Memorandum by The Market Research Society
The Market Research Society (MRS) is the world's
largest professional body for individuals employed in market research
or with an interest in it. Founded in 1946, it is the largest
body of its kind with over 8,000 members working in most organisations
currently undertaking market research in the UK and overseas.
The importance of the Census to market researchers
is fundamentalmuch of their work, in terms of population
surveys in this country, is underpinned by the Census.
This submission has been prepared on behalf
of the MRS by the Census and Geodemographics Group (CGG), which
is an advisory group of the Society. The CGG represents the views
and interests of census users in market research and CGG members
are drawn from a wide range of sectors, including market research
agencies, client users, users within census agencies and independent
consultants who advise on census applications.
The views expressed in this submission are supported
by the Demographics User Groupwhich represents fourteen
large commercial users of census and other government dataand
the Association of Census Agenciesan informal grouping
of census agencies.
The main focus of this submission is the usability
of outputs from the 2001 Census, however we will start with some
general comments on innovations from a user's perspective, in
order to "set the scene".
When planning for 2001, ONS undertook to "break
the mould" set by previous censuses. The 1991 Census had
followed the model used in 1981 and had generally been successful.
However, the 1991 results were called into question due to variations
in the coverage rates for different demographic groups and so,
for 2001, ONS made its goal to maximise population coverage and
produce estimates which had been pre-adjusted for non response
(the One Number Census).
At the same time, ONS were able to take advantage
of technological advances at key stages in its data capture and
As a result, a number of significant innovations
were made in the 2001 Census:
2.1 Data Collection
(a) Geographic Information System (GIS) technology,
combined with the Ordnance Survey AddressPoint product, was used
to plan the enumeration districts and provide address lists to
enumeratorsthis we welcome.
(b) The majority of forms were collected
by post back from respondents; this was followed by enumerator
visits in order to obtain forms from non-respondents. As census
users, we were concerned about how well this procedure would work
and the quality of the information that would be obtained. In
previous censuses, the forms were personally collected by enumerators,
who were thus able to check and rectify the form before leaving
In the event, the postal response was far greater
than had been expected and, overall, ONS estimates that forms
have been returned from some 98 per cent of householdswhich
is similar to the coverage level in 1991. However, the data collection
procedure has led to knock-on problems, such as the flood of calls
to the Census Helpline and highly adverse publicity in the national
In terms of quality of data collected, users
will have to wait until the Census Quality Report is published
in 2003. However, preliminary information on question non-response
rates has just been published on the ONS web site. This shows
that for the majority of questions asked in both 1991 and 2001,
the non-response rate was higher in 2001typically 1 to
5 per cent in 2001 compared with less than 1 per cent in 1991.
For questions that were harder to answer (eg employment related
topics) the non-response rate was typically 5 to10 per cent, with
a highest rate of 10 to 15 per cent for professional qualifications
and company size.
Therefore the new data collection procedure
appears to have had an adverse impact on question response rate,
which we regret.
(a) We strongly welcome the move to 100 per
cent coding of question responses; also the reorganisation of
processing to publish outputs for tranches of statistics across
all areas of the country, instead of on an area-by-area basis
as done previously.
(b) The One Number Census should lead to
improved outputs, but adds significantly to the processing time-scale.
(c) We currently expect to see results delivered
to users in the first half of 2003, assuming that no major problems
arise, and this is equivalent to the delivery time achieved on
the 1991 Census (after the impact of a six month delay due to
a data coding problem).
(a) We also welcome the new geography of
Output Areas which will be created for the 2001 Census, together
with the associated package of geographical boundaries and mapping
data to go with them.
(b) We appreciate the efforts put into consulting
with users on the output products, however these have largely
been negated over the last few weeks following the ONS announcement
about new measures for disclosure control (see section 3 below).
(c) We welcome the agreement to include,
at our request, Census Area Statistics for Approximate Social
Grade (the ABC1 classification used in market research and advertising).
(d) We welcome, of course, the Census Access
Project which will provide census results to users effectively
free of charge. We expect that this, combined with availability
of data over the Web, will significantly extend the user base
and increase the uses made of the 2001 Census.
OF 2001 CENSUS
The combination of output products and geography
planned by ONS, in consultation with users, has been designed
to obtain maximum benefit out of the 2001 Census, subject to confidentiality
requirements. ONS have held four rounds of output consultations
over the last three years and have made every effort to elicit
the views of census users.
However, this has all been placed at risk following
the recent decision by the National Statistician over disclosure
control, which was made public to users on 12 November 2001.
The National Statistician has now decided to
introduce additional disclosure control measures, which we are
concerned could seriously degrade the quality and usefulness of
All counts in all Census outputs issued for England
and Wales will be rounded either to zero or to numbers divisible
This rounding will apply to every cell count
and every total in each output table, therefore the effects will
The counts in each table will be
either zero or multiples of threethis will be incomprehensible
to ordinary users.
If those counts are expressed as
percentages, the percentage figures will "clump" on
particular values and so will appear to be incorrect.
The cell counts will not generally
agree with the row and column totals, which in turn will not agree
with the overall total, ie the table will be internally inconsistent.
We believe that the rounding is unnecessary
and will create confusion amongst inexperienced users and ridicule
for ONS. It comes on top of perfectly adequate existing blurring
measures built into the Census operation:
1. The swapping of records in the Census
database, before producing outputs, which users had accepted as
a method of blurring results while maintaining consistency between
and within tables.
2. In addition, the database will already
have been blurred as a result of the One Number Census processes
and imputation modelling applied in order to fill gaps in the
3. The data will also be automatically blurred
by the passage of time, since it will be two years out of date,
when first results are made available to users.
We understand that the Registrar General for
Scotland has decided that the rounding measure is not necessary
for the Census in Scotland (and that no decision has been taken
as yet for Northern Ireland). Therefore why has this decision
been taken for England and Wales, and why has it been made at
this late stage without any user consultation?
We are not alone in our belief that the National
Statistician has made an uncharacteristically bad decision over
disclosure controlwe understand that he has received criticism
from all major census users. As a result, he has arranged to meet
and discuss the issue with representatives of the user groups
on Thursday 13 December.
Finally, as a further confidentiality measure,
the National Statistician has decided to raise the population
threshold for output tables to 100 resident persons and 40 resident
households for the release of Census Area Statistics (CAS). We
feel that these levels are unnecessary, as the Output Area production
system has been designed to create areas containing 100-125 households.
However, in rural areas such as many parts of Wales, the planned
threshold is too high and would result in suppression of their
In 1991, the population threshold was 50 persons
and 16 households. Prior to the recent decision by the National
Statistician, the previous consultation paper by ONS had proposed
a threshold of 50 persons and 20 households, which we feel would
have been adequate.
10 December 2001