The Unearned Increment

Thirty years of London terraced house sales reveal a city splitting at the seams

Imagine it’s 1995. You and your insufferable cousin, Julian, are both shopping for your first homes. Julian is doing a bit better than you - he’s a few rungs higher in the corporate pecking order, has a better-fitting suit, whiter teeth, the works.

He pays £118,521 for this beautiful terrace next door to a semi-retired rock star in Richmond Upon Thames, within walking distance of the deer in Richmond Park.

You scrape by for the downpayment on a £85,000 dingy terrace near a derelict gasworks on the canal in Hackney. You are embarrassed to invite your mother over and your daily commute is a sweaty, soul-crushing nightmare on the North London Line. Julian spends the next decade mocking you for your “dodgy postcode” and “questionable life choices” at every family gathering.

Fast forward to April 2026. Julian’s house is now worth £872,500. He has 7.4x-ed his money. In any other asset class, he’d be flogging a course by now.

You, on the other hand, are the proud owner of a Hackney terrace worth £1,132,500. Your house has 13.3x-ed. Julian mutters something about you “just getting lucky with location” into his Chardonnay.

And he’s not even entirely wrong.

Figure 1

You didn’t do anything to regenerate Hackney. You didn’t open the coffee shops, throw the acid techno raves at Hackney Wick or lobby for the Overground extension. You bought a house and went to work every day - as did Julian. The £260k gap between you is not a reward for how much nicer your house is - Julian has an objectively better home in every possible metric. You won a lottery ticket, denominated in postcode. Your house didn’t change, didn’t magically become 13.3x better. What did change, however, was the land beneath it and everything that the community around it created

Henry George called this the unearned increment: the rise in land value that owes nothing to the landowner’s effort and pretty much everything to the neighbourhood around them. A new Tube station opens, a decent school fills its catchment area, maybe even a farmers’ market appears on Saturday mornings - and the land price goes parabolic. The PTA of the school and the artists living in squalid sublets don’t collect. But you, as the freehold owner, do.

What follows is an attempt to measure how unevenly that increment has been distributed across London - using 1,085,883 terraced house sales from the HM Land Registry spanning 1995 to 2025.

How this works. First, we filter exclusively for terraced houses – comparing a Kensington mansion flat to a Dagenham semi is meaningless, so fixing the property type isolates the land value. We then group these sales into roughly 1,820 hexagonal cells (about 460m across) and smooth the data using a 5-year rolling average. To calculate the drift index, we first determine a neighbourhood’s starting position in the late 90s (1995–2000) as a percentage of the London-wide median. We then calculate its current status the same way, and divide where it is today by where it started. Because house values across all of London grew over this period, this metric strips out general inflation. It shows us whether an area was lagging behind in the race or storming ahead: a score of 1.0 means it kept pace, above 1.0 means it outperformed, and below 1.0 means it fell behind.

The map

Have a play with the slider. Blue means the area is pulling ahead of London’s average; red means it’s falling behind. Hover over a hex for the area name and exact drift index.

Figure 2

One thing I want to make crystal-clear: this is not a map of house price growth. Every single area in this city got more expensive in nominal terms. The red hexes didn’t get cheaper - they got cheaper relative to the rest of London. If you bought a home in a red zone in ’95, you still made a respectable amount of dosh. Just not as much as others.

The most “perfectly average London terrace” in 2025 - the hex cell with the drift index closest to 1.0 - is in Ilford. At the top, Newham has a drift index of 2.35: it has more than DOUBLED its position in the city’s land-value hierarchy since the late nineties. At the bottom sits Bromley at 0.53, meaning it has lost nearly half of its relative standing against the London median.

Figure 3

The gap

The spread between London’s most and least expensive terraced-house neighbourhoods grew by 339% between 2000 and its peak in 2016.

Figure 4

If you scroll back up to the map and set the slider to 2001, the city is almost a uniform pale sea. That looks like common prosperity - and the P90/P10 ratio of 1.202 backs it up. In reality, this was an artefact of a universally distributed credit bubble. High-street banks were handing out 125% mortgages and self-cert liar loans to anyone who walked in. The tsunami of cheap credit hit every postcode simultaneously, compressing the gap between Barking and Belgravia, purely by making money cheap enough that nobody had to think too hard about where they were buying.

When the Global Financial Crisis hit and the credit evaporated, that illusion shattered. The underlying geography of London’s land values - centuries-old, stubborn, deeply structural - reasserted itself with a vengeance.

Why it happened

Figure 5

Each hex is sorted into one of four buckets by its price position in 2000 and tracked through twenty-five years. The top line is the most expensive quartile – your Westminsters, Barnets, Camdens. The bottom is the cheapest – the likes of Havering, Bexley, Hillingdon. The dashed line at 1.0 marks the baseline: where you’d be if your neighbourhood had kept pace with its own 1990s position.

The post-GFC decade is the story here. From 2009, the quartiles sprint apart. With cheap credit gone, the underlying geography of London land values was once again legible: expensive areas surged (the top quartile hit 1.18 by 2015, meaning those neighbourhoods were 18% further ahead of London’s average than they’d been in the 1990s), while the cheapest sank to 0.89. The mask was ripped off.

In comes 2016. April - Osborne announces a 3% SDLT surcharge on second homes. Brexit uncertainty arrived in June. Prime London – overvalued, reliant on foreign capital, laden with buy-to-let debt – had been building toward an overshoot since the post-Olympics boom. The top quartile collapses from 1.18 to 0.96 by 2025. Estate agents started calling it a “repricing of prime”, which is industry euphemism for oh God.

The convergence since 2017 is real – the P90/P10 ratio has been falling steadily for eight years. But it has closed primarily through the slackening at the top, not creation at the bottom. Kensington got cheaper; Barking didn’t get richer.

Most interesting is where the four lines end up in 2025. The cheapest and most expensive quartiles have converged at roughly 0.96. The middle two quartiles – Q2 and Q3 – are both lower, at around 0.90. The extremes held their relative positions better than the middle did. There’s a structural logic to this: prime areas are supply-inelastic (you cannot build more Kensington; every Georgian terrace is listed, every mews is full, and planning rules prevent densification), so when demand retreats there is no supply response to deepen the correction – the floor holds. The cheapest areas are supply-elastic: you can build in Barking, convert offices in Croydon, and the displacement of Zone 2 priceouts into Zone 5 provides genuine economic demand underneath the numbers. The squeezed middle got the worst of both worlds – not scarce enough to hold value, not cheap enough to benefit from this displacement.

All four lines end up below 1.0 in 2025. This is a stat worth interrogating before you take it at face value.

Figure 6

The picture is more complex than the clean story about land and supply would suggest. All four property types follow the same temporal arc – rising divergence to a 2013-14 peak, then falling back after the SDLT surcharge and Brexit – which tells you that the cycle was driven by macro forces that hit every asset class simultaneously. Cheap credit in; policy shocks and rising rates out.

What differs is the level and the amplitude. Flats and detached houses run above terraced in absolute P90/P10 terms for most of the period, which is initially surprising. But the flat stock in London is not a single asset class – a studio bedsit in Dagenham and a three-bedroom in Mayfair are both “flats” in the Land Registry data. The dispersion reflects stock heterogeneity as much as geography. The same is true of detached, in spades. Semi-detached is the most internally standardised type and shows the lowest ratio throughout – probably because two semi-detacheds in different boroughs are genuinely more similar to each other than two flats are.

Terraced houses show the largest proportional increase in spatial divergence over the period: from around 3.4 in 2000 to a peak of roughly 4.85, a 43% rise. That’s consistent with the land hypothesis – a Hackney terrace and a Bromley terrace are genuinely similar structures, so price divergence between them reflects the land beneath more cleanly than any other type. But the supply elasticity argument for flats doesn’t hold up neatly in this data. What the chart is really showing is that macro forces set the tempo, and structural geography determined who felt it most.

The Bank of England raised this directly in a 2019 Staff Working Paper: is this really a land story, or just a cheap credit story? Their argument – if a house price is the present value of expected future rents, and the risk-free discount rate falls from 5% to 1%, the price mechanically quintuples even if rents stay flat. But if cheap credit were the whole explanation, we’d expect broadly uniform appreciation across all postcodes, with spatial inequality staying stable. What we actually see – a tripling of dispersion – points to something more structural than monetary policy. The credit environment set the overall level; the land determined who benefited.

Checking our own homework

There’s something that should nag at you about all four quartile lines ending up below 1.0. The drift index divides each hex’s price by the London-wide transaction median. But that median isn’t spatially neutral – it’s volume-weighted. If expensive areas generate more transactions per hex than cheap areas, they pull the median upward compositionally, even if no individual neighbourhood’s prices changed. This is a potential Simpson’s paradox: the aggregate can move in a direction that no subgroup actually moved.

To check, we can compute two versions of “the London median”: the standard volume-weighted one, and a spatial one – median price inside each hex, then the median of those hex medians. Equal weight to every neighbourhood, regardless of how many sales it had.

Figure 7

The answer is largely reassuring. In 2000, the transaction median was 0.95x the spatial median – the two tracked almost identically. The gap widened to 1.06x by 2012 before compressing back to 1.01x in 2025. That peak deviation of about 6% is modest: expensive areas were briefly over-represented in the transaction count during the post-Olympics boom, pulling the volume-weighted median upward. But the composition effect alone cannot explain a drift below 1.0 that is visible in every quartile, including the most expensive.

Let’s dig into the volumes.

Figure 8

Transaction volume across London has been remarkably stable by quartile over twenty-five years. The most expensive neighbourhoods generate roughly the same share of total sales in 2025 as they did in 2000. The composition effect is not driving the story. All four quartile lines ending below baseline doesn’t seem to be a statistical artifact, but rather reflects the concentrated gains at the extreme top of the distribution.

The league table

Here’s every London borough, ranked by how much a terraced house appreciated between 1995 and 2025. Find yours and see how it did.

Figure 9

Your neighbourhood

Pick your area from the dropdown. The line is the drift index: above the dashed baseline means the area has outperformed its own 1990s position relative to London; below means it has fallen behind. The bars show transaction volume – thin bars mean thin data, treat those years accordingly. Most areas peak in 2014-2016 and soften after; whether your area’s trajectory is idiosyncratic or part of the city-wide pattern is the interesting question.

Figure 10

So what?

Here’s the Georgist bit, and I’ll keep it short because George already wrote a whole book about it – which, true story, was the second best-selling book in the world after the Bible for a time.

The £660,000 gap between Hackney (13.3x) and Sutton (6.8x) is not a premium on Victorian craftsmanship or the quality of the pointing. It’s a premium on land – and on everything the land is adjacent to: schools, parks, restaurants, and the ‘social composition’ of the street. In essence, it is the ransom price of a postcode that says you’re ‘creative’ while you work in middle-management at an advertising strategy agency that churns out slide decks about Gen Z – all while you completely ignore your own children.

None of that was created by the person who owns the freehold. It was created by the community, by public investment, by thirty years of accumulated decisions by millions of people about where to live and work and eat. George’s argument was simple: the landowner who captures that value without creating it is engaged in something that is, at a structural level, robbery. Whether or not you buy the full prescription (a single tax on land values replacing all other taxation – either the most elegant policy idea in history or completely deranged, depending on your priors), the data is worth sitting with.

The dispersion of terraced-house land values across London more than tripled between 2000 and 2016. It has been narrowing since – but because the top fell, not because the bottom rose. That convergence, if it continues, will have been achieved through the destruction of one kind of inequality while leaving the structural conditions for another firmly in place.

Julian still isn’t talking to you. But at least now you know why you both got rich.

---

Data sources. HM Land Registry Price Paid Data (complete dataset, 1995-2025). ONS Postcode Directory (November 2025) for geocoding. Spatial aggregation via Uber H3 hexagonal grid at resolution 8. All code available on request (hey at aeronautes dot net)

Limitations. ONSPD coordinates are current-vintage, applied retrospectively to historical postcodes – a (hopefully) small number of early transactions may be in the wrong place where postcodes were retired and re-allocated. The 1995-2000 base period is data-thin for some outer London hexes; a minimum of 10 transactions in the base period and 5 in any rolling window is enforced. I’m not a Londoner and I don’t have the local knowledge to interpret every anomaly in the data – if you see something that looks interesting, feel free to investigate and let me know what you find. I’d love to hear from you.