Review History
Underrepresentation of women in the senior levels of Brazilian science

Thanks for your careful attention to the minor corrections.

Thank you for your careful attention to the feedback on your original submission. The PeerJ system generates the main letter, but I am pleased to tell you that I am happy to accept the paper subject to the minor revisions proposed by reviewer 2, plus attention to the following minor points:

line 225: refer to 'rows' rather than 'lines'
line 355 'in particular'
line 372 delete comma after 'areas'
lines 388-389: the quote has a type , need for -> of, ie: 'ensure they do not opt out of inorganic fields...'
line 400: environments

The reviewers have commented on aspects of the paper where greater clarification is required and I think you will find their comments helpful in indicating where some revision is needed. I have some minor comments about stylistic issues below.

My main response to this paper, however, is that the dataset has potential which is not fully exploited. You could analyse the data to give more of an indication of possible reasons for the gender differences that you describe. But I also have a concern that in places the presentation of results may be misleading.

As you note, a sharp decline in the percentage of women in science as one advances in seniority has been reported in other countries, but there is a range of explanations. These are not mutually exclusive, but include a) effects of implicit or explicit bias against women in competitive situations, b) women’s self-exclusion from competition, and time out for maternity etc. c) It has also been suggested that the very fact of being in a minority makes women more likely to drop out. d) In addition, it is important to remember that there may be generational effects: if the more senior grades are people who started careers a long time ago, then there may have been different pressures and opportunities for them. I think you should spell out these options: even though an observational study like this cannot unambiguously choose between them, there will be clues in the data.

My concern is that, at several points you make statements like “Female scientists were more frequently awarded PQ scholarships at the lower level of the system” etc (Abstract and also para 1, p 10)– this makes it sound as if the scientists apply for a PQ and then the BNC decides the level of award. My impression from your account, though, is that the specific level of award is very dependent on stage of career. It is a subtle difference, but my concern is that it is misleadingly described, as if it is the funder who decides the award level of an applicant. I suspect that, if this is similar to other countries, the applicant decides to apply for the award. This would mean that the success rate at different grades will be a function of two things: the proportion how many men and women apply at each grade, and the success rate at each grade. It really is not possible to tell from such data whether there is bias against women unless you have data on the number of applicants of each gender, and so can see if success rates differ by gender. I think this point needs to be made explicitly to avoid confusion. A related concern was the account of lower funding for women in para 2 of p 10. See point 2 below.

Despite this ambiguity, I think there are interesting features in the data which could be used to throw light on possible explanations.

1. First you have already compared patterns of gender imbalance across three broad disciplines, and these do give very different pictures. As reviewer 2 points out, the presentation of data in Table 1 is unhelpful, as what we need to see are the percentages by rows (i.e. the % female) and not by columns. This would make it easier to see that in ‘exact sciences’ you start with a low percentage of women at the lowest grades, which just declines as they become more senior. In life sciences, it starts with a fairly equal gender distribution at the lowest grades, but women again drop off as positions become more senior. This is not the case, however, for the humanities, where the initial roughly equal split between male and female at the most junior levels remains pretty constant as the grades move up in seniority. This is an important result, as it would seem to rule out any general explanation of the gender gap at senior grades being due simply to women selecting themselves out because of competing pressures of motherhood etc – these presumably would be similar in all disciplines.

2. The second opportunity that you have with these data is to consider whether the differences in grant income between genders are explicable in terms of seniority or whether there are additional biases. The most direct way to do this would be to link the information about grant level to the information about level of scholarship. I looked at your data and found you could match the names from the PQ Holders to the Edital Universal dataset for about 16% of cases – which is still a large sample size given the overall numbers. This means that, for those where PQ stage is known, you can compare income levels by gender and PQ stage, to see whether the lower income for women is explicable by their lower PQ stage. As far as I could tell from a brief look at the data, this was a sufficient explanation: i.e. it looked as if women had lower grant income because they were more likely to be at an earlier career stage. This is not entirely clearcut, because it’s unclear whether those who do not appear in the Edital database are people who applied for funding unsuccessfully or if they did not apply. But it suggests that there is no additional gender bias when funding decisions are made.

I think if you consider these further points, it would make a stronger paper. In particular, the striking differences in retention between Humanities and Life Sciences, both of which start from a similar baseline % women, raises questions about what is different for women in these two broad disciplines.

Yet another point considers the analysis of the ABC (p 8-9): if this is an elite institution, then one would predict that the people who are elected would come from the highest level of PQ, i.e. the 1A band. You note that the % of women in the ABC is lower than the % women with PQ. But if you are looking for evidence of bias, it would seem appropriate to focus on % women in band 1A . Here, it looked as if the % women in 1A and in the ABC were fairly comparable in Hard Sciences and LIfe Sciences, but the reduction in % women that you described was very pronounced in Humanities. I think this point emerges from the analysis at top of p 9, but it was a rather convoluted approach to analysis.

Overall then, I think the paper needs reshaping so you have :

1. Introduction: covers evidence for inequality and then delineates possible reasons for this
I did not find the hypotheses listed in lines 132-141 very convincing: they seemed more like redescriptions of the data. As this is essentially a descriptive study, I think it may be preferable to remove these hypotheses, but note instead that differences between subject areas may be informative about the underlying causes of gender inequality and how to tackle them.
2. Methods
3. Results (to incorporate additional analyses as above)
4. Discussion: How results inform our understanding of different causes. (see reviewer 1). You could move to Discussion the material that you currently have on p 3 paras 2-3, which focuses on possible ways of addressing gender inequality.

More minor points
Remember that you need to use full stop, rather than comma, for all decimal points. This applies to figures as well as tables.
I agree with reviewer 2 that you should put Methods before Results.
line 51: Although -> Despite
line 74: Herein -> Here
line 170: less -> fewer
line 312. Brazilian -> The Brazilian
line 316 taken literally, it sounds as if you are saying we are seeking equality between sexes in violence against women – clearly that is not so and this needs rewording
line 322-323: material is repeated and this needs rewording
p 23; I found Figure 2 very hard to understand. What do the bars represent here?
Figure 3 is probably not needed: especially as I think these data only make sense if related to career stage and discipline, both of which will affect funding.
I shared reviewer 2’s confusion about ‘adjusted residuals’. Are these the difference between expected and observed frequencies in the chi square? They are always equivalent but different in sign for male and female and it does not seem necessary to report both.