Estimating underlying demand for housing is an inexact science, because people’s preferences about the number of people per household, or the proportion of vacant dwellings, can never be observed. At a nationwide level, our longstanding model of underlying demand for new housing is premised on the theory of revealed preferences: that over the medium-term, average outcomes for occupancy or vacancy rates will reflect people’s preferences.

This analysis takes a step away from those assumptions,1 and instead looks at what demographics might tell us about how the supply of housing compares to demand.

From the early 1970s until the Global Financial Crisis, the occupancy rate (number of people per occupied dwelling) was closely approximated by allowing one home for every two people aged between 20 and 79, and one dwelling for every person aged 80 or over. But from 2009, the relationship broke down (see Chart 1).

The number of people per household trended upwards again, lifting from 2.678 to 2.752 by early 2020. At first, between 2009 and 2011, we put the change down to the economic downturn – when the economy is weak, young people keep living with their parents for longer, people are more likely to take on boarders, and two-family households also become more common.

But as the divergence between the expected occupancy rate and actual occupancy rate persisted and then widened further even as the economy improved, it became clear that something else was behind the change.

Of course, this period coincided with a significant worsening in housing affordability and, in the second half of last decade, a rapidly expanding waiting list for social housing. Housing became more expensive relative to both incomes and other goods and services, effectively encouraging people to purchase less of it.

This pressure manifested itself through the shrinking size of new sections and new dwellings, as well as a greater number of people per household, effectively sharing these higher housing costs.

Chart 2 shows how the size of the dwelling stock might differ if instead, the simple relationship outlined above between the age profile of the population and the number of occupied dwellings had held true over the last 30 years. Between 1991 and 2008, by this standard there was sometimes an undersupply of dwellings and sometimes an oversupply of dwellings, but never more than 15,000 homes in either direction across the country. However, between 2008 and 2020 the relative supply position only moved in one direction, as continued strong population growth meant that the construction industry was never able to catch up for the period of very weak building activity in the aftermath of the Global Financial Crisis.

The closure of the borders between 2020 and 2022 saw a collapse in population growth, at the same time as very low mortgage rates and a booming housing market drove residential consent numbers to a record high. This combination saw a reduction in the housing undersupply of as much as 60,000 homes over a 2½-year period.

Because the occupancy rate has been so far away from the above simple demographic modelling for so long, our current working assumption of the occupancy rate is different. We instead allow one dwelling for every 2.037 people aged between 20 and 85, and one dwelling for every person aged 86 or over. The differences might sound small, but they effectively allow for 90,000 fewer households due to more young people living with their parents, more people with boarders, and increased life expectancy meaning that people are widowed later in life.

As Chart 3 shows, these assumptions would not have been appropriate during the 1990s or 2000s, because they would have implied a longstanding oversupply of around 50,000 dwellings.

However, they provide a better indication of the undersupply or oversupply during recent years given current land and house prices (more on that qualifier in a moment). In particular, population growth averaging 0.4%pa between March 2020 and September 2022, coupled with 95,900 new homes added to the dwelling stock, saw a reduction in the housing undersupply of almost 68,000 homes. Since then, faster population growth (averaging 2.7%pa) has seen the undersupply increase again, by 26,400 homes.

CoreLogic’s quarterly index shows that house prices fell 15% between December 2021 and June 2023, and there was less of a sense of demand stress within the housing market. However, even with this house price correction, it seems implausible that the dwelling stock got into a position of oversupply during 2022, as suggested in Chart 3. Long-term comparisons indicate that house prices and debt-servicing costs are still highly elevated compared to incomes.

Instead, we believe that there was an oversupply of housing in 2022, given how far house prices had risen in 2021 and the evolving expectation in 2021/22 that mortgage rates would need to rise substantially. If more affordable conditions were to eventuate (for example, house prices were 30-40% lower so that house-price-to-income ratios were in line with their average of the last 40 years), actual demand for housing would be significantly and sustainably higher. This hypothesis lines up with recent statements by the Minister of Housing, Chris Bishop, who stated that house prices need to fall because “average house prices to the average household income are too high by any objective measure. They are severely unaffordable by international standards.” In other words, there is latent demand for housing from people that will remain hidden while housing remains severely unaffordable.

The government has indicated that improving the supply of housing is one of its major priorities, through addressing zoning regulations, reforming the Resource Management Act, and providing funding for key infrastructure around housing developments. In our view, freeing up the supply of land is the crucial element in addressing the housing crisis and making housing more affordable again. However, we’re also conscious that neither of the most recent National-led or Labour-led governments over the last 15 years have made any sustained headway in slowing the upward march of house prices, let alone achieved a lasting improvement in affordability metrics.

For now, our conclusion is that the nationwide undersupply of housing is somewhere between the 20,400 and 113,000 homes indicated by our two sets of modelling assumptions. Most critically, though, this undersupply has worsened over the last 18 months as population growth has accelerated ahead of the supply of new housing.

1 These assumptions underpinned our work in 2019 that suggested the nationwide undersupply of dwellings could be as small as 2,500 homes, rather than the 130,000 estimated by Kiwibank. As our new analysis shows, our 2019 figure is likely to have been too low, even considering the suppression of demand for housing by unaffordably high house prices. However, we note that our 2019 work also showed considerable variation by region, with the biggest housing supply issues concentrated in Auckland (which, along with Queenstown-Lakes, is probably still the case).