Child Poverty Stats: June 2025

Child poverty statistics: Year ended June 2025 - technical appendix sets out the methodology used to prepare estimates of Aotearoa New Zealand's child poverty rates for the year ended June 2025. These estimates are derived from Stats NZ's Household Income and Living Survey (HILS): Year ended June 2025. This technical appendix provides detailed information about the design and operation of HILS, as well as investigations into data quality.

This is the first year that child poverty statistics have been produced from HILS. In previous years these were produced using the Household Economic Survey (HES).

Household income and housing cost statistics - HES (Income) - DataInfo+ provides summary information and metadata about HES series.

Household income and housing cost statistics - HILS (Income) - DataInfo+ provides summary information and metadata about HILS series.

Note that rounding has been applied to counts and related statistics in this document as per confidentiality standards.

Child Poverty Reduction Act (2018)
Household Income and Living Survey (HILS)
Output and analysis methodology
Quality of administrative data
Very low income
Updates made to previous release
Data quality
Glossary
Appendices

Very low income

Equivalised household income provides a useful indicator of economic wellbeing. However, there are challenges in measuring the income distribution accurately, particularly at the extremes of the distribution; that is, the people with the very lowest and highest incomes. The number of people in these categories becomes smaller and their circumstances can be much more diverse. The challenge with conducting surveys that measure these extremes is in selecting enough sample to reflect this diversity, achieving good response rates from people who may be harder to find and less willing to participate, and possible misreporting of actual income.

Data from HES has shown there are households with extremely low and even negative income, which places them well below the safety net of income support provided by government benefits, New Zealand Superannuation, and other income support. Some possible reasons for this are:

• households may under-report their incomes in the survey at all income levels, including low-income households
• households may correctly report low levels of income if they have incurred losses in their unincorporated business or have negative returns from other investments
• many households containing older individuals report low incomes but have expenditure levels which suggest the use of assets to maintain a higher standard of living than implied by their incomes alone
• households may include individuals who have only received income for part of the year (for example, recent migrants)
• respondents may be linked to incorrect or incomplete admin income data.

As the Ministry of Social Development has outlined in section on (page 161) of Child Poverty in New Zealand, an unexpectedly high proportion of very low-income households report a material standard of living higher than expected, when compared with other low-income groups.

While a relatively small proportion of the total population tends to fall into this group, including this group in analysis may lead to misleading results, particularly for:

• analysis where the population of interest is the low-income group
• analysis where household income is used as an indicator of material wellbeing and consumption possibilities for households
• income distribution analysis
• reported child poverty rates, as children living in these very low-income groups may not be experiencing poverty.

Child poverty statistics: Year ended June 2021 - technical appendix provides further information about this issue, including work undertaken to understand the characteristics of the very low-income group.

More recent investigation by Stats NZ compared the incomes of those in very low-income households - as calculated in HES - against alternative admin income records in the IDI. This research showed that many of the very low-income cases in the 2018/2019 HES had reported a higher annual income in the 2018 Census and indicated that the HES method of retrieving income is, in some cases, drawing on incomplete admin income data (for example, from tax records).

Stats NZ has added a 'catch-all' household income question to the 2025/2026 HILS, similar to what is used in the census. This will enable us to better understand the financial situation of people identified as being in the very low-income group and explore options to treat very low-income households.

In the meantime, we have decided that no treatment will be applied when producing poverty rates to try and correct for this group of people with very low income. However, users of the data should be aware of this issue when analysing this end of the distribution and may want to apply their own treatment depending on the purpose of their analysis.

Updates made to previous release

The most recent year of HES and HILS will always include the most up-to-date data available. This may mean we need to update previous years to preserve the time series. This decision to update, and re-release, previous estimates will take the child poverty reporting and monitoring framework into account and will be based on the statistical significance and magnitude of the change.

Applying the Stats NZ revisions policy to outputs from the Household Economic Survey provides more detail.

Statistics from the 2023/2024 HES, described in Household income and housing-cost statistics: Year ended June 2025 and Child poverty statistics: Year ended June 2025 have been revised since the previous year and are identified in the release tables with an 'R'.

The revisions include more complete admin data and updated population estimates.

Statistics from 2018/2019 through to 2023/2024 have also been updated as part of a population rebase.

Household Economic Survey population rebase: Year ended June 2019 to 2024 provides further information on the rebase.

Users should note that the rebased estimates published on 12 February 2026 do not include the updated admin data. Therefore, there are three versions of estimates for the year ended June 2024, those published in:

• Child poverty statistics: Year ended June 2024 - published February 2025 (no revisions)
• Household Economic Survey population rebase: Year ended June 2019 to 2024 - published 12 February 2026 (revised due to population rebase only)
• Child poverty statistics: Year ended June 2025 - published 26 February 2026 (revised with up-to-date admin data and population estimates).

Table 1 shows the differences between the primary child poverty measures for each release containing data for 2024.

Updates to administrative data

Admin data is updated on a regular basis in the IDI as more recent and more accurate data for the time period becomes available through the tax system. The originally published estimates for the year ended June 2024 used admin data from the October 2024 refresh of the IDI. The revised estimates for the year ended June 2024 (and the first estimates for the year ended June 2025) used data from the October 2025 IDI refresh.

Using a more recent refresh can lead to changes due to new data being made available, older data being updated, corrections, or changes in the links to the spine (the linking process is redone for each refresh).

Between the October 2024 and October 2025 refresh, the main change that affected the income data extracted for the 2023/2024 HES was related to WFF tax credits - the refresh includes a larger proportion of tax returns and therefore uses less data from the previous tax year .

Updates to population estimates

Population estimates are produced as provisional estimates and then revised each month until they are finalised 16 months later, when it can be confirmed that migrants have met the criteria of having been in (or out of) Aotearoa New Zealand for 12 months or more. The data used for the 2023/2024 HES population benchmarks at the time of first publication (February 2025) was provisional but had been finalised by the time of the 2024/2025 HILS release.

The final population benchmarks for 2024 have been used to revise statistics for the year ended June 2024.

Dwelling and household estimates: December 2025 quarter includes the finalised population estimates.

Impact of updates

Methodology tables 14 and 15, in the Excel file available under Download data, show the combined impact of all the data updates to the child poverty estimates and household incomes, respectively, for the year ended June 2024. These did not change the overall trend and all revisions were within sample error on the original estimates.

The magnitude of the revisions to low-income poverty measures was between a reduction of 0.1 percentage points and an increase of 0.5 percentage points.

Data quality

As with any release, we aim to report statistics that are as reliable as possible and be transparent about challenges faced during data collection and measurement that impact data quality.

Survey collection for the 2024/2025 HILS performed as designed, unlike in the HES collection years 2019/2020 through 2022/2023 where collection challenges meant sample size reductions and lower response rates. Nevertheless, in addition to standard validation activities, we continued to assess data quality using the Total Survey Error (TSE) framework, to allow for comparisons of quality over time, and to ensure the 2024/2025 HILS dataset can be used to produce high-quality, representative statistics. We assessed three types of statistical error:

1. Validity error
   • Is the entire reference period adequately represented in the survey period?
2. Measurement error
   • Is there evidence of measurement error in concepts of interest due to the use of multiple collection modes?
3. Non-response error
   • Did the 2024/2025 HILS represent the demographic variation of Aotearoa New Zealand?

For each error type we determined the presence and magnitude of errors and compared against previous years. The analysis and results for each of the three error types are briefly summarised below.

Analysis and results

Validity error

We investigated whether the statistics produced in 2024/2025 adequately represented the full reference period, rather than only part of the period as in earlier years that had disrupted data collection.

Figure 6 shows the proportion of households interviewed by collection month, for each year since 2021/2022. In 2021/2022, there were months where no interviews occurred due to COVID-19 restrictions. However, the subsequent years demonstrate consistent proportions across the interview months, meaning that for each of the collections the full two-year reference period was well represented.

Figure 6. Percentage of interviewed households by collection month, year ended June 2022-2025

Measurement error

The personal questionnaire in the 2024/2025 HILS survey, which is completed by all responding members of a household aged 15 years and over, included an item for our data collection specialists to indicate whether the questionnaire was primarily conducted over the phone or in person via computer-assisted in-person interviewing. It showed that overall, 27 percent of completed interviews were carried out over the phone, though the rate of phone interviewing differed by demographics such as region and age group. This compares with the 2023/2024 HES survey, where 42 percent were phone interviews.

Measurement error is, nevertheless, challenging to assess without being able to compare how the same respondents would respond - and how the interviewer hears and records responses - in each interview mode. We therefore used more indirect means to understand measurement error.

First, we examined imputation rates by mode, finding no systematic differences for the key concepts of interest and demographic variables. We then conducted internal consistency checks.

Internal consistency checks

We compared the relationship between income and material wellbeing for the 2024/2025 HILS by mode. Differences might suggest a change in respondent behaviour depending on the interview mode.

Figure 7 shows the relationship between the mean MH-18 score and HEDI-BHC decile for the 15 years and over population. The downward trend indicates that people in higher income deciles have higher material wellbeing, and this pattern is consistent across interview modes.

Figure 7. Weighted mean MH-18 score by HEDI-BHC decile, by interview mode, for people aged 15 and over, year ended June 2025

Non-response error

Each survey year, non-response error is thoroughly investigated among the three error types.

With an achieved sample size of 17,892 households and high response rate, it was expected that the degree of non-response would be smaller than in previous years with smaller sample sizes and disrupted collection. Since the collection exceeded the designed sample size, the high-level key demographics, such as age, ethnicity, and disability, should be well represented.

Response rates

Response rates provide a measure of how successful we were in collecting responses from our sample. They are an indicator of potential non-response error - they measure the magnitude of non-response, but not the degree to which non-respondents differ from respondents on key concepts.

The overall response rate for the 2024/2025 HILS was 82 percent, higher than the previous year's 78 percent, and similar to the 2018/2019 and 2019/2020 collection years.

Figure 8 shows that regional response rates were generally consistent with the previous year, with notable increases in Auckland and Taranaki.

Figure 9 indicates that the response rates by NZ Deprivation Index decile remained similar, or slightly improved compared to 2023/2024, including for higher deprivation areas.

Figure 8. Response rate by regional council area, year ended June 2024 and 2025

Figure 9. Response rate by NZDep2018 Index level, year ended June 2024 and 2025

Technical enquiries

Household Financial Statistics
[email protected]