Polar wander, the secular motion of the Earth’s rotation axis relative to its surface, has been studied for many years. Dynamical arguments(1-3) show that polar wander can arise from the redistribution of mass in a plastic deformable Earth, the rate depending on both the rate of mass redistribution and the rate at which the Earth’s rotational bulge can readjust to the changing geoid modelling(4), a mantle flow field consistent with tomographic anomalies(5), and time-dependent lithospheric plate motions(6) to calculate the advection of mantle density heterogeneities and corresponding changes in the degree-two geoid during the Cenozoic era. We show that the rotation axis will follow closely any imposed changes of the axis of maximum non-hydrostatic moment of inertia. The resulting path of the rotation axis agrees well with palaeomagnetic results(7), with the model predicting a current rate of polar motion that explains 40% of that observed geodetically(8).