Review: Invisible Women: Data Bias in a World Designed for Men

I read a non-fiction book a few months ago that left a sour taste in my mouth; it wasn't because I disagreed with the author, but rather that they made lots of claims that either had no citation, were based entirely on a single anecdote, were practically impossible to empirically prove, or were very provably wrong with a cursory search on the internet. I couldn't help but take notes and look things up, which led to exasperation and quitting about two thirds of the way through. I complained about it to my roommates, and one of them recommended I read Invisible Women: Data Bias in a World Designed for Men for the polar opposite experience.

I finally got around to it, and found it very good. It's definitely heavy on the citations, and makes a compelling case for its central arguments, namely:

Men are usually the "default model" when designing something, which leads to it not working for women
A lack of sex disaggregated data (or lack of data in general) leads to a lack of information about how things impact women
A lack of knowledge about how things impact women lead to ineffective treatments, safety measures, workplace conditions, and more
Men are often responsible for data collection and decision making, and so typically do not consider the needs or wants of half the population
Collecting data about women is often perceived to be too difficult (despite making up 50% of the population)
Even when data is collected and properly classified, it's sometimes just ignored

There's about ten billion examples, ranging from bathroom design, to car safety, to drugs and medicine, to workplace dynamics, to politics, and more. I ended up falling into the same cycle as the other book, where I took copious notes, but in this case, it was because there were just so many interesting things that I would be remiss to read and then forget forever. I've added a bunch of them at the bottom, but it's by no means everything I took down, which is also a tiny subset of what the book covers; if you don't feel like reading the entire book, it might make a decent TL;DR.

There were a couple moments where alarm bells went off in my head, and I went exploring to validate them. For example, she says:

Tech’s love affair with the myth of meritocracy is ironic for an industry so in thrall to the potential of Big Data, because this is a rare case where the data actually exists. But if in Silicon Valley meritocracy is a religion, its God is a white male Harvard dropout. And so are most of his disciples: women make up only a quarter of the tech industry’s employees and 11% of its executives. 11 This is despite women earning more than half of all undergraduate degrees in the US, half of all undergraduate degrees in chemistry, and almost half in maths.

I have two misgivings with this paragraph. The first is that it is citing a New Yorker article for the "quarter" figure, but that article doesn't give any sources itself besides saying that "studies" say it. I haven't been able to track down their original source, despite an afternoon of trying; the closest I got were AnitaB reports from 2017 and 2020, which support the gist of the citation. I understand that sifting through every Guardian, New Yorker, Atlantic, etc. article for primary sources would have taken until the heat-death of the universe to complete, but it would also give peace of mind that the author is not being led astray by misquotes or deliberate misinformation, and allow readers who want to explore a topic to do so more easily.

The second is that it feels a little disengenuous/oversimplifying. There is no denying that women making up a quarter of tech employees demonstrates something (or many somethings) very wrong along the childhood-to-career-in-tech pipeline, and the author points some of these out; for example, she tackles ignoring female students in academic settings, bias in hiring, the effect of unpaid work on paid work commitment, and the mistreatement of women in tech pushing them to leave the field entirely, all of which supports the implied statement that tech is misogynistic. I don't doubt the conclusion (which would require some olympic-level mental gymnastics to achieve). I do, however, take issue with the approach: the inclusion of overall degrees held in general and in chemical engineering implies that the proportion of women in tech should reflect the proportion of women in degrees in general, or in chemical engineering. The reality is that many fields do not have an even distribution of all degree-holders; mechanical engineers, for example, are likely woefully under-represented in nursing, while tech is dominated by software programming/engineering, electrical engineering, and other adjacent engineering fields. Chemical engineering just isn't that statistically relevant; even if they were present in tech in proportion to the number of total jobs in the USA regardless of field, there are still 10 times more software jobs, so even if every single chemical engineering job went to a woman, they'd still be under-represented compared to the glut of men in software. It's made crystal clear that there's everything wrong with every part of the pipeline, from education, to hiring, to the workplace, but the author's choice of statistics to support her critique of tech feels like picking the largest, most bombastic number, and not the most relevant one. This leaves me wondering where else data has been purposfully mis-used to prove a point, especially outside topics I have a passing familiarity with.

In other instances where I thought something was iffy, it turned out I was wrong. One such example is about married couple income taxes:

In a married couple’s joint tax return, the couple must ‘stack’ their wages. The higher earner (given the gender pay gap this is usually the man) is designated the ‘primary earner’, and their income occupies the lower tax bracket. The lower earner (usually the woman) becomes the ‘secondary earner’, and their income occupies the higher tax bracket. To return to our couple earning $60,000 and $20,000, the person earning $20,000 will be taxed on that income as if it is the final $20,000 of an $80,000 salary, rather than all she earns. That is, she will pay a much higher rate of tax on that income than if she filed independently of her higher-earning husband.

I was a little flabbergasted reading that paragraph; surely stacking low income on high can't be real, and surely even if it were, the larger tax brackets for married couples would make sure the lower earner doesn't pay disproportionately more taxes. I did the math, but turns out, she's right on both counts; in her $60,000 and $20,000 income example, the lower earner ends up going from $2,500 to $4,000 in income taxes, and income stacking for married couples is actually a thing!

I'm a bad person for these types of books: I like analyzing data, I like checking citations, and most importantly, I like to be as annoying as possible when it comes to questioning whatever is being presented, regardless of whether I agree with it or not; I wouldn't call myself a contrarian (not if someone called me one, at least). I came out of this book with exactly three instances of "I want to look into this", and only one of them failed a deeper look (if you can even call my anectdotal experience with tech and quick Google search "deep").

This can mean one of two things: the first is that I'm simply not seeing other instances where questionable information or deceptive data is being provided, either because I lack the expertise in the given field, or because I agree with what the author is saying, and am less inclined to go source-hunting if my brain is already happily nodding along. The second is that there are very few rough edges, and minus an oversight or two, it's airtight. I really can't praise the veracity of a book any more than that, my lack of knowledge in most of the topics presented not-withstanding (something something Dunning-Kruger). Likewise, the absolute breadth and depth of topics covered is gargantuan, and in itself an achievement. The fact that the author was able to present it in a manner which is actually fun is a cake-sized cherry on top of an already delicious dessert.

Overall, I can highly recommend Invisible Women if you're looking for a thoroughly researched analysis and depressing smack-down of basically every facet of society with respect to women.

A side-note on citations and the internet, which is by no means an indictment of the author herself, but rather the peculiar situation in which the internet finds itself: the citations provided in this book are a combination of scientific literature, international body reports, government websites, news articles, and occasional smatterings of other content. I perhaps associate too much longevity to the former, but I've experienced firsthand the ephemeral nature of the others, and here is no exception; we are in 2024, five years after the publication of this book, and already some of the links provided have begun to rot. Take this citation from Chapter 16, reference 4. It is a United Nations page that no longer exists, or has been relocated. I wanted to read more on the topic of marriage customs in Rwanda and tried to visit another source the author cited. That page has been replaced with a more general one, although I managed to find what I think the original may have been, an article in the Journal of the International African Institute titled "Child Marriages in Rwandan Refugee Camps", which was a very interesting read. The Internet Archive's Wayback Machine is an excellent tool to find old versions of websites that no longer exist; that citation from Chapter 16, for example, can be found here (assuming I selected the right snapshot). I've personally been collecting PDF and HTML copied of web-pages I find interesting as my own mini version of the Wayback Machine, and can highly recommend doing the same.

The Big List

A collection of snippets from the book.

Representation:

Gendered languages that default to male when trying to be gender neutral are still overwhelmingly interpreted as being male, not gender neutral
G-rated films between 1990 and 2005, only 28% of speaking roles went to women; they made up only 17% of crowd scenes
2007 study found only 13% of non-human characters were female in children's shows
3.3% of protagonists at E3 2016 were female
White and male is assumed to be the default state of being, and everything else is an aberration, and that the former is not an identit like women or non-white is
There's an idea that more female representation in media is over-representation or pandering, but the reality is that women are still way under-represented; people assume that it's an even split, so more women = imbalance

Society:

British austerity measures in 2010s cut funding for buses (women-dominated), but not roads (men-dominated)
Equal floor space bathrooms make no sense, since urinals take more space, more elderly and disabled people are women, women have periods, are more likely to bring in children, all requiring more time
Upwards of 90% of women in France who took public transit were sexually assaulted, but many fear their reports not being taken seriously. Treating them as hate crimes in Nottingham dramatically increased the rate of reports
Statistically uninformed solutions are useless, like CTV cameras and panic buttons on buses, despite being 3x more likely to be assaulted at the bus stop than on the bus
Dixing a lot of issues by actually listening to women and properly collecting and using statistics would be cheaper, since doing so would reduce consequential costs of sickness, injury, etc.
90s Vienna found parks were male-dominated even by age 10; by creating more entrances (boys congregated around single-entrances), more informal spaces (boys dominated the formal spaces like basketball courts), and more small spaces, attendence of girls increased
Homelessness statistics typically measured at homeless shelters, but women underutilize these due to abuse/assault, so under-represented in homelessness statistics

Unpaid Work:

When unencumbered (no care responsibilities), men and women have similar health outcomes at 48 hours of work, but when in a care position, women significantly more likely to have life-threatening diseases
US men have an extra hour of leisure time compared to women
Women do vast majority of unpaid work (childcare, cleaning, cooking, etc.)
If a woman works, the man in relationship doesn't do more unpaid work, so woman's overall work hours go up
Australian study found that housework time for single men and women is about even, but when cohabitating, men's housework time goes down women's goes up
More unpaid work, so women work less paid hours to/more likely to take part-time work
Pay gap grows to 33% over 12 years since the birth of a child
Women earn between 31% and 75% less than men over life, but pensions and social nets don't account for this at all -> older women tend to be more poor
Married women with children 35% less likely to get tenure track than married men with children
90% of single parents are women in the UK, and over 80% in US
Cutting childcare support and social care budgets results in less women working, as they now must take up the extra responsibilities, resulting in less economic contribution, actually making things worse. That, and it's just shifting paid work to unpaid work for women

Workplace:

No pregnant parking by Google building until Sandberg herself became pregnant
Metabolic rate of women tends to be lower than men significantly, and typical office is 5 degrees too cold for women
Work expenses often pay for things like drinks, but not the babysitter to be able to go have the drinks
Businesses rely on unpaid work women do, but there's no compensation for this hidden operating cost
Men who believe they're more objective in hiring tend to prefer male applicants over identically described female ones
Managers in "meritocratic" organizations favour male employees over equally qualified female employees
The demographic most likely to believe in meritocracy are young white upper class americans

Academia:

Female students significantly less likely than similar male students to receive funding, get professor meetings, be offered mentorship, or get a job
Men are less likely to cite women in research
If gender neutral naming is used, men are ten times more likely to assume the author is male than vice versa
In economic papers, men routinely refer to male contributors as the lead author, despite female authors being the lead
Students more likely to turn to female professors for emotional problem resolution, to ask for extensions, rule-bending -> all extra time-consumers male professors don't have to deal with as often
Women asked to do more (undervalued) admin work than male colleagues
If half a classroom of students are led to believe their professor is male, and the other half that the professor is female, the "male" grades faster

Medicine:

Vast majority of pain studies conducted on exclusively on male mice
Studies on cancers found significantly more in women due to exposure to chemicals in female-dominated jobs (cleaning, salons, etc.) are virtually non-existant
Some studies try to minimize the effect of oestradiol or progesterone on study outcome, ignoring that these indeed exist for women, and the effect of drugs can vary significantly throughout course of a month
Women's anatomy is different to men on essentially every level, but they're severely underrepresented (if not non-existent) in basically all studies, and so medication and treatements tend to either not work or be detrimental
Women show symptoms differently, but these differences are often disregarded during diagnosis
Women less likely to be treated identically to men, or even taken seriously