Introduction

Our study makes two primary contributions. First, we develop a parsimonious quantitative measure for summarizing and examining changes in access to scientific opportunity over time: the childhood SES rank of top scientific achievers. We focus on the most elite scientists making breakthrough discoveries, for whom misallocation could be most socially consequential (Romer, 1990; Aghion and Howitt, 1992; Galor and Tsiddon, 1997; Griffith et al., 2004; Akcigit et al., 2017) . Second, we develop a long time series and provide new evidence that access to scientific opportunity has expanded substantially from the 19th century to the post-war era.

The study of parental occupations of scientists goes back as far as Galton (1874) and has been of perennial interest. Prior work has shown that scientists' parents are disproportionately likely to have PhDs, white collar jobs, or work as educators or in the sciences (Zuckerman, 1967; Berry, 1981; Akcigit et al., 2017; Morgan et al., 2022) .3

In related work, Stansbury and Rodriguez (2024) study current relationships between career success in academia (and specifically in economics in Stansbury and Schultz (2023) ) and parental education, finding a substantial class gap, but do not examine changes over time. Several papers study racial and gender gaps in academic advancement and recognition (Ginther et al., 2011; Ginther et al., 2018; Sarsons et al., 2021; Hengel, 2022; Ross et al., 2022; Koffi et al., 2024) .

There is a long literature on intergenerational mobility, which has most often focused on measuring average outcomes for children born into poor families (Solon, 1992; Clark and Del Bono, 2016; Chetty et al., 2014; Zimmerman, 2019; Jia and Li, 2021) . Access to opportunity in the sciences is particularly important because of the high social value of scientific discovery: identifying and attracting more talented people into the sciences could affect the direction and quantity of research itself (Murphy et al., 1991; Hsieh et al., 2019; Hoogendoorn et al., 2013; Freeman and Huang, 2015; Hunt and Gauthier-Loiselle, 2010; Gaulé and Piacentini, 2013; Koning et al., 2021) .

Several other studies provide useful benchmarks for our findings. In concurrent work, Abramitzky et al. (2024) estimate that U.S. academics at U.S. universities in 1900–69 had an average father SES rank of 72. Airoldi and Moser (2024) report the median U.S. scientist in 1921 had a 78th percentile father, with the median 'star' scientist (rated among the 15% most successful by their peers) ranking slightly higher. At the 89th percentile, the average Nobel laureate is more elite than these broader groups. Laureates are also from more elite families than inventors; Akcigit et al. (2017) and Bell et al. (2019) show mean U.S. patent-holder ranks in the 60s and 70s, with little SES gradient in patent importance.4 Relative to prior work on scientists, to our knowledge ours is the first to measure how their socioeconomic representation has changed from the late 1800s to the present; even in the mobility literature, it is rare to find a 123-year time series with a comparable measure of success or childhood SES over time in a socially important domain.5

Our work also relates to a broader literature connecting childhood and early career circumstances to scientific achievement. Bell et al. (2019) and Aghion et al. (2017) study the childhood socioeconomic ranks of U.S. and Finnish inventors, respectively, in a similar quantitative framework to ours, while Kim and Moser (2021) examine the role of motherhood in scientific careers. A related literature studies the networked relationships between scientists who receive major prizes (Ma and Uzzi, 2018; Jin et al., 2021) .