Why is $v_4/(v_2)^2$ larger than predicted by hydrodynamics?

The second and fourth Fourier harmonics of the azimuthal distribution of particles, $v_2$ and $v_4$, have been mesured in Au-Au collisions at the Relativistic Heavy Ion Collider (RHIC). The ratio $v_4/(v_2)^2$ is significantly larger than predicted by hydrodynamics. Effects of partial thermalization are estimated on the basis of a transport calculation, and are shown to increase the ratio by a small amount. We argue that the large value of $v_4/(v_2)^2$ seen experimentally is mostly due to elliptic flow fluctuations. However, the standard model of eccentricity fluctuations is unable to explain the large magnitude of $v_4/(v_2)^2$ in central collisions.