Gender Bias and the Journal of Roman Studies

Reflecting on present unease about structural biases in the discipline, and aiming to offer a data-rich response to some recent criticisms of this Journal, the Editorial Board has undertaken a study of the representation of female scholars in the Journal of Roman Studies. To that end, we have gathered data on publications, submissions and JRS Editorial Board membership for the past fifteen years, from Volume 95 (2005) through to the present volume, Volume 109 (2019). The data are set out in the final section (VII), following a brief review of the main results. Our goal here is neither to present a definitive analysis, nor to offer a commentary on the underlying causes of the patterns revealed (on which we expect much fruitful discussion elsewhere). Rather, the JRS Editorial Board aims to make key data available both to inform a much wider debate within the profession as a whole and, importantly, to inform this Journal’s policies, procedures and active outreach. The Board is also acutely aware that any analysis of gender bias needs to be framed carefully — both by an awareness that there are other under-represented groups in the discipline (on which our data in their current form would regrettably only offer a most imperfect picture), and by a sensitivity to the limitations of a conception of gender as a simple binary.


II EDITOR AND EDITORIAL BOARD
Since 2005, the post of Editor has been held by Alison Sharrock (2005)(2006)(2007), Greg Woolf (2008-2013), Catherine Steel (2014-2017 and Christopher Kelly (2018Kelly ( -2019. The Journal has had a female editor for seven of the fteen years covered by this review. The Editor is assisted by an Editorial Board which conducts the bulk of peer review of submissions. The size of the Board has gradually increased from eight/nine in [2005][2006][2007][2008][2009] to ten in 2010-2017 and twelve in 2018-2019. The average representation of women on the Board (including the Editor) was 41 per cent over the whole period. There is some natural uctuation from year to year as the Board renews itself and tries to maintain a balance of expertise (as well as seniority and geographical distribution), with the annual gure for the representation of women ranging between 33 per cent (in 2005-2016 and again in 2017-2018) and 50 per cent (in 2009-2010, 2013 and again in 2019) (Fig. 1). Overall, the proportion does not seem out of line with the representation of women among mid-career/professorial scholars at UK departments (from which the Board is normally recruited).

III ARTICLES
In recent years (2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016)(2017)(2018)(2019), the Journal has published an average of nine articles per year. Each article is reviewed by two to three members of the Editorial Board (those accepted and published often by ve). Highly specialist pieces are occasionally referred to an external reviewer for an additional opinion, as are all submissions from members of the Editorial Board itself. All reviewing is strictly double-blind: the identity of the author is known to the Editor, but not to any of the members of the Editorial Board reading it (save for a very small number of cases where anonymisation is impossible).
Over the fteen volumes from 2005 to 2019 inclusive, the Journal has published 129 articles, excluding commissioned surveys and reviews, which are discussed separately below. 22 per cent were written by women. For the purpose of this calculation, joint authors are each counted as an appropriate fraction. For example, Lisa Eberle and Enora Le Quéré, joint authors of an article in JRS 2017, each count as 0.5 for the tally. There has been a gradual increase in the representation of female authors over the period. Given the small number of articles published each year, the annual gure is highly volatile: 2018 was anomalously low with 0 per cent, but 2019 jumped back above trend with 40 per cent (Fig. 2). Nevertheless, a gradual upwards trend is clear (see the trendline in Fig. 2). Aggregating the data to smooth out uctuations, The Journal publishes one or two substantial 'review articles' each year, covering particularly important new books or larger clusters of volumes. These 'review articles' are typically 5,000-8,000 words in length, and are directly commissioned by the Reviews Editor in consultation with the rest of the Editorial Board.
Since 2005, the Journal has published a total of nineteen 'review articles'. Fourteen of these 'review articles' were authored by male scholars (in one instance by a pair of male scholars); two were authored by a male-female pair; and three were authored by female scholars. Over the entire period, 21 per cent of 'review articles' have been female-authored (counting each member of a reviewing pair as 0.5 of the review's author). This is approximately equal to the proportion of articles (22 per cent), but signicantly below that for reviews (40 per cent).
A similar imbalance is observable in the substantial 'survey articles' that have in the past occasionally been commissioned to provide a survey of recent scholarship on a particular topic (the most familiar being the series on 'Roman Inscriptions', the last and nal of which appeared in 2012). rates for reviews and articles, it does appear that the Journal's Editorial Board has historically been considerably less willing to commission substantial review articles and survey articles from female scholars.

V REVIEWS
The Journal of Roman Studies has a single Reviews Editor responsible for commissioning book-reviews. Since 2004, the post of Reviews Editor has been held by Christopher Smith (2004-2006), Greg Woolf (2007-2008), Catherine Steel (2009-2013, Christopher Kelly (2014Kelly ( -2017 and Peter Thonemann (2018Thonemann ( -2019. The Journal has had a female reviews editor for ve of the fteen years covered by this review. The Journal typically publishes between seventy and a hundred reviews each year. Reviews of single volumes are normally 900-1,200 words long. Each year the Journal publishes a handful of reviews covering multiple volumes on a single theme, which can be up to 2,400 words in length. The more substantial 'review articles' are discussed separately above. At present the Journal typically carries only a single review from each reviewer per issue, but in the past it was more frequent for individual reviewers to publish several discrete reviews in a single issue; for example, JRS 103 (2013) includes four reviews by Michael Crawford (covering a total of thirteen volumes).
A handful of reviews were co-authored by two individuals. For the purpose of analysis, each of the reviewing pair is counted as 0.5 of the review's author, as for articles and submissions. (In 2018, for example, the number 'seventy-eight' in fact represents eighty discrete reviewers, since one review was co-authored by Valentina Grasso and Garth Fowden, another by Martin Dinter and Astrid Khoo.) Where a single reviewer produced several separate reviews in a single issue, these are counted separately (so Michael Crawford is in fact four of the forty-nine male reviewers in 2013).
Over the entire period 2005-2019, the Journal has published 1,255 reviews (excluding review articles), of which 754.5 (60 per cent) were male-authored and 500.5 (40 per cent) female-authored. Fig. 3 tracks the gender breakdown of authors of reviews between 2005 and 2019. It is noteworthy that the shift from a male to a female Reviews Editor in 2008 coincided with a signicant increase in the proportion of female-authored reviews from 2009 onwards (the time-lag between commissioning and publication is typically between one and two years).
The overall trend is towards a gradual increase in the proportion of female-authored reviews (see trendline in Fig. 3) that appears broadly in line with the proportion of female postholders in Classical Studies in US and UK universities.

VI SUBMISSIONS
Any potential bias in the editorial process can only be identied by comparing published articles to submitted manuscripts (while noting that the balance of submissions may itself be affected by other factors, including the image and reputation of the Journal). The rate of female authorship in the 642 submissions was 30 per cent, almost identical to that of 28 per cent in the ninety-two articles published in volumes 2010-2019 (and somewhat higher than the 22 per cent in [2005][2006][2007][2008][2009][2010][2011][2012][2013][2014][2015][2016][2017][2018][2019]. This indicates that the imbalance in published articles is almost entirely due to a similar imbalance in submissions. There is a small discrepancy in success rates: whereas the overall average acceptance rate was 14.3 per cent, the rate for female authors was 13.0 per cent and that for male authors 14.6 per cent. This discrepancy is not, however, signicantly signicant (the hypothesis of a gender bias is rejected with a chi-squared test at the 0.05 level of signicance). In other words: in a sample of this size, the slight imbalance could easily have arisen by chance, even if male and female authors had equal chance of having their submissions accepted.
The overall gure may nevertheless obscure some chronological developments. As with publications, the annual acceptance rates are highly volatile (Fig. 4). Fig. 5 presents the three-year rolling average success rates to smooth out the uctuation. It can be observed that the rates for men and women are broadly similar and that the rate for women is more volatile year-to-year (as expected, given the small sample size). But there was a period 2014-2017 in which relative success rates were inverted (with 17 per cent for women vs 13 per cent for men, compared to 10 per cent for women and 17 per cent for men in 2010-2013). It may be signicant that this largely coincides with the time in which the Editorship was held by a woman (Catherine Steel in 2013-2016. But any causal effect must have been complex given the double-blind reviewing system. The pattern may also have arisen by chance. The high volatility in 2018-2019 makes more recent trends hard to discern. Overall, female authors do appear to have been signicantly under-represented in JRS articles over the past fteen years (compared to their representation among postholders), but the imbalance seems to be almost entirely attributable to a comparable imbalance in submissions to the Journal. If there has been any gender bias, it appears to have been relatively small and transient. The Editorial Board sees no need to change its current practice for reviewing submissions and certainly no grounds for departing from the principle of double-blind reviewing. Indeed, the results appear to conrm the merits of the double-blind principle in reducing gender bias in reviewing. We are, however, actively reviewing our current policies and procedures as part of a rm and on-going commitment to addressinsofar as we are ablethe imbalance in submissions to the Journal, while recognising that progress also depends on wider discussion and reform in the profession as a whole.