ESA/STAT/AC.84/15
6 July 2001
English only
Symposium on Global Review of 2000 Round of
Population and
Housing Censuses:
Mid-Decade Assessment
and Future Prospects
Statistics Division
Department of Economic and Social
Affairs
United Nations Secretariat
New York, 7-10 August 2001
Traditional
censuses versus alternatives *
Ireland
Aidan Punch **
* This document was reproduced without formal
editing.
** Central Statistics Office, Ireland. The views
expressed in the paper are those of the author and do not imply the expression
of any opinion on the part of the United Nations Secretariat.
Traditional censuses versus alternatives
1. Background
There are many reasons for conducting Censuses of Population. At the
most fundamental level a census provides a population headcount at national
level. Depending on how the census is organised, this national total may be
broken down by detailed geographic area. Indeed, the taking of a census may
even be enshrined in the Constitution of a country for the purposes of
determining political representation. In addition, budgetary allocations
invariably hinge on census outcomes. However, notwithstanding the legal or
quasi-legal nature of the census objectives at national level, the richness of
the data provided and the small area dimension of the census are its greatest
assets. Against this backdrop therefore the decision as to the best means of
conducting the census is one which national statistical institutes do not take
lightly.
At opposite ends of the methodological spectrum are traditional
censuses and those carried out using registers. In between are various
combinations of the two i.e. mixed models, supplemented in some cases by sample
surveys. Traditional censuses are commonly taken to mean those in which members
of the public are requested to respond to a census questionnaire and where a
field force of enumerators is involved in the field operation either delivering
blank forms or collecting completed ones, or both. In its purest form a
register-based census will, through the use of various statistical techniques,
combine results from existing administrative registers at the micro level to
produce census type information.
Which is the best method to use? What considerations should inform our
decisions? What price accurate statistics? These are some of the questions
addressed in the remainder of this brief note.
2. Have traditional censuses had their day?
Staging a traditional census is a multi-faceted project involving some
or all of the following components:
·
Holding a census pilot to test new question wordings or amendments to
existing ones.
·
Getting Government agreement to the final census questions.
·
Determining appropriate census geography to ensure manageable
work-loads for enumerators.
·
Recruiting and training a field force to carry out the census.
·
Handling support activities such as field force payments, warehousing,
transportation, logistics.
·
Conducting a public awareness campaign to highlight the importance of
the census and the need to comply with its provisions.
·
Using technology to ensure the speedy processing of the information
returned on the census forms.
·
Making the results available as speedily as possible using up to date
methods which are responsive to users’ needs.
Cost is normally cited as the main disadvantage of the traditional
census. Indeed this has now almost become part of the lore in some
international organisations with NSIs being admonished to move to the more
enlightened (and cheaper) register based approach. If this assertion is
repeated often enough then traditional censuses may just eventually be seen as
a bad thing. However, while not wishing to downplay the importance of cost, the
investment made (and that is precisely what expenditure on a census is) has to
be viewed in the context of the likely benefits which will flow from this
investment. As it is not possible to formulate policy in a vacuum, the
provision of comprehensive and timely information, such as that provided by the
census, is a pre-requisite for informed decision making. It is highly
questionable whether a move to a less preferred (and maybe less accurate)
method of carrying out the census in order to reduce expenditure is the most
cost effective solution in the longer term.
So, precisely what are the advantages of traditional censuses? First is
the fact that the NSI has control over the operation which means that it can be
organised in a streamlined and uniform manner conducive to optimising its
statistical potential. The comprehensive nature of the census in terms of topic
coverage should not lightly be overlooked when compared with some of the
alternative approaches. And there is the positive spin-off which the census
generates for official statistics in general, through its public awareness
campaign. As well as engendering a sense of national participation it presents
an opportunity to the NSI through its spokespersons to highlight the importance
of objective and timely statistics for society at large.
On the downside is the aforementioned cost argument, which cannot be
ignored especially when public finances are under pressure. To illustrate that
cost is a real issue there are many instances of censuses being cancelled
because of budgetary cut-backs. This happened in Ireland in the mid-seventies,
unfortunately (with the benefit of hindsight) at a time of major demographic
change.
A successful census field operation requires not alone detailed and
painstaking planning but also a modicum of good luck. This was singularly
absent in the case of the planned 2001 census in Ireland when because of the
precautions in place to prevent the spread of foot and mouth disease it was decided
to postpone the census for a year. The revised census date is 28 April 2002.
Respondent burden is also a real issue to be contended with. On the
basis of crude estimates, the opportunity cost of completing the Irish census
questionnaire may be as high as a quarter of the overall census budget.
However, this is part of the price to be paid for quality information for
policy formulation and assessment.
The fact that censuses are carried out infrequently gives rise to major
organisational headaches. However, if they are carried out every five years, as
is the case in Ireland, then the institutional memory both in the statistical
office and in the field means that the learning curve is not quite as steep as
it otherwise would be if starting from scratch every ten years.
Finally, traditional censuses are finding it hard to grapple with
certain sub-populations such as the single young mobile population, people
living in ghettos or shanty towns and marginalised sections of the community.
While these groups pose major problems for census takers it is unlikely whether
any other method of enumerating them would be any more effective.
3. Administrative registers – a statistical panacea?
The major positive factor associated with using administrative
registers for statistical purposes is the fact that the same information is not
collected more than once. Where recourse can be had to existing data holdings
then, depending on their quality and coverage, they can be collated in order to
make them amenable to statistical reporting. They can also be merged with other
data sources in order to optimise the output being produced. This reduces the
burden on the respondent by making use of whatever data is already in the
system. The immediate saving is financial. Costly field procedures are avoided
and most effort is concentrated on improving the quality of the registers
themselves and using relevant statistical techniques to derive the required
output.
The NSI has greater control as well over timing issues. In theory data
sets can be merged at any time allowing great flexibility in terms of reference
date and periodicity.
However, there may be drawbacks to using administrative registers as a
source of statistical data. At the basic level is the degree of compliance with
the register on the part of the public. This has an impact on both the coverage
of the register and on the accuracy of the information contained in it. Where
the administrative rationale for registering correctly and on time is strong
then the accuracy of the register is likely to he high. But is this always the
case? In societies which are highly regulated the public may be used to and
accept the need to register for various public schemes. Even in these cases
there may be reasons why individuals would not want to register. A lot depends
therefore on the degree to which the administration maintains its register and
sorts out its shortcomings.
A definite drawback is that the information provided may not be
precisely what is required for statistical purposes. In some cases no
information at all may be available on certain topics. However, it is a choice
for the NSI to either settle for what may be a sub-optimal source of data
because of cost considerations or to mount a statistical operation which will
have direct financial and indirect response burden implications.
The public also has to be assured of the confidentiality of the
statistical process. In other words they have to understand the distinction
that individual identifiable data on administrative registers may be used by
the NSI for statistical purposes (indeed merged with other such data) but that
no such information passes from the statistical office to any outside agency
(the so-called one-way street). Not alone do they have to understand such
subtleties but they also have to implicitly trust the NSI to do the right
thing. Direct statistical inquiries have the merit that the methods used can be
easily understood by members of the public. Complex data merging and register
cleaning does not have the same simplicity.
The way forward
Censuses are national instruments paid for out of tax payers’ money. It
is for the countries themselves to determine what offers the best value for
money, taking account of factors such as direct financial cost, data quality,
burden on respondents, public acceptability etc. Indeed in some countries there
may be no real choice as the traditional census may have fallen into disrepute
for various historical reasons.
Where the choice exists between mounting an expensive field operation
or using excellent data from registers then this will be a simple matter.
However, life is rarely that simple. In countries where good administrative
data is limited it may not be feasible either politically or culturally to
rectify the situation in the short term. Neither, is it sufficient to say that
registers should be put in place in order to reduce the cost of collection of
statistics. Experience shows that there has to be a very strong administrative
rationale for developing and maintaining register based information. Without
that the level of compliance by the public will render any information which
may flow from the register to be of limited use for statistical purposes.
The real choice is not between one end of the spectrum and the other
(i.e. traditional versus register). In reality many NSIs use combinations of
direct collection and administrative information. For instance, use may be made
of a national address databases to assist in the field operation. Ultimately,
NSIs are accountable for the money they spend. They therefore have to justify
to Government that the census will be carried out in the most cost-effective
way possible. If the Government sanctions this expenditure (even if it is the
cost of conducting a traditional census) it does so bearing in mind that it is
an investment in the knowledge necessary to be able to formulate public
policies which are well informed and of long term benefit to citizens.
Aidan Punch
Senior Statistician
Demographic Statistics
Central Statistics Office
Ardee Road
Rathmines
Dublin 6.
Ireland
Phone +353 1498 4316
Fax +353 1496 4597
Email aidan.punch@cso.ie
________________