So, before we get to the actual facts (numbers) we will start with a few concepts outlined to form a basic understanding of lexicon etc.
First this writing is specific to Western Society, specifically the United States. I did not research the topic in Canada or England or any country OTHER then the US so that is what is being addressed here. A worldwide version would be, I feel difficult in the extreme to get numbers for to say the least.
There are different types of evidence, and I will attempt to identify which types are used but I admit this idea of evidence type identification occurred to me as an afterthought and I may well forget that part. But to put a hierarchy numbers are more important then hypothetical reasons which is something to keep in mind.
There were several studies put forward I didn't bother looking at a study from 2008 because that was a decade ago and the more current information the better because we are really concerned about moving forward which starts from right now as opposed to "back then" so I used 2 studies one from 2016 and one from 2018. As it is 2018 right now that study is the most relevant, also the 2016 study was of less value as it looked only at median values. It is impossible to be inside the heads of every person involved to know why they are picking the jobs they are and it is useless to compare a surgeon's pay to a dental hygienist so a median value based on gender as the only difference is not really a good measure for scientific method it is more meant to induce hysteria because it will generate the largest number.
To be more clear there are two numbers (in a broad sense) that are used when talking about gender pay gap Adjusted and Unadjusted. Also Controlled or Uncontrolled is another common term used.
In Controlled that is the median example, meaning it only accounts for the gender of the two values it does not account for things like job differences, experience differences and of course time off. Without accounting for multiple variables the value is largely meaningless.
Uncontrolled numbers are taking into account as many variables as they can. (Which you would think the names are backwards) So meaning the comparision would be between two data entry personnel with similar levels of experience and job requirements at companies of similar size etc where the only factor different was gender. When looking at how much impact gender has on pay differences this is the best way to do it, try to make gender be the only difference.
2018 payscale.com data indicates Uncontrolled paygap of 2% between the genders.
Now first reaction is usually 2% wow that's way different then the 20+% that usually gets put forward, and the second reaction is typically "on the other hand if they do the exact same work why should there be any difference at all?" and that reaction is totally fair but even the Uncontrolled study didn't quite manage to cover every aspect of the differences which I will go into now.
Now the primary difference involves time off, women are more likely to take time off (not like as in vacation days but as in a break from time in the workforce/job in question) then men are. When they do take time off women are usually doing something unrelated to the company where as when men take time off they are more likely to be increasing their education which naturally makes those particular men more valued to the companies in question. Even without the men doing that with their time off, the women taking time off (sometimes a year or more) from work of course will cause a bit of financial loss in terms of raises/seniority.
Some bosses (male or female) may also look at women who have taken time off to start a family as less valuable workers based on their desire to take care of their families. This of course would be unfair to the women in question but is also not really quantifiable with hard numbers/proof. It is my feeling that the result there would more likely be not hiring the lady at all and not to just pay her 2% less BUT that is not in any way proven it is just my rational thinking.
Fun Fact, 2% is a very common yearly raise!
Given these factors a 2% difference between the Genders is not really something to fret over, in fact if you consider men make up 86% of workplace fatalities it could easily be argued that on average men are under paid.
As an amusing aside, what married woman does not spend more then 2% of the money her husband is making anyway? Like I said that's just a tidbit it isn't really the concern of the companies but as the overall hysteria is meant to have you fear for women overall.....
By way of additional evidence you should consider looking into the Male Variability Hypothosis. Basically it indicates that men make up the bulk of extreme examples (the very smart/good and the very dumb/bad) while women make up the bulk of the middle (or mediocre)
This would help to explain why you find more men at the upper levels and why you find more men doing the jobs society views as....distasteful lets say?
There is potentially a great deal more to say but I think I am in danger of being overly long winded, I will put in the comments links to different studies etc. I may include one for the Male Variability Hypothosis as well but that is a side issue not the main point of the writing so I also may not. If I do not it is easily looked up under the name I have provided despite my typo's.