The Thai Demographic and Health Survey (TDHS) was a nationally representative sample survey conducted from March through June 1988 to collect data on fertility, family planning, and child and maternal health. A total of 9,045 households and 6,775 ever-married women aged 15 to 49 were interviewed. Thai Demographic and Health Survey (TDHS) is carried out by the Institute of Population Studies (IPS) of Chulalongkorn University with the financial support from USAID through the Institute for Resource Development (IRD) at Westinghouse. The Institute of Population Studies was responsible for the overall implementation of the survey including sample design, preparation of field work, data collection and processing, and analysis of data. IPS has made available its personnel and office facilities to the project throughout the project duration. It serves as the headquarters for the survey.
The Thai Demographic and Health Survey (TDHS) was undertaken for the main purpose of providing data concerning fertility, family planning and maternal and child health to program managers and policy makers to facilitate their evaluation and planning of programs, and to population and health researchers to assist in their efforts to document and analyze the demographic and health situation. It is intended to provide information both on topics for which comparable data is not available from previous nationally representative surveys as well as to update trends with respect to a number of indicators available from previous surveys, in particular the Longitudinal Study of Social Economic and Demographic Change in 1969-73, the Survey of Fertility in Thailand in 1975, the National Survey of Family Planning Practices, Fertility and Mortality in 1979, and the three Contraceptive Prevalence Surveys in 1978/79, 1981 and 1984.
Kind of Data
Sample survey data
Unit of Analysis
- Women age 15-49
Unit of Analysis
- Women age 15-49
The population covered by the 1987 THADHS is defined as the universe of all women Ever-married women in the reproductive ages (i.e., women 15-49). This covered women in private households on the basis of a de facto coverage definition. Visitors and usual residents who were in the household the night before the first visit or before any subsequent visit during the few days the interviewing team was in the area were eligible. Excluded were the small number of married women aged under 15 and women not present in private households.
Producers and sponsors
Authoring entity/Primary investigators
Institute of Population Studies (IPS)
Institute for Resource Development (IRD)
U.S. Agency for International Development
Division of Nutrition
Ministry of Public Health
United Nations Economic and Social Commission for Asia and the Pacific
SAMPLE SIZE AND ALLOCATION
The objective of the survey was to provide reliable estimates for major domains of the country. This consisted of two overlapping sets of reporting domains: (a) Five regions of the country namely Bangkok, north, northeast, central region (excluding Bangkok), and south; (b) Bangkok versus all provincial urban and all rural areas of the country. These requirements could be met by defining six non-overlapping sampling domains (Bangkok, provincial urban, and rural areas of each of the remaining 4 regions), and allocating approximately equal sample sizes to them. On the basis of past experience, available budget and overall reporting requirement, the target sample size was fixed at 7,000 interviews of ever-married women aged 15-49, expected to be found in around 9,000 households. Table A.I shows the actual number of households as well as eligible women selected and interviewed, by sampling domain (see Table i.I for reporting domains).
THE FRAME AND SAMPLE SELECTION
The frame for selecting the sample for urban areas, was provided by the National Statistical Office of Thailand and by the Ministry of the Interior for rural areas. It consisted of information on population size of various levels of administrative and census units, down to blocks in urban areas and villages in rural areas. The frame also included adequate maps and descriptions to identify these units. The extent to which the data were up-to-date as well as the quality of the data varied somewhat in different parts of the frame. Basically, the multi-stage stratified sampling design involved the following procedure. A specified number of sample areas were selected systematically from geographically/administratively ordered lists with probabilities proportional to the best available measure of size (PPS). Within selected areas (blocks or villages) new lists of households were prepared and systematic samples of households were selected. In principle, the sampling interval for the selection of households from lists was determined so as to yield a self weighting sample of households within each domain. However, in the absence of good measures of population size for all areas, these sampling intervals often required adjustments in the interest of controlling the size of the resulting sample. Variations in selection probabilities introduced due to such adjustment, where required, were compensated for by appropriate weighting of sample cases at the tabulation stage.
The final sample of households was selected from lists prepared in the sample areas. The time interval between household listing and enumeration was generally very short, except to some extent in Bangkok where the listing itself took more time. In principle, the units of listing were the same as the ultimate units of sampling, namely households. However in a small proportion of cases, the former differed from the latter in several respects, identified at the stage of final enumeration:
a) Some units listed actually contained more than one household each
b) Some units were "blanks", that is, were demolished or not found to contain any eligible households at the time of enumeration.
c) Some units were doubtful cases in as much as the household was reported as "not found" by the interviewer, but may in fact have existed.
Equal sample sizes for each domain were originally intended, due to population growth, particularly in urban areas, the number of households selected varied slightly by region. The total number of target households is highest in the central region followed by Bangkok, the north, south and northeast.
In general the response rates of both household and individual interviews in the TDHS were relatively high. For the country as a whole, 96 percent of the selected households were successfully interviewed. The main reason for non-response in the household survey is that either no one at all or no adult was at home. The household response rates vary by region being highest in the northeast (99 percent) and lowest in Bangkok (92 percent). However the total number of households interviewed was greatest in the central region and lowest in the south.
The overall TDHS response rate is 90 percent. As expected Bangkok yielded the lowest success rate while the north and northeast had the highest success rate. The response rate for the eligible woman sample is lower than the household response rate. About 94 percent of eligible women identified were successfully interviewed. The main reasons for non-response in the eligible women survey were that the targeted respondent was not at home and/or refused to be interviewed. Regional differences in the response rates of the individual interviews were similar to the household interviews. The highest response rate for eligible women was in the north (98 percent) and the lowest in Bangkok (87 percent).
The generally high response rates for both household and women interviews were due mainly to the strict enforcement of the rule to revisit the originally selected household if no one was at home initially. No substitution of the originally selected households was allowed. Interviewers were instructed to make at least 3 call-backs if contact with the household or eligible woman had not been made or the interview was incomplete. In many instances revisits were made until the team had moved out of the province.
The survey indicates a low ratio of the number of eligible women per household. On the average there are about 80 eligible women per I00 households interviewed. This is much lower than found in SOFT, conducted in 1975, where the ratio was 96 per 100 households. At least in part this could be attributable to the increasing age at marriage. There is some regional variation in terms of number of eligible women per household. The ratio is highest in the northeast (83 per i00) and lowest in the south (75 per lO0). This lower ratio of number of eligible women per i00 households explains why the total number of eligible women interviewed was lower than the number targeted (6,775 versus 7,000).
Sample cases are weighted for the following reasons:
- to compensate for differences in sampling probabilities,
- to compensate for differences in response rates, and
- to make the regional and urban-rural distribution of the sample correspond to the distribution according to the most recent population projections and evidence available from other, supposedly more reliable, sources.
Each of these is described below in turn.
1. Design weights. These refer to the weights which compensate for differences in selection probability. They are inversely proportional to design probabilities of selection, but can be scaled arbitrarily such that the average weight is 1.0 per case for the sample as a whole. Firstly, these weights differ by sampling domain since domains were sampled at different rates to yield nearly constant sample sizes despite differences in domain size. Secondly, to a lesser extent, sampling rates differed among blocks and villages in cases where it was necessary to introduce this variation to improve control over sample takes, given the inaccuracies in the available measures of size.* Design weights are applied at the level of the block or village identically to all households and individual women in the area.
2. Weights due to differential non-response. Because of generally high response rates, with the exception of Bangkok to some extent, the application of weights to compensate for non-response was in itself not very important. However, since it was already necessary to apply design weights at the level of the block/village, these latter could be easily modified to take into account non-response as well. The adjustment consisted of multiplying the design weights by the inverse of the response rate in the block/village. The final weights were scaled so that the average weight was, again, 1.0 per case.
Since the overall response rate for individual interviews was lower than that for household interviews in the same area, the weights were not exactly the same for the two types of units.
3. Adjustment of regional anH urban-rural distribution of the sample: Finally, the above weights were adjusted to make the sample distribution of the population and of eligible women correspond to the best available "standard" distributions of these at the level of the major reporting domain (region and urban-rural sectors). The external standards were obtained from (a) NESDB projections of the total population for 1987 by region; (b) proportion urban of total population by region from Ministry of Interior registration figures for 1985; (c) the corresponding NESDB projections for the numbers of women aged 15-49; and (d) estimates for proportions ever-married among women aged 15-49 in each domain from the 1984 Survey of Population Change. Multiplication of (a) and (b) provides estimates of the total population distributed simultaneously by region and urban-rural status while multiplication of (b), (c) and (d) provides estimates of the number of ever-married women age 15-49 (i.e. eligible women) distributed simultaneously by region and urban-rural status. The former distribution is used to derive correction factors for the few tabulations that refer to the total population of individuals as enumerated in households while the latter distribution serves as the basis for determining the correction factors to be applied to tabulations referring to information derived from the eligible woman questionnaire.
Both for the total population and for ever-married women aged 15-49, the joint distribution by region and urban-rural status consists of nine mutually exclusive and exhaustive categories: one for Bangkok (which is treated as entirely urban) and one each for the urban and rural sectors of the four remaining regions. Table A.3 of the Final Report compares the distribution of the entire population and of ever-married women aged 15-49 as enumerated in the TDHS sample both before any weighting and after being weighted for sample design and non-response with the corresponding standard distribution. Tables A.4a and A.4b in the Final Report illustrate the derivation of the standard distributions for the total population and the population of ever-married women aged 15-49 respectively. The ratio of the proportion in each of the nine categories in the appropriate standard distribution to the corresponding proportion in the distribution of the sample population after weighting for sample design and non-response represents the multiplication factors to be applied to obtain the final weights. Adjusting the weights in this manner ensures that the regional and urban-rural distribution of the weighted sample agrees with the external standard. This adjustment has no effect on the survey results for the individual sampling domains when taken separately except in the case of the provincial urban domain, in which case it ensures that the regional distribution of the weighted provincial urban sample agrees with the external standard. Moreover, when results are presented by region, it ensures that the urban-rural distribution of the weighted results within a region corresponds to the external standard.
There is a substantial difference both in the cases of the total sample population and the eligible women population between the unweighted distribution and the distribution after weighting for design and non-response. This is as expected based on the nature of the sample design. There are also some differences, however, between the latter substantially larger than official estimates and projections indicate. However, no firm conclusion can be drawn concerning this based on a sample of the scale of the TDHS. Therefore it is appropriate to accept the standard estimates as a basis for adjusting the sample weights.
Dates of Data Collection (YYYY/MM/DD)
Mode of data collection
All supervisors and assistants were instructed to closely observe and supervise the interviewers particularly during the first few days of the fieldwork. This procedure was enforced strictly so that any misunderstanding in the questionnaires and errors made could be detected and corrected at an early stage. The field director also visited the teams to help with any problems each team had as well as to deliver any supplies each team needed and bring back completed questionnaires.
Completed questionnaires were submitted to the supervisor or assistant immediately following interview. The questionnaires were edited in the field to the extent feasible. If possible, inconsistencies and errors were clarified and corrected and re-interviews on the questions for which answers were omitted or inconsistent were made.
The task load of supervisors and assistants was very heavy in the fieldwork. They were responsible not only for the overall management of the team, which included making all contacts, assigning the households to the interviewers, editing the questionnaires, and planning daily work, but they were also assigned to do the anthropometric measurements and the community survey including the visits to the health and family planning service outlets. In retrospect, this workload was excessive. To improve fieldwork quality, it would have been advisable to have had a separate team carry out the time consuming community survey component. One result of the this excessive workload was that it became impossible for the supervisor and assistants to fully edit all the completed questionnaires in a timely manner in the field.
Type of Research Instrument
The DHS core questionnaires (Household, Eligible Women Respondent, and Community) were translated into Thai. A number of modifications were made largely to adapt them for use with an ever- married woman sample and to add a number of questions in areas that are of special interest to the Thai investigators but which were not covered in the standard core. Examples of such modifications included adding marital status and educational attainment to the household schedule, elaboration on questions in the individual questionnaire on educational attainment to take account of changes in the educational system during recent years, elaboration on questions on postnuptial residence, and adaptation of the questionnaire to take into account that only ever-married women are being interviewed rather than all women. More generally, attention was given to the wording of questions in Thai to ensure that the intent of the original English-language version was preserved.
a) Household questionnaire
The household questionnaire was used to list every member of the household who usually lives in the household and as well as visitors who slept in the household the night before the interviewer's visit. Information contained in the household questionnaire are age, sex, marital status, and education for each member (the last two items were asked only to members aged 13 and over). The head of the household or the spouse of the head of the household was the preferred respondent for the household questionnaire. However, if neither was available for interview, any adult member of the household was accepted as the respondent. Information from the household questionnaire was used to identify eligible women for the individual interview. To be eligible, a respondent had to be an ever-married woman aged 15-49 years old who had slept in the household 'the previous night'.
Prior evidence has indicated that when asked about current age, Thais are as likely to report age at next birthday as age at last birthday (the usual demographic definition of age). Since the birth date of each household number was not asked in the household questionnaire, it was not possible to calculate age at last birthday from the birthdate. Therefore a special procedure was followed to ensure that eligible women just under the higher boundary for eligible ages (i.e. 49 years old) were not mistakenly excluded from the eligible woman sample because of an overstated age. Ever-married women whose reported age was between 50-52 years old and who slept in the household the night before birthdate of the woman, it was discovered that these women (or any others being interviewed) were not actually within the eligible age range of 15-49, the interview was terminated and the case disqualified. This attempt recovered 69 eligible women who otherwise would have been missed because their reported age was over 50 years old or over.
b) Individual questionnaire
The questionnaire administered to eligible women was based on the DHS Model A Questionnaire for high contraceptive prevalence countries.
The individual questionnaire has 8 sections:
- Respondent's background
- Health and breastfeeding
- Fertility preference
- Husband's background and woman's work
- Heights and weights of children and mothers
The questionnaire was modified to suit the Thai context. As noted above, several questions were added to the standard DHS core questionnaire not only to meet the interest of IPS researchers hut also because of their relevance to the current demographic situation in Thailand. The supplemental questions are marked with an asterisk in the individual questionnaire. Questions concerning the following items were added in the individual questionnaire:
- Did the respondent ever experience a miscarriage or abortion? If so, how many? Educational attainment and expectations for each of respondent's living children age 6 or above.
- Did the respondent ever use contraception subsequent to marriage and prior to first pregnancy? If so, how long after marriage did she first use contraception?
- Information on whether or not users of oral contraceptives forgot to take the pill any time during the last month and if so, how many times.
- Information on the type and timing of first contraceptive method used since last birth including a probe on whether contraceptive use was initiated prior to or subsequent to the return of menses
- The place of the respondent's last delivery.
- Whether the respondent's marriage was registered; whether the marriage was marked by a ceremony.
- Did the couple live with any set of parents following marriage? If so, with whose parents did the couple reside following marriage?
- Does the respondent consider a lower high school education sufficient for young people nowadays?
- Secondary occupation of husband.
- Information on respondent's current work, employment status and type of payment.
- Height and weight of mothers of children 3-36 months of age.
c) Community questionnaire
TDHS community questionnaire was based on the model DHS community questionnaire. Again it was modified to suit the situation in Thailand. The community survey was conducted in all 192 sample clusters (villages) of rural areas but not in urban areas. The community questionnaire focuses on information on village characteristics, accessibility to health and family planning services, and availability to public services nearest to the cluster.
The community was defined according to official administrative boundaries. A group interview was used as the mode of data collection for the community survey. The interview was conducted by the team supervisor. The respondents were a group of community leaders (typically 3-5 persons). Persons qualifying as respondents included current or former village headmen, or their assistants, village health volunteers, village health communicators, members of existing associations (groups) in the village, and other village leaders who have been residing in the community for five years or more. Visits were also made to all government health and family planning service outlets within a 30 kilometer radius from the cluster to collect information from the personnel about services.
Institute of Population Studies
All completed questionnaires have been sent to IPS for office editing. It was originally planned that the team supervisors and some assistants would be retained as office editors and keyers. Unfortunately, most of the temporary team supervisors and assistants left the project at the end of the fieldwork. Therefore, five new editors and keyers had to be hired. These new editors and keyers are graduates from various universities in Thailand with a bachelor degree in social science or a related field. They received intensive training on the content and logic of the questionnaire. To further improve their ability to edit the questionnaires, they conducted interviews with households of the sample clusters that required revisits in Bangkok and the central region.
Office editing of questionnaires was supervised by the field director and two IPS research associates who had also been TDHS team supervisors. The editing was done by the five new editors/keyers, two project assistants, and two permanent research assistants who had also served as team assistants. All the questionnaires were given numbers and sorted by sample cluster number.
The data entry of TDHS started in early July, 1988. The data were directly transferred from the questionnaires to micro-computers, using the ISSA program developed by DHS. Two programmers from DHS were sent to IPS to help set up the ISSA program and train IPS data processing staff on how to work with the program. Office editing and data entry were completed by the first week of January 1988. The tabulations for the preliminary and country report were then prepared with the assistance of the DHS programmer.
Nonsampling error is due to mistakes made in carrying out field activities, such as failure to locate and interview the correct household, errors in the way questions are asked, misunderstanding of the questions on the part of either the interviewer or the respondent, data entry errors, etc. Although efforts were made during the design and implementation of the TDHS to minimize this type of error, nonsampling errors are impossible to avoid and difficult to evaluate statistically.
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
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.