Household Survey on Corruption and Social Assistance 2014

The primary target population is all regular (non-institutional) households in the Kyrgyz Republic. Other target populations are all population oldest 16 years living in non-institutional households. The survey population is identical to the target population; the survey covers all areas within the national borders.

The primary sampling frame will be the list of census enumeration areas (EA) from the Census 2009. There are in all 13 297 enumeration areas in the Republic. Many of the rural EAs are formed around settlements (villages) so the EA coincides with the settlement. Larger settlements contain two or more EAs.

For each EA there is information on total number of people (by sex) and total number of households .There is also administrative information on urban/rural classification, municipality, district and oblast as well as a sketch map of the area. The census maps are kept at the municipality offices. There have been no changes in boundaries between administrative units (municipality, district, and region) since the census, but 50 EAs have been reclassified from urban to rural. The frame has been updated accordingly. The Kyrgyz project employs a stratified two stage design.

The first stage of sampling entails sampling of areas. In the second stage a sample of households are selected in each selected area. The census enumeration areas serve as first stage sampling units - Primary Sampling Units (PSU). Within selected PSUs a sample of households is drawn.

The survey domains are the eight administrative regions plus Bishkek City. It was decided to stratify the sample on survey domain by urban/rural area and Bishkek City. There are therefore 3 strata in all. It was discussed if a deeper stratification- further stratification within region - would give further gains in precision of the estimates. It was concluded that the gains would be small so no further stratification was done. There is, however, an implicit geographical stratification within each province. This is achieved by a geographical ordering of the PSUs and systematic sampling of PSUs in the first stage of sampling.

Number of Sample Households per Cluster

The project had a cluster size of 12households. This is a rather small cluster size compared to what is used in many other surveys focusing on demographic indicators. With a cluster size of 10 households altogether 108 PSUs would be needed to achieve a sample of 1080 households. A larger cluster size than 10 would mean that fewer PSUs than 108are needed. Consequently, the field work costs per household would be lower. On the other hand, the standard errors would be larger. To find out the theoretically "optimum" cluster size - the cluster size that gives the best precision per unit of cost - detailed data on field work costs would be needed. These data are not at hand but some crude calculations could still be done using "guesstimates" on average time for transports between PSUs, time for listing of households in sample PSUs and time for interview per household.
Based on the calculations described above it was decided to increase the cluster size from 10to 12 households. This size may well differ from the theoretical optimum value but it is suitable from a practical field work point of view. The total sample size is1080 households. So, the number of PSUs will be 1080/12= 90.

Sampling of PSUs

The PSUs are selected by systematic PPS sampling (PPS= sampling with probabilities proportionate to size). The PSUs are ordered in geographical sequence within the stratum before the selection of PSUs is done. The size measures are the number of households in the PSU from the Census 2009.
It was discussed within the sampling group whether there was a need for adjusting the size measures of some PSUs. This should be done in areas where it is known that substantial changes have taken place since the census (e.g. new large scale housing projects or clearance of squatter areas). The conclusion of the discussion was that updating of size measures should not be done. The opinion was that there should be rather few cases of radical population changes and also, that it will be difficult/costly to obtain updated information at the PSU level.
The sampling of PSUs will be done in Excel by a standard procedure which is used for all household surveys.

Sampling of Households

For each selected PSU, the starting points will be defined by regional supervisors. Big cities are divided into several territorial units, these units placed in the program Excel, and then the program will randomly select starting points.
In the rural area starting point is also determined by supervisors. The interviewer reaches selected from the sample village and then interviewer describes the layout of the streets supervisor and administrative buildings (the number of streets, crossing the street, the number of administrative buildings). This all is entered in the program Excel and then the program will randomly select starting point for this village. The interviewer has no right to select the starting point, or change it.

Starting from the given address/point, an interviewer will follow strict rules to select a household and a respondent within selected household. The random route method using the right-hand rule is used with the predefined interval of three to select the household (counting each third household, excluding the starting point). Never move on the left side! In the deadlock - Interviewers cover only the right side of the street. In apartment building interviewers begin to move from the top floor down in a clockwise direction, also adhere to step 3.

Each third household is considered as main household, where up to three contacts must be attempted at different times of the day, days of the week, and the weekend within the fieldwork period to conduct a successful interview. In areas where the interviewer will not be able to return on a different day, the interviewer will make attempts with at least a two-hour gap between each attempt before substituting the household.

If the interviewer cannot obtain an interview at the main sample household, the interviewer selects the household to the immediate right of the main household as the first substitute. In the event that the attempt at the substitute household also fails, then the interviewer selects the house immediately to the left of the initial/main household as the second substitute. In the event that an interviewer fails to obtain an interview at all three households, the interviewer selects another main household continuing with the same interval and numbering sequence of questionnaires can be saved.