Where Ireland Actually Builds: The Geography of What Gets Through

Irish cities have the tightest housing markets in the country — 5.9% vacancy, the most people, the highest rents. They receive 5.1 commenced housing units per 1,000 residents from schemes of ten or more units. Towns of 1,500 to 10,000 people, with 8.8% vacancy, receive 11.6. The places where housing pressure is most acute get half the scheme housing per capita of medium-sized towns.

This is not a planning allocation failure. Seventy-nine percent of high-growth, low-vacancy small areas received at least one scheme grant between 2018 and 2022. The system is granting permission where the demand is. What fails is activation: schemes in cities commence at lower rates because they carry more constraints — and the constraints are measurable, rankable, and site-specific.

The strongest of them is the one this series has documented since Part 5: an An Bord Pleanala case. But two new spatial findings sharpen the picture. Schemes nearest to public transport — exactly where national policy wants density — commence at the lowest rates. And proximity to a protected structure carries a commencement penalty in the same range as an ABP appeal. The constraint map is the viability map, and it runs directly against the policy map.

Does Ireland build where the people are?

Not in proportion to them. The clearest way to see it is scheme housing granted and commenced per 1,000 residents, by settlement type.

Towns deliver 11.6 commenced scheme units per 1,000 people. Cities deliver 5.1, despite having the lowest vacancy and the most acute demand. These are 10+ unit scheme units only — one-off houses are counted separately, and they reinforce rather than offset the pattern: towns and villages received 10,025 one-off grants in this period against 766 scheme grants. Outside the cities, the one-off house is the housing delivery mechanism.

The rates are lower bounds — building control matching is incomplete, particularly in Dublin's core councils. But the same method applies to every settlement type, so the gap between cities and towns is informative even where the absolute levels are understated.

Why do schemes near transport do worst?

National planning policy since the 2018 framework has pushed transit-oriented development: build density where people can reach a bus or a train. The market does the opposite, and the data is unambiguous about it.

Schemes within 500 metres of transport commence at 55% (n=979). Schemes 1-2 kilometres away commence at 68% (n=202). The transit-adjacent sites carry 98 amenities within a kilometre against 17 for the 1-2km band — they are the dense, complex, high-friction urban sites where appeals, heritage, and apartment economics concentrate. The schemes that succeed are close enough to reach transport but far enough to escape the urban friction premium.

The same inversion shows up in deprivation. The most affluent areas — deprivation score below -2, vacancy at 5% — commence at 54.6% (n=383), a few points below the national rate. The tightest markets see slightly lower activation, not higher. Demand is highest exactly where the friction is highest.

What determines whether a scheme gets built?

We tested 30 spatial layers — environmental, heritage, infrastructure, amenity — against commencement for 1,906 residential scheme grants. Most of them are noise. Protected species proximity, groundwater vulnerability, radon risk, water advisories, building density, coal-restricted zones: all produce spreads under five percentage points, consistent with random geographic variation.

Six variables produce spreads of nine points or more, and they have credible mechanisms. The dominant one is the appeal process. ABP involvement outside Dublin drops commencement from 67% to 49% — an 18-point penalty (n=364), independent of dwelling type and scheme size, consistent with Part 5 and Part 6. When you isolate it in a controlled gradient — taking clean 61-100 unit sites and adding one factor at a time — ABP is a 30-point marginal penalty. Nothing else in the dataset comes close as a single factor.

A clean site — no ABP, no heritage, no flood zone — commences at 67% outside Dublin (n=768). Heritage proximity and an ABP case together drop it to 39% (n=69). The 28-point spread between the best and worst measurable profiles is set before a developer hires an architect. The constraint profile of the land predicts close to a third of the outcome.

The heritage layer the series had missed

Proximity to a protected structure is the finding this article adds to the series. It is not the strongest predictor — ABP is — but it operates independently of the others, and it lands in exactly the places policy targets for growth.

A building on the National Inventory of Architectural Heritage within 100 metres of a site: 46% commencement versus 60% for sites with no heritage nearby (n=217). A 15-point penalty, comparable to the ABP effect. The penalty for proximity to a structure on the Record of Protected Structures is steeper still — 38% (n=53) — but that sample is small enough to treat as indicative rather than precise.

The reason heritage matters is that it survives the obvious objection. Heritage-adjacent sites are urban, and urban sites are where apartments and appeals cluster — so the penalty could simply be those other things in disguise. It isn't. The heritage penalty holds at roughly nine points for houses (-8.7pp) and apartments (-9.0pp) alike. It holds at 15.7 points even among schemes with no ABP case. And it is worst not in Dublin but in towns (-15.3pp), where historic cores are precisely the areas development plans zone for densification. A protected structure triggers Section 57 obligations, conservation-officer involvement, and design constraints that add cost, time, and appeal risk.

One-grant developers are overrepresented on these sites: 78% of heritage-adjacent schemes are held by first-timers, against 69% of non-heritage schemes. Portfolio developers see the heritage flag and choose a different site. First-time developers walk into it — because they don't check, or because their one available site happens to sit there.

Is it Dublin, or is it Dublin's sites?

The most useful finding for anyone buying land is that Dublin is not the problem. Its sites are.

A clean mid-size scheme of 61-100 units in Dublin commences at 83% — the highest rate in the country. A clean site in Fingal or South Dublin, where building control data is more complete, commences at 68%, the same as a clean site anywhere else. Strip the constraint layers and Dublin outperforms.

The aggregate rate is low because the constraints stack. Only 14% of Dublin City Council scheme sites are clean. Next door in Fingal, 69% are. DCC carries 55% ABP involvement, 61% heritage proximity, and 50% apartment schemes; Fingal carries 25%, 11%, and 27%. Same housing market, same demand. Fingal commences at 65%, DCC at 46%. Some of that 19-point gap is a data artifact — DCC's building control linkage runs near 23%, which suppresses its measured rate — so the cleanest comparison is the clean-site one: clean Dublin-outer sites match clean sites nationally. On this evidence, Dublin has a constraint-stacking problem, not a location problem.

The city-versus-county pattern repeats. Cork City commences at 55%, Cork County at 66%. Galway City at 59%, Galway County at 74%. In every pair the county outperforms the city by 11-15 points, and the constraint profile — heritage, appeals, apartment share — explains the gap. The constraint is the urban fabric itself.

What gets built, and where

The sites that work share a profile: clean constraint status, 61-100 units, residential or recently-rezoned land, outside the historic core. Schemes on recently-rezoned agricultural land commence at 64%; on residential-zoned land, 57%; on mixed-use town-centre land, 44%. The system delivers on the edge of town and struggles wherever the existing urban fabric adds friction.

Carrigtohill in Cork County is the poster child: five schemes, 281 units, 100% commenced — planned suburban growth where the infrastructure was ready. Newtown in Limerick: ten schemes, 646 units, 100%. Carlow town: seven schemes, 204 units, 86%.

Where it fails is equally specific. Tullamore: seven schemes, 298 units, 0% commenced — a county town that should be viable, except the median new-build price there (€273,000) sits below the delivery cost floor documented in Part 1. Heritage-constrained town centres in Louth commence at 33% despite strong demand. These schemes were not refused permission. The constraint environment made the granted permission undeliverable.

One caveat worth stating plainly: a flood zone is the exception where the data is thin. Sixty-four schemes were granted in Flood Zone A and commence at 42%, but they average 89 units against 44 elsewhere — so the penalty partly reflects scheme size, a known low-commencement profile, not flood risk alone.

The policy contradiction

National policy targets compact growth in urban centres — the places with the most heritage, the most appeals, the most constraint stacking, and the lowest commencement. The market delivers on greenfield edge-of-town sites, where the constraint layers thin out.

One-off houses, which policy explicitly discourages, commence at 70.5% nationally with 2.5% ABP involvement. Urban apartment schemes, which policy encourages, commence at 43% in cities with 46% ABP involvement. The most reliable housing product and the policy-preferred housing product are opposites.

Neither policy is wrong in isolation. Compact growth makes environmental and infrastructure sense. Heritage protection preserves irreplaceable buildings. The appeal system provides democratic oversight. Each layer is rational. The compound effect is a system that delivers at the margins — literally, at the edges of towns and cities where the constraint layers run out.

What this means before you submit

If you are evaluating a site, check the constraint profile before anything else. The variables that move the outcome are the appeal history of the council, heritage proximity (NIAH and RPS within 100m), flood zone status, and zoning category. A clean site in any region commences at 67%. A heritage-adjacent site with an ABP case: 39%. The 28-point spread is knowable before you commit capital.

If you are building in a city or town centre, heritage is the constraint to assess first, because it is the one most likely to be missed. In Dublin City, 61% of scheme sites carry it. The permission may be grantable; the question is whether the design, cost, and timeline of heritage compliance leave a deliverable scheme.

If you are a first-time developer, you are statistically the one walking into these sites — 78% of heritage-adjacent schemes are held by first-timers, because the portfolio developers who know the system avoid them. A single query returning the heritage, appeal, and flood status of a site tells you whether you are looking at the 67% profile or the 39% one.

Methodology

Population. 1,906 residential permission grants of 10 or more units, received January 2018 to December 2022, across all 31 local authorities. Permission-type applications only (excluding extensions of duration, compliance, retention).

Commencement. Linked National Building Control Office records. Dublin core (DCC, DLR) linkage runs near 23% versus 54-66% elsewhere, so Dublin-core absolute rates are lower bounds. All headline constraint comparisons use data from all regions; where Dublin-core figures are cited they are flagged as lower bounds and are not used to support the constraint-stacking conclusion, which rests on Dublin-outer (Fingal, South Dublin) and outside-Dublin data.

Per-capita figures. Granted and commenced 10+ unit scheme units divided by Census 2022 population, by settlement classification. These count scheme housing only. One-off houses (10,025 grants in towns and villages versus 766 schemes) are excluded from the per-capita rates and reported separately; including them widens the town-versus-city gap rather than narrowing it. CSO completion totals exceed our scheme counts in every council because they also capture one-offs, small schemes, social housing, and conversions — this analysis isolates the highest-friction slice of the pipeline, not total delivery.

Spatial variables. Heritage: NIAH within 100m (48,000 rated buildings), RPS within 100m (32,917 structures from 30 development plans), SMR within 100m. Environmental: flood zone (OPW CFRAM), knotweed within 250m (NBDC/GBIF citizen-science records, not exhaustive), protected species (GBIF, 1km), EPA-licensed facility within 500m. Infrastructure, ground, energy, location, and zoning layers as listed in sources. Heritage proximity indicates a protected or rated building near the site, not on it — the constraint arises from setting, character, and design implications.

Variable selection. Of 30 variables tested, six produce commencement spreads of nine points or more; the remaining 24 fall under five points. The six each have a credible causal mechanism (heritage adds design cost and appeal risk; ABP adds time and uncertainty; flood zone adds insurance and engineering cost; knotweed blocks mortgage lending; EPA proximity affects buyer perception; area BER proxies for prior delivery success). Multiple-testing risk is low because the significant variables sit an order of magnitude above the noise floor.

Heritage magnitude. The robust figure is NIAH within 100m: -14.7pp (n=217), comparable to the ABP penalty. The RPS-only figure (-22.6pp, n=53) is the steepest but smallest-sample case and is reported as indicative. Controlling for dwelling type, the heritage penalty is ~9pp for both houses and apartments; controlling for ABP, ~16pp among schemes with no appeal. In an isolated marginal gradient (clean 61-100 unit sites, one factor added) heritage moves commencement only a few points while ABP moves it ~30 — ABP is the dominant friction; heritage is a real, independent, secondary one.

"Clean" definition. No ABP case linked, no NIAH or RPS within 100m, no flood zone. 59.3% of sites outside Dublin are clean (67.1% commencement, n=768); 37.5% in Dublin. This single definition is used throughout.

ABP penalty. 18.1pp outside Dublin (n=364), stable across scheme sizes. The aggregate ABP penalty varies by year (near zero in the SHD era and during Covid, 19-22pp in 2022); all ABP figures are period averages.

Flood zone. 64 schemes in Flood Zone A, 42.2% commencement (n=64). These average 89 units versus 44 for non-flood schemes, so the penalty is confounded with scheme size and should be read as directional.

Transit and deprivation. Distance-to-transport and Pobal HP deprivation index from Census 2022 small areas joined by spatial lookup. Transit bands have samples of 110-979; the gradient is robust.

Figures deliberately excluded. Dublin-core clean-site commencement (30.4%, n=23) is not cited as a headline due to the linkage baseline. A conditions-count gradient was dropped after roughly 16% of the population proved to be amendments to existing permissions (which carry fewer conditions and may not file separate commencement notices); excluding them, the spatial constraint findings are unaffected or strengthened, but the conditions signal is too contaminated to publish. The "prior successful scheme nearby" cluster effect was not testable within query performance limits.

Sources