UNICEF developed the Multiple Indicator Cluster Survey (MICS) to monitor goals established at the World Summit for Children (WSC) held in New York in 1990. But it is also consistent with many monitoring needs of the Millennium Development Goals (MDG), as well as those of the more recent World Fit for Children (WFFC) that many countries are now using for human development planning into the 21st Century. MICS was created especially to meet the needs of developing countries lacking reliable routine sources of statistics and/or experience in carrying out reliable household surveys to measure performance relative to the WSC and, now, the Millennium Development and WFFC Goals.
The MICS is particularly important for Timor-Leste both because of the relevance of the these goals and the indicators that can be estimated from MICS data for policy formulation and planning and because the MICS can be used to establish more up-to-date baseline conditions in this newly independent country to help to more clearly define the challenges that lie ahead. The last comparable survey (Indonesian Demographic and Health Survey (DHS)) was carried out in 1997. The recently completed World Bank Poverty Assessment contains some relevant information, but not the depth of attention to child health, development and human rights provided by the MICS.
The Government has also established its own vision and agenda for the county?s future. Supporting that agenda through the provision of reliable base-line information is one of the main reasons for undertaking the MICS at this time.
Results of the survey clearly show the extent of current problems, but they can also assist Government in planning the way forward. Strengthening basic health and education infrastructure, including the all-important human resource base of teachers and trained medical personnel, will necessarily be a major priority. But there is also considerable scope for institution building at the community level and working with the people themselves to build demand and strengthen knowledge and capabilities to ensure a sound and healthy environment for children to develop and prosper. There are significant levels of ignorance and lack of power on the part of families and communities to adequately safeguard the rights of children that, in fact, contribute to the current conditions along with deficiencies in the service networks. Building or renewing relevant community-based institutions, particularly through the effective empowerment of women who are the main caregivers, can be a critical vehicle to improve overall conditions for the development of Timor-Leste's future generations.
Kind of Data
Sample survey data [ssd]
Major topics covered in the MICS report include:
- Population and Household Characteristics
- Infant and Child Mortality
- Water and Sanitation
- Child Malnutrition
- Child Health
- Reproductive Health
- Other Child Rights
Producers and sponsors
National Statistics Office
United Nations Children's Fund
European Community Humanitarian Aid Office
United Nations Children's Fund
A multi-stage sample design was used to select a total of 4000 households in 200 clusters of 20 households each. These choices are generally consistent with the guidelines in the MICS Manual, but were also predicated on practical considerations of time, budget and expected interviewer workloads. The sample was designed to produce national estimates of key MICS variables with acceptable levels of sampling error as well as more indicative estimates for various strata within the country. Strata of interest were those defined for the 2001 Suco Survey and included breakdowns by rural and urban residence, major region (East, Central and West) and highland or lowland location with the latter defined in terms of Suco with a majority of their area above or below a 500-meter elevation. Among urban areas, special attention was also given to the two major urban centers of Dili and Baucau where it was thought conditions would be likely to be different from those in smaller towns and rural areas in the country.
The first stage involved assignment of clusters to village level units (Suco) using systematic random sampling with probability proportional to population size and using populations from the 2001 Suco Survey. Suco were listed in serpentine fashion within Sub-districts (Posto), Sub-districts in serpentine fashion within Districts (Distrito), and Districts in serpentine fashion for the entire country in order to minimize any spatial bias in the selection process. Sampling was also stratified between Suco in the two major urban centers of Dili and Baucau (40 clusters) and those in the rest of the country (160 clusters). A total of 187 Suco were selected, with 13 Suco having more than one cluster due to their large population size.
The second stage involved selection of primary sampling units, which were defined in terms of sub-village units (known as Aldeia). Aldeia within the selected Suco were listed according to sequence numbers from the Suco Survey along with their populations and one or two aldeia (the latter for the Suco with multiple clusters) were chosen with probability proportional to population size, again using populations from the 2001 Suco Survey as the guide. In a few cases, very small Aldeia were grouped together before selection to ensure a sufficient base number of households for drawing the final sample of 20 households in the field.
Sampling of households was carried out as part of the fieldwork just before interviewing took place. In most cases, this was based on a new listing of households prepared in the field based on any listings held by and/or discussions with Aldeia heads that were then confirmed by quick observation in the field. In the major urban centers, however, greater care was taken, and here a door-to-door listing operation was performed. Once listing had been completed, random samples of 20 households were selected in the field using random number tables. This resulted in a non-self weighting sample at this level. However, the preferred alternative (using fixed sampling intervals) was not feasible given the great uncertainty over the actual population sizes reported in the Suco Survey and the need to keep interviewer workloads as consistent as possible.
More detailed information on sampling procedure is available in Annex III of the report.
Out of 4000 households in the sample, 3982 were located and successfully interviewed for the household questionnaire giving a response rate of 99.6 percent. Non-responses here included 5 outright refusals, 9 cases where no one was at home during the times that the interviewers could attempt an interview and 4 cases where the dwelling unit could not be found. There were a total of 4802 women aged 15-49 in the interviewed households of which women?s questionnaires were able to be completed for a total of 4606 (95.9 percent). However, almost all eligible women who were caregivers of children under age 5 for whom data were obtained were included in the group successfully interviewed. Of 4493 children under age 5 in the interviewed households information was obtained for 4454 or 99.1 percent. There was little variation in response rates across strata. All of these are consistent with high rates of response for surveys of this type and thus add confidence to the credibility of the overall data collection process. Response rates to the various questions were also high with numbers of missing values seldom exceeding a few percent of the total. Of particular interest was the ability of interviewers to obtain complete information on birth dates for nearly two-thirds of eligible women aged 15-49 and nearly 95 percent of eligible children under age 5. This greatly facilitated correction of reported ages in the household questionnaire, particularly for young children. Even though there may be some reporting errors remaining, this is extremely important due to the need for at least reasonable accuracy in age reporting for a number of the key child indicators in the MICS.
Dates of Data Collection
Data Collection Mode
Data Collection Notes
The design provided for 10 field teams with each team made up of four interviewers, one supervisor, and one person to do anthrompometric measurement and 1 driver. This was based on considerations of vehicle capacity and timing. It was hoped that at an average of about 5 interviews per day per interviewer, a team could complete one cluster (20 households) in about one-day's work and that with 10 teams and about 20 clusters per team, field operations could be completed over a period of between three and four weeks. In actual practice, this proved to be somewhat optimistic given difficulties with local terrain and needs to accommodate listing and anthropometric measurement as part of the process. However, delays were manageable and all fieldwork was completed within just over five weeks after the commencement of formal field operations.
Female interviewers were recruited locally with about one-fifth having had previous experience on the Suco Survey. Supervisors were almost entirely male (9 of 10) and came from the Timor-Leste Statistics Office staff. Anthropometrists were local female health personnel. Training was carried out over a 4-day period (5-8 August 2002), with the interviewers, supervisors and anthopometrists divided into two groups with each one containing members of five of the field teams. Even through team members had different jobs, grouping teams for the training turned out to be useful. This was because the individual teams had to be able to work together in the field and to cover for each other if necessary. Training in this way allowed the teams to start building a sense of teamwork and cooperation before the actual fieldwork began. However, separate training was still organized for anthropometrsts (by UNICEF) and for supervisors (by IHS), with the latter focusing on household listing and field management procedures.
Lack of sufficient time for practicing fieldwork procedures (due to time pressures on completing field work during the dry season) led to a decision to have all teams start work in one area (the District of Liquica) and to cover the first 10 clusters (200 households) under close supervision. This was followed up by a one-day classroom session to review problems encountered and to provide reinforcement of training where necessary. This worked well. The close supervision meant that quality could still be controlled, while interviewers, supervisors etc. were able to gain confidence in their abilities that could serve them well when they were then sent out to undertake further fieldwork on their own.
Remaining fieldwork was done in two overlapping stages. The first stage included all remaining sample clusters outside Dili, with teams having varying workloads ranging from around 10 to 19 clusters depending on lines of communication (road networks) and the nature of terrain. The second stage was Dili, with workloads divided so that teams with fewer clusters outside Dili did more work in the capital. The first teams were able to return, and start fieldwork in Dili during the last week of August (slightly more than two weeks after leaving for the field) and all fieldwork was able to be completed by around 12 September (including Dili), roughly 5 weeks after the teams had initially left for the field.
Insan Hitawasana Sejahtera
Separate questionnaires for households, women and children were adapted directly from the relevant question modules contained in the MICS Manual. All of the modules were utilized with the exception of optional modules dealing with food fortification, child disability and maternal mortality. Following recommendations from UNICEF it was also decided to include additional questions on housing conditions and household assets that could be used to produce a "wealth index" using methods developed by Filmer and Pritchett that had been applied in previous Demographic and Health Surveys, as well as a few MICS, in other countries. Besides the anthropometrical measurement of children (under fives), the Timor-Leste MICS also provided for measurement and weighing of mothers/caregivers aged 15-49 with children under-five years of age in the sampled households.
Draft questionnaires were assembled from the modules and then translated into Indonesian (Bahasa Indonesia), which was the language to be used in the field. They were then translated back into English to check for consistency. Pre-tests were also carried out; an initial pre-test in Jakarta, Indonesia (near to the consultant's offices) to get a rough idea on timing and potential interviewing difficulties, and a larger pre-test in Manatuto, Timor-Leste that included not only tests of the instruments, but also of field procedures (listing, sampling) to be used during the final field operations. Both pre-tests resulted in minor modifications to the instruments. However, major benefits were in reaching decisions on the most appropriate approaches to general field procedures, including household listing and sampling operations.
Data entry was done using a package program, CSPro, which was set up for MICS by IHS staff with advice from BPS. Batches of questionnaires returned from the field were first manually reviewed for overall consistency by advisors and senior Statistics Office staff in Dili and then were subject to double entry (independent entry by two different data entry personnel). Input data sets were matched to uncover any inconsistencies and these were then resubmitted for correction (referring back to the questionnaires) by advisors and senior statistics staff. Once this was completed, the various data sets (household, women and children) were linked and merged (where relevant) for processing using SPSS.
This was a relatively complex and time-consuming process, but was fundamental to ensuring maximum quality. Limitations on numbers of computers available in the Statistics Office in Dili and some limitations in the input program CSPro that had to be dealt with during data input also contributed to the pace of progress.
Office editing operations were completed on 30 September, data entry (double entry) on 3 October, matching operations on 4 October, final editing of inconsistencies on 8 October, and final correction of files 16 October bringing to an official end this phase of survey operations.
Final merging and processing of the raw data files was carried out in the offices of the principal consultants in Jakarta, Indonesia starting in the last week of October 2002. This included final data cleaning operations as well as preparation of tables and other output for the report. Involvement of local staff was accommodated by providing for two persons from the Timor-Leste Statistics Office to work with the consultants in Jakarta during this part of the operations.
The cleaning operation basically involved extensive evaluation of data consistency across questions and sections of the questionnaire (for example consistency in age reporting - particularly for women and children reporting dates of birth) that would not have been picked up during data entry and machine editing of inconsistent responses. This took about 3 weeks to complete.
Once a clean data set had been produced, sample weights for each of the 200 clusters were assigned and creation of tables and other statistical outputs was done using SPSS.
Estimates of Sampling Error
Sampling errors, including the value of the statistic (R), its standard error (SE), the number of weighted and un-weighted cases, the design effect (DEFT), the relative standard error (SE/R), and the 95 percent confidence intervals (R+2SE, R-2SE) for each variable are available in Annex Vii of the report. Separate tables are prepared for the total sample and for the principal domains used for calculations presented in the report.
MICS Programme Manager
Use of the dataset must be acknowledged using a citation which would include:
- the Identification of the Primary Investigator
- the title of the survey (including country, acronym and year of implementation)
- the survey reference number
- the source and date of download of the data files (for datasets obtained on-line)
Disclaimer and copyrights
The user of the data acknowledges that the original collector of the data, the authorized distributor of the data, and the relevant funding agency bear no responsibility for use of the data or for interpretations or inferences based upon such uses.