We jointly model 22 early-type gravitational lens galaxies with stellar dynamical measurements using standard CDM halo models. The sample is inhomogeneous in both its mass distributions and the evolution of its stellar populations unless the true uncertainties are significantly larger than the reported measurement errors. In general, the individual systems cannot constrain halo models, in the sense that the data poorly constrains the stellar mass fraction of the halo. The ensemble of systems, however, strongly constrains the average stellar mass represented by the visible galaxies to $0.026\pm0.006$ of the halo mass if we neglect adiabatic compression, rising to $0.056\pm0.011$ of the halo mass if we include adiabatic compression. Both estimates are significantly smaller than the global baryon fraction, corresponding to a star formation efficiency for early-type galaxies of $10%-30%$. In the adiabatically compressed models, we find an average local B-band stellar mass-to-light ratio of $(M/L)_0 = (7.2\pm0.5)(M_{\sun}/L_{\sun})$ that evolves by $d\log(M/L)/dz = -0.72\pm0.08$ per unit redshift. Adjusting the isotropy of the stellar orbits has little effect on the results. The adiabatically compressed models are strongly favored if we impose either local estimates of the mass-to-light ratios of early-type galaxies or the weak lensing measurements for the lens galaxies on 100 kpc scales as model constraints.

Microlensing is the only known direct method to measure the masses of stars that lack visible companions. In terms of microlensing observables, the mass is given by M=(c^2/4G)\tilde r_E \theta_E and so requires the measurement of both the angular Einstein radius, \theta_E, and the projected Einstein radius, \tilde r_E. Simultaneous measurement of these two parameters is extremely rare. Here we analyze OGLE-2003-BLG-238, a spectacularly bright (I_min=10.3), high-magnification (A_max = 170) microlensing event. Pronounced finite source effects permit a measurement of \theta_E = 650 uas. Although the timescale of the event is only t_E = 38 days, one can still obtain weak constraints on the microlens parallax: 4.4 AU < \tilde r_E < 18 AU at the 1 \sigma level. Together these two parameter measurements yield a range for the lens mass of 0.36 M_sun < M < 1.48 M_sun. As was the case for MACHO-LMC-5, the only other single star (apart from the Sun) whose mass has been determined from its gravitational effects, this estimate is rather crude. It does, however, demonstrate the viability of the technique. We also discuss future prospects for single-lens mass measurements.