Why Are Houses So Expensive? — Gary Stevenson's thesis, made intuitive

1.The puzzle: everyone kept being wrong, in the same direction

After the 2008 crash, interest rates across the developed world fell to nearly zero. Every year afterwards, financial markets predicted they would soon return to "normal". Every year, they were wrong. Stevenson's favourite example: in September 2014, the governors of the US Federal Reserve — the people who set the interest rate — predicted where it would be over the next three years. Not one of the governors, for a single year, guessed too low. All of them were wrong, by a lot, in the same direction.

3.54%

Where Fed governors expected their interest rate to be by end-2017, on average (predicted in 2014)

1.38%

Where the rate actually was at end-2017 — the prediction was wrong by a factor of ~2.5

2.9 → 4.8

UK average house price as a multiple of a mortgage-borrower's income, 1992 → 2017 (ONS)

<3 → 10+

The same ratio for London alone, 1996 → 2017

Meanwhile asset prices did the opposite of what cheap-money-is-temporary thinking implied: UK houses, UK farmland (roughly doubling in price per acre between 2007 and 2016), and stock markets (the S&P 500 growing over 12% a year from 2009, before dividends) all raced far ahead of wages.

Stevenson's starting point is simple: when the smartest, best-incentivised predictors in the world are unanimously wrong for a decade, they are missing a cause. His candidate for the missing cause is rising wealth inequality. He has some standing to say so — as an interest-rates trader he bet, against the consensus, that rates would stay low precisely because inequality was rising, and in 2011 became his bank's most profitable trader in the world. The thesis is his attempt to turn that trading instinct into formal economics.

2.One idea you need first: an interest rate is just an asset price, upside down

This is the key that unlocks the whole thesis, and it needs no maths.

Suppose a house earns £10,000 a year in rent. If the house costs £100,000, an investor earns 10% a year. If buyers bid the same house up to £500,000, the same £10,000 of rent is now only a 2% return. Nothing about the house changed — the "interest rate" fell purely because the price rose.

The return on an asset = what it pays out ÷ what it costs. Push the price up, and the interest rate falls automatically. They are the same phenomenon seen from two sides.

So the two great puzzles of the post-2008 era — "why are asset prices so high?" and "why are interest rates so low?" — are really one puzzle: why is there so much money desperately trying to buy assets that it will accept almost no return for holding them? Answer that, and you have answered both.

3.The toy world: a game of Monopoly with two kinds of stuff

Economists test ideas by building tiny artificial worlds and seeing what happens inside them. Stevenson's world is deliberately as simple as possible, resting on three deliberate design choices:

Design choice 1

There are two kinds of stuff: things you use up, and things you own

The world contains consumption goods (food, haircuts, holidays) and a fixed stock of assets — think of it as land. Crucially, you cannot turn one into the other: no amount of saved wages manufactures a new plot of prime London. In most economic models saving automatically creates new capital, which quietly makes it impossible for asset prices to detach from everything else. Stevenson removes that assumption, because the decade after 2008 — savings piling up, interest rates at zero, companies buying back shares rather than building — looks exactly like a world where saving does not smoothly create new assets.

Design choice 2

The richer you are, the less of each extra pound you spend

Give a struggling family an extra £1,000 and they will spend nearly all of it. Give a billionaire an extra £1,000 — or an extra million — and they will spend almost none of it; it gets saved, i.e. used to buy assets. This is well supported by evidence, and it is the engine of the whole thesis. In the model it is built in mathematically: people get satisfaction not just from consuming, but from owning wealth itself, and the desire to consume satiates while the desire to hold wealth keeps going.

Design choice 3

One dial controls inequality

Society is split into a rich group and a poor group. A single dial, E (for Equality), sets what fraction of society is in the rich group — and the rich group owns all the assets, split evenly among themselves. E = 1 means everyone owns an equal share (perfect equality). E = 0.05 means 5% of the people own everything. Everybody works the same amount either way. Crude — but it lets you turn inequality up and down like a thermostat and watch what happens.

What happens when you turn the dial

Turn E down — concentrate the wealth. Each rich person now owns more, so each rich person is further into "spends almost nothing of each extra pound" territory. Society's total income hasn't changed, but more of it now flows to people who want to hold wealth with it rather than spend it. So more buying power chases the same fixed stock of assets. There is only one way that can resolve:

Wealth concentrates a smaller group owns everything, and each member is individually much richer

↓

Society's appetite shifts from spending to owning income lands in the hands of people who save nearly all of it

↓

More money chases the same fixed assets nobody can conjure new land, so competition only moves the price

↓

Asset prices rise relative to wages houses cost more multiples of a salary

↓

Interest rates fall — automatically same rents divided by higher prices = lower yield (section 2)

↓

Non-owners fall further behind they must buy dear assets that pay them almost nothing, while owners' paper wealth grows

⟲ Feedback: step 6 is a further rise in wealth inequality — which feeds back into step 1. High inequality can be self-reinforcing.

4.Try it yourself: the actual model, live

Below is not an illustration — it is Stevenson's model, solved in your browser using the equations from the thesis. Drag the equality dial and watch asset prices (as a multiple of wages) and the interest rate respond.

The inequality machine

Equality dial E — share of society owning all the wealth 10%

Capital's share of income (a) — how much of national income goes to asset owners 38%

–

Asset price ÷ wage (how many "years of pay" one unit of wealth costs)

–

Effective interest rate (what the asset pays ÷ what it costs)

Model 1 is the thesis's chapter-4 model (equations 4.10–4.14, solved exactly). Model 2 is the chapter-5 extension, solved numerically exactly as in the thesis appendix. Units are the model's own (total wealth, labour and technology all normalised to 1); the shapes, not the absolute numbers, are the finding. The default a = 0.38 matches the OECD capital-share figure the thesis cites.

What you should notice

The left-hand side of the chart is the whole story. As E heads towards zero — wealth concentrating into ever fewer hands — the asset-price-to-wage ratio doesn't just rise, it takes off towards infinity, and the interest rate collapses towards zero. And it accelerates: each step of further concentration does more damage than the last. If the real world sits somewhere on that left slope, then rising inequality produces exactly the twin phenomena of section 1 — and predicts they will persist as long as inequality does.

The wrinkle on the right, and how the thesis irons it out. In Model 1 you may spot that at high equality (E near 1) the curve can gently tip the "wrong" way, depending on the capital-share slider. Stevenson diagnoses this as an artefact of a crude assumption — Model 1 hard-wires the poor to spend everything, so as E rises, poor people "flip" into rich savers, which nudges prices up. Switch to Model 2, where everyone has the identical psychology and the rich save more only because they have more: the wrinkle vanishes. More equality then means cheaper assets and higher interest rates, smoothly, everywhere, for any capital share — even when the assets pay no income at all. The fix makes the result stronger, not weaker.

5.The four models are one argument under increasing attack

The thesis reads as four separate models, but it is more intuitive to see it as one claim surviving three rounds of objections — each model built to answer the sceptic's next question.

Prove the mechanism works at all.

MODEL 1 — the basic machine (static model). Two groups, fixed assets, the rich crave wealth. Result: concentrate wealth and asset prices explode relative to wages while interest rates sink towards zero — exactly the pattern the machine above shows.

✔ Mechanism established — with a wrinkle at low inequality.

You cheated — you hard-wired rich and poor to be different species. Of course you got a class effect out.

MODEL 2 — same psychology for everyone. Give every person the identical utility function; the rich save more only because they have more. The result gets cleaner and stronger: prices now fall with equality smoothly across the whole range, for any parameter values — even when assets are pure unproductive tokens earning nothing. Inequality alone can make worthless things expensive.

✔ Result survives and strengthens

In the real world people can build new assets. Your fixed pile of land is doing all the work.

MODEL 3 — the dynamic model, with buildable capital added. Now there are two assets: capital you can create by saving, and land you cannot. Run it over infinite time. The result survives — land prices still rise and its yield still falls as inequality rises — provided the rich specifically want to hold the fixed asset. If instead they only care about total net worth and treat the two assets as interchangeable, the effect dissolves. This is the honest hinge of the thesis: the mechanism needs the rich to want ownership of the scarce stuff itself — land, property, equity — not merely an abstract number. Stevenson argues a decade of zero rates and record share buy-backs looks like exactly such a world.

◑ Survives, conditionally — the thesis's most important caveat.

Fine — assets are expensive and yields are low. Why should anyone who doesn't own assets care?

MODEL 4 — the lifetime model (overlapping generations). Make the poor realistic: they live two phases, work-and-save then retire, and they care only about consumption — no craving for wealth at all. To retire they must buy assets from the rich while working and sell them back when old. High inequality forces them to buy dear and hold low-yielding assets, so a lifetime of work simply purchases less retirement. The harm needs no envy and no ideology: expensive assets are a toll on everyone who has to buy their savings from those who already own everything.

✔ The welfare punchline

6.Why it matters, if he's right

It's not a bubble. High asset prices would be the natural resting state of a highly unequal economy — not froth waiting to pop. Waiting for the "correction" would be waiting for something the model says isn't coming.

Rates aren't "normalising", ever — unless inequality falls. A decade of unanimous misprediction (section 1) stops being mysterious: forecasters kept expecting reversion to a normal that inequality had permanently moved.

The spiral. High inequality raises asset prices; high asset prices enrich owners and strip non-owners (Model 4), which raises inequality further. Stevenson sees the UK's falling home-ownership rate as this loop in action, and warns of the endpoint: a society where your position depends almost entirely on inheritance rather than work — his phrase for it is a new feudalism.

The policy target changes. If housing unaffordability is a symptom of wealth concentration, then planning reform, interest-rate tinkering and help-to-buy schemes treat symptoms. The lever that matters is the distribution of wealth itself — which points, though the thesis stops short of prescribing it, towards taxing wealth rather than work.

7.What the thesis does not prove

Being fair to the argument means being clear about its edges — most of which Stevenson flags himself.

It is a theoretical possibility proof, not an empirical test. It shows inequality could produce these phenomena in a consistent model — it does not measure how much of the real-world movement inequality actually explains, against rival suspects (ageing populations, a global savings glut, quantitative easing, secular stagnation).
The fixed-asset assumption carries a lot of weight. Defensible for prime land and for a zero-rate decade; much less obviously right over long horizons in which economies do build capital.
Model 3's hinge. The result requires the rich to specifically want the fixed asset. Plausible, but it is an assumption about taste, not a derived result.
Two classes, one dial. Real wealth distributions are continuous; a rich/poor binary is the bluntest possible representation, chosen for tractability.
Status of the work. An MPhil thesis (Oxford, 2019) — serious, supervised work, but not a peer-reviewed journal publication.

None of these breaks the core intuition. They mark the distance between "here is a coherent, worked-out mechanism the profession has ignored" — which the thesis delivers — and "this is proven to be the cause" — which it does not claim.

8.The whole argument on a postcard

The rich spend little of each extra pound; they buy assets instead. When wealth concentrates, more of society's income flows to asset-buyers, chasing an essentially fixed pool of things that can be owned. Prices of those things must rise relative to wages — and since an interest rate is just an asset's payout divided by its price, interest rates must fall. Both effects are permanent for as long as the inequality lasts, and both transfer lifetime wellbeing from people who must buy assets to people who already own them — pushing inequality higher still. That is why houses got expensive, why interest rates defied a decade of predictions, and why neither will "normalise" on its own.