We use a globally unique dataset that scores every individual academic's holistic research performance in New Zealand to test several common explanations for the gender pay gap in universities. We find a man's odds of being ranked professor or associate professor are more than double a woman's with similar recent research score, age, field, and university. We observe a lifetime gender pay gap of~NZ$400,000, of which research score and age explain less than half. Our ability to examine the full spectrum of research performance allows us to reject the 'male variability hypothesis' theory that the preponderance of men amongst the 'superstars' explains the lifetime performance pay gap observed. Indeed women whose research career trajectories resemble men's still get paid less than men. From 2003-12, women at many ranks improved their research scores by more than men, but moved up the academic ranks more slowly. We offer some possible explanations for our findings, and show that the gender gap in universities will never disappear in most academic fields if current hiring practices persist.