Submitted by n_n in Feminism (edited )

In 2011, officials in the Swedish town of Karlskoga decided to clear snow from pavements and public transport routes before clearing roads, rather than the opposite. This meant that women, whose travel patterns tended to be more complicated than men, as they picked up children and did the shopping as well as going to work, suffered fewer accidents in the winter.

Before 2011, pedestrians were injured three times more often than motorists in the winter, and more than two-thirds of these were women. The estimated cost of all these falls in a single winter season was SKr36m, about $4m. The switch in the priorities of the snow-clearing schedule did not disadvantage commuters — it is easier to drive a car through three inches of snow than a buggy — and it ended up saving the local authorities money.

This is one example cited by Caroline Criado Perez, the British journalist and activist who led the campaign for Jane Austen to appear on the new £10 note, in Invisible Women: Exposing Data Bias in a World Designed for Men to demonstrate the benefits of including data about women when designing policy. Her book is mainly full, however, of the consequences of excluding them in areas as disparate as healthcare, parliaments, town planning, offices, factories, academia, agriculture, peace talks and humanitarian disasters.

While some of these “gender data gaps” are well-known, others are strikingly unexpected. Women do three-quarters of unpaid work, irrespective of the proportion of household income they bring in; carers and cleaners can lift more in a shift than a construction worker or a miner; indoor air pollution from domestic open-fire stoves is the single biggest environmental risk factor for female mortality globally. The book covers a huge range of examples of how data are biased against women — from industrial design to healthcare systems to disaster responses.

As Criado Perez says, most, if not all, of these examples did not come about because men deliberately excluded women from the data, but because they just didn’t think about them.

Many resonated with me personally. I, like the author, have been asking my employer and other events organisers for years to provide microphones that don’t require a suit jacket and trousers to wear comfortably. But the consequences of the gender data gap are far more serious than such minor irritations. Treating men as the “default human” in the scenarios described by Criado Perez means that women are not just being treated unfairly — but are actually dying unnecessarily.

The lack of sex-disaggregated data in clinical trials, for example, affects the ability to give women sound medical advice. The electrical wave threshold below which a pacemaker is fitted in the US is the correct one for men but should be lower for women. Some drugs that are commonly prescribed to treat high blood pressure have been found to lower men’s mortality from heart attacks — but to increase it in women.

Healthcare is far from being the only area where leaving women out of the data can affect their lives. Drivers’ seats in cars are designed for men and they protect women less well in a crash. Providing food but not fuel or water during the Ebola epidemic in Sierra Leone in 2014 meant that women were forced to leave quarantine areas and spread the disease.

Most worrying, perhaps, is Criado Perez’s argument that the gender data gap is getting worse. The introduction of Big Data, she contends, “can magnify and accelerate already existing discriminations”.

Algorithmic scanning of CVs is a particularly problematic area. She cites the example of Gild, a tech hiring platform, which uses algorithms to analyse candidates’ online presence to identify the best computer programmers. According to Gild, frequenting a particular Japanese manga site is a “solid predictor of strong coding” — despite the fact that women have less leisure time to spend online than men and often don’t like manga sites, which are dominated by men.

It is not just CV-scanning software that is trained on data that under-represent or misrepresent women. So do algorithms used in voice-recognition technology and translation. Satnavs in cars recognise male voices better than female ones; Google’s speech recognition software is 70 per cent more likely to recognise male than female speech, while Google Translate will assign stereotypical genders to, for example, Turkish gender-neutral pronouns, defaulting to “he is a doctor” and “she is a nurse”. Male-biased databases, Criado Perez argues, are not just reflecting but amplifying biases.

Even when policymakers are aware that women are being excluded from their data sets, the responses are often inadequate — or plain wrong. Recommended FT Podcast Why the exclusion of women from data matters Friday, 1 March, 2019

In the 1990s and 2000s some US universities adopted a policy that gave parents an extra year per child to earn tenure, in order to help women find it more feasible to achieve. Analysis of economics department hiring, however, showed that the policy resulted in a 22 per cent decline in their chances of gaining tenure at their first job — because women used the year to look after their children, while men used it to dedicate more time to their research.

Gender pay gap reporting, a requirement on UK employers designed to narrow the difference between what women and men are paid, has been introduced at the same time as a retrograde benefit system that pays a “universal credit” to the main earner in each household, who is often a man and who, years of research have demonstrated, is less likely to spend money on his children.

Data determine how resources are allocated. Bad data lead to bad resource allocation. Criado Perez hammers home this message with example after example, and a lack of evidence is certainly not a criticism that could be levelled at the book. There are 69 pages of references, and one worry is whether the author can have done adequate due diligence on the quality of all the research she cites.

This, and the repeated assumption that correlation (a fall in female legislators, for example) equals causation (a comparable drop in education spending) are likely to give ammunition to any readers determined to reject her thesis.

It is also hard at times not to feel overwhelmed by the drumbeat of examples of how women have been routinely left out of the data on which the most important decisions — in disasters, in hospitals, in factories, on our roads — are made.

But instead of being overwhelmed, I felt angry. Criado Perez comprehensively makes the case that seemingly objective data can actually be highly male-biased, and that public spending, health, education, the workplace and society in general are worse off as a result. Policymakers everywhere should take heed.

6

Comments

You must log in or register to comment.