Measuring these things by outcomes is always wrought with errors. For example, how do you compare women's interest in medicine 50 years ago to their interest in comp sci today? If raw interest is low even a trivial scale issue might dissuade you.

It's also dangerous to treat all software development the same. I'm my experience women tend to be more interested in human facing parts of software like web and ui development, but less interested in the back end parts. Is this the result of sexism? Maybe, but regardless I bet the gender stats are very different if you consider front end as different than back end development.