Frequently Asked Questions

Not yet! The paper is conditionally accepted at the Review of Economics and Statistics. But we think researchers should be publishing open data and exhibits like this before their papers get published. The kind of visibility we’re promoting here should be essential to the review process. In fact, we did already share an interactive online appendix with reviewers that included the raw data and some of the other exhibits in here.

Authors might be afraid to do this—what if reviewers find and harp on the one sub-sub-result in an extraneous exhibit and use that as grounds to reject the paper? Hostile review is a risk, but we think authors should get credit for being maximally transparent in their work. We want to get to a world where everyone recognizes that making every detail of your work visible in fact raises the credibility of the research.

We can think of many reasons why this would be true. Business owners (the most common laureate father occupation) can be very rich or very poor (we observe both in the data). A high-SES father can choose an occupation with low average SES (like Carl Wieman, whose highly educated father took a job as a sawyer in Oregon).

Father’s occupation is nevertheless widely used as a proxy for SES in economic history papers, because it is one of the few correlates of household SES that is widely available in historical data. It was perhaps most famously used by Long and Ferrie (2013) in their study on Intergenerational Mobility in Great Britain and the U.S. since 1850.

Almost every social science paper uses some imperfect proxy like this that is correlated with the outcome of interest but is subject to some measurement error. We acknowledge the inevitable uncertainty, but don’t see a strong reason to think that where laureate fathers sit within their occupation SES distribution has changed substantially in either direction, but we can’t rule it out.

This is undoubtedly a source of error, but is less concerning when you consider that 576 out of 714 of the laureates in our sample come from USA or Western Europe, which we expect to have broadly similar occupational distributions. When laureates come from countries that were much poorer, we use the occupational distribution from older U.S. years that were more likely to be comparable to those countries. The transition of occupational structure with economic growth is a consistent and predictable pattern across countries.

Nevertheless, we acknowledge that this is a source of measurement error, as is the use of father’s occupation as a predictor of SES. We would of course prefer to use different occupational distributions for all countries if we had them. 

We also post several results just by occupation, which show a similar story. Table A2 shows that children of technical professionals are over-represented by a factor of 12, while laborers and farm workers are similarly under-represented. Table A4 shows how these over-representation ratios have changed over time.

We agree that genetic traits can create a correlation between parental income and child scientific success. Is this link strong enough to fully explain the unequal distribution of Nobel laureate childhoods?

We discuss this in the concluding discussion of the paper. The answer depends on (i) the heritability of cognitive skills; (ii) the extent that heritable cognitive skills matter in producing parent income and child scientific success; (iii) the probability that children with high genetic cognitive inheritances choose to enter the basic sciences (rather than engineering, or business, or any other domain that rewards cognitive skills); (iv) the failure rate of very good scientists (not every brilliant scientist enters the right field or wins a prize), among other factors.

In an earlier draft of the paper, we tried to proxy all of these factors and ran a simulation to see what kind of distribution of laureate childhoods you would get from genetic factors alone, using heritability estimates from the literature. Our conclusion was that you need relatively extreme assumptions to reproduce the empirical distribution of laureate childhoods; genes could explain 25-30% in our mainline scenarios. Referees found this too speculative (it’s true, it was speculative), but we’re working on refining and posting it as we think it’s interesting and worth discussing.

The genetic argument against our findings is least compelling on the cross-country dimension. Fewer than 3% of laureates were born in Africa (N=9), India (N=7), or China (N=6), who now represent a combined 54% of the global population. It seems extremely likely to us that there are a large number of individuals born in these regions who would have the genetic capability to become Nobel laureates if pathways to scientific opportunity had existed for them.

This was the question that motivated our work on this paper. We think access to opportunity is particularly important in the sciences, because important discoveries benefit all of humanity.

In working on the paper, we realized that you need more assumptions to make this claim definitively, so the current revision of the paper is more measured on this point.

For example, if scientists are made and not born, then it doesn’t matter who becomes a scientist. You just need enough sufficiently talented people to enter the training pipeline. It’s probably true that some discoveries don’t depend on a single genius way ahead of their field; whenever there are discovery races, it means that more than one scientific team could have found a major discovery. For some laureates—e.g., Einstein, Feynman, Schrödinger—it’s harder to imagine that a different highly talented person would have been just as productive in their positions.

We would have loved to put a statistic in the paper on the number of additional scientific discoveries that could have been made if the world was perfectly efficient at identifying and nurturing early scientific talent. But this would take so many assumptions that it wouldn’t be credible. So we just share the highly unequal distribution of laureate childhoods, and let readers interpret them through their own assumptions. There is a fantastic and important literature on identifying and nurturing early outlier talent; you can read some of the current papers in this space here.