Revisions and FAQ

Referee Comments and Responses

This section describes this paper's path through review. We share the substantive parts of editors' reports (with permission), and explain how they shaped our changes to the paper. We tried to use AI to extract and share the substantive content of the referee reports, but after many attempts, we could not coax even the best language models to adequately summarize the main ideas.

What we submitted (Feb 5, 2025) Manuscript ↗

Similar to submission 4 to AER, but shortened to meet AER:I's length limits.

Editor's comments (Apr 6, 2025)

R1: Reject
R2: Reject
R3: Conditional accept
R4: Reject

I have now received four reports on your paper from knowledgeable referees. R3 is the most positive and recommends that I conditionally accept the paper. The other three all recommend rejection -- R1 and R2 very decisively, and R4 a bit closer to the line. They are mostly concerned with the leaps required to interpret these facts as measures of access of opportunity, as well as various details of the empirical analysis.
After reading the paper myself, I have decided to follow the advice of the more critical referees and reject. I really enjoyed reading the paper. The data work is admirable, and I find the basic facts convincing and clear. However, my view of the interpretation of these facts is closest to that of R1 and R2. To be over our bar, I think the paper would need to show convincing estimates of opportunity gaps rather than just unequal distribution of outcomes, and I don't think what you have quite gets us there.
If I could offer one piece of advice, it would be to be more circumspect in your interpretation. I don't think most readers will come in with the prior that talent is equally distributed in the population (certainly not after people have completed their educations). You currently write as if that is a self-evident fact, or at least a compelling premise.
I'm very sorry to have to deliver a disappointing decision. Our policy of making up or down decisions on the first round leaves us limited flexibility to address major concerns through the revision process. Due to space constraints, we are able to publish less than 5 percent of the manuscripts we receive, and we have to reject many papers that make important contributions.

How we changed the paper

This was a disappointing rejection, because it seemed like the empirical component of our paper would have been above the bar, but the referees felt we had over-interpreted it.

The substantive critique was valid. We began this project from the premise that if half of laureates come from top 5% families, then there is misallocation, there are many "missing Einsteins" and we have fewer scientific discoveries as a result.

The feedback here and in some seminars led us to think about this more carefully, and we realized that you need a number of steps to get to misallocation, for example:

Scientists need to have low substitutability for each other. If anyone with a moderate level of talent can make the next discovery, then who becomes a scientist doesn't matter as much.
Scientific talent has to come from raw individual talent, rather than educational investment, i.e. scientists must be born and not trained.
The relevant raw ability factor can't be too correlated with parent income, or else the right people are already becoming scientists.

In later versions of the paper, we were more careful about these claims, and were more clear about the circumstances in which our findings do and don't lead to misallocation.

This version of the paper had also emphasized an equal opportunity benchmark with the median laureate at the 50th percentile. Referees reasonably argued that any genetic inheritance of cognitive skills would make the benchmark higher, and parental investments in education complicated the story, and it was difficult to know the true equal opportunity benchmark.