We present measurements of the void probability function (VPF) at z ∼ 1 using data from the DEEP2 Redshift Survey and its evolution to z ∼ 0 using data from the Sloan Digital Sky Survey (SDSS). We measure the VPF as a function of galaxy color and luminosity in both surveys and find that it mimics trends displayed in the two-point correlation function, ξ; namely that samples of brighter, red galaxies have larger voids (i.e. are more strongly clustered) than fainter, blue galaxies. We also clearly detect evolution in the VPF with cosmic time, with voids being larger in comoving units at z ∼ 0. We find that the reduced VPF matches the predictions of a 'negative binomial' model for galaxies of all colors, luminosities, and redshifts studied. This model lacks a physical motivation, but produces a simple analytic prediction for sources of any number density and integrated two-point correlation function,ξ. This implies that differences in the VPF across different galaxy populations are consistent with being due entirely to differences in the population number density andξ. We compare the VPF at z ∼ 1 to N -body ΛCDM simulations and find good agreement between the DEEP2 data and mock galaxy catalogs. Interestingly, we find that the dark matter particle reduced VPF follows the physically motivated 'thermodynamic' model, while the dark matter halo reduced VPF more closely follows the negative binomial model. The robust result that all galaxy populations follow the negative binomial model appears to be due to primarily to the clustering of dark matter halos. The reduced VPF is insensitive to changes in the parameters of the halo occupation distribution, in the sense that halo models with the sameξ will produce the same VPF. For the wide range of galaxies studied, the VPF therefore does not appear to provide useful constraints on galaxy evolution models that cannot be gleaned from studies ofξ alone.