So it's April 2020, and I start seeing headlines like this. Okay. And they carry the same message, female scientists are getting disproportionately affected by the global pandemic. But I wasn't really convinced, because there was a mostly kind of anecdotal evidence is look like a rumors, or very small data sets.

So we decided to test that hypothesis and approach the problem with kind of more rigor, let's say. So first things first, we get a data, we get the data from bioRxiv, medRxiv, and Springer Nature journals.

And we collect approximately the data on 83,000 papers that were published between January 2019 and May 2020. Together with all this metadata.

Together with that we have the metadata about the half a million authors of those papers. We did not get the data from archive. And the reason for that is archive does not have information on affiliation, that was important for our analysis. And the second step, we genderize the names. So we use genderized API to determine the most likely gender of the author.

And we also identify the author's location, based on their affiliation name. So, we have the papers we have the authors we have the genders of the authors we also have the location so now let's, what I like to say crunch the data. Okay.

So our approach in, in this analysis is, is as this so this, we usually observe something what happened in the past, and we don't compare it directly to what happened in the future but we actually project, what we expect to see in the future here.

And we compare those things that we expect to see to the things that we actually see. So we observe the things, what happened before the pandemic, we predict what's going to happen during the pandemic. And then we compare it to the, to the ground truth to what actually happened.

And we do that with everything with a number of papers number of authors also the proportion of females in the, in the science. So first results show that actually we see more papers there is highly high surge in number of papers that appear during the pandemic.

So usually in our presentation here when you can see that green bars are expected numbers. This is what we expect with our models to see, and the orange bars are what we actually see. So, for everything we see more papers, and also we see more authors, consequently,

it goes also for meta hive and bio archive a little bit less for Springer. But that's not the main main result the main results regarding the proportion of females, female authors.

And, and this is kind of interesting because we see the drop of female authors we expect to see something like this, but we actually observe this, so it's 6.8% less female authors than expected. It seems like not much, but it's significant. Okay.

But that's that's for all papers, but let's focus a little bit on this middle figure here. Those are the papers about COVID-19. This is all papers about any topic in those preprint servers, and this is about COVID-19. Okay.

And we expect to see this and we see this so it's like it's 37.7% job than expected. And it's, it's large and it's significant. And if we desegregated by the publication see across the publications in be archive in meta archive.

And the speaker nature we didn't have enough data for that was too early. And so, our guess is that during the pandemic.

This gender disparity increase due to an increased publication rate of papers about COVID-19, altered mostly by men. So what happened. So there is COVID pandemic starting, and suddenly we observe a large surge of papers that deal with the topic.

And those papers are mostly altered by men. And that is actually the main driver of this, this effect. So we go a little bit deeper so we see actually the men who already have a publication in our data set are 70% more likely to publish a paper about COVID-19.

So we see in our previous data set we have 1.7% of men that published about COVID-19. It's compared to 1% of women. Also, of course we have more new authors which did not see in our previous database. So, are women getting excluded from the critical research about COVID-19.

I don't know. I'm just asking. We still don't know that. And we do some kind of country level analysis. So, since we have affiliations we have the location of the author we know where they are affiliated with.

And here, let's let's focus on the on the middle panel here. So, orange points mark the percentage decrease in proportion of female authors and green points mark the increase. And those horizontal lines are standard errors that you can read more about that in paper in our paper.

So we here actually observe a greater gender, like a gender gap than papers on other research topics. So this is about this. Those are the papers about COVID-19. Those are the papers about something else. And, for example, in Italy, the real, the relative drop in the proportion of female authors is 40%.

And that indicates that male scientists affiliated with Italian institutions are publishing disproportionately more than their female colleagues about COVID-19.

And similar considerations applied to Australia, Brazil, United Kingdom, France, Netherlands, Germany, and so on. And the opposite is actually true for Switzerland and Japan.

So are you some kind of similarities between those countries, and we see that actually there is a correlation between the GDP per capita, and this effect. Actually the countries with higher GDP per capita are more resilient to the effect of COVID-19 to gender imbalance.

And actually this suggests that women experience bigger life disruptions in more poor countries, and that actually affects their productivity. And additionally, actually, women get more excluded from COVID-19 research in poorer nations.

So what are the main takeaways here. So we find that women get excluded from COVID-19 related research, we measure 37% drop, and increased gender gap in publishing is persistent across the 11 countries that we observe.

And also women get more excluded from COVID-19 research in poorer nations. Of course, we have to take into consideration some possible errors in gender identification, although we were very careful there. Possible errors in identifying the author's location as well, but we checked for the most common locations.

And we also observed only first four months of the pandemic, so we still don't know what happened later. And all the data and reproducible code is available on this link here. And the data is formatted, you can just download the code, run it, and get all those fancy plots immediately.

And all the modeling is there as well. Thank you very much, and now it's time for questions.