Comparing Dynamical and Photometric Mass Estimates of Low- and High-Redshift Galaxies: Random and Systematic Uncertainties^*

We determine the importance of redshift-dependent systematic effects in the determination of stellar masses from broadband spectral energy distributions (SEDs), using high-quality kinematic and photometric data of early-type galaxies at z ~ 1 and z ~ 0. We find that photometric masses of z ~ 1 galaxies can be systematically different, by up to a factor of 2, from photometric masses of z ~ 0 galaxies with the same dynamical mass. The magnitude of this bias depends on the choice of stellar population synthesis model and the rest-frame wavelength range used in the fits. The best result, i.e., without significant bias, is obtained when rest-frame optical SEDs are fitted with models from Bruzual & Charlot. When the SEDs are extended to the rest-frame near-IR, a bias is introduced: photometric masses of the z ~ 1 galaxies increase by a factor of 2 relative to the photometric masses of the z ~ 0 galaxies. When we use the Maraston models, the photometric masses of the z ~ 1 galaxies are low relative to the photometric masses of the z ~ 0 galaxies by a factor of ~1.8. This offset occurs both for fits based on rest-frame optical SEDs and fits based on rest-frame optical+near-IR SEDs. The results indicate that model uncertainties produce uncertainties as high as a factor of 2.5 in mass estimates from rest-frame near-IR photometry, independent of uncertainties due to unknown star formation histories.