关联。
Global Stratification Economics (GSE): A Primer - Definitions and Research Implications
Authors: Alan A. Aja, Anita Alves Pena, Mary Lopez
Abstract: We coin the term Global Stratification Economics (GSE) to extend Stratification Economics (SE) to international contexts where mechanisms emerge that can be identified as (1) universal but matched to positioning along a country’s development path or (2) context-specific and historically dependent in ways different from the United States and other international environments. Building on SE’s core insight that groups compete over scarce resources and that dominant groups shape rules to preserve advantage, we define GSE as an approach that foregrounds historical genesis (e.g., settler colonialism, slavery, caste, extractive governance, military intervention), local institutions that structure dominance and resistance, and transnational mechanisms (e.g., imported stratification and the lateral mobility hypothesis) that link origin- and destination-country hierarchies. Drawing on evidence from Latin America and the Caribbean, Africa, South and Southeast Asia, and the global macroeconomy, we synthesize research on intergenerational mobility, wealth and asset inequality, labor-market segmentation, colorism, caste and ethnicity, gendered and intersectional constraints, spatial inequality and property regimes, and global value chains. We conclude by outlining how GSE reshapes measurement and international policy evaluation by centering group-differentiated outcomes, institutional transformation, and pillars of recognition, redistribution, and repair and by offering a checklist for researchers to apply GSE.
摘要: 我们创造了“全球分层经济学”（GSE）这一术语，将分层经济学（SE）扩展到国际背景中，其中出现了可以被识别为（1）普遍存在但与一个国家的发展路径相匹配的机制，或（2）特定于特定环境且在历史上与美国和其他国际环境不同的机制。基于SE的核心观点，即群体之间为争夺稀缺资源而竞争，主导群体塑造规则以保持优势，我们将GSE定义为一种方法，强调历史起源（例如，殖民主义、奴隶制、种姓制度、榨取性治理、军事干预）、构建主导和抵抗的地方制度，以及跨国机制（例如，进口分层和横向流动性假设），这些机制将原籍国和目的地国的等级制度联系起来。我们借鉴了来自拉丁美洲和加勒比地区、非洲、南亚和东南亚以及全球宏观经济的证据，综合了关于代际流动性、财富和资产不平等、劳动力市场分割、肤色主义、种姓和种族、性别和交叉性限制、空间不平等和财产制度以及全球价值链的研究。我们最后概述了GSE如何通过以群体差异化结果、制度转型以及认可、再分配和修复的支柱为中心，重塑测量和国际政策评估，并为研究人员提供应用GSE的清单。
Brexit and UK Migration: Germany-Benchmarked Evidence on Structural and Reallocation Shifts
Authors: Assaf Razin
Abstract: This paper examines how Brexit reshaped the volume and composition of UK migration flows. To address the inherent endogeneity of migration responses to political change, the study employs a Difference-in-Differences design that benchmarks the United Kingdom against Germany, a stable EU member unaffected by Brexit-related institutional shocks. Treating the 2016 referendum and the 2021 end of free movement as sequential shocks, the results show that migration volume increased after both shocks, but through different channels: the referendum reduced emigration relative to Germany, while the post-2021 points-based system triggered a reallocation of inflows away from EU free-movement migrants toward non-EU migrants from the rest of the world, with a markedly stronger skill-selective profile. The paper analyzes how Brexit reshaped migration flows into and out of the United Kingdom by exploiting the 2016 referendum and the 2021 termination of free movement as sequential institutional shocks. Using a Difference-in-Differences design with Germany as a stable EU benchmark, the results show that Brexit acted as a structural regime shock, not a border-tightening event. Net migration into the UK rose after both shocks—first due to a sharp post-referendum decline in emigration relative to Germany, and later due to a substantial expansion of non-EU immigration under the new points-based system. Immigration patterns exhibit a clear regime shift, with EU inflows contracting sharply after 2016 and globally sourced, skill-selective inflows rising after 2021. In the later period, a modest but notable increase in UK-born emigration relative to Germany also emerges, reflecting frictions in post-Brexit mobility. Overall, Brexit reoriented—rather than reduced—UK migration flows, transforming both their scale and composition.
摘要: 本文研究了脱欧如何重塑了英国移民流动的规模和构成。为了解决移民对政治变化反应的内在内生性问题，研究采用了双重差分设计，将英国与德国进行对比，德国是一个稳定的欧盟成员国，未受脱欧相关制度冲击的影响。将2016年公投和2021年自由流动结束视为连续冲击，结果显示两次冲击后移民流动量均有所增加，但通过不同渠道：公投相对于德国减少了外迁，而2021年后的基于积分的系统促使移民流入从欧盟自由流动移民转向来自世界其他地区的非欧盟移民，并且具有明显更强的技能选择性特征。本文通过利用2016年公投和2021年自由流动终止作为连续的制度冲击，分析了脱欧如何重塑英国的移民流入和流出。使用德国作为稳定欧盟基准的双重差分设计，结果显示脱欧起到了结构性制度冲击的作用，而非边境收紧事件。两次冲击后英国的净移民均有所上升——首先是因为相对于德国的外迁在公投后急剧下降，后来是因为在新的基于积分的制度下非欧盟移民大幅增加。移民模式显示出明显的制度转变，欧盟流入在2016年后急剧收缩，全球来源、技能选择性的流入在2021年后上升。在后期，与德国相比，英国出生的外迁也出现了适度但值得注意的增加，反映了脱欧后流动性的摩擦。总体而言，脱欧重新定位了英国的移民流动，而非减少，改变了它们的规模和构成。
State and Time-Dependent Pricing
Authors: Philip Bunn, Nicholas Bloom, Craig Menzies, Paul Mizen, Gregory Thwaites, Ivan Yotzov
Abstract: We present new evidence on how firms set prices using direct questions from a large, economy-wide survey of UK firms. Since 2023, 54% of firms report setting prices in a state-dependent manner, as opposed to changing prices at fixed intervals. In contrast, 44% of firms used state-dependent pricing in 2019. Smaller firms, those with a higher share of non-labour costs, and those reporting higher subjective uncertainty around sales and prices are more likely to be state-dependent. We then analyse the implications of price-setting behaviour for inflation dynamics. State-dependent firms experienced a sharper increase in price growth over 2022-2023, and also a faster subsequent decline. Using evidence from a randomised survey experiment, firm-level forecast errors, and local projections, we show that prices of state-dependent firms respond faster to cost shocks. The difference between state-dependent and time-dependent firms is furthermore larger for bigger shocks, consistent with theoretical predictions.
摘要: 我们通过一项针对英国企业的大型经济范围调查，提出了关于企业如何设定价格的新证据。自2023年以来，有54%的企业报告称以状态依赖的方式设定价格，而不是在固定间隔内改变价格。相比之下，2019年有44%的企业使用了状态依赖定价。规模较小的企业、非劳动力成本占比较高的企业，以及报告称销售和价格主观不确定性较高的企业更可能采用状态依赖定价。然后，我们分析了定价行为对通货膨胀动态的影响。状态依赖型企业在2022-2023年期间经历了价格增长的更急剧上升，随后也出现了更快的下降。利用随机调查实验、企业层面的预测误差和局部投影的证据，我们展示了状态依赖型企业的价格对成本冲击的反应更快。此外，对于更大的冲击，状态依赖型企业与时间依赖型企业之间的差异更大，这与理论预测一致。
Weighing the Impacts of GLP-1s: Quasi-Experimental Evidence From Provider Adoption
Authors: Sam Bock, Jasmin Moshfegh, Jonathan Zhang
Abstract: The arrival of GLP-1 medications has been described as one of the most important health care innovations in recent years. We provide large-scale quasi-experimental evidence on their real-world impacts by exploiting variation in the eventual prescribing propensities of patients’ pre-existing primary care providers. Using a panel intent-to-treat design, we compare outcomes for 1.4 million diabetic or obese veterans based on their 2018 provider’s eventual propensity to adopt GLP-1s, leveraging comprehensive electronic health records and biomarker data from the Veterans Health Administration, a setting with minimal insurance attrition and low-cost access to these drugs. Patients whose providers become higher propensity adopters experience substantial improvements in glycemic control and clinically meaningful weight loss; our treatment-on-the-treated estimates closely match estimates from clinical trials. Despite these metabolic benefits, we find no statistically significant effects on emergency department utilization, mental health and substance use outcomes, or non–GLP-1 medical spending through 2024.
摘要: GLP-1类药物的到来被描述为近年来最重要的医疗保健创新之一。我们通过利用患者现有的基层保健提供者最终开处方倾向的变化，提供了关于它们在现实世界中影响的大规模准实验证据。采用面板意向治疗设计，我们比较了基于2018年提供者最终采用GLP-1类药物倾向的140万糖尿病或肥胖退伍军人的结果，利用退伍军人健康管理局的全面电子健康记录和生物标志物数据，这是一个保险流失最小且这些药物低成本获取的环境。那些提供者成为更高倾向采纳者的患者的血糖控制有了显著改善，并且临床上有意义的体重减轻；我们对治疗后的治疗估计与临床试验的估计非常接近。尽管有这些代谢益处，我们发现到2024年，对急诊部门使用、心理健康和物质使用结果，或非GLP-1医疗支出没有统计学上显著的影响。
Experiences, Expectations, and Asset Prices
Authors: Stefan Nagel
Abstract: In models with subjective beliefs, the dynamics of investor beliefs are central to explaining asset price movements. Allowing subjective beliefs to deviate from full-information rational expectations requires a model of how investors form beliefs. I review recent research that explores experience-based belief formation. The experience-based model closely resembles Bayesian updating but allows individuals to overweight data observed during their own lifetimes when forming beliefs about the parameters of the data-generating process. Because individuals of different age learn from different histories, the model predicts heterogeneity between age groups that evolves over time with the path of realized data individuals have experienced, closely mirroring the age-related variation observed in survey microdata. At the aggregate level, experience-based expectations formation can explain asset price dynamics. In the stock market, learning from experience about long-run cash flow growth generates valuation cycles and return predictability. Experience-based formation of long-term inflation expectations can explain secular changes of real interest rates. I conclude by outlining directions for future research.
摘要: 在具有主观信念的模型中，投资者信念的动态是解释资产价格变动的核心。允许主观信念偏离完全信息下的理性预期，需要一个模型来说明投资者如何形成信念。我回顾了最近的研究，这些研究探讨了基于经验的信念形成。基于经验的模型与贝叶斯更新非常相似，但允许个体在形成关于数据生成过程参数的信念时，过分重视自己一生中观察到的数据。由于不同年龄的个体从不同的历史中学习，模型预测了随时间演变的年龄组之间的异质性，这些异质性与个体经历的实现数据路径密切相关，与调查微观数据中观察到的年龄相关变化非常相似。在总体层面上，基于经验的预期形成可以解释资产价格动态。在股票市场中，从经验中学习长期现金流增长会产生估值周期和回报可预测性。基于经验形成的长期通胀预期可以解释实际利率的长期变化。最后，我概述了未来研究的方向。
Authors: Augustin Landier, Parinitha R. Sastry, David Thesmar
Abstract: Do investors care about why a firm behaves responsibly, over and above what the firm does and how much cash it earns? In our main experiment, participants value shares in companies that reduce pollution relative to peers, while CEOs articulate the company’s intentions: prosocial or for-profit. Investors are willing to pay more for shares in a company that takes a given prosocial action if it is motivated by social concern rather than profit maximization. This “intention premium” is distinct from the valuation effects of cash flows and environmental externalities. Respondents who are deontological, politically liberal, female, or less individualistic value the company’s intent the most. Additionally, and somewhat surprisingly, respondents view favorably firms that embrace both profit and social objectives, without ranking them. Overall, our findings temper claims that when there is no financial trade-off between doing good and doing well, shareholder value and stakeholder value maximization are equivalent.
摘要: 投资者是否关心公司负责任行为背后的原因，而不仅仅是公司做了什么以及赚了多少钱？在我们的主要实验中，参与者对那些相对于同行减少污染的公司股票给予更高的估值，而首席执行官则阐述了公司的意图：利他主义还是盈利。如果一家公司采取特定的利他行动是出于对社会关切的动机，而不是为了利润最大化，投资者愿意为该公司的股票支付更高的价格。这种“意图溢价”与现金流和环境外部性的估值效应是不同的。那些持有义务论观点、政治上较为自由、女性或不那么个人主义的受访者最看重公司的意图。此外，令人有些惊讶的是，受访者对同时拥抱盈利和社会目标的公司持积极看法，并没有对它们进行排名。总体而言，我们的发现缓和了这样一种说法，即当做好事与做好事之间没有财务权衡时，股东价值和利益相关者价值最大化是等价的。
The Impact of Preschool Entry Age on Children’s Behavioral and Developmental Health in Medicaid
Authors: Maya Rossin-Slater, Adrienne Sabety, Aileen Wu
Abstract: We find that preschools facilitate early diagnosis and treatment of conditions that can hinder learning. Low-income children born shortly before their state’s school-entry cutoff date are 16.9, 9.3, and 14.8 percent more likely to be diagnosed with Attention Deficit Hyperactivity Disorder, a speech or language disorder, and a hearing or vision condition at ages three and four, compared to children born after the cutoff. They are also more likely to receive downstream services. Findings emphasize the role of earlier and longer exposure to public preschool in driving diagnostic gaps previously attributed to elementary school-entry and within-grade peer comparisons.
摘要: 我们发现，学前班有助于早期诊断和治疗可能妨碍学习的条件。与在截止日期后出生的儿童相比，低收入家庭中在他们所在州入学截止日期前不久出生的儿童，在三、四岁时被诊断出患有注意力缺陷多动障碍、言语或语言障碍以及听力或视力问题的可能性分别高出16.9%、9.3%和14.8%。他们也更有可能获得后续服务。这些发现强调了更早、更长时间接触公立学前班在推动之前归因于小学入学和同年级同伴比较的诊断差距方面的作用。
Do GLP-1 Medications Pay for Themselves?
Authors: Coady Wing, Sih-Ting Cai, Daniel W. Sacks, Kosali I. Simon
Abstract: Glucagon-like peptide-1 receptor agonists (GLP-1s) represent a major improvement in treatment of diabetes, obesity, and cardiovascular risk reduction, but they are also among the most expensive drugs in widespread use and the subject of significant policy debate. The high price of these drugs may overstate their net cost if the health improvements they produce lead to reduced downstream health care use and medical spending, that is, cost offsets. We estimate such offsets using insurance claims data, examining the effects of GLP-1 initiation on subsequent GLP-1 use and spending, and on other non-GLP-1 spending. We use a stacked difference-in-differences design, comparing patients initiating GLP-1 medication to not-yet-treated controls who initiate GLP-1s several months or years later, allowing us to control for underlying time trends and baseline characteristics. Overall, we do not find a reduction in downstream medical spending. Although GLP-1 initiation reduces spending on other diabetes medications, total non-GLP-1 spending increases, driven by higher outpatient health care use; GLP- 1 drug spending rises mechanically. For health care payers, the relevant cost of GLP-1 initiation therefore extends beyond the sticker price of the drug. We find similar results across subgroups of GLP1 initiators including those with prior cardiovascular disease and those without diabetes (consistent with obesity indication). Our main results examine spending responses over the first year after initiation. However, we also estimate longer run effects in a smaller sample and find no cost offsets even five years after GLP-1 initiation. Taken together, these results suggest that payers facing the costs of GLP-1 coverage are unlikely to see large savings from reduced spending on other care. If GLP-1 therapies ultimately yield cost savings, they are likely to occur only over longer horizons or through non-medical channels.
摘要: 胰高血糖素样肽-1受体激动剂（GLP-1s）在糖尿病、肥胖症和降低心血管风险的治疗方面代表了重大进步，但它们也是广泛使用中最为昂贵的药物之一，并且是重大政策辩论的主题。这些药物的高昂价格可能会夸大它们的净成本，如果它们带来的健康改善导致下游医疗使用和医疗支出的减少，即成本抵消。我们使用保险索赔数据来估算这种抵消效应，检查GLP-1启动对随后GLP-1使用和支出的影响，以及其他非GLP-1支出的影响。我们使用堆叠差异-差异设计，比较开始使用GLP-1药物的患者与几个月或几年后才开始使用GLP-1s的未治疗对照组，使我们能够控制潜在的时间趋势和基线特征。总体而言，我们没有发现下游医疗支出的减少。尽管GLP-1启动减少了其他糖尿病药物的支出，但由于门诊医疗使用增加，总非GLP-1支出增加；GLP-1药物支出机械性上升。因此，对于医疗支付者来说，GLP-1启动的相关成本超出了药物的标价。我们在GLP1启动者的各个亚组中发现了类似的结果，包括那些有心血管疾病史的人和那些没有糖尿病的人（与肥胖症适应症一致）。我们的主要结果检查了启动后第一年内的支出反应。然而，我们也在较小的样本中估计了长期效应，发现即使在GLP-1启动五年后也没有成本抵消。综合这些结果，它们表明，面临GLP-1覆盖成本的支付者不太可能从减少其他护理支出中看到大量节省。如果GLP-1疗法最终产生成本节省，它们可能只会在更长的时间范围内或通过非医疗渠道发生。
Firm Emissions and Credit Allocation
Authors: Grace Weishi Gu, Galina Hale, Bhavyaa Sharma, Jinhong Wu
Abstract: Do banks help or hamper green transition? To answer this question, we analyze the dynamics of bank lending to firms in the US, EU, and separately Denmark in relation to the borrowers’ emissions of CO2. We evaluate the allocation of bank loans across industries and within industries across firms, allowing for heterogeneity of firm emissions and changes in these emissions. To facilitate green transition, bank lending needs to flow to greener and greening firms, but not out of high-emission industries that need funding to transition to cleaner production methods. Using syndicated loan data, we find that for US borrowers, bank lending was likely hampering green transition, while in the EU bank lending is more likely to facilitate it. Zooming in on Denmark, for which we have data on the full universe of firms and banks, we find more significant credit reallocation to greener firms, especially within industries. However, the reallocation of funds to green firms is, to a large extent, a byproduct of green firms becoming bigger. We do not find any evidence consistent with banks active stewardship of green transition.
摘要: 银行是帮助还是阻碍绿色转型？为了回答这个问题，我们分析了美国、欧盟以及单独的丹麦银行对企业的贷款动态，这些贷款与借款者的二氧化碳排放量有关。我们评估了银行贷款在不同行业以及同一行业内不同企业之间的分配，允许企业排放量的差异性以及这些排放量的变化。为了促进绿色转型，银行贷款需要流向更绿色和正在绿化的企业，但不是从需要资金转型到更清洁生产方法的高排放行业中流出。利用银团贷款数据，我们发现对于美国借款人来说，银行贷款可能阻碍了绿色转型，而在欧盟，银行贷款更有可能促进它。当我们聚焦于丹麦，我们拥有关于所有企业和银行的完整数据，我们发现信贷更显著地重新分配给更绿色的企业，尤其是在行业内部。然而，资金向绿色企业的重新分配在很大程度上是绿色企业变得更大的一个副产品。我们没有发现任何与银行积极管理绿色转型一致的证据。
Transformed Intermediation: Credit Risk to NBFIs, Liquidity Risk to Banks
Authors: Viral V. Acharya, Nicola Cetorelli, Bruce Tuckman
Abstract: We argue that the rapid asset growth of nonbank financial intermediaries (NBFIs) relative to banks is the outcome of transformations of risks between banks and NBFIs that increase the interconnectedness of the two sectors. These transformations are consistent with avoiding tighter, post-GFC bank regulation while harnessing the funding and liquidity advantages of bank deposit franchises and access to safety nets. Specifically, we show that banks fund NBFIs through senior loans and credit lines, which NBFIs use for acquiring junior credit claims, warehouse financing, and liquidity management. We empirically demonstrate that shocks experienced by NBFIs spillover to the banks that provide them with credit lines, particularly in times of stress. We then describe a number of policy approaches that are consistent with our transformation view and conclude with suggestions for future research.
摘要: 我们认为，非银行金融中介机构（NBFIs）相对于银行的资产快速增长是银行与NBFIs之间风险转换的结果，这种转换增加了两个部门之间的相互联系。这些转换与避免更严格的、后全球金融危机（GFC）时期的银行监管相一致，同时利用银行存款特许经营的资金和流动性优势以及获得安全网的途径。具体来说，我们展示了银行通过高级贷款和信贷额度为NBFIs提供资金，NBFIs则利用这些资金获取次级信贷债权、仓储融资和流动性管理。我们从实证上证明了NBFIs所经历的冲击会溢出到为它们提供信贷额度的银行，特别是在压力时期。然后，我们描述了一些与我们转换观点一致的政策方法，并以对未来研究的建议作为结论。
Platform Work: Evidence from Drivers in India, Indonesia, and Kenya
Authors: Achyuta Adhvaryu, Martin Atela, Valentina Brailovskaya, Priyanka Dua, Jenny Susan John, Pratibha Joshi, Rivandra Royono
Abstract: Using surveys and administrative data from representative samples of drivers working on three leading gig platforms in India, Indonesia, and Kenya, we document the composition, economic experiences, and labor market trajectories of platform workers. Combining platform-based earnings with operating cost data, we estimate earnings net of costs (in PPP-adjusted terms) in each context. We find that the flexible nature of platform work enables drivers to work substantially more than the full-time equivalent, generating higher monthly net earnings than low-skill or casual employment, despite comparable or lower hourly net earnings relative to these outside options. Drivers who exit platform work in India and Indonesia do so to take up better-paying full-time positions. In contrast, Kenyan drivers often exit involuntarily, returning to offline driving with adverse financial consequences. One-third of drivers across countries rely on platform work to supplement earnings during emergencies or slow work periods, suggesting that platform work may play an important role as a financial safety net.
摘要: 利用来自印度、印度尼西亚和肯尼亚三个主要零工平台的代表性司机样本的调查和行政数据，我们记录了平台工人的构成、经济经历和劳动市场轨迹。结合基于平台的收入和运营成本数据，我们估算了在每个情境下扣除成本后的收入（以购买力平价调整）。我们发现，平台工作的灵活性使得司机能够比全职工作更长时间地工作，从而产生比低技能或临时工作更高的月净收入，尽管与这些外部选择相比，每小时的净收入相当或更低。在印度和印度尼西亚退出平台工作的司机这样做是为了获得更高薪酬的全职职位。相比之下，肯尼亚司机经常被迫退出，回到线下驾驶，带来不利的财务后果。各国三分之一的司机依赖平台工作来补充紧急情况或工作缓慢时期的收入，这表明平台工作可能在作为金融安全网方面发挥重要作用。
Re-Examining Geographic Variation in Health and Health Care
Authors: Amy Finkelstein, Matthew Gentzkow
Abstract: A large literature has documented widespread variation in health care spending per capita across areas of the United States without correspondingly better health outcomes. Recent work has used mover designs to estimate the causal impact of place on both health care spending and mortality. In this paper, we investigate whether places that increase health care spending also tend to be places that increase health. We find that they do not and discuss the implications.
摘要: 大量文献记录了美国各地区人均医疗保健支出的普遍差异，而这些差异并没有带来更好的健康结果。最近的研究利用搬迁者设计来估计地点对医疗保健支出和死亡率的因果影响。在本文中，我们调查了增加医疗保健支出的地方是否也倾向于增加健康。我们发现并非如此，并讨论了其含义。
Authors: Philip Oreopoulos, Oliver Keyes-Krysakowski, Deepak Agarwal
Abstract: Computer-assisted learning (CAL) platforms frequently underperform at scale not because the technology is ineffective, but because schools face substantial implementation frictions: teachers and administrators must overcome initial technical hurdles, reorganize instructional routines, manage competing scheduling pressures, and do so while uncertain about the technology’s effectiveness—conditions that often lead to low and unproductive student engagement. This study explores whether strengthening implementation structure can raise both the quantity and quality of CAL usage in 83 residential government middle schools in Uttar Pradesh, India and, in turn, learning gains. All schools had access to Khan Academy, but randomly selected treatment schools received on-the-ground lab-in-charges whose sole responsibility was to ensure high-fidelity implementation by securing reliable connectivity, simplifying student rostering, protecting weekly practice time, supervising in-class use, coordinating content with teachers, and monitoring progress. The intervention increased platform usage from 7.2 to 47.4 minutes per week. Mathematics achievement rose by almost half a standard deviation over 31 weeks, with gains broad-based across achievement levels and question difficulty. These results show that the central constraint on effective and scalable CAL is not technology or content, but the presence of organizational structures that ensure sustained, productive instructional use.
摘要: 计算机辅助学习（CAL）平台在大规模应用时常常表现不佳，并不是因为技术本身无效，而是因为学校在实施过程中面临大量的阻力：教师和行政人员必须克服最初的技术障碍，重新组织教学流程，管理竞争性的日程安排压力，同时还要在对技术效果不确定的情况下进行这些工作——这些条件常常导致学生参与度低且无效。本研究探讨了加强实施结构是否能够提高印度北方邦83所寄宿制政府中学CAL使用的数量和质量，进而提高学习收益。所有学校都能访问Khan Academy，但随机选择的处理组学校获得了实地实验室负责人，他们的唯一职责是确保通过确保可靠的连接性、简化学生名册管理、保护每周练习时间、监督课堂使用、与教师协调内容和监控进展来实现高保真度的实施。这项干预将平台使用量从每周7.2分钟提高到47.4分钟。在31周内，数学成绩提高了近半个标准差，而且成绩提高在各个成绩水平和问题难度上都是广泛的。这些结果表明，有效且可扩展的CAL的核心限制不是技术或内容，而是确保持续、有生产力的教学使用的组织结构的存在。
The first 1000 days and beyond: The process of child development.
Authors: Orazio Attanasio
Abstract: This paper reviews recent developments in the economics of human development, focusing on the early years of life as a critical period for shaping long-term outcomes. Early childhood development is inherently multidimensional: cognitive and socioemotional skills evolve dynamically and interact with health, nutrition, and environmental influences. Economists have contributed to this field by providing a conceptual unifying framework that highlights how key drivers of development reflect the choices of individuals operating under incentives and constraints. Within this framework, the paper emphasizes two central challenges: understanding the interactions among multiple dimensions of development and identifying causal links - particularly the effects of different inputs at different ages. Measurement issues are a recurring theme, given the difficulty of assessing young children and the need for comparability across contexts. The paper also stresses these issues’ policy relevance for poverty reduction and social mobility by discussing early childhood interventions in both developed and developing countries.
摘要: 本文回顾了人类发展经济学的最新进展，重点关注生命的早期阶段作为塑造长期结果的关键时期。儿童早期发展本质上是多维的：认知和社会情感技能动态发展，并与健康、营养和环境影响相互作用。经济学家通过提供一个概念性统一框架，强调发展的关键驱动因素如何反映在激励和约束下个体的选择，从而为这一领域做出了贡献。在此框架内，文章强调了两个中心挑战：理解发展多个维度之间的相互作用，并确定因果关系——特别是不同年龄阶段不同输入的影响。鉴于评估幼儿的困难和跨情境比较的需要，测量问题是一个反复出现的主题。文章还通过讨论发达国家和发展中国家的儿童早期干预措施，强调了这些问题对减贫和社会流动的政策相关性。
The Effects of Pension Reforms on Health Inequality in Italy
Authors: Agar Brugiavini, Raluca Elena Buia, Giacomo Pasini, Guglielmo Weber
Abstract: Using data from the Survey on Health, Ageing and Retirement in Europe (SHARE), we explore whether the recent pension reforms enacted in Italy had an impact on the inequality in health among individuals. We construct a set of indexes measuring several health outcomes (both physical and mental) and analyze their evolution along time (across SHARE waves) from 2004 to 2022, by relating them to the households’ income. While we can observe some differences in the health-income gradient between genders, we do not find evidence of a positive relationship between the introduction of the pension reforms and an increase in inequality.
摘要: 利用来自欧洲健康、老龄化和退休调查（SHARE）的数据，我们探讨了意大利最近实施的养老金改革是否对个体之间的健康不平等产生了影响。我们构建了一系列衡量多种健康结果（包括身体和心理）的指标，并分析了它们从2004年到2022年随时间（跨SHARE波次）的演变，并将它们与家庭收入联系起来。虽然我们可以观察到性别之间健康-收入梯度的一些差异，但我们没有发现养老金改革引入与不平等增加之间存在正向关系的证据。
Automation Experiments and Inequality
Authors: Seth Gordon Benzell, Kyle R. Myers
Abstract: Many experiments study the productivity effects of automation technologies such as generative algorithms. A key test in these experiments relates to inequality: does the technology increase output more for high- or low-skill workers? However, the theoretical content of this empirical test has been unclear. Here, we formalize a theory that describes the experimental effect of automation technologies on worker-level output and, therefore, inequality. Worker-level output depends on a task-level production function, and workers are heterogeneous in their task-level skills. Workers perform a task themselves or delegate it to the automation technology. The inequality effect of improved automation depends on the interaction of two factors: (i) the correlation in task-level skills across workers, and (ii) workers’ skills relative to the technology’s effective skill. In many cases we study, the inequality effect is non-monotonic — as technologies improve, inequality decreases then increases. The model and descriptive statistics of skill correlations generally suggest that the diversity of automation technologies will play an important role in the evolution of inequality.
摘要: 许多实验研究了自动化技术，如生成性算法，对生产率的影响。这些实验中的关键测试与不平等有关：这项技术是否更多地增加了高技能或低技能工人的产出？然而，这一实证测试的理论内容一直不清晰。在这里，我们提出了一个理论，描述了自动化技术对工人层面产出的影响，因此也影响了不平等。工人层面的产出取决于任务层面的生产函数，而工人在任务层面的技能是异质的。工人自己执行任务或将其委托给自动化技术。改进的自动化对不平等的影响取决于两个因素的相互作用：（i）工人之间任务层面技能的相关性，以及（ii）工人的技能相对于技术的有效技能。在我们研究的许多案例中，不平等的影响是非单调的——随着技术的进步，不平等先是减少然后增加。模型和技能相关性的描述性统计通常表明，自动化技术的多样性将在不平等的演变中发挥重要作用。
Price Discrimination with Costless Resale
Authors: Joshua S. Gans
Abstract: This paper shows that income effects create an endogenous barrier to arbitrage, allowing price discrimination to survive costless resale. A monopolist sells an indivisible good to consumers with heterogeneous incomes who can freely resell. When the good is strictly normal, a consumer’s reservation price to resell increases as the purchase price decreases—lower prices leave buyers wealthier and raise their valuation of the good. The monopolist exploits this by subsidising low-income consumers to raise their reservation prices to a target that high-income consumers must also pay. The optimal schedule increases dollar-for-dollar with income in the subsidised segment, weakly dominates uniform pricing, and achieves the first-best allocation when the entire market is served. We show the mechanism extends beyond income effects: low substitutability with market alternatives generates large reservation-price responses even when income sensitivity is modest. Sustaining discrimination requires market power at the individual level—consumer-specific quantity limits—not merely aggregate output restrictions. Extensions examine multiple monopolists and endogenous privacy choices.
摘要: 本文表明，收入效应创造了一个内生的套利障碍，允许价格歧视在无成本转售的情况下生存。一个垄断者向可以自由转售的、收入异质的消费者销售一个不可分割的商品。当商品是严格正常商品时，消费者转售的保留价格随着购买价格的下降而增加——较低的价格使买家更富有，提高了他们对商品的评价。垄断者通过补贴低收入消费者来利用这一点，提高他们的转售保留价格，使其达到高收入消费者也必须支付的目标。最优的价格表与受补贴部分的收入等比例增加，弱于统一定价，并在服务整个市场时实现第一最佳分配。我们展示了这一机制超越了收入效应：即使收入敏感性适中，与市场替代品的低替代性也会产生大的保留价格反应。维持歧视需要在个体层面上的市场力量——消费者特定的数量限制——而不仅仅是总体产出限制。扩展部分考察了多个垄断者和内生隐私选择。
Effective Protection with Global Value Chains
Authors: Emily J. Blanchard, Mitchell W. Boice, Robert C. Johnson
Abstract: The effective rate of protection (ERP) is a widely-used tool to evaluate the joint impact of input and output tariffs on the net protection offered to individual sectors. Nonetheless, the theoretical foundations underlying the traditional ERP are decades out of date. In this paper, we provide fresh guidance for policy analysts that seek to measure effective protection, with two core elements. First, we develop a new ERP concept, which measures how much tariffs shift the derived demand for real value added at the sector level. The ERP then has an intuitive interpretation: it is the effective subsidy to buyers of sectoral value added that replicates the impact of the tariff structure on demand. Second, we demonstrate how to compute our ERP index accounting for global value chain linkages. The result is an ERP for the GVC age, in which effective protection depends on the structure of value chains across countries and sectors, and the structure of all tariffs applied globally. We demonstrate the usefulness of our approach by analyzing how recent tariff changes have altered effective protection in the United States and the rest of the world.
摘要: 有效保护率（ERP）是一个广泛使用的工具，用于评估输入和输出关税对各个部门净保护提供的联合影响。尽管如此，传统ERP背后的理论基础已经过时了几十年。在本文中，我们为寻求衡量有效保护的政策分析师提供了新的指导，包含两个核心要素。首先，我们发展了一个新的ERP概念，它衡量关税在部门层面上对实际增加值的派生需求的转移程度。然后，ERP具有直观的解释：它是对部门增加值购买者的有效补贴，复制了关税结构对需求的影响。其次，我们展示了如何计算我们的ERP指数，同时考虑到全球价值链的联系。结果是一个适用于全球价值链时代的ERP，其中有效保护取决于跨国和跨部门的价值链结构，以及全球应用的所有关税结构。我们通过分析最近的关税变化如何改变了美国和世界其他地区的有效保护，展示了我们方法的实用性。
With or Without U? Binning Bias and the Causal Effects of Temperature Extremes
Authors: Benjamin Jones, Jacob Moscona, Benjamin A. Olken, Cristine von Dessauer
Abstract: Estimates of climate impacts show that extreme temperatures have large and wide-spread effects. To estimate these effects, a common approach counts days in different temperature ranges and considers how exposure to these distinct ‘bins’ affects outcomes. This often produces non-linear, U-shaped results, in which high and low temperatures have the largest effects. We show that non-linear approaches like these can generate spurious findings. Specifically, global warming induces trends in extreme temperature exposure that correlate mechanically with a location’s baseline temperature. Substantial bias emerges if trends in the outcome variable also correlate with baseline temperature for any reason. We demonstrate this problem theoretically, in simulations, and with real outcomes. We then develop solutions. In applications using US data, some results in the literature are unaffected by these corrections, while other results change substantially.
摘要: 气候影响的估计表明，极端温度具有重大且广泛的影响。为了估计这些影响，一种常见的方法是计算不同温度范围内的天数，并考虑暴露于这些不同的“区间”如何影响结果。这通常会产生非线性的、U形的结果，其中高和低温度具有最大的影响。我们展示这种非线性方法可能会产生虚假的发现。具体来说，全球变暖导致极端温度暴露的趋势与一个地区的基线温度机械相关。如果结果变量的趋势也因任何原因与基线温度相关，就会出现显著的偏差。我们从理论上、在模拟中以及用实际结果来展示这个问题。然后我们开发了解决方案。在使用美国数据的应用中，文献中的一些结果不受这些修正的影响，而其他结果则发生了重大变化。
The Impacts of Parole Supervision
Authors: Luke D. Brinkman, Andrew Jordan, Derek Neal
Abstract: We study the impacts of a reform to parole supervision in Illinois. The reform reduced the term of supervision for some parolees from 12 months to 6 months while many similar parolees continued to receive 12 months of supervision. We evaluate the impact of the reform using a standard difference-in-differences estimator, and we find clear evidence that the reform reduced prison re-entry rates. Sharp drops in rates of technical revocations drove these reductions. Rates of prison re-admissions linked to new crimes did not change. We merge data from Cook County Courts with our data on state prison admissions and releases, and we find no evidence that the reform increased crime rates among parolees. The reform reduced both the cost of incarceration and the cost of parolee supervision without creating harms to public safety.
摘要: 我们研究了伊利诺伊州对假释监督改革的影响。这项改革将一些假释犯人的监督期限从12个月减少到6个月，而许多类似的假释犯人继续接受12个月的监督。我们使用标准的双重差分估计器来评估改革的影响，我们发现有明确证据表明改革降低了监狱再入狱率。技术性撤销的急剧下降推动了这些降低。与新犯罪相关的监狱重新入狱率没有变化。我们将库克县法院的数据与我们关于州监狱入狱和释放的数据合并，我们发现没有证据表明改革增加了假释犯人的犯罪率。改革降低了监禁成本和假释犯人监督成本，而没有对公共安全造成损害。
How Globalization Unravels: A Ricardian Model of Endogenous Trade Policy
Authors: Jesús Fernández-Villaverde, Tomohide Mineyama, Dongho Song
Abstract: We study how uneven gains from globalization can endogenously generate protectionism as a political equilibrium. Using U.S. data, we document that regions more exposed to import competition display stronger opposition to globalization, especially among households with little financial wealth, and that firms in trade-exposed sectors sharply increase lobbying expenditures. To interpret these patterns, we develop and quantify a general equilibrium Ricardian model with heterogeneous households, input-output linkages, and endogenous trade policy shaped by voting and lobbying. Distributional shocks reallocate political support among voters, while lobbying propagates through production networks, generating strategic complementarities that sustain protectionism. Calibrated to U.S.-China sectoral data from 1991–2019, the model accounts for rising inequality, declining support for globalization, and key aggregate trends in consumption and trade.
摘要: 我们研究了全球化带来的不均衡收益如何内生地产生保护主义作为一种政治均衡。利用美国数据，我们记录了更多暴露于进口竞争的地区对全球化的反对更为强烈，尤其是在金融财富较少的家庭中，以及在贸易暴露部门的公司大幅增加了游说支出。为了解释这些模式，我们开发并量化了一个具有异质性家庭、投入产出联系和由投票及游说塑造的内生贸易政策的一般均衡李嘉图模型。分配冲击重新分配了选民之间的政治支持，而游说通过生产网络传播，产生了维持保护主义的战略互补性。该模型校准至1991-2019年间美中部门数据，解释了不平等的上升、对全球化支持的下降以及消费和贸易的关键总体趋势。
Does Advertising Expand the Market for Hospital Services? Evidence from Medicare
Authors: Abby E. Alpert, Atul Gupta, Michael R. Richards, Sarah D. Schutz, Christopher M. Whaley
Abstract: Direct-to-consumer advertising is pervasive in US healthcare markets, but little evidence exists on the effects of advertising by hospitals, second only to drug manufacturers in medical marketing. Advertising may help facilities increase market share by stealing existing patients, expand the market for hospital care, or do both. Regardless, it has important public finance implications due to the large sums of taxpayer funds spent by federal and state governments to subsidize hospital operations and finance care through public insurance programs. This paper provides the first causal evidence, to our knowledge, on the market expansion effects of hospital advertising. To obtain causal estimates, we leverage the fact that spikes in political advertising significantly crowd out hospital advertising in the same market, motivating an instrumental variables design. Using claims data on the universe of Traditional Medicare beneficiaries, we find that advertising expands aggregate patient volume and spending on inpatient care – though to a modest degree (implied elasticities of 0.06 and 0.05, respectively). Although the overall effect of advertising on hospital outpatient care is muted, for-profit hospitals obtain higher outpatient Medicare volume and revenue with greater advertising. Across both care settings, therefore, Medicare spending increases with hospital advertising.
摘要: 直接面向消费者的广告在美国医疗市场中无处不在，但关于医院广告效果的证据很少，医院在医疗营销中仅次于药品制造商。广告可能帮助医疗机构通过吸引现有患者来增加市场份额，扩大医院护理市场，或者两者兼而有之。无论如何，由于联邦和州政府花费大量纳税人资金来补贴医院运营并通过公共保险计划资助护理，因此它具有重要的公共财政意义。据我们所知，本文首次提供了关于医院广告市场扩张效应的因果证据。为了获得因果估计，我们利用政治广告激增显著挤出同一市场中医院广告的事实，激发了一个工具变量设计。使用传统医疗保险受益人的全部索赔数据，我们发现广告扩大了住院护理的总患者量和支出——尽管程度适中（分别为0.06和0.05的隐含弹性）。尽管广告对医院门诊护理的整体影响较为温和，但营利性医院通过增加广告获得更高的门诊医疗保险量和收入。因此，在这两种护理环境中，医疗保险支出随着医院广告的增加而增加。
Legalizing Cannabis: Implications for Child Maltreatment
Authors: Weiwei Chen, Johanna Catherine Maclean, Michael T. French
Abstract: Cannabis is the most used illicit drug in the United States. Though cannabis possession and consumption are prohibited federally, states are increasingly implementing laws that legalize this substance, initially for medical use and, more recently, for recreational use. We study the impact of recreational cannabis laws (RCLs) that allow all adults 21 years and older to legally consume and possess this substance on reports of child maltreatment and post-investigation service use. To do so, we combine difference-in-differences methods with administrative data on child maltreatment reports 2010-2022. Overall, we find no evidence that child maltreatment reports or service use are influenced by RCL adoption, though we find that physical abuse reports decline. Our findings indicate that recent efforts to legalize cannabis for recreational consumption have not led to an increase in child maltreatment reports.
摘要: 大麻是美国使用最广泛的非法药物。尽管在联邦层面上，大麻的持有和消费是被禁止的，但各州越来越多地实施法律，使这种物质合法化，最初是用于医疗用途，最近则是用于娱乐用途。我们研究了允许所有21岁及以上成年人合法消费和持有这种物质的娱乐性大麻法律（RCLs）对儿童虐待报告和调查后服务使用的影响。为此，我们将差异中的差异方法与2010-2022年儿童虐待报告的行政数据结合起来。总体而言，我们没有发现儿童虐待报告或服务使用受到RCL采纳的影响，尽管我们发现身体虐待报告有所下降。我们的发现表明，最近为娱乐消费而使大麻合法化的努力并没有导致儿童虐待报告的增加。
Authors: Natee Amornsiripanitch, Philip E. Strahan, Song Zhang
Abstract: Older home sellers receive lower returns than younger home sellers. Homes sold by older people have fewer major renovations but higher rates of poor upkeep. Older sellers are also more likely to sell off-MLS (“pocket listings”) and to sell to investors, leading to lower prices. These patterns suggest that older sellers may be disproportionately disadvantaged by agents’ incentive to maximize fees through generating high sales volume instead of maximizing sale prices. Age-related cognitive decline makes the elderly more vulnerable. For causal evidence, we show that reforms making pocket listings more transparent reduced both the prevalence of pocket listings and the magnitude of the age gap in returns.
摘要: 老年房主的回报低于年轻房主。老年人出售的房屋进行的重大翻新较少，但维护不善的比率较高。老年卖家也更有可能通过非MLS（“口袋房源”）出售房产，并且更倾向于卖给投资者，导致价格较低。这些模式表明，老年卖家可能因代理人通过产生高销售量而不是最大化销售价格来最大化费用的动机而受到不成比例的不利影响。与年龄相关的认知衰退使老年人更加脆弱。为了提供因果证据，我们展示了使口袋房源更透明的改革既减少了口袋房源的普遍性，也减少了回报中年龄差距的大小。
World Trading System For Whom? Evidence from Global Tariffs
Authors: Rodrigo Adão, John Sturm Becko, Arnaud Costinot, Dave Donaldson
Abstract: We use global tariffs to reveal the weights that nations implicitly place on the welfare of their trading partners relative to their own. Our estimated welfare weights suggest that formal and informal rules of the world trading system make countries internalize the impact of their policies onto others to a substantial extent, though not fully. On average, countries place 25% less value on transfers to foreigners than transfers to their own residents. Across nations, we find that countries that put higher welfare weights on the welfare of foreigners also tend to receive higher weights from them, consistent with a general form of reciprocity among nations. Using our estimated welfare weights, we provide a first look at what countries stand to lose, or gain, from the dissolution of the world trading system as we know it.
摘要: 我们利用全球关税来揭示各国在相对自身福利的情况下，对贸易伙伴福利的隐含权重。我们估计的福利权重表明，世界贸易体系的正式和非正式规则使得各国在很大程度上内化了他们的政策对其他国家的影响，尽管并非完全如此。平均而言，各国对外国人转移支付的价值比对本国居民的转移支付低25%。在国家之间，我们发现对外国人福利赋予更高权重的国家也倾向于从他们那里获得更高的权重，这与国家之间的一般互惠形式相一致。利用我们估计的福利权重，我们首次审视了各国可能因世界贸易体系的解体而遭受的损失或获得的收益。
The Economics of Scaling Early Childhood Programs: Lessons from The Chicago School
Authors: John A. List
Abstract: Many ideas show remarkable returns in small-scale trials but often disappoint when scaled to broader populations and contexts. Using early childhood investment as a case study, this study develops a dynamic human capital formation model that integrates complementary skill investment with “Option C thinking” on scaling challenges. The model is stylized in the Chicago tradition: micro-founded with optimizing agents, dynamic skill production, and a policymaker evaluating scaling decisions. It formalizes how naive extrapolation from pilot studies systematically overestimates policy efficacy by ignoring “voltage drops,” declining treatment effects due to unrepresentativeness at scale. The model demonstrates that optimal scaling policy requires mechanism-based design that anticipates these failures through backward induction from implementation realities. The scientific insights from a set of recent studies provide valuable perspectives on the model.
摘要: 许多想法在小规模试验中显示出显著的回报，但在扩大到更广泛的人群和环境中时常常令人失望。以儿童早期投资为案例研究，本研究开发了一个动态人力资本形成模型，该模型将补充技能投资与“选项C思维”在扩大挑战上的思考结合起来。该模型以芝加哥传统风格呈现：以微观为基础，包含优化行为主体、动态技能生产和评估扩大决策的政策制定者。它正式化了如何从试点研究中简单外推系统性地高估政策效果，因为忽略了“电压下降”，即由于规模上的非代表性导致治疗效果下降。该模型展示了，最佳的扩大政策需要基于机制的设计，通过从实施现实向后归纳来预见这些失败。一组最新研究的科学见解为模型提供了宝贵的视角。
Dementia and Long-run Trajectories in Household Finances
Authors: Jing Li, Kathleen M. McGarry, Lauren Hersch Nicholas, Jonathan S. Skinner
Abstract: Existing evidence suggests that wealth may decline before dementia onset, but the mechanisms underlying these reductions are poorly understood. Using longitudinal data from the Health and Retirement Study, we compare household finance trajectories for individuals who later develop dementia and those who do not. We find that wealth divergence between the two groups is not explained by reduced earnings, higher healthcare spending, intentional “spend-down” to qualify for Medicaid coverage, state-dependent utility, or reverse causation by which wealth declines cause dementia. Instead, our results point to impaired financial decision-making beginning about six years prior to clinically recognizable dementia.
摘要: 现有的证据表明，在痴呆症发病前，财富可能会下降，但这些减少背后的机制尚不清楚。利用健康和退休研究的纵向数据，我们比较了后来发展成痴呆症的个体和没有发展成痴呆症的个体的家庭财务轨迹。我们发现，两组之间的财富差异不能通过降低收入、增加医疗支出、故意“花光”以获得医疗补助保险、状态依赖效用或财富下降导致痴呆症的反向因果关系来解释。相反，我们的结果指向了大约在临床可识别的痴呆症前六年开始的财务决策能力受损。
Authors: Marc A. C. Hafstead, Roberton C. Williams III
Abstract: The full economic costs and macroeconomic effects of microeconomic damages associated with pollution damages, climate impacts, and similar negative shocks are not well understood. Beyond the direct damages, market distortions, supply chains, changes in capital and labor costs, etc. can potentially magnify (or diminish) the costs of these shocks. In this paper, we use analytical and numerical general equilibrium models to investigate the channels through which microeconomic shocks affect the macroeconomy. We model different types of shocks, evaluate how anticipation affects the macroeconomic implications of shocks, and measure the overall welfare costs relative to the value of direct damages. We find that preexisting taxes magnify the overall welfare costs relative to direct damages, that savings and investment decisions driven by anticipations of certain types of shocks can interact with tax distortions to lower welfare costs, and that the ratio of welfare cost to direct damages varies substantially across sectors of the economy.
摘要: 与污染损害、气候影响以及类似负面冲击相关的微观经济损害的全部经济成本和宏观经济效应尚未被充分理解。除了直接损害之外，市场扭曲、供应链、资本和劳动力成本的变化等，可能会放大（或减少）这些冲击的成本。在本文中，我们使用分析性和数值性的一般均衡模型来研究微观经济冲击如何影响宏观经济的渠道。我们模拟了不同类型的冲击，评估预期如何影响冲击的宏观经济含义，并相对于直接损害的价值衡量总体福利成本。我们发现，预先存在的税收放大了相对于直接损害的总体福利成本，由预期某些类型冲击所驱动的储蓄和投资决策可以与税收扭曲相互作用，降低福利成本，以及福利成本与直接损害的比率在经济部门之间差异很大。
Authors: Jason Abaluck, Robert Pless, Nirmal Ravi, Anja Sautmann, Aaron Schwartz
Abstract: We deployed large language model (LLM) decision support for health workers at two outpatient clinics in Nigeria. For each patient, health workers drafted care plans that were optionally revised after LLM feedback. We compared unassisted and assisted plans using blinded randomized assessments by on-site physicians who evaluated and treated the same patients, as well as results from laboratory tests for common conditions. Academic physicians performed blinded retrospective reviews of a subset of notes. In response to LLM feedback, health workers changed their prescribing for more than half of the patients and reported high satisfaction with the recommendations, and retrospective academic reviewers rated LLM-assisted plans more favorably. However, on-site physicians observed little to no improvement in diagnostic alignment or treatment decisions. Laboratory testing showed mixed effects of LLM-assistance, which removed negative tests for malaria but added them for urinary tract infection and anemia, with no significant increase in the detection rates for the tested conditions.
摘要: 我们在尼日利亚的两个门诊诊所为卫生工作者部署了大型语言模型（LLM）决策支持。对于每位患者，卫生工作者起草了护理计划，这些计划在LLM反馈后可选择进行修订。我们通过现场医生对同一患者进行评估和治疗的盲法随机评估，以及对常见疾病进行实验室测试的结果，比较了未辅助和辅助计划。学术医生对一部分记录进行了盲法回顾性审查。作为对LLM反馈的响应，卫生工作者为超过一半的患者改变了他们的处方，并报告对建议感到高度满意，而回顾性学术审查者对LLM辅助计划的评价更为有利。然而，现场医生观察到在诊断一致性或治疗决策方面几乎没有或没有改善。实验室测试显示LLM辅助的混合效果，它消除了疟疾的阴性测试，但增加了尿路感染和贫血的测试，对于测试条件的检出率没有显著增加。