Current methodology - Building Consents Issued from March 2015

Scope

We only include construction work that requires a building consent in these statistics. Some civil engineering works, such as roads, require resource consents but not building consents, so are excluded.

The Building Act 2004 determines the scope of work requiring a building consent. Its main parts came into force in 2005, replacing the building Act 2001. The 2004 Act introduced measures to provide greater assurances to consumers, such as registration of building consent authorities, and the licensing of building practitioners. The Act was reviewed in 2009. The review broadened the scope of work that may proceed without a building consent.

See Exemptions for changes to the Building Act 2004 and building work that does not require a building consent, on the Ministry of Business Innovation and Employment's website, third edition published 1 March 2014. ISBN:978-0-478-41705-0

The Canterbury Earthquake Recovery Authority had legislative power to undertake work without a building consent. For example demolition work, and temporary repairs.

We exclude consents that are predominantly for demolition work, and consents valued below $5,000.

The building consent value excludes the land value.

Seasonal adjustment additive outlier treatment change from April 2023

We have changed the seasonal adjustment methodology for Building consents issued to automatically detecting unusual data (outliers) for special treatment. This is a more systematic approach for seasonal adjustment across Stats NZ following the impact of COVID-19. This treatment is applied from March 2020 month onwards.

Seasonal adjustment and automatic outliers in time series after COVID-19 describes this treatment.

The seasonally adjusted number of new dwellings series (references BLDM.SFTZ1100A1S and BLDM.STTZ1100A1SPC) published on 30 May 2023 included updates across the whole series due to this change. Under the automatic additive outlier treatment April 2020 was no longer specially treated as an outlier and this, combined with the Easter (trading day) effect, caused updates to the whole time series. The Easter holiday occurs in April most years (including 2019-2023) but sometimes in March. Easter and the number of weekends in a month affect the number of trading days available for building consents to be issued.

The seasonally adjusted number of new houses consented series (references BLDM.SFTZ1110A1S and BLDM.STTZ1110A1SPC) were unaffected as April 2020 month was not specially treated.

Seasonally adjusted series

Seasonal adjustment removes the estimated effect of regular seasonal events, such as summer holidays and pre-Christmas purchasing, from statistical series. This makes figures for adjacent periods more comparable.

The seasonally adjusted series are re-estimated monthly when each new month's data becomes available. Figures are therefore subject to revisions, with the largest changes normally occurring in the latest months.

We use the X-13 ARIMA-SEATS seasonal adjustment program, developed at the US Census Bureau, to produce the seasonally adjusted and trend estimates.

Trend estimates

Trend estimation removes the estimated effect of regular seasonal events and irregular short-term variation from statistical series. This reveals turning points and the underlying direction of movement over time.

The trend series are re-estimated monthly when each new month's data becomes available. Figures are therefore subject to revision, with the largest changes normally occurring in the latest months. Revisions can be large if values are initially treated as outliers but are later found to be part of the underlying trend.

We use the X-13 ARIMA-SEATS seasonal adjustment program to produce the seasonally adjusted and trend estimates. Irregular and short-term variation is removed by smoothing the seasonally adjusted series using optimal weighted moving averages.

To reduce distortions, we estimate the monthly and quarterly trend series for the value of non-residential buildings after removing consent values of $100 million or more from January 2006, and $25 million or more before 2006. However, monthly non-residential building consent values are still volatile with no stable seasonal pattern, and therefore a stable trend for the monthly series is slow to emerge.

Seasonal adjustment at Statistics New Zealand has more information.

Interpreting the data

Values for new buildings include conversion costs. For example, if a hotel is converted to apartments, we treat them as new dwellings in the statistics. Consent values for new buildings sometimes include the cost of demolishing or removing the previous buildings.

Some consents, particularly for large projects, are issued in stages across more than one month. We collect value data at each stage but floor areas and dwelling or building counts are normally recorded at the first large stage. The difference in timing can affect calculations of average prices.

Care should be taken in using building consents data for individual building types at small geographic areas, as it may contain errors and omissions that are not significant at the national level. We may not have detected these errors during our editing processes.

Trading day adjustments

An aim of time series analysis is to identify movements that are due to actual changes. Seasonal adjustment is done to remove systematic calendar-related variation. Specific adjustments can be made to remove variations due to trading day differences, which are not accounted for in a standard seasonal adjustment.

Some of the apparent movement in building consent statistics is due to trading day differences between months. For example, a month with four weekends has more trading or working days than a comparable month with five weekends. This can affect monthly figures, even though there may be no difference in the length of the month or difference in the rate at which consents are issued.

We quantify and remove trading day effects (including Easter) when they are estimated to be statistically significant.

Trend estimates versus month-on-month comparisons

Trend estimates reveal the underlying direction of movement in statistical series. In contrast, comparisons of unadjusted data between one month and the same month in the previous year/s do not take account of data recorded for the intervening months, and are subject to one-off fluctuations. Reasons for fluctuations include changes in legislation, economic variables such as interest rates, and trading day composition of months.

Coding

Until February 2015, all building consents were manually coded.

From March 2015 to August 2018 we used an automated rule-based process to classify building consents.

In the September 2018 release, we moved to a machine learning approach for automatically classifying building consents. We use a generalised linear model to decide how to classify each consent based on available information (eg the presence or absence of specific words and phrases in the consent's job description). The model is trained using manually coded data, and refined over time.

After the building consents data has been automatically coded, we manually check higher value and unusual consents to validate the data.

The variables that are automatically coded are building nature, building type, unit count, institutional sector, and institutional control.

New dwellings consented per 1000 residents

The annual number of new dwellings consented per 1000 residents are available monthly by region, territorial authority, and Auckland local board areas. National dwelling numbers per 1000 residents after December 1991 will use quarterly population estimates, using the latest available quarterly estimates (DPEA.SG2CTOTM). When dwelling numbers are published before the relevant quarterly population estimate, the new dwellings consented per 1000 residents calculation will revert to using the previous quarter's estimate. This will be recalculated with the newest quarterly population estimate once it is available. National dwelling numbers before December 1991 will use annual population estimates, given for each December year (DPEA.SA2CTOTD).

For dwellings consented per 1000 residents at the regional, territorial authority, and Auckland local board level, annual estimates are produced using the appropriate sub-national annual population estimates (as at June). Similar to the national estimates, this will involve using the population estimates with the same reference year as the year-ended annual dwelling counts. Regional estimates are based on the population series DPEA.SFR&, territorial estimates DPEA.SF[IJKLMNOP]&, and Auckland local board areas DPEA.SJS& (where & is the region, territorial authority, or Auckland local board area code).

The survey has 100 percent coverage of the target population so there is no sampling or sample error.

Non-sample errors

Non-sample errors can occur when there is incomplete or incorrect information on consent forms, or when information is incorrectly delivered, interpreted, or classified (including automatically coded building consents). While we make much effort to minimise these errors, they will still occur, and we cannot quantify their effect.

New Zealand