The Most Restrictive Ten

The companion paper drew a distinction it did not fully work out: the difference between a borough that is hard because an adopted policy makes it hard, and a borough that is hard because it has decided to be. This note takes that distinction to the ten boroughs with the largest negative residuals, the ten that, given what they received, produced the fewest approvals. Running the plan audit across all ten returns a sharper result than a binary. Restriction does not divide in two. It divides in three, and the three are not equally defensible.

I will argue that six of these boroughs can name a policy that accounts for the shape of their restriction, and that four cannot. Five of the six protect family-sized homes from being carved into flats. One, Lambeth, applies a general amenity and garden policy across the whole small-sites cohort. Of the four without a written account, Waltham Forest is the sharpest case: its adopted plan promotes small-site intensification, and its outcomes run the other way.

I Why these ten: the boroughs that say no

The differential is observed approval minus the rate London-average decision behaviour would have produced on the same caseload, holding site type, public-transport accessibility, conservation status and unit count constant. The ten boroughs below are the ten with the largest negative residual in the 4 June 2026 build. Four of them are among London's cheapest. The interest of the other six is that they are not: Barnet, Kingston and Waltham Forest sit in the middle of the price distribution, and Lambeth is a relatively expensive inner-south borough. Whatever produces their restriction, it is not poverty.

The colour split does not follow price, and it does not follow the size of the residual. Barnet at −8.7 has a clearer policy basis than Waltham Forest at −15.3. It follows whether, when the adopted plan is opened, a policy accounts for the restriction the data records. For six of these boroughs it does. For four it does not.

II What the number hides: two signatures of restriction

As at the other ends of the table, the borough figure resolves into something more legible once it is broken to site type, with cells below thirty decisions suppressed. Here the disaggregation does more than sharpen the picture. It sorts the ten into two shapes, and the shape carries most of the verdict before a single plan is read.

The uniform signature is a whole row below expectation, and its tell is the demolish-and-rebuild column: it too runs negative. Croydon is the textbook case, depressed on all six measurable typologies (conversions n=327, demolish-and-rebuild n=258). Havering, Waltham Forest, Kingston and Lambeth share the shape on the cells they have volume in. A restriction spread this evenly is hard to read off a development plan, because plans constrain kinds of development, and a borough hostile to every kind at once is exercising a general disposition. The conversion signature is the opposite: a single red column over conversions, at Barking & Dagenham (−29, n=65), Newham (−20, n=142), Hounslow (−22, n=131), Tower Hamlets (−16, n=62) and Barnet (−15, n=174), beside a rebuild figure that is neutral or positive. That is the fingerprint of a family-housing policy. The two shapes do most of the sorting. The plan audit finishes it, because among the uniform boroughs one, Lambeth, can point to an adopted amenity policy where the others cannot.

III The unexplained four

Four boroughs carry the uniform signature, and in each the plan audit comes back the same way: there is no adopted small-sites policy that compels the restriction the data records. Havering defends its restraint openly, but the defence is a town-centre growth strategy. Its development-management policies are the ordinary character and amenity tests almost every borough carries, with no density cap, garden bar or conversion restriction in the adopted text. When a backland refusal at 20 Parsonage Road was appealed (APP/B5480/W/24/3355037, allowed 30 May 2025), the Inspector found the scheme complied with Havering's own Policies 10, 7 and 26. Croydon revoked its suburban design guide in July 2022 and stayed low and uniform on a plan that is broadly conformant. Kingston has no anti-rebuild policy, no loss-of-family-housing policy and no garden embargo.

Waltham Forest is the sharpest of the four, because its plan does not merely fail to compel the restriction; it points the other way. Local Plan Part 1, adopted 29 February 2024, promotes character-led intensification (Policy 8) and states that proposals for new homes on small sites will be supported where they infill, intensify and re-model existing housing (Policy 19). Policy 20 then gates conversions. When a small scheme is refused and appealed, the dismissal has rested on amenity and family-housing standards, not on the intensification policy. A borough whose plan says yes, whose outcomes say no, and whose refusals are defended on everything except the policy that is supposed to govern them, is restricting on a posture its own plan does not contain. That is the case the companion paper said belonged in a transparency regime rather than a plan-examination defence.

IV The legitimate six: family housing and amenity

The other six boroughs can each point to an adopted policy that accounts for the shape of their restriction, though the policies protect two different things. Five protect family housing. Barking & Dagenham's DMH5 and Newham's longstanding family-housing priority resist the loss of dwellings capable of housing a family with children. Tower Hamlets Policy S.H1 protects the supply of self-contained and family-sized homes. Barnet's Policy HOU03, adopted 4 March 2025, permits the conversion of a family house into flats only where at least one three-bedroom, four-bedspace home is re-provided with access to a dedicated garden, and only where the existing house meets a 135 square metre floorspace threshold; a borough-wide Article 4 direction has, since 2016, removed the permitted-development route to small HMOs. Hounslow's Policy SC6 requires a minimum original internal floor area of 130 square metres and a ground-floor family unit, and will not treat a house that has been extended up to that threshold as qualifying. In each the conversion column is the red one, the policy names conversions as the thing it constrains, and the restriction is the visible cost of a deliberate choice to keep family homes whole. That is a defensible distributional aim, and it is the kind of legitimate local choice the companion paper said the differential must leave room for.

One protects residential amenity, and it is the most contestable of the six. Lambeth's Local Plan supports housing supply in general terms under Policy H1, then gates the specifics through the garden, amenity, access and servicing tests of Policy Q14. Unlike the family-housing boroughs, its restriction is not concentrated on a single typology. It is uniform: conversions −15 (n=100), demolish-and-rebuild −19 (n=64), mid-terrace −36 (n=31). The restriction is plan-traceable and, on its own terms, lawful. The thing it protects is amenity and garden land, not a named housing need, and a borough whose general design policy does across every kind of small site the work a targeted policy would have to justify is making a contestable public-interest claim. Lambeth sits at 74 per cent on the Housing Delivery Test. Plan-backed and beyond reproach are not the same verdict.

V The verdicts

VI Restriction is a choice, not a shortfall

A natural defence of a restrictive borough is that it is simply struggling to deliver: that low approvals reflect a hard market, not a hard posture. The Housing Delivery Test, which measures each borough's completions against its requirement, removes that defence for much of this group. Waltham Forest restricts small sites while passing at 119 per cent; Croydon at 160; Hounslow at 108; Barnet at 104. These are not boroughs failing to build. They are boroughs meeting their housing numbers on larger sites and declining the small-site homes on top.

The boroughs on the other side of the test are in a different position, and not a more flattering one. Havering at 61 per cent, Barking & Dagenham at 66, Lambeth at 74, Kingston at 48 and Newham at 61 are all under the threshold that triggers the presumption in favour of sustainable development, and all are refusing the small-site homes that would help close the gap. For the family-housing boroughs that is a coherent trade: they have judged family housing worth the delivery cost, and the plan says so. For the unexplained ones it is harder to defend. A borough below its housing target, refusing compliant small schemes on a posture its plan does not contain, is paying twice: once in homes unbuilt, and once in the appeal costs and presumption exposure that an undocumented restriction invites.

VII The appeal record: refusals stick

The appeal record is easy to misread in both directions. Restrictive boroughs are not, on the whole, being overturned. Across the matched cohort the hardest boroughs lose around 16 per cent of decided appeals, below the permissive boroughs' 24 per cent, and several of these ten sit lower still: Newham's conversion refusals were upheld outright (0 of 24), Waltham Forest is overturned at 12.2 per cent, Tower Hamlets at 16.7. About six in a hundred refusals are appealed in the first place. A refusal in these boroughs is, in practice, the end of the road.

A low overturn rate does not license the unexplained boroughs. When Waltham Forest's refusal is upheld, it is upheld on amenity and family-housing standards, not on the intensification posture that actually drove the borough below expectation. There is almost always some generic policy on which a defensible dismissal can rest, which is why the appeal rate cannot distinguish a borough applying a real policy from a borough applying a posture and reaching for a policy afterward. The Inspectorate confirms the floor: these are not capricious boroughs whose decisions collapse on review. It cannot draw the line between the legitimate and the unexplained. Only the plan can.

VIII The spine: Waltham Forest, whose plan says yes

Tower Hamlets and Waltham Forest sit close together on the table and far apart on legitimacy. Both come down hard on conversions. Tower Hamlets restricts on Policy S.H1, an adopted policy that protects family-sized homes. Waltham Forest restricts on nothing its plan contains, a plan that, read plainly, encourages the small-site intensification the borough's officers decline. Same region, opposite verdict: one borough has put its restriction in writing and can be held to it at examination; the other has kept it out of writing and cannot. The question the companion paper framed for Havering applies with equal force here, and it is an enquiry rather than an accusation: your plan says yes, so what are you doing that it does not?

IX Conclusion

The companion paper proposed that a negative differential could be legitimate or not, and left the test for later. Run across London's ten most restrictive boroughs, the test returns three answers rather than two. Five boroughs restrict to protect family housing, on policies named in the plan and upheld on appeal. One restricts through a general amenity policy that bites hardest on the awkward plot, plan-traceable but resting on a public-interest claim that is far more contestable. Four restrict on nothing their plans contain, uniformly, across every typology, in one case in flat contradiction of a plan that promotes the very thing the borough refuses, and several do it while comfortably meeting their housing targets, which removes the excuse of necessity. The differential found the restriction; the type cut sorted it into its two signatures; the plan audit decided which were licensed. The league table called all ten of these boroughs restrictive. It could not tell you that six of them are restricting on a policy you can read and four on a disposition you cannot, and that distinction, not the ranking, is the one a borough has to answer for.

Sources & confidence

Outcome differentials and type-disaggregated cells: LSSPD, build 4 June 2026 (borough_differential.csv, type_differential_cells.csv). Appeal overturn: LSSPD matched-appeal cohort, London 21.0%; Waltham Forest 12.2%, Tower Hamlets 16.7%, Newham 0%, Kingston 25.9%, Barnet 27.3%. Housing Delivery Test, 2023 measurement (hdt_small_sites.csv): Waltham Forest 119% Pass, Barnet 104% Pass, Hounslow 108% Pass, Croydon 160% Pass; Havering 61% Presumption, Barking & Dagenham 66% Presumption, Lambeth 74% Presumption, Kingston 48% Presumption, Newham 61% Presumption; Tower Hamlets 92% Action plan. Plan audit: Waltham Forest LP1 (adopted 29 February 2024, Policies 8/19/20); Tower Hamlets Local Plan 2031 (adopted 15 January 2020, Policy S.H1); Barnet Local Plan (adopted 4 March 2025, Policy HOU03; HMO Article 4 in force 2016); Lambeth Local Plan 2020–2035 (adopted 22 September 2021, Policies H1/Q14); Hounslow Local Plan (adopted 15 September 2015, Policies SC4/SC6); Havering appeal APP/B5480/W/24/3355037 (20 Parsonage Road, allowed 30 May 2025).

What this can and cannot tell you. The descriptive claims are firm: these ten boroughs approve below their caseload-expected rate, their restriction takes one of two type-level shapes, and their adopted plans range from actively promoting small sites (Waltham Forest) to actively constraining conversions (Tower Hamlets, Barnet, Hounslow). The legitimacy verdict is an interpretation of the adopted plan text, not a legal ruling. Newham's family-housing reading is provisional pending adoption of its emerging Policy H4. Political control is stated for the 2022–2026 data window; Waltham Forest and Lambeth passed to Green control and Barnet to no overall control in May 2026, after the last decision in the data. The audit locates where each restriction comes from. It does not predict the outcome of any individual application.