A survey of the median house prices around the world has revealed Auckland to be among the five least affordable cities to buy a house. The annual Demographia survey, released today, compares prices to incomes in 367 cities. Auckland is one of the worst in the world due to extremely high house prices coupled with moderate wages.
We’ve often talked about the issues with how Demographia produce their results. They take an overly simplistic view of the discussion, and exclude important factors. But while the scale of the issue is likely wrong, that doesn’t mean the general outcome – that housing affordability needs to be improved – isn’t correct.
We also disagree with their proposed solution of unfettered greenfield development. For Demographia, it seems that opening up greenfield land is always the solution, regardless of the question being asked. While land supply is an issue, they like to conveniently ignore the impact of planning regulations on existing urban land that prevents development across most of Auckland. They also like to ignore the cost to tax and ratepayers of providing the infrastructure needed to enable that greenfield development. For example, based on Auckland Transport’s figures it will cost about $67,000 per dwelling to provide the roads needed in the major new greenfield areas that are proposed.
Many people may want a home with a large backyard on the fringe of town, but many just want a home. A lot are prepared to forgo a large backyard for the added amenity of living closer to the city or other urban centres – but they are unable to do so, as so much development has been restricted.
This brings me to the main point of the post, the media (especially the Herald) who want to have it both ways.
While today they’re lamenting house prices, the Herald has spent much of the last few years championing opposition to one of the key tools that will help address housing supply, the Auckland Unitary Plan. From when the draft plan was released almost three years ago, they’ve given countless space to those opposing any change in Auckland. They’ve deliberately misled the public and recently they’ve even become so absurd as to call two-storey townhouses “Highrise” in their bid to whip up fear and anger over the plan.
Of course politicians of all stripes shouldn’t escape blame. Whether they’re also trying to whip up fear, generally oppose change or just have it happen in some other neighbourhood they are as much to blame. They also seem to me to have less desire to actually fix problems. After all, which of them are really going to stand up to house owning voters and say they’ll enact policies which could result in existing house prices falling or at best stagnating for many years as a result of changes.
Regardless of what you think the solutions are, it still feels like we’re some way off any real changes happening.
On a related note: I suspect we could see John Key include housing announcements in his announcement on Wednesday when he also announces support for the CRL to start in 2018. There have been suggestions the government have been talking to the council and CCOs like Watercare looking at what other big infrastructure projects could be brought forward to help speed up housing supply.
Demographia state that planning regulations – or, more precisely, metropolitan urban limits – cause houses to be unaffordable. Without planning, they claim that the median house would cost no more than 3 times the median household income.
Is this claim actually true? Last week, I took a look at some recent research by New Zealand economists that found that Auckland’s planning regulations had a much lower cost than Demographia think. (But still a substantial cost, especially for apartments.) So that’s evidence against Demographia’s claims.
But to really test the robustness of their analysis, I’m going to attempt what mathematicians call a proof by contradiction. My aim is to show that it cannot be the case that planning is the only factor that is pushing the median multiple above three for standalone houses. In order to prove that I will have to show that Demographia’s claims are mathematically impossible.
Somewhat to my surprise, this is not particularly difficult.
The first step in the proof is to consider the channels through which planning regulations can impact the cost of housing. They can:
Constrain the supply of bare land for development, which will tend to drive up the price of land
Limit the number or size of dwellings that you can build on a site
Impose cost, delay, and uncertainty in the regulatory process, which may scare some people off entirely.
Consequently, I’m going to look at what houses would cost if only construction costs mattered. No land costs, no consenting requirements, or anything else. That way, we can be certain that we’re not accidentally including any costs that arise as a result of planning regulations.
If the cost to physically build an average house is greater than three times the median household income, then planning regulations cannot possibly explain all of the difference between observed prices and Demographia’s preferred median multiple of three. For Demographia to be correct, land costs would have to be negative in the absence of planning regulations. As this is a logical impossibility, this would be sufficient to prove that Demographia’s estimates of the cost of planning regulations in New Zealand cities are too high.
Without further ado, here are my calculations. I’ve drawn this data from several reliable sources:
Construction cost data for different cities from a recent edition of Rawlinsons Construction Cost Handbook, a reliable source on building costs. I’ve used the midrange of their estimated costs for a single-storey weatherboard home – the most common, cheapest option on the market.
Statistics NZ’s building consents data, which shows that over the last 10 years the average size of new residential dwellings has been around 190 m2. (This is actually likely to under-state the size of new houses, as it also includes apartments. Also, note that it’s necessary to go into Infoshare to actually access the data.)
Median household income (2013)
Average build cost ($/m2)
House size (m2)
Total build cost ($/house)
Median multiple of build costs
Fairly remarkable results. The costs of physically building an average-sized house in four New Zealand cities ranges from 3.5 times the median household income (Wellington) to 5.7 times (Dunedin). The “median multiple of build costs” seems to always be above three.
We have just shown that planning regulations cannot fully explain the difference between house prices in New Zealand cities and the prices that would be implied by a median multiple of three. That is, in fact, a logical impossibility. Demographia seems to be over-stating their case, probably substantially. Quod erat demonstrandum.
However, it does look like construction sector productivity might be a bit of an issue. Different sources disagree on whether New Zealand faces higher construction costs than Australia and other countries. In its 2012 Housing Affordability Inquiry, the Productivity Commission suggested that pay more for construction. On the other hand, a rather excellent NZIER report that benchmarked construction costs against Australia found that we don’t.
So who knows what to think? Economists, eh?
That being said, it’s clear that productivity trends in the building industry are not flash. Here, for example, is the Productivity Commission’s graph of growth in construction industry productivity since 1978. In short, there hasn’t been much:
To summarise, there are many factors driving New Zealand’s house prices. Planning regulations are part of the picture, but as I’ve shown here they cannot possibly explain everything about our high housing costs. We’ve got to also consider other supply-side factors, such as construction productivity and the availability of building tradesmen, and myriad demand-side factors ranging from urban amenities to tax policy to interest rates. And, of course, we might want to ask why people aren’t choosing to respond to high building costs by building smaller houses. (Perhaps it’s an income effect – wealthier people spending more on big houses? Or perhaps there are some rules that make it difficult to build small, efficient dwellings?)
What do you make of the data on residential construction costs?
Since I moved to Auckland, I’ve been trying to make sense of local trends in house prices. Why have they risen over the last decade? Will they keep going up, or crash back down to earth? What’s driving all this?
Others are more confident that planning was what done it. For example, the annual Demographia Housing Affordability Survey states, quite confidently, that planning regulations – an “institutional failure at the local level” – are the main cause of high house prices:
The purpose of the Demographia Surveys is to alert the public and policy-makers if housing exceeds 3.0 times annual household incomes, that there is institutional failure at the local level. The political and regulatory impediments with respect to land supply and infrastructure provision must be dealt with.
Former Reserve Bank governor (and former leader of both National and ACT) Don Brash made matters even clearer in his foreword to the 2008 edition:
The affordability of housing is overwhelmingly a function of just one thing, the extent to which governments place artificial restrictions on the supply of residential land.
In short, planning rules – especially metropolitan urban limits – are bad. Very bad. They are the main reason that houses in some places are unaffordable, which Demographia defines, somewhat arbitrarily, as median house prices that are more than three times the median household income:
Now, according to Demographia, Auckland currently has a “median multiple” of 8.2: “severely unaffordable”. Based on their figures, the median house price in the city at the end of 2014 was $613,000. (This actually seems low – perhaps they’ve converted it to USD?)
In order for Auckland to meet Demographia’s definition of an “affordable” market, house prices would have to fall by $390,000, or 63%. Given that they think that urban planning/metropolitan urban limit policy is the main cause of high house prices, we can take this as their estimate of the cost of those policies. $390,000, per house.
This is obviously a very large number. If house prices fell back to the “affordable” level, it would have ruinous effects on NZ’s financial system and household wealth. (Frankly, this does not seem like a very good outcome.)
So that’s Demographia’s estimate. Recently, several New Zealand economists have taken a more detailed look at the costs of planning regulations in Auckland. For the most part, they analyse the impact of planning rules that were put in place by previous councils:
Motu’s paper on the costs of planning regulations concludes that they add between $32,500-$60,000 to the cost of a new standalone house and $65,000-$110,000 to the cost of a new apartment. These figures were sourced from a survey of developers and accounted for the impact of a number of individual rules ranging from balcony rules to section size controls.
NZIER’s paper models the impact of urban limits and height/density limits. (This paper uses a similar approach to the one I discussed here.) It concludes that these two rules cost households the equivalent of $1800 per annum. In present value terms, this is somewhere on the order of $30-40,000 per household (depending upon what discount rate you use).
[Disclaimer: I know the authors of both studies and have a great deal of respect for their work. In a professional capacity, I provided comments on earlier drafts of both papers. This is somewhat unavoidable given the size of New Zealand…]
Here’s a chart comparing Motu and NZIER’s estimates of the cost of planning regulations with Demographia’s estimate:
In other words, these estimates suggest that Auckland’s planning regulations explain only 10-30% of the difference between Auckland house prices and a median multiple of 3. Even if we add together the estimates from the two papers, we don’t get anywhere near explaining the gap.
So: whose estimates of the cost of planning should we believe?
Personally, I trust Motu and NZIER’s analysis, as it’s backed up by empirical research and/or economic modelling, whereas Demographia’s is mainly justified by rather repetitive and self-referential haranguing. Consequently, I don’t think we can conclude that planning regulations are a sufficient explanation for Auckland’s relatively high house prices.
That doesn’t necessarily mean that we should be sanguine about planning regulations. While things will change under the Unitary Plan, the costs associated with existing rules are reasonably large. And, in contradiction to Demographia’s claims that constraints on greenfield land supply are the biggest problem, both papers find that existing regulations place higher costs on higher-density developments.
The Motu paper finds that existing planning rules add twice as much cost to apartment developments as to standalone houses. Similarly, the NZIER paper finds that height limits (and other controls on density) are slightly more costly than urban limits. (This is partly because Auckland is mostly surrounded by water, meaning that preventing land from being used efficiently is much more deleterious than it would be in a city with more land.)
A few weeks ago, I took a look at property taxation in the US, Canada, and New Zealand. I found that Auckland homes are taxed lightly by comparison – rates average 0.39% of house value. Property tax rates are twice as high in most of the other cities I looked at. In some cases – e.g. Houston, where property taxes average 2.31% of home value – they are much, much higher.
This should come as no surprise to anyone who’s read the literature. For example, Grimes and Coleman (2009) find that New Zealand under-taxes property:
McLeod et al (2001; p.26) showed that the proportion of taxation raised through property taxes was lower in New Zealand than in Australia or the United States. Taking into account all levels of government (federal, state and local), Grimes (2003) found New Zealand’s share of property taxes in government revenue was relatively low, at 5.7%, compared with a (20 country) OECD average of 8.3%. As a share of GDP, New Zealand’s property tax share was also relatively low at 1.8% compared with the average rate of 2.4%.
They go on to suggest that introducing even a relatively small land tax could result in fairly large changes to land prices. It’s intuitively sensible that higher property taxes would discourage people from bidding up prices – the more they pay for land, the more they pay in taxes!
Stu recently took a look at the same research, and some of the broader trends, and concluded that higher property taxes could take some heat off the housing market. But how much does property tax really matter for housing affordability?
To get a rough sense of the relationship, I’ve put together a chart showing the relationship between property tax rates (x axis) and Demographia’s “median multiple” measure of house prices relative to incomes (y axis). It includes data on all 59 US, Canadian, and New Zealand cities with a population over 1 million.
Notice the substantial negative correlation between property taxes and median multiples. There is a strong tendency for places with lower property taxes to have higher house prices, and vice versa. The least “affordable” places all have relatively low property taxes.
Overall, this chart suggests two things: First, New Zealand’s relatively low property taxes may contribute to our relatively high house prices. Notice how Auckland’s house prices seem to fit the overall trend in the data.
Second, as I’ve previouslyargued, Demographia’s analysis is largely meaningless as they have failed to account for the full range of explanatory variables, from interest rates to tax policies to economic fortunes.
Here’s another view on the same data. I’ve taken the natural logarithm of both variables to smooth out the relationship, and put a trend-line through the data points. This simple bit of analysis suggests that:
About 44% of the variations in median multiple can be “explained” by differences in property tax rates. For a bivariate regression, this is quite high.
The slope of the regression line suggests that, within this sample of cities, a 10% increase in the property tax rate is associated with a 4.6% reduction in the median multiple. Again, that’s a quite strong relationship.
I don’t think we can draw any firm policy conclusions from this data, but it certainly suggests that our low property taxes are worth investigating as a cause of our high house prices. In the words of xkcd, “Correlation doesn’t imply causation, but it does waggle its eyebrows suggestively and gesture furtively while mouthing ‘look over there’.”
Finally, it’s worth taking a closer look at the four US cities with the highest median multiples – the top data points on the left hand side of the first chart. They are all large cities in California – San Francisco, Los Angeles, San Jose (i.e. Silicon Valley), and Sacramento (the state capitol). They provide a great illustration of why failing to account for multiple, correlated explanatory variables can undermine an analysis of house prices.
There’s also a great irony underlying the use of LOS [traffic Level of Service] as part of CEQA’s environmental impact checklist. It seems self-evident that bike projects are favorable to the environment, but the use of LOS to evaluate them can sometimes imply quite the opposite. The person who filed the 2005 lawsuit against the San Francisco master bike plan, for instance, suggested that because bike lanes raise LOS they also raise congestion and car idling, and thereby cause pollution.
That’s not the only contradictory aspect of LOS. Case in point: a developer whose building fails an LOS threshold can mitigate the environmental impact by widening the street, which of course would attract more cars and pollution. So instead of encouraging dense development and lower vehicle mileage — the hallmarks of a transit-first city — San Francisco’s use of LOS as part of CEQA actually discourages livable design. In a three-part series on LOS at Streetsblog, one transportation consultant called LOS the “single greatest promoter of sprawl and the single greatest obstacle to transit oriented development” in California.
However, CEQA and its absurd requirements for traffic assessments (etc) aren’t the only thing going on in California. State-level laws have also constrained local governments’ ability to raise property taxes. Proposition 13, a citizen-initiated referendum passed in 1978, caps property tax rates at 1% and fixes them to the value of the house at the time it was purchased, plus a 2% annual increment for inflation.
This has had a number of perverse effects, including stripping away funding from California’s formerly excellent primary and secondary education and setting it on a path of decline. (Prop 13 is basically exhibit A in the case against binding referenda.) It has also distorted the housing market. Because home-owners know their property taxes won’t increase if the value of their house increases, they may be more willing to speculate on capital gains.
A 1982 paper by economist Kenneth Rosen offers empirical support for this hypothesis – he found that reductions in property tax rates were almost immediately followed by proportional increases in house prices.
Consequently, it would be foolish to analyse the Californian housing market without attempting to control for both taxation and planning policy. If you only looked at one policy, your conclusions would be biased by mis-attributing the effects of the other policy. (That’s precisely what Demographia seems to have done, by the way.)
What’s true for California is also true for New Zealand. I find it hard to take seriously the claims of people who attribute housing affordability solely to regulatory policy and fail to consider the potential impact of our low property taxes.
What do you make of the data on property taxes and house prices?
My lasttwo posts about Demographia’s analysis of house prices prompted quite a bit of discussion. I thought that it may be worth expanding on my points and clarifying why they mean that we should take Demographia’s conclusions with a large grain of salt.
Economists (and statisticians) have a term for what Demographia has done: “omitted variable bias”. This can occur when there are multiple variables that have a causal impact on an outcome. If we attempt to understand that outcome without considering all explanatory variables, we run the risk of biasing our conclusions.
Economists are trained to identify and address issues arising from omitted variable bias. Here, for example, is a comment on the topic from a widely used undergraduate econometrics textbook:
Now suppose that, rather than including an irrelevant variable, we omit a variable that actually belongs in the true (or population) model. This is often called the problem of excluding a relevant variable or underspecifying the model. We claimed in Chapter 2 and earlier in this chapter that this problem generally causes the OLS [ordinary least squares regression] estimators to be biased.
Now, I realise that’s a bit obscure, so let me be more specific. Here’s a list of (some) factors that can influence house prices, with a view on their expected impact:
Expectations for future population and economic growth
Future expectations tend to be capitalised into house prices – i.e. prices will tend to be higher in areas with better growth prospects
Consumer and natural amenities
People are willing to pay more to live in nicer places
Interest rates (and availability of finance)
Lower interest rates enable people to afford a larger mortgage on a given level of income
Construction sector productivity
Lower productivity will increase the cost of supplying new dwellings
Tax policies, including capital gains taxes and mortgage interest deductions
Taxation of capital gains will reduce willingness to invest in housing for capital gains
Tax subsidies for mortgage-holders will tend to push prices up
Other housing market policies, such as rent subsidies or public housing provision
Rent subsidies tend to be captured by landlords and thus will tend to push up prices
Ongoing construction of state housing will add supply to the low end of the market and thus constrain price increases
Urban planning policies
Policies that constrain the development of new housing in desirable areas, or make it more costly or uncertain to develop, will reduce supply and push up prices
Demographia only addresses one of those seven variables. Because they fail to account for the other six potentially explanatory variables, their estimates of the welfare impact of urban planning policies are not likely to be reliable. Without controlling for other potential explanations for high housing prices, it’s impossible to say whether their conclusions about any individual city are correct or not.
Consequently, my recommendation to people seeking to understand the impact of urban planning policies on housing costs is simple: Ignore Demographia and go read the relevant economics literature instead. For those who are interested in doing so, here are a few papers that I have learned a lot from. They apply a range of modelling approaches, but what they have in common is that they undertake a detailed analysis of rules, rather than making sweeping and unsupported generalisations:
Glaeser, Gyourko and Saks (2005) studied the impact of building height limits in Manhattan by looking at the gap between observed sale prices and the marginal cost to add another floor to high-rise buildings. They find that constraining the supply of high-rise apartments imposed a significant “regulatory tax” on residents.
Grimes and Liang (2007) took at look at land prices around Auckland’s Metropolitan Urban Limit, finding evidence of a “boundary discontinuity”. They interpreted this as evidence that the the MUL is overly restrictive.
Kulish et al (2011) developed a hypothetical model of the impact of several factors on housing and transport costs. They modelled density restrictions as well as increased transport costs and lower building productivity, finding that building height limits raise housing costs and increase sprawl. I have previously discussed their findings.
MRCagney (2013) examined the impact of minimum parking requirements in Auckland, finding that they impose a loss on developers and businesses by forcing them to over-supply parking. They also cause worse congestion, meaning that not even drivers benefit. Luke C briefly discussed this study in a post on the Unitary Plan parking rules.
My favourite economics paper on planning regulations is Cheshire and Sheppard’s 2002 study on the impact of greenbelt rules in Reading, UK. The authors observed that greenbelts have both positive and negative effects. On one hand, they limit the supply of land for new housing, which drives up costs. On the other hand, they give residents access to public open spaces, which people like. Rather than ignoring this trade-off, they used house price data to model it.
Overall, Cheshire and Sheppard did find that allowing development in Reading’s greenbelt would make people better off. However, they also found that a failure to consider the amenities produced by planning rules would have resulted in too high an estimate of the gains in wellbeing. In other words: right direction, wrong magnitude.
In light of the evidence, my view is that failing to account for urban amenities and other explanatory variables in an analysis of the impact of supply restrictions can result in two errors:
First, it can make us over-optimistic about the degree to which loosening rules will affect housing prices. That’s not to say that less restrictive planning regulations couldn’t make us better off – just that we should not expect house prices to fall by 60-75% as Demographia implies when it says that Auckland should have a median multiple of 3.
Second, it can lead to perverse outcomes, by encouraging us to eliminate rules that are serving a useful purpose. Often (although not always) planning rules are managing the external social or environmental costs associated with some developments. A proper cost-benefit analysis – which Demographia has not done – will consider both the pluses and minuses of rules.
In short, housing markets are complex, and any analysis needs to consider that fact. To reiterate my point from last week: Demographia takes an inappropriate, overly simplistic view of house prices. This may be good for grabbing headlines, but it’s not good economics.
To recap: urban economics suggests that differences in the level of the median house price to median household income ratio between cities can be interpreted as differences in livability. All else equal, people should be willing to pay more to live in cities that offer better quality of life.
But how should we interpret changes to the median multiple from year to year? If a city’s median multiple rises from 5 to 5.5, does that mean that the city suddenly got 10% more livable? Or did something else happen?
Demographia is very certain that higher median multiples are the product of one thing, and one thing only: limits on sprawl into greenfield areas. Here’s Don Brash, former Governor of the Reserve Bank of New Zealand and former head of several right-wing political parties, laying out that view in Demographia’s 2008 report:
Once again, the Demographia survey leads inevitably to one clear conclusion: the affordability of housing is overwhelmingly a function of just one thing, the extent to which governments place artificial restrictions on the supply of residential land.
With that in mind, Demographia’s data seems to indicate that housing in Los Angeles and Las Vegas (as well as many other US cities) suddenly became a lot more affordable in the late 2000s. It’s obvious that they must have removed their Metropolitan Urban Limits – how else to explain such a big drop? Oh, wait…
(It’s slightly disturbing that our Reserve Bank used to be run by a man who doesn’t believe that interest rates and credit conditions can affect house prices. But I digress.)
Here’s a graph of changes in Demographia’s median multiple estimate for Auckland since 2004. We haven’t seen the same drastic swings as in the US, where the housing bubble and bust was pronounced, but house prices have risen relative to incomes. (Although, as Stu found in his analysis of different measures of housing costs, this hasn’t flowed through to rents or mortgage payments, due in part to changes in interest rates.)
This change has been especially pronounced in the last few years. Since 2012, Auckland’s median multiple has risen roughly 22%. Does this mean that we’ve become 22% more “livable” during that time?
With all due respect to the good work done by Auckland Council and Auckland Transport since their inception, probably not. So we need to look for alternative explanations, of which there are several. I’ll focus on three in particular.
The first potential explanation is that there has been a market failure. Residential construction slumped massively during the Global Financial Crisis, with the most significant reductions occurring in the supply of apartments and attached dwellings. Here’s a graph that John Polkinghorne put together illustrating that trend:
Incomes and employment have mostly come back from the recession, but construction has been a bit slow to respond. I suspect this reflects technical constraints within the development sector, as it takes a while to organise finance, find sites, and hire the cranes, bulldozers, and blokes/blokesses in hardhats. Until they get into gear, there may be a bit of an undersupply – but one that will tend to sort itself out.
The second potential explanation is that the introduction of Auckland’s Unitary Plan has caused people to expect housing supply to be more constrained in the future. While the Unitary Plan envisages the gradual expansion of the city’s Metropolitan Urban Limit to meet new demand for greenfield suburbs, it maintains extensive controls on the supply of new dwellings in accessible, high amenity areas. Moreover, the plan actually got significantly more restrictive following the consultation process.
Transportblog has highlighted this issue a number of times before. To recap, here’s a map (from Koordinates) that shows where the Unitary Plan got more restrictive as a result of consultation. The areas in red have been down-zoned to restrict development, while areas in green have been up-zoned. The large orange areas show future greenfield land. As you can see, there are not a lot of opportunities to develop new dwellings in the isthmus and lower North Shore, while West Auckland has been happy to facilitate growth:
Timing is important here. Demographia’s figures suggest that there was a jump in house prices relative to incomes between the end of 2012 and the end of 2013. This coincides with the notification of the Unitary Plan in September 2013, which, as described above, will ease greenfield land supply while limiting development opportunities in the inner suburbs.
However, there is also a third potential explanation: that our rising house prices reflect increasing awareness of Auckland’s great quality of life. For most of the last decade, our city has been near the top in Mercer’s Quality of Living Survey. It’s been ranked third for five years running.
So maybe the story is that potential migrants and investors have observed that, by international standards, Auckland offers high quality of life at an affordable price. And they are in the process of arbitraging that away.
I’ve illustrated that process in the following graph, which shows the relationship between Demographia’s median multiple (X axis) and Mercer’s Quality of Living Survey (Y axis). The trend-line is estimated based on 2012 data. I’ve also plotted Auckland’s median multiple and Mercer index in 2015.
The red dot that represents Auckland is moving towards the trend-line, suggest that its prices are catching up with its livability.
Previous studieshave found that growth in New Zealand house prices is strongly correlated with net migration – i.e. migrants tend to bid up house prices. Net migration to New Zealand did, in fact, start picking up in 2013 – around the time that Demographia’s estimate of the price to income ratio began to rise. Perhaps this is evidence for the “amenity arbitrage” hypothesis?
So, which explanation is true? Honestly, it’s impossible to say without a lot more detailed analysis. My point in writing this is not to argue that there is a single explanation for changes in Auckland’s house prices, but to point out that there are many possible explanations. Housing markets are affected by a number of factors, and it’s inappropriate to focus on one without controlling for the rest.
This is, essentially, why Demographia’s analysis fails. Rather than articulating a model that encompasses all of the potential explanatory factors, they have settled on a single number and insisted that it must be interpreted in a single way. It’s hard to see the value in that approach. And it’s definitely not good economics.
Every year since 2005, pro-sprawl think-tank Demographia has published a new edition of its “International Housing Affordability Survey“. They report a “median multiple” measure of housing affordability that compares median house prices to median household incomes within a number of cities, mostly in the English-speaking world.
Demographia’s aim, in publishing this data, is to argue that “if housing exceeds 3.0 times annual household incomes, that there is institutional failure at the local level. The political and regulatory impediments with respect to land supply and infrastructure provision must be dealt with.” By this, they mean building car-dependent suburbs on the urban fringe – and nothing else.
Another Demographia-approved urban paradise.
A number of people, including Todd Litman and Stu Donovan (on Transportblog), have taken aim at Demographia’s empirical analysis and choice of metrics. Unfortunately, Demographia is unwilling to open up its analysis and methodology for an independent peer reviewed, so it’s difficult to referee those claims.
Here, I want to take a look at the issue from a different perspective. Basically, the urban economics literature suggests that Demographia’s chosen measures do not mean what they think they mean. And they almost certainly do not prove the case they’re trying to make.Before I explain why, let’s start out with a quick look at the data. According to Demographia’s 2015 report:
The most “affordable” cities included the likes of Detroit (median multiple of 2.1), Cleveland (2.6), and Houston (3.5)
The “unaffordable” cities included most large Australian cities, including Sydney (9.8) and Melbourne (8.7), many “coastal” North American cities, such as Los Angeles (8.0), San Francisco (9.2), Vancouver (10.6), New York (6.1), and Boston (5.4)
All New Zealand cities were on the “unaffordable” end of the spectrum, ranging from Palmerston North (4.1) and Dunedin (4.6) to Christchurch (6.1), Tauranga (6.8) and Auckland (8.2).
In other words, there’s a quite large range of median multiples. This raises a quite obvious question: Why are people willing to pay so much more to live in some places? Why live in “unaffordable” San Francisco when “affordable” Houston is just down the road? Why live in Auckland when housing is relatively cheaper in Dunedin?
Why would anyone want to live in a large, multicultural city located between two beautiful harbours in a subtropical climate? Sheer madness.
Urban economists have studied this phenomenon in detail, and observed that there is an omitted variable in Demographia’s equation: the differing amenities offered by different cities. If a city offers good natural amenities or consumer amenities, people will be willing to pay more to live there. Conversely, if a place isn’t particularly nice, people won’t be willing to pay much for houses there. (Common sense, really.)
It is quite common in discussions of housing affordability to focus on the share of income being spent on housing, as if this is a natural measure of the degree to which housing affordability is a problem within an area. The spatial equilibrium assumption suggests that this measure is not particularly meaningful or helpful.
In short, urban economics suggests that we should interpret a high median multiple as an indication that a city offers great amenity for its residents, rather than an indication of bad policies. I tested this hypothesis by looking at the correlation between the (2012) Demographia median multiple figures and two international quality of living rankings. I found that there was a positive correlation between median multiples and livability.
Here’s the correlation between the median multiple (X axis) and Mercer’s 2012 Quality of Living Survey (Y axis; lower numbers indicate higher rankings). Once again, a positive correlation, with 31 data points:
In other words, high house prices relative to incomes are a good indicator that a city is a nice place to live. Rather than proving that Metropolitan Urban Limits inevitably push up house prices, Demographia’s median multiple seems to simply measure cities’ relative levels of amenity. When they argue that all cities should have a median multiple of under three, they are arguing for an absurdity: that all cities should offer the exact same level of amenity to their residents.
If we wanted to accomplish that, we’d have to destroy most of the things that make great cities great. This might make housing cheaper, but it wouldn’t make us any better off in a broader sense. That’s because it would require us to:
Bulldoze the Waitakere Ranges and use the spoil to fill in the Hauraki Gulf – to ensure that Auckland didn’t have any natural advantages over a flat, inland city like Hamilton
Dynamite the historic boulevards of Paris and replace them with American-style subdivisions and malls – to ensure that Paris didn’t offer anything that Houston doesn’t
Ban any venture capital or startup activity in San Francisco, to ensure that it doesn’t offer any agglomeration economies that don’t exist in Detroit
Build large screens over sunny cities like Tauranga and Brisbane – to ensure that they don’t have nicer weather than Moscow or Toronto.
But Demographia’s not aware of this. Their analysis is overly simplistic. The only thing it reveals is the authors’ grievous failure to understand the basics of urban economics. It’s no wonder that Demographia has never tried to have its studies peer reviewed or published in academic journals. Their claims aren’t supported by any valid conceptual model. But I guess that’s what happens when you get an urban planner and a former property developer to do an economist’s job…
Demographia is a pro-sprawl think tank in the USA that publishes density and house price data for cities across the world. They’re often seen using their statistics to argue that the only way to deliver affordable housing is with suburban fringe expansion into greenfields land.
Demographia’s approach to calculating density is simple but misleading. They have simply calculated the total number of people in each city and divided it by the total land area covered by that city – including unpopulated areas like parks and reserves. This measure of average density is actually quite irrelevant. For example, Demographia uses it to claim that Los Angeles is more dense than New York.
As Peter outlined in this recent post, what is much more relevant is the density of the neighbourhood that the average person lives in, rather than the density of the average acre of land in the city.
For argument’s sake, consider a village of one hundred acres with one hundred residents. By the measure of area weighted density the density is simple, one person per acre. Does this represent the reality of how people live? Well it could, if everyone lived alone in a separate house on an acre of land. But what if they didn’t? What if everyone lived in a single apartment block in the middle of town that sat on one acre of land? Well then the density the people actually live at would be 100 people per acre. That’s a hundred times more dense… but the same density by Demographia’s measure!
And what if it were something more complex? What if a quarter of the town lived in the one apartment building, half lived in eighth-acre sections around it, and the remaining quarter were spread out over the rest of the land? Doesn’t that sound a bit closer to the reality of most cities? One person per acre means nothing for this theoretical town, half the people live at eight times that density and a quarter at fifty times the density.
Just in case you’re still unsure, in the image below the dots represent dwellings and each box has the same number of dots in it. Overall they have the same average density however in reality they would feel like two very different places.
In short population-weighted density is a much better indicator of the density of the neighbourhoods people chose to live in, and a much better way to describe cities and housing.
With that in mind, it was a simple task to take Peter’s data and throw it on a chart. In simple terms these show how many people live in neighbourhoods (Census meshblocks to be precise) grouped by density in Auckland, Wellington and Christchurch. I’ve also picked out the density level that Demographia says each city is.
It’s easy to see a few things here. First of all we can see that neighbourhood density can vary quite a lot within cities. One number just can’t describe how people live. Second, we can see there is something of a bell curve. Most people live within the middle range of densities, those living very low or very high are small in number, but that middle is actually quite broad. Third, we can see how far off Demographia is. Their supposed summary statistic isn’t anywhere near the middle of the curve, it’s actually near the bottom in each case.
For example it seems the Demographia figure describes the density at which roughly five percent of Aucklanders live. Nine percent live at lower densities, and 86% live at higher densities. Many people live in neighbourhoods that are two or three times more dense than Demographia’s misleading average. In short, Demographia’s figures are irrelevant for the vast majority of Aucklanders (and Wellingtonians, and Christchurchers). They don’t reflect how the majority of people choose to live.
My first post suggested Demographia’s primary findings were not supported by independent evidence, such as alternative “rent-income” and “home affordability” indicators. My second post then outlined some issues with their “median-multiple” indicator (calculated as the median house price divided by the median household income).
This post will now refine some of these criticisms, before outlining some of my own ideas on the causes of housing affordability issues in New Zealand. First I wanted to tease out some of implicit assumptions that underpin Demographia’s “median-multiple” indicator, namely:
Median-matching: This issue was best articulated by James H: “the median-multiple indicator carries an assumption: that the income-earners at and around the median are the same group who demand the houses priced at and around the median, and that result can be extrapolated for other price-income pairings. As mentioned in the post, that excludes measurement for a large group of earners at many income levels who are in fact happy to rent for a range of reasons. Also I doubt whether median houses are often bought by median earners for a variety of reasons including life stages, geographical differences etc.”
Independent inputs: This issue relates to the fact that the two inputs into the median-multiple indicator (house prices are income) are actually not independent of each other. Consider a situation, for example, where most of the houses in New Zealand were being bought and sold by relatively wealthy households, and that these households subsequently experienced high income growth, while incomes for the general population remained broadly unchanged. In this situation the median house price (and median-multiple indicator) would rise simply because income growth was concentrated within the same people that were purchasing properties, rather than because housing was becoming less affordable.
The second issue is quite important, because it implies that changes in income may in fact impact on house prices. The figure below illustrates the input data used by Demographia for Australia. In this graph, we find a strong positive correlation between median household income (x-axis) and median house prices (y-axis).
This suggests that locations with high incomes have more expensive housing (surprise surprise!). More specifically, it suggests that for every $1 increase in median household income there is a corresponding $5.1 increase in median house price. While that sounds like a lot, note that income is measured p.a. whereas house prices are “total.”
My final comment is on the relevance of Demographia’s indicator. i.e. the “so what” question? It seems that the parts of New Zealand with the highest median-multiple ratios, such as Auckland, are actually attracting the fastest population growth, as illustrated below. Of course, this may reflect other factors that are at play, but it does suggest that our housing affordability (at least as measured by the median-multiple indicator) is not yet significant enough to drive people away.
Thus, population growth in these places seems to be going the other direction from what Demographia would expect – we are increasingly moving to areas that they consider to be unaffordable. Based on this evidence, I’d suggest that the median-multiple indicator used by Demographia is not a good measure of housing affordability. Instead, it seems to measure:
The degree to which income growth is invested in housing; and
Population growth (which will tend to push up property prices but suppress income growth).
It’s not clear to me that the median-multiple indicator measures housing affordability, and nor is it clear to me that the urban containment policies pursued by local governments are binding to the degree that they have major impacts on property prices. They may be – but Demographia’s indicator does not, and cannot, tell you that.
My personal view is that the primary impact of local government regulations is not through the constraints they place on land supply (i.e. urban containment), but actually through the barriers they create to the development of more compact and affordable housing. Here’s some examples of regulations pursued by local governments in New Zealand that seem likely to restrict the supply of affordable housing:
Minimum lot sizes – i.e. “all ye who have less money shall be forced to purchase land you don’t want.”
Minimum apartment sizes – i.e. “all ye who have less money shall be forced to purchase living space you don’t want.”
Minimum parking requirements – i.e. “all ye who have less money shall be forced to pay for vehicles you don’t own”.
Maximum height limits – i.e. “all ye who chose to live like rats are consigned to perish like rats – on the street.”
Heritage protections – i.e. “all ye who don’t have the money to renovate a villa shall live elsewhere.”
In my experience these policies are often more binding constraints than the availability of land. So my suggestion is that housing affordability has less to do with policies that favour urban containment (as Demographia and the National Party would have you believe) than they are to do with the plethora of policies that suppress more intensive and affordable housing. I’d go as far as to say that most of our policy settings have a systematic bias against the development of compact and affordable housing.
In this light, it seems that recent political announcements have missed the mark. National are deluding themselves into thinking that the release of land on the urban periphery will deliver meaningful and sustained reductions in the cost of land, and by extension housing. Labour and the Greens, meanwhile, seem intent on using government capital to build our way out of the problem – which is not only expensive but also runs the risk (at least on the surface) of building the wrong kinds of houses in the wrong places. None of these three parties seems to yet acknowledge that some of our issues with housing affordability may be the result of policies that prevent urban intensification.
So instead of writing the foreword to next year’s (deeply flawed) Demographia report, I’d suggest that Bill English – and other National cabinet Ministers – should be writing letters in support of proposals to develop apartments, town houses, and units in places like Milford and Orakei. And more importantly, they should be making submissions on aspects of the draft Unitary Plan that support and/or prevent more compact and affordable accommodation options. Onya Bill.
Yesterday’s post considered the recently released Demographia survey on housing affordability. Thanks to everyone who commented; the discussion was useful for honing my thoughts on follow-up posts. Such as this.
But first let’s re-cap: Demographia’s key findings were 1) New Zealand has increasingly unaffordable housing and 2) this is the direct result of urban containment policies.
The main issue I took with the Demographia report in yesterday’s post was 1) the lack of strong economic justification/references supporting their housing affordability indicator of choice (namely the median-multiple ratio) and 2) the lack of discussion/investigation of potential alternative indicators.
Indeed, my quick web search threw up at least two alternative indicators of housing affordability, namely the rent-multiple ratio and the home affordability index, neither of which appeared to lend much support to Demographia’s findings. Of course, this does not prove their conclusions are incorrect, but it does suggest they are premature.
In this post I wanted to look beneath the hood of Demographia’s housing affordability indicator a little more. The reason being that when you do you start to see what they are measuring and, perhaps more importantly, what they are not measuring. In Demographia’s case, they calculated their housing affordability indicator as follows:
Median-multiple = median house price / gross median household income per annum
This then measures, in a simple sense, the cost of the median home relative to the median household income. While that may sound reasonable enough on the surface, the devil is in the detail. Two of the more obvious issues with Demographia’s indicator that spring to my mind are discussed in the following paragraphs.
Demographia’s definition of “income” excludes taxes and transfers. This is pertinent for at least two reasons:
First, some taxes have direct impacts on property prices, e.g. local rates. These will simultaneously tend to affect property prices (higher rates = lower property prices) and post-tax income (lower), but not gross income. Somewhat perversely, this would mean that jurisdictions with higher property taxes would tend to exhibit more affordable housing, at least according to Demographia’s indicator.
Second, most taxes directly impact on a household’s disposable income and in turn affects their ability to afford housing. In New Zealand tax rates have changed considerably over time, especially for different segments of the population. Consider for example the impact of Working for Families on demand for certain types of housing.
Such issues mean that the median-multiple housing affordability indicator, as it appears that Demographia have applied it will not pick up on relevant differences in taxes and transfers, both spatially and temporally.
The spatial differences are likely to be fairly minimal within a country like NZ – where local taxes don’t vary that much from place to place – but this is certainly not the case when making international comparisons. Many countries have much higher rates of property taxes (and even local income taxes) that will tend to impact on house prices and thereby affect their housing affordability relative compared to New Zealand.
On the other hand, the temporal differences introduced by changes in domestic tax and transfer policies are likely to be fairly large, even within a country. The potential impacts on housing affordability of recent tax changes to the top personal tax rate, ability to claim capital depreciation on properties, and commercial tax rates are hard to predict in advance. Tax impacts may well spill over national boundaries as well; NZ’s lack of capital gains tax, for example, is frequently quoted by my Australian colleagues as a primary driver of their decision to invest in New Zealand’s property market.
These issues would make me extremely cautious about drawing broad, sweeping conclusions on trends on housing affordability both within and between countries simply based on the median-multiple indicator.
You don’t measure the affordability of cookies based on the cost of buying the cookie factory.
The point is that housing is a actually a type of good, or more specifically a service, which is “produced” by a house. You can gain access to housing without necessarily buying the factory that produces it, i.e. rent a house. Obviously, some people do this already and they’re called “renters.” Like me.
Even in New Zealand many people rent by choice. And in many countries in central and northern Europe renting is even more prevalent. But the key takeaway message is that the affordability of housing, which is what Demographia sets out to investigate, is probably better measured (from an economic perspective) using rents rather than house prices. This is especially true for low income households that are more likely to rent.
And that’s why I’d place more emphasis on the graph produced by the Productivity Commission, which calculated the ratio of rents to household disposable income over time than the median-multiple indicator presented by the Demographia study. This showed the rent to income ratio in New Zealand declining since the 1990s, contrary to Demographia’s findings and casting some not inconsiderable doubt on their conclusions.
My preference for using rents is also related to the first point on the impacts of taxes on house prices: Unlike houses, which are an asset, rents measure the cost of housing services. I suspect it’s far easier to “net out” the impact of services taxes in various jurisdictions, i.e. GST, on rents than it is to adjust for changes in the myriad of other income and asset taxes that might affect house pricing.
That’s all for tonight, but tomorrow’s another day and I’m already fomenting ideas on the next Demographia post; in the meantime I’d welcome your comments/suggestions/criticisms.