Almost no one lives in the house of their dreams. We rent the flat we can afford, not the one we'd choose. We buy further out than we wanted, or smaller than we planned, or later than we meant to. The spare room becomes a bedroom. The study becomes a nursery and the nursery stays a nursery. The adult child who would have moved out is still down the hall. We tell ourselves the commute is fine. We build a granny flat to accommodate ageing parents. Whether we rent or buy, almost all of us are living in a compromise, and we made it because of what housing costs - be it the purchase price, the mortgage repayment cost, planned renovation costs, maintenance costs or rental costs.
That ordinary, universal experience is the reason a claim now circulating cannot bear the weight being put on it. The claim, in its various forms, is that we are building enough houses, or that at some lower rate of population growth the current rate of construction would be sufficient. The evidence offered is usually a ratio: dwellings against households, or population growth against completions, or population per dwelling with the conclusion that if the two roughly match, or match some historical figure, then demand has been met and there is no real housing shortage.
I am not going to argue the opposite. I am not going to tell you there is a shortage of a particular size, because I do not think anyone can honestly tell you that. I am going to argue something narrower, and I think harder to dismiss. You cannot cleanly conclude adequacy from a ratio like that, in either direction, because the quantities in it are not independent of the thing you are trying to judge. And what we most need to know, how much housing demand has been compromised away, and at what price it would reappear, is not in the data at all.
How we normally read a market
Demand responsiveness, in an ordinary market, reveals itself through a decision you can see. The price of something rises. Some people keep buying it. Others decide it is not worth it and stop. The quantity that changes hands moves, and that movement is the signal. When coffee gets dear and people buy less of it, the market has told you something measurable about how much people wanted coffee at the margin. The purchase that does not happen is itself a recorded fact, because the absence shows up in the volume.
This is the implicit model behind almost every confident statement about whether we have enough housing. Count the dwellings. Count the households. Compare. If they roughly match, the market has cleared, and clearing means demand was met. If there were a real shortage, the reasoning goes, it would show up as households without dwellings.
The trouble is that housing is not coffee, and the difference is not a detail. It is the whole argument.
Housing has no walk-away
Everyone needs shelter. Almost nobody responds to expensive housing by consuming no housing at all. Homelessness is avoided; people work hard, often desperately, to stay housed. The buy-or-don't-buy decision that reveals responsiveness in an ordinary market is, for housing, essentially closed. You cannot opt out of being housed the way you can opt out of a coffee.
So the responsiveness does not disappear. It has to go somewhere, and the only place it can go is into the kind of housing people end up accepting: how much, of what quality, in what location, shared with whom, and beginning when. The demand response that an ordinary market would record as "did not buy" is, in housing, expressed as a great many quiet adjustments that nothing writes down. The adult child stays in the childhood bedroom rather than renting. Two people who would each have formed a household share one instead, and form none. The couple who would have moved stay put. The household that would have rented close in rents two zones out, and someone else's compromise cascades behind them. The separation that, a generation ago, would have produced two homes now produces one home and a long stay on a friend's couch.
Every one of these is a demand response to price, exactly as real as the decision not to buy something that got too expensive. And not one of them is a transaction the market records. The household that does not form does not appear anywhere as unmet demand. It appears, if it appears at all, as an extra person in a dwelling that was already counted. The responsiveness is fully present and almost entirely invisible, because in housing it is revealed not through a decision you can see but through a sea of compromises you cannot.
Housing demand, in other words, is enormously elastic. Just not in the way elasticity is usually measured. The measure looks at the one margin housing demand cannot use, and concludes from the quiet there that demand is unresponsive, when in fact the response has simply gone somewhere the measure does not look.
Why economists have had to reach for unusual methods
This is not a new observation, and it is not mine. Housing economists have known for decades that the standard quantity-price approach to measuring demand responsiveness fails for housing, and the careful literature has reached for a series of unusual methods precisely because the textbook one returns the wrong answer.
Some build a richer quantity variable. Instead of counting dwellings, they construct a measure of housing services, hedonically adjusted for size, quality, age and location, so that when households downgrade what they consume under price pressure, that downgrade shows up as a fall in services even though the dwelling count is unchanged. Some study household formation directly, looking at how the rate at which young adults leave the parental home, or at which separating couples form two homes rather than one, moves with local housing costs. Some exploit natural experiments, places where supply was unexpectedly increased by zoning reform, or families whose budget constraint moved exogenously through inheritance or a voucher programme, and watch the formation and consumption responses. Others build structural models in which household formation is itself a choice inside a budget constraint, and use observed choices across people facing different housing costs to recover what people would choose if the constraint were eased.
In Australia, there is a strand of housing economics that studies precisely this: the edges of home ownership, intergenerational housing wealth, the family-level adjustments that affordability forces, and the wellbeing consequences of housing stress. The thread connecting these methods is the same recognition: the response to expensive housing happens at the edges, in the families and the share houses and the wellbeing data and commute times, not at the cash register.
The reach for these methods is the evidence that the textbook one is looking in the wrong place. If the standard measure of housing demand responsiveness gave a sensible answer, no one would need to construct hedonic services indices, study formation rates among young adults, or trace wellbeing impacts through household surveys. The literature reaches because it has to, because the elasticity is real and the standard measure does not see it. And when the careful methods are used, they consistently find what the structural argument predicts: housing demand is materially responsive, just on the margins where ordinary markets do not record their responses.
But the methods that find the responsiveness are not the ones doing the work in the public adequacy debate. The arguments that we are building enough, or would be at lower migration, almost never invoke the formation literature or the natural experiments or the structural models. They invoke a ratio.
Why the ratio cannot certify itself
This is the precise reason a dwellings-to-households, population-to-dwellings, or completions-to-population ratio cannot settle the question. It is not a crude tool that a better one would replace. The problem is structural and it survives refinement.
Take the household-based version first. It compares the dwelling stock to the number of households. But the household count is not an independent fact about demand. It is an outcome of the price of housing, produced by exactly the mechanism above. The number of households that exist is the number that could afford to form, given what shelter cost. Comparing that number to the dwelling stock and concluding that demand was met is circular: you have measured demand after price has already finished suppressing it, and then announced that there was no suppression. A ratio sitting at some comfortable level is not evidence that the level is adequate. It is equally consistent with a market in which the pressure was absorbed, invisibly, as compromise rather than as price or as visible shortage. The ratio cannot tell you which world you are in, because it reads the same in both.
The population-based versions, which are the ratios actually most used in the public debate, fail for a related but broader reason. A population-to-dwellings or completions-to-population ratio looks more robust because population is not as obviously price-endogenous as the household count. People do not refuse to be born when rent is high. But the problem is not that the population number is wrong. It is that the ratio compresses a multi-dimensional demand into a single one-dimensional quantity. What people want from housing is not just a roof. It is enough space, the right location, acceptable quality, autonomy from people they did not choose to live with, the option of forming a household when they are ready rather than when they can afford to. Each of these is a dimension along which households absorb price pressure invisibly. Some by sharing more. Some by living further out. Some by accepting a smaller or older or worse dwelling than they would have chosen. Some by deferring formation, or forgoing it. A dwellings-per-head ratio sees none of this. It assumes there is some right number of dwellings per person, treats hitting that number as success, and is silent on every margin along which the population was forced to consume less housing than it wanted.
Household formation is the most vivid of these margins because it is the one a reader recognises from their own family, and it is real and material, but it is one channel among several. The deeper point is that any ratio of dwellings to people is a one-dimensional summary of a multi-dimensional adjustment. Whatever the mix of compromises in a given year, the ratio cannot see them by construction, because it has nowhere to record them. The numerator counts dwellings and the denominator counts people, and the entire content of "did demand get met" lives in the dimensions the ratio does not have.
It does not help to make the comparison demographically sophisticated, to allow for smaller households, ageing, more people living alone. Those trends are real, and some of that change is genuine preference that would hold at any price. But the patterns you actually observe are the patterns that survived affordability, not the patterns people would choose if housing were cheaper. The demographic inputs you would use to refine the ratio are themselves outcomes of the price the ratio is trying to evaluate. The refinement inherits the same blindness as the thing it was meant to fix.
Nor does it help to anchor to a past year and argue that the market was in balance then, so we need only build back to that. There was no uncompromised year. The adult child in the childhood bedroom was there in 2019, and in 2016. Whatever reference point you choose was itself a price-suppressed equilibrium. It may well have been less compromised than now, that is plausible, but less compromised is not uncompromised, and restoring a past ratio restores a past compromise, not an adequacy.
And the appeal to a past baseline does a second thing that goes unexamined: it assumes that the past was adequate without offering any proof that it was. The argument is almost always implicit. The ratio sat at some value, prices were lower, therefore demand was met. But the same circularity applies to the historical reading as to the present one. Whatever the past ratio looked like, it was the ratio that survived whatever level of compromise was operating then, not evidence that the level was zero. Pointing to it as a target is not an empirical claim about a state of adequacy ever achieved; it is a wish, projected onto a number.
It is a wish that also has to ignore something difficult, which is that what people want from housing has not weakened since. If anything, the demands on the housing stock have grown. Family sizes have fallen, so more dwellings are needed for the same population. People partner later and separate more, so the same lives produce more household-years. Life expectancy has lengthened, so people occupy their homes for longer at the end of life. Expectations about space, location, quality and the option of one's own front door have not regressed. None of this means the past was bad; it means there is no version of the past whose ratio, applied to today's people with today's preferences, would produce adequacy. The reference is wrong on its own terms and wrong relative to what it is being asked to deliver.
The mechanism that hides the shortfall was not switched on by recent events. It is a property of any market for a thing no one can refuse to consume, and it has always been operating.
A fact the simple story struggles with
You can see the shadow of all this in one stubborn observation. Through a period of sharply rising interest rates, borrowing capacity was crushed. The clean account that says prices are mostly a function of what people can borrow predicts that prices should have fallen substantially and stayed down.
That is not what happened. Prices wobbled, paused, and resumed rising while rates were high and capacity was low. A market in which demand can walk away does not behave like that. A market in which demand cannot walk away, and instead re-expresses itself by accepting less and worse housing per dollar, behaves exactly like that. I cannot give you a number for the elasticity, because the paths it travels are many and individually invisible, which is the entire point. But prices holding into a tightening cycle is the aggregate trace of a great many private compromises happening at once. It is the closest thing to a directly visible footprint of a mechanism whose individual steps leave none.
There is a timing asymmetry underneath this worth naming. The compromise margin adjusts in months. A household can consolidate, take someone in, or defer a move almost immediately. Supply adjusts in years: approvals, finance and construction do not respond on the timescale of a demand shock. So whenever demand and price move, the compromise margin necessarily absorbs the first-order adjustment, because it is the only margin that can move in time. The quiet adjustments do the clearing, every time, invisibly, and the next shock tends to arrive before supply has caught up with the last.
The objection I would rather raise myself
Here is the strongest reply to all of this, and I would rather state it than wait for it. If the suppressed demand cannot be measured, then a shortage cannot be asserted either. The argument cuts both ways. By my own logic, "there is a large shortage" is as unsupported as "we are building enough."
On scale, that is correct, and I will not claim a scale. I do not know how large the gap is and neither does anyone else. But the burden of proof is not symmetric, because the measurement error is not symmetric. Suppressed household formation is invisible. The compromises are invisible. The only fully countable failure is the extreme tail, the people with nowhere at all. There is no mechanism that pushes these estimates the other way, no force that makes a dwelling count systematically overstate the shortfall. So two things are true at once: we do not know how large the imbalance is, and we do know which direction the error runs. Together, they put the burden on the adequacy claim. "We are building enough" is the single conclusion this kind of evidence is structurally least able to support, because it is blind in exactly the direction that matters.
You could still answer that rising prices, more people per dwelling, and a broad sense of housing distress are each separately explainable: the rate cycle for one, demography for another, the mood of the times for the third. You could. But that asks for three separate explanations for three things that arrived together, when one explanation covers all of them: demand absorbed as compromise rather than recorded as transaction. Parsimony is not proof. Combined with a measurement error whose direction is certain, it is enough to make adequacy the least credible position in the room rather than the default one.
What we actually don't know
Strip it back and the honest statement is small. There is a preference: for one's own household, for enough room, for a workable location, formed when people want to form it. Some unknown share of that preference has been compromised away by price. We do not know how large that share is. And we do not know at what price it would manifest, what fall in cost would call those absent households out of the spare rooms and the share houses and into being.
Both of those unknowns exist for the same reason. We only ever observe the outcome after price has done its filtering. The filtered-out demand left no record, because the way people absorb unaffordable housing is to need less of it, not to be counted wanting it. This is why claims of adequacy cannot be tested. Not because the current tools are weak, but because the question asks about a quantity that the act of being unaffordable removes from view.
There is a test, in principle, though no one has run it. A sustained fall in real prices or rents (brought on by an increase in supply) should bring latent households out of the woodwork, a surge in formation that no fixed-demand model would predict, occupancy falling faster than demography can explain. Until that happens, the size of what was suppressed, and the price that would release it, remain exactly the things we cannot see.
So when someone says we are building enough, or would be at some lower rate of growth, the useful reply is not a rival number. It is that the comparison being offered cannot answer the question, because one side of it is produced by the very price the question is about, and the thing we would most need to know, how much demand has been quietly given up, and what it would take to get it back, is not in the data and was never going to be. Almost no one lives in the house of their dreams. Until that stops being true, what we are counting is the compromises, not the demand, and a generation is quietly wearing the difference.
No comments:
Post a Comment