NBER Weekly Digest

We examine the growth and performance of AI-driven investing. Using investment advisers' regulatory disclosures, labor market data, and fund strategy descriptions, we document that AI-driven investing has grown steadily since the early 2010s and is concentrated among hedge funds. AI hedge funds outperformed non-AI hedge funds in the early years, but this outperformance declined over time, even among early adopters. Contrary to concerns about AI-driven strategy homogeneity, AI hedge funds exhibit lower return comovement than non-AI peers. Our findings highlight both the alpha-generating potential and the limitations of AI as a source of investment performance.

Formal childcare and public schooling are widely understood to be central to mothers’ labor market attachment. Yet, a large literature finds that even substantial expansions of formal childcare produce modest or null effects on maternal employment. We study the role of stochastic caregiving demand shocks that cause childcare constraints to bind intermittently even when formal care is available. Using plausibly exogenous variation in wildfire smoke exposure, we show that smoke increases school closures and student absenteeism, generating caregiving demand. Further, by comparing mothers whose youngest child varies in reliance on school-based childcare, we find that these shocks impose substantial contemporaneous costs and, when accumulated, reduce employment. Employment losses are fully offset among mothers working outside male-dominated industries, consistent with workplace tolerance moderating the consequences of intermittent caregiving demands.

Extensive discussion centers around the question of whether affirmative action (AA) in higher education “works,” despite AA being in practice a continuous policy instrument. We study the dose effects of AA on its intended beneficiaries by exploiting the ethnic minority bonus-point policy in China's centralized college admissions system, where targeted students receive explicit point advantages. Exploring cohort-level fluctuations in exam score distributions that generate plausibly exogenous variation in the effective dose of AA, we show that higher AA doses increase recipients' admissions to colleges and elite colleges, but lower their academic rankings, shift them toward lower-earning majors, and reduce their satisfaction with college life. From data linking college entrance exam (CEE) records to comprehensive card transaction records 10–23 years after CEE, we find that low AA doses increase long-term consumption by 19% but high doses reduce it by 25%. Such non-monotonic dose responses exist across expenditure categories, including child-related expenditure, suggesting potential impacts on future generations. Heterogeneity analysis reveals that while the positive effects of low AA doses persist across student groups, students from stronger academic backgrounds are more resilient to the negative effects of high AA doses. Overall, our results suggest that careful calibration of AA policies is crucial for achieving their intended goals.

Why did U.S. manufacturing productivity stop growing after 2010? Productivity growth disappeared, evaporating from an annual rate of +3.3 percent during 1987-2010 to -0.3 percent from 2010 to 2023. This paper shifts attention from 2010 as the start of the puzzle to a decade earlier when output stopped growing. This cessation of output growth in 2000 is attributed to the invasion of imports that closed domestic plants, destroyed jobs, and squeezed profits. Then followed a chain of causation that ultimately undermined productivity growth – from falling capacity utilization, to lower investment in fixed capital and R&D, and to an erosion of innovation. Beyond the import invasion, the paper identifies a set of handicaps ranging from self-inflicted wounds by private manufacturing firms to a marked reduction in government-funded R&D spending. Corporate funds were diverted from productive investment to share buybacks. Investment was distorted by environmental, health, safety, and fuel economy regulations. Innovation slowed not only because of diminishing returns to R&D, but also because of a decline in public R&D, and a diversion of private R&D from basic science and process improvements to product refinements and brand extensions. Skilled worker shortages have plagued manufacturing for decades in the absence of sufficient public and private investment in vocational training.

We revisit the employer size wage effect (ESWE)—arguably the most basic and influential departure from the law of one price for labor. Our main result is that this canonical fact disappears completely across establishments within the same firm, even though they operate in different local labor markets. We uncover and dissect this fact by including a firm fixed effect in otherwise standard cross-sectional regressions of wages on establishment size. We implement this demanding specification in population-wide triple-linked firm-establishment-employee data in Germany. This result is new in the ESWE literature (for which our paper also provides the first systematic meta-analysis). This wage-size decoupling is hard to square with the view that employment is determined along a finitely elastic employer-specific labor supply curve—i.e., employers pay exactly the minimum needed for the quantity of labor, but no more—the foundation of the monopsony view. By contrast, large multi-establishment firms (MEF) appear to hire off their labor supply curves (or those curves are very elastic), pay wage premia above the monopsonistic minimum, and leave excess labor supply. We find some evidence for a reemergence of the ESWE within low-premium MEFs. Overall, at least for the 25% of German employment in large firms for which the ESWE disappears, wage setting and employment determination may be better accounted for by alternative models, namely accommodating above-market-clearing wage premia and rationing of labor supply, such as efficiency wage theories.

We study whether exposure to frontier knowledge in college affects student outcomes. Combining 459,415 syllabi from seven Texas public universities with 107 million publications and linked student records, we measure each course’s proximity to recent versus older research in its field. Exploiting syllabus updates unobserved at enrollment, we find that frontier exposure increases completion, GPA, graduate-school attendance, and earnings, and reduces time-to-degree. Completion, GPA, and progression gains are broad, while graduate-school and earnings returns are larger for students with stronger preparation and family resources. The evidence suggests two mechanisms: frontier content keeps students engaged, and sustained exposure builds labor-market skills.

We document and explain the gap between measures of AI exposure and measures of AI adoption in the workplace. This leads us to propose a new AI adoption index based on comparative advantage. Using the representative German DiWaBe employee survey linked to worker and establishment information, we compare worker-reported AI use to prominent exposure measures and find that the relationship is weak. Motivated by this gap, we develop a framework in which adoption depends not only on technical feasibility (i.e., AI’s absolute advantage measured by exposure) but on profitability (i.e., AI’s comparative (dis)advantage relative to a specific worker), balancing AI productivity against AI user costs and worker productivity against wages. We operationalize this framework at the task level by (i) estimating worker productivity relative to pay, (ii) mapping exposure indices into AI productivity, and (iii) inferring task-specific AI user costs from revealed-preference adoption. The resulting occupation-level index accounts for almost 60% of cross-occupation variation in observed AI adoption, compared to 14% for an exposure-only model. The two approaches diverge substantially for approximately 30% of workers, highlighting that comparative advantage—not exposure alone—is crucial for assessing AI’s labor-market impact.

We document the intergenerational mobility of housing capital in the United States using a new dataset linking Decennial Census data to administrative property and income-tax records for over 3.4 million families. We find that housing capital is substantially more persistent across generations than earnings. Moreover, the gap in housing capital between White and Black children widens sharply across the parental distribution, more so than the earnings gap. Using a capital accumulation and transmission framework, we study how assets and earnings jointly shape intergenerational housing capital persistence. Less than half of this persistence operates through children's earnings, leaving substantial scope for direct transmissions of capital assets and knowledge. Differences in earnings explain most of the White-Black housing gap at the bottom of the parental distribution, but less than half at the top. Finally, we present quasi-experimental evidence that local housing supply constraints amplify intergenerational persistence by widening homeownership gaps between children from lower- and higher-resource families. Together, our findings indicate that economic resources are more concentrated across generations than studies focused on income alone would imply.

This paper studies the role of economic participation via the labor market in enabling the political mobilization of an underrepresented group. Specifically, we study the wives and daughters of disabled Union Army soldiers after the U.S. Civil War. Linking Union Army enlistment records to the 1860 and 1870 U.S. censuses, we find that the wives and daughters of disabled veterans were significantly more likely to participate in the labor force than those of non-disabled veterans. Historical evidence suggests that disabled veterans were also more exposed to postwar alcohol and substance abuse, increasing the household burdens faced by women. Town-level data show that increases in women's labor force participation combined with higher shares of disabled veterans predict more Temperance Crusade activity in 1873-74. Information provision via newspapers and proximity to other protest towns amplify these effects. Using unit-level disability rates as an instrument for veterans' disability status supports a causal interpretation of the labor market effects. Our results suggest that labor force participation can be an important enabling factor for the political mobilization of underrepresented groups.

We construct a new long-run series of subjective expected equity returns from an independent equity analysis firm, spanning 1956-2024. These expected returns are strongly positively correlated with the earnings-price ratio, respond negatively to past returns, and predict future returns. These patterns contrast with subjective expectations of individual investors and professional forecasters, which are weakly or negatively correlated with our new measure. Disagreement between sophisticated and individual investors is associated with higher trading volume. Our findings are consistent with a model of heterogeneous beliefs, where naive investors extrapolate past returns, rather than past dividends, while sophisticated investors are close to rational.