Write a comment

Overall Summary:

This preprint compares actual postdoc salaries across the U.S. (sourced from available data for postdocs on H-1B visas) with cost-of-living in various locations, and with salaries for the wider labor market in those locations, to highlight that postdoc salaries do not adjust for cost-of-living. This data is also put into context with a calculation the ratio of tenure-track faculty to students produced in a city. This work provides insight into use of a previously overlooked source of data for postdoctoral salaries and generates interesting new analyses to further research into compensation of early career researchers.

Major comments:

The presentation of real and median salaries with the regional price parities is particularly compelling in Figure 1; likewise in Figure 2 the conversion of annual salaries to real salaries illustrates in particular the “postdoc participation tax” exacted on postdocs carrying out research in the higher cost (and likely higher “prestige”) locations.

The H-1B data from 2015-2020 provides a good approximation of postdoctoral salaries, and a particular strength is the ability to get data from private universities, for which there is no requirement otherwise to make salary data available. The use of this data does come with some caveats, but I am not concerned that these would affect the overall results significantly:

• Postdocs on H-1Bs represent a subset of postdocs, namely foreign postdocs (that is, those who are not U.S. Citizens or U.S. Permanent Residents) who are more likely to be later in their postdoctoral training: that is, it is common for foreigners to progress to H-1Bs after exhausting F-1-OPT or J-1 options (depending on when in their education they came to the U.S.). These options are also cheaper for universities than the H-1B in terms of application forms (and also there is some burden to explain and justify salary levels on H-1B applications, while there may not be an actual requirement to a salary threshold for universities and other non-profit organizations hiring postdocs). So, for example, for a foreign researcher with a PhD coming to the U.S. to postdoc, they are likely to spend 5 years first on a J-1 visa, before then transferring to an H-1B. The sum of all this is that there may be some nuances to this sub-population that may currently be uncertain. For example, we might expect the salaries in this dataset to represent some of the highest postdoc salaries, if we were to assume that salaries at institutions increase with years of experience (which is an expectation, but likely not a common practice). We may also assume that foreign postdocs tend to be paid less than domestic postdocs generally, and so the salaries overall for this group could even be lower. Overall they are still likely a good indicator for the purposes of this work, however, of postdoc salaries generally.
• The range of 2015-2020 includes the major change in postdoc salaries that occurred as a result of changes to accommodate updates to the Fair Labor Standards Act in December 2016, which, though they were never actually enacted, caused an essentially irreversible change in policies at universities and the NIH in late 2016. This dataset therefore likely includes a large increase in salaries within the range. This has been somewhat accounted for by the conversion of salaries to 2021 levels, but it remains the case that salaries in 2015 and 2016 were likely to be substantially lower, and as NIH NRSA stipends have increased at one of the most consistent rates in their history since 2016, it may be worth breaking out analyses by year, as 2015 data in particular may bring down overall averages.

There are further nuances to be considered with the specific population in this sample, but overall this approach and the analyses within the preprint are an excellent insight into postdoc salaries and provide a solid basis for future work.

Minor comments:

The use of the Clauset et al. dataset is helpful for providing an indication of production of tenure-track faculty as a proxy for academic career prospects at an institution. As the preprint points out, postdocs are likely to be a more important determinant of academic career progression success. However this comparison seems fair as a proxy for institutional prestige, that is likely to track across fields, and so . This may also track with recent work demonstrating that 80% of faculty come from 20% of institutions, with one in 8 coming from just 5 institutions (see https://www.nature.com/articles/s41586-022-05222-x#Sec9) and it could be interesting to continue this analysis as data emerges on tracking of outcomes. It would also be interesting to consider the distribution of postdocs across institutions; the data from the NSF’S NCSES GSS (Graduate Student Survey) is a census which includes an attempt to count the number of postdocs at institutions, and while this data is to be used with great care (https://www.biorxiv.org/content/10.1101/171314v3) the overall data on number of postdocs could be interesting, as postdocs are distributed in a very biased way, with most postdocs being at a handful of research intensive institutions.

One emergent conversation since the publication of this preprint is the “Great Postdoc Shortage” reported by faculty seeking to recruit postdocs. This is likely in part due to increased demand for those with biomedical and/or computational skills in biotech and the wider data analytics/data science labor demand - it may be that some of the H-1Bs in the dataset are in industry/other roles, and not at universities, and it could be interesting to see what their salaries are and how they track. Obtaining real salary data for these roles in the wider labor market may be difficult, but it could be possible to find salary expectations in those roles and compare them with postdoc salaries to see how wage disparities may be affecting academic retention. This is discussed briefly in the discussion already; and it is the case that there has long been a pay disparity, but perhaps a recent change in demand/job availability in these better-paid roles. This would be interesting to uncover to see how much effort academia may have to expend to retain talent. There is already a comparison for this, in computer science postdocs, who are paid much more than any other discipline given their recent employability in the private sector in recent years.

Some more detail on where the data was extracted from, and how, would be very helpful, for those wishing to undertake future similar analyses - this may be in the Github repository but it could be helpful in the text for the methods too.

Conflicts:

I have no conflicts to report; I have discussed this work with the author previously but did not know the author prior to seeing this work. I have not been involved in any way with the current version of the work that is published here. I do not stand to gain or suffer financially or otherwise from this publication.

License:

This review is published under a CC-BY license.