Household Income and Expenditure Survey 2022

The sampling design of the Tuvalu 2022 HIES consists in the random selection of the appropriate numbers of households (within each strata urban and rural) in order to be able to disaggregate HIES results at the strata level (in addition to National level).
The urban strata of Tuvalu is made of the island of Funafuti (as a whole) and the rest of the country (all outer islands) compose the rural strata. The statistical unit used to run this sampling analysis is the household.
The sample procedure is based on the following steps:

• Assessment of the accuracy of the previous 2015 HIES in terms of per capita total expenditure (variable of interest) and check whether the sample size at that time were appropriate and correctly distributed among both stratas,
• Update this assessment process by using the most recent population count to get the new sample size and distribution,
• Proceed to the random selection of households using this most recent population count.
  The sampling frame (most recent household listing and population count) used to update and select is the 2021 Tuvalu Household Listing conducted by the Central Statistics Division of Tuvalu.
  At the National level, the 2015 Tuvalu HIES reported a good accuracy of the per capita total expenditure (less than 5%) but the disaggregation results by strata showed a lower quality of the result in Tuvalu urban.
  The Tuvalu 2021 household listing provides the most recent distribution of the households across all the islands of Tuvalu. This step consists in updating the accuracy of the previous 2015 HIES by using this recent household count and get the appropriate RSE by changing the sample size.
  For budget constraint, the total sample size cannot get increased, as the funding situation does not allow higher sample size. It means that the only parameter that can be modified is the distribution of the sample across the strata.
  Sample size by stratum:
  -Urban: 350 (out of 1,010 urban households as per the 2021 listing)
  -Rural: 310 (out of 835 rural households as per the 2021 listing)
  -National: 660 (out of 1,845 total households as per the 2021 listing)

2015 per capita mean total expenditure (AUD):
-Urban: 3,190
-Rural: 2,780
-National: 3,000

Relative Standard Error (RSE):
-Urban: 5.1%
-Rural: 4.1%
-National: 3.3%

It results from this new sample design a new distribution that shows an increase in Funafuti urban, mainly due to:

• The low quality of the survey results from the 2015 HIES,
• The number of households that have increased by more than 15% between 2015 and 2020 in Tuvalu urban area.

The household selection process is based on a simple random procedure within each stratum:

• The 350 households in Funafuti are selected using the same probability of selection across all villages of the islands
• The 310 household in rural Tuvalu are distributed proportionally to the size of each rural island of Tuvalu. This proportional allocation of the sample across rural Tuvalu islands generates the best accuracy at the strata level.

Distribution of sample accross strata:
Urban: Funafuti 350
Rural: Nanumea 42
Nanumaga 37
Niutao 46
Nui 39
Vaitupu 75
Nukufetau 45
Nukulaelae 23
Niukalita 4

Non-response is a problem in surveys, and it is crucial that the field teams interview the selected households (the location on the map and the name of the household head are used to help to determine the selected households).
During the first visit, interviewers must do their best to convince the household head to participate in the survey (and get his/her approval to proceed to interview). It may happen in the field that the first visit results in:
I. A refusal: the household head does not show any interest in the survey and is reluctant to participate,
II. The house is empty (household members away at the time of the visit).

(I) Refusal: if the interviewer cannot convince the household head to participate, he has to liaise with the survey management, and the supervisor will help in the discussion to convince the household head to respond. In this case, it is important to mention that all responses are kept confidential and insist on the importance of it for the benefit of Tuvalu population.
(II) Empty house: the interviewer must investigate (checking with neighbours) whether or not the house is still inhabited by the family:
o If it is not the case, the dwelling is then vacant, and the replacement procedure must be activated.
o If the dwelling is still occupied, interviewer must come back later the same day or the day after at different time

Only in extreme cases of persistent refusal or empty house (household members away during the time of the collection) the replacement procedure must be activated.
The replacement procedure consists in changing the selected household to the closest neighbour who is available.