As I recall there have been a number of studies done on this... and they fall into the "technically true" if you looked specifically at gender within a given work pool and discounted all other factors then this is the answer you arrive at.

Unfortunately, every single one of these that I have personally read ... all suffered from the fact that other factors play a part in that somewhat disingenuous number. If roles are factored in - these numbers begin to fall apart. As mentioned elsewhere in this thread: women have maternity leave... and following that can look to exit the workforce or move to part time. Compensation can be different between these categories. Continuing down this path: in a household that is dual income - it has been traditional to see the woman leave the workforce for child rearing opposed to the man. So looking at a given workforce, specifically at a given role in that group may still have a disparity in experience and time in the position (and thus compensation.) Lastly there is the bane of all - starting compensation negotiations. It is my understanding that generally men are more aggressive / assertive during this phase in the hiring process.

In short: this is stupidly difficult to generate fair and correct numbers for this type of metric and RARELY does it behoove the party running that inquiry to get the details right. The more accurate the results: the less sensational the number. Now to be clear: I do believe that there are cases where there are unfair practices taking place - but they are the exception... not the rule.

At the end of the day - if we made it commonplace to be acceptable to discuss compensation.. And put some more workers rights laws into place... We'd have a system where everyone could have a fair shake in a job, equally.

I'd be happy to be proven wrong with some numbers that have actually factored in these variables. With regard to OPs statement: that number looks strikingly familiar to one horrifyingly old and incorrectly run survey.