Gill, R (On the Application Of) v Cabinet Office [No.3]

Every 10 years since 1801 (except during WWII), there has been a census of the population of the United Kingdom. Since the 1991 census, information as to the ethnicity of respondents has been sought. Responses to the ethnic group question in the census form are provided by means of ticking the appropriate box (“tick-box”). If there is no tick-box considered appropriate, then the respondent can insert the appropriate ethnic group in a box marked, “Other”. Many within the Sikh community have long campaigned for the presence of a Sikh tick-box as a distinct available response to the ethnic group question, claiming, amongst other things, that the absence of such an option leads to significant undercounting of the Sikh population. The Claimant, Mr Amrik Singh Gill, is the Chair of the Sikh Federation UK, which is one of the leading Sikh community groups pursuing a Sikh tick-box option.

For the 2011 Census, the Office for National Statistics (“ONS”) did not recommend the inclusion of a Sikh tick-box and none was included. The same recommendation was made in respect of the forthcoming 2021 Census, and this was reflected in the Census Order 2020 made by Her Majesty in Council on 20 May 2020 (“the Census Order”). The Claimant challenges the failure of the Census Order to include a Sikh tick-box as unlawful and seeks a declaration to that effect and a quashing order.

It is important for the reader to note that this judgment is not concerned with whether or not there should be a Sikh tick-box in the census form or with the respective merits of the arguments for and against such a tick-box. Such matters are not for the Court to determine. This judgment is concerned solely with the question whether, as alleged by the Claimant, the Cabinet Office, which has responsibility for laying the necessary legislation for the 2021 Census, has acted unlawfully in the process leading to the making of the Census Order. This is solely a question of law.

As to that question of law, I have found against the Claimant. In summary:

Ground 1 of the claim fails because the ONS did not, as alleged, fail to apply the published policy in respect of assessing tick-boxes and nor did it apply another hidden or secret policy to that exercise. In my judgment, the ONS conducted its assessment in accordance with published criteria.

Ground 2 of the claim fails because the ONS did what it said it would do in that it applied an updated version of a prioritisation tool developed for the 2011 Census.

Ground 3 of the claim fails because the ONS did not apply its criteria inconsistently as between different topics.

Ground 4 of the claim fails because a report on which the ONS placed some reliance did not contain, as alleged, any “legally unsustainable” conclusions.

The ONS is the executive office of the UK Statistics Authority (“UKSA”), which is a non-ministerial department sponsored by the Defendant. The UKSA is empowered to conduct a census in England and Wales (the census being a devolved matter with the Scotland and Northern Ireland governments conducting their own censuses). The UKSA makes recommendations to the Minister and the Cabinet Office about the content of a proposed census and how it should operate.

The ONS is the UK’s independent National Statistics Institute and is widely recognised as a world leader in the production of official statistics. It publishes statistics in relation to a wide range of matters including the UK’s National Accounts (including GDP), prices (including the retail and consumer price indices) and on births, marriages and deaths.

The ONS undertakes the census every 10 years in England and Wales. The census provides valuable data which informs decision-making regarding the distribution of Government funds, the supply of grants and the provision of local services. The information obtained also provides a valuable insight into the social condition and fabric of the population. Such information enables providers of local services and employers to gain a better understanding of the populations with which they deal in respect of issues such as public safety, fair recruitment and the provision of appropriate services.

The ONS draws on its experience and expert technical knowledge and that of external experts in assessing and advising upon the design and content of proposed census questionnaires. The questionnaires seek information under different subject headings

There are different methods of capturing answers to questions in the census form: either a list of possible answers set out in separate tick-boxes, a blank space in which the respondent can enter their own answer (“write-in option”), or a combination of tickboxes and a write-in option. Tick-boxes assist respondents by providing a quick and convenient means of identifying the desired answer, and they promote consistency of response. Most topics are mandatory, that is to say, the respondent must provide a response to the question(s) asked under that topic. Only one topic, namely “religion”, was voluntary in the 2011 Census.

The majority of topics to be covered by the census remain largely the same over time. However, where a need is identified though consultation, research and evidence gathering, new topics may be added. For example, as stated above, the topic of ethnic group was added to the census in 1991. In 2001, “Religion” was added, as were the topics of “Passports held”, “National Identity” and “Language”. The 2021 census adds three new topics, namely, “Veteran Status”, “Sexual Orientation” and “Gender Identity”. It is clear, given the sensitive nature of many of the topics, that care needs to be taken in the design of the question; a badly worded or inappropriate question might cause a respondent not to engage further with the census or produce poorer quality responses. Topics can also be discarded if no longer required or other sources of information are available. For example, the census ceased asking about outdoor WCs in 2001.

For each census, the ONS undertakes a formal non-statutory consultation exercise on which topics to include. It does not undertake a formal consultation on the response options, but it engages with stakeholders about the questions to be asked and the response options that are to be made available for those questions.

The last census conducted in the UK took place in March 2011. It was the first census that could be completed online. For the topic of “Religion”, the answers available to respondents included a “Sikh” tick-box. A total of 423,158 respondents ticked the Sikh tick-box for religion.

For the topic of “Ethnic Group”, respondents were asked the question, “What is your ethnic group?”. The respondent could choose from one of five sections, “A” to “E”: A – White; B – Mixed/multiple ethnic groups; C – Asian/Asian British; D – Black / African / Caribbean / Black British; and E – Other ethnic group. For sections A to D, respondents were asked to “tick one box to best describe your ethnic group”, from a list of tick-box response options. The options available under C – Asian / Asian British were: Indian; Pakistani; Bangladeshi, Chinese; and then a write-in option for “Any other Asian background”. Selecting E – Other ethnic group, enabled the respondent to write in their ethnic group.

Whilst there was no Sikh tick-box, “Sikh” could be entered as a write-in response under either “C – Asian / Asian British” or “E – Other ethnic group”. In the 2011 census, a total of 83,362 respondents identified their ethnicity as Sikh, 76,500 of whom also identified their religion as Sikh. The number of respondents choosing to identify only as ethnically Sikh was therefore 6,862.

The Claimant considers that the absence of a Sikh tick-box option leads to substantial undercounting of the Sikh population, estimated by the Sikh Federation to be in the region of 7-800,000. That figure is said to be based on the Sikh Federation’s

The Sikh Federation, amongst others, campaigned to have a Sikh tick-box response option included under the ethnic group question for the 2011 Census. The evidence of the ONS, given by Mr Iain Bell, Deputy National Statistician and Director General of Population and Public Policy at the ONS, is that whilst there was “some demand” for such a tick-box, mainly from Sikh community organisations, other views were also expressed. Mr Bell refers to cognitive testing commissioned by the Scottish Government for the Scottish 2011 Census to determine whether a Sikh tick-box should be included. That exercise, which involved test questionnaires that included the Sikh tick-box option, revealed that most of the seven Sikh respondents interviewed ticked the “Indian” response option and either did not notice the Sikh response option, or had noticed and been uncomfortable about having to choose between the two. Some of those interviewed felt that Sikhism was a religion and were confused by the inclusion of “Sikh” as an option under ethnic group.

In March 2009, the ONS published an Information Paper entitled, ‘Deciding which tickboxes to add to the ethnic group question in the 2011 England and Wales Census’ (“the March 2009 Paper”). Its recommendation was that only two tick-boxes be added for the

As we shall see, it is the underlined criterion (referred to by the ONS as “acceptability”) under the third theme at 3.2 of the 2011 PT that is of particular importance for the purposes of this claim.

Each criterion could result in a score of ‘0’ (low level of evidence), ‘1’ (medium level of evidence) or ‘2’ (high level of evidence). Each response option was scored against each criterion and those scores were then weighted in accordance with the scheme of the 2011 PT. For acceptability, the Sikh tick-box response option was scored ‘0’, meaning that there was a low level of evidence that its addition is “clear and acceptable to respondents … and provides the required information to an acceptable level of quality”. The rationale for this score was set out at 5.9.4 of the March 2009 Paper:

The 2011 PT was applied to 22 different ethnic identities being considered for inclusion in the 2011 Census. The two that were selected for inclusion, namely “Gypsy or Irish Traveller” and “Arab” scored the highest. The Sikh response option came 13^th out of the 22.

In January 2010, the Sikh Federation commissioned a report entitled ‘Can the ONS be trusted with the 2011 Census?’ The report was highly critical of the ONS’s analysis of the Sikh response option (although not of the PT itself) stating as follows:

The ONS commenced consultation on the 2021 Census topics (“the Topic Consultation”) in June 2015. As with the 2011 Census, the Topic Consultation considered new topics as well as whether existing topics were to be retained. The Government’s response to the consultation, ‘Assessment of initial user requirements on content for England and Wales – response to Consultation’, which was published in May 2016 (“the May 2016 Consultation Response”), stated that its aim was:

The evaluation criteria used in the Topic Consultation to determine which topics should be included were split into three groups: “User Requirements”, “Other Considerations” and “Operational Requirements”. These topic evaluation criteria (“the Topic Criteria”) were summarised in a table on p.4 of the May 2016 Consultation Response. Five criteria were included under “Other Considerations”: Data Quality; Public Acceptability; Respondent Burden; Financial Concerns; and Questionnaire Mode.

These criteria were further explained in Table 3 of the Consultation Response. The full title of the second criterion under “Other Considerations” was “Impact on Public Acceptability”. It was described as follows:

The Claimant’s challenge is concerned directly with this criterion of “(impact on) public acceptability” (“the PA Criterion”), his case being that the ONS stated repeatedly from May 2016 onwards that it would apply the PA Criterion (amongst others) in determining not only the topics but also which response options to include in the 2021 census, and that it failed to do so in that it applied an unpublished version of the 2011 PT instead.

Applying the Topic Criteria to the various topics under consideration led the ONS to propose three new topics and to retain the existing ones, including the ethnic group topic, which was considered to have a low impact on public acceptability. Under a section entitled “Next Steps” (p.27 of the May 2016 Consultation Response), the ONS described the proposed development activities for each of the topics. In relation to the ethnic group topic, the ONS stated that it would “consider whether there is sufficient need for any additional response categories in the ethnic group question”.

Further detail of the ONS’s analysis was provided on the topic specific report on the ethnic group topic, published in May 2016 (“the May 2016 Topic Report”). The introduction to the May 2016 Topic Report referred to the Topic Criteria and stated that these “largely reflect those used in the 2011 Census topic consultation and have undergone expert review within ONS and via the census Advisory Groups for use in the 2021 Census topic consultation.” (my emphasis). Mr Wolfe QC, who appears for the Claimant in this matter with Ms Christie, submits that the underlined words reveal a tendency on the part of the ONS to underplay quite important variations and changes from earlier iterations of criteria that the ONS employs. I do not accept that submission. I do not consider that describing the Topic Criteria as “largely reflect[ing]” those in the 2011 Census, in which similar headings were used, was inaccurate or misleading. It is to be expected that criteria will be developed and adjusted over time, particularly with such a large period between censuses, and the ONS appears here to be acting transparently in referring to the “expert review” of the criteria resulting in some changes such that they are not identical to those previously applied. There is no suggestion in the May 2016 Topic Report that the Topic Criteria (including the PA Criterion) would be used in relation to the assessment of anything other than topics.

The May 2016 Topic Report noted that, whilst the consultation was, at that stage, about topics and not specific responses, the ONS nonetheless received a number of requests for additional response options. These included a request for a Sikh tick-box option. In a section headed “Next Steps”, the ONS referred to these suggested response options and stated as follows in a series of bullet points:

ONS intends to undertake a review of the ethnic group response options, and will consider this alongside the national identity and religion response options. This will involve consultation with stakeholder groups that have expressed an interest in this question.

The review will follow a similar format to that undertaken prior to the 2011 Census whereby response options were prioritised. This methodology is described in the [March 2009 Paper] [Hyperlink to that paper]. This methodology will be reviewed and updated to reflect current legislation. This will involve engagement with key stakeholders to ensure data needs to support the Public Sector Equality Duty under the Equality Act 2010 are well understood.

…” (Emphasis added)

The methodology described in the March 2009 Paper is the 2011 PT. Notwithstanding that, Mr Wolfe submits that this is not a reference to the 2011 PT because the words

The Claimant submits that “From May 2016 onwards, the ONS made a clear and repeated commitment to apply its published “public acceptability” criterion to the assessment of whether to include a Sikh ethnic group tick-box”. The only relevant documents published as of May 2016 were the May 2016 Consultation Response and the May 2016 Topic Report, neither of which in my judgment, for reasons set out, made any such commitment at all.

At a meeting with the Sikh Federation on 17 August 2016, the ONS presented a slide which was materially the same as the page in the May 2016 Topic Report describing next steps. In particular, it referred to the ONS’s intention to review the proposed ethnic group response options using the methodology (as updated) used for the 2011 Census. The “methodology” was a reference to the 2011 PT. There was no suggestion at this meeting that the Topic Criteria, and in particular the PA Criterion, used to determine which topics would be included in the census form, would be applied to tick-box response options.

The 2011 PT was (as prefaced in the May 2016 Topic Report) the subject of an internal review conducted in Autumn 2016 to update it for the 2021 Census. In November 2016, the updated PT (having been considered by a group of experts from the ONS and representatives from the Scottish and Welsh governments, and other national statistics bodies), was discussed with the Equalities and Human Rights Commission and the Ministry for Housing Communities and Local Government. In January 2017, the ONS created the 2021 Census Ethnic Group Assurance Panel (“the Assurance Panel”) to gather stakeholder views on the development of the ethnic group question. The Assurance Panel included representatives from many organisations both governmental and non-governmental, among them two Universities. The updated PT was shared with the Assurance Panel in January 2017.

The updated PT (“the 2021 PT”) (subsequently published in June 2019 in the Information Paper: The ethnic group prioritisation tool; 2021 census in England and

Comparability with 2011 data

This is the relevant criterion for present purposes and, as can be seen, there is no material change between that used in 2011 and that for the 2021 Census.

A follow-up survey was conducted by the ONS in November 2016 (“the November 2016 Follow-Up Survey”). This was described (in the document introducing the survey, the survey itself being accessed via a hyperlink) as the “next step in the development of the ethnic group question for the 2021 Census”. Under a heading, “Evaluation Criteria”, the November 2016 Follow-Up Survey states that:

The document goes on to list certain criteria that “will be used to decide on the ethnic group question’s design and outputs:”. This list does not include “public acceptability” or the PA criterion.

The Claimant contends that the November 2016 Follow-Up Survey confirms that the PA Criterion would be used to assess all aspects of the ethnic group question, including the response options. The Defendant disagrees and submits that a distinction is to be drawn between the design of the question, i.e. how the question is to be phrased and what terminology is used, and the assessment of new response options. The November 2016 Follow-Up Survey itself (which only provides a link to the actual survey questions) is ambiguous. It is certainly not unreasonable to interpret the references to the “question’s design and outputs” as encompassing both the questions and the response options. The survey is not in the documents before the Court. However, the ONS’s response to the follow-up survey, which was published on 20 March 2019 (“the March 2019 Survey Response”) is in the papers, and is instructive. This sets out the questions actually asked in the survey and analyses the responses. The March 2019

In my judgment, reading the November 2016 Follow-Up Survey with the survey questions themselves (as described in the March 2019 Survey Response) would make it tolerably clear to the reasonable reader that this survey was about the design of the question, including the extent to which the ethnic group question ought to be subdivided into the existing categories, and whether the existing terminology was acceptable. These are, as the Defendant submits, matters concerned with the design of the question under

The ONS’s view was that it is not possible, for various (perfectly understandable) reasons, to include 55 new tick-boxes in the census form. The ONS therefore engaged in a process of research and engagement with numerous communities to decide which, if any, of the 55 proposed responses should be included. There was extensive engagement with the Sikh community involving approximately 9 meetings over a period of 3 years from January 2016. I mention just some of those meetings here.

On 12 October 2017 Mr Bell met with the Crossbench Life Peer, Lord Singh of Wimbledon, who explained that the Sikh community does not speak with one voice, and that there was more than one view on the issue of whether there ought to be a Sikh tick-box option. Lord Singh expressed the view that the views of the Sikh Federation UK on this issue were not representative of the community as a whole.

On 23 October 2017, the ONS hosted a meeting to discuss statistics about the Sikh community. The purpose of the meeting was to provide an update on the development of the ethnicity question and to seek the views of the Sikh community on whether or not to have Sikh tick-box as a response option to the ethnic group question. A note of the meeting shows that it began with an overview showing the “journey from the topic consultation to where we are now in the process towards finalising recommendations on the topics and questions for the 2021 Census. This included details of the evaluation criteria and timelines for finalising questions.” The slides shown at that meeting included one at p.6 of the presentation:

Consultation launched in June 2015 collecting evidence of information needs from the 2021 Census

279 responses from organisations, 816 responses from individuals

Assessed evidence against a range of evaluation criteria

The slide continues with a table which sets out the Topic Criteria. Amongst the criteria highlighted was “Public acceptability”, the PA Criterion. The Claimant’s submission is that the highlighting of the PA Criterion as being one of those criteria of “most relevance to today”, at a meeting convened to canvass the views of the Sikh community on tick-box options amounted to a clear indication that the PA Criterion under the Topic Criteria would be applied in assessing the Sikh tick-box option. I do not accept that submission. It was made clear from the outset of the meeting that it would be in two parts: the first providing an update on the process so far; and the second seeking the community’s views on response options. The reference to the Topic Criteria (and therefore to the PA Criterion) was in the first part of the meeting. Moreover, it is apparent from the content of the slide that it is describing past events and assessments already undertaken using the Topic Criteria. There is no reference here to the assessment of response options. The following slide (at p.7 of the presentation) discusses further points about topics, many of which are general and not specific to the Sikh community at all. Then (at p.9 of the presentation) the ONS describes what further work remains on a “few topics” and confirms that “No decision has been made on whether or not to make changes to the ethnic group or religion tick-boxes for the 2021 Census.” It is only at p.11 of the presentation that the discussion moves to the second part of the meeting. I do not see, therefore, any real basis on which it could be said that the ONS was indicating or promising that the Topic Criteria in general, or the PA Criterion in particular, would be used in the exercise, yet to be carried out, of deciding upon the response options. Indeed, at p.15 of the presentation, the ONS refers to its “published next steps in 2016” for a review of the ethnic group response options. As set out above, that review was, according to the material published in 2016, to be conducted by applying the updated 2011 PT. Finally, at p.25 of the presentation, under the heading, “Making a decision”, the ONS indicated that the decision would be based on evidence gathered so far and “views gathered today on the acceptability of those options to the Sikh community” (my emphasis). The language of “acceptability … to the Sikh community” reflects the language of the 2011 PT: “clear and acceptable to respondents”, and not that of the PA Criterion under the Topic Criteria. The use of the word “today” (at p.6 of the presentation) would not, for these reasons, lead the reasonable reader or attendee to conclude that the highlighted criteria (including the PA Criterion) would be used to determine the response option issue.

Three response options were considered at the meeting in order to stimulate debate: (1) include a Sikh tick-box in both the ethnicity and religion questions; (2) only include it in the religion question; and (3) only include it in the ethnicity question. There was strong support for option 1 with all but two of those present voting in favour. However, concerns were expressed by some individuals after the meeting that “dominant voices left no room for debate” and that the views voiced “were not representative of the full Sikh community”. The meeting concluded with the ONS stating that it would use “published evaluation criteria together with legal advice to make an assessment for any additional tick-boxes to include on the 2021 Census”. The Claimant submits that the underlined words can only refer to the Topic Criteria as those were the only criteria published as at that stage. The Defendant responds that this was a reference to the 2011 PT as had been made clear in earlier documents referring to a review of the response options. In my judgment, any reasonable member of the public, aware of what had been expressly stated in the May 2016 Consultation Response, the May 2016 Topic Report and other documents, would be in no doubt that the reference to “published criteria” was a reference to the 2011 PT. At no stage up to that point had the ONS indicated that it would be applying the Topic Criteria, used to determine what topics are to be included, to the different and entirely separate question of which response options to include. The 2011 PT had been published, albeit in the course of the 2011 Census, and a link to the 2011 PT had been provided in the context of describing next steps in reviewing the response options during the consultation for the 2021 Census. I return to this issue below when considering Ground 1 of the Claimant’s claim.

The ONS also commissioned research from various bodies to inform its analysis. The ONS commissioned an online survey of two areas with a high Sikh population: Hounslow and Wolverhampton. This survey, conducted in 2017, involved asking participants to respond to census questions that contained a Sikh tick-box option for the ethnic group question. The survey revealed that all respondents who selected the Sikh ethnic group tick-box also selected Sikh as their religion. This was considered to be consistent with the results of the 2011 Census in which only a small proportion of respondents identified only as ethnically Sikh. The ONS concluded from this survey that there was no indication that the religion and ethnic group questions would capture significantly different Sikh populations, and that the Sikh option under religion was a good proxy for the ethnic group question.

In December 2017, the ONS published an update report (“the December 2017 Update

The December 2017 Update Report goes on to report the ONS’s progress on the 55 tick-box options that had been proposed. An initial prioritisation evaluation based on strength of need had reduced the number of options to eight: Gypsy; Irish traveller; Sikh; Somali; Jewish; Roma; Korean; and Kashmiri. These were then evaluated against additional criteria including the availability of alternative data sources, data quality and comparability. This left four groups – Jewish, Roma, Sikh and Somali - in respect of which further work was required before any recommendations could be made.

It will be apparent from that brief summary of the relevant parts of the December 2017

In March 2018, the ONS commissioned Kantar Public (“Kantar”) through open competition to conduct a qualitative study in respect of the four remaining ethnic group response options. Kantar’s research was conducted by way of focus groups sampled according to (self-identified) ethnic and religious identity. The Sikh focus group comprised 53 participants from a “demographic, and where appropriate, generational mix”. Mr Bell describes this sample size as “large” and “robust” for this type of qualitative study. He further notes that the purpose of the study was to obtain an indepth understanding of how a respondent reaches an answer for a given question, and that it was not a quantitative study from which one could infer that a percentage of the population would act in a particular way. The Claimant, whilst not seeking to challenge Mr Bell’s evidence in this regard, appears to seek to diminish the value of the study by reason of the small number of participants. Similar criticisms are made of the Scottish study referred to above in respect of the 2011 census and which had only seven participants. In the absence of any challenge to the veracity of what Mr Bell says or any evidence to the contrary, I accept that the sample size was adequate for the purposes of this type of study. In any event, the Claimant’s challenge is not based on any suggestion that it was impermissible for the ONS to take account of qualitative research.

ONS instructed Kantar to “assess acceptability, clarity and quality of tick-boxes” and provided a list of research questions for Kantar to consider using. These included: “Is a Sikh ethnic tick-box acceptable?” and “Are respondents uncomfortable with the term?”. The criteria of acceptability, clarity and quality (“the ACQ Criteria”) are consistent with those of the 2011 and 2021 PTs although those tools are not mentioned expressly in the instructions. The instructions also provided Kantar with the existing 2011 version of the ethnic group question and responses, with three further options relevant to the Sikh tick-box response: Sikh tick-box under “Asian”; Sikh tick-box under “other”; and no Sikh tick-box but with the option of Jewish included under “White”.

The focus group sessions lasted for 90 minutes each with 5 such sessions being conducted with the Sikh groups. The results of Kantar’s research were published in a report dated July 2018 (“the Kantar Report”). Section 1.3 of the Kantar Report provides:

The RAG rating scheme was as follows:

Green: More than 2011 ii) Amber: No change from 2011

Section 1.3.2. of the Kantar Report sets out Kantar’s findings for the Sikh focus groups:

The RAG ratings for the Sikh tick-box were as follows:

Acceptability – Red: “Seen as unacceptable – particularly amongst younger, second-generation participants”

Quality – Red: “Causes confusion as to whether participants had to choose between ‘Indian’ and ‘Sikh’ ethnic identity”

Clarity – Amber: “Generally participants did not identify with this tickbox apart from a small group of order, male participants”

Of the four ethnic group response options considered, only one, “Roma”, achieved Amber and Green ratings for all three criteria.

Around this time – July 2018 – the results of a survey of Gurdwaras (“the Gurdwara Survey”) conducted by the All Party Parliamentary group for British Sikhs were sent to the ONS. The survey question, which I have not seen, but apparently directed to whether the respondent was in favour of a Sikh tick-box response option under the ethnic group question, was sent to all Gurdwaras in the UK. According to the Sikh Council UK, there are 249 listed Gurdwaras. 112 Gurdwaras (48%) responded to the survey indicating their support for the Sikh tick-box option. It appears that the others did not respond at all. The Claimant states that the Gurdwaras have an “official membership” of 107,000 and an estimated congregation of 460,000. There is no evidence as to how or on what basis that estimate for congregation size was reached. I note in passing that the “official membership” figures for each Gurdwara, which in the majority of cases come to very round numbers such as 10,000 (there are three with a membership of precisely 10,000 and ten with a membership of precisely 1,000 or

The ONS states that it has taken the Gurdwara Survey into account, although it expected that such a survey would be undertaken in line with established social survey practices. The implication of that expectation being, as Mr Auburn submits, that the Gurdwara Survey was not considered by the ONS to comprise sound statistical research.

The ONS also conducted a piece of work to consider whether the additional tick-boxes would be acceptable to the general population. In May 2018, an external research agency conducted a quantitative online survey of 2,412 respondents. This survey found that 88% of the general population surveyed regarded the presence of a Sikh tick-box as acceptable.

On 14 December 2018, the ONS produced its Census topic research update (“the December 2018 Update Report”). Its recommendation was that a tick-box be included in the ethnic group topic for Roma as this had been assessed as having the strongest case for inclusion. Whilst Sikh, Jewish and Somali tick-box options would not be included, the online Census would include a “search-as-you-type” functionality to help respondents enter their ethnic group under the write-in option.

The December 2018 Update provided a detailed report on each of the topics considered for inclusion, including ethnic group. This summarised the research undertaken and stakeholder engagement in relation to the additional ethnic group response options, confirming that there was a continued need for data on ethnic group. At p.36 the ONS stated that it had “committed to undertake a review of the ethnic response options, involving a consultation with stakeholder groups that have expressed an interest in this question”. It also confirmed that it “…would use similar methodology to that used prior to the 2011 census. The prioritisation evaluation considered user need, alternative sources, data quality, public acceptability and comparability with the previous census.” At p.37, the ONS stated that it had “also evaluated effect on data quality, public acceptability and clarity (respondent burden) of the potential use of tick-boxes for the four identified groups.”

The Claimant points to these two instances of the use of the term “public acceptability” in the context of evaluating tick-box responses as being supportive of his case. The Defendant submits that it is clear from the context that the criterion being referred to is that of “acceptability” within the meaning of the PT and that the inclusion of the word “public” was no more than infelicitous wording in a document – one of many hundreds produced by the ONS – not intended to be subject to Chancery-drafting style scrutiny. In my judgment, the Defendant’s submissions are to be preferred. The documents published up to this point do not, as discussed above, indicate that the tick-box response options would be assessed against the PA Criterion of the Topic Criteria; instead there had been express reference to the use of the PT. If the December 2018 Update Report indicated otherwise, that would represent a major departure from what has transpired hitherto. The statement on p.36 of the Report seeks to summarise past events and refers to the fact that the ONS had “confirmed that [it] would use similar methodology to that used prior to the 2011 Census” and also to the “prioritisation evaluation”. These are clearly references back to the statements of intent in the May 2016 Topic Report which

The December 2018 Update Report also included the findings of some exploratory research undertaken by ONS. The purpose of this research was to see “whether the surnames captured in the 2011 Census could be used to improve [the ONS’s] understanding of the Sikh community from what is already derived from the religion question” and whether surnames could be used as an indicator of persons who may wish to identify as ethnically Sikh. The suggestion to carry out this research emerged from the ONS’s meetings with Sikh community groups. The research was based on a number of untested assumptions as to the correlation between surnames and Sikh ethnicity and/or religion. The modelling techniques used by the ONS in this research produced

In December 2018, the Defendant published the White Paper ‘Help Shape Our Future - The 2021 Census of Population and Housing in England and Wales’ (Cm 9745). The White Paper does not recommend including a Sikh tick-box under the ethnic group question. The reasons are as follows:

ONS has always provided an "other, specify" box within the ethnic group question, to allow respondents to answer as they wish to (such as defining their ethnicity as Sikh). With the online census in 2021 ONS is developing the "search-as-you-type" capability which will make it easier to use this option, making it easier for respondents to self-define their ethnic group (when a specific response option is not available).

The 2021 Census will continue to include a religion question, with a specific Sikh response option. Flexible data outputs will allow analysis of those who define their religious affiliation as Sikh (through the religion response option) and those who define their ethnic group as Sikh through the use of the "search-as-you-type" capability on the online ethnic group question.

ONS will estimate the Sikh population using alternative data sources to assess the numbers who may declare themselves of Sikh background but not through the religion question. ONS will strengthen the harmonisation guidance on the collection of religion alongside ethnicity data across government. ONS will also increase the analytical offering and outputs for all ethnic groups, through flexible outputs.

The proposals on utilisation of the Digital Economy Act 2017 (see Chapter 3, paragraph 3.183) will ensure that data on the Sikh population is available across public services. ONS will work with members of the Sikh population to encourage wider participation in the census and raise awareness of the options of writing in their identity in the ethnic group question.

ONS does not propose adding an additional specific response option to the 2021 Census ethnic group question because of the evidence that this would not be acceptable to a proportion of the Sikh population. ONS considers that the estimates of the Sikh population can be met through data from the specific response option in the Sikh religion question.

The proposals meet the user needs expressed to ONS and follow extensive investigation. Leaders of Sikh groups have provided information which has fed into the analysis. There are differing views within the Sikh population as to whether a

ONS received information from a survey of Gurdwaras enquiring about acceptance of a Sikh ethnic group tick-box, which showed a high acceptance for inclusion. The survey gave ONS more insight into the views of leaders of Sikh groups, alongside ONS's other research. Independent research was undertaken for ONS to further understand the acceptability of the Sikh response option within the ethnic group question.

Focus groups were conducted, with over 50 participants from Leicester, Birmingham and London who were spread across age, gender and life stages. These found:

that the inclusion of Sikh tick-box, without other religion tickboxes, within the ethnic group question was viewed as unacceptable – particularly amongst younger, second-generation participants

younger second-generation participants wanted to express their Sikh background through the religion question as this was how they expected Sikh identity to be recorded

a small number of older, male participants were keen to express their Sikh identity with an ethnicity Sikh tick-box and many stated that it was one of the most important aspects of their background

there was increased respondent burden with some participants confused about having to choose between an Indian and Sikh identity, and felt that they were being asked to make a choice when they felt they were both

The Equality Impact Assessment, which was published with the White Paper, stated:

As mentioned above, in March 2019, the ONS published the March 2019 Survey Response, and in June 2019, the ONS published the 2021 PT used to assess the ethnic group tick-boxes.

This is the Claimant’s third judicial review challenge relating to the absence of a Sikh tick-box in the 2021 Census. The first judicial review application (“JR1”), Gill (on behalf of the Sikh Federation UK) v Cabinet Office [2019] EWHC 3407 (Admin), was made in July 2019 before the draft Census Order was laid before Parliament. The Claimant proceeded at that stage because of his view, reasonably held, that early resolution of this matter was preferable. The Claimant was granted permission on the papers by Thornton J and the claim was expedited. It was heard by Lang J at a full hearing in November 2019. Whilst the Defendant resisted the application on the merits,

The Claimant sought permission to appeal.

On 2 March 2020, the draft Census (England and Wales) Order 2020 was laid before Parliament.

On 16 April 2020, the Claimant issued his second Judicial Review claim (“JR2”). The appeal against the decision in JR1 was withdrawn, that claim having been rendered academic by the laying of the draft Census Order before Parliament.

On 7 May 2020, Foster J refused permission on the papers in JR2.

On 20 May 2020, the Census Order was made by Her Majesty in Council. It did not include a Sikh tick-box response for the ethnic group question.

The Claimant then withdrew his claim in JR2 as it too had been rendered academic by the making of the Order in Council.

On 11 June 2020, the Claimant issued proceedings in the present case and sought expedition. The matter came before me on the papers on 16 June 2020. I granted expedition, abridging the time for the Defendant’s Summary Grounds of Resistance.

The matter came before Lewis J on 6 July 2020 and permission was granted on the papers.

That procedural background is set out in some detail because there was extensive argument as to its relevance to the relief that should or should not be granted in this matter, should I find the grounds made out.

The 1920 Act makes provision for the taking from time to time of a census.

Section 1(1) of the 1920 Act enables the Queen, by Order in Council, to direct the taking of a census:

the date on which the census is to be taken; and

the persons by whom and with respect to whom the returns for the purpose of the census are to be made; and (c) the particulars to be stated in the returns:

an order shall not be made under this section so as to require a census to be taken in any part of Great Britain in any year unless at the commencement of that year at least five years have elapsed since the commencement of the year in which a census was last taken in that part of Great Britain; and

no particulars shall be required to be stated other than particulars with respect to such matters as are mentioned in the Schedule to this Act."

The Schedule to the 1920 Act sets out the matters in respect of which particulars may be required. Paragraph 3 provides for "Nationality, birthplace, race, language" and paragraph 5A for “religion”. Paragraph 6 includes "Any other matters with respect to which it is desirable to obtain statistical information with a view to ascertaining the social or civil condition of the population."

Section 2(1) of the 1920 Act provides that it is the duty of the Statistics Board (previously the Registrar-General) "to make such arrangements and to do all such things as are necessary for the taking of a census in accordance with the provisions of this Act and of any Order in Council or regulations made thereunder …".

The UKSA holds the statutory powers of the Statistics Board, and acts through the ONS as its executive office (established under section 32 of the Statistics and Registration Service Act 2007). The ONS has responsibility for the delivery of the census.

Section 2(2) of the 1920 Act provides that the Statistics Board (in effect, the ONS) is subject to the control of the Minister. The Minister has responsibility for the drafting of the census secondary legislation and laying it before Parliament.

Section 3(1) of the 1920 Act empowers the Minister to make regulations providing for the conduct of the census, including the forms to be used in the taking of the census (at (f)). Section 3(2) requires the regulations to be laid before both Houses of Parliament and to be subject to the negative resolution procedure.

The Claimant contends that the Census Order, which does not include the Sikh ethnic group tick-box response, is unlawful. This is because the Census Order was based on recommendations made by the ONS which were reached following an unlawful evaluation process, and based on unlawful reasoning. Four specific grounds are relied upon:

The ONS failed to apply its published “public acceptability” evaluation criterion (the PA Criterion) to the assessment of whether to include a Sikh ethnic group tickbox option in the 2021 Census;

Alternatively, even if, contrary to its published statements, the policy to be applied was that set out in the 2011 PT (updated in light of the Equality Act 2010), the ONS unlawfully failed to apply even that policy, and instead applied a materially different one, which remained unpublished until June 2019, which was long after the White Paper;

The ONS failed to apply the PA Criterion consistently across questions/response options considered for inclusion under various topics/sub-topics in the 2021 Census; and

The ONS’s reliance on the Kantar Report was unlawful because it applied different evaluation criteria to those promised by the ONS and the Kantar Report also contains material internal inconsistencies and inaccuracies.

I shall deal with each ground in turn.

The legal basis of this challenge is said to be the decision of the Supreme Court in R

Mr Wolfe relies principally on the second and third of those propositions. Lord Dyson considers the second proposition further at [26]:

As to the third proposition, namely the need for policies to be published to afford the potential object of any discretionary decisions made under a policy the right to make representations, Lord Dyson said as follows at [35] to [38]:

Mr Wolfe submits that the clear, published policy for evaluating ethnic group tickboxes for inclusion in the 2021 Census was one which included the PA Criterion. However, instead of applying that policy and that criterion, the ONS instead applied another policy, unpublished at the time, containing different criteria and a different version of “acceptability”. This was, submits Mr Wolfe, a clear Lumba-style failure to follow published policy. The definition of “public acceptability” under the published policy required the ONS to assess whether the questions:

were “sensitive or potentially intrusive” and as such would have a negative impact on responses;

“may lead to respondents giving socially acceptable rather than accurate answers”; or

would result in the collection of “data that would deliberately promote political or sectarian groups, or sponsor particular causes”.

Mr Wolfe submits that none of these matters were assessed. He further submits that the distinction which the ONS seeks to draw between the Topic Criteria and those for tickbox response options is a false one. That is because the ONS continued to refer to the PA Criterion in relation to tick-boxes long after the decision to include the ethnic group topic had been taken, thereby suggesting that the same criterion would be applied to both issues. As there has been no lawful assessment of the “impact on public acceptability” of the Sikh tick-box option, it was unlawful for Her Majesty to make the Census Order in its present form.

The Defendant, represented here by Mr Auburn of Counsel, submits that the Claimant’s claim is based on a fundamental misunderstanding as to the distinction between the evaluation criteria for determining which topics/sub-topics should be included in the Census, and those used to determine which tick-box response options should be included. The Claimant bases his claim on an alleged failure to apply the Topic Criteria (which relate to the topics and question design), when what he is really complaining about is the omission of a Sikh tick-box (which relates to the choice of responses). The claim is, accordingly, misconceived and rests upon a misreading of the relevant statements made by the ONS in consultation and other documents, and in correspondence.

Mr Auburn further submits that, in the absence of any challenge based on legitimate expectation, failure to consult or irrationality, the Claimant is left with, at best, an argument that the ONS did not communicate its policies as clearly as it might have done. That, however, falls far short of a Lumba-style failure of applying an unpublished or hidden policy.

I begin by addressing the difference between topics/sub-topics on the one hand, and tick-box response options on the other. Mr Bell’s evidence about this was as follows:

It is for that reason, submits Mr Auburn, that the criteria for evaluating topics/sub-topics and those for tick-box response options are different notwithstanding that both sets of criteria do have an element of “acceptability”. The Claimant does not challenge Mr Bell’s evidence in this regard. Mr Wolfe’s submission is not that there is no distinction between the two sets of criteria as such, but that the ONS had clearly stated it was going to apply the former set of criteria to both topics and response options. Mr Wolfe did not advance any reason as to why the ONS, which plainly used separate sets of criteria for the 2011 Census, would for the 2021 census ignore the different contexts applicable to topics and response options and instead apply the same set of criteria to both.

Both Counsel engaged in a detailed analysis of the documents and correspondence, with each vigorously arguing that these are to be interpreted in support of their respective cases. It must be borne in mind that the various documents in question are, in the main, consultation and engagement documents, produced for the purpose of informing members of the public and interested stakeholders as to the processes involved in the design of the 2021 Census and seeking their views. The documents must therefore be read and examined with that purpose in mind. As stated by Dove J in R (Stephenson) v Secretary of State for Housing, Communities and Local Government [2019] EWHC 519 (Admin) at [44]:

That is the approach taken by this Court to the documents in the present case.

Many of the documents in question have already been considered above. To avoid unnecessary repetition, the conclusions set out below as to how the reasonable reader would view the documents are to be read with the analysis set out above.

May 2016 Consultation Response: This document does not support Mr Wolfe’s case, as set out in his Skeleton Argument, that from May 2016 onwards the ONS repeatedly stated that it would apply the PA Criterion to evaluate whether to include a Sikh tick-box response. In fact, this document was solely concerned with topics. Mr Wolfe conceded in oral argument that that was the case. That concession was rightly made. There is nothing in the May 2016 Consultation Response to suggest that the PA Criterion would be used for anything other than assessing topics.

May 2016 Topic Report: The same conclusion applies in respect of this document. The focus at this stage was very much on topics. In fact, far from supporting the Claimant’s case, it seems to me that this document directly contradicts it in that there is express reference (in the section, ‘Next Steps’) to using the methodology used in the 2011 Census for reviewing response options. Not only that, but the reader is directed by a hyperlink to the PT used in that exercise. There can be no doubt that any reasonable reader would understand that the approach that would be taken to the response options would be similar to that undertaken previously. In my judgment, these references are more than adequate to discharge the Defendant’s obligations to publish the applicable policy within the meaning of Lumba. The reasonable reader would be aware, at least in broad terms, of the approach that would be taken by the decision-maker and could direct their representations accordingly.

November 2016 Follow-Up Survey: There is a degree of ambiguity in this document as discussed above. However, when it is read with the March 2019

23 October 2017 – Open Meeting: The slides used at this meeting, which was clearly in two parts, do not support the Claimant’s case. The first part of the meeting was an update on the progress so far on topics, one of which was the ethnic group topic. The second part was specifically about the tick-box options. The Claimant’s case that the reference to the use of “published criteria” in the published summary of the Open Meeting can only have been a reference to the Topic Criteria ignores the express indications in previous documents (in particular, the May 2016 Topic Report) to the PT.

December 2017 Update Report: The references to “public acceptability” in this document are made in relation to topics, and in particular to question design. To the extent that mention is made of “additional response options”, there is express reference to evaluating these “using a tool to prioritise requests for additional response options”. There would be no need to mention such a “tool”, which plainly denotes the PT, if the Topic Criteria were to be used.

19 February 2018 – Mr Bell’s letter to Preet Kaur Gill MP regarding the Gurdwara survey:

In this correspondence, Mr Bell states that he has a different recollection to that of Ms Gill MP of a recent meeting. He goes on to say:

The Claimant contends that the reference here to “public acceptability” is a clear indication that the Topic Criteria would be used in assessing the Sikh tick-box. Mr Bell explains that his use of the term here was a reference to the “acceptability consideration in the tick-box criteria.”

The ONS has on several occasions (this being one of them) used the term

It is also noteworthy that the reasonable member of the public, being aware of the existence of two sets of criteria, would readily appreciate that the Topic Criteria are not apt for the purpose of assessing whether to include a particular tick-box. The Topic Criteria include such matters as the need to avoid asking sensitive or potentially intrusive questions that may have a negative impact on response, and the need to avoid enquiring about opinions or attitudes. It is difficult to see how these could be applied to an assessment of tick-box response options, where the question has necessarily already been formulated.

7 September 2018 – Letter from Mr John Pullinger, National Statistician to Ms Gill MP: In this letter, Mr Pullinger states as follows:

13 December 2018 – Further letter from Mr Pullinger to Ms Gill MP; This letter stated that new information received since the UKSA Board meeting in June 2018 had been “evaluated against the published criteria”. Insofar as this letter was written in the context of Ms Gill’s concerns about a Sikh tick-box, there is nothing inconsistent in Mr Pullinger referring to published criteria. If this is construed as referring to the criteria for tick-boxes, the ONS had, as I have found, discharged its obligations by previously publishing the relevant policy containing the PT.

14 December 2018 – December 2018 Update Report: There are two references to “public acceptability” in this document in the context of evaluating tickboxes. This is perhaps the high-water mark of the Claimant’s case (with Mr Wolfe describing it as the “killer document”). However, for reasons explained more fully above, it is apparent when one reads the document in light of what has gone before that the reasonable reader would not draw the conclusion that the ONS had suddenly decided not to use the PT (albeit updated) that it had used for the 2011 Census and had instead opted to use an inapt set of criteria designed for another purpose. Furthermore, there are references in the December 2018 Update Report to the PT being used in evaluating response options. It would be odd for the ONS to refer to the use of “…similar methodology to that used prior to the 2011 Census” if it was in fact trying to communicate a desire to use an entirely different evaluation method, namely the Topic Criteria.

14 December 2018 – White Paper: Whilst the Claimant’s skeleton argument includes the White Paper as one of the documents in which it is said the ONS stated its intent to use the Topic Criteria to evaluate tick-box response options, no specific passages of it were impugned, and Mr Wolfe did not pursue the point in submissions. Where the White Paper does refer to the tick-box issue (at 3.101 to 3.109), it relies specifically on the criteria under the PT.

20 March 2019 – March 2019 Survey Response: My attention was drawn to the summary of findings in which reference is made to the “public acceptability of terminology being used in [the ethnic group] question”, the suggestion being that this was yet another statement of intent to apply the Topic Criteria to the response options. However, on any fair reading of the passage, it would be apparent that this is about the terminology used in the question and is not about response options at all.

2 April 2019 – Meeting between Sikh Federation UK, APPG for British Sikhs and the ONS: This was a follow-up meeting to an earlier one held with Chloe Smith MP on 14 January 2019. The minutes of this meeting refer to two members of the Sikh Federation UK asking, “What quantitative data did ONS have which showed that there was not public acceptability for a Sikh Tick-box on the ethnicity question?” It is not clear why this document is relied upon against the Defendant as the reference to public acceptability was made by members of the Sikh Federation UK, not the ONS. The ONS did state (item 6

13 May 2019 - Meeting between Sikh Federation UK, APPG for British Sikhs and the ONS (Mr Pullinger): Once again, the minutes record a member of the Sikh Federation UK referring to “public acceptability” in relation to tick-boxes. The same points as for the previous meeting apply.

It is my judgment that the ONS did not at any stage indicate to the Claimant or to the public more widely that it intended to apply the Topic Criteria or the PA Criterion in particular to the evaluation of the tick-box response options. The Claimant’s case is based, as Mr Auburn submits, on a misconception that any reference to the term “public acceptability” regardless of context must be to that term as defined under the Topic Criteria. The ONS had, at an early stage, made clear its intent to use an updated version of the PT used in the 2011 Census. The Claimant would have been well aware of that PT as the Sikh Federation had engaged closely with the ONS during the 2011 Census when the same tick-box issue was considered, and indeed had produced a detailed critique of the ONS’s application of the PT. It is therefore somewhat odd that the Claimant should have laboured under the misapprehension, apparently for several years, that the ONS was seeking to abandon that PT for something designed for an entirely different purpose.

I accept that the terminology of “public acceptability” was used on occasion in ways that might be apt to confuse. However, an inadvertent lack of clarity in the communication of intended policy, or an occasional lapse in so doing, is nowhere near enough to make good a Lumba-style challenge that the ONS had failed to apply a stated policy, or, even more seriously, applied a hidden or secret policy inconsistent with the published one.

The ONS has, in my judgment, conducted its assessment in accordance with its own published criteria. That assessment may not be to the Claimant’s liking, and it may even be thought that the outcome is in some respects surprising, given the apparent (though not universal) strength of feeling in support of a Sikh tick box. However, the challenge was not that the outcome was surprising (which would fall well short of “irrationality” and would not give rise to any sustainable grounds for judicial review in any event) but that the Defendant applied unpublished policies. That challenge is not made out.

For these reasons, Ground 1 of the Claim fails.

This ground is relied upon in the alternative to ground 1. Mr Wolfe submits that even if the Court were to accept that the updated PT was indeed the policy to be applied in respect of tick-box options, the ONS unlawfully failed even to apply that policy and instead applied a different secret/unpublished policy. Mr Wolfe goes further and makes the bold submission that the policy published in June 2019, some 6 months after the White Paper, which contains a version of the PT, “appears … to have been drafted as a retrofit” to match what Kantar had actually done in their focus group research. The 2021 PT, submits Mr Wolfe, contains material differences to methodology that entirely contradict the ONS’s stated intent to apply the 2011 PT, updated only to take account of changes in legislation (in particular the Public Sector Equality Duty (“PSED”) under the Equality Act 2010). The changes to the 2011 PT, which include changes to weighting and to the scoring methodology - whereby a ‘stop/go’ RAG rating is applied to the “acceptability” criterion instead of a point score that contributes to a cumulative total under all criteria – are not related to the PSED.

Mr Auburn submits that the ONS did what it said it would do and has been transparent. It said that it would apply a similar methodology to that used for the 2011 Census, “reviewed and updated to reflect current legislation”. The changes that were made were minor. In relation to the “acceptability” criterion itself, the changes in wording – from “clear and acceptable to respondents…” to “acceptable to respondents, clear…” – merely clarify and do not affect the substance. Insofar as other parts of the PT were amended, the changes were minor and arose out of the review which the ONS had been open about all along. The changes to weighting meant that more emphasis was placed on “user need”, which would only benefit the claim for a Sikh tick-box. The RAG rating was not significantly different from the numerical scoring system. In any event, submits Mr Auburn, there is clear and uncontested evidence from Mr Bell that these changes are minor with nothing from the Claimant to gainsay that.

As stated by Lord Dyson in Lumba, “What must … be published [in respect of a policy] is that which a person who is affected by the operation of the policy needs to know in order to make informed and meaningful representations to the decision-maker before a decision is made.” The policies being considered in Lumba related to the continued detention of foreign criminals who had served their sentences, and affected the liberty of the subject. Even in that context, it was not necessary to publish drafts when policy is evolving or to publish details which are irrelevant to the substance of decisions made pursuant to the policy.

Although the Claimant does not seek to bring his challenge on the basis that there was unfair consultation, the ONS’s documents were produced in the context of consultation and it is instructive to consider the words of Lord Woolf MR in R v North and East Devon Health Authority Ex p Coughlan [2001] QB 213 at [108]:

Whilst the ONS’s publications relating to the census will not affect liberty or directly affect the day-to-day life of an individual member of the public, they must at the very least enable any interested party to make informed and meaningful representations before a decision is made. In that context, publication of the criteria to be used in selecting between various options will be key. However, the weight to be attached to individual criterion under those criteria will not always be so; small shifts in emphasis on different criteria are unlikely to affect the representations that would be made in respect of each criterion. Such shifts could arise for any number of rational reasons which the decision-making body might consider important. The position might be otherwise where a weighting system is so radically amended as to require a change in the nature or scope of the representations made. Thus, if a criterion that was previously highly weighted were suddenly given no weight at all, or its weighting were diminished to such an extent as to make its influence on the outcome negligible, the decision-maker could normally be expected to publish that change before a decision is made in order that those with an interest could focus their representations on other criteria. Similarly, minor adjustments to a scoring system, such as using 5, 10 and 15 points instead of 1, 2 and 3 to reflect a “low”, “medium” and “high” assessment, would be less likely to affect the representations made than an adjustment that was more substantive.

In the present case, the ONS did not give any assurance that the PT for the 2021 census would be identical to that used in 2011; the ONS, as far back as May 2016, made clear its intent to use similar methodology to that used in 2011, reviewed and updated. In doing so, the ONS was not fettering its ability to make changes that did not arise directly out of an application of the PSED; the reference to reviewing and updating makes that apparent.

The 2011 PT underwent an expert review between the Autumn of 2017 and updated versions of the PT were shared with various governmental and non-governmental bodies and panels, including two Universities. That review process was hardly secret, albeit that the final updated PT was not published until June 2019. The publication date of June 2019 was a year before the legislation being challenged was laid in Parliament, and matched the timescale of the previous 2011 PT, which was similarly published two years prior to the 2011 Census.

Furthermore, there was unchallenged evidence from the ONS that the changes to the 2011 PT were minor. Of course, that may be viewed as self-serving opinion, but it is significant that there was no request to challenge or undermine that evidence by an application to cross-examine Mr Bell. In these circumstances, considerable weight may be attached to the evidence of Mr Bell, who is well-versed in these matters. However, even in the absence of such evidence, I would have reached the same conclusion, namely that the changes to the 2011 PT were minor and were not such that any interested party would need to know so as to be in a position to adjust any representations made. In fact, the key criterion in question (indeed the only criterion on which any attention was focused in submissions) is that of “Acceptability, Quality and Clarity” (the ACQ Criterion). The acceptability element of the ACQ Criterion in the 2021 PT remained very similar in wording, and substantially the same in meaning, to that which appeared in the 2011 PT. As for the other changes:

The criteria were organised slightly differently in the updated version. However, this aided clarity and did not alter the substance of the matters to be assessed.

The weight to be attached to the different criteria was altered but not to the extent that the value of any single criterion was radically inflated or diminished. The criteria that had an increased weighting in the 2021 PT were ones relating to user need. That would have favoured the case for a Sikh tick-box as compared to the 2011 PT.

The scoring system remained the same as before for most of the criteria with a score of 2, 1 or 0 being allocated according to whether the strength of the supporting evidence was high, medium, or low. In 2021, the ACQ Criterion was not scored numerically or weighted because it was the final criterion to be applied. As stated in the June 2019 PT:

The Claimant submits that the RAG rating was decisive such that even a high score on the other criteria might not suffice if the tick-box option was otherwise considered to be unacceptable or such that it does not provide the required information to an acceptable level of quality. However, the final sentence of the passage, which states that consideration is given to the assessed scores from the other criteria, leads to a different conclusion. The ACQ Criterion is clearly important but is not dispositive as the other scores will be taken into account. That would suggest that even if a tick-box option had been rated Green or Amber, that would not necessarily override a low score for say “user need” or “data quality”.

Taking all of these matters into account, it is my judgment that the changes to the PT were, as the Defendant submits, not material ones in the sense that the Claimant would not have been materially disadvantaged in knowing what representations to make.

The upshot is that there was no Lumba-style failure to apply a published policy or the application of an unpublished policy. The ONS had at all times stated its intent to apply an updated version of the 2011 PT, and that is what it did.

The Claimant submits that the PA Criterion in the Topic Criteria was, in any event, not applied consistently in relation to questions and response options considered for inclusion in the 2021 Census. Whereas “public acceptability” was considered for both questions and response options for the sexual orientation topic, the same approach was not taken for the ethnic group topic. It is also said that ,whereas for other topics, the likelihood of a question having a negative impact on responses was not treated as fatal to satisfying the PA Criterion, the existence of a few negative responses within focus groups was treated as fatal to satisfying the same criterion in respect of the Sikh ethnic group question. For example, although about 5% of respondents in England and 10% in Wales said that they would skip a gender identity question, that was not treated as a failure to satisfy the PA Criterion. Similarly, some negative responses to the sexual orientation question did not lead to the conclusion that such a question was unacceptable whereas the existence of some negative views amongst small focus groups was treated as being decisive against the Sikh tick-box response option. That was notwithstanding evidence available to the ONS of the high level of support for that option from within the Sikh community.

The Defendant submits that this ground suffers from the same flaw as the first, which is that it is based on the same failure to recognise the distinction between topics/questions and response options. It further submits that no comparison can be made between decisions on the inclusion of, and question design in relation to, the sexual orientation topic on the one hand, and the addition of a Sikh tick-box response option under the ethnic group topic on the other. The former is concerned with the inclusion of an entirely new topic in respect of which no census data currently exists, whereas the latter involves data that can (and has been) collected in other ways (e.g. under the religion question and/or via the write-in option). In suggesting that there has been inconsistency, the Claimant is not, it is submitted, comparing like with like in that there is an alternative source of information in relation Sikhs (via the religion question) whereas there is no such source for new topics such as sexual orientation.

The Defendant’s submissions are to be preferred. The Claimant is not comparing like with like.

The Claimant’s challenge in relation to the ONS’s approach to sexual orientation arises out of the following passage in the December 2017 Update Report:

70% of respondents in England and Wales found it acceptable to include a question on sexual orientation on the 2021 Census

only 1% would stop completing the census form altogether if sexual orientation question was included in the 2021 Census as a voluntary question

less than 1% of the public in England and Wales would provide an inaccurate answer or request an individual form

The Claimant’s submission is that the analysis of a response option, namely the “prefer not to say” option, under the sexual orientation topic, is a clear indication that, contrary to the Defendant’s case, the Topic Criteria were being applied to response options as well. I do not accept that submission. It is clear on any reasonable reading of the passage cited that the context of this analysis is whether or not a question on sexual orientation should be included at all and, if so, what it should say. That is an issue of topic/question design. One of the criteria used to determine that issue is the PA Criterion which includes consideration of whether the census is asking sensitive or potentially intrusive questions that have a negative impact on response. It is apparent from the ONS’s research that some respondents did consider the question on sexual orientation unacceptable, in some cases leading to the respondent ceasing to complete the form altogether. However, the research also revealed that the inclusion of a “prefer not to say response” significantly increased the acceptability of the question. It seems to me that this sort of research, which involves testing different question designs including one providing an option effectively not to answer the question at all, is all about topic inclusion and question design. It is, in my judgment, quite different from a prioritisation exercise whereby positive response options are assessed and ranked. It must also be remembered that the Topic Criteria were used to assess public acceptability across all respondents, whereas the focus of the PT was on acceptability (as defined) amongst those respondents to whom a particular response option might be relevant. That distinction is reflected in the exercise described here in which public acceptability of the sexual orientation question was assessed in respect of all respondents, not just those in respect of whom a particular sexual orientation may be relevant.

Viewed in that way, there is nothing inconsistent about the ONS’s approach to the negative views expressed by some Sikh respondents in the focus groups. That exercise was assessing acceptability across a particular subset of respondents in respect of whom a particular tick-box response option might be relevant. The Claimant’s submissions are also flawed in so far as they seek to adopt a quantitative approach to research that is clearly qualitative. The Kantar research, designed to understand how respondents within a particular group understand and react to the inclusion or exclusion of particular response options, cannot be directly compared with a quantitative study.

For these reasons, Ground 3 fails and is dismissed.

Under this ground, the Claimant submits that the conclusions in the Kantar Report are “legally unsustainable” on account of material internal inconsistencies, inaccuracies, and a failure to apply Kantar’s own stated definitions and methodology. The Claimant makes three specific criticisms:

First, it is said that Kantar assigned ratings to its “Clarity” sub-criterion, which assessed the extent to which respondents identified with the Sikh tick-box option, inconsistently with the findings of its own focus groups. Given that the RAG rating system applied was defined on a “less than 2011”, “no change from 2011” and “more than 2011” basis, if any respondent identified with the Sikh tick-box then the rating should have been “Green” because there had previously been no Sikh tick-box.

Second, it is said that Kantar failed even to properly apply its own published “Quality” sub-criterion. This sub-criterion involved an assessment of whether the addition of the Sikh tick-box resulted in more or fewer respondents being unsure/uncertain/confused about which box to tick. The Claimant submits that instead of focusing on confusion about which box to tick, Kantar impermissibly took account of confusion about other matters.

Third, it is submitted that Kantar’s conclusion that a Sikh tick-box was not acceptable was based upon the legally erroneous view that Sikh is not an ethnic group. Reliance is placed on the decision of the House of Lords in Mandla v Dowell Lee [1983] 2 AC 548, in which it was determined that Sikh is an ethnic group for the purposes of the then Race Relations Act 1976. (“the 1976 Act”)

The Defendant makes the general point that none of these issues raised grounds of illegality and that at most they are fine-grained merits points about the work of a research company, whose input was only one part of the material considered by the ONS. The Defendant submits that the Claimant’s first and second challenges under this ground are based on a misunderstanding of the criteria applied and are misconceived. The third point is equally misconceived as census forms are completed on the basis of self-identification, not legal definitions.

In R (Law Society) v Lord Chancellor [2019] 1 WLR 1649, the issue was whether a decision to reduce the amount payable to criminal defence solicitors under the

In dealing with irrationality in relation to the process by which a decision is reached, the Divisional Court said as follows at [98]:

It is clear from these passages that the threshold of unreasonableness or irrationality in relation to the process by which a decision is reached is a high one. In order to undermine a report such as the Kantar Report, upon which a public body places reliance, it is not enough merely to identify an error. The Claimant must establish that the errors amount to serious logical or methodological errors, that they go to the heart of or are central to the conclusions reached, and are such as to make a substantive difference to the outcome.

The Claimant has not adduced any expert evidence in this case to challenge the Kantar Report. Mr Wolfe submits that such evidence is not necessary because the errors in this case, arising as they do in the Kantar Report, are of a kind which would be obvious to an untutored layperson (which would for these purposes include a judge). I do not agree. In my judgment there are no significant errors at all, and even if there were, these are far from being so serious as to undermine the material conclusions expressed. I deal with each of the three categories of error relied upon by the Claimant in turn.

The first alleged error is that the rating for the Clarity sub-criterion should have been

That said, there is a difficulty with the way in which the RAG rating system is defined insofar as it applies to the Clarity sub-criterion: whereas it is possible to make an assessment (in respect of the Acceptability and Quality sub-criteria) as to whether the inclusion of a tick-box is more / less / similarly acceptable to respondents than the position in 2011 (where there was no tick-box), it is not really possible to make a similar relative assessment for “Clarity”. That is because that criterion is premised on there being a tick-box, which is identified with and used: “If the tick-box is available…”. It would not be possible for any respondent to identify with or use a tick-box that is not there. If a strictly relative comparison with the 2011 position is undertaken then it would be correct to say, as the Claimant submits, that any identification with and use of the tick-box should result in a “more than” outcome and a “Green” rating. However, to take that approach would be absurd and would probably have resulted in a “Green” rating for all four tick-boxes being considered (Jewish, Sikh, Somali and Roma) as none of those were previously available.

When one examines the actual rationales given for each rating in respect of the Clarity sub-criterion, it becomes apparent that the researchers did not take that approach. Instead, as the following sample of rationales demonstrate, they consistently considered whether there was, in general, little or no identification with the tick-box, which would attract a “Red” rating; some identification with it, which would be rated “Amber”; or strong identification with it, in which case it would be “Green”:

Jewish tick-box - Red: “Participants did not identify with Jewish as an ethnicity…”

Sikh tick-box – Amber: “While many recognised that Sikh was not a marker of ethnicity, some were likely to identify with the tick-box – particularly older male participants”

Somali tick-box – Green : “Somali participants identified with this tick-box and were clear this was for them”

Roma tick-box – Green: “Participants identified with the Roma tick-box and would tick as a result.”

Kantar’s qualitative approach, whilst not strictly following the relative assessment suggested by its own RAG rating system for this sub-criterion, does appear to have been applied consistently and has produced a rational assessment for that sub-criterion. On that basis, the amber rating for Clarity, based as it is on the presence of some identification with and use of the Sikh tick-box, is apt.

Kantar’s brief was to “Assess acceptability, clarity and quality of tick-boxes”. There was no statutory or other requirement as to how those matters were to be assessed. Kantar’s approach to the assessment of the Clarity sub-criterion was, notwithstanding its apparent failure to follow the letter of its own rating system, rational and consistently applied. The reasoning did not involve any serious logical or methodological error that could be said to undermine the purpose or the results of the analysis.

The second alleged error is that Kantar wrongly took into account confusion in relation to matters other than over what box to tick. I consider this argument to be misconceived.

In the first place, the Claimant seeks to test the criteria applied by Kantar against the wrong published criteria. The Claimant’s skeleton argument, after setting out Kantar’s definition of the “Quality” sub-criterion, states as follows:

The underlined wording reflects the PA Criterion under the Topic Criteria and not the ACQ Criterion in the 2021 PT. The latter merely refers to whether the addition of the tick-box “is acceptable to respondents, clear…, and provides the required information to an acceptable level of quality.” The Claimant’s challenge is therefore based on an incorrect premise.

In any event, the Claimant’s challenge is unfounded because it appears to be based on the notion that Kantar should be constrained to apply an unduly narrow interpretation of its own criteria in assessing “Quality”. No issue appears to be taken with Kantar’s decision to assess the level of confusion that might result from having a Sikh tick-box option. That is not surprising since, on any view, a confusing response option would be likely to affect the quality of the information obtained. If it is accepted that the likelihood of confusion is an appropriate matter to assess, then the researchers cannot be criticised for going about that assessment in any rational manner that they, in their

The Claimant’s third and final challenge is that Kantar’s conclusion that a Sikh tickbox was not acceptable was based upon the legally erroneous view that Sikh is not an ethnic group. He submits that the House of Lords having determined in Mandla v

This argument has no merit. It is based on a misunderstanding of the way in which the

For these reasons, Ground 4 also fails and is dismissed.

For the reasons set out above, none of the Claimant’s grounds of challenge succeeds and the application for judicial review is dismissed.

There were extensive submissions from both sides as to the approach the Court should take in exercising its discretion as to remedy should any of the grounds have been made out. As none of them has been made out, it is not necessary to deal with those submissions. Nevertheless, I deal with some of them very briefly in view of the importance to the parties of some of the issues raised.

The Defendant submitted that if any of the grounds had been upheld, the Court must apply s.31(2A) of the 1981 Act and refuse relief. Section 31(2A) provides:

Mr Auburn submitted that it was highly likely that the outcome for the Claimant would not have been substantially different, and that the Court should not shirk its responsibility in assessing whether that is the case: R (Gathercole) v Suffolk County Council [2020] EWCA Civ 1179 at [38]. He relied upon evidence from Mr Bell to the effect that the application of the Topic Criteria, and the PA Criterion in particular, to the tick-box options would have been highly likely to yield the same result, namely that the tick-box should not be included.

Mr Wolfe reminded the Court that it should be cautious of straying into the merits of a decision under challenge, because, as stated by the Court of appeal in R (Plan B Earth) v Secretary of State for Transport [2020] EWCA Civ 214 at [273]:

Mr Wolfe also submitted that there is a danger in relying upon post-decision speculation which may not be based on material in existence at the time of the original decision: see R (Logan) v Havering London Borough Council [2015] EWHC 3193 Admin at [55] per Blake J; and that views expressed in the course of litigation as to how a decisionmaker might have acted are “rarely, if ever, of assistance in considering whether a decision might have been different if the decision-maker had considered the issues properly”: see Pemberton International Limited v London Borough of Lambeth [2014] EWHC 1998 (Admin) at [72], per Lewis J.

In my judgment, the Court can, with due caution being applied, take account of evidence as to how the decision-making process would have been approached if the identified errors had not occurred. Section 31(2A) is not prescriptive as to material which the Court may consider in determining the “highly likely” issue: see R (Enfield LBC) v Secretary of State for Transport [2015] EWHC 3758 at [106], per Laing J. Furthermore, a witness statement could be a very important aspect of such evidence: R (Harvey) v Mendip District Council [2017] EWCA Civ 1784 at [47], per Sales LJ.

I have, accordingly, and with due caution considered Mr Bell’s evidence in this regard. However, having done so I am not satisfied that it would have been highly likely that the outcome for the applicant would not have been substantially different if the conduct complained of had not occurred. The issue here was, according to the Claimant, the application of entirely the wrong assessment criteria. The two sets of criteria in question were, as Mr Bell states in his first witness statement, intended to assess different things in relation to different target groups. Indeed, the Topic Criteria are in some respects wholly unsuited to the task of prioritisation between different tick-box options, and it

Mr Auburn submits that even if there had been any errors of law, quashing of the Census Order should be refused because of the colossal detrimental effect that such relief would have on good administration. There is unchallenged evidence before me that if the Census Order is quashed the March 2021 census would have to be postponed to March 2022 at a cost of about £250m. That sum arises from the need, amongst other things, to scrap 3 million paper questionnaires, to continue paying 300 workers that have already been recruited, and to meet the costs of the many commercial contracts which have already been entered into to facilitate the 2021 Census. Mr Auburn also points out that delaying the Census by a year would have great implications for the data gathered, for example:

it would break the 10-year time cycle which has stood since the early 19^th century;

it would affect the quality of the data and any long-term analysis based on the statistics gleaned; iii) it would damage the “rebasing” of mid-year estimates;

it would damage the integrity of the statistics and the confidence of others in decisions made in reliance upon those statistics; and

relevant decisions over the next year after March 2021 would be based on increasingly out of date data.

Mr Auburn submits that it is especially important that Government and other decisionmakers have accurate and timely data during a time of such economic and social upheaval, arising from both the pandemic and the consequences of leaving the EU.

Mr Wolfe submits that the difficulties caused by postponement would appear to be exaggerated. He points out that the Scottish 2021 Census has already been postponed until 2022 and the 2021 Census itself is on a “knife edge”. He reminds me that in most cases the relief upon a finding of illegality is the quashing of the relevant decision, in this case the Census Order, and that “… the discretion of the court to do other than quash the relevant order ‘where such excessive exercise of power is shown’ is very narrow.”: see Edwards v Environment Agency [2008] UKHL 22 at [63], per Lord Hoffmann. Mr Wolfe submits that it is wholly inappropriate for the Defendant to argue

In my judgment, had the court been called upon to consider, in its discretion, whether to grant or refuse relief, it is highly likely that the desired relief of a quashing order would have been refused. That is because of the very serious detriment to good administration that would result. The consequences of a quashing order, as identified by Mr Auburn, are of an order of seriousness, both in terms of the detriment to good administration and the additional cost, not often seen in this Court. The fact that the Scottish census has been delayed does not mean that the consequences in England and Wales of a delay cannot be as serious as stated. The decision in Scotland, in respect of a far smaller population, was taken on the basis of its own planning and preparedness, and there is nothing before me to gainsay the powerful evidence adduced by the Defendant as to the consequences of a postponement in England & Wales. That huge detriment, which is almost unparalleled, very substantially outweighs, in my judgment, the potential prejudice to the Claimant of not having a Sikh tick-box in the 2021 Census. In coming to that view, I do not underestimate the importance to the Claimant of having a specific Sikh tick-box under the ethnic group question. However, as Mr Auburn points out, the Census, as currently designed, will not prevent any respondent who wishes to do so from identifying as ethnically Sikh. The write-in option, with the auto-fill function in the online version, will enable the respondent to do so.

There are many allegations, from both sides, of delay in relation to the various proceedings. It is not necessary to make findings on these allegations, as the application before me, which targets the Census Order, was brought in time. I accept that the judgment of Lang J in JR1 (at [108]) does suggest that the Defendant could have adduced evidence as to relief in previous proceedings if it was concerned about the uncertainty and delay that would result if a legal challenge was brought after the Order in Council. However, it was not obliged to do so and was entitled to resist the claim on the grounds that it did. I do not consider its failure to adduce such evidence previously to be inappropriate or such as to amount to a lack of candour, particularly as the issues relating to relief may not have seemed nearly as acute in 2019 as they appear now with just five months left before the 2021 Census. I also do not accept Mr Wolfe’s suggestion that the Defendant ought effectively to have put the brakes on the 2021 Census once it knew of the Claimant’s application. It cannot be right that any challenge to executive decision-making should require the decision maker to cease ongoing work, particularly in respect of a project of the scale of a national census. The Claimant had the option of applying for interim relief if that was considered desirable. It does not appear that there was any such application.

As well as alleging a lack of candour in relation to the relief sought (which I have considered above), the Claimant asserts that there was a lack of candour on the Defendant’s part in the way that it has sought to defend these proceedings and the disclosure that it has provided (or failed to provide). Mr Wolfe relied upon the decision in Citizen’s UK v Secretary of State for the Home Department [2019] 1All ER 416 in which Singh LJ referred, at [106], to the principle that:

Mr Wolfe submits that the documents in this case “have trickled out”, that the Defendant has given “ever-changing explanations” for its conduct, and that it kept “changing its story”. The difficulty with these submissions is they are mostly premised on the Claimant being correct in his principal contention that the ONS stated that it would apply the Topic Criteria, all the while applying some other criteria. In fact, as I have held above, that contention is not correct. At most it can be said that the Defendant was not as clear on occasion as it might have been as to the different sets of criteria being applied. However, a lack of clarity in communications does not amount to a lack of candour any more than it amounts to an error of law.

There is a specific allegation in respect of the report produced in June 2020, namely the Ethnic Group Prioritisation Report (“the June 2020 Report”). The Claimant submits that the June 2020 report discloses yet more errors, which, had they been disclosed earlier, would have enabled the Claimant to rely upon additional grounds of challenge. I do not accept that submission. As Mr Auburn points out, the June 2020 report was only completed in June 2020 and could not have been disclosed any earlier. Furthermore, despite knowing about the June 2020 report for several months, the Claimant has not sought to amend its claim to raise any further allegations arising from its contents. This complaint, therefore, goes nowhere.

I conclude by expressing my thanks to all Counsel for their helpful and focused submissions and skeleton arguments.