The Urban Land Market: Location, Rents, and Prices

Why does a home near the city centre cost more than an otherwise identical home farther away? One explanation is that the central home saves its resident time and money on travel. Urban land economics asks how that saving becomes rent, how a city decides where to stop expanding, and why expectations about tomorrow can change land prices today.

Learning goals

Explain a rent gradient without jargon, separate housing rent from land rent, calculate a city boundary, and distinguish current rent from the value of future growth. Then explain when different households outbid each other for central land.

1. A deliberately simple city

Imagine a circular city where everyone works at one central station. This employment centre is the Central Business District (CBD). Every household has the same income, preferences, and fixed lot size q. Homes have identical structures. Commuting costs increase linearly with distance, and households can move freely. At the edge, urban land competes with farming.

This is a Ricardian monocentric-city model: differences in location generate differences in land rent. It isolates one mechanism. Real cities can have several employment centres, uneven amenities, variable density, transport networks, and restrictions on building.

Symbol	Meaning	Units used in the worked town
d, b	Distance from CBD; city boundary	Kilometres
q	Land per household	Square kilometres
N	Number of households	Households
V	Fraction of the circular area available for housing	Between 0 and 1
k	Annual commuting cost per unit of distance	Currency per household per year per kilometre
R(d), c	Annual housing rent; annualised structure-service cost	Currency per home per year
r(d), r_a	Annual urban land rent; agricultural land rent	Currency per square kilometre per year
i, g	Continuous discount rate; city-radius growth rate	Per year

Treat N as households, rather than people, when q is a lot per household. If N counts people, convert through household size first. Keep area units consistent: a boundary in miles produces square miles, so q in acres must be divided by 640 in the area equation. The model works with either unit system, but not with a mixture.

2. The indifference principle: pay the landlord or pay for travel

If identical households could obtain the same home with lower total housing-and-travel cost at one location, everyone would want it. Competitive bidding raises its rent until that advantage disappears.

At the border b, housing rent covers the agricultural opportunity cost of the lot plus the structure:

R(b) = r_a q + c.

At an interior location d, residents save k(b - d) in annual commuting costs relative to the edge. Landlords can collect that saving as location rent:

R(d) = R(b) + k(b - d).

Equivalently, R(d) + kd = R(b) + kb everywhere. Lower housing rent farther out compensates for higher travel cost. This equality is the economic argument behind the formula.

A worked town

Let b = 10 km, q = 0.001 square km per household (1,000 square metres), r_a = 1,000,000 per square km per year, c = 9,000 per home per year, and k = 500 per year per kilometre. Thus the lot’s agricultural rent is 1,000 and edge housing rent is 10,000.

Distance d	Housing rent R(d)	Annual commuting cost kd	Combined annual cost
0 km	15,000	0	15,000
5 km	12,500	2,500	15,000
10 km	10,000	5,000	15,000

The central home is not a better structure in this example. Its resident pays 5,000 more rent to avoid 5,000 of commuting cost. Moving one kilometre outward reduces annual rent by k = 500. The housing-rent gradient therefore has slope dR/dd = -k.

Teach it back: Why is the cheaper home at the edge not automatically the better bargain?

Self-check: compare the complete cost

Its rent saving is exactly offset by additional travel costs under these assumptions. If travel costs, amenities, or house quality differ from the model, the comparison can change.

3. Housing rent includes a building; land rent is the residual

Housing rent pays for both the structure and its location. To find land rent, subtract annual structure cost and divide the remaining rent by lot area:

r(d) = [R(d) - c] / q = r_a + k(b - d) / q.

At d = 5 in our town, the home rents for 12,500. Subtract c = 9,000 to leave 3,500 per lot per year. Divide by q = 0.001 to get r(5) = 3,500,000 per square km per year. That large number is a per-area measure, not the household’s rent bill.

The land-rent slope is dr/dd = -k/q. Housing rent falls 500 per home per year for each kilometre outward; land rent falls 500,000 per square km per year for each kilometre outward.

Smaller lots fit more households on each unit of land, concentrating their willingness to pay for access. Thus, holding k fixed, smaller q steepens the per-area land-rent gradient. It does not steepen the housing-rent slope, which remains -k. When lot size changes, the city boundary and rent levels can change too; do not confuse a slope comparison with a complete equilibrium comparison.

Teach it back: Why can two identical housing-rent slopes correspond to different land-rent slopes?

4. Where does the city stop?

Households need Nq of residential land. A circle of radius b has area pi b squared. If a uniform fraction V is available, the area balance is:

Nq = pi b^2 V, giving b = sqrt[Nq / (pi V)].

For example, with N = 100,000 households, q = 0.001 square km, and V = 0.5, the boundary is approximately 7.98 km. The town needs 100 square km of usable residential land, inside a circular envelope of 200 square km.

Change, holding other inputs fixed	Boundary effect	Rent implication in the fixed-lot model
N increases	b rises with the square root of N	At an unchanged interior d, larger b raises location rent
q decreases	b falls	Housing-rent slope remains -k; land-rent slope becomes steeper
V decreases	b rises	Less usable land requires a larger envelope and raises rent at unchanged interior d

Quadrupling N doubles b, rather than quadrupling it. Halving q multiplies b by sqrt(0.5), approximately 0.707. Halving V multiplies b by sqrt(2), approximately 1.414.

A larger radius when V falls does not mean more land becomes usable. It means households spread over a larger envelope to obtain enough usable parcels. This simplified V assumes a uniform buildable fraction; a coastline, mountain barrier, or zoning map can have a more complicated geometry. If outward expansion is impossible, the fixed-q model needs modification.

At fixed k, population growth shifts the housing-rent curve upward without changing its slope. With a smaller q, housing rent at a fixed common interior location falls when r_a, c, and k are fixed, because both r_a q and b fall. The level of per-area land rent need not move in the same direction everywhere: a steeper slope alone does not establish that every parcel becomes more valuable.

Teach it back: How can a geographically constrained city have a larger radius but less usable land within each ring?

5. Prices buy future rents

Rent is payment for current use. Price capitalises expected future net benefits. Two homes can have different price-to-rent ratios if their future rent paths differ.

Assume deterministic population growth N(t) = N(0) exp(2gt) with fixed q and V. Since radius is proportional to the square root of population, b(t) = b0 exp(gt). Thus g is the radius growth rate; population grows at rate 2g.

For an already developed home at fixed d no farther out than b0, assume constant k, r_a, q, and c, and a constant continuous discount rate i > g, with i positive. Then:

R(t,d) = r_a q + c + k[b0 exp(gt) - d].

The home’s present value is the integral of R(t,d) exp(-it) over future time. Integrating the constant terms gives their value divided by i; integrating exp[(g-i)t] gives 1/(i-g). Therefore:

P0(d) = (r_a q + c - kd)/i + kb0/(i-g).

An equivalent decomposition is:

P0(d) = r_a q/i + c/i + k(b0-d)/i + kb0g/[i(i-g)].

The four terms are agricultural lot value, structure value, current location value, and the premium for future city expansion. The growth term has denominator i(i-g), with no extra factor of i. That corrects the transcription in the supplied notes and makes the decomposition consistent with the price-to-rent formula below.

Here the cash-flow convention abstracts from taxes, operating expenses, depreciation, and replacement complications. The structure term assumes a maintainable perpetual service stream with costs treated consistently. An actual valuation must use net cash flows rather than blindly capitalising gross rent. Also, i here is a discount rate, not automatically the observed current-income cap rate.

A price example

Return to the 10 km town and use i = 5% and g = 1% per year. These imply population growth of 2% under continuous compounding. The growth premium at every currently developed location is:

500 x 10 x 0.01 / (0.05 x 0.04) = 25,000.

Location	Current annual rent	No-growth value R0/i	Value with growth	Price/current-rent multiple
Centre, d = 0	15,000	300,000	325,000	21.67
Midpoint, d = 5	12,500	250,000	275,000	22.00
Current edge, d = 10	10,000	200,000	225,000	22.50

The edge home is cheaper in absolute terms but has a higher multiple. The same 25,000 growth premium is a larger share of its lower current rent and base value.

6. Why the edge can have a lower current yield

Dividing the value equation by current rent gives:

P0(d)/R0(d) = 1/i + kb0g/[i(i-g)R0(d)].

With g = 0, the multiple is 1/i everywhere. With positive g, the growth term is larger relative to current rent at outlying locations. The corresponding current-income yield, R0/P0, is lower there. Do not mistake that lower initial yield for a lower discount rate: i is the same throughout this example.

As g approaches i from below, the value becomes extremely sensitive to growth. If g is at least i, the assumed infinite-horizon integral does not converge. The sensible response is to reconsider perpetual growth assumptions, not to interpret an infinite model value as a market price.

Teach it back: Can an expensive central home have a lower price-to-rent multiple than a cheaper edge home?

Self-check: price level versus multiple

Yes. Our central home costs 325,000 at a multiple of about 21.67, while the edge home costs 225,000 at 22.50. Absolute price and price relative to current income answer different questions.

7. Vacant land: buying the future arrival of the city

For undeveloped land beyond today’s boundary, d > b0, the expanding city reaches the parcel when b0 exp(gT) = d. For g > 0:

T = ln(d/b0)/g.

At d = 12 km, b0 = 10 km, and g = 1%, development arrives after approximately 18.23 years. At g = 2%, it arrives after about 9.12 years. This illustrates why changing growth expectations can matter even when current agricultural income is unchanged.

Let p0(d) be the price per unit of vacant land area. The parcel earns agricultural rent until T. At T, its developed land value contains both agricultural opportunity value and future location-rent growth. Discounting both pieces gives:

p0(d) = r_a/i + exp(-iT) x [k b(T) g / (i(i-g)q)].

Since b(T) = d, the second term can also use kd in place of k b(T). This expression values land, so it excludes a house’s structure value. It assumes deterministic outward growth, conversion at the edge, and no additional conversion frictions beyond the model’s structure-cost treatment.

To see why agricultural value appears only once, add agricultural income before T, r_a[1-exp(-iT)]/i, to the discounted agricultural component after T, exp(-iT)r_a/i. Their sum is r_a/i. The remaining term is the development-related premium.

With g = 0, land outside a fixed boundary is never reached in this model and remains worth r_a/i. For positive growth, fringe land can be very sensitive to revisions in g because both development timing and future location income change. Calling it universally the “most volatile” land goes beyond this deterministic model: measured volatility also depends on uncertainty, finance, planning, and which locations are compared.

8. Who wins central locations?

Now introduce two household groups with different commuting costs. With equal fixed q, a group with higher k loses more from every extra kilometre, so it is willing to bid more aggressively for central access.

For example, if k1 = 800 and k2 = 400, moving five kilometres inward saves Group 1 an annual 4,000 but Group 2 only 2,000. Under the standard two-group equilibrium, with both groups present and otherwise suitable assumptions, the higher-k group occupies the inner ring and the lower-k group the outer ring. Their land bids meet at an internal boundary m.

With an available fraction V and equal lots, the inner group’s area condition is N1 q = pi m^2 V. At m, neither group can profitably displace the other at prevailing rent. This is the highest-bid principle: competitive allocation goes to the user willing and able to pay most for a parcel, not necessarily the use with the greatest social benefit.

If groups have different lot sizes, compare k1/q1 with k2/q2, their per-area bid-rent slopes. Higher commuting cost alone no longer establishes who occupies the centre. Income, housing consumption, amenities, discrimination, credit access, and regulation can also affect observed sorting.

The model shows that sorting can arise without zoning. It does not show that actual segregation is caused only by voluntary competition or that policy and historical exclusion are irrelevant.

Teach it back: Why might someone who values time highly still choose a suburban home if they also demand a much larger lot?

9. Transport can change the value of distance

Separate commuting cost per distance into money cost and time cost. A simple expression is k = M + wH/v, where M is annual monetary cost per kilometre, w is the value of an hour, H counts annual commute legs, and v is speed. Faster travel reduces the time-cost component when other inputs stay fixed.

The supplied historical comparison uses walking at about 3 mph and cars at about 25 mph: roughly an eightfold speed difference. Eight times the distance at eight times the speed takes the same time. This arithmetic explains how better transport can offset some effects of outward expansion.

It does not establish that transport completely prevented historical real house-price growth. Congestion, fares, fuel, household incomes, amenities, construction costs, and changing density also matter. Nor does an eightfold reduction in travel time cost imply an eightfold reduction in total k when money costs remain.

In this closed, fixed-N, fixed-q city, lower k flattens rent gradients while the area equation leaves b unchanged. To explain transport-induced suburban expansion, allow lot sizes, population, or land availability to respond. State what is allowed to change before telling the historical story.

Feynman study lab

Explain it simply

Imagine identical homes arranged around one workplace. The nearer home saves travel, so residents bid up its rent. Subtract the building’s service cost to find what belongs to the land. More households require a larger developed area. Prices also depend on future expansion, so a parcel can become valuable before the city reaches it.

Teach it back without looking

Close the notes and explain the worked town in two minutes. Include one household’s rent-plus-travel budget, the distinction between a home’s rent and land rent per unit of area, and why a price is forward-looking. Then answer:

Why can a house at the urban edge have a higher price-to-rent multiple than an identical central house even though it has a lower price?

Check your explanation after you try

In this growth model, each currently developed location gets the same absolute growth premium kb0g/[i(i-g)]. At the edge, current rent is lower, so that premium adds more to the price-to-current-rent ratio. The discount rate has not changed across locations.

Transfer questions

1. Using the 10 km town, calculate housing rent, land rent per unit of area, and commuting cost at d = 8 km.
2. If households quadruple with q and V fixed, what happens to b?
3. If k doubles at fixed q and b, how do the two rent slopes change?
4. If a parcel is at 1.5 times today’s radius and g = 2%, when is it reached?
5. What assumption fails in the perpetual-value calculation when g is at least i?
6. If Group 1 has twice Group 2’s k and three times its q, which has the steeper per-area land bid?

Check the transfer questions

1. Housing rent is 11,000; land rent is (11,000-9,000)/0.001 = 2,000,000 per square km per year; commuting costs 4,000. Housing plus travel remains 15,000.
2. The radius doubles.
3. Both double in magnitude: housing slope becomes -2k, and land slope becomes -2k/q.
4. T = ln(1.5)/0.02, approximately 20.27 years.
5. Discounting no longer dominates the assumed perpetual growth, so the integral diverges.
6. Group 2. Group 1’s k/q is only two-thirds of Group 2’s. Comparing commuting costs without comparing lot demand would mislead.

Find the gap and try again

Identify the exact step you could not explain: compensation for travel, dividing by lot size, square-root city growth, or discounting future income. Rework that example, then close the notes and explain it without formulas. Change one assumption, predict the effect, and check your prediction. Tomorrow, rebuild the rent equation and price decomposition from memory before looking.