Urban Land Dashboard

Technical notes

Calculations, data sources, caveats for the rural urban differential (RUD), price to cost ratio (PCR), rental price index (RPI), rental affordability index, building consents per 1000 people and mortgage interest rates.

Rural urban differential (RUD)

Calculations

There are two steps in calculating RUD. First defining a cluster and second inferring an urban-rural boundary (fringe) based on land zoning.

Clusters

The unit of property for the Land Value Model (LVM) is a "cluster". This is a collection of District Valuation Roll (DVR) properties (like, rating assessments) and titles which are connected to the same parcels.

For example, a multi-unit dwelling might be held on cross-lease titles, and each unit will be its own DVR property, but they will all sit on a single parcel. These will combine to form a single cluster. Another common case is for shared driveways, where many houses (each their own DVR property and titles) have a share in a driveway parcel. All of these houses will be joined as a single cluster, to ensure that the land area/value of the driveway parcel is not counted multiple times. This will have the side-effect of averaging out the land value across these properties, but this is acceptable for the LVM.

Some records are excluded for cluster creation:

• Non-ownership titles (Gazette Notice, Life Estate, Timeshare, Records Embodied in the Register, Supplementary Record Sheet)
• Non-land titles (titles for minerals, coal, timber rights, which are recorded as "Freehold" titles)
• Non-land parcels (water bodies, rights, residual parcels)
• Road parcels and parcels with unknown intents

Clusters are created using DVR records for the last valuation period before the snapshot date (30 September in the snapshot year), and combined until no record (title, parcel, DVR) is in more than one cluster. Clusters are time-specific, and change over time. For example, A house can be subdivided into multiple units with their own DVR records, but remain linked to the same parcel.

Capital and land values for each cluster uses the first record for a given valuation period. Subsequent changes for that valuation period are ignored.

Attributes for each cluster is a "best-guess" summary of the underlying records. For example, if there are three DVR properties in a cluster, and two of them have a land-zoning code of "residential", then the cluster is considered to be "residential".

Land-zoning inferred boundary (LZIB)

The LZIB is an algorithmically generated boundary that determines which clusters are included in the rural/urban buffers. This is not an official administrative boundary, but reflects land zoning as defined in the DVR records.

Residential, commercial or industrial properties are defined as "urban". Lifestyle or rural properties are "rural" and all other codes are treated as "other_lz". Māori land is treated as its own category, and this supersedes urban or rural designations. This is the simple lz_state.

The LZIB is generated from "urban" and "rural" clusters (like Māori land and "other_lz" are excluded entirely). Clusters are also excluded if they are too small (<50 square metre), too big (>1000 square kilometres) or too odd (usually narrow strips of accessway, berms or likely to be erroneous matches).

The urban or rural clusters are combined to create a set of boundaries:

• combine clusters of the same type within 50m of each other into boundaries
• remove small notches and holes (<0.5km squared) from these boundaries
• "Snip" boundary necks that are <20m wide.

Urban boundaries will only be included in the RUD if:

• They are >5km squared and have more than 100 DVR properties... ("Urban areas")
• ...or are >0.5km squared and have more than 20 DVR properties and are within 2km of an urban area defined above ("Satellite urban areas")
• ...or are within 200m of an "urban area" or "satellite urban area" defined above ("Nearby outliers")

Rural boundaries will only be included in the RUD if:

• They are >5km squared ... ("Rural areas")
• ...or are >0.5km squared and are within 200m of a "rural area" defined above ("Disconnected rural areas")

Some areas will be neither rural or urban (for example, Māori land or "other_lz") or be mismatched (for example, Urban cluster in a rural boundary, due to the boundary creation algorithm or administrative errors in the underlying record). These are considered "outliers", and excluded from analysis. This is recorded in the lzib_state attribute.

Finally, we calculate the distance of each eligible urban cluster to the nearest eligible rural cluster and vice versa. This is the dist_to_lzib, which is used to determine whether a cluster is within the buffer_dist. For example, an eligible urban cluster is "within 1000m of the LZIB" if its dist_to_lzib is <1000 (Such as it is within 1000m of the nearest eligible rural cluster).

A cluster can be on the edge of the LZIB but not be within the buffer. For example, the Auckland CBD is on the edge of the LZIB since the LZIB ends at the sea, but since there are no valid rural boundaries nearby, it will not have a dist_to_lzib within the threshold. The same mechanism applies for large parks or reserves, and the small-boundary exclusion means that the rural boundaries will have to be sufficiently large before it can "flip" the area.

All LZIBs are time-specific, since land-zoning changes over time and therefore the LZIB changes over time.

Rural urban differential estimation

RUD calculation runs are defined using a selection of LZIB urban boundaries and a specific snapshot date. For example, "For the urban boundaries in the Porirua Territorial Authority on 30-09-2022, what is the RUD between the clusters 1km within the boundary (i.e. Urban) and outside the boundary (i.e. Rural)?" Note that this will capture all rural clusters within 1km, even if the clusters are in a different Territorial Authority (TA).

Urban boundaries for multiple territorial authorities (for example Wellington City + Porirua + Hutt City + Upper Hutt City + Kapiti Coast) can be combined in a single analysis. Note that this will create issues where the TAs have significantly different valuation dates, and care should be taken for interpretation.

Linear regression models are applied to the lv_per_sqm and parcel_per_sqm values in the rural and urban buffer (clusters on either side of the inferred boundary within a given distance), accounting for known property attributes (distance to centre, distance to water, mean_slope, median income).

The rural and urban lv_per_sqm coefficients are used to derive a modelled rural and urban difference in land values.

A further adjustment is made by combining the parcel_per_sqm modelled difference with the estimated development costs of the property. Development costs are estimated using a default value ($95,000 per housing equivalent unit (HUE) in 2010 dollars).(external link) This value is then CPI-adjusted for the snapshot date and further adjusted by the modelled parcel_per_sqm to account for large rural parcel sizes. The adjusted development costs is subtracted from the modelled difference in land value.

Data sources

• District Valuation Roll records (land value and land zoning)
• Census 2013/18 median incomes
• Land Information New Zealand (LINZ) title or parcel data
• LINZ coastline, river, lakes maps
• LINZ digital elevation model (2012 8m resolution)

Caveats

RUD for an area is estimated when a new round of valuations becomes available for that council area, which means the RUD values for a Territorial Authority could be lagged by up to 3 years.

Small changes to land zoning can push boundaries over/under the LZIB algorithm’s thresholds. Maps and summary tables are generated as part of calculation, and these should be used to assess whether there have been significant changes to the LZIB boundaries which may impact the final RUD results.

There are data quality issues in the underlying DVR data (specifically, in DVR-title matching tables). The current strategy is to remove poor quality matches from the data, but efforts are being made to improve these matches to increase data completeness. This may slightly change the results in the future.

On RUD charts we have flags on the quality of data points. ‘Incomplete data’ which means more than 50% of the data is from previous revision periods. ‘Irregular period’ which means the revision period is not 3 years from the previous one.

Price to cost ratio (PCR)

Calculations

The PCR is a ratio of the new build construction cost per square metre and the sale price of new builds per square metre. “New builds” are defined as dwellings sold within 5 years of its estimated build date. One benefit of excluding older buildings is that their prices often do not reflect current building costs. By including relatively new houses in this estimate, it is gives a better comparison of house prices to the cost of building, which gives a clearer picture of how responsive housing supply is to demand.

Construction costs are calculated using the building consent value plus an additional 30%, in which 25% is construction cost buffer and 5% for real estate fees and other costs of buying a home. A 25 percent buffer is used based on advice from quantity surveyors and industry experts (Ministry for the Environment, 2017, National policy statement of Urban Development Capacity(external link))

PCRs for all territorial authorities and by typology can be found in the data download.

Data sources

• District Valuation Roll sales records
• Stats NZ building consents data

Building consents data

Only counts for new consents (not consents for alterations) are used. StatsNZ dwelling typologies are mapped as:

• Houses: "Houses"
• Apartments: "Apartments"
• Units: "Townhouses, flats, units, and other dwellings"
• Total residential: "Total residential"

Sales data

The following dwelling typologies are used based on DVR property codes associated with sales

• Houses: Residential dwellings (RD)
• Apartments: Apartments (RA)
• Units: Ownership home units (i.e. Flats) (RF)
• Other: Other residential (R* not RD/RA/RF)
• Total residential: Sum of all types above

Sales are only included where:

• Sale type is S11 (sale of a whole rating unit that is a freehold title which is an arm's length sale at market price and matched to a DVR record, see p71 of LINZ Ratings Valuation Rules 2008)
• DVR dwelling type information is available
• Sales took place no more than 5 years after it was built
• Sales decade is the same (or one-off) the decade of the built-year
• Sales not classified as "vacant lot"
• Sale price is between $50,000-$10,000,000
• Floor area is between 40-2000 square metres

Caveats

A degree of caution is needed when comparing the price-cost ratio (PCR) across cities, because differences may reflect underlying local conditions rather than supply responsiveness alone. Context matters. For example, geographic constraints—such as coastlines, mountains, or limited developable land—can naturally elevate land values and, in turn, PCRs, even where planning systems function relatively well.

In addition, variations in urban form, infrastructure provision, regulatory settings, and demand pressures (e.g., population growth or income levels) can all influence the relationship between prices and construction costs. As a result, a higher PCR in one city does not automatically imply more severe supply constraints than in another. Comparisons should therefore be interpreted alongside local context and supporting indicators, rather than in isolation.

Rental price index

Calculations

The methodology uses a Fixed-Effects Window-Splice (FEWS) model to track price changes.  Rental prices are sourced from Tenancy Bonds data relating to the private-sector. These are representative of the rental costs of new tenancies (a ‘flow’ concept). . Timeseries use a quality-adjusted rental price index which controls for changes in the ‘quality-mix’ of properties newly rented over time. The index methodology (a property fixed-effects regression estimator) is an internationally recognised approach and consistent with that used for the New Zealand Consumers Price Index, and Rental Price Index released by Stats NZ (see Stats NZ, 2019; Bentley, 2022).

Data sources

The Rent Price Index is determined via weekly rent included from Bond Lodgement data provided via Tenancy Service (Ministry of Business, Innovation, and Employment)

Caveats

Bond data is collected by Tenancy Services from tenants and landlords/property managers. Therefore, the values (and quality) are based on their respective inputs.

Rental affordability index

The change in rental affordability indicator compares changes in rental prices for new tenancies with the growth in median household disposable (after tax) income. 

Calculations

Rental Affordability Index = Change in Income Index (median household income) / Change in Rent Price Index (rent price)

Data sources

Income index is derived from Census median household income, Household Economic Survey, Linked Employee Employer Data, and Tax data (Business employment data) from Stats NZ.

Rent Price Index is calculated using Tenancy Services Bonds data as described in Rental Price Index above.

Caveats

This indicator shows experience of those on median incomes, not individual experiences of affordability. It compares the change of median incomes with overall rent movements. Households not experiencing median income growth will experience different changes in affordability. 

The median household income data for recent periods is estimated using tax data and are revised as further income data becomes available. 

Some of the indicators are affected by seasonal change. For example, rent prices tend to peak in the first quarter of each year. For short term analysis, it’s best to use the same time of year for the start and end of your comparison if you are using indices from data download. 

The indicators track whether affordability is improving or worsening in an area but not how affordable an area is at a point in time. This means it is not possible to compare the level of affordability between areas. Affordability in one area can improve more than another’s over a period of time, while the level of affordability remains worse.  

Building consents per 1000 people

Calculations

A measure of the number of residential building consents, against a proportional population expressed per thousand.

Data sources

Building Consent from Stats NZ.

Caveats

Government consents (e.g. Kāinga Ora builds) are included in these numbers, which affects the assessment of market responsiveness. An updated version of this indicator is currently under development to exclude government consents and help provide a clearer view of the market response to the housing demand.

Mortgage interest rate

Calculations

Two-year fixed special mortgage interest rate published by Reserve Bank of New Zealand.

Data sources

Reserve Bank of New Zealand

Caveats

The interest rates recorded by RBNZ are simple averages of the special advertised (or ‘discounted’) first mortgage interest rates offered by banks to new borrowers for residential property. Special rates or discounts are offered to borrowers who meet specific lending criteria, terms or conditions (e.g. at least 30 percent equity).  May not be available to all borrowers.

Notes

We are continuing to refine the dashboard. We expect to add more indicators to the suite. The current indicators are provisional and may be revised as further feedback is received. Upcoming improvements include an updated Price to cost ratios that excludes Community Housing Provider and Kāinga Ora consents to better assess market responsiveness. 

For feedback, please contact us on HUD.Insights@hud.govt.nz