Housing Markets: Turnover, Prices, User Cost, and Supply

A city can gain very few new households while millions of existing households move, buy, sell, or change tenure. A house can also become more expensive to purchase while appearing cheaper to own in a model that assumes large future capital gains. Understanding housing markets requires separating these different quantities rather than treating sales, demand, prices, and affordability as interchangeable.

Learning goals

Distinguish gross turnover from net household growth, explain vacancy and sales duration, compare price indices with transaction medians, and calculate a simplified ownership user cost. Trace how expectations, credit, construction, and maintenance can reinforce or dampen market changes.

1. Housing services and housing assets are different markets

A tenant pays to use a home. An owner-occupier owns the asset and consumes its housing services. Both need shelter, but only one necessarily makes a property purchase. Ownership demand can therefore change without an equivalent change in the total number of households needing homes.

Think of a theatre: the building is an asset, while an evening’s performance is a service. Selling the theatre to another investor does not create more seats. Likewise, selling an existing home does not by itself add to the housing stock.

Tenure means the arrangement through which a household occupies housing, such as renting or owning. Changes in incomes, household composition, access to credit, preferences, and the relative costs of renting and owning can change tenure choices.

2. Gross flows can dwarf net growth

The supplied historical lecture snapshot reports about 105.5 million US households, roughly 69.8 million owner households and 35.6 million renter households, an average household size of 2.62, more than 7.4 million renter moves, and about 5 million annual single-family sales. These are historical lecture figures, not current estimates, and their rounding and measurement bases differ.

There is also an arithmetic distinction: 0.83% of 105.5 million is about 0.876 million, while 0.83 million divided by 105.5 million is about 0.787%. Treat the supplied growth count and percentage as approximate figures rather than exact equivalents.

A small-city example

A city starts with 100,000 households and gains 1,000 net households during a year. During that same year, 8,000 existing households move within the city. Those internal moves can create substantial demand for viewings, removals, leases, purchases, and temporary vacancies without adding 8,000 net households.

Measure	What it tells you	What it does not establish
Net household growth	Change in households needing accommodation	Total moves or home sales
Moves	Gross changes of residence	New households or new construction
Property sales	Asset transactions	Occupancy changes one-for-one
Tenure transitions	Moves between owning and renting	Equivalent growth in total housing demand

Do not simply add renter moves and home sales to obtain unique moving households: their populations, units of measurement, and overlap must be checked. A move chain can connect several transactions, and an investor purchase may leave the tenant in place.

Teach it back: How can a city with only 1% household growth have a busy estate-agency market?

3. Moving creates a timing problem

An existing homeowner often needs to sell one home and buy another. Buying first provides certainty about the next home but can create overlapping mortgage payments, taxes, insurance, and maintenance. Selling first reduces overlap but can require temporary accommodation or storage.

Bridge financing can span the funding gap, but adds cost and exposes the household to sale delays. Suppose extra carrying costs are 2,000 per month. An unexpected three-month delay costs 6,000, before price changes, financing fees, or other differences. Longer selling time can therefore weaken demand from households already holding an unsold home.

Demographics affect this process. The supplied rental propensities of about 56% for single households and 22% for married couples are historical lecture comparisons, not fixed behavioural laws. Age, income, family structure, cohort experience, and local housing institutions interact. Population ageing does not mechanically imply a specific ownership rate or maintenance choice.

Self-check: does every sale increase occupied housing?

No. Existing owners can exchange homes, investors can buy occupied rentals, and a sale can be followed by temporary vacancy. Net occupancy growth and gross sales measure different processes.

4. Vacancy and selling time reveal market frictions

Vacancy is an unoccupied part of the housing stock under a specified definition. Time on market measures the marketing duration of listed property. A homeowner can live in a home while selling it, so a slow sale does not necessarily create a vacant unit. Conversely, not every vacant unit is actively for sale.

The supplied lecture contrasts historical vacancy and sales duration across cities such as Boston, Minneapolis, Phoenix, and Dallas. Its ranges of about 1-2% homeowner vacancy and 6-9% multifamily rental vacancy are historical context, not universal current bands. Owner and rental vacancy measures use different eligible inventories; compare their definitions before comparing percentages.

One hypothesis posits a relatively stable structural vacancy rate, V*: normal empty space needed for search, matching, renovation, and moving. More flexible prices help clear demand around that benchmark. A sticky-price hypothesis instead allows unexpected shocks to show up more strongly in vacancy, listings, or transaction time while prices adjust slowly.

Both are simplifying explanations. Residential markets can also have sticky asking prices, lengthy searches, credit constraints, and slow supply. Commercial markets can adjust through concessions as well as quoted rents. Stable observed vacancy alone does not prove instant price adjustment or unusually disciplined construction.

Why a small demand change can matter

In an illustrative stock of 10,000 eligible units, 200 are vacant and 9,800 occupied. If 100 additional households occupy existing empty units before construction or other changes, vacancy falls to 100: 2% becomes 1%. A demand change of only 1% of stock has halved the vacancy buffer. Search becomes harder even though the aggregate percentage change looks small.

Potential household formation is also not identical to realised occupancy. Prices, incomes, credit, and sharing decisions affect whether people establish separate households. Distinguish desired demand before adjustment from transactions and occupancy observed afterward.

5. Price, housing quality, and expenditure

For a particular bundle, write E = P x Q, where P is price per unit of quality-adjusted housing and Q is the quantity or quality consumed. E is expenditure on that bundle. In this section it refers to purchase expenditure when the bundle is an asset, not automatically an annual housing expense.

Suppose quality-adjusted price rises 10% and the bundle becomes 20% larger or better. Expenditure rises 1.10 x 1.20 - 1 = 32%, not 10%. For small changes, expenditure growth is approximately price growth plus quantity growth; the exact calculation also includes their interaction.

A median transaction price measures the middle sale in that period’s sample. It changes with both underlying prices and the mix of homes sold. If only larger homes transact this month, the median can rise even when each individual home’s value is unchanged. It is therefore neither a pure constant-quality price index nor a direct measure of every household’s expenditure change.

Approach	How it tries to isolate price	Remaining issue
Hedonic index	Controls observed attributes and estimates period effects	Omitted quality, specification, and changing attribute values
Repeat-sales index	Compares sale prices of properties observed more than once	Renovation, depreciation, transaction selection, and excluded new homes
Transaction median	Summarises actual sale prices	Strong sensitivity to the changing sales mix

In a log-price hedonic regression, a time-dummy coefficient b implies a fitted price ratio exp(b), with an exact percentage difference of 100[exp(b)-1] relative to the omitted period, conditional on the model. Repeat sales, including Case-Shiller-style methods, reduce some composition problems but do not make quality literally immutable.

Teach it back: Can the median sale price rise while a constant-quality index stays flat?

6. User cost: owning combines consumption and investment

Owning provides shelter and exposure to asset gains or losses. A stripped-down annual user-cost equation from the notes is:

u = PQ[i(1-t) - a_e], and U = u/Q = P[i(1-t)-a_e],

where P is price per housing-quality unit, Q is the quality quantity, i is a nominal financing or opportunity-cost rate, t is an assumed effective marginal tax benefit on that interest component, and a_e is expected price appreciation, not a guaranteed future change.

The supplied interpretation of t relies on an interest-deductibility assumption. Actual eligibility, limits, itemisation, financing structure, and jurisdiction can differ. Treat this as a model parameter rather than assuming every buyer receives a tax deduction. For a fully financed stylised purchase, i can represent the mortgage rate; otherwise debt and equity need their respective costs and tax treatment.

A numerical example

Let total property value PQ = 500,000, i = 6%, t = 25%, and expected appreciation a_e = 2%. The simplified user-cost rate is 6%(1-25%)-2% = 2.5%, giving annual user cost of 12,500.

If expected appreciation rises to 4%, that model cost falls to 2,500. Yet the house still costs 500,000, the interest bill does not disappear, and the buyer still needs funding. Expected capital gains generally do not pay this month’s mortgage instalment.

User cost is not the mortgage payment or a complete affordability measure. Principal repayment builds equity rather than being pure consumption expense, but it is still a cash outflow. A household may face down-payment, debt-service, and borrowing limits even when a model gives a low user cost.

7. What the short formula leaves out

A broader stylised rate includes property tax, maintenance, depreciation, insurance, risk, and relevant transaction costs. Write, for illustration:

Annual user cost approximately = PQ[financing/opportunity cost + recurring cost rates + risk adjustment - expected appreciation] + annualised transaction costs.

Avoid double-counting: whether appreciation is measured before or after physical depreciation affects how depreciation enters. Debt and equity financing costs must also be weighted consistently, and nominal rates should be paired with nominal appreciation expectations.

Adding 2% of recurring costs and risk allowance to the first example raises annual cost from 12,500 to 22,500, before separately treated transaction costs. The omitted components can be material.

With equal-quality services and the necessary market assumptions, annual rent can be compared with ownership user cost. This suggests a model relationship P/rent approximately equals 1/user-cost rate when the rate is positive and all inputs are consistently defined. It is not automatically a cap-rate identity or a dependable trading rule.

A negative calculated user cost under very optimistic appreciation does not mean ownership is free, safe, or scalable without limit. It signals how strongly the result depends on expectations and on omitted risks, constraints, and costs.

Teach-back self-check: why is expected appreciation different from income?

It is an uncertain change in asset value. It may not occur, can reverse, and may not be available as cash without a sale or additional borrowing. It can influence the economic appeal of ownership while leaving immediate payment obligations unchanged.

8. Expectations and credit can reinforce a boom

In a positive-feedback story, recent gains raise expected appreciation; lower perceived user cost encourages buying; stronger demand can shorten marketing time and raise prices; new gains then reinforce expectations. A dynamic model and evidence are needed to establish a bubble, rather than merely noting that prices increased.

Affordability constraints, revised expectations, weaker income, tighter lending, or new supply can interrupt the process. A reversal can work in the other direction: lower expected gains raise perceived ownership cost, delaying purchases and lengthening sales chains.

Credit expansion is best understood as a possible amplifier, not itself a mechanical bubble-popping event. Easier finance can let constrained buyers bid more; tighter credit or repayment stress can weaken that demand. The supplied 2000-2006 subprime discussion is historical lecture context, not a current lending assessment or proof that credit alone explains the housing cycle.

A higher price-to-rent ratio may reflect expectations, financing conditions, risk, taxes, or changing fundamentals. It can be a warning indicator without uniquely identifying mispricing. A falling ratio likewise does not prove that every property is affordable to every household.

Teach it back: How can easier borrowing increase asset demand without immediately adding housing services?

9. Supply moves through a pipeline

New homes move through stages: starts, construction in progress, and completions. A start is not a completed, occupiable home. Project delays, cancellations, and construction duration affect when new supply arrives.

For example, if 100 homes enter construction, 20 already under construction are completed, and there are no cancellations or other adjustments, the under-construction inventory rises by 80. Completions add finished units; net usable stock also depends on demolition, conversions, and losses.

Supply responsiveness differs by place. Land availability, infrastructure, finance, labour, materials, and regulation influence the response. In constrained markets, demand can raise prices more than construction; in more responsive markets, additional stock can moderate pressure after a lag. The four-quadrant model links this adjustment to rents and asset prices.

The supplied notes describe monopolistic competition among differentiated homebuilders, larger builders in higher-volume markets, and more numerous smaller builders under fragmented planning. These are lecture hypotheses or empirical patterns, not universal industry identities. Local scale economies, acquisition networks, competition, and regulatory processes require evidence before making a causal claim about firm size.

10. Existing homes also supply future housing services

Maintenance preserves service quality. Remodeling can improve quality or add floor area without creating a new dwelling. Increases in housing quality and increases in dwelling count therefore need separate measures.

An owner should compare the expected discounted benefit of an improvement with its cost, including additional services, resale value, disruption, and financing. Suppose a repair costs 10,000 today and creates benefits worth 12,000 in present value: its net benefit is 2,000 under those assumptions. Higher expected property appreciation does not automatically make every project worthwhile, and a low purchase price alone does not establish high returns to renovation.

The supplied figure that major improvements equal roughly 30% of new-construction volume is a historical lecture comparison requiring a consistent date and spending definition. It is not a current fixed ratio. Likewise, lower average maintenance spending among elderly owners, if observed in a sample, can reflect income, liquidity, expected tenure, preferences, or property mix. It does not mean every older household deliberately consumes its equity or neglects repairs.

Feynman study lab

Explain it simply

Housing is both shelter and an asset. Most transactions rearrange who lives in or owns existing homes, while net household growth changes how many homes are needed. Sale prices mix market changes with the quality of the homes sold. Ownership cost includes financing, upkeep, and uncertain gains. Credit and expectations can move demand quickly, but building and maintaining usable housing take time.

Teach it back without looking

Explain the 100,000-household city and the 500,000 ownership example aloud in two minutes. Define turnover, vacancy, constant-quality price, and expected appreciation. Then answer:

Why does a higher expected capital gain lower modelled user cost without necessarily making the house easier to finance?

Check your explanation after you try

Expected appreciation offsets part of the economic cost in the model, but it does not reduce the purchase price, guarantee a gain, or remove current payments and borrowing limits. Economic valuation and immediate cash affordability are different questions.

Transfer questions

1. A city gains 500 net households but records 6,000 internal moves. Are those 6,000 additional households needing newly built homes?
2. Quality-adjusted price rises 5% and the purchased bundle’s quality rises 10%. How much does expenditure rise?
3. A property costs 400,000, i = 5%, assumed effective t = 20%, and expected appreciation is 1%. What simplified annual user cost results?
4. Add recurring costs of 1.5% of property value to that calculation. What is the revised annual amount?
5. Of 20,000 eligible units, 400 are vacant. If 100 become occupied with stock unchanged, what is the new vacancy rate?
6. Why can a repeat-sales index still be affected by quality change?

Check the transfer questions

1. No. Internal moves can redistribute existing homes without adding net households.
2. 15.5%, since 1.05 x 1.10 = 1.155.
3. 12,000: 400,000 x [0.05 x 0.80 - 0.01].
4. 18,000, adding 6,000 of recurring costs.
5. 1.5%, since 300/20,000 = 0.015.
6. The same property can be renovated, extended, or depreciate between sales; repeat identity does not guarantee unchanged quality.

Find the gap and try again

Record where you hesitated: gross versus net flows, vacancy versus listing duration, price versus quality, or user cost versus cash payment. Rework that example, then explain it without formulas. Change one expectation or credit assumption and predict what can adjust immediately versus slowly. Tomorrow, reconstruct the user-cost argument and name three costs omitted from its shortest form.