Much progress has been made in measuring black hole (BH) masses in (non-active) galactic nuclei using the tight correlation between stellar velocity dispersions (sigma) in galaxies and the mass of their central BH. The use of this correlation in quasars, however, is hampered by the difficulty in measuring sigma in host galaxies that tend to be overpowered by their bright nuclei. We discuss results from a project that focuses on z~0.3 quasars suffering from heavy extinction at shorter wavelengths. This makes it possible to obtain clean spectra of the hosts in the spectral regions of interest, while broad lines (like H-alpha) are still visible at longer wavelengths. We compare BH masses obtained from velocity dispersions to those obtained from the broad line region and thus probe the evolution of this relation and BH growth with redshift and luminosity. Our preliminary results show an offset between the position of our objects and the local relation, in the sense that red quasars have, on average, lower velocity dispersions than local galaxies. We discuss possible biases and systematic errors that may affect our results.