Wednesday, September 30, 2015

How to analyse housing markets

Housing costs are typically 30% of household income, while about 43% of household savings are tied up in the value of owner-occupied dwellings. There is really no more important market for the general public when it comes to their cost of living and their ability to save for the future.

But simply talking about the housing market as if it is some monolithic beast will lead you to the error of conflating three distinct markets that must be considered independently if you want to really understand what is happening. These markets are

1: The land asset market
2: The housing service market (annual occupancy from rent or ownership)
3: The residential construction market

When you buy a home on the second hand market (rather than a new home), you are actually buying a bundled good which includes a land asset along with a durable housing product which lasts the life of the building. A close analogy would be buying a car bundled with an equity share in the vehicle manufacturer - you get the vehicle for its useful life, and the equity asset in perpetuity.

So when we talk of high demand for housing, home prices increasing, and housing bubbles, we must be clear about whether we are talking about the market for the land asset component of the bundled housing good, or the market of occupying homes themselves. Conflating these is the most common error in housing market analysis, and it leads to conclusions that make little sense in reality.

For example, take the frequent commentary about the effect of population growth on home prices. To me it is utterly confusing. If we are talking about the land asset market, the question then arises about why we don’t talk about the population effects on equity and debt markets, derivatives markets, and other asset classes that could equally see effects. The reason being that more people means more buyers AND sellers of the same assets.

You can see from the graph below that population effects don’t seem to be driving the growth in land asset prices, or at least can’t be a major contributor if areas with a 10-15% population decline can still see 70% growth in home prices.

Of course like other asset markets the reason for the land price increases has a lot to do with the systematic reduction of interest rates in the past 20 years. Asset prices are just the capitalised value of future claims on incomes, and a lower interest rate increases that asset value compared the value of the future incomes. This means that comparing prices of the bundle of house and land asset to incomes makes no sense at all. It would make just as much sense to compare the price of an equity share in Woolworths bundled with a kilos of bananas as a way to measure food inflation. Why not measure the food itself?

Luckily, we do have a market for housing as a produced good that we consume on an annual basis quite apart from the land asset; the rental market. If we measure how much of our incomes we spend on rent, and the quality of the homes we reside in (in terms of sqm per person), we can apply the supply and demand model to the market. If there really is something going on with population and housing production, it must be observable in the rental market. Looking at the chart below we can see that in fact the rent to income ratio declined all the way through the land price boom of the early 2000s, as did the occupancy rate (fewer people per home) indicating that in fact we were building more new homes than new people.

So sure, use your supply and demand analysis on the market for produced durable housing goods, but remember that home prices aren’t the price in that market. Rents are the price in the housing market, while home prices mostly reflect prices in the land market.

Lastly, we can look at the construction market, which is driven by trends in other markets, including speculation on land markets. Here the idea of supply and demand also works fine, as periods of high demand for new construction result in increasing construction prices (as demand shift to the right against a resource-constrained upward sloping supply curve for construction services). But again, the construction market and construction prices are not the contributor to growth in home prices. In fact, higher construction costs will decrease the value of the land asset, as they provide an additional cost to capturing future income flows.

The situation now in Australia is that asset market dynamics, including lower interest rates, international buying, and simple cyclical timing of investments, are driving up land prices in some capital cities. In some areas, when this asset buying occurs in new homes it also increases demand for construction, pushing up prices in that market as well. And in the housing service (i.e. rental) market, the additional supply is suppressing rents.

This is the way to analyse housing markets. Don’t be drawn into the monolithic view by conflating behaviour in these distinct markets.

Thursday, September 10, 2015

Doing the housing supply maths

Laurence Murphy is a top property economist at the University of Auckland. I met him last night after a presentation in Sydney where he took on the myth that planning constraints are a major determinant of current home prices in Australia and New Zealand.

He said it is very easy to demonstrate mathematically how little impact even a large increase in the rate of supply would have on prices. But when he shows this analysis to government officials, planners, and engineers who have bought into the supply-side narrative their response is often

“I see you calculations. I follow the logic. But I don’t believe it!”

So I wanted to try the ‘basic supply-side maths’ for myself on the blog to see what sort of effects radical changes to the rate of new housing supply could have, and see if I generate some of the same responses.

Here’s how the maths work. I take the number of new dwelling completions from the ABS for the past 20 years, which is shown in quarterly figures in the blue line of the chart below. Since 1995 new housing supply has been 146,546 dwellings per year on average, which is about a 2% increase in the stock annually, though this moves with the business cycle.

I then add 10% to this number every year to generate a counterfactual world where supply has been much higher over a sustained two-decade period (green line). Then I add 20% just to take an extreme scenario (yellow line). Note that in this exercise I don’t ‘elastify’ supply, which would have higher construction in boom periods, and lower construction in slump. When I run the numbers of more elastic supply that responds to both booms and slumps more I get fewer home built compared to what actually happened! This is because when completion rate falls, it falls faster, offsetting all of the gain from the previous boom. I show a twice as elastic scenario in the next graph in red, which actually results in 8,000 fewer dwellings built in the past 20 years. ‘Elastifying’ supply can’t really be what is desired by those advocating for supply-side reforms. 

Any supply-side housing initiative should simply aim to get more homes built, year in, year out. This is what I capture in my counterfactual scenarios of 10% and 20% higher construction over two decades.

So here is question. How many more houses would there be now in these counterfactual worlds? And what would the price impact be?

Well, if we had built 10% more new home each year for the past 20 years Australia would have around 300,000 more homes. At a 20% higher rate of completions that's 600,000 more. Sounds terrific! That must have a massive impact on prices.

Well. No.

You see Australia’s current housing stock is somewhere above 9million homes. Around 8.8million occupied, and many second homes, holiday homes, and so forth that are traditionally about 8% of the housing stock. Let’s say that there are 9.3million dwelling in the country right now. These additional homes in my 20 year supercharged supply scenarios represent just a 3.2% and 6.4% increase in total stock respectively.

The price impact of a 3% increase in supply is a 3% reduction if demand elasticity is unity. That’s it. The price reduction could be less if there are countervailing income effects that lead to outbidding for superior locations.  So twenty years of supercharged supply provides somewhere between 0% and 3% lower prices, which suggests to me that focusing on the supply side is close to a waste of time. In the 20% higher housing completions scenario the effect is somewhere between zero and 6%. About the same as two and a half years of rental price growth.

To put it another way, after 20 years of a 10% higher rate of new supply, rents today would be then same as they were in early 2014.

We can alternatively look at raw measure of the gains to the amount of floor space per person. Taking  average floor size of homes, which is about 180sqm, and add 3%, and assign it to the average of 2.6 occupants, to get an additional 2sqm of floor space per person.

Or alternatively we can think of it in terms of occupancy rates, which would be 2.51 instead of 2.6 with the same size homes under the 10% higher supply scenario.

That’s all you get for 20 years worth of sustained housing supply stimulus. And you get none of that simply from more elastic supply only.

The point being that current massive price increases, in the order of 17% per year in Sydney and Melbourne, simply cannot be explained by anything like unresponsive supply. Not only that, any supply-side effect on prices takes many decades to have any effect, and only enters the price equation via effects on rents.

If we want cheaper housing we need to reform legal structures to shift bargaining power to tenants from landlords, curb speculation through financial controls (and keep stamp duties!), and stop rewarding political parties who promise housing supply as any sort of solution to current prices.

Unfortunately, very few people actually want housing to become more cheaper. Around 70% of households are homeowners, around 30% are property investors who come from the wealthier part of society, while most politicians also have a huge share of their wealth tied up in residential property. It suits all of these interests to point the finger at supply because they know it sounds attractive in a naive economic way, but won’t actually reduce the value of their housing portfolios.

As Professor Murphy explained, the consensus around new housing supply as a solution to housing affordability problems is a political construct. This unfortunate political reality is best summarised in this tweet. 
Dear reader I hope you see my calculations, follow the logic and believe it!