The percentage of women in computer science classes was growing until 40% in 1984, but has been declining ever since then. Back in 1960, about half of people who called themselves "programmers" were women.
NPR's explanation:
http://www.npr.org/blogs/money/2014/10/21/357629765/when-women-stopped-coding
In brief: It was around 1984 that first personal computers appeared, such as Apple, and they were marketed predominantly to boys, so young women didn't get to play with computers while growing up. Also, the popular culture portrayed software as an all-male field. The result was that women had a disadvantage in co.sci. classes compared with men, and started quitting the field because they felt unwelcome.
However, it's not clear why women would feel unwelcome in a field they participated so actively up until 1980. The popular culture of those times reflected this:
A quote from "The Cosmopolitan" in 1967:
“Now have come the big, dazzling computers — and a whole new kind of work for women: programming. Telling the miracle machines what to do and how to do it. Anything from predicting the weather to sending out billing notices from the local department store. And if it doesn’t sound like women’s work — well, it just is.”
Source: http://www.digitaltrends.com/business/where-did-all-the-women-in-computer-science-go/
Up until 1980 there were TV shows portraying women as software engineers:
http://www.cnn.com/2015/07/31/tech/women-computer-science-halt-catch-fire-feat/
My first observation would be that it is not 1984 but around 1980 that the curve of women's participation in co.sci. classes stopped growing at the previous rate.
Indeed, a decisive change occurred in the software industry around 1980 - the advent of personal and/or hobbyist computing. (One of the first very successful personal computers was TRS-80, introduced in 1977.) However, this does not explain why women stopped being interested in software. Why wouldn't women simply continue programming with the new computers?
Also, participation in co.sci. programs is not the same as employment as coders. Starting from about 1985, many universities adopted EE/CS programs rather than pure CS programs; additionally, many other majors such as physics, chemistry, etc., started requiring programmer's skills. So the number of pure CS (i.e. mathematics) degrees dwindled, as people primarily interested in coding moved elsewhere. Nowadays, many people employed in software industry have not actually studied computer science. The total number of CS graduates has declined after 1985.
http://cs.stanford.edu/people/eroberts/cs181/projects/women-in-cs/graphs/csbachelors.jpg
My main hypothesis about the decline in female software developers is that after 1980 the software field itself has changed.
In the early years, the job of a "programmer" was to write code and enter that code into the computer and check that there are no typos. It was a very tedious job, akin to secretarial word processing, except that it was about computers. You also had to load punchcards, tapes and so on. All these menial jobs were performed predominantly by women. This was the same in all countries where computer industry started: there was initially a high female participation. For example, in the USSR, the (often male) scientist would give equations to the (often female) programmer, and she would code them for the computer. The programmers did have to know some math, - just as they do now - but most of the job was rather tedious compared to what it is now.
Quote: My mother, who is 70, started working as a programmer in the mid 1960s. She says that at least half the programmers in her department were women and almost all the operators (loaders of tapes, punch cards, paper, etc.) were female. On the other hand, she left school at 14 with no qualifications so wouldn't have shown up on this graph [of computer science degrees]
Source: https://www.reddit.com/r/dataisbeautiful/comments/2joexk/what_happened_to_women_in_computer_science_of/
Around 1980 the tasks have changed to game programming and more generally to user interface programming. The programming languages became more powerful, and the entire process of working with computers got more efficient. The hobbyist computers such as the TRS-80 or the Commodore required a motivation that many men found very compelling ("assemble a mechanism from parts and play with it in a competitive way"). However, most women didn't find this kind of activity interesting back then, - just as they don't today. As the traditional women's "punchcard programmer" jobs declined, and as most jobs became tasks that resembled engineering (design and construction of mechanisms), so did the participation of women in the software industry decline.
Today's programming resembles design and construction of large mechanisms a lot more than word processing. The majority of women just don't find these things as interesting as men do.
I don't think the arguments about male-dominated advertising and discrimination are relevant. Computers did not suddenly become much less "human-friendly" or "inaccessible" in 1980; on the contrary, the advent of personal computing has made computers more easily accessible in people's homes. The big computer companies in the 1960s and 1970s were certainly much more male-dominated and gender-rigid than the software industry in 1990-2010. The society in general was also much less permissive and flexible with respect to gender roles back in 1960-1970 than it was in 1990 or later. Starting from 2000s we have had a number of advertising campaigns targeting specifically women for science and engineering careers. And yet, the female software engineers became less numerous relative to male engineers. (In absolute numbers, they probably became more numerous though.)
Discrimination and gender bias certainly exist in some people's behavior, but I believe that no amount of private (i.e. not government-imposed) discrimination can produce such huge differences in representation, especially since the market forces are always efficient remedies against discrimination, and since the software industry is very lucrative and competitive. The fact that women were initially 50% of the computer programmers is ample evidence that the comparatively rigid gender roles in the 1950s-1960s have not significantly influenced the software field. One of the first software companies was founded by a woman (Elsie Shutt) in 1958, and employed women who all worked at home because they wanted to take care of their children. http://ethw.org/Oral-History:Elsie_Shutt
Women have been much more represented in computer linguistics and natural language processing throughout the recent decades than in, say, computer system administration or application software development. I doubt that there were any ads specifically about user interface programming being somehow more suitable for boys while language-related computations being more suitable for girls. Advertising isn't tech-savvy enough to make these fine distinctions.
At the moment, the amount of encouragement for women to enter the engineering and programming fields is unprecedented; the gender bias and the rigidity of assumptions about gender roles are the lowest in history; the flexibility of the jobs and the wide availability of free training materials for software engineering is unprecedented. Yet, the fraction of coders who are female is variously estimated to be between 15% to 20%. I think that's about it; this is how many women actually are able and want to code today.