We present a sample of 77 optical afterglows (OAs) of Swift detected gamma-ray bursts (GRBs) for which spectroscopic follow-up observations have been secured. Our first objective is to measure the redshifts of the bursts. For the majority (90 ) of the afterglows, the redshifts have been determined from the spectra. We provide line lists and equivalent widths (EWs) for all detected lines redward of Lya covered by the spectra. In addition to the GRB absorption systems, these lists include line strengths for a total of 33 intervening absorption systems. We discuss to what extent the current sample of Swift bursts with OA spectroscopy is a biased subsample of all Swift detected GRBs. For that purpose we define an X-ray-selected statistical sample of Swift bursts with optimal conditions for ground-based follow-up from the period 2005 March to 2008 September; 146 bursts fulfill our sample criteria. We derive the redshift distribution for the statistical (X-ray selected) sample and conclude that less than 18 of Swift bursts can be at z > 7. We compare the high-energy properties (e.g., y-ray (15-350 keV) fluence and duration, X-ray flux, and excess absorption) for three subsamples of bursts in the statistical sample: (1) bursts with redshifts measured from OA spectroscopy; (2) bursts with detected optical and/or near-IR afterglow, but no afterglow-based redshift; and (3) bursts with no detection of the OA. The bursts in group (1) have slightly higher y-ray fluences and higher X-ray fluxes and significantly less excess X-ray absorption than bursts in the other two groups. In addition, the fractions of dark bursts, defined as bursts with an optical to X-ray slope fox <0.5, is 14 in group (1), 38 in group (2), and >39 in group (3). For the full sample, the dark burst fraction is constrained to be in the range 25 -42 . From this we conclude that the sample of GRBs with OA spectroscopy is not representative for all Swift bursts, most likely due to a bias against the most dusty sight lines. This should be taken into account when determining, e.g., the redshift or metallicity distribution of GRBs and when using GRBs as a probe of star formation. Finally, we characterize GRB absorption systems as a class and compare them to QSO absorption systems, in particular the damped Lya absorbers (DLAs). On average GRB absorbers are characterized by significantly stronger EWs for Hi as well as for both low and high ionization metal lines than what is seen in intervening QSO absorbers. However, the distribution of line strengths is very broad and several GRB absorbers have lines with EWs well within the range spanned by QSO-DLAs. Based on the 33 z > 2 bursts in the sample, we place a 95 confidence upper limit of 7.5 on the mean escape fraction of ionizing photons from star-forming galaxies.