We study how population variation in testosterone levels impacts male labour market earnings using data from the UK Household Longitudinal Study between 2011 and 2013. We exploit genetic variation between individuals as instrumental variables following a Mendelian Randomization approach to address the endogeneity of testosterone levels. Our findings show that higher testosterone levels have a strong positive impact on earnings. Importantly, these findings are limited to men belonging to the lower quartile of the testosterone distribution and working in higher-paid jobs. We show that differences within rather than between occupations drive these findings, whereas we find limited support for selection into occupation or mechanisms involving individual characteristics, including personality traits and education.