Women in STEM fields

Jessica Hanzlik recently posted a primer on gender and science on this blog. Her primer is a pretty good overview of the various theories in this regard, but there are plenty of additional points I’d like to make on the topic. Hence, this blog post.

First, a bit of terminology. As is customary in many such discussions, I’ll use the acronym STEM for “Science, Technology, Engineering and Mathematics” — a broad array of related fields that are roughly characterised by the use of the scientific method and a mathematically based language of discourse. When I talk of people in STEM fields, I am including jobs both in academia and industry, and jobs that include both research and more mundane application of prior research (such as a job as an engineer in a construction company). Also, I won’t draw a fine distinction between declared gender and biological sex, since, for the vast majority of people, the two are the same.

The main factual claim is that, across the world, to varying degrees, the proportion of females in STEM fields is less than their proportion in the population or the relevant demographic segment of the population. The proportion of women in a large national population at any given age is close to 50% (about 51% of naturally occurring births are male, but females have higher survival rates through childhood in the absence of significant differences in access to nutrition or health care). So, as a reasonable rule of thumb, if the proportion of females in a given STEM field is much less than half, women would be called “under-represented” in that field.

Under-representation: why is it a problem?

Women are under-represented in STEM fields across the world. Why, though, is this under-representation a problem? Before you protest that this is a silly question, note that some people don’t see it as a problem, and even the people who do see it as a problem often have many different reasons for seeing it as a problem. What are these? Here are some:

• Individual tragedy: A particular woman who is capable of entering a STEM field and enjoying/benefiting from it more than the alternative she chooses, does not do so. This “individual tragedy” may be due to discrimination, due to stereotypes, because she never even knew she could do it, etc. Note that the individual tragedy scenario could occur even without discrimination by educational institutions, though discrimination of various forms of course leads to individual tragedies. Note also that the simple fact of a woman capable of a STEM field not choosing it is not in itself an individual tragedy — may be she is choosing something that works even better.

• Collective loss: If large numbers of females who can make great contributions to, say, physical chemistry, do not enter this field, then the field suffers, and there is less of the technological progress that would arise from breakthroughs in the field.

• Skewed perspective: This is based on the idea that each group brings its own perspective on the subject to the table, and a lower representation of one group leads to an under-representation of a particular perspective and hence derails the progress of the subject. For instance, a journalist who comes from a poor rural third-world country may be able to offer more insight when writing articles or filing reports about events in poor rural third-word countries, thanks to his personal experiences. Similarly, a journalist who spent her school years high on drugs and sex may be able to offer more personal insight when reporting on drug abuse by school-going teenagers. While journalism offers an obvious example where perspective could matter, some argue that groups with different experiences have different approaches to subjects such as mathematics or physics and that each approach has some unique merits.

Women elsewhere

To put the discussion of women in STEM fields in context, here are some facts. In the United States, more than half of undergraduate degrees are awarded to females. I’ve got a collection of links plus speculation on the meaning of this. The finding is not limited to the United States — it holds in many developed countries, and even in some Middle Eastern Islamic countries. The latest Council of Graduate Schools report (with data from 2009) also indicate that females outnumber males in graduate study and doctoral degrees in the United States — see here, here, and here, for instance. While the female majority in doctoral degrees is slim (50.4%) and 2009 is the first year this has happened, the overall female majority in graduate programs (58.9%) is pretty robust.

Of the 11 fields of graduate study reported in the CGS report, women outnumber men is 7 of the 11 fields. What are the remaining four? Business (female 45.7% — less than half, but hardly a significant skew) and the STEM fields: engineering (22.3% female), math and computer science (29% female) and physical sciences (37.3% female). In other words, when we are looking at data from the United States, the only fields with a significant under-representation of females are the STEM fields.

Why worry?

With this context, we can ask: why the obsession with STEM fields? Why is there so much concern about male majorities in STEM fields, rather than female domination of the field of education and the health sciences, at both the undergraduate and graduate and professional levels? If we go back to why under-representation could pose a problem, reasons appear thin. If we are concerned about individual tragedies, is there any reason to believe that there are fewer males who are missing out on the pursuit of sociology or political science or education research than there are females who are missing out on mathematics and computer science? Under the assumption that both males and females are equally suited on average for every job, the statistics don’t bear such a conclusion.

Is it the case that there is greater collective loss to society if a talented person capable of mathematics or engineering stays away, and less of such a loss if the same happens with sociology or history? This seems plausibly true on average — which is one reason why the most talented people are generally more strongly encouraged, through both reputational and financial incentives, to enter STEM fields than the social sciences or humanities. But it is unclear if this is true at the margin, i.e., whether more people should switch into STEM fields given their current levels. We should keep in mind that most of the females who could switch into STEM fields would be drawn, not from the lowest rungs of academic talent, but from other high-end disciplines that currently enjoy female domination. An influx of female engineers would come, not from females who had been planning on becoming sales clerks, but from females who had been planning on becoming doctors or social scientists.

The third point, which is about skewed perspectives, could be of some concern in STEM fields, but considerably less than in most other fields, since, for the most part, STEM fields offer a number of checks and balances to ensure rigour and verifiability. These fields are more constrained by empirical evidence and mathematical reasoning. More importantly, though, I am not aware of any specific assertions about particular female advantages in perspective that are of importance to STEM fields. (This is not to say that there are no ability advantages — slightly better average female verbal ability may be a slight advantage in understanding some things in the subject, but that I have seen no hypothesis on a difference of perspective that would affect the nature of the new ideas that are developed).

A man’s game

An alternative explanation for why female under-representation in STEM fields seems more worrisome than male under-representation in other fields is provided by Robin Hanson in this blog post:

So why do both men and women prefer sports that emphasize male hunting type physical skills, over female gathering type skills? Looking for parallels, I notice that women are said to look good in male-style clothes (e.g., suits), far more than men are said to look good in female-style clothes (e.g., dresses). Women also earn more respect succeeding at male-dominated professions than men earn by succeeding at female-dominated professions.

The general pattern in all three cases is that we seem to respect women doing well at what mostly men do far more than we respect men doing well at what mostly women do. For better or worse, male abilities seem to more define which abilities count most for high status. Doesn’t seem fair to women, but there it is.

The causes of female under-representation

Setting aside the question of why it should matter, we may still be curious about understanding the causes for female under-representation in STEM fields. Jessica’s post suggests that there are two competing theories: the sociological theories and the biological theories. Biological theories claim that there are genetically influenced differences in the population distributions of a number of traits for males and females. These traits may include various aspects of cognitive abilities (such as verbal skills, visuo-spatial memory, visuo-spatial manipulation skills, short-term memory), personality/character traits (perseverance/conscientiousness, openness to experience, extraversion, neuroticism), desires/goals/ambitions in life, etc. Differences in the distribution of these is a factor in the different degrees to which males and females take up different kinds of jobs.

First, it is worth noting that just because a trait is genetically influenced, and because the distribution of that trait differs between males and females, does not imply that the difference in distribution between males and females can itself be attributed to genetic/biological factors. However, a nonzero heritability does open up the possibility of biologically influenced sex differences. For almost all the traits under consideration, heritabilities are significantly greater than zero. Heritability estimates for cognitive abilities range between 0.5 and 0.8, with the most reliable kinds of studies such as the Minnesota Study of Twins Reared Apart yielding estimates near 0.75 for cognitive ability and up to 0.5 for various personality traits. [Caveats: (i) These are broad-sense heritabilities and including emergenic phenomena. Narrow-sense heritabilities, which are used to predict correlations between the arithmetic mean of parents' value and the child's value, are closer to the 0.5-0.6 range for cognitive ability and probably lower for most other traits; (ii) these heritability estimates are obtained within a relatively narrow band of socio-economic status and wealth in the United States. When comparing people across highly diverse socio-economic conditions, such as Bill Gates versus a starving rural Indian, heritability estimates would be a lot lower.] This, of course, proves nothing about sex differences per se, but it means that the naive counter-argument which says that intelligence is purely environmentally determined simply doesn’t hold water.

One common argument made by those who posit a 100% social origin of the under-representation of women in STEM fields is that the degree of female representation differs radically between countries. This would be a knock-down argument against 100% biological determinism, but it certainly does not indicate that the biological equilibrium is a 50-50 split. There is no reason to believe that, in the realm of the abilities and character traits needed for STEM fields, males and females would have identical distributions. (Note: The abilities and character traits needed themselves keep changing as the nature of the fields changes, which further complicates the issue but does not affect the underlying point). For almost all physical and physiological traits (height, weight, BMI), and many acknowledged character traits, the distributions of males and females differ. Even in cases where they are nearly equal, there are hidden differences. As mentioned earlier, about 51-52% of naturally occurring births are male. In fact, about 60-65% of conceptions are male, but males have a much lower chance of making it in the womb during those crucial nine months. Again, social factors can have an impact — high status families produce more male offspring, and females under higher stress produce more female offspring. Despite this, the sex of one’s offspring is not considered socially determined.

As Jessica pointed out, males do outperform females at the upper end of the SAT score distribution in the United States, and the precise reasons are unknown. On the other hand, females outperform males in high school grades, and averaging out high school grades and SAT scores typically provides a tool with equal predictive validity for both sexes as far as college grades are concerned. Interestingly, the situation with SAT scores is reversed in some segments of the US population, such as Black Americans, as well as in some other countries, such as Iceland. As also pointed out, there are profile differences in the score distributions of male and female abilities in verbal and visuo-spatial skills on IQ tests — and many of these are fairly subtle differences that simply cannot be explained by stereotypes (because the stereotypes are too crude to predict these very specific patterns). For instance, in an EDGE debate between Steven Pinker and Elizabeth Spelke, Pinker summarises some of the differences revealed by research:

Indeed, in cases where there are differences, there are as many instances in which women do slightly better than men as ones in which men do slightly better than women. For example, men are better at throwing, but women are more dexterous. Men are better at mentally rotating shapes; women are better at visual memory. Men are better at mathematical problem-solving; women are better at mathematical calculation. And so on.

Personally, I am not convinced that better male performance at the upper end of the SAT score distribution is a significant reason for male domination in STEM fields, but it is certainly a possibility that has not been ruled out. Other possibilities (which may be part-biological and part-socially influenced), as documented by Steven Pinker in the same debate, include: males are more likely to chase status, males are generally more risk-taking (partly a consequence of testosterone — females with higher than usual levels of testosterone are greater risk takers too), males are more likely to enjoy abstract rule systems (even assuming an equal level of ability to handle them), males are somewhat better (at the tails) at three-dimensional visuo-spatial manipulation (as opposed to visual memory), and males are somewhat more variable.

If something should be done, what is it?

Assuming that you see female under-representation in STEM fields as a significant problem, what can or should be done about it?

First, a minimisation of individual tragedies, in the sense that females (or for that matter, males) who would enjoy a STEM career more than the best alternative career, get adequate exposure to, and accurate information about, the STEM career. The key point here is accurate information. More on this later.

Second, some aspects of the bureaucracy of STEM fields, both in academia and industry, may create unnecessary inconveniences for both sexes, but ones that particularly hinder the choice of such a career by women who want to have children. For instance, the “publish or perish” tenure track in academia may be one such example. This would definitely be something to think about.

Third, an emphasis on people identifying their individual strengths and desires and not being unduly influenced by or bothered about knowledge of group averages. Jessica, and many others, have referred to “stereotype threat” theories that claim that females, when reminded of stereotypes that females cannot do math, do worse on a test. While I am sceptical of such theories having sweeping power outside of their narrow experimental context, they do point to one uncontroversial area where there is scope for improvement: reducing females’ identification with and emotional concern about broad-based statement about gender averages (whether such statements are true or false). Personally, I find it ridiculous that somebody should spend their emotional energies worrying about how statements about a large heterogeneous involuntarily chosen group such as “women” (or for that matter “children” or “Africans”) might reflect on them or predict anything about them personally. [Things are a bit different for voluntarily chosen group identifications such as "Marxist" or "Muslim" or "Wikipedian"].

Fourth, it is unhelpful, and probably counter-productive, to distort science beyond recognition so as to make it appear sexy to women, often based on (possibly true and possibly false, but decidedly crude) stereotypes about what will attract women. While such works may be useful to help women with mathematical difficulties grasp school-level mathematics (for instance, this kind of book) it is not at all the kind of approach that should be used to attract people into professions that demand serious mathematics.

For instance, around the late 1980s and early 1990s, there was a craze for “reform calculus” at the freshman college level in the United States. The problem was that large numbers of American freshmen were failing their introductory college calculus courses, and even those who passed didn’t seem to have learned or appreciated much of it. Further, people who did poorly in introductory college calculus had many mathematically intensive options, particularly STEM fields, closed to them. “Women and minorities” were somewhat worse hit. Enter calculus reform. The idea was to reduce the algebraic skills needed in the calculus courses, increase calculator use, introduce more “group work”, reduce rote memorisation of formulas, and have more real-world projects and applications. The claim was that original formula-based rote memorisation and the competitive plug-n-chug environment was not suited to modern needs, and further, that it disadvantaged women who were more cooperative and preferred to work in groups.

I’m not sure whether these claims were true, but regardless of whether they are true, they miss the point — which is that an introductory course in a subject should provide an honest and accurate picture of how the subject is done, rather than pander to what people might find sexy and appealing. (For more on my take on this subject, see here. For articles on reform calculus, see this, this, and this (PDF).