Examination of Witnesses (Questions 80-94)|
WEDNESDAY 24 OCTOBER 2001
80. Were the enumerators trained to interview
(Mr Pullinger) Yes, they were.
81. In your memorandum you say, "... Census
processing is taking longer than originally scheduled, although
the Census offices, with the contractors, are invoking contingency
options in order to meet the deadlines for output delivery within
the overall Census budget". Why is the processing behind
schedule? What are the contingency options?
(Mr Pullinger) The first reason is that we extended
the period for enumeration in a number of areas in order to make
sure that we really did get into the most difficult places and
knock them up. We had an issue of completing the enumeration so
we were determined to get the job done as well as possible and
there was an extension to the time period we had right up front.
The information was then collected and sent to the processing
centre at Widnes where processing began. We found at the early
stages of that processing more need to fine tune the system than
we had anticipated. We had gone through a significant test of
the whole system in 1999. We started off with an initial batch
of forms from one area and ran them through the system. It did
take longer than expected. The first deliveries came later. We
now have deliveries. We have now tuned the system. Our contractors
have responsibility for this and we have been working very, very
hard with them to develop a catch-up strategy, which we now have.
We have a plan which takes us to final delivery on the right schedule
and the efforts we have made with Lockheed Martin, the main contractor,
and ICL, the contractor at Widnes, have created both a plan which
enables us to catch up, which we have now shown to work for the
first areas we have, and a series of contingency options to make
sure that stays on track. We know it will deliver.
82. How much is that going to take you over
your original budget?
(Mr Pullinger) That will not take us over our original
budget at all, because of the nature of the contract we have with
Lockheed Martin, which is for the delivery of information to a
83. You said earlier this system you are using
for counting was already tested and tried in the United States.
(Mr Pullinger) Yes.
84. Why were the lessons you are now learning
not learned then?
(Mr Cook) What Lockheed Martin would say is that it
has taken them until last month to achieve the keying rates in
the United Kingdom that they achieved much earlier on in the United
States. As a consequence what they have over the next six months
is the need to catch up the unkeyed material because they took
so much longer in the United Kingdom to get to the keying rates
they had achieved from the start in the United States.
85. Why was that?
(Mr Cook) I can only report what happened. Presumably
there are issues around the way in which expectations were formed,
about what the performance of keyers would be. There may well
be issues about how performance is recognised in the work force
which are different here from the United States.
86. My colleague has already mentioned that
some of the times do seem inordinately long before people see
these figures. The memorandum you have given us says that the
centre piece of local government finance, the Standard Spending
Authority, will not have necessary results available until 2003.
People will be operating over the next couple of years with data
which is 12 years old.
(Mr Cook) Yes and these were targets which were accepted
87. They seem to be rather prolonged.
(Mr Cook) Yes, but the trade-off which was preferred
at that time was very much that people would rather have a final
set of figures in early 2003 than a series of numbers in between
which were subject to some revision.
88. Are you sure you are not making the best
the enemy of the good? In other words you are refining these statistics
so much, thereby engendering a delay, when something which would
have been useful sooner could have been produced.
(Mr Cook) In all statistical work there is a trade-off
between accuracy and timeliness and availability. For us that
trade-off was made in a particular way during the 1999 period
when all these decisions were made about the processing path which
was to be followed. I am sure that users of statistics have a
different view of the timescale they are prepared to accept four
years before the results are available than when they are a year
89. What was the lowest response rate achieved?
In which area was the lowest?
(Mr Pullinger) The lowest areas were in the centre
90. What was that?
(Mr Pullinger) The difficulty of estimating the response
rate is that we only know who we have captured. Until we have
completed the process we will not know what we do not have.
91. How do you know it was the lowest then?
(Mr Pullinger) Because we know the forms we sent out,
we know the number of addresses we expected to find from our address
lists, we know any additional addresses that enumerators have
found and we can count the number of forms we got compared with
the number of those addresses. When you add those up you get to
the number of 98 per cent which has been quoted. The smaller the
areas you go to with that, the more chance that the difficulties
you have are going to be very substantial with understanding the
addresses and understanding the people in those areas. We have
a spread of different answers but they will not be authoritative.
They will not give an indication of what the actual final response
is. We believe we have managed to do better than we did in 1991,
but the scale of additional response understanding we will get
when we have done the coverage survey will be such that saying
to you anything below the national level would be to fool you.
Whether that is being the best enemy of good or not I do not have
a number in front of me I can give you.
(Mr Cook) In small areas we have two lots of response
rates: one is the response rate from our Census coverage survey
in relatively broader small areas; secondly in each of the Census
areas we have the difference between the national average and
what we actually achieve. What that number means is subject to
some interpretation. What we can certainly do is tell you what
we believe happened in the ten worst areas, if that would help
92. That would be helpful. Will we all know
at the end of the whole operation?
(Mr Pullinger) Yes.
93. What the rate was for each area?
(Mr Pullinger) Absolutely; yes. At the end of this
we shall publish a general report on all aspects of the Census
including issues such as the questions we had on the helpline.
Those reviews are in train and there will be a very comprehensive
quality report which will enable people to understand how much
of this has been estimated rather than counted.
94. Could you very briefly just describe the
main problems you had with the Post Office?
(Mr Cook) Fundamentally the Post Office, although
it had a very major role in the Census, refused to give us a contract,
which I discover is not unusual for the Post Office. So there
was no way we could get a guarantee of performance out of it.
What happened was that whilst we received the forms from the Post
Office in aggregate in a reasonably effective manner, the variability
of performance around the country was far greater than was anticipated
and it created huge problems for us locally in terms of the differing
impact of that poor response on the ability of area managers to
plan when the enumerators would go back into households. We had
the difficulty in some areas of people training people to receive
forms which were sitting for two more weeks in the Post Office.
We had people going back to households where forms had been returned.
In the areas where the Post Office was a problem to us it was
quite a significant problem. I would say the Post Office problems
were the largest difficulties that we faced in the conduct of
the Census. It was mainly the resilience of our local managers
that allowed them to keep on top of what were huge variations
in performance provided to them which should not have happened.
Chairman: Thank you very much indeed.