Stellar kinematics of elliptical galaxies have turned up only 3 unambiguous cases where dark matter is needed to fit the data : NGC 2434, NGC 7507 and NGC 7626 (Kronawitter et al. 2000), partly because the data is limited in radius to 1 - 2 Re. Planetary nebulae can now be used as a tracer further out, and is becoming an industry. Interestingly, constant M/L models can explain the new data, though M/L values are rather high (e.g. Romanovsky et al. 2003).
For field ellipticals, evidence for dark matter does come from HI studies, which show for several galaxies a flat rotation curve from ~ 0.3 Re out to 5 - 6 Re (e.g. Oosterloo et al. 2002). This implies a global (M/L)B ratio ~ 25. This value is in good apparent agreement with a similar value found based on X-ray studies. Yet the new Chandra and XMM data show a wealth of detail in the images of the X-ray gas of individual galaxies, so much so that one can question the validity of the hydrostatic equilibrium equation used to evaluate masses. A further complication is the contribution to the X-ray flux of low mass X-ray binaries, occuring presumably in globular clusters. Indeed, Paolillo et al. (2003) argue that the core of the X-ray emission is associated with the stellar distribution (gas from mass loss of evolving stars), and that the extended X-ray emission traces really the group or cluster potential rather than the potential of the elliptical galaxy itself. This agrees with studies using globular clusters as a tracer, e.g. for NGC 4472 and M87, which show that there is need for dark matter at larger radii there (see Kronawitter et al. 2000 and references therein).
A most interesting development is the use of lensing data, combined with more classical spectroscopy, to estimate the mass of ellipticals. A very good example of this is the study of the lens 0047-281 by Koopmans & Treu (2003), who find indications that 1) the total density follows a power law with slope -1.9, 2) a constant M/L model can be eliminated using velocity dispersion data, and 3) the inner dark matter slope can be as shallow as -1.1. For some lenses, the flux ratios cannot be matched with a smooth potential, and therefore substructure needs to be added to the models to produce a fit (e.g. Bradac et al. 2002). This is seen as consistent with the CDM prediction.
Acknowledgments. I thank Lia Athanassoula for frequent discussions, and Erwin de Blok and Stacy McGaugh for fruitful joint work on LSB galaxies.