This study replicates the World Health Orgnization (WHO) multi-country study on Women’s Health and Domestic Violence against Women. The study was designed to:
-Estimate the prevalence of physical, sexual and emotional violence against women, with particular emphasis on violence by intimate partners;
-Assess the association of partner violence with a range of health outcomes;
-Identify factors that may either protect or put women at risk of partner violence;
-Document the strategies and services that women use to cope with violence by an intimate partner; and
-Assess the association of partner violence withabuse against children.
Kind of Data
Sample survey data [ssd]
Version 01: Cleaned, labelled and de-identified version of the Master file.
Women’s Health and Domestic Violence against Women
National coverage through 5 strata:
- three for Gilbert Islands,
- one for Line and Phoenix Islands,
- one for South Tarawa.
Unit of Analysis
Households and Individuals.
In each selected household only one woman was randomly selected to be interviewed for the survey among all eligible women 15–49 years of age.
The survey sample design was developed by the National Statistics Office with technical assistance from the Secretariat of the Pacific Community (SPC). This design was based on a systematic sample of 1500 interviews, 750 on South Tarawa (urban) and 750 from the outer islands. It was noted that due to the sensitivity of the survey the number of non-response cases may be particularly high.
To adjust for a possible reduction of the actual sample size due to non-response, the sample size was inflated by 25%. With this adjustment, the total sample size of households to be visited for this survey was 2000. The sample size represents 15% of all households in the Kiribati and 8% of the female population aged 15-49 in Kiribati.
This design strayed from the World Health Organization (WHO) recommended multi-stage strategy, which consists of 3000 interviews: 1500 from the urban area and 1500 from the rural area. The justification for the reduced sample was that Kiribati had a much lower population than most of the places in the world where the WHO method had been used. Also, because the study used a listbased systematic sample, updated lists of households would have to be produced for each selected island and a systematic skip would be needed to produce the required sample.
Despite it being necessary to only generate estimates at the urban/rural level, the outer islands were further stratified so that they could be better represented. The stratification adopted for the 2006 household income and expenditure survey (HIES) was therefore applied and three strata for the Gilbert Islands (northern, central and southern) and one for the Line and Phoenix Islands were selected. South Tarawa was left as the 5th stratum. Because of cost issues only two islands were selected randomly from each stratum, except the Line and Phoenix Islands where only one was selected.
For sampling in South Tarawa, a two-stage sample design was adopted. For the first stage, systematic sampling was adopted and a predetermined number of 77 enumeration areas were specified to achieve the sample size of 1000 households. The households were then allocated proportionately to each enumeration area based on the overall number of households per area. A systematic skip was then run through the updated list of households for each selected enumeration area to produce the final sample.
The sample for the outer islands (also 1000 households) was proportionally allocated to each stratum based on the number of 15-49-year-old females counted during the 2005 census. Within each stratum, the sample was then proportionally
allocated to each island once again based on the number of 15-49-year-old females counted during the 2005 census. On each selected island, an updated household listing exercise was performed for all villages. The sample required from each village was determined based on the proportion of households that village contained. A systematic skip was then run through the village to achieve the final sample. In all, 1010 households were finally selected from the outer islands.
The target population was women aged 15-49. In each selected household only one woman was randomly selected among all eligible women to be interviewed. WHO ethical guidelines stipulate that there should be a maximum sampling density
of 25% in each cluster (island); that is, no more than one in four households on an island should be sampled. This in order to promote confidentiality and ensure that the nature of the survey (i.e. that it asks about violence against women) does not spread around the island too quickly, as this could put the safety of both the respondent and the interviewer at risk and reduce the likelihood of open and honest reporting by women. This density was maintained in most places except for a couple of enumeration areas in South Tarawa.
There was no significant variation in response rates between South Tarawa and the outer islands. However, the outer islands had a higher percentage of households without eligible women. This is likely because in Kiribati, as in many other Pacific Island countries and territories, many women aged 15–49 migrate to urban areas for employment and education opportunities, often leaving only younger children and older people in the villages. Overall, 1769 women completed the questionnaire and the non-response rate did not exceed the 25% by which the sample was inflated by to account for possible refusals. This means that the size of the sampled exceeded the size needed to be nationally representative. Also, given the high individual response rate any possible participation bias is likely to be low.
The derivation of the person weights took into consideration two key elements:
-The probability of selection of the females who participated in the survey
-The best estimate of females in scope for the survey for each island.
For (a), the probability of selection was based on the various stages of selection, which included:
a) Probability of the EA being selected
b) Probability of the household being selected
c) Probability of the female being selected
d) Probability of the island being selected
e) Probability of the household being selected
f) Probability of the female being selected
For (b), a best guess estimate of the number of females in scope of the survey was derived using information from the survey only. These estimates where then compared to estimates using population projections derived from counts from the 2000 and 2005 censuses. Given there were significant differences between the two counts, it was decided to use the population projection figures to adjust weights to more appropriately reflect the total number of females in scope of the survey for Kiribati. The justification for this is because it was considered that the estimate of total number of females in scope coming from the survey would be more likely to contain errors because:
-Households tend not to account for all members of a household as rigorously in a sample survey, as opposed to a census.
-There was more likelihood of inter-island travel taking place at the time of the FHSS, as opposed to when the two most recent censuses took place.
It is anticipated that the weights for households will only be used in the production of table that produces estimates of the “household size”, “sex of household head” and “socio-economic status” by region and whether or not the interview was completed.
Dates of Data Collection (YYYY/MM/DD)
Mode of data collection
During the training, six field teams of different sizes (proportional to the sample size of the area they had to cover) were formed. Each team had one supervisor/counselor, one field editor and one to three interviewers.
One team was assigned to each stratum except South Tarawa, which had two teams, and enumerators conducted interviews under the guidance of a supervisor.
Mechanism for quality control by supervisors:
-close supervision of each interviewer during fieldwork, including having the supervisor observe the beginning of a proportion of the interviews;
-random checks of some households by the supervisor, without warning, during which respondents were interviewed by the supervisor using a brief questionnaire to verify that the respondent had been selected in accordance with the established procedures and to assess the respondent’s perceptions of the initial interview.
Type of Research Instrument
The study questionnaire was based on the WHO multi-country study questionnaire, version 10, which was the outcome of a long process of international discussion and consultation.
The questionnaire was adapted to the Kiribati context through a stakeholder workshop with the regional and national team. However, an attempt was made to keep the changes to a minimum to ensure that international comparability was maintained.
The Kiribati study questionnaire, developed in English, includes the following 12 sections. The questionnaire replicates the World Health Orgnization (WHO) multi-country study questionnaire, version 10. Section KIR10 was added specifically for the Kiribati study to investigate potential emotional, physical and sexual abuse against the respondent’s children by her partner/s.
Below are the different sections of the questionnaire and the area they each cover:
1. Community data: community information, community social capital, geographic proximity between the residence of the interviewee and her relatives, her membership in local groups, and her demographic data.
2. General health: interviewee’s mental and physical health during the previous month and health-related lifestyle practices such as smoking
3. Reproductive health: interviewee’s history of pregnancy, miscarriage, contraceptive use, and male partners’ shared responsibility in family planning practices and condom use
4. Children: interviewee’s children, the time when she was pregnant and after delivering, and the children’s behaviours
5. Current or most recent partner: interviewee’s partner and his lifestyle (e.g. drug use and alcohol consumption, employment status and type)
6. Attitudes towards gender roles
7. Experience of violence: for ever-partnered women – relationships and experience with intimate partner violence, e.g., sexual, physical, and psychological, violence during pregnancy, and types and frequency of violence perpetrated by intimate partners
8. Physical injuries and treatment sought, or why no treatment was sought
9. Factors and situations preceding the violence by intimate partners, the consequences of violence, women’s coping strategies, and leaving the home
KIR10. Partner’s treatment of children
10A. Experience of physical, sexual or childhood sexual abuse by non-partners
11. Financial autonomy of respondent, possession of property, and ability to use household resources
12. Completion of interview and opportunity for anonymous reporting of childhood sexual abuse.
Sections 5, 7, 8 and 9 were administered to women ever or currently married or with a current regular partner (these women were considered ‘ever partnered’ in this study). They were not administered to women who had never been in a
relationship. Sections 8 and 9 were only for those who reported physical and/or sexual violence in Section 7. Sections 4 and KIR10 were only for women with children. The time required for each questionnaire interview was 30–90 minutes,
depending on the participant’s experience being in a relationship, intimate partner violence and violence during childhood.
Questionnaires in every batch were counted and checked once the supervisor received them.
Ministry of Internal and Social Affairs
Government of Kiribati
Data processing activities involved manual and automatic processes with a direct impact on the quality of the data. Listed below are the main procedures involved in the process:
1. reception and verification of questionnaires
2. data entry (double entry and verification)
3. secondary editing
4. recoding new variables
The data processing system was developed using CSPro 3.3 and was designed to run in a networkbased environment. The system included double data entry, data verification, data editing and tabulation. The data processing supervisor was responsible for implementing all procedures listed above.
In secondary editing a programme checked the structure of the questionnaire, validating individual data items and checking and testing consistency between items. This programme was run on every batch once secondary entry was completed and no more differences were found between the data sets. When data entry was completed, all batches were combined into a single data file and the programme was re-run on the combined data file to make sure that all errors were fixed and the data were ready to generate the final tabulation.
Every batch was entered twice, allowing 100% verification. The two data files were compared and the supervisors fixed any differences found. The data entered in the first round was used to run the secondary edits while the data entered in the second round was stored as raw data.
Ministry of Internal and Social Affairs
Government of Kiribati
"Kiribati National Statistics Office, Family Health and Support Study 2008 (FHSS 2008), Version 01 of the licensed dataset (November 2008), provided by the Microdata Library. https://microdata.pacificdata.org/index.php/home"
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.