Feasibility of innovative tools and methods to improve household surveys in complex urban settings: Multiple methods analysis of the Surveys for Urban Equity (SUE) study in Kathmandu, Dhaka, and Hanoi

* Corresponding author 1. Department of Social Statistics, University of Southampton, Southampton UK 2. Health Research and Social Development Forum-International, Kathmandu, Nepal 3. Nuffield Centre for International Health and Development, University of Leeds, Leeds UK 4. Hanoi University of Public Health, Hanoi, Vietnam 5. Centre for Injury Prevention and Research – Bangladesh, Dhaka Bangladesh 6. Advancement through Research and Knowledge Foundation, Dhaka Bangladesh 7. School of Medicine, The University of Notre Dame Australia, Freemantle WA Australia Preprints (www.preprints.org) | NOT PEER-REVIEWED | Posted: 28 October 2019 doi:10.20944/preprints201910.0320.v1


SUMMARY BOX
What is already known?
• Researchers and practitioners increasingly raise concerns about exclusion of vulnerable and mobile urban populations from LMIC household surveys on conceptual grounds: sample frames are usually outdated; typical two-stage designs require a long time gap between household listing (final sample frame) and interviews; and paper-based field tools and protocols developed 40 years ago are not well-suited to modern complex urban settings.• LMIC urban settings pose numerous challenges to survey fieldwork including atypical housing arrangements, large numbers of migrant workers, rapid expansion of new -often informal -dwellings, and high mobility of residents.• The challenges of conducting surveys in LMIC cities are only going to worsen as urbanization, population mobility, and socio-economic disparities increase, particularly in African and Asian cities.What are the new findings?
• This study quantifies rates of exclusion among vulnerable and mobile sub-populations in Kathmandu Valley, and areas of Dhaka and Hanoi where these populations concentrate.• We describe and evaluate innovative survey methods that might improve accuracy of household surveys in LMIC cities, including evaluation of feasibility.What do the new findings imply?
• Alternative sample frames, such as gridded population estimates, are a viable alternative to outdated or inaccurate census sample frames.New types of sample frames can enable new survey designs, such as one-stage sampling, which improve coverage of vulnerable and mobile urban populations in surveys.• Until urban areas can be stratified by deprived / not-deprived areas, or some other area classification that reflects urban disparities, household surveys are unlikely to accurately sample tent and shack dwellers in slum-like areas.

INTRODUCTION
In low-and middle-income countries (LMICs), household survey methods have remained consistent for over forty years.However, in the same period, population trends and available technologies to measure populations have changed substantially.Continued use of census sample frames, two-stage sampling, and paper-based mapping and listing in LMICs has likely led to exclusion of vulnerable and mobile urban populations, 1 particularly in Africa and Asia. 2 A global shift to urban living is dramatically changing the structure and nature of communities and households, 3 and survey methods must change in response.
The largest survey programmes in LMICs include the Demographic and Health Surveys (DHS), Multiple Indicator Cluster Surveys (MICS), and Living Standard Measurement Surveys (LSMS), which essentially use the same sample frames, survey designs, and implementation methods. 4Crucially, surveys are used to measure progress against one-fourth of the Sustainable Development Goal (SDG) indicators. 5If current survey methods systematically under-represent vulnerable and mobile urban populations, our understanding of progress towards the SDGs is fundamentally flawed.
7][8] The mapping-listing protocols influence which households are listed, and eligibility criteria and interviewer interpretations influence who is recorded as a household member. 9e DHS and MICS define household members as: (i) usual residents or slept in the dwelling (living space) the previous night, (ii) share living arrangements, and (iii)share food. 6,7The LSMS defines household membership as: (i) slept in the dwelling three or more of the last 12 months and (ii) share food. 8By all definitions, households in both residential and commercial buildings should be included, [6][7][8] guards and servants are subsumed into the household of their employment, [6][7][8] and seasonal and migrant populations are usually intentionally excluded. 10nversely, unintentional exclusion of vulnerable and mobile populations occurs in at least three ways.First, as a result of outdated EA sample frames.In LMICs, the urban population grew roughly 30 percent between 2005 and 2015. 2 Yet, since 2000, the average DHS sample frame was 7 years old with many exceeding 10 years. 11Second, two-stage sample designs require a gap of several months between the mapping-listing and interview activities, resulting in exclusion of recently settled households.a Third, exclusion can result from poorly-defined or difficult to operationalize mappinglisting protocols, for example assuming that one household occupies each dwelling. 12 address problems of unintentional exclusion of vulnerable and mobile households in surveys, the Surveys for Urban Equity (SUE) project piloted and evaluated three survey innovations in Kathmandu, Dhaka and Hanoi: (1) use of modelled gridded population data as a sample frame, (2) one-stage sample design, and (3) mapper-lister protocols including a script, OpenStreetMap and OpenDataKit tools, and a broadened household definition.Here, we present results of the pilot including the extent to which populations were unintentionally excluded from a standard survey design.Further, we evaluate the feasibility, cost and skills required to implement our novel methods in complex urban settings.
a In most surveys, if a household moves away and is replaced, then the new household is interviewed.If the household is not replaced, then members will be recorded as non-responders and be accounted for in the sample weights.

METHODS
We compared the ability of three survey innovations to identify different types of households and individuals than standard surveys.To establish feasibility of the innovations, we recorded costs and team skills required and conducted focus group discussions (FGDs) to explore enumerator experiences.

Setting
We selected Kathmandu Nepal, Dhaka Bangladesh and Hanoi Vietnam, as they typify different points on the urbanisation trajectory.The pace of growth in South Asia has particularly strained urban housing markets increasing the number of people living in atypical arrangements and locations. 2hile some poorer households live in informal settlements, others live in economically heterogeneous neighbourhoods. 2In Kathmandu and Dhaka, for example, it is common for the building owner to occupy the top floor, rent the middle floor to a middle-class family, and rent the bottom floor to multiple low-wage workers.In Vietnam, old, cramped buildings continue to house the economically and socially vulnerable, while migrant labourers live in multiple-occupancy inadequate structures near work. 13We sampled the entire Kathmandu Valley, and purposefully chose to survey a slum and an economically mixed ward in Dhaka, and an economically mixed district with a large migrant population in Hanoi.The Hanoi survey occurred soon after a government campaign to evict illegal occupants.

Innovations
We used several innovative datasets, protocols and tools which aimed to improve representation of vulnerable and mobile populations in surveys (Figure 1).SUE survey planning and field manuals are available elsewhere. 14One-stage sample design One-stage sampling means that all households in a cluster are sampled, allowing the household listing and interviews to occur on the same day.One-stage sampling also allowed us to broaden the household definition to include all usual residents or people who slept in the cluster the previous night, including hostel-dwellers, long-term occupants of guesthouses, and street-sleepers.In the questionnaire, we collected information about living arrangements, meals, and length of time at the dwelling to identify individuals and households that met DHS/MICS and LSMS definitions.

Sample frame
We used WorldPop gridded population datasets as sample frames rather than older censuses.At the time of planning, the last censuses in Nepal (2011), Bangladesh (2011) and Vietnam (2009) were seven or more years old. 15WorldPop is modelled with a machine-learning approach that disaggregates population counts from administrative areas to approximately 100x100m grid cells based on dozens of spatial covariates derived from satellite imagery and GIS data. 16The small size of grid cells enabled one-stage sampling.

Mapping-listing tools
We replaced standard hand-drawn field maps with geographically accurate maps.Before fieldwork, we updated building, road, and pathway data for each cluster in OpenStreetMap using the iDeditor tool. 17In ArcGIS, we created a map for each cluster showing the OpenStreetMap base layer and cluster boundary. 18In the field, we noted changes on the paper map and updated OpenStreetMap accordingly.Both the household listing and interviews were collected in GeoODK, an OpenDataKitbased application. 19

Study design and protocol
In 2017 and 2018, we conducted three cross-sectional household surveys in Kathmandu, Dhaka and Vietnam. 20The survey in Kathmandu estimated depression and injury prevalence in Kathmandu targeting 1200 households in 60 clusters (sample size calculation 20 ).We used the Global Human Settlement (GHS) layer of 1x1km grid cells to define the city boundary because old municipality boundaries only included the city centre, while new municipality boundaries included rural communities beyond the Kathmandu peri-urban reach. 21We randomized half of the clusters to a one-stage arm and the other half to a two-stage arm to compare survey designs, and drew randomly from 18 (30%) backup clusters when a main cluster lacked residential buildings.The Kathmandu sample was drawn from the 2017 WorldPop dataset 15 using the GridSample R package 22 by selecting 100x100m "seed" cells with PPS and then "growing" clusters to a minimum of 200 households by randomly adding neighbouring cells.In clusters randomized to the one-stage arm, we used the 100x100m "seed" cell as the cluster boundary (Figure 2).The sample sizes in Dhaka and Hanoi were smaller, each targeting 400 households in 20 clusters with 6 (30%) additional backup clusters.The aim of these surveys was to evaluate transferability of our methods and tools across settings.An additional 4 clusters were sampled in Hanoi because more than 6 clusters were dropped.Both the Dhaka and Hanoi surveys used only a one-stage design, and clusters were selected with GridSample from 2020 WorldPop estimates with PPS. 15,22In Dhaka, clusters were each 100x100m cells drawn from one ward and one slum community, while in lessdense Hanoi, clusters were 200x200m cells drawn from one district.
Following standard survey methods, geospatial specialists mapped and listed household in Kathmandu's two-stage sample, while public health specialists conducted interviews later.In Kathmandu and Dhaka's one-stage samples, geospatial experts listed dwellings, but the household listing was performed by interviewers on the day of interview.In Hanoi, mapping, listing, and interviews were wrapped into one activity.The trainings for the mapping-listing(-interview) teams were each one-week and involved lectures, role-play, group discussion and a field test.
In the field, mappers-listers followed a script to approach residents, and upon request, distributed a written description of the survey.In all three surveys, respondents provided written informed consent, were 18+ years of age and usually a senior household member.The interviewers were mostly female, and read questions and recorded responses on a tablet.The household questionnaire collected demographics, assets, income/savings/expenditures, social capital, migration, and injury information.One adult in each household was randomly selected using the Kish method to complete an individual questionnaire with mental health and migration questions. 23ys worked by each staff member and costs were recorded by the survey coordinator in each country.Time spent by survey coordinators to develop and learn the novel methods was excluded from cost calculations.However, time spent training mappers-listers and interviewers was included.
In Kathmandu, we estimated costs for the one-stage and two-stage survey separately by holding constant costs of administration, training, and durable goods, and varying days of fieldwork.

Public involvement
Members of the public, including survey respondents, were not involved in setting the research questions, outcome measures, design, or implementation of the study, nor the dissemination of study results.

Statistical evaluation
Sample weights were calculated separately according to the SUE and DHS/MICS household definitions.In the one-stage samples in all cities, we evaluated whether use of the DHS/MICS household definition resulted in different estimates of individual and household characteristics compared to use of the SUE household definition using means or percentages, and linear or multinomial regression at 5% alpha level.In the Kathmandu sample, we used the same statistical techniques to compare whether individual and household characteristics differed in the one-stage versus two-stage sample; first, holding the DHS/MICS household definition constant, and second, comparing two-stage-DHS/MICS with one-stage-SUE households.Household characteristics included building type, household member configuration, slum household, and migration status of household head.Individual characteristics included age-gender groups, employment status, marital status, and highest level of education.A reference group was selected for each variable to make statistical comparisons, and observations were dropped if they lacked data to determine household definition eligibility.We analysed survey results in Stata 14.0, adjusting for sample weights and clustering of observations.The analysis in Kathmandu was stratified by arm (one-stage/two-stage), and the analysis in Dhaka was stratified by community (ward/slum).This meant that in Kathmandu, the twostage arm represented more weighted households than the one-stage arm because two-stage clusters had larger populations.

Qualitative evaluation
An FGD was held with each of mapping-listing teams using the same guide covering topics of OpenStreetMap enumeration, mapping-listing, and workflow.Additional questions exploring differences in one-stage and two-stage clusters were included in the Kathmandu FGD.FGDs were facilitated and audio recorded by two trained qualitative researchers, and conducted in the local language.The recordings were transcribed into the local language and then translated into English.
We performed a thematic Framework Analysis in NVivo 11, coding every line by theme and summarizing positive/neutral experiences, challenges, and recommendations.

RESULTS
In Kathmandu, 11% of clusters were dropped and replaced.No clusters were dropped in the targeted areas of Dhaka, and 30% were dropped in the Hanoi district where the sample frame was older (Table 1).Due to high density in Dhaka, and larger clusters in Hanoi, nearly all clusters in those cities required segmentation to achieve 20 households per cluster (Table 1).Household response rates were 96.8% in the Kathmandu two-stage arm, 88.3% in the Kathmandu one-stage arm, 98.7% in Dhaka, and 82.7% in Hanoi (Table 1).

Unintentional exclusion due to household definition
Across the one-stage samples, applying the DHS/MICS or LSMS household definition resulted in exclusion of approximately 10% of households compared to the SUE definition (Table 1).When we compared characteristics by household definition in one-stage samples, we found that a substantial portion of individuals from certain sub-groups were excluded by the DHS/MICS definition.In Kathmandu, nearly half (46.9%) of single adult households and sizable portions of migrant-headed households (6.7%), non-married (8.5%), unemployed (10.5%), disabled (9.3%), and studying (14.3%) adults were excluded by the DHS/MICS definition (Table 2).

Unintentional exclusion due to sample design
Applying the DHS/MICS household definition, we compare one-stage and two-stage samples in Kathmandu to understand how sample design might influence types of respondents (Table 3).We found average household size was smaller in the one-stage sample but dwellings had more occupants (household size: 3.5 vs. 3.9, dwelling size: 5.0 vs. 3.9) (Table 3).Further, the one-stage design had more non-family households (6.0% vs. 1.9%), but the two-stage design included more shack and tent dwellers (0.7% vs. 3.8%) (Table 3).

Unintentional exclusion due to sample design and household definition
Building off the previous analysis, we compared the one-stage sample with SUE definition and the two-stage sample with DHS/MICS definition in Kathmandu to understand the combined effects of survey design and household definition.In the one-stage-SUE sample, there were more single adult (10.4% vs. 4.5%) and non-family households (6.0% vs. 1.9%), plus inclusion of hostel dwellers (3.8%), street sleepers (1.0%), and long-term guesthouse residents (0.1%) who did not meet DHS/MICS household definition (Table 3).However, the two-stage-DHS/MICS sample included more shack and tent dwellers (0.6% vs. 3.8%) (Table 3).

Time and cost
In Kathmandu, a one-stage gridded population survey with a target of 600 households in 30

Skill mix
The skills required to plan and implement the SUE surveys were similar to standard household surveys.The main difference was skillset of the mapping-listing team.In a standard survey, mappinglisting staff are required to have a secondary school education. 25To use SUE tools and methods, the mapping-listing staff should additionally have training in geography, GIS, or related fieldwork, and be comfortable using mobile technologies for data collection and navigation.The skillsets of other staff including survey planners, trainers, and interviewers were identical to a standard household survey.At the time of planning, the GridSample R package was the only available tool for gridded population sampling and required intermediate R programming and GIS skills.However, a free point-and-click tool called gridsample.org was since released, allowing non-technical design and implementation of gridded population surveys.

Experiences
Feedback from the mapper-lister FGDs was generally neutral or positive, and staff resoundingly said they would prefer SUE tools and protocols to a conventional paper-based protocol.The SUE survey fieldwork, however, was not without limitations.

Key challenges.
In Kathmandu, the mapping-listing staff were comprised of university geospatial students.Several described approaching residents as their greatest challenge, as well as their greatest reward.One mapper-lister explained, "It was fun to work at the social level and interacting with the local people.We always used to be limited to using the computers before.This time we got to go to the field and visit the people in their homes to collect the data.This was a new experience for me."Other mappers-listers agreed and added that role-play and practical activities prepared them for fieldwork, though additional training on the survey aims would have helped to explain the survey's purpose to residents.In Nepal, mapping-listing staff were able to enumerate 20-30 households per day initially, and this increased to 40-50 households per day after a week.
The main challenges in Dhaka and Hanoi were different.In these cities, the survey planners were trained about field tools and protocols but were not able to accumulate field experience before training field teams.As a result, mapping-listing staff, including the geospatial students in Dhaka, described challenges learning how to use the tablet applications during the first few days of fieldwork.In Hanoi, staff additionally struggled with navigation.Several of the teams in Hanoi even enlisted local guides who often informally took on the role of approaching residents and introducing the survey.
Across cities, mappers-listers described working in pairs to be essential because it provided them with "mutual support" and allowed them to adapt to the moods and reactions of residents, interact in more languages, and to work faster and more accurately by supporting each other with navigation and recording details.Overwhelmingly, mappers-listers recommend that teams be comprised of one geospatial expert who focuses on navigation and mapping, and one public health expert who approaches residents and lists dwelling or household information.
Response rates.In all three cities, mapping-listing staff found that residents seemed to omit mention of neighbours who did not have official mortgages or rental contracts, presumably for fear of evictions or fines.This was a particular challenge in Hanoi where "people tended to answer our question following their household record book," an official registry of households administered by the government.On mapper-lister-interviewer explained, "For residents who were living in evacuated houses, they felt worry and scare as if something wrong could happen." The Hanoi planning team thus decided to hire several guides from the community to accompany teams.Together guides and teams returned to each cluster multiple times to build trust with residents and identify households not reported during previous visits.While the presence of guides likely improved response rates in Hanoi, it also meant that these survey teams were limited by the availability of guides.Most teams performed the listing and interviewed households in the evenings when guides were home from work, though this meant that residents were trying to eat dinner and often rushed to answer questions or refused to participate in the survey.Mapper-listers and interviewer in Kathmandu and Dhaka, however, performed most of their work during the day.
Residential building access was a problem across cities.The Hanoi teams faced secured apartment buildings without a guard.In these situations, the planning team contacted the building management boards and were usually able to gain access to these buildings, however once inside, mappers-listers-interviewers often found that residents knew little about their absent neighbours.
Kathmandu had wealthy "VIP" neighbourhoods, and mapping-listing staff also reported substantial scepticism and non-response in these neighbourhoods.
Travel.Mapping-listing staff commuted to clusters via bus, rickshaw, motorbike, and foot.In Kathmandu, most staff never travelled more than one hour from home to a cluster, however a team working in peri-urban Kathmandu spent three hours commuting one way to one particular cluster due to the absence of buses or taxis.In Dhaka, where traffic is notoriously bad, commute times from the office to clusters ranged from 1.5 to 3 hours.Across the three cities, mapping-listing staff recommended hired vehicles to save time.
One-stage versus two-stage clusters.Mappers-listers in Kathmandu reported different experiences in one-stage and two-stage clusters they visited.The two-stage clusters were, by definition, at least ten times the area of one-stage clusters, however due to variability in the WorldPop estimates, several two-stage clusters were even larger resulting in extra days of work and more physical barriers such as hills or rivers to navigate.In addition to being much larger, the two-stage clusters required much more information than one-stage clusters, resulting in longer interactions and higher levels of scepticism among residents."We had to explain everything to the people from the very beginning in the two-stage [clusters].That is because we required a lot of information in that stage.
On the contrary, we did not require a lot of information in the one-stage [clusters].We had to only ask them whether there was anyone living in those houses or not." Residents were generally willing to report number of apartments/dwellings per building, however, they were reluctant to specify the number of households per dwelling and to give household head names.In many two-stage clusters, Kathmandu teams approached a business owner in the ground level who gave number of dwellings on the above floors, but refused to give household-level information and instead directed the mapping-listing staff to the building owner.One way that mappers-listers addressed this challenge was to approach people at a local grocery store, and start a conversation about their household and neighbouring households away from their building.In this context, residents were less likely to feel they were speaking on behalf of the landlord and were more open.This approach allowed mappers-listers to gather helpful information about neighbouring buildings, as well.
Technology.Across sites, mapping-listing staff faced challenges with the tablet applications.While some challenges could have been averted with more, or better, training, other challenges were inherent to the tools and protocols used.First, although OpenStreetMap was updated by mapperslisters before visiting clusters, the map updates to various applications occurred on different schedules resulting in different versions of the same map in the field.Specifically, updates to ArcGIS (from which paper field maps were printed), GeoODK (where building GPS points were collected as part of the listing), OSMAnd and MAPS.ME (both used for navigation) were all updated 1 to 30 days after a change was made to OpenStreetMap.A mapper-lister in Kathmandu explained this challenge with an example, "When we looked at the OSMAnd, it showed us that there is a house at a location.But when we used to work with the plug-in of the GeoODK, there were no houses there.So, there was problem with the maps being updated in the GeoODK and OSMAnd.The paper maps were also prepared using old satellite images which does not match with the current on-field structures.So, it would be easier for us to work if there was data consistency in the three tools that we used." A second problem was the number of applications that the mapping-listing staff were expected to use.It was time consuming and confusing to switch between MAPS.ME to navigate to a cluster, OSMAnd to navigate within a cluster and visualize its boundary against the tablet location, GeoODK to manually record a GPS point over each building, and finally CamScanner to submit daily reports to the planning team.Despite the multiple applications plus paper map meant to help find each cluster, mappers-listers in all cities reported delays and difficulty navigating to clusters.Once in a cluster, however, mappers-listers did not report any challenges identifying the cluster boundaries, despite their blocky shapes.Mappers-listers across cities also found recording the listing data in GeoODK for by building was arduous, and they often took notes on paper when speaking to residents then entered the information into the tablet immediately after.
Third, the location precision within OSMAnd and GeoODK were poor, often showing a large circle up to 36 metres in which the tablet could be located, obfuscating the purpose of this feature.Location precision was a particular problem in high density areas (presumably with tall buildings blocking or refracting signals), and resulted in more than one instance of a mapping-listing team starting their work, and then realizing part-way through that they were recording data one or two streets away from the cluster.

DISCUSSION
By comparing DHS/MICS and SUE household definitions across one-stage samples, and by comparing a one-stage and two-stage sample in Kathmandu, Nepal, we found evidence that standard household survey methods unintentionally omit single adults and non-family households, both of which are more likely to represent disjoined households, or be mobile compared to stable nuclear family households. 10,26,27To our knowledge, this is among the first studies in a LMIC context to evaluate under-coverage due to survey design and methods in face-to-face surveys; such studies tend to be conducted in high-income countries. 12,28though the same protocols and household definitions were used to identify households in Kathmandu's one-stage and two-stage arms, the quality of the household listing data likely differed because interviewers listed household in one-stage clusters, rather than mappers-listers.Interviewers had more skills to interact with the public and substantially more time at each building (2.5 to 3 hours per interview).As LMIC urban contexts are increasingly defined by complex living arrangements and mobile and temporary residents, there might be a need to move the household listing responsibility to interviewers using one-stage survey designs.Indeed, this argument has been made by others who provide evidence that standard household definitions are no longer suitable in complex LMIC cities, and that individuals and communities are more appropriate units of measurement. 27,29thout urban strata, the two-stage sample in Kathmandu was better able to measure tent and shack dwellers than the one-stage sample, likely due to the larger area of two-stage clusters.The only way to ensure representative survey samples of shack/tent dwellers and other vulnerable populations concentrated in slums is to treat slum/non-slum areas as strata, in both one-stage and two-stage designs. 30 found that response rates in many one-stage clusters were lower than in two-stage clusters, particularly if the household witnessed a neighbour or landlord refusing participation.This may have been due to the greater proportion of vulnerable and mobile households measured in one-stage clusters if they were less willing to participate, more likely absent, or felt disempowered to respond.Readers who are interested in one-stage survey designs should take account of lower response rates and potentially higher design effects due to similarity among immediate neighbours when calculating sample size. 31The surveys conducted in Dhaka and Hanoi focused on vulnerable and mobile communities, so rates of exclusion may be higher than the general population.
Societal changes, particularly rapid urbanization in LMICs, have likely caused decay in survey data accuracy due to increased complexity in living arrangements, urban disparity, and population mobility.Not only are vulnerable and mobile populations at risk of unintentional, unmeasured exclusion from standard household surveys, their data are masked in urban averages when they are sampled.Given the importance of household survey data to policy-making, planning, and monitoring progress toward development goals, it is time to evaluate new survey tools and protocols that ensure inclusion of all households.

Figure 1 .
Figure 1.SUE survey coverage in Kathmandu, Dhaka, and Hanoi, and example two-stage and one-stage cluster field maps

Figure 2 .
Figure 2. SUE survey coverage in Kathmandu, Dhaka, and Hanoi, and example two-stage and one-stage cluster field maps 24

Table 1 .
Sample design, number of clusters, and households (unweighted) by survey clusters would cost approximately US$26,769, or US$45 per household, while a comparable two-stage survey would cost approximately US$35,284, or US$59 per household.One-stage survey costs per household in Dhaka (US$34) and Hanoi (US$76) differed due to cost of living and limited economy of scale due to smaller sample sizes.The main cost difference between Kathmandu's one-stage and two-stage survey was the mapping-listing activity; costs were 2.5 times greater in a two-stage survey due to larger clusters and need for an advanced mapping-listing team.

Table 2 .
Unintentional exclusion due to household definition: Percent of population who would be excluded using the standard DHS/MICS versus SUE household definition in Kathmandu, Dhaka, and Hanoi Households in each one-stage sample were split by those who (a) met the SUE and DHS/MICS household definitions, and (b) met the DHS/MICS household definition only.We present the percent of households excluded from the DHS/MICS household definition, and regression coefficient p-value comparing (a) and (b).with servants and/or extended family, sometimes with non-family household members as well ** defined as lacking improved water, improved sanitation, a durable structure, sufficient sleeping space (based on DHS/MICS household member definition), or insecure tenure † multinomial logistic regression N-wgtweighted count

Table 3 .
Unintentional exclusion due to sample design and household definition: Kathmandu sample characteristics comparing a) two-stage DHS/MICS versus one-stage DHS/MICS, and b) two-stage DHS/MICS versus one-stage SUE with servants and/or extended family, sometimes with non-family household members as well ** defined as lacking improved water, improved sanitation, a durable structure, sufficient sleeping space, or insecure tenure † linear regression coefficient (continuous) or multinomial logistic regression (categorical) N-wgtweighted count

Table 4 .
Comparison of time and budget to perform one-stage versus two-stage survey (estimated) in Kathmandu, Dhaka, and Hanoi