Non-white scientists spend more time under review – PNAS

Black scientists wait 22% longer to hear if their papers have been accepted for publication than do white authors of similar papers published in the same journal in the same year, says a new study published in the Proceedings of the National Academy of Sciences (PNAS) last month.

The study also found that even in Asia, Africa and South America, where the majority of the population is non-white, there are fewer non-white editors than would be expected relative to their share of authorship.

Further, PhD student Fengyuan Liu and Associate Professor Talal Rahwan, both of whom are affiliated with the computer science division of New York University Abu Dhabi, and Bedoor AlShebli, who teaches in the university’s social science division, found that black and Hispanic scientists are cited significantly less than are comparable white authors doing similar work.

The study titled “Non-white scientists appear on fewer editorial boards, spend more time under review, and receive fewer citations” analyses more than one million articles from 500 different journals edited by almost 65,000 editors and published by six publishers between 2001 and 2020.

The six publishers are Frontiers, Hindawi (a unit of John Wiley and Sons), the Institute of Electrical and Electronics Engineers (IEEE), the Multidisciplinary Digital Publishing Institute (MDPI), Public Library of Science (PLOS) and the PNAS.

“Many studies analyse much smaller datasets, consisting of only a few hundred data points. Such datasets are typically insufficient to provide conclusive evidence about certain phenomena,” says Rahwan.

“The new field of computational social sciences uses sophisticated techniques that can be used to analyse massive datasets such as ours. This emerging field is bringing new insights, unravelling findings that were hiding in plain sight, waiting for the right type of analysis to be revealed.”

After noting in the introduction of their paper that other studies have “found that editorial boards are dominated by scientists from North America and Europe”, the first part of Liu et al’s study provides a longitudinal examination that sets out the racial composition of editorial boards of each journal in relation to the racial make-up of the scientists published in that journal.

The six-colour map that shows percentage of authorship to editorship strikingly depicts the imbalance. North America, Oceania, parts of Western Europe and a few other parts of the map are green, while most of Asia, Africa and South America vary between beige and brown. These last colours signify the extent to which scientists from these non-white countries are under-represented on editorial boards.

In percentage terms, 35% of articles in these countries are written by scientists who live in these countries – but only 19% of editors are from these countries. Malaysia, China and South Korea have less than half the number of editorships that can be expected based on authorships, while Saudi Arabia, Mexico and Poland have significantly fewer editors than could be expected by the number of scientist authors.

When broken out by the publisher, the disparities become even more obvious. At MDPI, 93% of Asian, African and South American countries are under-represented on editorial boards, in contrast to 55% of American, Oceanian and European countries being over-represented. At Hindawi, the figures are 84% and 80%.

Using Google Scholar, Rahwan and his co-authors identified the top 20 journals in 19 disciplines. There were almost no editors from Asia, Africa and South America in the environmental science, economics, history, geography and medical journals examined.

The disciplines with the highest number of editors from Asia, Africa and South America were physics and art, with approximately 25% of the journals examined having editors from outside North America, Europe and Oceania.

A central finding of the study is just how skewed the racial makeup of the editorial cadre is. It is not difficult to imagine a critic, especially in the United States, seizing and admitting that NamePrism, the algorithmic tool they use to determine a scientist’s ethnicity or nationality, “is not a perfect classifier” and then arguing that their research is fatally flawed.

However, as the authors note, NamePrism is uncontroversial amongst social scientists. Further, the outsized percentage Rahwan and his co-authors found of white editors neuters such a naysaying argument.

Fully 57.5% of editors are white. Asians and Pacific Islanders account for 39.6%. Hispanics make up 2.8% of the editorial cohort and blacks 0.1%.

Even when Liu et al account for the fact that in recent decades blacks and Hispanics have made great gains in academe, the figures do not markedly change for blacks, who in 2020 account for approximately 0.2% of editors, while Hispanics account for slightly below 4%.

Citation distortion

Citation distortion – that is, citations not being distributed in accordance with textual similarity of papers submitted by pairs or groups – is much more than an academic complaining that her or his work has not been noticed. Citations count for PhD students applying for positions, promotion and tenure decisions and for grants and awards.

As Rahwan and his co-authors note, in 2020 in “Racial and ethnic imbalance in neuroscience reference lists and intersections with gender” (in bioRxiv), Maxwell A Bertolero et al reported that papers on which the first or last author had a white surname were cited 5.4% more than expected, while articles that have non-white first and last-named authors were cited 9.3% less than expected.

Even more strikingly, while scientists with East Asian last names make up between 14.3% and 33.6% of relevant authors in the journal Nature, they make up less than 7.7% of quotes in non-research articles, as NR Davidson and CS Greene reported in an article in bioRxiv in 2021.

Liu et al significantly deepen the analysis of citation disparity because, instead of looking at bibliometric characteristics (for example, publication year and venue, and the author’s discipline and academic age), they employ a sophisticated analytical tool called “citation lensing”.

Citation lensing begins with the assumption that the more textually similar papers are, the greater the probability there is of having one of them cite the other. In the context of the “Non-white scientists” study, Rahwan explained, “given the textual similarity of any two groups of papers (for example, those written by white authors and those written by Hispanic ones), we measure the number of citations that one would expect to see from one group to the other.

“Then, by comparing the actual citation rate to the expected one, we can quantify the degree to which any given group is over- (or under-)cited.”

To obtain a more holistic view, in this part of their analysis, Rahwan and his research partners focused on all US-majority papers in Microsoft Academic Graph, which provides records of 200 million scientists.

Their findings are striking. For the past four decades black and Hispanic scientists have been under-cited, while white and API authors have been consistently over-cited relative to what would be predicted by textual similarity of papers.

The data shows “that this phenomenon persists across four types of disciplines, namely 1) biomedical, behavioural and ecological sciences, ii) engineering and computational sciences, iii) physical and mathematical sciences, and iv) social sciences”.

Indeed, the gap in social sciences, which might be thought to be more accepting and supportive of black and Hispanic authors, is actually the largest.

“Citation is a proxy of how much attention the work of a scientist is garnering, and how much impact their work is making. While citation is an imperfect measure for attention and impact, it is heavily adopted by hiring committees and the general public as a measure to evaluate the ‘quality’ of a scientist,” says Rahwan.

Citational distortion, thus, becomes a contributor to the feedback loop that depreciates black and Hispanic scientists: fewer citations are taken to mean less important or original work which leads to the perception of lower merit, which leads to fewer citations.

Relative acceptance delay

Since the “relative acceptance delay” (RAD) varies by country – with Uganda, Vietnam, Mexico and Brazil being among the 20 countries with the longest delays – to eliminate what Rahwan and his co-authors refer to as the “confounder of countries”, in examining the impact of race on RAD, they focus on the US.

They justify this decision, further, by underlining that the US is a racially heterogeneous country and that it “contributes the largest number of papers to our dataset”.

Here, again, Liu et al’s data show that black scientists are significantly disadvantaged in relation to white scientists. The RAD for papers of similar authorship submitted by scientists with identifiable black surnames is 20% higher. In other words, these scientists wait 20% longer to hear if their papers have been accepted by journals.

While studies such as Bertolero’s et al’s “Racial and ethnic imbalance” provide evidence for an ‘old boy’s network’ that advantages white male scientists, Rahwan and his co-authors’ RAD data find no such evidence for non-white authors when their manuscripts are in the hands of non-white editors.

The reverse is, in fact, the case. The average RAD for papers submitted by non-white scientists to a journal with a non-white editor is six times longer than it is when a non-white scientist submits her or his paper to a journal with a white editor.

Scientists are trained to be dispassionate researchers, as exemplified by refusing to speculate, writing (as much as possible) in the passive voice and focusing on the data presented in the paper.

It is to Liu, Rahwan and AlShebli’s credit, and an indication of the power of what the longitudinal study reveals, especially about the RAD black scientists endure, that they situate the RAD in the racialised structure of American society.

Waiting, they show, is almost a way of life for black Americans. In the 2016 election, “residents of entirely black neighbourhoods spent more time waiting at voting stations compared to residents of entirely white neighbourhoods,” as they did, we should add, in the 2018 and 2022 midterms as well as the 2020 presidential election.

Blacks wait longer in emergency rooms because they are “less likely to be placed into the ‘Most Urgent’ category of the Emergency Severity Index”, which leads inevitably to blacks dying in the waiting rooms at a greater rate than do whites. In both the US and Britain, they note, blacks wait a significantly longer time for cancer diagnoses, which means they wait longer for life-saving treatment.

By placing their findings about the RAD of papers written by black scientists in this broader sociological context, Rahwan and his co-authors do more than indicate it is simply another instance of a phenomenon so common in the US that it is all but invisible – save for the long lines shown on television on election night.

Rather, they link their analysis to the fight for racial justice as they echo Martin Luther King’s “Letter from Birmingham Jail” written in 1963. “For years now I have heard the word ‘Wait!’ It rings in the ear of every Negro with a piercing familiarity. This ‘wait’ has almost always meant ‘never’. We must come to see with the distinguished jurists of yesterday that ‘justice too long delayed is justice denied’.”