How do their editing incomes compare ?

Background: The work of journal editors is essential to producing high-quality literature, and editing can be a very rewarding career; however, the profession may not be immune to gender pay gaps found in many professions and industries, including academia and clinical medicine. Our study aimed to quantify remuneration for journal editors from core clinical journals, determine if a gender pay gap exists, and assess if there are remuneration differences across publishing models and journal characteristics. Methods: We completed an online survey of journal editors with substantial editing roles including section editors and editors-in-chief, identified from the Abridged Index Medicus “Core Clinical” journals in MEDLINE. We analyzed information on demographics, editing income, and journal characteristics using a multivariable partial proportional odds model for ordinal logistic regression. Results: There were 166 survey respondents (response rate of 9%), which represented editors from 69 of 111 journals (62%). A total of 140 fully completed surveys were analyzed (95 males and 45 females); 50 (36%) editors did not receive remuneration for editorial work. No gender pay gap and no difference in remuneration between editors who worked in subscription-based publishing vs. open access journals were detected. Editors who were not primarily health care providers were more likely to have higher editing incomes (adjusted odds ratio [OR] 2.96, 95% confidence interval [CI] 1.18-7.46). Editors who worked more than 10 hours per week editing earned more than those who worked 10 hours or less per week (adjusted OR 16.7, 95%CI 7.0239.76). Conclusions: We were unable to detect a gender pay gap and a difference in remuneration between editors who worked in subscription-based publishing and those in open access journals. More than one third of editors surveyed from core clinical journals did not get remunerated for their editing work. Open Peer Review


Introduction
The number of academic journals continues to grow each year. In 2018, there were 5399 clinical journals tracked by Journal Citation Reports in comparison to only 3681 10 years prior 1

.
A rise in open-access journals is also evident; while in 2008 there were just 249 open-access journal titles in Web of Science, that number ballooned to 1431 in 2018 1 . Growth in the journal industry comes with more opportunities to become a journal editor. Editors are often scientists, researchers, administrators, and clinicians who have expertise in a particular field and competencies in evaluating articles' suitability for publication in scientific journals 2 . Editorial work can be a very satisfying full-time career or part-time job in addition to a clinical, research, or administrative careers.
Within medicine, substantial evidence of gender inequity in academia exists, including disparities in compensation [3][4][5] . For example, a 2012 US survey found that male physician researchers had higher average salaries ($13,399 USD; p=0.001) than females after adjusting for specialty, academic rank, and research productivity 6 . Other studies have shown salary deficits can be up to 12-28% for female physicians 7,8 when compared to male counterparts. Similar statistics can be found in the UK, where the average pay gap for consultant physicians, including professors, was 13% in favor of men 9 . Several countries have mandated pay transparency laws in recent years. Denmark 10 and Austria 11 passed laws requiring companies to internally report gender-based wage statistics to it employees in 2006 and 2010 respectively. In 2018, the United Kingdom pushed for national pay transparency and equity by mandating public annual reporting of gender pay gap for any organization comprising more than 250 employees 12 . France followed suit with mandatory reporting, while Germany and Iceland enacted pay equity and transparency laws.
While there is an expanding body of literature on pay gaps in of academia and of medicine in general 9,13 , there is currently very limited literature that evaluates how journal editors are compensated for their work. An international email survey of 88 editors of nursing journals found that their mean annual salary was $12,749 USD (ranging from $0 to $56,000) for a mean of 13.4 hours worked per week 14 . A total of 8% of survey respondents (7 of the 88 editors) did not receive any monetary compensation and only 31% of participants felt that their compensation was adequate 14 . It is concerning that the critical job of editors to uphold the integrity of academic literature can be low-paid or voluntary. With mandatory data reporting, The Lancet reported gender pay gaps of 13-40% favouring men over women in major UK publishing companies in 2018 15 . Specific data for journal editors were not available but some of these data do highlight the potential gaps.
There is substantial heterogeneity among journals, such as publishing platform, scope, publication frequency, and Journal Impact Factor 16 . Though open access publishing comes at a large cost to authors, its popularity is supported by many benefits that may include faster publishing times and the ability to reach bigger audiences compared to subscription-based publishing. It is unknown whether a journal editor's remuneration is affected by these journal variables.
With special interest in biomedical sciences, the objectives of this study were to quantify remuneration for journal editors from core clinical journals, determine if a gender pay gap exists, and assess if there are remuneration differences across publishing models (e.g., subscription-based or open-access) and other journal characteristics (e.g., publication frequency and Journal Impact Factor).

Design and participant recruitment
We completed an international online survey of full-time and part-time journal editors identified from the Abridged Index Medicus "Core Clinical" journals in MEDLINE, which represented 111 peer-reviewed core medical journals with high clinical impact. Our target population of journal editors were those with substantial editing roles including editors-in-chief, deputy editors, executive editors, senior editors, associate editors, and editors of a specialty section. Editors were identified through each journal's webpage in December 2018. We excluded roles classified as statistical editors, assistant editors, international editors and editorial board members. If the editors of a section or specialty section numbered greater than 200 for a single journal, these individuals were excluded based on the assumption that they may not have a substantial editorial role. Publicly available emails were found via the English-language search engine, Google. Major sources included journal web pages, academic institutional web pages, and corresponding author on recently published articles. Reporting of this online survey was guided by the CHERRIES reporting guideline 17,18 .

Survey development and pre-testing
The online survey was conducted using the subscription software, Qualtrics CoreXM Survey Tool 19 . We developed the survey to capture demographic data, editing remuneration in USD, editing experience, and journal characteristics. A blank copy of the survey is available as Extended data 18 . An online pilot test was sent out to our knowledge user team of three journal editors to identify poorly constructed questions, and to assess face validity before distribution. The survey included 10 questions on three pages, and adaptive questioning was used. Respondents were able to review their answers before submission. The following variables were collected in our survey and included in the final multivariable model: sex (male vs. female), gender identity, primary role (health care provider vs. other), academic rank (any professorship vs. none), editing role (section/specialty/associate editor vs. editor-in-chief/executive/ senior/deputy editor/other), years in editing (>10 vs. ≤10 years), hours/week in editing role (>10 vs. ≤10 hours), and years worked for current journal (>5 vs. ≤5 years). Journal characteristics of publishing model (subscription-based, vs. open or hybrid [open access option or open access for developing countries]) and frequency of publication (best fitted to monthly/bimonthly vs. weekly/biweekly) were extracted from the journal selected in the survey. A journal's 2017 Journal Impact Factor was obtained from Journal Citation Reports 1 .

Survey administration
Email invitations with the survey link were distributed via Qualtrics in February 2019. This was a voluntary survey, and no incentives were offered. We employed established methods to enhance survey completion rates with reminder emails at week 2 and week 4 20 .

Statistical analysis
Only completed surveys were analyzed. Dichotomous baseline characteristics for male and female journal editors were presented as frequencies and percentages and compared with the chi-square test or Fisher's exact test when expected sample sizes were 5 or less. Journal Impact Factor of the journal where male and female editors worked was a non-normally distributed continuous variable presented as a median with interquartile range. The Journal Impact Factor of each group was compared with the Wilcoxon rank sum test.
Our outcome of interest was journal editor salary, which was modeled as an ordinal variable with three categories: ≤$10,000 per year, $10,001 to $50,000 per year, and ≥$50,000 per year. We derived adjusted odds ratios (OR) and 95% confidence intervals (CI) for our outcome of interest from a multivariable partial proportional odds model for ordinal logistic regression in SAS version 9.4 (SAS Institute, Cary, North Carolina). The variables for editor sex and academic appointment did not satisfy the proportional odds assumption; therefore, these variables were assumed to have nonproportional odds in the final multivariable partial proportional odds model. Two-sided p-values were reported and p-values <0.05 were considered statistically significant.

Ethical approval
Institutional review board approval was obtained through the Unity Health Toronto Research Ethics Board in Toronto, Canada. The survey landing page included study information and consent to participate was implied by survey completion and submission.

Results
A total of 2165 editors were identified in December 2018, with 1948 having publicly available email addresses; of these, survey invitations were successfully sent to 1844 without bounce-back. A total of 193 surveys were started, and 166 surveys were submitted (overall response rate of 9%). We received survey responses from journal editors at 69 of 111 journals (62%). 140 completed surveys were included in our analyses ( Figure 1). Data that had small cell counts (n<5) were not reported to preserve the privacy of participants. A de-identified version of the dataset is available as Underlying data 18 . This was composed of 95 male and 45 female editors. All respondents identified as cis gender. A total of 50 (36%) editors did not receive remuneration for editorial work; 111 (79%) and 29 (21%) editors worked for a subscription-based journal versus an open/hybrid journal, respectively. A total of 90% of editors held an academic position. The median 2017 Journal Impact Factor was 4.9 (Interquartile range [IQR] 3.5-6.6 for males and 3.5-7.5 for females). There was a larger proportion of female survey respondents who were fulltime journal editors compared to male counterparts (10 females   In multivariable analysis (Table 2), there was no gender pay gap detected or remuneration differences identified between publishing models. We found that editors who were not primarily health care providers were more likely to have higher editing incomes (adjusted OR 2.96, 95%CI 1.18-7.46). Editors who worked more than 10 hours per week editing earned more than those who worked 10 hours or less per week (adjusted OR 16.7, 95%CI 7.02-39.76).

Disc ussion
In addition to a recent study by The Lancet highlighting the presence of a gender pay gap of employees at major academic publishing companies 15 , numerous studies in the past have highlighted similar inequities across academia [3][4][5][6][7][8] . In our study of part-time and full-time journal editors from a focused subset of core clinical journals, we did not detect a gender pay gap. We found that annual remuneration for editing was higher for editors working more than 10 hours per week and if their primary occupation was not a health care provider. Full-time journal editing positions were disproportionately more likely to be held by a woman. While we hypothesized that the different revenue structure of subscription-based versus open access journals may translate to a difference in editor's remuneration, we did not find it to be true after adjusting for predictor variables. A journal's Impact Factor did not affect an editor's remuneration. Although 36% of editors surveyed reported no direct earnings from their editorial work, there can be other benefits including subsidies for scientific meeting registration fees and travel costs.

Limitations
While there is robust data on gender pay gaps in academia 13 , our study only examined gaps specifically in biomedical sciences. Our study had several limitations. First, our results had limited generalizability due to a low response rate, which can magnify volunteer bias. However, our participants did represent 69% of the core clinical journals. Moreover, our sample size of 140 was comparable to an international survey of 148 scientific editors from biomedical journals to evaluate journal editing core competencies 21 . We were unfortunately not powered to compare subgroups of editing roles and journal characteristics. Second, given our approach to identifying email addresses for journal editors, our survey was biased towards editors who had either an English-language academic profile on an institutional website or if they were a corresponding author of a recently published article. Third, we were unable to make direct comparisons between for-profit and non-profit journals, because several companies publish journals on behalf of nonprofit organizations. This distinction may factor into a journal editor's desire to edit for a journal at a given remuneration rate. In addition, we did not assess non-financial reasons that factor into the decision to be a journal editor. Lastly, there was substantial heterogeneity in editing roles and the lack of standardization in editing titles made it difficult to draw conclusions that were generalizable to the population. Future studies can include qualitative interviews to gauge the various roles and responsibilities of editors, and remuneration practices.

Conclusion
We conducted an international survey of journal editors from core clinical journals to understand how remuneration varied across editor's demographics, professional experience, and journal characteristics. We did not detect a gender pay gap or a difference in remuneration between editors who worked in subscription-based publishing vs. open access journals. More than a third of editors surveyed were not remunerated for their work. fitted to monthly/bimonthly vs. weekly/biweekly), were extracted from the journal selected in the survey." I would specify where journal impact factor was extracted from -many journals no longer list this information on their website. In these cases was this coded as missing? Or, was the data searched for elsewhere?

Data availability
Competing Interests: I co-authored a paper with the third author (Diane Kelsall) in Sept 2017. This paper was a team science project with 31 authors. The Centre for Journalology, to which I am a member, has hosted Diane Kelsall to teach on a regularly held 'Introduction to Journalology' course for trainees. © 2020 Singleton C. This is an open access peer review report distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Carl Singleton
Department of Economics, University of Reading, Reading, UK This study aimed to quantify gender differences in pay for editing services within academia, specifically for "Core Clinical" journals. It also aimed to test for the presence of a gender pay gap, conditional on other observable (surveyed) characteristics of editors and their journals.
Overall, the article is well-written, the methods are clearly explained, and some of the limitations are fairly outlined.
Cobey (2020) has already provided a thoughtful review of this article, and I would echo those comments.
In addition, I would like to make the following further suggestions: In the introduction and the limitations, the authors may wish to dwell on the other reasons that editors choose to do what they do, besides direct financial gain (e.g., enjoyment, prestige, compensation from their main employer in the form of relief from teaching or other activities). These motivations are all potentially endogenous factors (omitted variables) in the regression model. While in theory they may not be correlated with gender in the population of editors, the sample size here is small and the response to the survey probably non-random.

1.
The literature cited for the gender pay gap is in academia is limited and very specific, focused on medical sciences. But there are good examples and literature reviews of gender pay gaps in academia more widely. One recent example I am aware of is Mumford & Sechel (2020) 1 in the British Journal of Industrial Relations. While that study focuses on the gender pay gap among economists, it also provides some prominent examples of the wider literature on the gender gaps in academia, including in Medicine and Science (e.g., Connolly and Holdcroft, 2009 2 ). The authors could look to reference more of this previous work, particularly since this journal is not field-specific.

2.
pay (and hours). While I appreciate that in this field odds models are common and often preferred, in my own field, as an applied economist/econometrician, we would prefer not to lose all the information contained in the pay data collected. For example, I would suggest also estimating a censored linear least squares regression model (i.e., tobit, censored because of the zeroes), where the mid-point of each and every range in the survey is imputed as the pay value, which becomes the dependent variable (normally in natural logs). Similarly, combining the narrow pay bands and narrower hours bands in the survey could allow the authors to look at a measure of hourly pay, which is what matters most economically, being the compensation for a unit of labour input (see Mumford and Smith, 2007 3 , for a relevant example of extracting more information from ordinal pay and hours data in this suggested way). Results following this alternative empirical strategy could be presented as a complement to those already shown in the article.
In the current ordinal logistic regression model approach, I would be curious what interacting gender with some the other covariates shows. For example, the estimation sample shows a large difference between the male and female in the proportion of editors whose main occupation is Health Care Provider. This variable looks as though it could be worth interacting with gender. This could be added as a third column of results in Table 2.