The S106 Lottery: Why the Same London Small Site Faces Eleven Times the Affordable Housing Threshold Depending on the Borough

I’ve been looking at S106 figures across London small sites and there is one observation worth pulling out before any of the cleverer ones. The affordable housing threshold, which is a single number written into each borough’s local plan, varies eleven-fold across the capital. Camden has set it at one unit. Most outer London boroughs have settled on ten. Bromley sits a little higher at eleven. That single choice does an enormous amount of the work in determining whether a small-site scheme attracts an S106 agreement, and through that, what the developer ends up paying.

You might expect schemes to differ in S106 exposure mostly because of the scheme: density, design quality, transport impact, perhaps the case officer’s mood that week. The data turns out to be more boring and more interesting than that. The borough is doing most of the discriminating, and the threshold is the thing doing it.

The threshold gap, in policy

Five boroughs, taken from their own published documents, in the order most developers visit them:

Borough	AH threshold	Target %	Sub-threshold
Camden	1 unit / 100 sqm GIA	50% (sliding)	Financial in lieu
Lambeth	10 units	35%	None
Brent	10 units	50%	£50,000/unit at 5 to 9 units
Croydon	10 units	50%	None
Bromley	11 units	35%	None

Camden’s sliding scale (2% per unit or per 100 sqm above a one-unit base) means a five-unit scheme is in for roughly a 10% affordable housing contribution, paid as a financial in-lieu sum. The same five-unit scheme in Bromley attracts nothing along that axis. Brent’s middle ground, the £50,000-per-unit small-site contribution at five to nine units, is doing similar work to a formal S106 obligation but lands ahead of the threshold the borough applies above ten. None of these are odd policies if you read them in isolation. They are slightly odder when you read them next to each other and notice that the variable producing the largest swing in obligation cost is not, in fact, anything about the scheme.

What the officer reports actually record

Across 12,993 small-site decisions in 27 boroughs, about 26.8% of officer reports record an S106 agreement as required. The borough-level figure runs from one in fifteen at the low end to roughly two in three at the top:

Bexley records S106 on 6.5% of small-site schemes. Camden, on roughly 65%. The gap is wide enough to look implausible until you remember that Camden has chosen to start charging at one unit and Bexley behaves a little like outer south-east London more broadly: a high threshold, a fairly approval-friendly culture, most small schemes sitting comfortably outside the obligation envelope.

The middle of the chart is, I think, the genuinely interesting part. Croydon records S106 on 44% of schemes despite a ten-unit affordable housing threshold that ought to leave small sites entirely alone. Westminster at 37% and Lambeth at 57% sit in similar territory. What’s happening in those boroughs is that the affordable housing line is only one of several reasons an S106 might attach. Transport contributions, public realm works, employment and training payments, sometimes a bit of education funding: these come and go at officer discretion, sit underneath the headline policy, and show up only when you read the decision notice, not the local plan.

Site type does the second-order work

Within the same borough, the S106 rate differs again by site type:

Site type	S106 rate	n
DMR — Demolition and rebuild	48.9%	1,612
MID — Mid-terrace	43.7%	471
EXT — Extension or upper-floor	42.4%	462
END — End-of-terrace	40.6%	347
BCK — Backland	40.2%	331
MIX — Mixed use	41.0%	173
CNV — Conversion	33.6%	3,304
INF — Infill	28.7%	160

Demolition-and-rebuild collects an obligation most often, conversions least. The conversion figure is the one I keep returning to, because conversions are commercially the most plausible play for an architect-led developer with a small balance sheet. The cross-borough range on conversions runs from 5.5% in Hounslow to 77% in Lambeth, on samples of 80 to 250 schemes per borough. A developer running the same conversion thesis across two outer boroughs is taking on an order-of-magnitude difference in S106 probability before they’ve drawn a single internal partition.

What follows for the bid

I don’t claim every developer should rebuild their model around this. The straightforward implication is narrower: the borough deserves to be a variable in the residual land value calculation, with a probability of S106 attached and a contribution range likely for the site type and unit count. Most small-site models treat the borough as a fixed input, set once when the deal lands on the desk. That sets up the same scheme to look more profitable in one borough than another for reasons that are visible in the decision data but absent from the assumption sheet.

Three things follow from that, none of them especially radical:

• Sites in low-threshold boroughs typically need to be bid somewhat lower than sites in high-threshold boroughs to deliver comparable returns. Camden a bit lower than Bromley, on average, for reasons that include S106 exposure but also a few others.
• For schemes at five to nine units, the sub-threshold policy is doing as much work as the headline figure. Brent’s £50,000 per unit at 5 to 9 units adds £350,000 to a seven-unit scheme that wouldn’t apply two boroughs over. That sits in the legal agreement, not in CIL, and most pre-acquisition models miss it.
• CIL is the easy line. It is rule-determined, predictable, and roughly fair across boroughs at this scale. S106 is the harder one. The gap between what the local plan implies and what the officer report records is widest here, and that gap is where most bid-model errors sit.

What I haven’t yet measured

The figures above record whether an officer report indicates an S106 agreement is required. They don’t record the £ amount, the breakdown between affordable housing, transport, public realm and the rest, or the negotiation between proposed and final terms. The route to capturing those is structured reading of the executed S106 deeds and the viability assessments, which most LPAs publish alongside the decision notice. That work is in train; the binary signal here is what’s defensible today, and it’s already enough to argue that the borough is doing more underwriting than the bid model usually credits it with.

One caveat I’d want to put on the record clearly: detection of S106 imposition from officer reports is rules-based, and the rules are imperfect. Six boroughs (Ealing, Richmond, H&F, Greenwich, Havering, City of London) sit outside the cohort because the detection isn’t reliable on the present sample. Among the rest, the figures I’d trust most are the mid-range ones (Bromley 15%, Hillingdon 20%, Redbridge 29%, Croydon 44%), where the policy and the officer-recorded outcome track each other neatly. The extremes (Bexley at the low end, Camden and Waltham Forest at the high end) carry more uncertainty, and I suspect a fraction of the highest-rate figures are amplified by template effects in the underlying officer reports.

The borough as the variable

The recurring observation when you read enough London small-site decisions is that the borough is doing more work in the outcome than developers tend to assume. Density, conservation status, the architect’s elevations: these are the variables most carefully optimised because they’re the ones the developer controls. The threshold in the local plan is set once, by people the developer never meets, and it produces an obligation profile that varies by an order of magnitude across schemes that are otherwise commercially identical. Worth, I think, treating the borough as a variable rather than a constant. The figures above are the part of the case that’s currently legible. The £ amounts, when we get to them, will probably make it more so.

Headline borough patterns are in the free dashboards at /boroughs/. The site-type breakdown and the per-borough rates are in there too.