Urban Informal Settlements Survey 2016

The survey collected data from a total of 600 urban slum households in the Dhaka City Corporation - 10 slum households from 57 medium and large size slum communities, and 5 slum households from a total of 6 small size slum communities. The sampling frame for this study came from the 2014 BBS Census of Slums and Floating Population. Urban slums or slum communities were classified into three different strata - small size slums (5-10 households); medium size slums (11-200 households); and large size slums (more than 200 households). These three strata were not all used as domains of study but rather defined based on field logistics operations - each stratum followed a slightly different field operation strategy to account for the fact that finding, listing, and interviewing households in small slums posed very different challenges for the field implementation than interviewing households in medium or large slums.

The Primary Sampling Units (PSU) in the survey were the slum communities. There were a total of 3,360 slum communities with more than 5 households in the Dhaka City Corporation. PSUs were equally allocated between strata 1 and 2 (small and medium size slum communities) and stratum 3 (large slum communities) using a combination of PPS and practical allocation. More specifically, PSUs were allocated across strata using PPS with the number of slum households used as the measure of size and rounding to account for the pre-determined cluster size of 10 slum households per PSU for medium size slum communities and 5 slum households per PSU for small slum communities. Using this rule we got an allocation of PSUs which very closely resembled the distribution of slum households across the three strata. Table 1 in the sampling document provided under Related Materials tab reports the number of slum communities and households in the BBS 2014 Census of Slums and Floating Population and the UIS sample by stratum. In the second sampling stage, the list of households in selected slum communities was updated as part of the field work. Using this updated list of households, 5 or 10 households were selected from each slum community using systematic equal probability sampling.

As mentioned previously, the selection of slum communities and households followed slightly different field implementation strategies across strata in three dimensions - creation of replacement slum communities, listing exercise, and selection of households. In terms of the creation of replacement slum communities, the first stage sampling included a ratio of 2:1 replacement slum communities for stratum 1 or small size slum communities (i.e. 2 replacement slum communities for each slum community that needed to be selected). For stratum 2 (medium size slum communities), the first stage sampling included a ratio of 1:2 replacement slum communities (i.e. 1 replacement slum community for each 2 slum communities that needed to be selected). For stratum 3 (large size slum communities), the first stage sampling included a ratio of 1:3 replacement slum communities (i.e. 1 replacement slum community for each 3 slum communities that needed to be selected).

The listing and selection of households in the final stage also followed different field protocols across strata. For stratum 1 (small size slum communities), all households were listed. If the selected slum community had only 5 households, all households were interviewed. If the selected slum community had more than 5 households, systematic equal probability sampling was used based on the updated list to select 5 of them that would be selected to be interviewed. For stratum 2 (medium size slum communities), all households were listed and 10 were selected to be interviewed using systematic equal probability sampling based on the updated list. Lastly, for stratum 3, slum communities were segmented into listing blocks of around 200-250 slum households each. Then one of the segments was randomly selected from each slum community to conduct the full listing exercise. 10 households were selected from the updated list to be interviewed using systematic equal probability sampling.

The survey was fielded over a period of 16 days, using 30 teams of two interviewers each. Each team was responsible for visiting 2 slum communities during this period and administer 10 household surveys in each community. Slum communities were allocated into teams based on practical considerations (e.g. distance of slum communities). During the first two days, interviewers conducted the listing and did the sampling using the field protocols described above for each of the stratum. In the next 14 days, each enumerator visited 5 households each day for a total of 7 times to collect 2-days recall consumption data. At the end of the 14-days period, each enumerator was expected to have completed 10 full questionnaires (5 from each of assigned slum community).