Tuvalu Long Form Census and Household Surveys Analysis

PACIFIC REGIONAL CONFERENCE ON HIES METHODS 28 TO 31 MARCH 2023 NADI, FIJI Tuvalu Long Form Census Presenter name: Lae Paleti Presenter title: Senior Statistician Presenter affiliation: Central Statistics Division, Government of Tuvalu PACSTAT project (P169122): Statistical Innovation and Capacity Building in the Pacific Island Countries

Contents Tuvalu population and history of census and HIES Challenges and constraints to conducting household surveys in Tuvalu The Tuvalu LFC project Study protocol: questionnaire design, field operations, methodological constraints (seasonality, census delay, de facto v de jure population, other)

Tuvalu population and history of household surveys 1,845 occupied dwellings (2021) 11,143 population (2021) 60% in Funafuti, 12% in Vaitupu and the rest distributed across another 6 inhabited atolls Census: eight conducted prior to the year 2000 Census cycle: 2002, 2012, 2017 (mini), 2022 (LFC*) HIES: five conducted in Tuvalu 1994, 2004, 2010, 2015, 2022 (LFC*) Other household surveys 2007 DHS and 2019 MICS Central Statistics Division is the National Statistics Office and has four full time staff

Challenges and constraints to household surveys High demand on data sourced from household surveys with census being fundamental to all population-based statistics (in lieu of a population register) and household surveys supplying fundamental to economic (e.g., household consumption, own account production, labour) and social (e.g., demographic, WASH, education) statistics Demand and disaggregation requirements continues to diversify Small NSO with, in addition to the household survey program, a multisector and multidisciplinary work program including regular statistical collections such as CRVS, trade and CPI, compilation of national accounts, population registers and administrative data, and statistics governance and development Small population spread over vast and remote areas of ocean adding enormous challenges to household surveys, including: Technical: sample design and representativeness, distribution of sample over survey period - in HIES, an even sample over 12-months for each strata is required in order to capture seasonality Logistics: no domestic flights, irregular ferry schedule, remote islands requiring inter-atoll boat transportation Infrastructure: limited internet access, electrification Labour: short labour supply, labour retention is a problem especially for HIES which has a 12-month field schedule Expensive (ref Table), resource intensive in NSO A burden on respondents, which compromises data quality Cost of household surveys in Tuvalu* Survey Tuvalu Long Form Census 2022 household listing 2019 MICS 2017 census (mini) 2015 HIES 2012 census 2010 HIES Total budget Sample (HHs) Cost per interviewed HH Cost per household Per capita $508,505 1,845 $30,016 1,845 #DIV/0! 1,626 $239,828 720 1,761 $125,551 590 $275.61 $275 $16 $0 $0 $129 $0 $68 $46 $3 $0 $0 $22 $0 $11 $16 *Figures are nominal and provisional and not to be cited $0 $333 $0 $213

The Tuvalu Long Form Census project As a result of the constraints to household surveys, Pacific statistics systems are falling behind when it comes to producing the necessary data to measure and monitor development goals, such as those relating to ending poverty and hunger and in achieving decent work and equality In response to this, the Tuvalu CSD, SPC and ADB have implemented an experiment designed to improve efficiency in the collection of the consumption data that are required to measure monetary poverty and undernourishment via integrating a consumption module in the questionnaire of the Tuvalu 2022 Population and Housing census and administer to a preselected sample (ref Table) The aim of the experiment is to collect the data that would normally be collected during an independent census and an independent consumption survey via one integrated Long Form Census . The experiment aims to test the data quality and applications in their own right, and the costs and benefits of an integrated survey. Distribution of non-HIES and HIES selected households to participate in the Tuvalu Long Form Census and time required to interview in each island Non-HIES HHs HIES sample Integrated instrument field plan Non-HIES households to enumerate (census only) House holds Days Days Weeks W1 W2 W3 W4 W5 Nanumea Nanumaga Niutao Nui Vaitupu Nukufetau Funafuti Nukulaelae Niulakita Total 33 36 70 39 121 56 664 4.1 4.5 8.8 4.9 10.1 7.0 13.3 0.0 6.0 58.6 75 60 60 60 75 60 210 60 18.8 15.0 15.0 15.0 12.5 15.0 8.4 15.0 0.0 115 3.8 3.3 4.0 3.3 3.8 3.7 3.6 2.5 1.0 29 33 36 48 39 72 48 300 48 12 672 24 24 22 24 49 6 300 12 24 24 24 24 36 24 64 24 12 24 12 36 24 150 3 12 3 12 60 0 12 0 1,031 660 588 456 306 234 PHC HIES HHs to interview per week (plus institutions) HHs to interview per week

Study protocol: field operations To not disrupt the census, an initial two-visit plan was conceived where the census and HIES interviews were conducted separately during two distinct visits to the dwelling. This approach proposed the questionnaire would include both the census and the HIES as separate sections where, following completion, of the census questionnaire for every household, the dataset would be used to randomly select the households to participate in HIES. The field workers would then conduct visit two to complete the HIES questionnaire for HIES participating households. The two visit approach was discarded because: The questionnaire logic and sequencing didn t make sense Requirement to fully complete the census prior to HIES sample selection Regular internet access on all islands required to synchronise the census data prior to HIES sample selection A one-visit approach was subsequently proposed whereby the census and HIES questionnaire was truly integrated . What this means is that the long-form census instrument disables HIES-specific questions for non-HIES participating households and enables them for HIES participating households. All non-HIES selected households are only asked the census questions, while for HIES participating households, HIES-relevant questions are enabled and the household consumption and income questions are enabled. The one visit approach resulted in a reduction of 56 days of field work (ref Table) Number of days in the field based on a one and two visit approach PHC enumeration days (visit 1) HIES enumeration days (visit 2) Integrated instrument (1 visit) Total two visit plan Island Nanumea Nanumaga Niutao Nui Vaitupu Nukufetau Funafuti Nukulaelae Niulakita Population Households HIES sample size 495 108 384 499 130 494 914 196 531 116 6,477 874 260 43 10097 1688 Enumerators Supervisors 4 4 4 4 6 4 25 4 1 56 75 60 60 60 75 60 210 60 1 1 1 1 2 1 6 1 0 13.5 12.0 16.3 12.4 16.3 14.5 17.5 7.1 18.8 15.0 15.0 15.0 12.5 15.0 8.4 15.0 32.3 27.0 31.3 27.4 28.8 29.5 25.9 22.1 22.9 19.5 23.8 19.9 22.6 22.0 21.7 15.0 6.0 173.3 96 99 57 12 0 6 0 6 660 14 115.6 114.7 230.2 Two visits Integrated

Exclusively census Common to census and HIES Cover: HH identification and geographic information Exclusively HIES Study protocol: questionnaire design Dwelling information: sample replacement information Person roster: usual household members and their location Dwelling information: institutions Person roster: list of household members, sex, age, relationship to household head Individual information: nationality, marital status, functional difficulties Health: acute and chronic illness Person roster: visitors Individual information: migration, biological parents, fertility and mortality Individual information: anthropometric Health: consumption of items in COICOP div.6 Education: consumption of items in COICOP div. 10 Communication and internet: internet consumption , data required to estimate consumption of mobile phone, digital tablets, laptops (COICOP div. 8) Food away from home: consumption of breakfast, lunch, dinner, snacks, soft drinks away from home (COICOP div. 11) Alcohol, tobacco, toddy and kava: consumption of products in (COICOP div. 2) Other individual consumption: clothing and footwear (COICOP div. 3) and other individual consumption (COICOP div. 12) Labour: labour market data required to estimate status of employment (ICSE-18) and income from wages and salaries Household details: data required to estimate consumption of rents and utilities (COICOP div. 4), consumption of telephone, internet, online streaming services (COICOP div. 7), Household assets: data required to estimate consumption of fixed assets(furniture = COICOP div. 5; vehicles = COICOP div. 7) Education: attendance and attainment Common to both census and HIES Communication and internet: internet access, mobile phone use Labour: labour market data required to estimate basic indicators Exclusively census Exclusively HIES Household details: dwelling characteristics, WASH, energy source, household communication access Household assets: ownership of fixed assets Household items and services: other consumption (household maintenance = COICOP div. 5; travel = COICOP div. 9; household services = COICOP div. 4; miscellaneus HH items = COICOP div. 12; legal services = COICOP div. 12) and transfers (financial support, Church donations, remittances) Primary activities: cash income from production; intermediate expenditure Food consumption: volume and value of in- house food consumption (COICOP div. = 1) Meal partakers: non-household members participating in meal events Non-food recall: consumption of miscellaneous goods and services (COICOP div. = 12) Long Form Census Instrument Primary activities: agriculture, livestock, fisheries, handicrafts and home processed food Elective modules: food insecurity experience scale, financial inclusion, cigueratera fish poisining, water tank size, Market survey: unit values for food data processing

Study protocol: other methodological considerations Seasonality Consumption surveys are designed to be implemented over a 12-month period to capture seasonal fluctuations in household consumption patterns. For example, there may be periods of the year where certain seasonal fruit or fish are available, or times where consumption is high or low for one reason or another. Total household consumption is supposed to be representative of a year for applications in CPI, GDP, poverty and food security. Data applications for the Tuvalu Long Form Census will be limited given the field operations occurred over a period of a couple of months De facto v de jure population In Tuvalu, the census aims to estimate the de jure population, which collects information on every person who slept in the dwelling during census night, including visitors and institutions. Consumption surveys, however, collect information on usual residents of the household, including those who are away during the time of the interview. The questionnaire was, as such, designed to be able to produce both de jure and de facto population estimates depending on the user needs.

Thank you Address/post: Pacific Community (SPC), 95 Promenade Roger Laroque, BP D5 98848 Noumea, New Caledonia Phone: +687 262000 | Fax: +687 263818 | Email: complaints@spc.int | Web: https://spc.int/accountability