This is an HTML version of an attachment to the Official Information request 'Spirituality and religion in the Census'.









From: [redacted - section 9(2)(a)] 
Sent: 15/06/2017 8:36:05 a.m. 
To: [redacted - section 9(2)(a)] [redacted - section 9(2)(a)] [redacted - section 9(2)(a)] 
Subject: RE: Religious affiliation - 2017 Census test results and steps forward 
I’ve got no concerns after the discussions yesterday. Challenge will be to get some 
meaningful examples to replace what is currently on the paper form.  And  the write-in sits 
above the no religion and object tick boxes. Multiple response would be helpful but I don’t 
think there are significant impacts on the overall data with the constraints of the AYT and 
ICS versus the paper. Key thing is that we should get much better quality data as respondents 
can give more detail on what used to be the tick boxes which was a key criticism that came 
through in the recent classification refresh. 
 
[redacted - section 9(2)(a)] 
[redacted - section 9(2)(a)] 
[redacted - section 9(2)(a)]
 
 
  
  
 
 
From: [redacted - section 9(2)(a)] 
Sent: Thursday, 15 June 2017 7:31 AM 
To: [redacted - section 9(2)(a)] [redacted - section 9(2)(a)] [redacted - section 9(2)(a)] 
Subject: RE: Religious affiliation - 2017 Census test results and steps forward 
 
Thank you [redacted - section 9(2)(a)].  Not having “as-you-type” options for responses 
beyond the first is a pity but understand the problem very well – and the incidence of multiple 
response is low enough for this not to be a problem (people committed to adding more than 
one response will do so anyway). [redacted - section 9(2)(a)] will confirm if this is likely to 
create any problems but as far as I can see it is the best solution in the circumstances.  The 
“no religion” and the required “object” options are important to retain (the latter is legally 
required I believe). 
From: [redacted - section 9(2)(a)] 
Sent: Wednesday, 14 June 2017 3:38 PM 

To: [redacted - section 9(2)(a)] [redacted - section 9(2)(a)] [redacted - section 9(2)(a)] 
Subject: RE: Religious affiliation - 2017 Census test results and steps forward 
 
Hi [redacted - section 9(2)(a)] 
 
Thanks for your comments. We had a discussion about this topic at a form development 
meeting this morning where [redacted - section 9(2)(a)]was present. 
 
We discussed the option of a paper format closer to the 2013 version – QMD has drafted up a 
version of this but it is ‘messy’ with very cramped routing. 
We also discussed the option of a question with the write in box with the necessary additions 
of ‘no religion’ and ‘object to answer’ response options. 
 
It was decided that the best approach would be to proceed with the write-in as primary 
response option, as ultimately this is the information that we want to collect from 
respondents. Given that the 2013 format already has issues as you have identified, recreating 
a less respondent friendly design of this seems unideal. 
 
We also had to consider some of the constraints of building the online version, with regards 
to multiple response for this question. All write-in boxes on the ICS with as-you-type 
functionality currently have the option for the respondent to write more text after they have 
chosen a response from the drop down suggestions. It was indicated that to build the 
functionality to ‘add another religion’ as is available for the iwi question would be very 
costly to build. The current default functionality as described would allow multiple response 
on the online form, if the respondent wished to do so however as-you-type functionality 
would not be available after they had selected their first response.  
 
If you are comfortable with this approach, I will then discuss this with Gareth Meech and 
Denise McGregor before QMD and Respondent Interaction begin developing the revised 
question format. 
 
Cheers, 
[redacted - section 9(2)(a)] 

From: [redacted - section 9(2)(a)] 
Sent: Wednesday, 14 June 2017 10:44 AM 
To: [redacted - section 9(2)(a)] [redacted - section 9(2)(a)] [redacted - section 9(2)(a)] 
Subject: RE: Religious affiliation - 2017 Census test results and steps forward 
 
Hi [redacted - section 9(2)(a)] 
 
This is a seriously difficult question, thanks to the paper-size – and it is a pity the pair of 
questions failed.  Census is an on-line first survey, so the paper form is secondary, but if we 
get a large number of paper forms then we need to understand how the two modes perform 
differently.  The key points have been made in the email trail below – the purpose in 
collecting this information is not satisfied by the questions in the test form.  I would also 
question how you know the data in the first question was of high quality: it is notoriously 
difficult to establish data quality in a box-ticking exercise and I would suggest that it is not 
good quality because the routing clearly did not work (your call backs will almost always 
evoke an “expected” response because people like to appear compliant face-to-face [voice-to-
voice].   
 
The problem is that the level of information on religious affiliation elicited by Q16 is not 
what is required since at least two of the tickboxes (Christianity and Other) aggregate many 
dozens of very distinct religions.  I do agree that the prompts in Q17 should have provided 
enough hint of what we need but the damage is already done by the tickbox question 
above.  Perhaps the wording “more detail” is not helping because I can imagine there would 
be a tendency to wonder ”what more detail? Do you want an essay?”. Having ticked 
“Christianity” I would imagine that many Presbyterians (as an example) would simply react 
that that is not more detail but a different question altogether. It is also more likely to produce 
an address rather than affiliation. 
 
Definitely comparability is an extremely important consideration but not the only one.  We 
are living in a rapidly diversifying society and census is one of the extremely few 
opportunities we have to be able to get some measure of this – so we not only need 
comparable data but we need contemporarily useful information that we can derive from that 
data.  Unfortunately q16 does not come close to doing this on its own, and the paired 
questions have failed. 
 
The online would be no problem – maybe just use the 2013 question, corrected of course to 
use nouns as the Q16 below has.  The problem is the paper form and I am wondering whether 
one workable solution might be to reformat the 2013 question so that the call-out boxes 
become a subset below the headings – but this would lengthen the question significantly to 
the detriment of neighbouring questions (especially the 1yr ago question that we discussed 


the other day and which is crucially important to get right with a write-in for country).  My 
main question then is: does the length required to do this exceed the total length used by Q16 
and Q17 together? 
Given the space available, the potential for drop down categories etc on-line, the best solution 
would perhaps be a write-in as Andrew suggested.  There will still be the old problems on a 
paper form of people who give their religion as “St Stephens” and we can only guess which 
temple they mean from their geographic locality (especially where the same names are used 
in multiple religions or the same building doubles service as Anglican, Catholic, Muslim 
etc).  While it will be flying blind in terms of testing I am sure that QMD have experience 
with this type of question and know the most effective format for write-in.  One thing to 
watch for though would be allowing enough characters to adequately capture a descriptor 
since we would need to collect up to, ideally, 4 religions (given the current low rate of 
multiple response I wonder whether we could drop this to 3? Thoughts, [redacted - section 
9(2)(a)]?). 
[redacted - section 9(2)(a)] 
From: [redacted - section 9(2)(a)] 
Sent: Wednesday, 14 June 2017 10:00 AM 
To: [redacted - section 9(2)(a)] [redacted - section 9(2)(a)] [redacted - section 9(2)(a)] 
Subject: Religious affiliation - 2017 Census test results and steps forward 
 
Hi all, 
 
The 2017 Census Test data has been finalised and the data for the religious affiliation quested 
indicates that the question performed poorly in being able to produce comparable data to 
previous collections of this concept in the census. 
 
 


 
 
While the data quality collected from question 16 – the first component was of high quality, 
the response rate to question 17 for those who stated a religion in question 16 is very low at 
33.9%. This is consistent across denominations, for example Anglican responses making up 
4.1% of the total respondent population compared to a prevalence of 11.8% in 2013. 
 
We also called back some respondents (6) who did not respond to question 17 to get more 
information on why they did not respond. The overall feedback that they had already 
responded and were happy with the response they had given, and did not feel the need to 
respond to the further optional question. 
 
If we proceeded with this question format, we would be unable to produce comparable data 
for any category other than those collected in question 16. 
 
There are obviously space constraints in revising the question to a preferable format.  
 
I have had a think about this issue might be resolved and the factors that need to be 
considered. I have copied in our previous discussion on this question design change prior to 
the 2017 test. 
 
-  The ‘soft’ routing to question 17 is not working successfully – a revised question 
needs to have direct routing to the write in box 
-  If we were to have a combination of tick boxes and write in box, given the limited 
space this would likely need to have a mix of concepts collected in the tick boxes 
 
-  The other option is to include a write-in only as [redacted - section 9(2)(a)] had 
previously mentioned. 

It would be good to have a discussion about how we might best proceed with this question in 
order to output quality information on the concept. One major constraint is that we don’t have 
another major test in order to test any revised format. There will be usability testing 
undertaken by QMD however. Happy to hear your thoughts. 
 
Cheers, 
[redacted - section 9(2)(a)] 
Thanks for the comments. 
[redacted - section 9(2)(a)] was not too keen on the Christian tickboxes given the 
mixing of levels being collected. We did also discuss a pure write in question which 
I agreed would work best in theory – especially if we are asking all religious 
respondents to write in detail anyway. 
I will forward on the suggestions of the plural wording and the size of the write in 
text box – currently the column in which the religion question sits on our marked 
up proposed form is very vertically full but until the sign off of topics for the March 
2017 Census Test is done, we cannot be certain of the full space constraints! 
Cheers, 
[redacted - section 9(2)(a)] 
From: [redacted - section 9(2)(a)] 
Sent: Wednesday, 28 September 2016 8:31 AM 
To: [redacted - section 9(2)(a)] 
Cc: [redacted - section 9(2)(a)] 
Subject: RE: Religion question - Census 3 column IF format 
Hi [redacted - section 9(2)(a)] 
Is looking much better. Personally I’d be happy with just a write-in box alone – it 
isn’t that hard a variable to code, we’ve reduced the codefile by half and 
improved the match rate and quality (using the proposed new version that we have 
created as part of the current refresh of the classification.) But I would go with 
this new design if at all possible. 
I would note that the key selected stakeholders that we have consulted  during the 
current refresh have all made some comment about the inability to collect 
multiple responses, and to collect sub-groupings (for want of another word) of 
many groupings such as Judaism, Islam, Buddhism as we lose the information by 
encouraging the tick-boxes. Religion is not about collecting the information to 
create a ranking system of the most popular, it is a measure of our cultural 
diversity and social-connectedness which is where a lot of the research is wanting 
the better detail. I can only do so much to improve the classification (which I have 

at this point) but it sort of becomes pointless if the question doesn’t allow for 
some better coverage. 
Whatever you do, there will be criticisms – the agnostic/atheist groups will want a 
tick-box, the jedis as well, and then there is the whole dump coding to Christianity 
which [redacted - section 9(2)(a)] talks about (and which I agree with). Having 
Christianity loses the information we want about the emerging and non-traditional 
groupings. Nobody really cares about how many Catholics and Anglicans there are 
(other than those two groups). So that’s my two cents as Senior Researcher for 
S&D and as the person doing the review of Religious Affiliation at the moment.. 
From a C&S perspective, the proposed change is still in keeping with the spirit and 
intent of the Statistical Standard so I can’t see that there are any issues for making 
a change, 
Cheers 
[redacted - section 9(2)(a)] 
[redacted - section 9(2)(a)]| Senior Researcher 
Statistics New Zealand - Tatauranga Aotearoa | BNZ Centre, 120 Hereford St,  Private Bag 4741, 
CHRISTCHURCH 8011 
Email: [redacted - section 9(2)(a)]| DDI: [redacted - section 9(2)(a)]| www.stats.govt.nz 
From: [redacted - section 9(2)(a)] 
Sent: Wednesday, 28 September 2016 7:35 AM 
To: [redacted - section 9(2)(a)] [redacted - section 9(2)(a)] 
Subject: RE: Religion question - Census 3 column IF format 
Thanks [redacted - section 9(2)(a)]. I think this is a much improved version and is in 
fact a better question format than 2013 – the primary thing I would suggest here is 
to reinforce that this is a multiple response question so we must provide the ability 
for people to provide more than 1 (we do collect up to 4), so change “religion” to 
“religion(s)” and add a couple of rows. 
I like your options since you have been careful to include a variety of Islam and 
two other religions.  
However, what really worries me with this format is that people will be likely to 
take the easy option of not giving text responses and if they do there is both a 
processing overhead and loss of key information.  This is why I would have much 
preferred to see some tickboxes other than (preferably instead of) “Christianity” 
since people are not interested in the high level grouping category but in counts 
and characteristics of people who are Catholic, Anglican, etc etc. This would not 
actually save space, because we would still have to have write in options for other 
religions, but it would simplify processing. 
Please keep me in the loop with this one. 


Cheers 
[redacted - section 9(2)(a)] 
From: [redacted - section 9(2)(a)] 
Sent: Tuesday, 27 September 2016 3:57 PM 
To: [redacted - section 9(2)(a)] [redacted - section 9(2)(a)] 
Subject: Religion question - Census 3 column IF format 
Hi [redacted - section 9(2)(a)]and [redacted - section 9(2)(a)] 
I am aware that [redacted - section 9(2)(a)]has already contacted you about some of the form 
changes we are considering for the 2018 Census individual paper form design as a result of 
the decision to use a folded A3 paper size. 
[redacted - section 9(2)(a)] and myself have been scoping out some of these changes in order 
to fit the recommended content into a form this size. In general this has been quite successful 
– but an issue with the religious affiliation question has arisen due to reliance on horizontal 
spacing for the 2013 format of this question. 
If a three column format is used for the 2018 Census individual paper form, the format for 
this question must be altered. 
While vertical space is not a huge constraint for design of this question in the first instance, 
reactions from the Census SLT and Carol Slappendel were negative towards a second 
separated question on Christian denominations which would most closely replicate the 2013 
format.  
[redacted - section 9(2)(a)] and myself had an initial attempt at thinking about how to best 
design a question given this constraint in mind, this being below.  
 
 












Feedback on this approach would be appreciated - with regards to retaining quality and 
comparable data for this question within the constraints identified above. [redacted - section 
9(2)(a)] – [redacted - section 9(2)(a)] had already passed myself your comment about 
collecting the primary Christian denominations as tickboxes within the primary question, let 
me know if you think that approach would be most appropriate. 
Cheers, 
[redacted - section 9(2)(a)] 
[redacted - section 9(2)(a)]| Statistical Analyst/Kaitätari Tatauranga | 2018 Census 
Customer Needs and Data 
Statistics New Zealand - Tautauranga Aotearoa 
Phone: +[redacted - section 9(2)(a)]| Statistics House, The Boulevard, Harbour Quays, 
PO Box 2922, WELLINGTON 614 
[redacted - section 9(2)(a)]