Missing Stellar Mass in SED Fitting: Spatially Unresolved Photometry can Underestimate Galaxy Masses

We fit model spectral energy distributions to each pixel in 67 nearby (<z>=0.0057) galaxies using broadband photometry from the Sloan Digital Sky Survey and GALEX. For each galaxy, we compare the stellar mass derived by summing the mass of each pixel to that found from fitting the entire galaxy treated as an unresolved point source. We find that, while the pixel-by-pixel and unresolved masses of galaxies with low specific star formation rates (such as ellipticals and lenticulars) are in rough agreement, the unresolved mass estimate for star-forming galaxies is systematically lower then the measurement from spatially-resolved photometry. The discrepancy is strongly correlated with sSFR, with the highest sSFRs in our sample having masses underestimated by 25% (0.12 dex) when treated as point sources. We found a simple relation to statistically correct mass estimates derived from unresolved broad-band SED fitting to the resolved mass estimates: m_{resolved} = m_{unresolved}/(-0.057log(sSFR) + 0.34) where sSFR is in units of yr^{-1}. We study the effect of varying spatial resolution by degrading the image resolution of the largest images and find a sharp decrease in the pixel-by-pixel mass estimate at a physical scale of approximately 3 kpc, which is comparable to spiral arm widths. The effects we observe are consistent with the "outshining" idea which posits that the youngest stellar populations mask more massive, older -- and thus fainter -- stellar populations. Although the presence of strong dust lanes can also lead to a drastic difference between resolved and unresolved mass estimates (up to 45% or 0.3 dex) for any individual galaxy, we found that resolving dust does not affect mass estimates on average. The strong correlation between mass discrepancy and sSFR is thus most likely due to the outshining systematic bias.