From Sloan Digital Sky Survey ugriz imaging, we estimate the stellar masses of the host galaxies of 70 low redshift SN Ia (0.015 < z < 0.08) from the hosts absolute luminosities and mass-to-light ratios. These nearby SN were discovered largely by searches targeting luminous galaxies, and we find that their host galaxies are substantially more massive than the hosts of SN discovered by the flux-limited Supernova Legacy Survey. Testing four separate light curve fitters, we detect ~2.5{sigma} correlations of Hubble residuals with both host galaxy size and stellar mass, such that SN Ia occurring in physically larger, more massive hosts are ~10% brighter after light curve correction. The Hubble residual is the deviation of the inferred distance modulus to the SN, calculated from its apparent luminosity and light curve properties, away from the expected value at the SN redshift. Marginalizing over linear trends in Hubble residuals with light curve parameters shows that the correlations cannot be attributed to a light curve-dependent calibration error. Combining 180 higher-redshift ESSENCE, SNLS, and HigherZ SN with 30 nearby SN whose host masses are less than 10^10.8 solar masses in a cosmology fit yields 1+w=0.22 +0.152/-0.143, while a combination where the 30 nearby SN instead have host masses greater than 10^10.8 solar masses yields 1+w=-0.03 +0.217/-0.108. Progenitor metallicity, stellar population age, and dust extinction correlate with galaxy mass and may be responsible for these systematic effects. Host galaxy measurements will yield improved distances to SN Ia.