Poverty Monitoring Survey 1996 - Kyrgyz Republic

Abstract
---------------------------

The main purpose of these surveys is to provide data for the study of multiple aspects of household welfare and behavior, analysis of poverty, and understanding the effect of government policies on households.

Geographic coverage
---------------------------

National coverage

Analysis unit
---------------------------

- Households
- Individuals
- Community

Kind of data
---------------------------

Sample survey data [ssd]

Sampling procedure
---------------------------

In order to expedite the survey process, NATSTATCOM used much of the same sample design and survey instruments as those used for the 1993 Baseline Survey. However, the Fall 1996-1998 KPMS surveys used a new sampling frame based on the Kyrgyz Household Registration System. This system was taken from the Census Posts intended for use by the first National Census of the Kyrgyz Republic. Using this system, NATSTATCOM updated the central household registration files effective January 1, 1996, and the information that was used for the sampling frame was as up to date as possible. The procedures followed in the stratification and identification of Primary Sampling Units (PSUs) were similar for all rounds of the KPMS as discussed below.

Formation of Strata

Initially the country was divided into seven (7) strata defined by oblasts (Oblasts are administrative divisions of the country which in turn are sub-divided in to Rayons) and by residence location (i.e. urban vs. rural) within oblasts. The rural portion of Bishkek oblast was combined with the rural portion of neighboring Chui oblast for stratification purposes as Bishkek has practically no rural population.

Selection of PSUs and Households

A total of 198 PSUs were identified for the whole of the Kyrgyz Republic of which 138 were in urban and 60 were in rural areas. The total number of households in the Kyrgyz Republic, as of January 1996, was about 1.1 million of which about 442,000 were classified as urban. It was initially targeted to select clusters of 6 responding households from each urban PSU and 20 responding households from each rural PSU (which would give us a total of 138*6 + 60*20=2,028 sample households). It was initially assumed that a 90 percent response rate would be attainable (though given the higher response rates obtained in the prior surveys, it could even be higher). The overall adjusted sampling rate was set at 1/500. It was then concluded that this overall sampling rate, combined with the projected response rate of somewhat above 0.90 would yield a sample size of close to 2,000 respondent households.

Once the strata and PSUs were formed and identified, selection of sample PSUs and households was then carried out in the following order:

1) Selection of large and small towns12
[Note: For the 1998 KPMS, large towns were defined as those with a population size of 41,125 or larger. Small towns are those with population less than 41,125. This number, according to a NATSTATCOM document was calculated as follows: n=4.7*350*25. This calculation was based on an estimated household size of 4.7, an estimated interval rate of 350 and an average work load per interviewer of 25 households. No further information is available regarding the bases of such an assumption. At the moment, we do not have information about the cut off number that separates large towns from small ones for the other two KPMS.]

2) Selection of Census Posts in urban areas

3) Selection of Ayil Kenshes (village authorities) and population points in rural areas, and

4) Selection of households from selected Census Posts and Ayil Kenshes.
In the rural stratum of each oblast, villages were used as the listing units and within these listing units, equal probability sampling methods were used to select the ultimate sampling units (households).
In urban areas, the centralized computer listings from various sources of household registration were used for the selection of households. These lists are categorized into four:
Type 1 - Private house resident households listed by BTIs
Type 2 - Public house residents listed with other organizations with dormitories only
Type 3 - Public and private households listed by JSKs
Type 4 - Public and private households listed by all other organizations.
In some cases, private households were included in the last three public categories (Types 2, 3 and 4). However, only public households were selected from these types since it was believed that any private households listed in these category types were also included in the Type 1 category. The counts for Type 2, 3, and 4 lists were then adjusted based on the oblast estimates of all urban households.13 Prior to actual household sample selection, lists from types 2 to 4 were updated and adjusted to remove private households, so that any potential double eligibility was eliminated. Urban strata were then formed within each oblast based on type of household listing. In most cases, types had to be combined to form strata of a reasonable size.

Within the limits of rounding and requiring at least one sampling unit per stratum, the allocation of sampling units to urban strata was proportional to the number of households projected for that stratum after allowing for removal of duplicates (private households appearing on a BTI and other lists).

As for rural households, selection of urban households was done using systematic random sampling within each stratum except that more subdividing of urban lists was required before selecting the final list sample that defines each sampling unit.

Even though the list sources were identified and sampled using data as of January 1, 1996 (and using projections of unduplicated counts in some cases), the final listings were updated in the field just prior to the survey period. Therefore, the sample households in selected areas were drawn from the most current available listings.

Mode of data collection
---------------------------

Face-to-face [f2f]

Research instrument
---------------------------

The KPMS surveys were carried out using a household questionnaire and a community (population point) questionnaire. The household questionnaires were used to collect demographic information on the composition of the household, housing, household consumption including home production, as well as economic activities in agricultural and non-agricultural sectors. For each household member, individual level data on health, education, migration and labor was collected using the household questionnaires. Community questionnaires were used to collect price data and the presence of social services and infrastructure in the community (population point) where the sampled household is located.

The household questionnaire was extensive and required several hours of intense interviewing to gather all that was needed from each household and its embers. The household questionnaire was split into two parts. The first part was used to collect data through a face to face interview on household roster, dwelling, education, health, migration, etc. At the end of the first part, members who shop for food for the whole household and those who know most about income, expenditure and savings of other household members were identified and designated as respondents for the next part (second round). The second round of interview was administered two weeks after the first half and collected data on crops, food and animal products produced by the household, food expenditure and home produced food consumption.

Some sections of the household questionnaire such as those that deal with dwelling and expenditure information were administered to the person most knowledgeable of the family's overall expenditures, income and other finances as well as about the family's business activities and employment. In other sections, each adult in each sample household was interviewed individually. The information gathered from each household included extensive data on education, health, employment, migration, reproduction and reproductive health (for women aged 15 to 49), land use, expenditure, revenue and other financial matters, as well as anthropometric measurements (for children 5 years and younger). Information about children under 14 years of age was collected by asking the relevant questions to the adult household member who is primarily responsible for each child's care.

The community (Population Point) questionnaires were administered to each sample cluster. They were used to collect data on prices of goods and services, distance to schools, shopping and medical facilities, types of housing, commercial and private land use and availability of infrastructure.

HOUSEHOLD QUESTIONNAIRE

The KPMS household questionnaires generally contain 15 major sections, and each of these sections covers a separate aspect of household activity. In some cases, the section has sub-sections. These household questionnaires were designed to better assess the changing environment brought about by the advent of a market economy and to enable a more in depth analysis of topics such as housing, health, and education. The various sections of the KPMS household questionnaire are described below.The household questionnaires administered in the KPMS surveys are more or less similar with minor modifications and additions in the successive rounds of the KPMS.

POPULATION POINT QUESTIONNAIRE

The community (population point) questionnaire was used to collect information and data that are relevant to the community/population point where the household is located. The questionnaire was designed to be administered in the geographical area of each sample cluster. It was used to collect data regarding prices of goods and services in the local area and data on community infrastructure. Respondents to these questionnaires are those believed to be well informed members of the community that the interviewers identified by going to the rayon, city, oblast administration or other governmental agency located in the population point6. The questionnaire also contains sections to be administered to retail outlets in the neighborhoods that sell various products such as food, drinks, tobacco products and fuel. Other data collected using the population point questionnaire includes distance to schools, distance to shopping and to medical facilities, commercial and private land use in the community, availability of electricity, water, communication and other infrastructure. Similar population point questionnaires were used in all KPMS. The population point questionnaires were completed by the field supervisors. The population point questionnaire contains nine (9) major sections

Response rate
---------------------------

Total Sample Households Selected: 2,193
Minus households found to be vacant: - 128
Minus households found to be demolished or uninhabitable: - 18
Minus households found to be used for commercial purposes: - 4
Minus households found to be ineligible for other reasons: - 8
Total Sample Households Eligible for Interview: 2,035
Minus households that refused to be interviewed (2.7%): - 56
Minus households that were unable to be contacted (1.0%): - 20
Minus households that did not respond for other reasons (0.4%): - 8
Total Households That Completed an Interview (95.9%): 1,951

Response rates were examined during the 1996 KPMS period to see if differential response rates for different strata might justify unequal weighting to reduce any potential bias during data analysis. the differences in response rates among strata were small, as the rates range from 0.905 to 1.000. The overall response rate was high (0.959) and, as expected, the response rate for rural strata was slightly greater than the one for urban strata. Under these circumstances, weight adjustment for nonresponse appears optional. For most analyses, it was believed that it would have little impact and would unnecessarily complicate the analysis. No such weighting adjustment was therefore done for all the KPMS surveys.

Data appraisal
---------------------------

There are no significant data quality problems, but the following deserve mentioning.

1) In discussions with NATSTATCOM staff, it was learned that the household registration that was used in the sampling process may not actually cover all resident persons. As more migration occurs, some residents may either be homeless or occupy housing units not included in the household registration system. It was pointed out that trends in household registration coverage need to be monitored in the future. If this becomes a serious problem, a move toward strict area probability sampling might be the only alternative that would provide near complete household population coverage.

ii) A review of the sample selection process was conducted after the survey by a senior statistician on site in Bishkek. According to the review, field sampling steps were completed according to plan, but problems were encountered in four clusters with classifying households into the four types:
Type 1 - Private house resident households listed by BTIs
Type 2 - Public house residents listed with other organizations with dormitories only
Type 3 - Public and private households listed by JSKs
Type 4 - Public and private households listed by all other organizations.
As a result, too many households were selected and interviewed in these clusters. To insure appropriate level of representation in the sample from these clusters, only a subsample of the interviews from these clusters were retained for the final data file.