Symposium
2001/44 25 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
Statement from Malaysia *
Harbans Singh **
CONTENTS
1. Strategies for involving stakeholders in census
activities
3. Adapting new technologies to census operations
4. Maintaining census-related activities during
intercensal years
5. Identifying and resolving problems of census mapping
6. Post-enumeration surveys: Are they worth it?
1.
The 2000 population and housing census of Malaysia
(or Census 2000) was the fourth decennial census to be conducted since the
formation of Malaysia in 1963. The
three earlier censuses were conducted in 1970, 1980 and 1991. The history of census taking goes as far
back as 1750 when colonial counts were made for the state of Malacca.
2.
Under the constitution of Malaysia, census
taking is a federal responsibility and the legal basis for the census is
provided by the Census Act of 1960, which empowers the government to conduct a
census from time to time. The responsibility for conducting the 2000 population
and housing census was vested in the Department of Statistics (DOS), which
implemented the project in collaboration with the state governments of
Malaysia.
3.
The 2000 population and housing census of
Malaysia, which had 5 July as its census day, was undertaken (using the de jure
approach) over a period of some 17 days.
Discussed briefly below are some of the experiences from the 2000 census
with relevance to the six symposium issues at hand.
4.
The 2000 census was funded by the government of
Malaysia through the Economic Planning Unit (EPU) under the development budget.
A total of RM190 million was allocated for this project.
5.
The Census 2000 project was under the direct
purview of the Census Steering Committee (CSC) chaired by the chief secretary
to the government. Members of the
committee were drawn from key ministries, departments and central agencies. The
CSC was responsible for providing policy guidelines in relation to finance and
manpower requirements as well as towards the planning and implementation of the
project. The chief statistician of
Malaysia was appointed the commissioner of census.
6.
In addition, technical support was also provided
by the Census Technical Committee chaired by the director-general of the
Economic Planning Unit (EPU). The committee consists of officials from various
relevant ministries, government departments and agencies which are the main
users of census data. This committee
provided the forum to users to air their views on what should be incorporated
in the census questionnaire. As would be expected, each user agency put forth a
list of questions/topics which were of interest to its own agency. When added
up, if all topics requested had been included in the 2000 census, the
questionnaire would have been too long and would have overburdened respondents
as well as enumerators, thereby involving huge costs.
7.
Alternative methods were suggested to some user
agencies on their data requirements.
Some of the data could be obtained from existing administrative records,
while others could best be obtained through sample surveys. For other requests,
questions were included in the census pre-tests to assess their viability; if
the results were found to be inaccurate or unreliable, it was suggested to
users that the census enquiry was perhaps not the best vehicle for obtaining
the required data.
8.
In the final questionnaire the topics added
(when compared with the 1991 census) were (i) disability, (ii) year of first
arrival in Malaysia of the foreign-born, (iii) literacy, (iv) field of study of
those with tertiary education, (v) rental, (vi) additional household items such
as handphone, personal computer (PC) and Internet, (vii) number of bedrooms,
(viii) garbage collection facility and (ix) fertility/mortality questions for
Sabah and Sarawak only.
9.
The items dropped in the 2000 Census were (i)
place of usual residence at time of census, (ii) number of hours worked, (iii)
previous employment of currently unemployed and (iv) year of construction of
the housing unit. In summary, the total
number of topics canvassed in the past censuses were 53 (1970), 50 (1980), 33
(1991) and 43 (2000). The 43 topics canvassed in the 2000 census were arrived
at given the constraints of the length of the questionnaire. The EPU is the
lead government agency in the use of census data; it makes use of these data as
inputs in the preparation of the five-year development plans, the latest being
the Eighth Malaysia Plan (2001-2005).
10.
Stakeholders also forwarded their requests
through meetings of the Census Technical Committee on the type of tables that
they would require for purposes of planning, implementation and evaluation of
programmes initiated by the respective government agencies.
11.
Given that the stakeholder’s requirements are
data at very small geographical levels, it is inevitable that the census
approach is the only method that would satisfy their requirements. The
undertaking of sample surveys would not meet their requirements.
12.
The registration-systems approach has not been quite
developed and is not accurate at the present moment in Malaysia. Information on
births and deaths is not a big problem, apart from some underreporting and late
reporting, especially in the states of Sabah and Sarawak. The more intractable
problem lies in obtaining data on internal migration and external migration.
Malaysians moving to different states, districts or addresses are required to
register the change in address, but this is seldom done and the accuracy and
reliability of these data are rather questionable. The problem is even more
intractable in the case of cross-border movements, especially of illegal
foreign workers, mainly from neighbouring countries.
13.
Census 2000 was conducted using the canvasser
approach, whereby some 54,000 enumerators went house to house and filled in the
census questionnaire through personal interviews of any responsible member of
the household. The planners of the 2000 census were faced with the task of
deciding the data-capture and processing strategy, which has important
implications for the nature and design of the census questionnaire. The 1970
and 1980 censuses used Optical Mark Readers (OMR), whereas in the 1991 census,
key-in through PCs was employed. After reviewing the pros and cons of the
key-in approach and OMR/scanning devices, it was decided that the key-in
approach using PCs, which had successfully been carried out in the 1991 census,
should continue to be used in Census 2000.
The aim of this approach was to make data available to users in the
shortest possible time without sacrificing accuracy.
14.
In Census 2000, the
strategy for the processing activity was decentralized, similar to that in
1991. Thirteen processing centres were set up, and about 2,800 persons were
recruited; 403 units of personal computers (PCs) were used in the processing
activity.
15.
At each of the 13
centres, the main processing activity was divided into two phases, that is,
manual processing and computer processing. Phase one started in August 2000;
the questionnaires underwent checks for completeness for non-coded questions
such as birthplace. In phase two, the questionnaires underwent coding of
questions relating to education, migration, occupation and industry. Quality-control checks were imposed, and if
the errors in the sample examined exceeded the acceptance level, then a
complete check of questionnaires was made.
16.
The computer package
used in the processing activity was the Integrated Microcomputer Processing
System (IMPS), where the data-entry module CENTRY was loaded in all 13
processing centres. The CONCOR module was designed to take care of consistency
and correction of data during the process of validation and imputation. During phase one, imputation was done on
four variables—namely, age, race, sex and citizenship—at the processing
centres.
17.
During the
census-processing stages, CENTRACK, a module of IMPS, was used as a management
information system. It enabled the progress of work at various processing
stages to be monitored both by the management in the processing centre as well
as in headquarters to identify bottlenecks in any of the processes with a view
to redeployment of resources to create a smooth and efficient flow of work.
18.
The strategy
of a two-phase processing approach meant that data from the first stage could
be released much earlier.
19.
In terms of
dissemination of data, summary data from the 2000 census were posted on the
Internet at the Department’s web site (www.statistics.gov.my.) It is planned to have the data available
also on CD-ROM, as well as in other digital media to satisfy the needs of
particular users.
20.
In terms of
outsourcing, the 1980 census saw the outsourcing of the housing questionnaire
for data capture by a private company, the cost of which was borne by the
Ministry of Housing and Local Government. For reasons of confidentiality, DOS staff were stationed
at this centre.
21.
Census 2000, needless
to say, a huge statistical undertaking, covered a period of five years
(1998-2002) and encompassed such activities as planning, pre-tests, trial test
and census proper, as well as analysis and publication of census reports.
Census-related activities that can be maintained during the intercensal years
are described below.
a.
Documenting the experiences from the 2000 census
22.
It is important that
every aspect of the census activity be documented, the good experiences as well
as the shortfalls and weaknesses, so that future census takers will not have to
“reinvent the wheel”. This is especially so since census officials involved in
Census 2000 may have retired or been promoted or transferred to other
divisions. All census documents and materials should be collated and bound and
kept in the library for reference by future census takers.
b.
Writing of census monograph series
23.
Another intercensal
activity that can be carried out is the writing of a monograph series based on
the richness of data collected in the census. “Data mining” will enable junior
officers to acquaint themselves with the techniques of analysis.
c. Examination of sample 2000 census
questionnaires
24.
While the bulk of
Census 2000 questionnaires will be destroyed after the census, a sample of
these questionnaires has been kept. This will allow the examination of
responses to certain open-ended questions, as well as the descriptions written
for the “other” category. It is hoped that this exercise will help to identify
the weaknesses in questionnaire design that can be improved in future censuses.
d.
Updating of census household frame
25.
This activity is a
continuous activity and does not stop when the census is completed. An up-to-date
frame is important for use not only in household surveys but also as
preparation for the next census.
e. Undertaking a mini-census
26.
Another possible
intercensal activity is the undertaking of a mini-census five years after the
actual census. The mini-census would inevitably have to be small-scale, with
only basic demographic information being obtained. This is especially relevant
in Malaysia because of the timing of the five-year development plans, which
require many data inputs.
a.
Decentralized mapping
27.
In Census 2000, the
mapping activity was decentralized and undertaken by the 13 state branches with
standardized guidelines issued by headquarters. This approach resulted in better
maps, since staff at the state branches had very good localized knowledge of
the geographical areas on the ground.
b.
Mapping as an ongoing activity in the intercensal period
28.
Mapping should be
treated as an ongoing activity even though the census has just been completed.
This is important, since a good frame is required for household surveys which
are conducted on a regular basis. With this in mind the 13 state branches
update the frame based on local knowledge of the growth areas in their respective
states. Given this ongoing activity of updating of the census frame, it is
hoped that there would not be a last-minute rush to produce census maps.
c.
Automated census maps through Geographic Information System (GIS)
29.
For Census 2000,
enumeration block (EB) maps were manually drawn. The Malaysian Government has set up a National Land Information
System (Nalis), whereby government agencies can share digital geospatial
information with such layers as roads, rivers and so forth. It is hoped that this digital information
from other agencies can be used to generate automated enumeration block maps,
as well as other census maps, through GIS in the future.
a. Census Coverage Evaluation Survey
30.
In Malaysia, post-enumeration
surveys (PES’s) have been conducted since 1980. The Census Coverage Evaluation Survey
(CCES) was carried out in August 2000, i.e., one month after census
enumeration. It was aimed mainly to
estimate the extent of coverage error and content error (based on citizenship,
ethnicity, sex and age).
Geographically, CCES covered the population residing in
non-institutional living quarters in the whole of Malaysia. A total of 1,800
EBs were selected.
b.
Independence of survey
31.
To ensure independence, the following approaches
were adopted:
·
Staffing. The operation of the CCES
was managed by a group of permanent officers who were not involved in census
activities. Field enumeration and
office processing were done by specially recruited temporary enumerators under
the supervision of the permanent officers of DOS.
·
Operations. The four phases—i.e.,
transcription, listing and enumeration, office matching and field matching—were
carried out by different sets of temporary enumerators.
c. Adjustment of
census results
32.
In the 1991 census,
the underenumeration rate of the population was 4.4 per cent for Malaysia as a
whole. Adjustment of basic demographic
data was done. In Census 2000 it is envisaged that similar adjustments will be
carried out.