The overwhelming majority of university computer science graduates are male, according to a newly published study which monitored universities in 21 countries.

However the report, which was coauthored by Maria Charles, professor of sociology at the University of California, San Diego, and Karen Bradley of Western Washington University, found significant country-to-country differences in the gender gap that imply "much more than genetics is at work".

"Restrictive government practices that minimise choice and prioritise merit may actually result in more gender-neutral distribution across fields of study, " the researchers stated.

Charles and Bradley analysed data compiled in 2004 by the Organisation for Economic Cooperation and Development (OECD) on higher education degrees awarded in 2001. Examining seven fields of study, including engineering and maths/physical sciences, Charles and Bradley calculated representation factors for each country by comparing male-to-female ratios in these subjects to those same ratios in other academic disciplines.

They found, as expected, that on the whole women predominate in such traditionally female-typed fields as education and health, and lag behind in stereotypically masculine fields. In computer science, females are under-represented in all 21 of the industrialised countries considered.

They also found that the extent of the difference in male-to-female ratios varies a great deal. Males are over-represented among computer science graduates by a minimum factor of 1.79 in Turkey, and a maximum factor of 6.42 in the Czech Republic.

In the United States, the "male over-representation factor" is 2.10 and in the United Kingdom, 3.10.

"The ubiquity of women's under-representation attests to the persistence of deep-seated and widely shared beliefs that men and women are naturally different and that they are suited for different occupations," Charles said.

"But the fact that there's so much cross-national variability suggests there's lots of room for country-specific cultural and social influences to play out."

There is little evidence, though, Charles added, for standard arguments of social evolution: the most economically developed countries do not produce the greatest numbers of women in computer science. Nor is there a strong correlation with more women in the workforce or in high-status jobs or in higher education generally.

Girls' higher maths achievement does not equate with better representation in stereotypically male fields, the researchers found.

Broad cultural support for equal opportunity is also not a good predictor. None of the study's highest-scoring nations ­ Turkey, South Korea and Ireland ­ the authors noted, is particularly known for gender-egalitarian attitudes or practices.

"There is no doubt that collective beliefs that men are naturally 'better' at maths and science are major factors that influence women's choices of college majors ­ and determine the climate in maths and science programmes worldwide," Charles said.

"When we emphasise choice and hold up self-realisation as an educational goal, girls will often freely choose poorly paid, female-typed fields of study that are in line with a conventional feminine identity and stereotypes about what girls are good at," she added.

What countries with the best female representation in computer science seem to have in common, Charles and Bradley observed, are governments that "exert strong control over curricular trajectories" and require substantial maths and science coursework.

The policy implications, Charles said, are clear: "Rather than letting people take what they expect to love (or expect to be good at), educational systems should insist on more maths and science for all students."