Longtime readers of this blog will recall that I have previously written about the census in some detail. And if you’ve been about the last day or so, you’ll have seen the fuss over the news that the Office of National Statistics might make declaring your sex optional in the 2021 census.

The full details are not known as the Sunday Times story is from a leaked report and contains few details. This did not stop the Times breathlessly rushing to get quotes from two well-known anti-trans “feminists”. It seemed no trans people were asked for quotes.

The misreporting around this issue is even more widespread than usual for trans-related stories. Misreporting caused by concerns that facts will get in the way of a good headline that further demonises trans people. So here are some myths already doing the rounds, debunked.

Myth: Gender will no longer be recorded in the census, or we will have inaccurate data on gender
There are many forms in existence in which make gender optional, and most people still tick the appropriate option. Religion is a far more sensitive issue and even when the question was made optional in the 2011 census, only 7% of people chose not to provide an answer.

And the Office for National Statistics are likely to use “imputation” to fill in the gaps – a system they routinely use and causes problems analysing statistics for minority communities, such as trans people and poly households. In a nutshell, if you put your name as Mary and don’t tick “Female”, the ONS may still record you as female in statistics.

Myth: This will erase women’s identity
Tick the box, or don’t tick the box. Your choice, nobody else’s. Established religion has not disappeared since the question was made optional in 2011.

Myth: Trans people and only trans people will not answer the question, which is why it’s optional
Most trans people identify more strongly with one gender than the other, and are likely to simply tick the box they most closely identify with. The groups most likely to skip this question are, roughly in order of likelihood:

  1. Those who think it is “none of your damn business”. (A view that’s been held by some feminists quite separate from trans concerns for some time)
  2. Respondents who didn’t understand the question, perhaps because of language issues
  3. People who simply couldn’t be bothered.
  4. Non-binary people

I predict that we will see some attempts to extrapolate the non-binary population of the UK based on 2021 data. I doubt such extrapolations will be valid.

Did the Office for National Statistics get this right?
In a word, no. At least, it doesn’t look like they have but it is hard to say until the final report is published – it may be that there has been selective quoting from the report in the original article (£) and the ONS considered other factors more important. In particular, I find it a little over the top for a government report to state that asking about sex rather than gender is reallys “unacceptable”. But that may have been one side point in a longer article, or reporting the views of a focus group.

And “Other” was rejected because it was “thought to homogenise trans people and differentiate them from the rest of society“. That statement suggests that whoever wrote the ONS report has a tentative grasp of trans issues at best, as it is only true if you start from the assumption that all or most trans people will tick “other”. Liberal Democrat Conference speakers cards have an “Other/Prefer Not To Say” option on them for gender for over a year, and we have not had any complaints about that. If we were going to get complaints about getting an equality issue wrong, that’s precisely the environment in which I would expect them to surface.

And finally, there seems to be some conflation in the quotes between non-binary and intersex issues. It is unclear why the ONS has lumped intersex people in with non-binary people in this way, as they are overlapping but still distinct groups in much the same way as being Irish and having red hair are.

From the 2011 England & Wales Census, it has been revealed that 39,200 relationships were edited/”corrected” due to an apparently invalid combination of relationships that would indicate a Polyamorous household.

Firstly, I’ll note they only did this where there was also a marriage or civil partnership involved – I.e. a triad or V where one side of the relationship has been legally recognised, so the figures will be low. Secondly, the reason the Census office edits these figures is because they believe in the majority of cases this is probably an error by those filling in the form.

Still, the figures are potentially interesting, so lets see how they stack up.

I can not find any data on the number of relationships recorded by the Census, but we know there were 23.4 million households. To get some idea of the kind of proportion of households that have indicated a Poly relationship, lets assume a maximum of one poly family per household: that gives around 0.1%-0.2%.

I am deliberately not going to be any more accurate than that because the data I’m working from is vague at best. As well as not including families without any marriage/civil partnership involved (Pushing the number down) there may be households where a couple have not yet divorced but a new partner has moved in. (Which would push the number up)

How likely is it that much of this is an error in filling in the forms? We have no real way of knowing. But I also asked for the number of marriages/civil partnerships that were edited due to an invalid combination including gender, such as indicating a marriage between a same-sex couple. That gave 50,400 edits. There are just over 100,000 civil partnerships in the UK and I can well believe somewhere around half of respondents would be bloody-minded enough to tick “marriage” and not civil partnership. There will also be Trans issues affecting this figure, but at a low enough rate to not be significant.

The above is, of course, not statistically sound in any sense and shouldn’t be used for the basis of policy or, well, anything really.

However, it’s enough to indicate that there may be a non-trivial number of Poly households out there and that warrants further study. Perhaps the Office for National Statistics can be persuaded to report in more depth on the data – they are far better statisticians than I for starters – or to conduct follow-up surveys. If the number proves to be large enough, it may even justify better investigation in the 2021 census.

Following on from my recent post on how the Office for National Statistics “corrects” peoples gender and marital status in the Census, I asked them how often this had happened in the most recent three censuses.

They don’t know.

I’m finding it a little strange that the ONS would have their computers correct data in this way and not track how often it is occurring. The ONS previously claimed that “…the majority of respondents recording themselves as being in a polygamous relationship in a UK census do so erroneously, for example, ticking the wrong box for one household member on the relationships question.” Given that they do not track any information on such responses, their “recognition” that such relationships don’t exist may boil down to “Well, nobody who works here is in such a relationship, so they can’t possibly exist”.

(The answer they provided makes reference to impudation, but it’s a generic measure tracking overall error rates in the census, based on a more detailed follow up survey to check the results. It does not appear particularly relevant to the question.)

Yes, there’s also a follow up FoI request I have just submitted, asking what other corrections are made and how they are justified.

It appears that the Office for National Statistics, when reporting on census data, will simply get their computer try to guess at what people intended.

If you’re a gay man and ticked the married box then the Office for National Statistics will ensure the “mismatch is… resolved using a probabilistic statistical system [to] alter one or more variables to make the response consistent“. And yes, they specifically state this could result in the system “changing the sex of one individual“.

This completely ignores the fact that married – not just Civil Partnered – couples of the same sex are entirely valid. Perhaps it’s a foreign marriage, or you’re Trans and married but don’t have a GRC for any of a whole host of reasons?

Oh, and they might divorce you anyway: If you indicate multiple relationships (A Poly household for example) they’ll just pick one to ensure everyone is in nice neat couples for their system. I guess anything else would just be Too Complicated for the statisticians.

I’ve submitted a followup FoI request to find out how common this has been in previous years. (Although previous years will not have had Civil Partner as an option)

Lies, damned lies and statistics?

I rather suspect that for both these categories, the demographic is so vanishingly small that it will disappear into the noise, but having filled in the Census last night I was interested in how some of the data is processed, so I fired off a Freedom of Information request.

Firstly, poly households:

1a) If a household consists of a polygamous relationship, will the data be accepted by the Census system? For example, if P1 and P2 are married or in a civil partnership and P3 indicates they are a partner of both P1 and P2, is this considered a valid response or will the “partner” response be ignored and not entered?

1b) If the data will be accepted and entered into the census computer systems, will it be either reported on or (In summary form) available via an FoI request?

In short: Will you be able to answer the question “How many poly households are there in the UK?” I doubt many people will accurately answer the question so the ONS figures won’t mean much. However, I don’t know if there have been any previous surveys in this area. (If anyone is aware of any, let me know!)

Having just started to read The Girl with the Dragon Tattoo, I do wonder if there are more people quietly getting on with their lives in such relationships than is usually reported on. Even more so than Trans issues, there doesn’t seem to be much of a poly community because stable poly relationships tend not to be the kind of thing one can really seek out. They just kind of happen and people are either open to being poly or not.

Someone is bound to disagree with me on the previous paragraph so I’ll caveat it now: In my own opinion, of course.

(Our census has been done on paper. Does anyone know if the online version restricts your answers at all and would not allow the response I’ve described above?)

Secondly question:

2) If there is an apparent mismatch between the indicated sex and marital status of individuals, how will this be entered into the system and handled? For example, if two individuals indicate sex as female but also indicate they are married, will this be entered into the census system as a marriage or as a same-sex civil partnership?

Translation: How many people are so militantly pro-equal-marriage that they’ll tick “married” despite technically being civil partnered. (And Vice-Versa, but I guess that’s a smaller group) There will also be people who have transitioned but have not received a GRC as they do not wish to divorce or are unable to obtain a GRC for any of the myriad other reasons, but there is no way of distinguishing between these two groups from census return data.

I expect to be disappointed with the response – I generally am with FoI requests like this – but they are potentially interesting topics if something useful does come of it.

Stonewall do seem to struggle when figuring out which campaigns to support, don’t they?

No, they haven’t failed to support one this time. Instead they’ve decided to jump on last decades Jedi bandwagon by asking people to list their religion as “Lesbian”. This is a really bad idea for two reasons.

Firstly, putting any religion in there will mark you down as religious, no matter how silly it is. It’s far better, as the Census Campaign have been urging people to do, to tick “No Religion”. And secondly, the reason for the Census Campaign in the first place is that previous censuses have returned bad data – and the ONS survey results suggest that a census sexuality question would fare no better.

Yes, this is a bad enough idea that Amy Lame, who had the original idea, changed her mind about it.

I’ve seen a few people around the place wonder why the census does not ask about sexuality or transgender status. I, for one, am glad it does not. You’re probably all aware of the Census Campaign, encouraging people to tick the “No religion” in response to the optional question on religion.

Religion is, compared to sexuality or transgender status, usually (But not always) a less controversial topic. Some may recall the ONS study, “revealing” that only 1.5% of people are gay/bisexual. That was with some careful work done to ensure that others could not overhear the answers being given. Despite that, it came out with a surprisingly low figure with a significant number of unknowns/won’t says.

In comparison, the census is relatively public and can be seen by other members of the household. Yes, you could request an individual questionnaire but if someone really fears being out to their housemates that much, are they really going to risk revealing they have something to hide? Simple human nature – procrastination – will mean they probably can’t be bothered anyway. They’re just go for the easy option and tick heterosexual.

We can’t expect everyone to be an activist.

And that’s without even considering the problem of identity. How many people have had a same-sex relationship, but regard themselves as heterosexual because they are now in an opposite-sex marriage? Or started out “straight” before having a couple of gay relationships, so would tend to go for “gay” over “bisexual”. Or just don’t know yet.

We do not need another ONS-type result claiming there are fewer LGBT people in the UK than there really are. Officials would base their funding allocations on these numbers and there will be less available in the way of LGBT resources, in much the same way as the skewed religion result causes problems.

Inaccurate data is worse than no data.