I was studying Raman scattering and I computed the probability of anti-Stokes scattering (with density of states n, at the numerator) over Stokes (n+1 at denominator). The densities are based on the Bose-Einstein distribution because we are talking about phonons on a crystal.
At nearly 1000 K of temperature the probability of the two events are the same, as we see in the plot (fraction value = 0.5). The curve seems to have a raising behaviour after 1000 K, meaning that anti-Stokes effects are more probable over Stokes' ones.
Anti-Stokes scattering gives us back from the material a photon with higher frequency, thus, higher energy, at cost of a phonon (or more) being absorbed from the crystal lattice. Absorbing a phonon diminishes the "amount of movement" of the crystal atoms, decreasing their thermal velocities that, at over 1000 K is still pretty high. With some SERS effect, the intensity of the phenomenon could be large.
Does that mean that, under specific selection conditions, light is cooling down the material?