Thanks to Jean-Baptiste for a link to this 2006 Universities of Pittsburgh/Virginia study about the "consistently small... and declining" numbers of women in computer science.
Since many people have recently expressed the opinion that there are few women in computer science to begin with and that they doubted loss accounted for this, I'll cite some of the motivation for the study here:
...the dearth of women in IT, and disproportionate loss of women from undergraduate CS programs (relative to men) are well-documented phenomena (Camp, 1997; Camp et al., 1999; Crews & Butterfield, 2003; Freeman & Aspray, 1999; Gurer & Camp, 2002; Myers, 1999; NSF, 2000). Statics available from the U.S. Department of Education show that, from 1993-1994 through 1997-1998, the percentage of women earning bachelor's degrees in computer science decreased to a low of 26.7% (NCES, 2001). This drop in female CS bachelor's degree recipients stands in contrast to the steady growth of women earning degrees in other sciences and in engineering.
The study focuses on the relationship between achievement and persistence based on enrollment data from the University of Pittsburgh and reports findings that increased levels of math education and home computing experience imcrease likelihood of persistence. The study reveals that many high-achieving female students switch out of computer science due to loss of confidence and/or finding that the major does not meet expectations. Because of the small sample size and the relative homogeneity of the sample we should not conclude anything about general trends, but it reports some interesting findings in line with those reported by Unlocking the Clubhouse.