We are often asked about the impact of maintenance and capital improvements on direct residential property returns. More specifically, what are the annual performance leakages attributable to these costs? Today’s column will try and shed some analytical light on this murky subject.
As an aside, I will not examine how the same issues afflict the far smaller “commercial property” sector--there is less than $300 billion of commercial, industrial and retail real estate compared with circa $3.6 trillion of housing--which nevertheless assumes far higher weights in institutional investor portfolios to the latter’s detriment.
The very poor performance reported by Australian super funds in recent years has been in part driven by unnecessarily high exposures to risky Listed Property Trusts (or REITs), and unlisted commercial property. As but one illustration, my first chart below compares the performance of Australian houses (capital growth plus 50% of gross rents to control for costs), which is the red line, with the ASX/LPT Accumulation Index, which captures all REITs and is denoted by the blue line.
For some reason, Australian super funds have loaded up on the blue line but completely overlooked the red one. Go figure. And which of these two lines gives more pause in relation to the perennially emotive “bubble trouble” question? We read so much crap about “house price bubbles”, but almost never hear about the far bigger bubbles that one finds in commercial property prices. The hard data suggests that the real “irrational exuberance” was attributable to institutional investors, like super funds, chasing too much commercial, as opposed to residential, property.
Commercial property advocates try to dismiss the highly liquid “market pricing” afforded by the ASX LPT index and point to the much lower volatility (risk) numbers that derive from indices that purport to track “unlisted” portfolios of commercial real estate. The problem with these measures is that they are heavily influenced by “non-market” human valuations that typically ignore distressed sales (ie, those properties selling at discounts during times of crisis) and thus artificially smooth returns. That’s akin to saying, “We are going to remove all the really badly performing companies listed on the ASX because they are distressed.”
But enough venting. Let’s return to my central task today: quantifying the impact of capex and maintenance on residential property returns (the results are similar for commercial property). Some house price proxies, such as “median price” or “stratified median price” indices--of the kind reported by APM and the ABS--make no attempt to control for the bias induced by capital improvements (viz., renovations) and the other maintenance required to preserve the value of the underlying assets included in the index to mitigate depreciation.
The repeat-sales indices reported by Residex are probably hampered by similar problems: rather than plucking the median or middle sales observation every quarter, Residex only looks at homes for which it can find both a purchase and sale, and then measures buy-and-hold returns via a regression-based index. This approach has the obvious drawback of excluding all newly built homes. Another problem is that over the average owner’s holding period of 7-8 years, many will carry out some structural improvements in order to boost the value they extract from that asset, and/or to offset the effects of the inexorable depreciation of the dwelling.
Standard repeat-sales indices do not directly deal with this problem, and are therefore biased measures of house price changes over time. In the academic research literature, there are a number of modifications one can make to the basic repeat-sales methods used by the likes of Case-Shiller in the US to statistically control for non-linear changes in a dwelling’s value, which, it is assumed, arise as a function of capex (or depreciation in the absence of capex). But these are inherently imperfect.
Unfortunately, we cannot know exactly what Residex does because, unusually for a supplier of house price indices, they have historically refused to provide specific details on the methodology they use.
Within Rismark’s library of house price index products, we privately produce all of the known techniques, including adjusted repeat-sales models that control for the abovementioned biases.
By far the best, and most direct, way to address the problems induced by capex on time-series measures of housing costs is through the use of “hedonic” indices. These benchmarks are RP Data-Rismark’s preferred tools for evaluating changes in housing costs over time, and have become the gold standard for the RBA and most banks.
Critically, an hedonic index explicitly controls for the exact attributes of all the homes included in the index, ranging from their precise location, number of bedrooms, number of bathrooms, the property’s type and land size, whether it has a garage or pool, and its view, amongst other things.
As a filter in RP Data and Rismark’s data processing method, we also insist on being able to confirm the specific attributes of all homes within six months of the sales data that is used in the index. In this way, we know that the characteristics that the index relies upon—such as number of bedrooms, bathrooms, land size, etc—are up-to-date.
Hedonic indices are designed to very explicitly compare like-with-like. They effectively take the highly heterogeneous “apples and oranges” that are the homes that transact over time, and normalise them such that we can identify the true underlying rates of capital growth attributable to regions, cities, and the national market.
This brings me back to my original question about the return leakages deriving from both high-value structural capital investments (ie, DA-approved renovations), and other lower value, or non-DA approved, maintenance.
In what follows, I will try and quantify these for you. It is important to bear in mind that these costs need to be weighed against the asset’s total returns: that is, both capital growth and normal or imputed rental incomes. They should not be deducted exclusively from one or the other.
The ABS reports two measures of the value of alterations and additions (A+A) on a quarterly basis. The first is the Construction Work Done Series and leverages off the ABS’s Building Activity Survey. The Building Activity Survey (BAS) supplies estimates of the value of work done on alterations and additions with an approval value of $10,000 or more as collected via DAs submitted to Australian councils.
In the ABS National Accounts a second, far broader measure of residential A+A is published. In particular, the National Accounts upwardly adjusts the BAS estimates to reflect the fact that a substantial amount of alterations and additions activity is not covered in the BAS survey (eg, work under $10,000, and work not done with permits). The National Accounts also boost the BACS A+A by an “undercoverage adjustment” to control for some reporting biases (eg, DAs tend to be undervalued).
Our previous correspondence with the ABS suggested that the overall National Accounts adjustment factor, which varies on a state-by-state basis, is between four and five times.
In the chart below we show the smaller BAS estimate of A+A as a share of the value of the total private residential housing stock since 1986. This is calculated by multiplying the number of private dwellings each quarter with RP Data-Rimark’s measure of the all regions, all property types median dwelling price. Since 1986 the average BAS A+A has been 0.26% per annum.
In the next figure we move to the more comprehensive A+A proxy reported in the National Accounts. Consistent with our prior correspondence, this total A+A estimate is on average 4.95 times larger than the BAS result (see the penultimate figure below). That is, the ABS believes that all forms of dwelling maintenance and capital investments, excluding the construction of new buildings, amount to about 1.26% per annum.
A useful sanity test of these A+A estimates is to look at the value of depreciation as a share of the housing stock. In its National Accounts, the ABS spends considerable time computing the annual value of the depreciation of the private residential dwelling stock owned by individuals. In the final figure we have taken this depreciation proxy and divided it by the same estimate of the total housing stock value used above.
Unsurprisingly, depreciation, which averaged 1.2% per annum between 1986 and 2011, is almost exactly the same as average capital expenditure. In fact, it also tapers off during the 2000s, just as capex appeared to do. And it reconciles almost perfectly with the ABS’s estimate of the average dwelling life of 88 years.
To summarise, this analysis suggests that the total value of maintenance and A+A as a share of the value of the dwelling stock is around 1.2% per annum (rounded down), all else being equal.
