Wellbeing statistics: December 2020 quarter - Data Collection

##Survey content The Wellbeing statistics: December 2020 quarter information release is based on the wellbeing supplement to the Household Labour Force Survey (HLFS). An individual in the household completes the HLFS household questionnaire, which collects information about all the usually resident people in that household (eg family relationships and household income). An eligible individual from the household aged 18 years or over is then randomly selected to answer the wellbeing supplement questionnaire (SQ).

The wellbeing supplement questionnaire includes topics on

#Personal wellbeing

• Overall life satisfaction
• Life worthwhile
• Self-rated general health
• Mental wellbeing
• Generalised trust
• Institutional trust
• Family wellbeing
• Loneliness

#Material hardship

• Material hardship measured using the Dep17 index questions (enforced lack of essentials, economising behaviours, restrictions, financial stress and vulnerability)
• Financial wellbeing

#Housing costs and quality

• Rent
• Mortgage and loans
• Dampness and mould
• Heating

##Sample population

The wellbeing supplement adopted the same target and survey population from the HLFS December 2020 quarter.

An individual was eligible for the wellbeing supplement, if

• Aged 18 or over and
• From an 'eligible responding' households and
• Not a ‘short term visitor’ (not living in NZ for 12 months or more)

HLFS non-respondents are automatically considered to be non-respondents in the wellbeing supplement.

##Achieved sample size

There was a total of 12,464 usable records for the wellbeing supplement.

##Target sample rate As the wellbeing SQ was taken from the HLFS as a subsample of households which have respondents 18 years or over, its target sample rate is the same as the HLFS target sample rate, which is 76.0% (See Labour Market Statistics: December 2020 quarter).

##Achieved sample rate (ASR) The achieved sample rate for the wellbeing supplement was 78.0%. This is calculated using the weighted number's of eligible individuals that responded to the wellbeing supplement divided by the total weighted number of individuals sampled within the HLFS. Essentially, this indicates what percentage of the sample responded to the survey.

The calibration to the population benchmarks reduced the impact of any bias. Non-response was partly due to the increased burden of it being a supplement to the HLFS, and partly because proxy responses were not accepted. A proxy response is a response by one member of a household on behalf of another. The benchmarks were generated from the eligible responding population of the HLFS.

##Sample design information

The wellbeing supplement used the HLFS sample design. The sample was selected using a two-stage stratified cluster design. Households are sampled on a statistically representative random basis from rural and urban areas throughout the North and South Islands.

All eligible respondents within the HLFS eligible responding households are eligible for the wellbeing supplement.

Every selected household is enumerated via the HLFS household form, which lists all the household members and various demographics such as age, sex and whether the individual is in-scope.

One eligible person per household is selected for the wellbeing supplement at random.

##New HLFS Sample

Following every census, the HLFS sample is reviewed, and starting in the December 2020 quarter some updates into the sample design have been introduced.

The new sample will be rolled in over eight quarters. Each quarter, one-eighth of the households in the old sample is rotated out and replaced by a set of households in the new sample.

The primary sampling units (PSUs) of the HLFS are small geographic areas contain approximately 70-100 households. Over time these PSUs can change in size due to population movements, and therefore need to be recreated periodically. This quarter will be the first to be selected on newly created PSUs, which were last recreated in 2014.

The sample is selected from these new PSUs to ensure it is representative across different regions, urban areas, labour force participation rates and NZ Deprivation rates.

Every quarter we monitor the quality of the sample. This quarter extra monitoring was done to ensure that these changes to the sample were not driving unexpected movements in labour force outcomes. These updates do not include changes to the selection criteria or population of interest, nor to the HLFS methodology.

##Weights

The wellbeing supplement has person level weights attached.

##Reliability of survey estimates Data with high sampling errors should be used with caution. Estimates with RSEs between 50 and 100 percent are considered unreliable for most uses and are flagged with double asterisks **. Estimates with RSEs over 100 are also provided and are flagged with triple asterisks ***. They are deemed to not be useful.

The absolute sampling errors are available in the published tables.

Two types of error are possible in estimates based on a sample survey: sampling error and non-sampling error. Sampling error can be measured and quantifies the variability that occurs by chance because a sample rather than an entire population is surveyed. Non-sampling errors are all errors that are not sampling errors. These errors are not quantifiable and include unintentional mistakes by respondents, variation in the respondent's and interviewer's interpretation of the questions asked, and errors in recording and coding data. We endeavour to minimise the impact of these errors by applying best survey practices and monitoring known indicators (ie non-response).

We estimate sampling errors using a jack-knife method, which is based on the variation between estimates and on taking 100 mutually exclusive subsamples from the whole sample. Sampling errors are quoted at the 95 percent confidence level. For example, if the estimated proportion of the population is 20 percent, and the estimate is subject to a sampling error of plus or minus 2 percentage points (measured at the 95 percent confidence level), that shows there is a 95 percent chance the true proportion of the population lies between 18 and 22 percent.

High-level checks of the ethnic groups indicate the samples are broadly representative of the population. However, as the size of the sampled group decreases, the relative sampling errors (RSEs – sample error as a percentage of the estimate) will generally increase. For example, estimates for Pacific Peoples would have a larger RSE than those for New Zealand European.

##Suppressed estimates Some estimates are suppressed (replaced by 'S' in the tables) for reliability reasons. These suppressed estimates have a weighted value of less than 1,000 and reflect a low number of responses which are subject to larger relative sampling errors.