# Auckland housing growth: Audit 20,529 consents, rank by adjusted local yield

Hadley Sims · September 17, 2026

> Audit Auckland’s 20,529 housing consents and assess adjusted local yield, separating zoning capacity from additional permits and completed homes.

| Takeaway | Detail |
| --- | --- |
| Separate permission from production. | The supplied evidence does not identify 22% as an Auckland result. Expanded zoning capacity describes what could be built, not additional permits or completed homes. |
| Measure additional housing, not gross activity. | 46% has no documented metric in the supplied research. An adjusted yield must distinguish reform-induced permits from construction that would have occurred without rezoning. |
| Require a matching land denominator. | 70% is not established as a rezoned-land share in the supplied evidence. Calculating additional permitted homes per hectare requires the actual affected area and a matching observation period. |
| Rank only after validating local estimates. | The evidence supplies no 22% local-yield benchmark. A defensible ranking would require geographically matched consent records, rezoning boundaries, and estimates that account for displacement and spillovers. |

70% has no documented meaning in the supplied Auckland research—a warning against giving an unsupported percentage the authority of a planning benchmark. The headline consent tally presents a related problem: citywide activity cannot identify where the next hectare of upzoning should go. A useful audit must distinguish legal capacity, permitted dwellings, and completed homes before treating any of them as evidence of productive land-use change.

California YIMBY’s summary of Greenaway-McGrevy and Phillips reports strong evidence that Auckland’s upzoning stimulated construction, with an effect that remained statistically significant after accounting for possible spillovers. That supports a consequential claim: reform increased housing activity rather than simply relocating it within the city. It does not establish a neighborhood ranking or supply the matching rezoned-area denominator needed to calculate additional permitted homes per hectare.

From Hadley Sims’s computational-planning perspective, the useful legacy is therefore a testable allocation strategy. Match consent locations to changed zoning, align geographic boundaries and observation periods, and estimate additional activity against a credible baseline. Then compare adjusted local yields while reporting uncertainty. Until those inputs exist, selecting the next hectare remains a research task—not a conclusion hidden inside the headline.

![Auckland hillside suburb with compact timber townhouses beside](https://static.mm-ais.com/article-images-ai/auckland-housing-growth-audit-20-529-con-ai-3f75d57c.jpg)
Auckland hillside suburb with compact timber townhouses beside

## Measure the 2016 Reform’s Yield, Not Its Paper

Auckland's 2016 Unitary Plan trebled the number of dwellings that could be built within its residential areas, according to California YIMBY's 9 August 2022 summary of Greenaway-McGrevy and Phillips. That is a permitted-capacity figure. It is not a yield, and treating it as one is the error that makes every downstream expansion decision look easier than it is.

The Mixed Housing Urban zone shows why envelopes, not zone labels, drive outcomes. Its operative standard is generally an 11-metre building height — typically three storeys, not a tower. But the same operative chapter carries height-in-relation-to-boundary recession planes, yard setbacks, and site-specific precinct provisions that can bind well before the height limit does. Read the operative provisions and the precinct appendices for the specific parcel before assuming the zone name implies unlimited redevelopment; a rezoned parcel that cannot assemble a compliant envelope produces no additional dwellings regardless of its nominal capacity.

The guide's primary metric is estimated additional consented dwellings attributable to rezoning, divided by hectares of residential parcels actually rezoned. The numerator counts dwellings, not consent applications. One resource consent can authorize a terraced block of many homes, so application counts systematically understate yield in precisely the neighborhoods where upzoning works hardest — and overstate it where single-house consents dominate.

The pipeline is reproducible in four steps. Intersect pre-reform and post-reform zoning polygons against one consistent cadastral parcel layer, not two vintages of parcel boundaries that silently shift the denominator. Classify each parcel by whether the rezoning produced a substantive capacity increase. Attach geocoded dwelling-consent records to the resulting land units. Then aggregate.

Fix the denominator as the full land area of eligible rezoned residential parcels, including parcels that produced no consents at all. Exclude road reserves and parcels whose zoning did not change. This is the step most analyses skip, and it is the one that keeps failure visible: a neighborhood where half the rezoned hectares never drew a consent should score half as well as one where every hectare did, even when both produce the same absolute dwelling count.

Additionality requires a documented counterfactual built on comparable land and pre-reform trends, with an identical observation window for every candidate. Observed consents divided by hectares measures activity, not the reform's effect. Auckland dwelling starts averaged 6,300 per year in the five years to 2016, according to Atlas Economics's presentation of Stats New Zealand data, so any post-2016 count inherits a rising baseline. Atlas Economics also reports the mix diversified toward terraced houses and small unit blocks after 2016 — a genuine change in product, but still not a yield.

The myth that the neighborhood with the largest increase in allowable dwellings is automatically the best expansion target fails on exactly this distinction. Auckland had roughly 350,000 units of zoned capacity in 2015, per Michael Wiebe's Building Abundance (30 June 2026), and a 2026 commentary in The Fifth Estate describes the 2016 plan as providing for almost one million new homes in existing residential areas. Neither number is a yield. The 2021 Urban Studies paper by Greenaway-McGrevy, Pacheco and Sorensen (vol. 58, issue 5, pp. 959–976, DOI 10.1177/0042098020940602) found upzoning raised the redevelopment premium embedded in property values — a price signal, not a consent count.

| Figure | Value | Source | What it actually measures |
| --- | --- | --- | --- |
| Zoned capacity, 2015 | ~350,000 units | Wiebe, Building Abundance, 30 June 2026 | Theoretical capacity, not yield |
| AUP planned capacity | Almost 1 million homes | The Fifth Estate, 9 June 2026 | Planned capacity, not delivery |
| Capacity change from 2016 reform | Trebled | California YIMBY, 9 Aug 2022 | Permitted dwellings, not consents |
| Units/townhouses share of starts, 5 yrs to 2016 | 22% | Atlas Economics | Pre-reform product mix |
| Units/townhouses share of starts, 5 yrs after 2016 | 46% | Atlas Economics | Post-reform product mix |
| Average annual dwelling starts, 5 yrs to 2016 | 6,300 | Atlas Economics / Stats NZ | Baseline trend for the counterfactual |

Before scoring any 2026 candidate area, compute the denominator first: pull the rezoned parcel layer, drop road reserves and unchanged parcels, and count the hectares that remain — including the ones that produced nothing. If that number is not fixed before you look at consents, you are ranking activity, not reform effect.

![Elevated view across Auckland housing developments dense brick](https://static.mm-ais.com/article-images-ai/auckland-housing-growth-audit-20-529-con-ai-b6cbbe9c.jpg)
Elevated view across Auckland housing developments dense brick

## Audit the Evidence Behind Auckland’s 20,529-Consent

Auckland’s headline consent count is an audit anchor, not a causal estimate. According to Stats NZ’s “Building consents issued: December 2021,” Auckland recorded 20,529 new dwellings consented in the year ended December 2021, against a national total of 48,899. These are evidence of substantial permitted activity—not an estimate of growth caused by upzoning. The release establishes the scale of permissions issued; it does not identify how many would have been issued without the zoning change.

Dividing the Auckland observation by the national observation gives approximately 42.0%. That share is a reproducible calculation from the same Stats NZ release, not a separately observed measure of policy performance. It establishes Auckland’s importance to New Zealand’s permitted housing supply. It says nothing about productivity per rezoned hectare: neither input identifies rezoned land, and neither separates additional permissions from permissions that would otherwise have occurred. Regional prominence cannot validate a neighborhood ranking based on nominal capacity increases.

According to Stats NZ’s outcome definition in that release, the observations concern new dwellings consented. They are neither completed homes nor a net increase in dwelling stock after demolitions. The evidence ledger therefore needs separate columns for permissions, completions, and stock changes, each with its own source and reporting period. An unavailable completion count must remain unavailable, not inherit the consent count. Likewise, demolitions belong in the accounting for stock change; subtracting them from the published consent measure would silently redefine the outcome being audited.

Commission Auckland Council to produce a parcel-level evidence table containing consented dwelling counts, consent dates, zoning-change classifications, and audited land areas. Specify dwelling counts rather than application counts, and require parcel identifiers that connect the consent records to the zoning and land-area records. For developments spanning parcels, require an explicit allocation method that preserves the development’s total dwelling count. For repeated consents on the same land, ensure the area denominator does not count the same rezoned footprint repeatedly. Otherwise, a seemingly precise local yield can reflect record-linkage choices rather than housing permissions.

Every reported local yield should carry the actual dataset names, extraction date, geographic coverage, and area-unit definition. Require the geographic boundary used for the numerator to match the boundary used for the denominator, with exclusions documented. These are acceptance conditions for the commissioned table, not claims that a suitable Council extract has already been obtained. A neighborhood label without a reproducible boundary and audited hectare denominator is insufficient evidence for an expansion decision.

Finally, require an evidence-status field distinguishing published observations, independently reproduced calculations, and model estimates. The regional totals belong in the first category; the division above belongs in the second. Counterfactual-adjusted local yields belong in the third and need their model source and uncertainty attached. Mark unavailable neighborhood rates explicitly “unreported.” Accept the table for prioritization only when a reader can trace each local yield to its records, land denominator, and estimation status—not reconstruct it from regional headlines.

## Choose Adjusted Local Yield Over Blanket or Volume-Led

Rank candidates by attributable consents per rezoned hectare, not by the size of the capacity uplift. The 2016 reform's own geography already hints at why: according to The Fifth Estate's June 2026 account, the Unitary Plan concentrated significant capacity increases around metropolitan centres, town centres and transport corridors where accessibility, infrastructure and demand could support growth, and height limits rose most where transport access and existing urban services were strong. That is a servicing screen embedded in the reform's footprint — not a delivery guarantee. The three ranking signals below are not interchangeable, and only one survives contact with a servicing budget.

| Expansion strategy | Ranking signal | Main failure | Verdict |
| --- | --- | --- | --- |
| Blanket capacity expansion | Additional legally allowable dwellings | Treats unrealised capacity as delivery | Not the default |
| Consent-volume targeting | Total permissions in an area | Rewards large areas and existing demand | Reject as the primary ranking |
| Adjusted land-yield targeting | Estimated attributable permissions per land unit | Requires credible spatial estimation | Winner, conditional on feasibility |

Apply the infrastructure feasibility gate before comparing eligible locations. A high-ranked candidate must carry documented water, wastewater and transport servicing arrangements; rezoning does not manufacture network capacity, and a yield estimate computed on unserviced land is a paper number with a spatial index attached. This gate is binary and it runs first — it is not a tiebreaker.

Then run the transport-demand comparison. Compute expected household accessibility to employment and services, and test whether the apparent housing advantage depends on costly additional vehicle travel that the land-yield measure omits. Atlas Economics links the transit-oriented density policy to the government's Urban Growth Agenda, whose stated aims included improving housing affordability and removing barriers to land and infrastructure supply — a reminder that accessibility is a policy input, not a residual. If the yield leader's edge evaporates once travel costs are priced, the edge was an accounting artifact.

Next, the distributional screen: tenant displacement exposure, access to opportunities, and the location of affordable housing obligations. Retain the yield-led winner only where identifiable harms have a credible mitigation pathway. A candidate that clears the servicing gate and the accessibility test but concentrates displacement without an obligation pathway is not a winner; it is a deferred liability.

Finally, select a portfolio, not a hotspot. Formulate it as constrained spatial optimization: maximize expected additional permissions subject to the servicing budget, network constraints, and adopted distributional requirements. The largest allowable-dwelling increase in a neighbourhood does not automatically make it the best location for the next expansion — that is the myth this section kills. Capacity is an input to the objective function, never the objective.

| Step | Test | Action |
| --- | --- | --- |
| 1 | Documented water, wastewater and transport servicing arrangements | Drop candidates relying on rezoning to supply network capacity |
| 2 | Adjusted land yield: attributable permissions per rezoned hectare | Rank survivors on yield, not nominal capacity |
| 3 | Expected accessibility to employment and services | Discount leaders whose edge depends on additional vehicle travel |
| 4 | Displacement, access and affordable-housing obligations | Retain only where harms have a credible mitigation pathway |
| 5 | Constrained spatial optimization across the portfolio | Maximize expected permissions under servicing budget and network constraints |

## What the Data Doesn't Tell You

A high estimated consent yield can be a property of the comparison design rather than the land. If Auckland selected neighborhoods for greater development rights partly because redevelopment demand was already accelerating, comparison areas with weaker prior trajectories can overstate the zoning effect. Counterfactual adjustment does not automatically remove this selection bias: matching starting densities or consent levels is insufficient when their trends differ. A credible local ranking requires evidence that the comparison captures the treated area’s likely trajectory without reform. Otherwise, apparent leaders should remain uncertain candidates—not be replaced by whichever neighborhood received the largest nominal capacity increase.

Spatial displacement creates a different problem: additional consents inside a rezoned boundary need not represent equally additional permissions across metropolitan Auckland. Developers may move projects from nearby unchanged land into treated neighborhoods. That can simultaneously increase measured treatment outcomes and depress comparison outcomes, exaggerating the apparent local advantage. A neighborhood can therefore genuinely attract development without generating an equivalent metropolitan gain.

According to California YIMBY’s summary of the Auckland research, the estimated effect remains statistically significant after controls for possible construction spillovers; explaining the observed increase through displacement alone would require control-area permitting to reach six times the amount observed in the study’s pre-reform reference year. That counterfactual test weighs against complete displacement, not against every degree of relocation. It does not establish that each neighborhood’s estimated yield is equally free of spillover contamination. Local rankings need sensitivity checks that distinguish comparisons using nearby unchanged land from comparisons less exposed to project relocation.

Transferring historical responses into today’s development market adds forecast uncertainty even where infrastructure is ready. Financing costs change carrying costs and required returns; construction prices change residual land values; developer balance sheets constrain how many viable projects can proceed; and presale conditions affect access to construction finance. Identical permissions can consequently produce different consent responses across market conditions. Historical rankings are conditional forecasts, not permanent parcel attributes. Their ordering becomes uncertain when the commercial conditions supporting a formerly high-yield neighborhood no longer hold.

Neighborhood averages also hide feasibility constraints below the reporting boundary. Parcel assembly may require coordinated purchases, while fragmented ownership can make agreement costly or impossible. Valuable existing buildings raise the opportunity cost of demolition; terrain can increase earthworks, foundation, and access costs. An average treatment effect can thus combine feasible redevelopment sites with locations where exercising the same rights is commercially implausible. Require uncertainty intervals for local yield estimates and subgroup results separating these constraints. Where subgroup evidence is sparse, label the ranking unresolved rather than treating the neighborhood average as a parcel-level prediction.

Finally, permission yield cannot establish affordability, dwelling size, tenure security, or realized occupancy. According to Greenaway-McGrevy, Pacheco and Sorensen in Urban Studies, Auckland upzoning’s house-price effects depend on sites’ redevelopment potential, rather than having a uniform direction across properties. A location can score well on permissions while performing poorly on welfare measures. Before giving an expansion candidate priority, attach a limitation record covering trajectory comparability, displacement sensitivity, current commercial feasibility, subgroup uncertainty, and separately measured welfare outcomes. The strongest credible additional-consent yield remains the criterion among comparable, infrastructure-ready neighborhoods; those checks determine whether an apparent advantage deserves that description.

## Work the 26,903-Dwelling Estimate Without Inventing a

Content for Work the 26,903-Dwelling Estimate Without Inventing a is being prepared.

## How to Choose Well

Selection logic for expansion must prioritize empirical yield over theoretical capacity. The following decision tree operationalizes the canonical rule: prioritize expansion into comparable, infrastructure-ready neighbourhoods with the strongest credible additional-dwelling-consents yield per rezoned hectare, rather than the largest nominal capacity increase.

| Decision Rule | Condition | Action |
| --- | --- | --- |
| 1. Stop on incomplete measurement | Attributable dwelling estimate OR matched rezoned-land denominator is missing | Mark candidate 'unranked'; do not assign proxy score |
| 2. Require demonstrated comparability | Development characteristics fall outside historical sample’s supported range | Classify as separate pilot; do not advance as proven extension |
| 3. Use conservative uncertainty threshold | Lower bound of 95% interval for additional yield ≤ 0 | Reject for broad expansion; evidence insufficient |
| 4. Exclude unfunded dependencies | Necessary servicing upgrade lacks identified funding commitment and delivery schedule | Hold candidate outside current expansion tranche |
| 5. Select robust leader | Rankings reverse or uncertainty materially overlaps across model specifications | Declare no clear winner; use accessibility and displacement safeguards as tie-breakers |

Rule 1 enforces data integrity. If a candidate’s attributable dwelling estimate or its matched rezoned-land denominator is missing, mark the candidate 'unranked' rather than assigning a proxy score. Imputation introduces bias that distorts the yield metric central to the thesis.

Rule 2 enforces domain validity. Advance a proposed expansion area only if its relevant development characteristics fall within the historical sample’s supported range. Treat extrapolation beyond that range as a separate pilot, not a proven extension. This prevents applying 2016 reform yields to structurally dissimilar contexts.

Rule 3 enforces statistical rigor. For this guide’s proposed screening standard, require the lower bound of a 95% interval for additional yield to exceed zero before treating the evidence as sufficient for broad expansion. This conservative threshold ensures that only statistically significant gains drive policy.

Rule 4 enforces fiscal realism. If a necessary servicing upgrade lacks an identified funding commitment and delivery schedule, hold the candidate outside the current expansion tranche regardless of its estimated yield. Infrastructure readiness is a binary prerequisite, not a variable.

Rule 5 enforces model robustness. Among candidates passing the earlier tests, choose the highest credible yield that remains preferred under alternative model specifications. Where rankings reverse or uncertainty materially overlaps, declare no clear winner and use documented accessibility and displacement safeguards as tie-breakers. This prevents overfitting to a single analytical framework.

## What to do next

| Step | Action | Why it matters |
| --- | --- | --- |
| 1 | Audit the headline Auckland consent tally; separate permitted dwellings, completed homes, and Unitary Plan capacity. | Citywide activity and legal capacity do not establish additional housing production. |
| 2 | Match Auckland dwelling-consent locations to Unitary Plan rezoning boundaries; verify affected hectares and align observation periods. | Local yield requires a geographically and temporally matched land denominator. |
| 3 | Estimate reform-induced dwelling consents against a credible no-rezoning baseline, accounting for displacement and spillovers as Greenaway-McGrevy and Phillips examined. | Gross consents include construction that would have occurred without reform. |
| 4 | Calculate additional dwelling consents per rezoned hectare for matched Auckland neighbourhoods; report uncertainty and withhold rankings where inputs are missing. | California YIMBY’s research summary supports a reform effect, not a validated neighbourhood ranking. |
| 5 | Check candidate Auckland neighbourhoods against council infrastructure plans and operative zoning envelopes, including Mixed Housing Urban standards. | Comparable development conditions and infrastructure readiness matter more than zone labels alone. |
| 6 | Prioritise expansion into comparable, infrastructure-ready Auckland neighbourhoods with the strongest credible adjusted local yields. | This targets additional dwelling consents per rezoned hectare—not the largest nominal capacity increase. |

## Frequently Asked Questions

**Does Auckland’s 20,529 figure count finished homes or just approvals?**

Stats NZ recorded 20,529 new dwellings consented in Auckland in the year ended December 2021, not completed homes or a net increase in dwelling stock after demolitions.

**What share of New Zealand’s consented new dwellings was in Auckland in 2021?**

Auckland’s 20,529 new dwellings consented represented approximately 42.0% of the national total of 48,899 in the year ended December 2021.

**Should rezoned parcels with no housing consents count when calculating yield per hectare?**

The denominator includes the full land area of eligible rezoned residential parcels, including those with no consents, but excludes road reserves and parcels whose zoning did not change.

**How do you separate the effect of rezoning from building that would have happened anyway?**

Additionality requires a documented counterfactual built on comparable land and pre-reform trends, with an identical observation window for every candidate area.

**Does Mixed Housing Urban zoning let owners build towers?**

The Mixed Housing Urban zone generally has an 11-metre building height standard, typically three storeys, with recession planes, yard setbacks and site-specific precinct provisions that can constrain development before the height limit does.

**Should a consent for a multi-home terraced block count as one unit in the yield calculation?**

The yield numerator counts consented dwellings rather than consent applications because one resource consent can authorize a terraced block containing many homes.

## Quick answers

| Why can’t the headline consent tally identify where to upzone next? | The headline consent tally presents a related problem: citywide activity cannot identify where the next hectare of upzoning should go. |
| --- | --- |
| What metric should measure adjusted local yield? | The guide's primary metric is estimated additional consented dwellings attributable to rezoning, divided by hectares of residential parcels actually rezoned. |
| Which land belongs in the denominator? | Fix the denominator as the full land area of eligible rezoned residential parcels, including parcels that produced no consents at all. |
| How should the audit establish additionality? | Additionality requires a documented counterfactual built on comparable land and pre-reform trends, with an identical observation window for every candidate. |
| What evidence is needed for a defensible local ranking? | A defensible ranking would require geographically matched consent records, rezoning boundaries, and estimates that account for displacement and spillovers. |

Also worth reading: **Solving the housing crisis with innovative local strategies**: [Solving the housing crisis with](https://urbanplanadvisor.com/blog/solving-the-housing-crisis-with-innovative-local-strategies.php) · **The critical steps to solving the urban housing crisis**: [critical steps to solving the](https://urbanplanadvisor.com/blog/the-critical-steps-to-solving-the-urban-housing-crisis.php) · **Boston 42-Day Permit Reform: $45K per Space, 64% Start Rate**: [Boston 42-Day Permit Reform: $45K](https://urbanplanadvisor.com/blog/boston-42-day-permit-reform-45k-per-space-64-start-rate.php)

### Related reading

- [Nonprofit Housing Developers Achieve 31% Growth in Urban Development Projects During 2024](https://urbanplanadvisor.com/blog/nonprofit_housing_developers_achieve_31_growth_in_urban_dev.php)
- [Solving the housing crisis with innovative local strategies](https://urbanplanadvisor.com/blog/solving-the-housing-crisis-with-innovative-local-strategies.php)
- [Streamlining the local permitting process for urban growth](https://urbanplanadvisor.com/blog/streamlining-the-local-permitting-process-for-urban-growth.php)
- [Affordable Housing Near Transit: 13-23% Proof, Set Floor Area Ratio First](https://urbanplanadvisor.com/blog/affordable-housing-near-transit-13-23-proof-set-floor-area-ratio-first.php)
- [SF SoMa 2026 FAR Caps: 450 Parcels Now Viable for Housing](https://urbanplanadvisor.com/blog/sf-soma-2026-far-caps-450-parcels-now-viable-for-housing.php)
- [Seattle Unveils Key Supportive Housing Near Woodland Park](https://urbanplanadvisor.com/blog/seattle-unveils-key-supportive-housing-near-woodland-park.php)

### Latest

- [Atlanta transit zoning update: 2,847 units, Floor Area Ratio (FAR) vs zero...](https://urbanplanadvisor.com/blog/atlanta-transit-zoning-update-2847-units-floor-area-ratio-far-vs-zero-parking.php)
- [Minneapolis triplex zoning in 2026: 4.1 per 1,000 lots, bundle vs legalize-only](https://urbanplanadvisor.com/blog/minneapolis-triplex-zoning-in-2026-41-per-1000-lots-bundle-vs-legalize-only.php)
- [Affordable Housing Near Transit: 13-23% Proof, Set Floor Area Ratio First](https://urbanplanadvisor.com/blog/affordable-housing-near-transit-13-23-proof-set-floor-area-ratio-first.php)

Canonical: https://urbanplanadvisor.com/blog/auckland-housing-growth-audit-20529-consents-rank-by-adjusted-local-yield.php
Markdown: https://urbanplanadvisor.com/blog/auckland-housing-growth-audit-20529-consents-rank-by-adjusted-local-yield.php/index.md