London's new-build sales fell to a 30-year low in 2024
5,524 new-build homes were sold in standard market sales in London's 33 boroughs in 2024 — the lowest full-year figure in three decades of Land Registry records, and 67% below the 2016 peak. The 2024 count will still rise as late registrations arrive: it would need 20.1% more to match 2011, and so far no year since 2016 has gained more than 13.2% once 20 months had passed since its end.
Corrected · 8 and 10 October 2026
We first published this note on 17 July counting London as the 19 postcode areas that cover it, which gave 6,316 for 2024. On 8 October we recalculated every figure for the 33 London boroughs (5,524), and on 10 October we restated two comparisons with earlier years so that both years are counted at the same registration age. The 30-year-low conclusion held at each step. What changed, with the old figures.
5,524
That is how many newly built homes were sold in standard market sales (Land Registry category A) in London's 33 boroughs in 2024, according to HM Land Registry's Price Paid Data as released on 28 September 2026. The data covers residential sales for value registered in England and Wales (Right to Buy discounts, gifts and similar transfers are excluded), 4.0 million London records since 1995.
Before 2024, the lowest year since 1995 was 2011, with 6,637 new-build sales, 20% more than 2024's count so far. The 2016 peak was 16,832; 2024 landed 67% below it. Counted at the same point, 20 months after each year's end, 2024 was 18% below 2023 (6,772 at that point; 7,061 now, after another year of late registrations).
Scroll sideways for earlier years. London (33 boroughs) new-build market sales by year of completion, 1995–2025. Green = 2016 peak, amber = previous low (2011), red = 2024. The latest years are still being registered: 2024's count covers sales registered within 20 months of the year's end and will rise; 2023's covers 32 months. Source: HM Land Registry Price Paid Data, category A, release of 28 September 2026. CSV
A sharp break in 2023
This was not a gradual decade-long fade. From 2018 to 2022, London new-build sales held at 12,351–14,121 a year. The break came in 2023: on current counts, quarterly sales fell from 3,558 in Q4 2022 to 1,441 in Q3 2023, and have stayed far below the old level since (1,315–1,441 a quarter from Q3 2023 to Q4 2024). The 2024 quarterly average of 1,381 is 50% below 2022's 2,789 when both years are counted 20 months after their end, and 55% below 2022's 3,088 on current counts.
The temporary stamp-duty thresholds ended on 31 March 2025, and buyers rushed to complete before the deadline. To compare the two first quarters at the same point in the registration cycle, we count only sales that Land Registry had registered within 17 months of each quarter's end. On that basis, London's sales in Q1 2025 were 79% higher than in Q1 2024 (30,254 against 16,900). New-build sales were up 32% (1,575 against 1,191), less than half the market's rise.
This is mostly a flats story
95.7% of London's 2024 new-build sales were leasehold flats — new-build in the capital effectively means apartment schemes. New-build leasehold flat sales fell from 15,399 in 2016 to 5,289 in 2024, down 66%. As a share of all London sales, new-build went from 16.3% in 2016 to 6.6% in 2024. 2016 was the share's highest year: from 1995 to 2014 it ran at 5.3–10.0%, so 2024 is back within that range.
Why might this be happening?
We publish the numbers, not a policy position. Factors widely cited by developers and industry bodies for the post-2022 fall in London starts and completions include higher financing costs, build-cost inflation, and regulatory transition in the tall-buildings pipeline. We have no independent data on the weight of each.
What this data can and cannot see
Three caveats matter before quoting these figures, and one comparison.
Registration lag. Land Registry adds a sale when the transfer is registered. It says the gap is typically two weeks to two months, but new-build registrations can arrive many months later, so the 2024 figure is provisional and will rise. We measured how late earlier years' sales arrived: Land Registry's transaction IDs come in monthly batches, so each record shows the month it first appeared (method in the audit table). The data now covers the 20 months after the end of 2024. To tie 2011, 2024 would need another 20.1%, or 1,113 sales; the batches for April to August 2026 (published May to September) added 270. The table shows how much each earlier year grew after its own 20-month mark: at most 13.2%, in 2021. Late registrations were heaviest for 2020–2022 sales: in the first 12 months after month 20 they added 6.8–9.5%, against 4.3% for 2023. 2024 has also registered late: between months 12 and 20 it grew 20.5%, more than 2023 (16.5%), about as much as 2020 (19.4%) and less than 2021 (31.2%) or 2022 (59.0%). Even at 2021's rate, 2024 would end near 6,250, still below 2011. The 2025 full-year total (hatched in the chart) is far from complete and is not used for any claim above; the Q1 2025 comparison is matched for registration time.
| Year | Growth in months 12–20 | Growth in months 20–32 | Growth after month 20, to date | Months since year-end |
|---|---|---|---|---|
| 2016 | +0.3% | +0.2% | +0.3% | 116 |
| 2017 | +0.4% | +0.1% | +0.2% | 104 |
| 2018 | +0.3% | +0.2% | +0.3% | 92 |
| 2019 | +0.8% | +0.9% | +1.7% | 80 |
| 2020 | +19.4% | +6.8% | +11.8% | 68 |
| 2021 | +31.2% | +9.1% | +13.2% | 56 |
| 2022 | +59.0% | +9.5% | +10.7% | 44 |
| 2023 | +16.5% | +4.3% | +4.3% | 32 |
| 2024 | +20.5% | — | needs +20.1% to tie 2011 | 20 |
Build-to-rent is mostly invisible here. Price Paid Data records homes sold one by one. A block built or forward-funded for a rental operator never appears as individual sales, and new homes sold to companies, or to landlords with a buy-to-let mortgage, fall in category B, which we exclude. The fall in recorded sales may partly reflect London's shift toward build-to-rent; the data measures homes sold, not homes built.
Category B is excluded. We count category A, Land Registry's standard entries: single residential properties sold for value. Category B covers repossessions, buy-to-let purchases identified by a mortgage, sales to companies and other non-private buyers, and property type "Other". Land Registry has only identified B since October 2013, so a series back to 1995 has to use category A alone; the A series shows no step at that date. B has grown: 28.0% of London new-build records in 2024, against 15.5% in 2016. Including B, 2024 had 7,671 new-build records against 19,909 in 2016, a fall of 61% rather than 67%, and was still the lowest full year since B was introduced (9,300 in 2023, or 8,760 counted at the same 20-month point). Land Registry did not identify such sales before October 2013, so that total cannot be compared with earlier years: the 30-year comparison rests on category A.
Land Registry's House Price Index shows fewer (added 12 October 2026). The UK House Price Index, published by HM Land Registry with the ONS, counts 4,223 London new-build sales in 2024 (London region, July 2026 data). Its sales volumes are revised only for the most recent 12 months, so registrations that arrive later are not added; Land Registry says the volumes “reflect the number of sales that have been registered so far” and that “users should not use provisional UK HPI data volumes to infer trends in sales activity”. On that index too, 2024 is the lowest year from 1995 to 2024 (2023: 4,369; 2011: 6,614). Its 2025 figure, 3,882, is still provisional.
Check our arithmetic
Every figure above comes from HM Land Registry's yearly Price Paid files (pp-1995.csv to pp-2026.csv, release of 28 September 2026). Definitions: new-build = old/new flag Y (first sale of a newly built home); market sales = category A; year = date of transfer. London = the 33 London boroughs. We assign each sale by its postcode district (outcode): the 262 districts in which at least half of the residential postcodes we track lie in a London borough in the ONS postcode directory. This is the same rule as in Data Note #2 and on our live data pages; the list is in this CSV. Districts with almost no homes, such as EC2V in the City or TW6 at Heathrow, are not on it; they had no new-build sales in 2024. We count each sale once: Land Registry's files contain a few exact repeats (same address including the flat number, date, price, property type and tenure), and we keep one copy (the earliest-published, where the ID shows it). Among new-build sales that removes 13 records in 2024, all flats in one block in Harrow that were each published twice, and between 0 and 26 in other years (none in 2011); among all sales, 17–122 a year. Unnumbered flats at one address are never treated as repeats, since they can be different homes. Using Land Registry's own county field instead of districts (GREATER LONDON) gives 5,531 for 2024 and 6,618 for 2011, the same conclusion. All yearly, quarterly and lag figures: data-note-1-figures.csv.
Audit table — every number, its query, its result
-- Data: HM Land Registry yearly files pp-1995.csv … pp-2026.csv (release of -- 28 September 2026), loaded into a table ppd(id, price, date, postcode, type, -- new_build, tenure, paon, saon, street, locality, town, district, county, -- category, record_status). outcode = the part of the postcode before the space. -- london = the 262 outcodes in data-note-1-london-districts.csv (33 boroughs). -- 0. Exact repeats removed first: rows equal on postcode, PAON, SAON, street, -- date, price, type, new_build, tenure and category, where SAON is given or -- type is not F; keep one copy (the earliest batch where known). London category A -- new-build: 13 removed in 2024 (Orla Lodge, HA1 2BP), none in 2011. -- 1. Yearly new-build sales (chart; 2024 = 5,524; 2016 peak = 16,832) SELECT substr(date,1,4) y, SUM(new_build='Y') FROM ppd WHERE category='A' AND outcode IN london GROUP BY 1; -- 1995..2010: 10191 11198 12002 11333 12567 11672 11457 11690 10471 12631 11981 12625 8862 8240 7643 7576 -- 2011..2025: 6637 9300 10905 11403 13581 16832 14924 14121 12622 12446 12763 12351 7061 5524 3770* (*2025 incomplete) -- 2. Lowest year before 2024: 2011 = 6,637 → 2024 needs +20.1% (1,113 sales) to tie -- 2b. Same registration age (only sales in batches within 20 months of the year's end, -- step 7): 2016 = 16,787, 2022 = 11,154, 2023 = 6,772, 2024 = 5,524 -- → 2024 vs 2023 −18% (−22% on current counts); vs 2016 −67% either way -- 3. Quarterly new-build sales (2024 avg 1,381/q; 2022 avg 2,789/q at 20 months → −50%, -- 3,088/q on current counts → −55%) SELECT substr(date,1,4)||'Q'||((CAST(substr(date,6,2) AS INT)+2)/3) q, SUM(new_build='Y') FROM ppd WHERE category='A' AND outcode IN london AND date>='2022-01-01' AND date<'2026-01-01' GROUP BY 1; -- 2022: 2514 3142 3137 3558 · 2023: 2584 1607 1441 1429 -- 2024: 1365 1438 1315 1406 · 2025: 1575 1031 879 285 (2025 still arriving) -- Same quarters counting only sales in batches within 17 months of the -- quarter's end (step 7): 2023: 1867 1401 1269 1253 · 2024: 1191 1154 1116 1362 · 2025Q1: 1575 -- 4. Q1 2025 vs Q1 2024, like for like: only sales whose batch month (step 7) -- is within 17 months of the quarter's end (≤ 2025-08 for Q1 2024, -- ≤ 2026-08 for Q1 2025) -- all sales: Q1 2024 16,900 → Q1 2025 30,254 (+79%) -- new-build: Q1 2024 1,191 → Q1 2025 1,575 (+32%) -- (without the 17-month cut, Q1 2024 = 17,207 / 1,365) -- 5. Composition: leasehold flats = 95.7% of 2024 new-build SELECT ROUND(100.0*SUM(type='F' AND tenure='L')/COUNT(*),1) FROM ppd WHERE category='A' AND new_build='Y' AND outcode IN london AND substr(date,1,4)='2024'; -- new-build leasehold flats: 15,399 in 2016 → 5,289 in 2024 (−66%) -- 6. New-build share of all market sales: 16.3% in 2016 (the highest year) → 6.6% in 2024; -- 1995–2014 range 5.3–10.0% -- (all category A sales: 103,260 in 2016, 83,891 in 2024) -- 7. Month each record first appeared. Since September 2015 Land Registry's -- transaction IDs come in batches: the last three groups of the ID are -- shared within a batch (…-29C7-E063-4804A8C015BC = July 2026). The first -- group is a counter that wraps round once, in July 2023, when the fourth -- group changes to E063; so order the E050/E053 batches and the E063 -- batches separately, each by the first group. That gives 132 monthly -- batches, one per month from September 2015 to August 2026, plus a few -- one-off batches in 2015–16 (one re-issued IDs for older records). In 129 -- of the 132 the latest transfer date falls in that month; each of the -- other 3 holds one date amended later. A month's batch is published in -- the following month's release; our April 2026 copy holds batches up to -- March 2026 and none later. A record deleted and re-added under a new ID -- looks later than it was: 1 of the 340 London category A new-build -- records added in the latest monthly change file. -- Share of each year's current count that had appeared within 20 months of -- the year's end: 2016 99.7% · 2017 99.8% · 2018 99.7% · 2019 98.3% -- 2020 89.4% · 2021 88.4% · 2022 90.3% · 2023 95.9% -- Growth after month 15 (the original note's vantage point): 2020 +20.3%, -- 2021 +28.6%, 2022 +40.5% -- 2024 sales in the April–August 2026 batches: 72 41 77 50 30 (= 270) -- 8. Cross-checks. Land Registry's own county field instead of outcodes -- (county = 'GREATER LONDON'): 2024 = 5,531, 2011 = 6,618; same conclusion. -- Including category B: 2016 = 19,909, 2023 = 9,300 (8,760 at 20 months), 2024 = 7,671 (−61% on 2016); -- B share of new-build records 15.5% in 2016, 28.0% in 2024.
What we changed
Correction · 10 October 2026
Two comparisons set 2024 (20 months of registrations) against years that have had longer for late registrations, which this note's own method warns against. Counted at the same point, 20 months after each year's end, they change. The other comparisons, with 2011, the 2016 peak, the new-build share and the category B totals, give the same rounded figures either way.
- 2024 was 18% below 2023, not 22%: 2023 had 6,772 sales at that point and has 7,061 now. Measured at any point from 15 months on, the gap is 17–18%.
- We had compared the 2024 quarterly average (1,381) with 2022's on current counts: 55% lower. At the same point 2022 averaged 2,789 a quarter, a gap of 50%. 2022's sales were registered unusually late, so both figures are now shown.
Other changes: the "2024 vs 2023" row of the 8 October table uses the same-age figure; the chart caption says how long the latest years have had for registrations; the quarterly paragraph says which figures are current counts; we added that 2016 was the new-build share's highest year (from 1995 to 2014 it ran at 5.3–10.0%, and 2024 is back within that range); we write "standard market sales" instead of "sold on the open market"; the category B comparison with 2023 is also given at the same point (8,760); the headline reads "fell to a 30-year low in 2024" instead of "just hit a 30-year low"; and the figures CSV adds each year's count at 20 months since 2016. The 2024 figure of 5,524, the fall of 67% from the 2016 peak (which holds when both years are counted at 20 months) and the 30-year-low conclusion are unchanged.
Correction · 8 October 2026
We first published this note on 17 July 2026 counting London as the 19 postcode areas that cover it (E, EC, N, NW, SE, SW, W, WC, BR, CR, DA, EN, HA, IG, KT, RM, SM, TW, UB). Those areas also take in towns outside Greater London, such as Dartford, Esher and Cheshunt. We have recalculated every figure for the 33 London boroughs, the definition used in our Data Note #2 and on our press page, using Land Registry's latest release (28 September 2026, with sales registered to the end of August 2026). The 30-year-low conclusion still holds, now tested properly against late registrations (see the honesty box).
| Figure | Published 17 July | Corrected |
|---|---|---|
| 2024 new-build sales | 6,316 | 5,524 |
| Lowest year before 2024 (2011) | 8,060 | 6,637 |
| 2016 peak | 18,747 | 16,832 |
| 2024 vs 2016 peak | −66% | −67% |
| 2024 vs 2023 (same registration age; see 10 October) | −24% | −18% |
| Late registrations needed for 2024 to tie the previous low | more than 27% | 20.1% |
| Q1 2025 vs Q1 2024: new-build / all sales | +6% / +74% | +32% / +79% |
| Leasehold flats, share of 2024 new-build | 89.4% | 95.7% |
| New-build leasehold flats, 2016 → 2024 | 16,236 → 5,644 | 15,399 → 5,289 |
| New-build share of all sales, 2016 → 2024 | 15.3% → 6.4% | 16.3% → 6.6% |
| London transactions since 1995 | 4.6 million | 4.0 million |
Of the change in the 2024 figure, the boundary accounts for −1,074: on the April 2026 data we first used, the 33 boroughs had 5,242. Later data adds +282: 270 sales registered from April to August 2026 and a net 12 from other revisions to Land Registry's records. What else changed:
- We had written that 2023 "was already the weakest year on record at the time". It was not: 2011 was lower on either definition. Removed.
- We had written that late registrations would need to add more than 27% to 2024 to break the record, "well beyond typical revisions at this distance". That was wrong. From the same point, 15 months after the year's end, later registrations added 20.3% to 2020, 28.6% to 2021 and 40.5% to 2022. Five months on, 2024 needs 20.1% more. Growth after month 20, by year: 2016 0.3%, 2017 0.2%, 2018 0.3%, 2019 1.7%, 2020 11.8%, 2021 13.2%, 2022 10.7% (24 months after month 20 so far), 2023 4.3% (12 months after month 20 so far). In the same 12 months after month 20 it rose from 0.9% (2019) to 9.5% (2022), then fell to 4.3% (2023).
- We had written that the March 2025 stamp-duty rush "barely touched" new-build sales (+6%) and passed through them "almost without trace", which we read as a sign that the supply of finished homes was the constraint. That set a quarter with about 12 months of registrations against one with 24. Like for like (17 months each), new-build sales rose 32% while all sales rose 79%. We have withdrawn the claim and the interpretation.
- We had described a floor of seven quarters, mid-2023 to Q1 2025, at 1,550–1,700 sales a quarter. Sales have stayed far below the old level, but counted at the same registration age they kept drifting down to Q3 2024, so we no longer call it a floor.
- We had described category B, which we exclude, as "portfolio transactions and other non-standard records". It covers repossessions, buy-to-let purchases identified by a mortgage, sales to companies and other non-private buyers, and property type "Other". The honesty box now says so and shows the figures with B included.
- Quarterly figures and ranges are recalculated on the same basis. We removed a sentence quoting Molior's 2024 sales estimate, which we could not check against a published Molior source; softened a sentence on build-to-rent that went beyond the data; and removed a line describing these notes as weekly.