Did you know housing gets counted twice in GDP?

What makes housing so different?

The argument to include both housing construction (as a new capital investment good) and housing occupancy (as a consumption good) arises from a conceptual trick at the heart of national accounting. That trick is to separate out two types of ‘final’ goods when adding up the ‘value-added’ in the economy, which is what GDP does.

One good is a consumption good. These are goods (and services) that households consume, like clothes, food, entertainment, and so forth. All the value added at intermediate stages in the production chain of these goods can be captured by looking only at the final retail value of the goods. That value represents the total value-added across the economy to produce that good.

The other type of good is an investment good. This is a good that lasts a long time and contributes to future production. A new rail line, for example, is classified a new investment good, and the value of its production is counted in GDP, even though households don’t get any value from it until it is used to run trains.

Once the rail line is being used to run trains, the value of those travel services is also counted in GDP as a consumption good, which will include within it the value contribution of the rail line itself. Thus there is a type of double-counting when it comes to investment goods — you count them when they are made, and you count them again when they are used to make consumption goods.

This is intentional. The production of investment goods is a large share of GDP — between 20 and 40% in most countries. By ignoring this production, which is also the more volatile part of production over the business cycle, GDP loses much of its value as a measure of how economically active a country is.

The construction of new homes is, therefore, an investment good, which gets counted in GDP. But then the occupancy of these same homes gets counted gain as a consumption ‘home rental’ good each period after. This applies to the 70% of households (in Australia at least) who own their own home, not just the renters. Although they don’t pay themselves rent to occupy their home, GDP is calculated as if they do by ‘imputing’ the rent that homeowners would have to pay themselves if they instead rented their home.

What effect does this have on GDP?

Obviously, including an imputed amount that a home-owning household would pay themselves to rent their own home adds to the level of GDP. On average, this imputed value is 8% of GDP [1].

There are two ways to deal with the double counting of housing. The first is to subtract dwelling construction and let that be an intermediate part of the production of the final consumption good of housing. The left panels below show the effect of this on Australia’s GDP over the past 40 years or so. On average, GDP is 6% lower by excluding dwelling construction, although a little more volatile.

The second way to remove double counting is to remove the housing consumption good. This is a little more strange because it would remove the home rental market as well as the imputed rents owner-occupiers are estimated to pay themselves. The right panels show the effect of this on Australia’s GDP, and rather surprisingly, despite the two massive housing booms of the 2000s and 2010s, the share of GDP coming from housing consumption is relatively flat at around 12%.

While there is obviously an effect on the level measurement, the double-counting of housing doesn’t really bias any long term trends.

What can happen, however, is that during periods of high construction and/or high population growth, the short term GDP figures will stop correlating closely with the economic experience of households.

For example, the past four years in Australia has seen very high rates of population growth and dwelling construction. If we subtract dwelling construction, the average per capita rate of growth of GDP is 0.6%, compared to 0.9% using the standard metric (which itself is historically low).

A higher GDP also comes from new people occupying newly built homes as well as from building them. If we subtract the imputed dwelling occupancy, the average per capita rate of growth of GDP in the past four years in Australia has been 0.8% — again below the headline growth rate.

Over longer periods, the way housing is included in GDP doesn’t fundamentally change the patterns of growth. The average growth rate per capita is within 0.03% if we take a decade long view and compare the two alternative metrics. But over short periods it can disguise what is going on under the surface.

Lastly, the treatment of housing in GDP is just one of the many curiosities in this rather over-rated economic metric. In Europe since 2011, illegals drugs and prostitution have also been included using imputation methods. You may never trust GDP again!

Footnote: [1] This is not the only way housing is treated rather uniquely in the national accounts. The “fees paid to lawyers, fees and commissions paid to real estate agents and auctioneers, stamp duty, Title Office charges and local government charges” associated with buying and selling homes are also all counted in GDP, and in particular, as new economic capital. This is unusual because other legal fees and commission-based services are counted as consumption goods, and taxes are not counted in GDP.

Conflating stamp duty and transfer costs as actual economic capital (real buildings) leads to very strange analysis, like that reducing stamp duty increases output because it reduces the ‘cost of capital’. The graphs in this article also remove these transfer costs with the occupancy costs.