Lady Business

It's awesome that you've done all this. I think you need to consider the base rate, though; it really matters what percentage of books published are by women.

For instance, I went to amazon.com, selected SF&F paperbacks, sorted by date (which should be a fairly random selection) and counted the number of books in the top 100 with at least one female author. I counted 35. Now a sample of 100 isn't that many and the 95% confidence interval is about +/-9% and I'm not really happy with how I handled the multiple authors, but for the sake of argument let's work it through assuming 35% women and 65% men. If I denote the probability of someone reading a book by a same-gendered author by 'ps' and by an opposite-gendered author by 'pd' then

for women:

(pd/ps)*(65/35) = 42/58
pd/ps = 35*42/(58*65) = 0.39

for men:

(pd/ps)*(35/65) = 18/82
pd/ps = 65*18/(82*35) = 0.41

So these numbers suggest, bearing in mind all the caveats above, that if I pick a book entirely at random from the SF&F books being published, the women in your sample are 39% as likely to review it if it's by a man as if it's by woman, and the men are 41% as likely to review it if it's by a woman as if it's by a man. The difference doesn't make men seem so awful when you express it like that!

Now here's the thing: I bet the publishers have similar figures to yours. And I bet they're reading them the same way many people are, as 'women read male authors and men don't read female authors', since publishers generally aren't statisticians. And they may well be drawing the conclusion that it's better to publish books by men because they appeal to a wider audience. And they might be *completely wrong*.