The Four-Quadrant Model, Calibrated to a British Housing Market

The four-quadrant model, calibrated to a British housing market

Choose a place and a year. The rent, the price and the yield are the ones actually observed there. You supply the three behavioural parameters that no dataset contains, then shock the market and see whether the model's prediction survives contact with what happened next.

RE02 Real Estate Finance and Investment, Michaelmas Week 2. Built from around 294 million property search enquiries and the near-universe of listings for Great Britain, joined to HM Land Registry sold prices, aggregated to local authority and year.

1. Calibrate to a real market

Every figure in the band below is observed for this place and year, not assumed.

Area

Base year

Rent

—

Rent per year

—

Sold price

—

Implied gross yield

—

2. Supply what the data cannot

Parameters, not observations. The values are your judgement.

Rental elasticity of demand

How far occupiers economise on space when rent rises.

Price elasticity of new supply

How readily building responds to price. The planning regime enters here.

Replacement rate

The share of the stock lost each year, which building must replace to stand still.

3. Shock the market

Move one at a time before you move two.

Demand for space

Capitalisation rate

Construction cost

Against what actually happened

Model, long-run rent

—

from the shock you set

Observed rent change

—

Difference

—

model less observed

Quadrant variable	Base	Short run	Long run	Long-run change

What is real here, and what is not

Rent is the geometric mean advertised rent, and price the geometric mean Land Registry sold price, for the area and year shown. The yield is the ratio of the two. It is a gross yield on unmatched samples, so it is indicative of the level and reliable about differences between places, not a valuation input.
Sold-price coverage falls away after 2022 in the underlying extract, so base years stop there. Rents run to 2025 and are used for the comparison above.
The search-demand figures are biennial, which is why the actual demand change offered is measured between odd years.
The stock and construction quadrants have no counterpart in this data. They are generated from the parameters you set, which is exactly the distinction the lecture draws between a parameter and an observation.
This is residential data used to teach a framework first written for offices. The structure carries over; the demand driver is households rather than office employment.