Household information is
traditionally collected through the "canvasser
method" in Singapore. Field interviewers have
to be deployed across the island to make door
to door visits. The information is collected
through face to face interviews and the answers
are recorded on paper questionnaires.
The "canvasser method" meant that transport
costs and travelling time have to be incurred.
These costs are further increased when field
interviewers could not contact the selected
household, and repeat visits have to be made.
As the number of cases completed per man-day
is low, large field forces are required for
household surveys. In addition, a data entry
team is required to translate the data collected
from the field into electronic form.
With the tight labour market in Singapore,
it is clear that a more productive approach
is necessary for data collection in order to
reduce manpower needs, lower costs and speed
up the process of data collection. Following
the 1990 Census of Population, the Department
of Statistics embarked on an extensive study
to review its overall data collection strategy
for household surveys and to find new ways to
collect data more efficiently. The result in
re-engineering the interview process is the
Computer Assisted Telephone Interviewing (CATI)
system. It is an interactive front-end computer
system that aids interviewers to ask questions
over the telephone. The answers are then keyed
into the computer system immediately by the
interviewer.
II.
The Use of CATI in the 1995 GHS (General Household
Survey)
In the 1995 GHS, a mini-census
covering about one-tenth of Singapore's population,
the CATI system was used for the first time.
The GHS is believed to be the first large-scale
survey in the region to use CATI and the Department
of Statistics is proud to have pioneered this
technology which is developed in-house over
a period of 11 months.
The successful deployment of CATI in the GHS
saw many benefits, most notably significant
reductions in manpower, costs and time required
for the survey, with no compromise to the quality
of the data collected. Most households welcomed
the new approach, as they no longer had their
privacy disturbed by strangers coming to their
doors. Some two-thirds of the selected households
were covered by CATI.
III.
Features and Benefits of the 1995 CATI system
The CATI system had several
important features which enhanced its capabilities
substantially.
Direct Data Entry and Coding
CATI allowed the interviewers to perform multiple
tasks of interviewing, data entry and simple
coding simultaneously.
The interviewers keyed the information directly
into the computer system. As a result, data
entry as a separate process was no longer necessary.
Most of the questions were in multiple-choice
format, so that the CATI interviewer need only
point and click on the right answer. The answer
was then translated directly into a code by
the CATI system and updated in the database.
With the exception of "occupation" and "industry
employed in" which required descriptive answers,
most of the data captured by CATI could be coded
in this way. This reduced coding operations
in the survey substantially.
Automatic Branching of Questions
Unlike the traditional paper questionnaire,
the CATI housed what was effectively a smart
electronic form. This form tailored the questions
for each respondent in the household based on
key information given by the respondent during
the interview. For example, each full time student
was asked the level of education attending and
the field of study. The system then skipped
questions on economic activity and prompted
the interviewer to ask the question on transport
mode to school. As a result of this automatic
branching of questions, the interview time for
each household was much shorter (about 20 minutes)
compared with the canvasser method (up to 30
minutes).
On-line Verification Checks
The smart electronic form also performed on-line
checks on the responses keyed in and alerted
the interviewer of inconsistent or doubtful
answers. The interviewer could then verify with
the respondent on the spot, and correct any
errors immediately. This reduced the inconvenience
to respondents to verify their particulars at
a later date, and improved the overall quality
of data collected by CATI.
Automatic Scheduling and Dialling
The CATI system scheduled and dialled up households
by an automated process. To dial a household,
the interviewer need only click the dialling
button on the screen (see picture 1). The system
then searched for the next household to call,
based on certain priority rules. If the call
was not answered, the system automatically scheduled
the household to another time. If the interview
could not be completed, the system also allowed
interviewers to re-schedule the appointment
to a date and time favoured by the household.
Picture
1 CATI dial up screen
For households with listed
telephone number, the date and time of the initial
scheduled interview were according to that printed
on the Department's notification letter to them.
Households with confidential telephone numbers
were requested to dial a toll-free hotline number
for assistance on enumeration.
The CATI system was programmed to assign the
appropriate interviewer based on the predominant
ethnic group of each household. This reduced
any language barriers between the respondent
and the interviewer.
IV.
Set-Up of the GHS CATI call centre
Some 90 CATI terminals were
deployed during the 1995 GHS, over a period
of about 4 months. Unlike the traditional "canvasser
method" where field interviewers were based
in various regional offices to carry out fieldwork,
the use of CATI meant that the interviewers
could all be based in the survey headquarters.
This proved beneficial as supervision and logistic
support could be better given.
To support the various capabilities of the
CATI system, the set-up of the CATI call-centre
in the survey headquarters involved the following:
Network Setup
A Novell network was used to interconnect all
the PCs in the call centre together on one logical
LAN. Thus operations like backup, troubleshooting
as well as loading of data were done via the
LAN from a central administrator terminal.
A total of 118 PCs at the peak were connected
to this network. The database files (Microsoft
Access database files) were stored both locally
on each PC as well as on a file server dedicated
for the GHS operations at that time.
When the CATI PCs first start up in the morning,
the system will check for updates to the data
file and downloads it from the file server.
Any subsequent updates were made when the system
detects a change in the data on the file server.
Headsets and Phone Quest Cards
The Interviewers' PCs were not only installed
with the CATI software which enabled them to
dial-out to respondents, but also equipped with
Phone Quest hardware cards and headsets that
enabled hands-free dialing. The interviewers
could key-in responses from the interview into
the system directly.
Supervisor PC setup
Supervisors could have full access to the CATI
interviewers' system from their own PCs. They
were also given a simple monitoring module which
allowed them to look at various indicators relating
to the usage of the CATI system. Their PCs were
equipped to pick up any household records from
the database to do random checking & verification.
Hotline PC setup
Ten hotline PCs were available for respondents
to make appointments or reschedule their appointments.
Also, the hotline PCs could be used to conduct
on-the-spot interviews if so requested by the
respondents.
Administrative Information
The call-centre system tracked administrative
information, including the date & time of
the interviews as well as the duration of every
interview. The information was stored in a log
file for generation of management statistics.
Back-end Operation
At the end of each day when all the CATI interviewers
had completed their work, a LAN-based program
was initiated to consolidate the various data
files from all the CATI operators. This consolidated
file was then processed and incomplete records
were redistributed to the various PCs for the
following day's operations (see picture 2).
All completed records were extracted and subsequently
uploaded to the mainframe system which housed
the master household records. Mainframe jobs
were then initiated to update these records
back to the IBM DB2 database.
Picture
2 Movement of records between CATI PCs and database
Reports were printed at the end of each operation
to highlight the number of records outstanding
in the CATI system and the number of records
that have been completed. The report helped
supervisors in their workload allocation.
V.
The CATI system of the future
The Department of Statistics has achieved
another milestone with the successful deployment
of CATI in the 1995 GHS and is planning to adopt
CATI as one of the key data collection mode
in the 2000 Census of Population. The CATI system
for Census 2000 will be developed in partnership
with a professional call-centre, building upon
the success of the 1995 system, with additional
state-of-the-art features. The components included
in the new CATI system is given in picture 3.
They include intelligent call distribution,
calls tracking and a logical scheduling engine.
In addition, an even user-friendlier screen
would be designed, with comprehensive online
help features.
Picture
3 CATI Design for the 2000 Census of Population
To improve the efficiency
of the system and to reduce the backend operation
of updating records from PCs to a master database,
the system will be redesigned to take advantage
of the latest advances in client server technology.
Instead of storing a limited list of records
in each of the CATI PCs, the records will be
stored in a central server. All CATI terminals
would in effect draw from this central server
or "live database" of household records. This
would improve the efficiency of scheduling records
for dialling, as all the records will be in
one central queue, to be allocated to any available
CATI operator according to certain priority
rules.
The hotline operations will also be streamlined.
Hotline operators will be equipped to select
household records and accord them with the highest
priority for immediate interview by CATI operators.
This will free the hotline operators from the
need to conduct actual interviews.
VI.
Conclusion
Singapore will be one of
the first countries in the world to deploy a
state of the art CATI system for a Census Operation.
With manpower shortage on the one hand, and
excellent tele-computing infrastructure on the
other, the use of CATI would take over the canvasser
method as the main method of data collection
in household surveys in the foreseeable future.
