We estimate the fraction of mass that is composed of compact objects in gravitational lens galaxies. This study is based on microlensing measurements (obtained from the literature) of a sample of 29 quasar image pairs seen through 20 lens galaxies. We determine the baseline for no microlensing magnification between two images from the ratios of emission line fluxes. Relative to this baseline, the ratio between the continua of the two images gives the difference in microlensing magnification. The histogram of observed microlensing events peaks close to no magnification and is concentrated below 0.6 mag, although two events of high magnification, Δm ~ 1.5, are also present. We study the likelihood of the microlensing measurements using frequency distributions obtained from simulated microlensing magnification maps for different values of the fraction of mass in compact objects, α. The concentration of microlensing measurements close to Δm ~ 0 can be explained only by simulations corresponding to very low values of α (10% or less). A maximum likelihood test yields α = 0.05+0.09 -0.03 (90% confidence interval) for a quasar continuum source of intrinsic size r_{s_0}˜ 2.6 × 10^{15} cm. This estimate is valid in the 0.1-10 M sun range of microlens masses. We study the dependence of the estimate of α with r_{s_0}, and find that α lsim 0.1 for r_{s_0}≲ 1.3 × 10^{16} cm. High values of α are possible only for source sizes much larger than commonly expected (r_{s_0}≫ 2.6 × 10^{16} cm). Regarding the current controversy about Milky Way/LMC and M31 microlensing studies, our work supports the hypothesis of a very low content in MACHOS (Massive Compact Halo Objects). In fact, according to our study, quasar microlensing probably arises from the normal star populations of lens galaxies and there is no statistical evidence for MACHOS in the dark halos.