Getting all knitted up with childcare models : tagging and gender-bias

I've remarked in passing that the tagging of certain things in OpenStreetMap can be highly variable.

This post started because I just wanted to show a few examples from shops which were putatively related to gender of mappers. However, I've also got massively sidetracked into looking at the (notorious) childcare issue, largely because this is becoming the exemplar of gender biases in tagging.

My shop  examples (way down below the fold) we're chosen because they were ones which I would have expected to show gender biases in tagging:

• Likelihood of the object being mapped
• Precision of the tagging of  the object is mapped

I've done this because we have known for a long time that OSM contributors are predominantly male, and, furthermore, have strong technical backgrounds (see Yu-Wei Lin's first paper on the subject, based on interviews in 2010).

The childcare tag controversy

For the last year or two a meme about OSM preferring to map strip-clubs to childcare facilities has been used to highlight this discrepancy.

A Day Nursery in Rise Park, Nottingham
Source: the author via OpenStreetView

Whilst the latter is excellent for polemics, (and, indeed, initially fitted my own preconceptions both about mappers and wiki contributors), it's not really supported by actual mapping: approximately 125,000 facilities related to childcare (mainly pre-school establishments), to under 2000 brothels and strip clubs worldwide.


Using one wikipage does not substantiate that a gender bias in numbers of contributors automatically implies strong gender bias in what is mapped. Oddly, this very moving wiki page is not cited in the discussion, which is a shame: it demonstrates exactly how OSM can be used to enhance the life of kids and their parents.

Anyway the idea that men might not be interested in childcare either shows a strong prejudice or that people are living in the past or in places I'm unacquainted with.

Of my close circle of friends, in about a third of those with children the principle carer has been the male partner. I can think of at least one British OSM contributor in this category, too. Furthermore it is difficult to avoid the ramifications of the childcare needs of ones friends and relatives, even if childless oneself. I can give chapter and verse on the facilities used by nephews, nieces, god-children and other friends children over 20 years ago. I have numerous relatives, predominantly female, who work in pre-school and early years school education; and I'd be surprised if other male OpenStreetMap contributors didn't have a similar exposure to this aspect of the education/childcare sector as well. Childcare is talked about so often it almost rivals discussions of house prices in middle-class Britain. I benefited enormously when running a business from being able to employ someone part-time who was predominantly a full-time carer for her young children. Subsequently I've always been amazed at the reluctance of firms to make the simple changes needed to employ people in that situation.

Lastly, I would hope parents would use rather more sophisticated approaches to finding childcare facilities rather than using an open data map! There are plenty of other ways to discover this sort of information, even when standard arrangements break down. Expectant mums have difficulty in avoiding a deluge of information which might be relevant as Lucy Mangan explains.

Analysing Childcare OpenData for Nottingham

Polemic is all very well, but I tend to prefer something with a solid evidence base. When I first became aware of this controversy I noted that Nottingham council had released a set of childcare Open Data. I thought "Perfect, I'll map these, and show it can be done". In fact childcare facilities for children of school age are actually difficult to map, as are, what used to be, informal arrangements (childminders).  Using this data enables one to look at the range, type, location and features on a reasonably comprehensive basis.

Day Nurseries mapped in the Greater Nottingham area on OpenStreetMap
Source: OSM-Nottingham, copyright OSM contributors

 The data set which includes 270 registered facilities includes an amazing range of things:

• Breakfast Clubs (55, 20.4%) at schools. Nearly all at schools in less prosperous areas.
• After-school Clubs (44, 13.7%) at schools or nurseries. In prosperous areas these are mainly about education, in poorer areas they may be more related to childcare. Certainly those attended by children in my family have little or no childcare aspect: they are about providing extra tuition which may be unaffordable by poorer parents, for instance in playing a(n expensive) musical instrument or learning a foreign language.
• Holiday Clubs (37, 13.7%), again often at schools or day nurseries.
• Day Nurseries (55, 20.4%). These are all mapped using the amenity=kindergarten tag
• Playgroups (13, 4.8%). Usually held in community facilities (church halls, community centres) and with limited hours. These used to be much commoner, but have been supplanted by professional day nurseries and most primary schools incorporating nursery schools. Playgroups are usually volunteer run, perhaps with a qualified paid-for supervisor (I know a bit about this as my mother set one up and I used to help during school holidays when I was a teenager). The remaining ones in Nottingham appear to be mainly in prosperous parts of town.
• Nursery Schools (66, 24.4%). These are attached to most primary schools, usually in a separate space, but not accessible, unless one is a parent, and it is certainly not advisable for single male mappers to try and acquire this sort of information. Most are mapped as amenity=school without any distinction as to the age group served.

Most of these facilities are at locations which have already been mapped, with playgroups being the most neglected area.

An older school building in Bulwell which is part of the
St Mary's (CofE) Primary and Nursery School.
  © Copyright David Hallam-Jones and licensed for reuse under this Creative Commons Licence.

In many cases multiple facilities may exist at the same location, which often creates tagging problems for things which fall into the same category. Furthermore many have moderately complex rules about opening hours (in and out of school terms) and these data are not available as part of the open data set. (When I first looked at the data I wasn't aware of all the schemes which are now possible for opening hours). In short there was little that I could map sensibly in a way which would be useful over-and-above what was already in OSM.

Even playgroups, a category which is definitely under-mapped (or not mapped at all) have tagging conflict problems because the location will be used for many activities: from Girl Guides, Scouts, Meals-on-Wheels lunches, photography clubs, judo clubs over the course of a typical week.

There are other mainly pre-school type amenities related to childcare: I've already mention childminders (and will do so again).

We have mapped a number of soft play activity centres which are in permanent facilities around Nottingham, but I am aware that there are probably more ad hoc ones, using the same sort of places as playgroups (my sister-in-law used to run one), and happening one or twice a week. My nieces used to go to one called TumbleTots which appears to be a franchised operation. There are also Toddler Groups. I've noted one location for one of these on Westdale Lane in Gedling. Both these types of activities usually require parental (or equivalent) supervision.

Childminders

Child minding is the exception, it tends to have a strong locational component (is it easy to park, can I use the bus, does it fit in with my commute), and probably represents far more facilities than the 270 listed in the Nottingham Open Data file.

In fact, because child minding has recently become regulated in Britain I can find out from Ofsted figures how many child minders are registered (I'm sure there are still plenty of informal arrangements). To my surprise there are only 388 providers registered in Nottingham, giving a total of 5,800 available early years places. The number of places and providers suggests a strong overlap between the Ofsted registrations and some of the facilities in the Nottingham Open Data. Some councils, such as Southwark in London, provide a comprehensive listing, so that I can see that they have a total of 277 (64 in Bermondsey, 60 in Walworth, 68 in Camberwell and 85 in Peckham - figures based on 2 / page) most offering 1-3 places. Ofsted figures show 449 providers with 1769 early-years places. (The difference between an inner London Borough and Nottingham is rather intriguing).

However, although Ofsted allows registered childminders to make their address public, the majority of Southwark childminders haven't done so. It then gets tricky as these are private addresses and both the privacy of the childminder and their charges comes into play: in some European countries I would anticipate that providing this sort of information may be a prima facie breach of people's legal right of privacy.

On a more practical note: how do I map it anyway? If we apply an on-the-ground rule there is rarely any information to substantiate and verify that you can find a childminder at location X. So once again we run into the impasse which is that this type of thing is just difficult to map, even though it would straightforward to tag as amenity=childminder or childcare=childminder, the former with 1 occurrence in OSM.

Conclusions

Broadly speaking quite a bit of the information provided in the Nottingham data set is close to the boundary of what it is currently sensible to map within OSM. This is nothing to do with that it's related to childcare. The truth is that the geo-component of the data which is not already mapped is small in comparison with entirety of the data. These are classic datasets where linking to OpenStreetMap to provide the location is more sensible than trying to wrangle the OSM tags to support information which is not their forte (it is even debatable whether opening hours really belong in OSM). If we map playgroups then we should try and map some of the other activities held in the same place (I've been asked where to find the local Scout group before now), and that raises a host of issues about maintainability, verifiability and completeness.

Even when the locational component is strong there are significant barriers in the processes of acquiring data, verification, respect for privacy, and legal restrictions. These barriers are considerably higher for male mappers, and higher still for male mappers without children. I believe they are strong enough even to daunt a motivated female parent mapper from collecting more than a few sites.

For the most part existing tags cover the major categories described here: amenity=kindergarten (for both day nurseries and nursery schools, and we can discribinate by using kindergarten=*), amenity=playgroup, amenity=childminder, leisure=softplay, leisure=playground, social_facility=group_home (with social_facility:for=child, etc.). School clubs (breakfast, after-school, holiday) are probably not well catered for, although there are two example of amenity=after_school_club. As stated above I have strong reservations about the suitability of trying to add data about these directly to OSM.

Childcare has changed quite a lot in Britain in the past 20 years. Day Nurseries and Nursery School provision has burgeoned; informal sectors such as child minding have become formalised and regulated and others have declined. Much of the ability to map things relies quite heavily on these formal aspects regulated by government. In many parts of the world I imagine a lot of childcare activities are informal and therefore will be harder to categorise, and probably more ephemeral.

Lastly it turns out that we have quite a lot of existing tags to describe the range of facilities covered by the Nottingham Open Data (plus some others which I and other mappers have noted during surveys), but perhaps no-one has tried to pull it together in one place before. So despite the strong caveats I raise in the preceding paragraphs it does look as if we have had the basics for doing this kind of mapping for quite a while. Certainly I hope that by going through the typology and the related issues it can provide a basis for anyone who wishes to take it further.

Looking for other examples of gender-biased tagging

My original starting point was to look at types of shops which tend to have strongly gender-biased customer bases. My examples are chosen from retail outlets which I presume have a predominant customer base of young people (roughly from 13-25 perhaps extending up to 30) and/or ones which I expect to have customers who are either predominantly male or female. My examples are:

• Games Workshop:
• Accessorize & Claire's:
• Model Shops: shops selling scale models of ships, aeroplanes, cars, buses and trains. Models may range from small inexpensive plastic ones to metal die-cast models costing hundreds of pounds/euros/dollars.
• Wool Shops: shops selling wool, knitting patterns, knitting needles and other knitting accessories. May also sell paraphernalia related to sewing.

The first ones in this list are multi-national chains which sell roughly the same mix of products and services across their outlets in different countries. The other categories are shops which tend to be independent or smaller scale chains.

What follows are merely observations, rather than a controlled and rigorous study. So don't read too much into it. This bit of the post started out rather light-heartedly.

Before looking at these areas my expectations were that shops such as Games Workshop would be much better mapped than Accessorize and Claire's; and furthermore that they would be tagged more consistently. I had similar expectations for comparing Model Shops to Knitting Shops in terms of frequency of mapping, but anticipated that these would have less consistency in tagging. Using a brand name makes it much easier to examine this phenomenon than for independent shops within a category.

Chain Retailers

Games Workshop

I specifically chose Games Workshop because it grew very substantially in the early 1990s after being listed on the London Stock Exchange. Given the known demographics of mappers I thought there was a very good chance that most mappers would be males under 40, who were therefore likely to have been exposed to Games Workshop products as teenagers. I have carried out straw polls on IRC and (most recently) at the London OSM meeting at the Barley Mow in Marylebone. Certainly there is very high brand recognition for Games Workshop, some of its products and even key personnel.

Giant role-playing figure outside Games Workshops HQ in Nottingham
from Geograph by  John Sutton

However, my expectations were confounded. Not only are relatively few Games Workshop shops mapped (33 or 34 out of over 300), but there is an extraordinary range of tags used for this small number of shops with an identical range of products: game, games, toys, RPG (presumably for role-playing game, not rocket-propelled grenade, computer and hobby. The most commonly used tag is shop=toys which is quite censorious given that I was assuming many mappers might have been customers of the firm in the relatively recent past. Given that I concur with Yu-Wei Lin's statement &emdash;

"Being able to do something for a place which has special meaning is one of the key motivations for these contributors. Mapping, just like blogging, is a personal way of expressing, preserving memories, documenting one’s footprints, a way of travelling and a way of remembering events in everyday life." (p. 61 op. cit).

&emdash; this suggests that my assumption that many mappers would have played Games Workshop games as teenagers, is probably wrong.

Accessorize and Claire's

Accesssorize is another UK-based company with shops worldwide. According to their website they have 240 shops in Britain and over 800 in total. Of these 68 are mapped in OpenStreetMap, 32 of them in the UK. So once again a relatively small number of outlets mapped and again a massive range of different tags for them: "jewelry", "fashion",  "clothes", "bag", "boutique", "yes", "accessories", "babskie_pierdoły", "stationery". I'd expected the range of tags, but in terms of overall numbers mapped a chain which I'd expected would mainly have resonance with young women is about as well mapped as one where I thought the opposite applies.


Claire's Accessories is, as far as I'm aware, a similar but larger US-based chain with over 3500 shops, about half in North America and over a 1000 in Europe. Some may be tagged just as "Claire's" so its a bit harder to cross check, but there are about 42 mapped in total, and once again a wide range of tags: "accessoires", "accessories", "bijouterie_fantaisie", "boutique", "clothes", "clothing_accessories", "costume_jewelry", "fashion", "fashion_accessories", "fixme", "general"
"generic", "gift", "jewelry","shoes","yes". Some of these values may be false positives, but the same pattern is apparent.

Summary

Had I by chance chosen two different chains of shops which are intrinsically difficult to categorise? Had I completely misinterpreted the sorts of retail outlets which might be of interest to teenagers and 20-somethings (and remember, peer-pressure often forces people to feign greater interest at that age)? Or perhaps these aspects of peri-teenage life are something the typical OSM mapper is happy to forget? I certainly failed to find any obvious gender-bias in how these were mapped.

Speciality Retailers

What about particular hobbies which are usually associated with one or other gender. 

Knitting and Wool shops

14 knitting shops, and 34 of the more popular wool shops. There are also sewing shops (92), haberdashery (94), not forgetting general craft shops (452) and many department stores (14, 694), which often started out selling fabrics and wool. However, it is the specialist shops which will have a wider range of wools, more interesting patterns and will usually have like minded staff. I think we've mapped a couple in Nottingham and a total of 6 in the urban area of which 3 are knitting/wool shops. I should have checked with our local expert to see if we've missed any.

A lovely knitted Nautilus created by palaeontologist Hallucygenia on Flickr
copyright Hallucygenia.

As Nottingham is pretty well mapped for shops this gives a run-rate of about 1 wool shop for every 200,000 inhabitants, or 1 knitting/sewing shop / 100,000 inhabitants. In other words every reasonable sized town should have one.

These are definitely under-mapped, and once again we have quite a range of options for tagging these. In practice, only regular customers of such shops can say if a fine-grained tagging scheme is needed or not.

Model Shops

Building scale models of ships, trains, aircraft and so on, used to be a rite of passage for engineers. Anyone who has visited places like the Glasgow Transport Museum with it's intricate models built by engineering apprentices should be aware this was once a valued skill. There are plenty of people who derive great enjoyment from building models. They are catered for by specialist shops, often run by fellow enthusiasts. My impression is that this is an overwhelmingly male hobby.

Ship models in Liverpool Maritime Museum
via Wikimedia Commons

 So how many shops catering to this activity have been mapped? There 28 different tags containing the string "model" representing 84 outlets. Plain "model" is the most popular (21), with obvious variants such as modelling, model_train and model_trains following. This is a different situation from the other shop types I've mentioned so far as it is clear just looking at the tag values that these are clearly closely related concepts. The range of values here reflects that most mappers were probably not able to find any documentation on a suitable tag and created them ad hoc.

There are two model shops mapped in Nottingham, and another closed last Summer. This suggests a similar level of demand to wool/knitting shops as a basis for extrapolating some kind of likely total number of shops.

Summary

Once again trying to select shops which potentially might show a gender-bias in mapping does not obviously exhibit the phenomenon.  These specialist shops are clearly not very well mapped. If we do a straight extrapolation on population one might expect at least 200-300 knitting and model shops across the United Kingdom. Currently we have 13 of the former and 14 of the latter. In both cases under 1% of the likely total.

Mappers seem to be reasonably consistent: they ignore both types of shop!

Fin

This post is very long and an unfortunate mixture of the quite serious and stuff which was meant to be light-hearted. My apologies in retrospect.

Turning to the more serious side. I hope what I have written about childcare is of some use for people who wish to capture this sort of information, but I also hope it acts as a corrective to the somewhat corrosive story emanating from activity on one wiki page a couple of years ago.

I also hope that people will go and map more knitting and model shops "just because they are there".

I failed in my little experiment to find gender-bias with the examples I chose. Perhaps there are other examples which show this up more clearly. However, please do as I have and spend time examining how things have been mapped. Note that there are no wiki pages for my examples. This was not by design but it meant that I had to work with what the data told me.

Anecdotally, looking at these shops suggest that the self-selection of OSM mappers has other biases which are not so obvious. Perhaps elucidating these might also help in the shared goal of involving more women in OSM (and in the US things seem to be going very well in this direction).

My own feeling is that if OSM never breaks out of recruiting mappers from STEM disciplines it's going to be a long-hard struggle. We can work to get more women contributing as mappers, developers, designers, or users in GLAMs, for instance, but we sure as hell are not going to fix recruitment of women into STEM subjects on the way!