Sample Design

3.1.1 Sample Selection for Facilities

To cover the major sources of public and private outpatient health care and school types, we defined six strata of facilities to survey:

• Government health centers and subcenters (puskesmas, puskesmas pembantu) • Private clinics and practitioners including doctors, midwives, nurses, and paramedics (klinik,

praktek umum, perawat, bidan, paramedis, mantri) 16

• Community health posts (posyandu) • Community health posts for the elderly (posyandu lancia) • Traditional health practitioners • Elementary schools (SD) • Junior high schools (SMP) • Senior high schools (SMU) / Senior vocational high schools (SMK)

IFLS4 used the same protocol for selecting facilities as in earlier waves. We wanted the specific schools and health providers for detailed interviews to reflect facilities available to the communities from which household respondents were drawn. Rather than selecting facilities based solely on information from the village leader or on proximity to the community center, we sampled schools and health care providers from information provided by household respondents. We followed the strategy first used in IFLS3, to track households that moved to or near the EA (in the same village/ kecamatan) during the main field work period, rather than after main fieldwork was over. This enabled us to add facilities to the sample frame from locally- tracked households. This strategy was adopted since it was felt that the tracked household information would cover facilities in the EA.

Health Facility Sampling Frame. For each EA, we compiled a list of facilities in each health facility stratum from household responses about the names and locations of facilities the respondent knew about. Specifically, we drew on responses from book 1, module PP of the household survey, which asked (typically) the female household head if she knew of health facilities of various types, such as government health centers. The names and locations provided were added to the sampling frame.

16 Because of time and money constraints, IFLS2 and IFLS3 did not interview traditional practitioners, as did IFLS1. In IFLS4 we added them back in part because there were indications that they had become more important in recent

years. And whereas IFLS1 grouped doctors and clinics in a different stratum from midwives, nurses, and paramedics, those strata were combined in IFLS2 and IFLS3 because of the difficulty of categorizing practitioners correctly. An advantage of grouping all private practitioners in one stratum is that the mix of provider types interviewed within the stratum better reflects what is available in the community. For example, in communities where paramedics were more plentiful than doctors, the mix of interviewed providers reflects that fact.

Household respondents did not need to have actually used a health facility for it to be relevant to the facility sample. Though someone in the household may well have used a facility that was mentioned, any facility known to the respondent was relevant. Requiring actual use of a facility was rejected because it was judged that that approach would yield a more limited picture of community health care options (since use of health care is sporadic) and possibly be biased because the sample would then be choice-based.

School Sampling Frame. Names of candidate schools were obtained from household responses to book K, module AR, in which (typically) the household head verified the name and location of all schools currently attended by household members under age 25. Therefore, unlike the health facility sampling frame, each school in the candidate list had at least one member of an IFLS household attending.

Final Samples. Not all identified health facilities and schools were eligible for interview. A facility was excluded if it had already been interviewed in another EA, if it was more than 45 minutes away by motorcycle. The facilities that were located in another area were eligible for interview so long it was in our reachable area (about 45 minutes away by motorcycle). We set a quota of facilities to be interviewed in each stratum in each EA. The goal was to obtain, for each stratum, data on multiple facilities per community. The quotas were different for different strata. For example, a larger quota was set for private practitioners than for health centers because Indonesian communities tend to have more private practitioners than health centers.

Stratum Quota per EA Government Health centers and subcenters

Private clinics and practitioners

Community health posts

Community health posts for the elderly

Traditional practitioners

Community informants

ADAT book

Elementary schools

Junior high schools

Senior high schools

Two forms were used in developing the facility sample for each stratum. Sample Listing Form I (SDI) provided space to tally household responses and ascertain which facilities met the criteria for interview and were not duplicates of each other. Those facilities constituted the sampling frame and were listed on the second form, Sample Listing Form II (SDII), in order of frequency of mention. The final sample consisted of the facility most frequently mentioned plus enough others, randomly selected, to fill the quota 17 for the stratum.

Note that because we sampled randomly from sample frames constructed by householder knowledge of facilities in 2007, we may not necessarily have re-sampled facilities that were sampled in IFLS1, 2 or 3; however many facilities will be the same.

17 In some EAs the pooled household responses did not generate enough facilities to fill the quota. Then, information from the village/township leader or women’s group head was used to supplement the sample frame.

Community Informant and ADAT Sampling Frame. Sampling was also used to identify the informants to be interviewed for the community informant and ADAT books. Six potential informants were listed for the community informant book, out of which up to 2 were chosen randomly. The six were suggested by the community leader (kepala desa or kelurahan), one each in six categories: elementary school principal, religious leader, youth activist, political party activist and business leader. Two random numbers one to six per EA were generated by the RAND programmer before field work and those were used to choose the type of informants for that EA. For the ADAT book, we asked for ADAT leader to be mentioned, and that person was chosen. In communities with no ADAT leader, usually urban, religious leaders who know about ADAT were asked for and if needed elderly knowledgeable about ADAT. The extra issue for the ADAT book was for communities with multiple ethnic groups (ADAT tends to be specific to a group). If there was a group that comprised over 50 percent of the local population, then people from that group were solicited. If however there was no dominant group, we collected two ADAT books, one for each of the largest two ethnic groups.

3.1.2 Response Rates

Table 3.1 shows the number of community-facility respondents and facilities covered in IFLS1, 2, 3 and 4. In all waves we met our interviewing quotas. In IFLS4 over 950 public health clinics and sub-clinics; almost 1,600 private health facilities; over 600 community health posts and 300 health posts for the elderly and over 2,500 schools were interviewed. Table 3.2 shows the number of facilities interviewed in each province, by stratum.

Despite not being intended, a number of the same facilities interviewed in IFLS4 were also interviewed in IFLS3, 2 and 1. This was especially true for public health centers and sub-centers and for schools. For these groups the turnover rate is small and the number available to be sampled per community is also small. The lowest re-interview rate was in private health facilities. This is not surprising since there are numerous private facilities, so the sampling rates are smaller, plus the yearly turnover is larger. The re- interview rate could have been increased by deciding a priori to go back to the same facilities that we visited in the previous waves. However, we judged it important to refresh the sample in 1993 and 1997 to allow for new facilities, since the community-facility survey was intended to portray the current nature of the communities and the facilities in which IFLS households resided. Table 3.3 shows the number of facilities interviewed in IFLS4 for which IFLS1, 2 or 3 data also exist, and the number of new facilities interviewed only in IFLS4. The exception is community health posts (posyandu). No community health post interviewed in IFLS4 has the same ID as its previous IFLS counterparts. That is because both the locations and volunteer staff change over time, so determining whether an IFLS4 post was the same as a post in IFLS1, 2 or 3 is effectively impossible. It is perhaps more appropriate to regard a community health post as an activity rather than a facility. As one can see, many IFLS4 facilities were interviewed in at least one earlier wave, especially for government health clinics, primary and junior high schools.