Okay, let's give it a try! We'll use math, and calculate the upper and lower bounds.

Let's suppose "every" woman in the U.S. has been sexually harassed.

Let's suppose further that men, in true egalitarian fashion, distribute harassment-of-women duties evenly between them. Gay men will probably refuse (~5%), and the remainder are all needed because the gender balance isn't that far off save in a very few locations (e.g. Silicon Valley), so about 95% of men are harassers.

Now let's suppose that there are serial harassers; people manage to have sex over once a week on average, and surely there are far more opportunities to harass than to have sex, so maybe roughly daily is plausible; and there's a ~30 year window in which to do it, so that's about 10,000 harassments inflicted per serial harasser. This is too much harassment to keep track of, so it's presumably going to be distributed fairly randomly; if the escapers are a tiny minority (let's say 0.1%) then Poisson-distributed harassment yields an average of ~7 harassments per woman, which still could be done by 0.1% of men.

So, that's the math: somewhere between 0.1% and 95% of men harass women.

Want to make a less specious suggestion?