Comparing spectroscopic and photometric stellar mass estimates

The purpose of this Letter is to check the quality of different methods for estimating stellar masses of galaxies. We compare the results of (1) fitting stellar population synthesis models to broadband colors from the Sloan Digital Sky Survey (SDSS) and the Two Micron All Sky Survey, (2) the analysis of spectroscopic features of SDSS galaxies (Kauffmann et al.), and (3) a simple dynamical mass estimate based on SDSS velocity dispersions and effective radii. Knowing that all three methods can have significant biases, a comparison can help to establish their (relative) reliability. In this way, one can also probe the quality of the observationally cheap broadband color mass estimators for galaxies at higher redshift. Generally, masses based on broadband colors and spectroscopic features agree reasonably well, with an rms scatter of only similar to0.2 dex over almost 4 decades in mass. However, as may be expected, systematic differences do exist and have an amplitude of similar to0.15 dex, correlating with the Halpha emission strength. Interestingly, masses from broadband color fitting are in better agreement with dynamical masses than masses based on the analysis of spectroscopic features. In addition, the differences between the latter and the dynamical masses correlate with the Halpha equivalent width, while this is much less the case for the broadband masses. We conclude that broadband color mass estimators, provided they are based on a large enough wavelength coverage and use an appropriate range of ages, metallicities, and dust extinctions, can yield fairly reliable stellar masses for galaxies. This is a very encouraging result as such mass estimates are very likely the only ones available at significant redshifts for some time to come.