Fewer CS students, and less of them women

Jay Vesgo recently published an article on declining interest in computer science among incoming freshmen in Computing Research News. The article has been picked up by several other sites. Here's the punchline:

According to an analysis of results from a survey conducted by HERI/UCLA, the percentage of incoming undergraduates indicating that they would major in CS declined by over 60 percent between the Fall of 2000 and 2004, and is now 70 percent lower than its peak in the early 1980s.

Alarmingly, interest in CS among women fell 80 percent between 1998 and 2004, and 93 percent since its peak in 1982.

The results are summarized in this graph from the full article:

What the hell happened in the early 1980s? Before then, the percentage of interested female students seemed to lag behind male interest by a steady 1/2 to 1%. But since the big peak in 1983 (right around the time I started college), interest in CS among female freshmen has dropped steadily as a fraction of interest among male freshmen. The male:female interest ratio went from roughly 2:1 in 1984, to 3:1 in 1994, to 4:1 in 1998, to 5:1 in 2001, to 6:1 in 2004. It looks like high school girls are being actively discouraged from computer science, and worse, that discouragement appeared to have steadily increased over the last 20 years.

In light of that data, the third graph in Vegso's article, measuring the percentage of various degrees awarded to women, is almost a relief:

The percentage of CS degrees granted to women has dropped steadily since the same peak in the early 1980s, hovering just below 30 percent. That's sad, but compare it to the freshman interest data above. Despite steadily decreasing interest among incoming women, the ratio of male to female computer science bachelors degrees is hovering steady at about 3:1.

I don't know what to make of this discrepancy. According to Vesgo, “Freshmen interest levels at any given point have been an accurate predictor of trends in the number of degrees granted four to five years later,” but that's not reflected in the gender discrepancy. Are universities actually doing something right?

Or has NSF's degree data, which stops in 2001, just not caught up with the latest drop in interest among women?