If you're going to be this pedantic, 50 is greater than 48. They're not wrong.

It might have to do with rounding and the resolution of the cohort binning -- 48% being pretty close to 50%.

Someone further up made a similar comment, and got a non-pedantic response: wages don't include bonuses, but net compensation does. That's why it's not 50%.

Yeah I looked for a * or reference of some kind, but did not find one. I think they are using 30k as the threshold for the 48% number, while the median is the 31k and change number.

Sure. Median(Variable) = Probability(value<=Variable) = 50%.

Ties. It's probably because of rounding.

...which is exactly what they've described.