数字经济中的企业生产率与学习：来自云计算的证据

James M. Brand, Mert Demirer, Connor Finucane, and Avner A. Kreps #32938

Abstract: Computing technologies have become critical inputs to production in the modern firm. However, there is little large-scale evidence on how efficiently firms use these technologies. In this paper, we study firm productivity and learning in cloud computing by leveraging CPU utilization data from over one billion virtual machines used by nearly 100,000 firms. We find large and persistent heterogeneity in compute productivity both across and within firms, similar to canonical results in the literature. More productive firms respond better to demand fluctuations, show higher attentiveness to resource utilization, and use a wider variety of specialized machines. Notably, productivity is dynamic as firms learn to be more productive over time. New cloud adopters improve their productivity by 33% in their first year and reach the productivity level of experienced firms within four years. In our counterfactual calculations, we estimate that raising all firms to the 80th percentile of productivity would reduce aggregate electricity usage by 17%.

摘要：计算技术已经成为现代企业生产中的关键投入。然而，关于企业如何有效利用这些技术的大规模证据仍然较少。在本文中，我们通过分析近10万家企业使用的超过10亿台虚拟机的CPU利用率数据，研究了云计算中的企业生产率和学习情况。我们发现，企业之间和企业内部的计算生产率存在较大且持续的异质性，这与文献中的经典结果相似。生产力较高的企业对需求波动反应更灵敏，更加关注资源利用，并使用种类更多的专业化机器。值得注意的是，生产率是动态变化的，企业会随着时间的推移学会变得更加高效。新采用云计算的企业在第一年内将生产力提高了33%，并在四年内达到了有经验企业的生产力水平。在我们的反事实计算中，我们估计，如果将所有企业的生产率提升到第80百分位，将减少17%的总体电力消耗。

通胀不确定性对家庭预期和消费的影响

Olena Kostyshyna and Luba Petersen #32939

Abstract: We use a new Canadian household survey to examine how inflation uncertainty influences inflation expectations and spending. Through randomized information interventions, we provide inflation statistics with or without second moments, creating variations in households' inflation uncertainty. All information types effectively lower inflation expectations and uncertainty. While communicating inflation uncertainty does not affect expectations or uncertainty levels, it increases the probability assigned to expected inflation near communicated ranges. Using Nielsen IQ Homescanner data, we find that higher inflation expectations and uncertainty reduce household spending on goods. Communicating inflation statistics with ranges increases spending by lowering expectations and reducing uncertainty.

摘要：我们使用一项新的加拿大家庭调查来研究通胀不确定性如何影响通胀预期和消费。通过随机信息干预，我们提供了包含或不包含二阶矩的通胀统计数据，从而在家庭通胀不确定性上产生了变化。所有类型的信息都有效地降低了通胀预期和不确定性。尽管传达通胀不确定性不会影响预期或不确定性的水平，但它增加了家庭将预期通胀值分配到所传达区间内的概率。通过使用尼尔森IQ家庭扫描数据，我们发现较高的通胀预期和不确定性会减少家庭在商品上的消费。而传达包含区间的通胀统计数据通过降低预期和减少不确定性，增加了消费。

专利文本与长期创新动态：模型选择的关键作用

Ina Ganguli, Jeffrey Lin, Vitaly Meursault, and Nicholas F. Reynolds #32934

Abstract: As distorted maps may mislead, Natural Language Processing (NLP) models may misrepresent. How do we know which NLP model to trust? We provide comprehensive guidance for selecting and applying NLP representations of patent text. We develop novel validation tasks to evaluate several leading NLP models. These tasks assess how well candidate models align with both expert and non-expert judgments of patent similarity. State-of-the-art language models significantly outperform traditional approaches such as TF-IDF. Using our validated representations, we measure a secular decline in contemporaneous patent similarity: inventors are “spreading out” over an expanding knowledge frontier. This finding is corroborated by declining rates of multiple invention from newly-digitized historical patent interference records. In contrast, selecting another single representation without validating alternatives yields an ambiguous or even opposing trend. Thus, our framework addresses a fundamental challenge of selecting among different black-box NLP models that produce varying economic measurements. To facilitate future research, we plan to provide our validation task data and embeddings for all US patents from 1836–2023.

摘要：正如失真的地图可能会误导，自然语言处理（NLP）模型也可能产生错误表述。我们如何知道该信任哪个NLP模型？本文为选择和应用专利文本的NLP表征提供了全面的指导。我们设计了新颖的验证任务来评估多个领先的NLP模型，这些任务评估候选模型与专家和非专家对专利相似性的判断是否一致。最先进的语言模型显著优于传统方法，如TF-IDF。利用我们经过验证的表征，我们衡量了当代专利相似度的长期下降趋势：发明者正在“扩散”到一个不断扩展的知识边界。这一发现得到了新数字化的历史专利干涉记录中多项发明率下降的证实。相比之下，未验证其他选择而仅选择单一表征会导致模棱两可甚至相反的趋势。因此，我们的框架解决了在不同黑箱NLP模型中进行选择时导致经济测量差异的一个基本挑战。为了促进未来的研究，我们计划提供1836至2023年间所有美国专利的验证任务数据和嵌入。

为什么工人不喜欢通货膨胀？工资侵蚀和冲突成本

Joao Guerreiro, Jonathon Hazell, Chen Lian, and Christina Patterson #32956

Abstract: How costly is inflation to workers? Answers to this question have focused on the path of real wages during inflationary periods. We argue that workers must take costly actions (“conflict”) to have nominal wages catch up with inflation, meaning there are welfare costs even if real wages do not fall as inflation rises. We study a menu-cost style model, where workers choose whether to engage in conflict with employers to secure a wage increase. We show that, following a rise in inflation, wage catchup resulting from more frequent conflict does not raise welfare. Instead, the impact of inflation on worker welfare is determined by what we term “wage erosion”—how inflation would lower real wages if workers' conflict decisions did not respond to inflation. As a result, measuring welfare using observed wage growth understates the costs of inflation. We conduct a survey showing that workers are willing to sacrifice 1.75% of their wages to avoid conflict. Calibrating the model to the survey data, the aggregate costs of inflation incorporating conflict more than double the costs of inflation via falling real wages alone.

摘要：通货膨胀对工人的成本有多大？对这个问题的回答通常集中在通货膨胀期间实际工资的变化路径上。我们认为，工人必须采取代价高昂的行动（“冲突”）才能使名义工资追上通货膨胀，这意味着即使实际工资没有随着通货膨胀上升而下降，福利成本仍然存在。我们研究了一种菜单成本模型，在该模型中，工人选择是否与雇主发生冲突以争取加薪。我们展示了，在通货膨胀上升后，因更频繁的冲突而实现的工资追赶并不会提高福利。相反，通货膨胀对工人福利的影响由我们称之为“工资侵蚀”所决定——即如果工人的冲突决策没有对通货膨胀作出反应，通货膨胀将如何降低实际工资。因此，仅通过观察到的工资增长来衡量福利会低估通货膨胀的成本。我们进行了一项调查，显示工人愿意牺牲1.75%的工资以避免冲突。根据调查数据校准模型，纳入冲突后的通货膨胀总成本是单单通过实际工资下降计算的通货膨胀成本的两倍以上。

带有多层样本选择的Lee Bounds方法

Kory Kroft, Ismael Mourifié, and Atom Vayalinkal #32952

Abstract: This paper investigates the causal effect of job training on wage rates in the presence of firm heterogeneity. When training affects worker sorting to firms, sample selection is no longer binary but is “multilayered”. This paper extends the canonical Heckman (1979) sample selection model - which assumes selection is binary - to a setting where it is multilayered, and shows that in this setting Lee bounds set identifies a total effect that combines a weighted-average of the causal effect of job training on wage rates across firms with a weighted-average of the contrast in wages between different firms for a fixed level of training. Thus, Lee bounds set identifies a policy-relevant estimand only when firms pay homogeneous wages and/or when job training does not affect worker sorting across firms. We derive sharp bounds for the causal effect of job training on wage rates at each firm which leverage information on firm-specific wages. We illustrate our partial identification approach with an empirical application to the Job Corps Study. Results show that while conventional Lee bounds are strictly positive, our within-firm bounds include 0 showing that canonical Lee bounds may be capturing a pure sorting effect of job training.

摘要：本文研究了在公司异质性存在的情况下，职业培训对工资水平的因果影响。当培训影响工人对企业的选择时，样本选择不再是二元的，而是“多层次的”。本文扩展了经典的Heckman（1979）样本选择模型，该模型假设选择是二元的，而本文将其扩展到多层次的情境中，并表明在这种情境下，Lee界限集识别了一个总效应，该总效应结合了职业培训对各公司工资水平的因果影响的加权平均值与不同公司在相同培训水平下工资差异的加权平均值。因此，只有当公司支付同质工资和/或职业培训不影响工人在不同公司间的选择时，Lee界限集才能识别出与政策相关的估量值。我们为每个公司中职业培训对工资水平的因果影响推导出了利用公司特定工资信息的精确界限。我们通过对Job Corps研究的实证应用说明了我们部分识别的方法。结果显示，虽然传统的Lee界限是严格正值，但我们的公司内部界限包括0，这表明经典的Lee界限可能捕捉到了职业培训的纯粹排序效应。

较高的小费最低工资是否能减少餐馆工人的种族、民族或性别收入差距？

David Neumark and Emma Wohl #32964

Abstract: One of the arguments increasingly made to support large minimum wage increases is that they decrease wage or earnings gaps for minorities or women (e.g., Derenoncourt and Montialoux, 2021). The argument is often made with particular reference to higher tipped minimum wages for restaurant workers, because of discrimination in tipping that is immune to equal pay policy requirements. Of course, even if higher tipped minimum wages reduce hourly pay differences between groups, increases in tipped minimum wages can reduce employment or hours among restaurant workers (Neumark and Yen, 2023), and these effects could differ by race and gender, so implications for hourly earnings do not necessarily extend to overall earnings. We estimate the impact of variation in tipped minimum wages – or, equivalently, tip credits – on earnings of restaurant workers (which ignores employment variation but incorporates hours variation). We find that tipped minimum wages raise hourly earnings of women, but not of Blacks or Hispanics. But tipped minimum wages generally do not raise weekly earnings for these groups (because of hours declines for women). In contrast, regular minimum wages boost hourly and weekly earnings of all three groups of restaurant workers, with the effects arising from non-tipped workers.

摘要：近年来，支持大幅提高最低工资的论点之一是，这可以减少少数族裔或女性的工资或收入差距（如Derenoncourt和Montialoux，2021）。这一论点常常特别提到餐馆工人的较高小费最低工资，因为小费中的歧视行为无法通过同工同酬政策加以解决。当然，即使较高的小费最低工资减少了群体间的时薪差异，小费最低工资的增加也可能减少餐馆工人的就业或工作时长（Neumark和Yen，2023），而这些影响可能因种族和性别而有所不同，因此对时薪的影响不一定会延伸到整体收入上。我们估计了小费最低工资（或相应的小费抵免）对餐馆工人收入的影响（忽略了就业的变化，但考虑了工作时长的变化）。我们发现，小费最低工资提高了女性的时薪，但对黑人或西班牙裔工人没有影响。不过，小费最低工资通常不会提高这些群体的周收入（因为女性的工作时长减少）。相比之下，常规最低工资提高了这三个群体餐馆工人的时薪和周收入，这一效果主要来自于非小费工人。

居家办公与残疾人就业

Nicholas Bloom, Gordon B. Dahl, and Dan-Olof Rooth #32943

Abstract: There has been a dramatic rise in disability employment in the US since the pandemic, a pattern mirrored in other countries as well. A similar increase is not found for any other major gender, race, age or education demographic. At the same time, work from home has risen four-fold. This paper asks whether the two are causally related. Analyzing CPS and ACS microdata, we find the increase in disability employment is concentrated in occupations with high levels of working from home. Controlling for compositional changes and labor market tightness, we estimate that a 1 percentage point increase in work from home increases full-time employment by 1.1% for individuals with a physical disability. A back of the envelope calculation reveals that the post pandemic increase in working from home explains 80% of the rise in full-time employment. Wage data suggests that WFH increased the supply of workers with a disability, likely by reducing commuting costs and enabling better control of working conditions.

摘要：自疫情以来，美国残疾人就业率显著上升，其他国家也出现了类似的趋势。然而，这一增长并未在其他主要性别、种族、年龄或教育背景群体中发现。与此同时，居家办公的比例增加了四倍。本文探讨了两者是否存在因果关系。通过分析CPS和ACS微观数据，我们发现残疾人就业的增加主要集中在居家办公比例较高的职业中。在控制了人口结构变化和劳动力市场紧张度的情况下，我们估计居家办公比例每增加1个百分点，身体残疾人士的全职就业率就增加1.1%。通过简易计算，疫情后居家办公的增加解释了全职就业增长的80%。工资数据表明，居家办公增加了残疾人的劳动力供给，可能是因为减少了通勤成本并使他们能够更好地控制工作环境。

专利猎手

Lauren Cohen, Umit Gurun, Katie Moon, and Paula Suh #32965

Abstract: Analyzing millions of patents granted by the USPTO between 1976 and 2020, we find a pattern where specific patents only rise to prominence after considerable time has passed. Amongst these late-blooming influential patents, we show that there are key players (patent hunters) who consistently identify and develop them. Although initially overlooked, these late-blooming patents have significantly more influence on average than early-recognized patents and are associated with significantly more new product launches. Patent hunters, as early detectors and adopters of these late-blooming patents, are also associated with significant positive rents. Their adoption of these overlooked patents is associated with a 6.4% rise in sales growth (t = 3.02), a 2.2% increase in Tobin’s Q (t = 3.91), and a 2.2% increase in new product offerings (t = 2.97). We instrument for patent hunting, and find strong evidence that these benefits are causally due to patent hunting. The rents associated with patent hunting on average exceed those of the original patent creators themselves. Patents hunted are closer to the core technology of patent hunters, more peripheral to writers, and in less competitive spaces. Lastly, patent hunting appears to be a persistent firm characteristic and to have an inventor-level component.

摘要：通过分析美国专利商标局（USPTO）在1976年至2020年间授予的数百万项专利，我们发现了一种模式，即某些专利在经过相当长时间后才变得重要。在这些“晚开花”的重要专利中，我们发现了一些关键角色（即“专利猎手”），他们能够持续识别并开发这些专利。尽管最初被忽视，这些晚开花的专利平均上比早期获得认可的专利具有更大的影响力，并且与更多的新产品推出相关联。作为早期发现和采用这些晚开花专利的专利猎手，他们与显著的正收益相关联。专利猎手采用这些被忽视的专利后，公司销售增长率平均提高了6.4%（t=3.02），Tobin’s Q（企业市值与资产重置成本比率）提高了2.2%（t=3.91），新产品推出增加了2.2%（t=2.97）。我们对“专利猎取”行为进行了工具变量分析，发现强有力的证据表明，这些收益是由于专利猎取所带来的。专利猎取所带来的收益通常超过了原始专利发明人自身获得的收益。被猎取的专利更接近专利猎手的核心技术，离发明者的核心领域较远，并且通常处于竞争较少的领域。最后，专利猎取似乎是公司层面持久的特征，并具有发明者层面的影响。

购买美国货成本的上升

Matilde Bombardini, Andres Gonzalez-Lira, Bingjing Li, and Chiara Motta #32953

Abstract: The latest resurgence in the U.S. of policies aimed at reducing imports and bolstering domestic production has included the expansion of Buy American provisions. While some of these are new and untested, in this paper we evaluate long-standing procurement limitations on the purchase of foreign products by the U.S. Federal Government. We use procurement micro-data to first map and measure the positive employment effects of government purchases. We then calibrate a quantitative trade model adapted to include features relevant to the Buy American Act: a government sector, policy barriers in final and intermediate goods, labor force participation, and external economies of scale. We show that current Buy American provisions on final goods purchase have created up to 100,000 jobs at a cost of between $111,500 and $137,700 per job. However, the recently announced tightening of the policy on the use of foreign inputs will create fewer jobs at a higher cost of $154,000 to $237,800 per job. We also find scant evidence of the use of Buy American rules as an effective industrial policy.

摘要：美国近期再次推行旨在减少进口、促进国内生产的政策，其中包括扩大“购买美国货”条款的适用范围。虽然其中一些政策是新的且尚未经过验证，但本文评估了美国联邦政府长期以来对外国产品采购的限制。我们首先使用采购微观数据绘制并衡量政府采购对就业的积极影响。接着，我们校准了一个定量贸易模型，该模型纳入了与《购买美国法案》相关的特征：包括政府部门、对最终产品和中间产品的政策壁垒、劳动力参与率以及外部规模经济。我们发现，现行的购买美国货政策在最终产品采购方面创造了多达10万个就业岗位，但每个就业岗位的成本在111,500美元到137,700美元之间。然而，最近宣布的关于使用外国投入品的政策收紧措施将以更高的成本（每个岗位154,000美元至237,800美元）创造更少的就业岗位。此外，我们几乎没有发现使用“购买美国”规则作为有效产业政策的证据。

碳VIX：碳价不确定性与脱碳投资

Maximilian Fuchs, Johannes Stroebel, and Julian Terstegge #32937

Abstract: We study the effects of carbon price uncertainty on firms' decisions to decarbonize their operations. We first use information on the pricing of options on emission allowances in the European Emissions Trading System to create the Carbon VIX, a market-based high-frequency measure of carbon price uncertainty. Carbon price uncertainty is high, varies substantially over time, and experiences persistent shocks around major climate policy events. To explore the effects of carbon price uncertainty on expected aggregate decarbonization investments, we analyze its effect on the stock returns of firms that help other businesses decarbonize. To identify these "carbon solution providers," we extract common types of decarbonization investments from a large survey of firms, and then identify companies that offer the associated goods and services. We find that the stock returns of these carbon solution providers vary positively with carbon prices, but negatively with carbon price uncertainty. The effect of increases in carbon price uncertainty on our proxy for expected decarbonization investments is economically large and of similar magnitude as the effect of declines in carbon prices. These findings support predictions from real options theory that firms may delay investments in decarbonization when faced with uncertainty about the future costs of emissions.

摘要：我们研究了碳价不确定性对企业脱碳决策的影响。首先，我们利用欧洲碳排放交易体系中的排放许可期权定价信息，创建了“碳VIX”，这是一种基于市场的高频碳价不确定性指标。碳价不确定性较高，随时间大幅波动，并在重大气候政策事件期间经历持续冲击。为了探讨碳价不确定性对预期总脱碳投资的影响，我们分析了它对帮助其他企业实现脱碳的公司的股票回报的影响。为了识别这些“碳解决方案提供商”，我们从大量公司调查中提取常见的脱碳投资类型，并识别提供相关商品和服务的公司。我们发现，这些碳解决方案提供商的股票回报与碳价呈正相关，但与碳价不确定性呈负相关。碳价不确定性增加对我们用作预期脱碳投资的代理变量的影响在经济上是巨大的，其影响程度与碳价下降的影响相似。这些发现支持了实物期权理论的预测：面对未来排放成本的不确定性，企业可能会推迟脱碳投资。

通过结合参数数据、行政数据和调查数据评估宏观指标中的非响应偏差

Ori Heffetz, Guy Lichtinger, and Daniel B. Reeves #32935

Abstract: National surveys are crucial for estimating key economic aggregates, including the unemployment rate, labor force participation, and household expenditures. The accuracy of these indicators is increasingly under scrutiny due to declining response rates and the consequent risk of nonresponse bias. How can we assess nonresponse bias in these key economic aggregates? Using two Israeli national surveys, we propose a novel approach. First, using often-available paradata such as number of contact attempts and nonresponse reason, we create respondent and nonrespondent subcategories. Second, using rarely-available merged administrative records, we identify, for each nonrespondent subcategory, which respondent subcategory appears to resemble it most. We find that nonrespondents are a heterogeneous group: some—e.g., those temporarily unavailable—share administrative-record demographic and outcome profiles with harder-to-reach respondents, while others—e.g., refusals and withdrawals—are more similar in the administrative data to easier-to-reach respondents. Third, assuming that these resemblances would extend to survey outcomes, we impute (always-available) survey-based aggregates to nonrespondents within each paradata-based subcategory. We demonstrate that our method can help assess nonresponse bias in surveys lacking matched administrative records, using the U.S. Consumer Expenditure Survey as an example.

摘要：全国性调查对于估算关键经济总量（包括失业率、劳动力参与率和家庭支出）至关重要。然而，随着响应率的下降，这些指标的准确性越来越受到关注，因为非响应偏差的风险正在增加。我们该如何评估这些关键经济总量中的非响应偏差呢？利用两项以色列的全国性调查，我们提出了一种新方法。首先，使用常见的参数数据（如联系尝试次数和非响应原因），我们将受访者和非受访者划分为子类别。其次，利用较为罕见的合并行政记录，我们为每个非受访者子类别识别出最相似的受访者子类别。我们发现，非受访者是一个异质群体：一些暂时无法参与的非受访者在行政记录中的人口统计和结果特征与难以联系的受访者相似，而另一些拒绝或退出的非受访者在行政数据中则更接近容易联系的受访者。第三，假设这些相似性会延伸至调查结果，我们根据每个基于参数数据的子类别，对非受访者进行基于调查数据的总量估算。我们展示了这种方法如何帮助评估那些缺乏匹配行政记录的调查中的非响应偏差，并以美国消费者支出调查为例进行说明。

贸易的分配性后果：来自谷物入侵的证据

Stephan Heblich, Stephen J. Redding, and Yanos Zylberberg #32958

Abstract: We examine the distributional consequences of trade using the New World Grain Invasion that occurred in the second half of the 19th century. We use a newly-created dataset on population, employment by sector, property values, and poor law transfers for over 10,000 parishes in England and Wales from 1801–1901. In response to this trade shock, we show that locations with high wheat suitability experience population decline, rural-urban migration, structural transformation away from agriculture, increases in welfare transfers, and declines in property values, relative to locations with low wheat suitability. We develop a quantitative spatial model to evaluate the income distributional consequences of this trade shock. Undertaking counterfactuals for the Grain Invasion, we show that geography is an important dimension along which these income distributional consequences occur.

摘要：我们通过研究19世纪下半叶的“新世界谷物入侵”来探讨贸易的分配性后果。我们使用一个新创建的数据集，该数据集包含1801年至1901年间英格兰和威尔士超过10,000个教区的人口、按行业划分的就业情况、财产价值和救济法转移支付。在应对这一贸易冲击时，我们发现与小麦适宜性较低的地区相比，小麦适宜性较高的地区出现了人口减少、农村向城市的迁移、农业结构转型、福利转移增加和财产价值下降。我们还开发了一个定量空间模型，以评估这一贸易冲击对收入分配的影响。通过对“谷物入侵”进行反事实分析，我们表明地理因素是这些收入分配后果的重要维度。

往期精选：

      重磅：学术财经•学术笔记全汇总

重磅：学术财经全球价值链专题学术笔记大汇总（附下载链接失效学术笔记推文）

重磅| 基金申请内部学习和交流

【学术笔记·第37篇】风险和全球供应链：我们知道什么和我们需要知道什么

【学术笔记·第36篇】各有所长：全球供应链风险敞口（Exposure）测度

【学术笔记·第35篇】日本生产者价格指数理论与实证：最终需求-中间需求加总系统

【学术笔记·第34篇】北美自由贸易的贸易和福利影响估计

【学术笔记 第33篇】 英属印度时期铁路：交通基础设施的影响评估（AER）

【学术笔记 第32篇】 技术、地理和贸易(经典EK模型）

【学术笔记 第31篇】2018年贸易战对美国物价和福利的影响

【学术笔记 第30篇】全球价值链和国家部门层面实际有效汇率